sched.c 222.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 *  kernel/sched.c
 *
 *  Kernel scheduler and related syscalls
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 *
 *  1996-12-23  Modified by Dave Grothe to fix bugs in semaphores and
 *		make semaphores SMP safe
 *  1998-11-19	Implemented schedule_timeout() and related stuff
 *		by Andrea Arcangeli
 *  2002-01-04	New ultra-scalable O(1) scheduler by Ingo Molnar:
 *		hybrid priority-list and round-robin design with
 *		an array-switch method of distributing timeslices
 *		and per-CPU runqueues.  Cleanups and useful suggestions
 *		by Davide Libenzi, preemptible kernel bits by Robert Love.
 *  2003-09-03	Interactivity tuning by Con Kolivas.
 *  2004-04-02	Scheduler domains code by Nick Piggin
I
Ingo Molnar 已提交
19 20 21 22 23 24
 *  2007-04-15  Work begun on replacing all interactivity tuning with a
 *              fair scheduling design by Con Kolivas.
 *  2007-05-05  Load balancing (smp-nice) and other improvements
 *              by Peter Williams
 *  2007-05-06  Interactivity improvements to CFS by Mike Galbraith
 *  2007-07-01  Group scheduling enhancements by Srivatsa Vaddagiri
25 26
 *  2007-11-29  RT balancing improvements by Steven Rostedt, Gregory Haskins,
 *              Thomas Gleixner, Mike Kravetz
L
Linus Torvalds 已提交
27 28 29 30 31 32
 */

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

76
#include <asm/tlb.h>
77
#include <asm/irq_regs.h>
L
Linus Torvalds 已提交
78

79
#include "sched_cpupri.h"
T
Tejun Heo 已提交
80
#include "workqueue_sched.h"
81

82
#define CREATE_TRACE_POINTS
83
#include <trace/events/sched.h>
84

L
Linus Torvalds 已提交
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
/*
 * Convert user-nice values [ -20 ... 0 ... 19 ]
 * to static priority [ MAX_RT_PRIO..MAX_PRIO-1 ],
 * and back.
 */
#define NICE_TO_PRIO(nice)	(MAX_RT_PRIO + (nice) + 20)
#define PRIO_TO_NICE(prio)	((prio) - MAX_RT_PRIO - 20)
#define TASK_NICE(p)		PRIO_TO_NICE((p)->static_prio)

/*
 * 'User priority' is the nice value converted to something we
 * can work with better when scaling various scheduler parameters,
 * it's a [ 0 ... 39 ] range.
 */
#define USER_PRIO(p)		((p)-MAX_RT_PRIO)
#define TASK_USER_PRIO(p)	USER_PRIO((p)->static_prio)
#define MAX_USER_PRIO		(USER_PRIO(MAX_PRIO))

/*
104
 * Helpers for converting nanosecond timing to jiffy resolution
L
Linus Torvalds 已提交
105
 */
106
#define NS_TO_JIFFIES(TIME)	((unsigned long)(TIME) / (NSEC_PER_SEC / HZ))
L
Linus Torvalds 已提交
107

I
Ingo Molnar 已提交
108 109 110
#define NICE_0_LOAD		SCHED_LOAD_SCALE
#define NICE_0_SHIFT		SCHED_LOAD_SHIFT

L
Linus Torvalds 已提交
111 112 113
/*
 * These are the 'tuning knobs' of the scheduler:
 *
D
Dmitry Adamushko 已提交
114
 * default timeslice is 100 msecs (used only for SCHED_RR tasks).
L
Linus Torvalds 已提交
115 116 117
 * Timeslices get refilled after they expire.
 */
#define DEF_TIMESLICE		(100 * HZ / 1000)
118

119 120 121 122 123
/*
 * single value that denotes runtime == period, ie unlimited time.
 */
#define RUNTIME_INF	((u64)~0ULL)

124 125
static inline int rt_policy(int policy)
{
126
	if (unlikely(policy == SCHED_FIFO || policy == SCHED_RR))
127 128 129 130 131 132 133 134 135
		return 1;
	return 0;
}

static inline int task_has_rt_policy(struct task_struct *p)
{
	return rt_policy(p->policy);
}

L
Linus Torvalds 已提交
136
/*
I
Ingo Molnar 已提交
137
 * This is the priority-queue data structure of the RT scheduling class:
L
Linus Torvalds 已提交
138
 */
I
Ingo Molnar 已提交
139 140 141 142 143
struct rt_prio_array {
	DECLARE_BITMAP(bitmap, MAX_RT_PRIO+1); /* include 1 bit for delimiter */
	struct list_head queue[MAX_RT_PRIO];
};

144
struct rt_bandwidth {
I
Ingo Molnar 已提交
145
	/* nests inside the rq lock: */
146
	raw_spinlock_t		rt_runtime_lock;
I
Ingo Molnar 已提交
147 148 149
	ktime_t			rt_period;
	u64			rt_runtime;
	struct hrtimer		rt_period_timer;
150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
};

static struct rt_bandwidth def_rt_bandwidth;

static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun);

static enum hrtimer_restart sched_rt_period_timer(struct hrtimer *timer)
{
	struct rt_bandwidth *rt_b =
		container_of(timer, struct rt_bandwidth, rt_period_timer);
	ktime_t now;
	int overrun;
	int idle = 0;

	for (;;) {
		now = hrtimer_cb_get_time(timer);
		overrun = hrtimer_forward(timer, now, rt_b->rt_period);

		if (!overrun)
			break;

		idle = do_sched_rt_period_timer(rt_b, overrun);
	}

	return idle ? HRTIMER_NORESTART : HRTIMER_RESTART;
}

static
void init_rt_bandwidth(struct rt_bandwidth *rt_b, u64 period, u64 runtime)
{
	rt_b->rt_period = ns_to_ktime(period);
	rt_b->rt_runtime = runtime;

183
	raw_spin_lock_init(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
184

185 186 187 188 189
	hrtimer_init(&rt_b->rt_period_timer,
			CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rt_b->rt_period_timer.function = sched_rt_period_timer;
}

190 191 192
static inline int rt_bandwidth_enabled(void)
{
	return sysctl_sched_rt_runtime >= 0;
193 194 195 196 197 198
}

static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
{
	ktime_t now;

199
	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
200 201 202 203 204
		return;

	if (hrtimer_active(&rt_b->rt_period_timer))
		return;

205
	raw_spin_lock(&rt_b->rt_runtime_lock);
206
	for (;;) {
207 208 209
		unsigned long delta;
		ktime_t soft, hard;

210 211 212 213 214
		if (hrtimer_active(&rt_b->rt_period_timer))
			break;

		now = hrtimer_cb_get_time(&rt_b->rt_period_timer);
		hrtimer_forward(&rt_b->rt_period_timer, now, rt_b->rt_period);
215 216 217 218 219

		soft = hrtimer_get_softexpires(&rt_b->rt_period_timer);
		hard = hrtimer_get_expires(&rt_b->rt_period_timer);
		delta = ktime_to_ns(ktime_sub(hard, soft));
		__hrtimer_start_range_ns(&rt_b->rt_period_timer, soft, delta,
220
				HRTIMER_MODE_ABS_PINNED, 0);
221
	}
222
	raw_spin_unlock(&rt_b->rt_runtime_lock);
223 224 225 226 227 228 229 230 231
}

#ifdef CONFIG_RT_GROUP_SCHED
static void destroy_rt_bandwidth(struct rt_bandwidth *rt_b)
{
	hrtimer_cancel(&rt_b->rt_period_timer);
}
#endif

232 233 234 235 236 237
/*
 * sched_domains_mutex serializes calls to arch_init_sched_domains,
 * detach_destroy_domains and partition_sched_domains.
 */
static DEFINE_MUTEX(sched_domains_mutex);

D
Dhaval Giani 已提交
238
#ifdef CONFIG_CGROUP_SCHED
S
Srivatsa Vaddagiri 已提交
239

240 241
#include <linux/cgroup.h>

S
Srivatsa Vaddagiri 已提交
242 243
struct cfs_rq;

P
Peter Zijlstra 已提交
244 245
static LIST_HEAD(task_groups);

S
Srivatsa Vaddagiri 已提交
246
/* task group related information */
247
struct task_group {
248
	struct cgroup_subsys_state css;
249

250
#ifdef CONFIG_FAIR_GROUP_SCHED
S
Srivatsa Vaddagiri 已提交
251 252 253 254 255
	/* schedulable entities of this group on each cpu */
	struct sched_entity **se;
	/* runqueue "owned" by this group on each cpu */
	struct cfs_rq **cfs_rq;
	unsigned long shares;
256 257 258 259 260 261
#endif

#ifdef CONFIG_RT_GROUP_SCHED
	struct sched_rt_entity **rt_se;
	struct rt_rq **rt_rq;

262
	struct rt_bandwidth rt_bandwidth;
263
#endif
264

265
	struct rcu_head rcu;
P
Peter Zijlstra 已提交
266
	struct list_head list;
P
Peter Zijlstra 已提交
267 268 269 270

	struct task_group *parent;
	struct list_head siblings;
	struct list_head children;
S
Srivatsa Vaddagiri 已提交
271 272
};

273
#define root_task_group init_task_group
P
Peter Zijlstra 已提交
274

275
/* task_group_lock serializes add/remove of task groups and also changes to
276 277
 * a task group's cpu shares.
 */
278
static DEFINE_SPINLOCK(task_group_lock);
279

280 281
#ifdef CONFIG_FAIR_GROUP_SCHED

282 283 284 285 286 287 288
#ifdef CONFIG_SMP
static int root_task_group_empty(void)
{
	return list_empty(&root_task_group.children);
}
#endif

289 290
# define INIT_TASK_GROUP_LOAD	NICE_0_LOAD

291
/*
292 293 294 295
 * A weight of 0 or 1 can cause arithmetics problems.
 * A weight of a cfs_rq is the sum of weights of which entities
 * are queued on this cfs_rq, so a weight of a entity should not be
 * too large, so as the shares value of a task group.
296 297 298
 * (The default weight is 1024 - so there's no practical
 *  limitation from this.)
 */
299
#define MIN_SHARES	2
300
#define MAX_SHARES	(1UL << 18)
301

302 303 304
static int init_task_group_load = INIT_TASK_GROUP_LOAD;
#endif

S
Srivatsa Vaddagiri 已提交
305
/* Default task group.
I
Ingo Molnar 已提交
306
 *	Every task in system belong to this group at bootup.
S
Srivatsa Vaddagiri 已提交
307
 */
308
struct task_group init_task_group;
S
Srivatsa Vaddagiri 已提交
309

D
Dhaval Giani 已提交
310
#endif	/* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
311

I
Ingo Molnar 已提交
312 313 314 315 316 317
/* CFS-related fields in a runqueue */
struct cfs_rq {
	struct load_weight load;
	unsigned long nr_running;

	u64 exec_clock;
I
Ingo Molnar 已提交
318
	u64 min_vruntime;
I
Ingo Molnar 已提交
319 320 321

	struct rb_root tasks_timeline;
	struct rb_node *rb_leftmost;
322 323 324 325 326 327

	struct list_head tasks;
	struct list_head *balance_iterator;

	/*
	 * 'curr' points to currently running entity on this cfs_rq.
I
Ingo Molnar 已提交
328 329
	 * It is set to NULL otherwise (i.e when none are currently running).
	 */
P
Peter Zijlstra 已提交
330
	struct sched_entity *curr, *next, *last;
P
Peter Zijlstra 已提交
331

P
Peter Zijlstra 已提交
332
	unsigned int nr_spread_over;
P
Peter Zijlstra 已提交
333

334
#ifdef CONFIG_FAIR_GROUP_SCHED
I
Ingo Molnar 已提交
335 336
	struct rq *rq;	/* cpu runqueue to which this cfs_rq is attached */

I
Ingo Molnar 已提交
337 338
	/*
	 * leaf cfs_rqs are those that hold tasks (lowest schedulable entity in
I
Ingo Molnar 已提交
339 340 341 342 343 344
	 * a hierarchy). Non-leaf lrqs hold other higher schedulable entities
	 * (like users, containers etc.)
	 *
	 * leaf_cfs_rq_list ties together list of leaf cfs_rq's in a cpu. This
	 * list is used during load balance.
	 */
I
Ingo Molnar 已提交
345 346
	struct list_head leaf_cfs_rq_list;
	struct task_group *tg;	/* group that "owns" this runqueue */
347 348 349

#ifdef CONFIG_SMP
	/*
350
	 * the part of load.weight contributed by tasks
351
	 */
352
	unsigned long task_weight;
353

354 355 356 357 358 359 360
	/*
	 *   h_load = weight * f(tg)
	 *
	 * Where f(tg) is the recursive weight fraction assigned to
	 * this group.
	 */
	unsigned long h_load;
361

362 363 364 365
	/*
	 * this cpu's part of tg->shares
	 */
	unsigned long shares;
366 367 368 369 370

	/*
	 * load.weight at the time we set shares
	 */
	unsigned long rq_weight;
371
#endif
I
Ingo Molnar 已提交
372 373
#endif
};
L
Linus Torvalds 已提交
374

I
Ingo Molnar 已提交
375 376 377
/* Real-Time classes' related field in a runqueue: */
struct rt_rq {
	struct rt_prio_array active;
378
	unsigned long rt_nr_running;
379
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
380 381
	struct {
		int curr; /* highest queued rt task prio */
382
#ifdef CONFIG_SMP
383
		int next; /* next highest */
384
#endif
385
	} highest_prio;
P
Peter Zijlstra 已提交
386
#endif
P
Peter Zijlstra 已提交
387
#ifdef CONFIG_SMP
388
	unsigned long rt_nr_migratory;
389
	unsigned long rt_nr_total;
G
Gregory Haskins 已提交
390
	int overloaded;
391
	struct plist_head pushable_tasks;
P
Peter Zijlstra 已提交
392
#endif
P
Peter Zijlstra 已提交
393
	int rt_throttled;
P
Peter Zijlstra 已提交
394
	u64 rt_time;
P
Peter Zijlstra 已提交
395
	u64 rt_runtime;
I
Ingo Molnar 已提交
396
	/* Nests inside the rq lock: */
397
	raw_spinlock_t rt_runtime_lock;
P
Peter Zijlstra 已提交
398

399
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
400 401
	unsigned long rt_nr_boosted;

P
Peter Zijlstra 已提交
402 403 404 405
	struct rq *rq;
	struct list_head leaf_rt_rq_list;
	struct task_group *tg;
#endif
I
Ingo Molnar 已提交
406 407
};

G
Gregory Haskins 已提交
408 409 410 411
#ifdef CONFIG_SMP

/*
 * We add the notion of a root-domain which will be used to define per-domain
I
Ingo Molnar 已提交
412 413
 * variables. Each exclusive cpuset essentially defines an island domain by
 * fully partitioning the member cpus from any other cpuset. Whenever a new
G
Gregory Haskins 已提交
414 415 416 417 418 419
 * exclusive cpuset is created, we also create and attach a new root-domain
 * object.
 *
 */
struct root_domain {
	atomic_t refcount;
420 421
	cpumask_var_t span;
	cpumask_var_t online;
422

I
Ingo Molnar 已提交
423
	/*
424 425 426
	 * The "RT overload" flag: it gets set if a CPU has more than
	 * one runnable RT task.
	 */
427
	cpumask_var_t rto_mask;
I
Ingo Molnar 已提交
428
	atomic_t rto_count;
429
	struct cpupri cpupri;
G
Gregory Haskins 已提交
430 431
};

432 433 434 435
/*
 * By default the system creates a single root-domain with all cpus as
 * members (mimicking the global state we have today).
 */
G
Gregory Haskins 已提交
436 437
static struct root_domain def_root_domain;

438
#endif /* CONFIG_SMP */
G
Gregory Haskins 已提交
439

L
Linus Torvalds 已提交
440 441 442 443 444 445 446
/*
 * This is the main, per-CPU runqueue data structure.
 *
 * Locking rule: those places that want to lock multiple runqueues
 * (such as the load balancing or the thread migration code), lock
 * acquire operations must be ordered by ascending &runqueue.
 */
447
struct rq {
448
	/* runqueue lock: */
449
	raw_spinlock_t lock;
L
Linus Torvalds 已提交
450 451 452 453 454 455

	/*
	 * nr_running and cpu_load should be in the same cacheline because
	 * remote CPUs use both these fields when doing load calculation.
	 */
	unsigned long nr_running;
I
Ingo Molnar 已提交
456 457
	#define CPU_LOAD_IDX_MAX 5
	unsigned long cpu_load[CPU_LOAD_IDX_MAX];
458
	unsigned long last_load_update_tick;
459
#ifdef CONFIG_NO_HZ
M
Mike Galbraith 已提交
460
	u64 nohz_stamp;
461
	unsigned char nohz_balance_kick;
462
#endif
463 464
	unsigned int skip_clock_update;

465 466
	/* capture load from *all* tasks on this cpu: */
	struct load_weight load;
I
Ingo Molnar 已提交
467 468 469 470
	unsigned long nr_load_updates;
	u64 nr_switches;

	struct cfs_rq cfs;
P
Peter Zijlstra 已提交
471 472
	struct rt_rq rt;

I
Ingo Molnar 已提交
473
#ifdef CONFIG_FAIR_GROUP_SCHED
474 475
	/* list of leaf cfs_rq on this cpu: */
	struct list_head leaf_cfs_rq_list;
476 477
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
478
	struct list_head leaf_rt_rq_list;
L
Linus Torvalds 已提交
479 480 481 482 483 484 485 486 487 488
#endif

	/*
	 * This is part of a global counter where only the total sum
	 * over all CPUs matters. A task can increase this counter on
	 * one CPU and if it got migrated afterwards it may decrease
	 * it on another CPU. Always updated under the runqueue lock:
	 */
	unsigned long nr_uninterruptible;

489
	struct task_struct *curr, *idle;
490
	unsigned long next_balance;
L
Linus Torvalds 已提交
491
	struct mm_struct *prev_mm;
I
Ingo Molnar 已提交
492

493
	u64 clock;
I
Ingo Molnar 已提交
494

L
Linus Torvalds 已提交
495 496 497
	atomic_t nr_iowait;

#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
498
	struct root_domain *rd;
L
Linus Torvalds 已提交
499 500
	struct sched_domain *sd;

501 502
	unsigned long cpu_power;

503
	unsigned char idle_at_tick;
L
Linus Torvalds 已提交
504
	/* For active balancing */
505
	int post_schedule;
L
Linus Torvalds 已提交
506 507
	int active_balance;
	int push_cpu;
508
	struct cpu_stop_work active_balance_work;
509 510
	/* cpu of this runqueue: */
	int cpu;
511
	int online;
L
Linus Torvalds 已提交
512

513
	unsigned long avg_load_per_task;
L
Linus Torvalds 已提交
514

515 516
	u64 rt_avg;
	u64 age_stamp;
M
Mike Galbraith 已提交
517 518
	u64 idle_stamp;
	u64 avg_idle;
L
Linus Torvalds 已提交
519 520
#endif

521 522 523 524
	/* calc_load related fields */
	unsigned long calc_load_update;
	long calc_load_active;

P
Peter Zijlstra 已提交
525
#ifdef CONFIG_SCHED_HRTICK
526 527 528 529
#ifdef CONFIG_SMP
	int hrtick_csd_pending;
	struct call_single_data hrtick_csd;
#endif
P
Peter Zijlstra 已提交
530 531 532
	struct hrtimer hrtick_timer;
#endif

L
Linus Torvalds 已提交
533 534 535
#ifdef CONFIG_SCHEDSTATS
	/* latency stats */
	struct sched_info rq_sched_info;
536 537
	unsigned long long rq_cpu_time;
	/* could above be rq->cfs_rq.exec_clock + rq->rt_rq.rt_runtime ? */
L
Linus Torvalds 已提交
538 539

	/* sys_sched_yield() stats */
540
	unsigned int yld_count;
L
Linus Torvalds 已提交
541 542

	/* schedule() stats */
543 544 545
	unsigned int sched_switch;
	unsigned int sched_count;
	unsigned int sched_goidle;
L
Linus Torvalds 已提交
546 547

	/* try_to_wake_up() stats */
548 549
	unsigned int ttwu_count;
	unsigned int ttwu_local;
I
Ingo Molnar 已提交
550 551

	/* BKL stats */
552
	unsigned int bkl_count;
L
Linus Torvalds 已提交
553 554 555
#endif
};

556
static DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
L
Linus Torvalds 已提交
557

P
Peter Zijlstra 已提交
558 559
static inline
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
560
{
P
Peter Zijlstra 已提交
561
	rq->curr->sched_class->check_preempt_curr(rq, p, flags);
562 563 564 565 566 567 568

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
	if (test_tsk_need_resched(p))
		rq->skip_clock_update = 1;
I
Ingo Molnar 已提交
569 570
}

571 572 573 574 575 576 577 578 579
static inline int cpu_of(struct rq *rq)
{
#ifdef CONFIG_SMP
	return rq->cpu;
#else
	return 0;
#endif
}

580
#define rcu_dereference_check_sched_domain(p) \
581 582 583 584
	rcu_dereference_check((p), \
			      rcu_read_lock_sched_held() || \
			      lockdep_is_held(&sched_domains_mutex))

N
Nick Piggin 已提交
585 586
/*
 * The domain tree (rq->sd) is protected by RCU's quiescent state transition.
587
 * See detach_destroy_domains: synchronize_sched for details.
N
Nick Piggin 已提交
588 589 590 591
 *
 * The domain tree of any CPU may only be accessed from within
 * preempt-disabled sections.
 */
592
#define for_each_domain(cpu, __sd) \
593
	for (__sd = rcu_dereference_check_sched_domain(cpu_rq(cpu)->sd); __sd; __sd = __sd->parent)
L
Linus Torvalds 已提交
594 595 596 597 598

#define cpu_rq(cpu)		(&per_cpu(runqueues, (cpu)))
#define this_rq()		(&__get_cpu_var(runqueues))
#define task_rq(p)		cpu_rq(task_cpu(p))
#define cpu_curr(cpu)		(cpu_rq(cpu)->curr)
599
#define raw_rq()		(&__raw_get_cpu_var(runqueues))
L
Linus Torvalds 已提交
600

601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643
#ifdef CONFIG_CGROUP_SCHED

/*
 * Return the group to which this tasks belongs.
 *
 * We use task_subsys_state_check() and extend the RCU verification
 * with lockdep_is_held(&task_rq(p)->lock) because cpu_cgroup_attach()
 * holds that lock for each task it moves into the cgroup. Therefore
 * by holding that lock, we pin the task to the current cgroup.
 */
static inline struct task_group *task_group(struct task_struct *p)
{
	struct cgroup_subsys_state *css;

	css = task_subsys_state_check(p, cpu_cgroup_subsys_id,
			lockdep_is_held(&task_rq(p)->lock));
	return container_of(css, struct task_group, css);
}

/* Change a task's cfs_rq and parent entity if it moves across CPUs/groups */
static inline void set_task_rq(struct task_struct *p, unsigned int cpu)
{
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq = task_group(p)->cfs_rq[cpu];
	p->se.parent = task_group(p)->se[cpu];
#endif

#ifdef CONFIG_RT_GROUP_SCHED
	p->rt.rt_rq  = task_group(p)->rt_rq[cpu];
	p->rt.parent = task_group(p)->rt_se[cpu];
#endif
}

#else /* CONFIG_CGROUP_SCHED */

static inline void set_task_rq(struct task_struct *p, unsigned int cpu) { }
static inline struct task_group *task_group(struct task_struct *p)
{
	return NULL;
}

#endif /* CONFIG_CGROUP_SCHED */

I
Ingo Molnar 已提交
644
inline void update_rq_clock(struct rq *rq)
645
{
646 647
	if (!rq->skip_clock_update)
		rq->clock = sched_clock_cpu(cpu_of(rq));
648 649
}

I
Ingo Molnar 已提交
650 651 652 653 654 655 656 657 658
/*
 * Tunables that become constants when CONFIG_SCHED_DEBUG is off:
 */
#ifdef CONFIG_SCHED_DEBUG
# define const_debug __read_mostly
#else
# define const_debug static const
#endif

I
Ingo Molnar 已提交
659 660
/**
 * runqueue_is_locked
661
 * @cpu: the processor in question.
I
Ingo Molnar 已提交
662 663 664 665 666
 *
 * Returns true if the current cpu runqueue is locked.
 * This interface allows printk to be called with the runqueue lock
 * held and know whether or not it is OK to wake up the klogd.
 */
667
int runqueue_is_locked(int cpu)
I
Ingo Molnar 已提交
668
{
669
	return raw_spin_is_locked(&cpu_rq(cpu)->lock);
I
Ingo Molnar 已提交
670 671
}

I
Ingo Molnar 已提交
672 673 674
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
675 676 677 678

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

I
Ingo Molnar 已提交
679
enum {
P
Peter Zijlstra 已提交
680
#include "sched_features.h"
I
Ingo Molnar 已提交
681 682
};

P
Peter Zijlstra 已提交
683 684 685 686 687
#undef SCHED_FEAT

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

I
Ingo Molnar 已提交
688
const_debug unsigned int sysctl_sched_features =
P
Peter Zijlstra 已提交
689 690 691 692 693 694 695 696 697
#include "sched_features.h"
	0;

#undef SCHED_FEAT

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

698
static __read_mostly char *sched_feat_names[] = {
P
Peter Zijlstra 已提交
699 700 701 702 703 704
#include "sched_features.h"
	NULL
};

#undef SCHED_FEAT

L
Li Zefan 已提交
705
static int sched_feat_show(struct seq_file *m, void *v)
P
Peter Zijlstra 已提交
706 707 708 709
{
	int i;

	for (i = 0; sched_feat_names[i]; i++) {
L
Li Zefan 已提交
710 711 712
		if (!(sysctl_sched_features & (1UL << i)))
			seq_puts(m, "NO_");
		seq_printf(m, "%s ", sched_feat_names[i]);
P
Peter Zijlstra 已提交
713
	}
L
Li Zefan 已提交
714
	seq_puts(m, "\n");
P
Peter Zijlstra 已提交
715

L
Li Zefan 已提交
716
	return 0;
P
Peter Zijlstra 已提交
717 718 719 720 721 722 723
}

static ssize_t
sched_feat_write(struct file *filp, const char __user *ubuf,
		size_t cnt, loff_t *ppos)
{
	char buf[64];
724
	char *cmp;
P
Peter Zijlstra 已提交
725 726 727 728 729 730 731 732 733 734
	int neg = 0;
	int i;

	if (cnt > 63)
		cnt = 63;

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

	buf[cnt] = 0;
735
	cmp = strstrip(buf);
P
Peter Zijlstra 已提交
736

I
Ingo Molnar 已提交
737
	if (strncmp(buf, "NO_", 3) == 0) {
P
Peter Zijlstra 已提交
738 739 740 741 742
		neg = 1;
		cmp += 3;
	}

	for (i = 0; sched_feat_names[i]; i++) {
743
		if (strcmp(cmp, sched_feat_names[i]) == 0) {
P
Peter Zijlstra 已提交
744 745 746 747 748 749 750 751 752 753 754
			if (neg)
				sysctl_sched_features &= ~(1UL << i);
			else
				sysctl_sched_features |= (1UL << i);
			break;
		}
	}

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

755
	*ppos += cnt;
P
Peter Zijlstra 已提交
756 757 758 759

	return cnt;
}

L
Li Zefan 已提交
760 761 762 763 764
static int sched_feat_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, sched_feat_show, NULL);
}

765
static const struct file_operations sched_feat_fops = {
L
Li Zefan 已提交
766 767 768 769 770
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
P
Peter Zijlstra 已提交
771 772 773 774 775 776 777 778 779 780 781 782 783 784
};

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

	return 0;
}
late_initcall(sched_init_debug);

#endif

#define sched_feat(x) (sysctl_sched_features & (1UL << __SCHED_FEAT_##x))
I
Ingo Molnar 已提交
785

786 787 788 789 790 791
/*
 * 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;

P
Peter Zijlstra 已提交
792 793
/*
 * ratelimit for updating the group shares.
794
 * default: 0.25ms
P
Peter Zijlstra 已提交
795
 */
796
unsigned int sysctl_sched_shares_ratelimit = 250000;
797
unsigned int normalized_sysctl_sched_shares_ratelimit = 250000;
P
Peter Zijlstra 已提交
798

799 800 801 802 803 804 805
/*
 * Inject some fuzzyness into changing the per-cpu group shares
 * this avoids remote rq-locks at the expense of fairness.
 * default: 4
 */
unsigned int sysctl_sched_shares_thresh = 4;

806 807 808 809 810 811 812 813
/*
 * period over which we average the RT time consumption, measured
 * in ms.
 *
 * default: 1s
 */
const_debug unsigned int sysctl_sched_time_avg = MSEC_PER_SEC;

P
Peter Zijlstra 已提交
814
/*
P
Peter Zijlstra 已提交
815
 * period over which we measure -rt task cpu usage in us.
P
Peter Zijlstra 已提交
816 817
 * default: 1s
 */
P
Peter Zijlstra 已提交
818
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
819

820 821
static __read_mostly int scheduler_running;

P
Peter Zijlstra 已提交
822 823 824 825 826
/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
P
Peter Zijlstra 已提交
827

828 829 830 831 832 833 834
static inline u64 global_rt_period(void)
{
	return (u64)sysctl_sched_rt_period * NSEC_PER_USEC;
}

static inline u64 global_rt_runtime(void)
{
835
	if (sysctl_sched_rt_runtime < 0)
836 837 838 839
		return RUNTIME_INF;

	return (u64)sysctl_sched_rt_runtime * NSEC_PER_USEC;
}
P
Peter Zijlstra 已提交
840

L
Linus Torvalds 已提交
841
#ifndef prepare_arch_switch
842 843 844 845 846 847
# define prepare_arch_switch(next)	do { } while (0)
#endif
#ifndef finish_arch_switch
# define finish_arch_switch(prev)	do { } while (0)
#endif

848 849 850 851 852
static inline int task_current(struct rq *rq, struct task_struct *p)
{
	return rq->curr == p;
}

853
#ifndef __ARCH_WANT_UNLOCKED_CTXSW
854
static inline int task_running(struct rq *rq, struct task_struct *p)
855
{
856
	return task_current(rq, p);
857 858
}

859
static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
860 861 862
{
}

863
static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
864
{
865 866 867 868
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
	rq->lock.owner = current;
#endif
869 870 871 872 873 874 875
	/*
	 * If we are tracking spinlock dependencies then we have to
	 * fix up the runqueue lock - which gets 'carried over' from
	 * prev into current:
	 */
	spin_acquire(&rq->lock.dep_map, 0, 0, _THIS_IP_);

876
	raw_spin_unlock_irq(&rq->lock);
877 878 879
}

#else /* __ARCH_WANT_UNLOCKED_CTXSW */
880
static inline int task_running(struct rq *rq, struct task_struct *p)
881 882 883 884
{
#ifdef CONFIG_SMP
	return p->oncpu;
#else
885
	return task_current(rq, p);
886 887 888
#endif
}

889
static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
890 891 892 893 894 895 896 897 898 899
{
#ifdef CONFIG_SMP
	/*
	 * We can optimise this out completely for !SMP, because the
	 * SMP rebalancing from interrupt is the only thing that cares
	 * here.
	 */
	next->oncpu = 1;
#endif
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
900
	raw_spin_unlock_irq(&rq->lock);
901
#else
902
	raw_spin_unlock(&rq->lock);
903 904 905
#endif
}

906
static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
907 908 909 910 911 912 913 914 915 916 917 918
{
#ifdef CONFIG_SMP
	/*
	 * After ->oncpu is cleared, the task can be moved to a different CPU.
	 * We must ensure this doesn't happen until the switch is completely
	 * finished.
	 */
	smp_wmb();
	prev->oncpu = 0;
#endif
#ifndef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
L
Linus Torvalds 已提交
919
#endif
920 921
}
#endif /* __ARCH_WANT_UNLOCKED_CTXSW */
L
Linus Torvalds 已提交
922

923
/*
P
Peter Zijlstra 已提交
924 925
 * Check whether the task is waking, we use this to synchronize ->cpus_allowed
 * against ttwu().
926 927 928
 */
static inline int task_is_waking(struct task_struct *p)
{
929
	return unlikely(p->state == TASK_WAKING);
930 931
}

932 933 934 935
/*
 * __task_rq_lock - lock the runqueue a given task resides on.
 * Must be called interrupts disabled.
 */
936
static inline struct rq *__task_rq_lock(struct task_struct *p)
937 938
	__acquires(rq->lock)
{
939 940
	struct rq *rq;

941
	for (;;) {
942
		rq = task_rq(p);
943
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
944
		if (likely(rq == task_rq(p)))
945
			return rq;
946
		raw_spin_unlock(&rq->lock);
947 948 949
	}
}

L
Linus Torvalds 已提交
950 951
/*
 * task_rq_lock - lock the runqueue a given task resides on and disable
I
Ingo Molnar 已提交
952
 * interrupts. Note the ordering: we can safely lookup the task_rq without
L
Linus Torvalds 已提交
953 954
 * explicitly disabling preemption.
 */
955
static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
L
Linus Torvalds 已提交
956 957
	__acquires(rq->lock)
{
958
	struct rq *rq;
L
Linus Torvalds 已提交
959

960 961 962
	for (;;) {
		local_irq_save(*flags);
		rq = task_rq(p);
963
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
964
		if (likely(rq == task_rq(p)))
965
			return rq;
966
		raw_spin_unlock_irqrestore(&rq->lock, *flags);
L
Linus Torvalds 已提交
967 968 969
	}
}

A
Alexey Dobriyan 已提交
970
static void __task_rq_unlock(struct rq *rq)
971 972
	__releases(rq->lock)
{
973
	raw_spin_unlock(&rq->lock);
974 975
}

976
static inline void task_rq_unlock(struct rq *rq, unsigned long *flags)
L
Linus Torvalds 已提交
977 978
	__releases(rq->lock)
{
979
	raw_spin_unlock_irqrestore(&rq->lock, *flags);
L
Linus Torvalds 已提交
980 981 982
}

/*
983
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
984
 */
A
Alexey Dobriyan 已提交
985
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
986 987
	__acquires(rq->lock)
{
988
	struct rq *rq;
L
Linus Torvalds 已提交
989 990 991

	local_irq_disable();
	rq = this_rq();
992
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
993 994 995 996

	return rq;
}

P
Peter Zijlstra 已提交
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
#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.
 */

/*
 * Use hrtick when:
 *  - enabled by features
 *  - hrtimer is actually high res
 */
static inline int hrtick_enabled(struct rq *rq)
{
	if (!sched_feat(HRTICK))
		return 0;
1018
	if (!cpu_active(cpu_of(rq)))
1019
		return 0;
P
Peter Zijlstra 已提交
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
	return hrtimer_is_hres_active(&rq->hrtick_timer);
}

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());

1039
	raw_spin_lock(&rq->lock);
1040
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
1041
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
1042
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
1043 1044 1045 1046

	return HRTIMER_NORESTART;
}

1047
#ifdef CONFIG_SMP
1048 1049 1050 1051
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
1052
{
1053
	struct rq *rq = arg;
1054

1055
	raw_spin_lock(&rq->lock);
1056 1057
	hrtimer_restart(&rq->hrtick_timer);
	rq->hrtick_csd_pending = 0;
1058
	raw_spin_unlock(&rq->lock);
1059 1060
}

1061 1062 1063 1064 1065 1066
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
static void hrtick_start(struct rq *rq, u64 delay)
1067
{
1068 1069
	struct hrtimer *timer = &rq->hrtick_timer;
	ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
1070

1071
	hrtimer_set_expires(timer, time);
1072 1073 1074 1075

	if (rq == this_rq()) {
		hrtimer_restart(timer);
	} else if (!rq->hrtick_csd_pending) {
1076
		__smp_call_function_single(cpu_of(rq), &rq->hrtick_csd, 0);
1077 1078
		rq->hrtick_csd_pending = 1;
	}
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
}

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:
1093
		hrtick_clear(cpu_rq(cpu));
1094 1095 1096 1097 1098 1099
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

1100
static __init void init_hrtick(void)
1101 1102 1103
{
	hotcpu_notifier(hotplug_hrtick, 0);
}
1104 1105 1106 1107 1108 1109 1110 1111
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
static void hrtick_start(struct rq *rq, u64 delay)
{
1112
	__hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
1113
			HRTIMER_MODE_REL_PINNED, 0);
1114
}
1115

A
Andrew Morton 已提交
1116
static inline void init_hrtick(void)
P
Peter Zijlstra 已提交
1117 1118
{
}
1119
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
1120

1121
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
1122
{
1123 1124
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
1125

1126 1127 1128 1129
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
1130

1131 1132
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
1133
}
A
Andrew Morton 已提交
1134
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
1135 1136 1137 1138 1139 1140 1141 1142
static inline void hrtick_clear(struct rq *rq)
{
}

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

1143 1144 1145
static inline void init_hrtick(void)
{
}
A
Andrew Morton 已提交
1146
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
1147

I
Ingo Molnar 已提交
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160
/*
 * 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

1161
static void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
1162 1163 1164
{
	int cpu;

1165
	assert_raw_spin_locked(&task_rq(p)->lock);
I
Ingo Molnar 已提交
1166

1167
	if (test_tsk_need_resched(p))
I
Ingo Molnar 已提交
1168 1169
		return;

1170
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186

	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);
}

static void resched_cpu(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1187
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
1188 1189
		return;
	resched_task(cpu_curr(cpu));
1190
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
1191
}
1192 1193

#ifdef CONFIG_NO_HZ
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214
/*
 * 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;

	for_each_domain(cpu, sd) {
		for_each_cpu(i, sched_domain_span(sd))
			if (!idle_cpu(i))
				return i;
	}
	return cpu;
}
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
/*
 * When add_timer_on() enqueues a timer into the timer wheel of an
 * idle CPU then this timer might expire before the next timer event
 * which is scheduled to wake up that CPU. In case of a completely
 * idle system the next event might even be infinite time into the
 * future. wake_up_idle_cpu() ensures that the CPU is woken up and
 * leaves the inner idle loop so the newly added timer is taken into
 * account when the CPU goes back to idle and evaluates the timer
 * wheel for the next timer event.
 */
void wake_up_idle_cpu(int cpu)
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

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

	/*
	 * We can set TIF_RESCHED on the idle task of the other CPU
	 * lockless. The worst case is that the other CPU runs the
	 * idle task through an additional NOOP schedule()
	 */
1247
	set_tsk_need_resched(rq->idle);
1248 1249 1250 1251 1252 1253

	/* NEED_RESCHED must be visible before we test polling */
	smp_mb();
	if (!tsk_is_polling(rq->idle))
		smp_send_reschedule(cpu);
}
M
Mike Galbraith 已提交
1254

1255
#endif /* CONFIG_NO_HZ */
1256

1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
static u64 sched_avg_period(void)
{
	return (u64)sysctl_sched_time_avg * NSEC_PER_MSEC / 2;
}

static void sched_avg_update(struct rq *rq)
{
	s64 period = sched_avg_period();

	while ((s64)(rq->clock - rq->age_stamp) > period) {
1267 1268 1269 1270 1271 1272
		/*
		 * 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));
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
}

static void sched_rt_avg_update(struct rq *rq, u64 rt_delta)
{
	rq->rt_avg += rt_delta;
	sched_avg_update(rq);
}

1284
#else /* !CONFIG_SMP */
1285
static void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
1286
{
1287
	assert_raw_spin_locked(&task_rq(p)->lock);
1288
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
1289
}
1290 1291 1292 1293

static void sched_rt_avg_update(struct rq *rq, u64 rt_delta)
{
}
1294
#endif /* CONFIG_SMP */
I
Ingo Molnar 已提交
1295

1296 1297 1298 1299 1300 1301 1302 1303
#if BITS_PER_LONG == 32
# define WMULT_CONST	(~0UL)
#else
# define WMULT_CONST	(1UL << 32)
#endif

#define WMULT_SHIFT	32

I
Ingo Molnar 已提交
1304 1305 1306
/*
 * Shift right and round:
 */
I
Ingo Molnar 已提交
1307
#define SRR(x, y) (((x) + (1UL << ((y) - 1))) >> (y))
I
Ingo Molnar 已提交
1308

1309 1310 1311
/*
 * delta *= weight / lw
 */
1312
static unsigned long
1313 1314 1315 1316 1317
calc_delta_mine(unsigned long delta_exec, unsigned long weight,
		struct load_weight *lw)
{
	u64 tmp;

1318 1319 1320 1321 1322 1323 1324
	if (!lw->inv_weight) {
		if (BITS_PER_LONG > 32 && unlikely(lw->weight >= WMULT_CONST))
			lw->inv_weight = 1;
		else
			lw->inv_weight = 1 + (WMULT_CONST-lw->weight/2)
				/ (lw->weight+1);
	}
1325 1326 1327 1328 1329

	tmp = (u64)delta_exec * weight;
	/*
	 * Check whether we'd overflow the 64-bit multiplication:
	 */
I
Ingo Molnar 已提交
1330
	if (unlikely(tmp > WMULT_CONST))
I
Ingo Molnar 已提交
1331
		tmp = SRR(SRR(tmp, WMULT_SHIFT/2) * lw->inv_weight,
I
Ingo Molnar 已提交
1332 1333
			WMULT_SHIFT/2);
	else
I
Ingo Molnar 已提交
1334
		tmp = SRR(tmp * lw->inv_weight, WMULT_SHIFT);
1335

1336
	return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX);
1337 1338
}

1339
static inline void update_load_add(struct load_weight *lw, unsigned long inc)
1340 1341
{
	lw->weight += inc;
I
Ingo Molnar 已提交
1342
	lw->inv_weight = 0;
1343 1344
}

1345
static inline void update_load_sub(struct load_weight *lw, unsigned long dec)
1346 1347
{
	lw->weight -= dec;
I
Ingo Molnar 已提交
1348
	lw->inv_weight = 0;
1349 1350
}

1351 1352 1353 1354
/*
 * To aid in avoiding the subversion of "niceness" due to uneven distribution
 * of tasks with abnormal "nice" values across CPUs the contribution that
 * each task makes to its run queue's load is weighted according to its
I
Ingo Molnar 已提交
1355
 * scheduling class and "nice" value. For SCHED_NORMAL tasks this is just a
1356 1357 1358 1359
 * scaled version of the new time slice allocation that they receive on time
 * slice expiry etc.
 */

P
Peter Zijlstra 已提交
1360 1361
#define WEIGHT_IDLEPRIO                3
#define WMULT_IDLEPRIO         1431655765
I
Ingo Molnar 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370

/*
 * Nice levels are multiplicative, with a gentle 10% change for every
 * nice level changed. I.e. when a CPU-bound task goes from nice 0 to
 * nice 1, it will get ~10% less CPU time than another CPU-bound task
 * that remained on nice 0.
 *
 * The "10% effect" is relative and cumulative: from _any_ nice level,
 * if you go up 1 level, it's -10% CPU usage, if you go down 1 level
1371 1372 1373
 * it's +10% CPU usage. (to achieve that we use a multiplier of 1.25.
 * If a task goes up by ~10% and another task goes down by ~10% then
 * the relative distance between them is ~25%.)
I
Ingo Molnar 已提交
1374 1375
 */
static const int prio_to_weight[40] = {
1376 1377 1378 1379 1380 1381 1382 1383
 /* -20 */     88761,     71755,     56483,     46273,     36291,
 /* -15 */     29154,     23254,     18705,     14949,     11916,
 /* -10 */      9548,      7620,      6100,      4904,      3906,
 /*  -5 */      3121,      2501,      1991,      1586,      1277,
 /*   0 */      1024,       820,       655,       526,       423,
 /*   5 */       335,       272,       215,       172,       137,
 /*  10 */       110,        87,        70,        56,        45,
 /*  15 */        36,        29,        23,        18,        15,
I
Ingo Molnar 已提交
1384 1385
};

1386 1387 1388 1389 1390 1391 1392
/*
 * Inverse (2^32/x) values of the prio_to_weight[] array, precalculated.
 *
 * In cases where the weight does not change often, we can use the
 * precalculated inverse to speed up arithmetics by turning divisions
 * into multiplications:
 */
I
Ingo Molnar 已提交
1393
static const u32 prio_to_wmult[40] = {
1394 1395 1396 1397 1398 1399 1400 1401
 /* -20 */     48388,     59856,     76040,     92818,    118348,
 /* -15 */    147320,    184698,    229616,    287308,    360437,
 /* -10 */    449829,    563644,    704093,    875809,   1099582,
 /*  -5 */   1376151,   1717300,   2157191,   2708050,   3363326,
 /*   0 */   4194304,   5237765,   6557202,   8165337,  10153587,
 /*   5 */  12820798,  15790321,  19976592,  24970740,  31350126,
 /*  10 */  39045157,  49367440,  61356676,  76695844,  95443717,
 /*  15 */ 119304647, 148102320, 186737708, 238609294, 286331153,
I
Ingo Molnar 已提交
1402
};
1403

1404 1405 1406 1407 1408 1409 1410 1411
/* Time spent by the tasks of the cpu accounting group executing in ... */
enum cpuacct_stat_index {
	CPUACCT_STAT_USER,	/* ... user mode */
	CPUACCT_STAT_SYSTEM,	/* ... kernel mode */

	CPUACCT_STAT_NSTATS,
};

1412 1413
#ifdef CONFIG_CGROUP_CPUACCT
static void cpuacct_charge(struct task_struct *tsk, u64 cputime);
1414 1415
static void cpuacct_update_stats(struct task_struct *tsk,
		enum cpuacct_stat_index idx, cputime_t val);
1416 1417
#else
static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
1418 1419
static inline void cpuacct_update_stats(struct task_struct *tsk,
		enum cpuacct_stat_index idx, cputime_t val) {}
1420 1421
#endif

1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
static inline void inc_cpu_load(struct rq *rq, unsigned long load)
{
	update_load_add(&rq->load, load);
}

static inline void dec_cpu_load(struct rq *rq, unsigned long load)
{
	update_load_sub(&rq->load, load);
}

I
Ingo Molnar 已提交
1432
#if (defined(CONFIG_SMP) && defined(CONFIG_FAIR_GROUP_SCHED)) || defined(CONFIG_RT_GROUP_SCHED)
P
Peter Zijlstra 已提交
1433
typedef int (*tg_visitor)(struct task_group *, void *);
1434 1435 1436 1437 1438

/*
 * Iterate the full tree, calling @down when first entering a node and @up when
 * leaving it for the final time.
 */
P
Peter Zijlstra 已提交
1439
static int walk_tg_tree(tg_visitor down, tg_visitor up, void *data)
1440 1441
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
1442
	int ret;
1443 1444 1445 1446

	rcu_read_lock();
	parent = &root_task_group;
down:
P
Peter Zijlstra 已提交
1447 1448 1449
	ret = (*down)(parent, data);
	if (ret)
		goto out_unlock;
1450 1451 1452 1453 1454 1455 1456
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
1457 1458 1459
	ret = (*up)(parent, data);
	if (ret)
		goto out_unlock;
1460 1461 1462 1463 1464

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
P
Peter Zijlstra 已提交
1465
out_unlock:
1466
	rcu_read_unlock();
P
Peter Zijlstra 已提交
1467 1468

	return ret;
1469 1470
}

P
Peter Zijlstra 已提交
1471 1472 1473
static int tg_nop(struct task_group *tg, void *data)
{
	return 0;
1474
}
P
Peter Zijlstra 已提交
1475 1476 1477
#endif

#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
/* Used instead of source_load when we know the type == 0 */
static unsigned long weighted_cpuload(const int cpu)
{
	return cpu_rq(cpu)->load.weight;
}

/*
 * Return a low guess at the load of a migration-source cpu weighted
 * according to the scheduling class and "nice" value.
 *
 * We want to under-estimate the load of migration sources, to
 * balance conservatively.
 */
static unsigned long source_load(int cpu, int type)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long total = weighted_cpuload(cpu);

	if (type == 0 || !sched_feat(LB_BIAS))
		return total;

	return min(rq->cpu_load[type-1], total);
}

/*
 * Return a high guess at the load of a migration-target cpu weighted
 * according to the scheduling class and "nice" value.
 */
static unsigned long target_load(int cpu, int type)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long total = weighted_cpuload(cpu);

	if (type == 0 || !sched_feat(LB_BIAS))
		return total;

	return max(rq->cpu_load[type-1], total);
}

1517 1518
static unsigned long power_of(int cpu)
{
1519
	return cpu_rq(cpu)->cpu_power;
1520 1521
}

P
Peter Zijlstra 已提交
1522 1523 1524 1525 1526
static int task_hot(struct task_struct *p, u64 now, struct sched_domain *sd);

static unsigned long cpu_avg_load_per_task(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
1527
	unsigned long nr_running = ACCESS_ONCE(rq->nr_running);
P
Peter Zijlstra 已提交
1528

1529 1530
	if (nr_running)
		rq->avg_load_per_task = rq->load.weight / nr_running;
1531 1532
	else
		rq->avg_load_per_task = 0;
P
Peter Zijlstra 已提交
1533 1534 1535 1536 1537

	return rq->avg_load_per_task;
}

#ifdef CONFIG_FAIR_GROUP_SCHED
1538

1539
static __read_mostly unsigned long __percpu *update_shares_data;
1540

1541 1542 1543 1544 1545
static void __set_se_shares(struct sched_entity *se, unsigned long shares);

/*
 * Calculate and set the cpu's group shares.
 */
1546 1547 1548
static void update_group_shares_cpu(struct task_group *tg, int cpu,
				    unsigned long sd_shares,
				    unsigned long sd_rq_weight,
1549
				    unsigned long *usd_rq_weight)
1550
{
1551
	unsigned long shares, rq_weight;
P
Peter Zijlstra 已提交
1552
	int boost = 0;
1553

1554
	rq_weight = usd_rq_weight[cpu];
P
Peter Zijlstra 已提交
1555 1556 1557 1558
	if (!rq_weight) {
		boost = 1;
		rq_weight = NICE_0_LOAD;
	}
1559

1560
	/*
P
Peter Zijlstra 已提交
1561 1562 1563
	 *             \Sum_j shares_j * rq_weight_i
	 * shares_i =  -----------------------------
	 *                  \Sum_j rq_weight_j
1564
	 */
1565
	shares = (sd_shares * rq_weight) / sd_rq_weight;
1566
	shares = clamp_t(unsigned long, shares, MIN_SHARES, MAX_SHARES);
1567

1568 1569 1570 1571
	if (abs(shares - tg->se[cpu]->load.weight) >
			sysctl_sched_shares_thresh) {
		struct rq *rq = cpu_rq(cpu);
		unsigned long flags;
1572

1573
		raw_spin_lock_irqsave(&rq->lock, flags);
1574
		tg->cfs_rq[cpu]->rq_weight = boost ? 0 : rq_weight;
P
Peter Zijlstra 已提交
1575
		tg->cfs_rq[cpu]->shares = boost ? 0 : shares;
1576
		__set_se_shares(tg->se[cpu], shares);
1577
		raw_spin_unlock_irqrestore(&rq->lock, flags);
1578
	}
1579
}
1580 1581

/*
1582 1583 1584
 * Re-compute the task group their per cpu shares over the given domain.
 * This needs to be done in a bottom-up fashion because the rq weight of a
 * parent group depends on the shares of its child groups.
1585
 */
P
Peter Zijlstra 已提交
1586
static int tg_shares_up(struct task_group *tg, void *data)
1587
{
1588
	unsigned long weight, rq_weight = 0, sum_weight = 0, shares = 0;
1589
	unsigned long *usd_rq_weight;
P
Peter Zijlstra 已提交
1590
	struct sched_domain *sd = data;
1591
	unsigned long flags;
1592
	int i;
1593

1594 1595 1596 1597
	if (!tg->se[0])
		return 0;

	local_irq_save(flags);
1598
	usd_rq_weight = per_cpu_ptr(update_shares_data, smp_processor_id());
1599

1600
	for_each_cpu(i, sched_domain_span(sd)) {
1601
		weight = tg->cfs_rq[i]->load.weight;
1602
		usd_rq_weight[i] = weight;
1603

1604
		rq_weight += weight;
1605 1606 1607 1608 1609 1610 1611 1612
		/*
		 * If there are currently no tasks on the cpu pretend there
		 * is one of average load so that when a new task gets to
		 * run here it will not get delayed by group starvation.
		 */
		if (!weight)
			weight = NICE_0_LOAD;

1613
		sum_weight += weight;
1614
		shares += tg->cfs_rq[i]->shares;
1615 1616
	}

1617 1618 1619
	if (!rq_weight)
		rq_weight = sum_weight;

1620 1621 1622 1623 1624
	if ((!shares && rq_weight) || shares > tg->shares)
		shares = tg->shares;

	if (!sd->parent || !(sd->parent->flags & SD_LOAD_BALANCE))
		shares = tg->shares;
1625

1626
	for_each_cpu(i, sched_domain_span(sd))
1627
		update_group_shares_cpu(tg, i, shares, rq_weight, usd_rq_weight);
1628 1629

	local_irq_restore(flags);
P
Peter Zijlstra 已提交
1630 1631

	return 0;
1632 1633 1634
}

/*
1635 1636 1637
 * Compute the cpu's hierarchical load factor for each task group.
 * This needs to be done in a top-down fashion because the load of a child
 * group is a fraction of its parents load.
1638
 */
P
Peter Zijlstra 已提交
1639
static int tg_load_down(struct task_group *tg, void *data)
1640
{
1641
	unsigned long load;
P
Peter Zijlstra 已提交
1642
	long cpu = (long)data;
1643

1644 1645 1646 1647 1648 1649 1650
	if (!tg->parent) {
		load = cpu_rq(cpu)->load.weight;
	} else {
		load = tg->parent->cfs_rq[cpu]->h_load;
		load *= tg->cfs_rq[cpu]->shares;
		load /= tg->parent->cfs_rq[cpu]->load.weight + 1;
	}
1651

1652
	tg->cfs_rq[cpu]->h_load = load;
1653

P
Peter Zijlstra 已提交
1654
	return 0;
1655 1656
}

1657
static void update_shares(struct sched_domain *sd)
1658
{
1659 1660 1661 1662 1663 1664
	s64 elapsed;
	u64 now;

	if (root_task_group_empty())
		return;

1665
	now = local_clock();
1666
	elapsed = now - sd->last_update;
P
Peter Zijlstra 已提交
1667 1668 1669

	if (elapsed >= (s64)(u64)sysctl_sched_shares_ratelimit) {
		sd->last_update = now;
P
Peter Zijlstra 已提交
1670
		walk_tg_tree(tg_nop, tg_shares_up, sd);
P
Peter Zijlstra 已提交
1671
	}
1672 1673
}

P
Peter Zijlstra 已提交
1674
static void update_h_load(long cpu)
1675
{
P
Peter Zijlstra 已提交
1676
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
1677 1678 1679 1680
}

#else

1681
static inline void update_shares(struct sched_domain *sd)
1682 1683 1684
{
}

1685 1686
#endif

1687 1688
#ifdef CONFIG_PREEMPT

1689 1690
static void double_rq_lock(struct rq *rq1, struct rq *rq2);

1691
/*
1692 1693 1694 1695 1696 1697
 * fair double_lock_balance: Safely acquires both rq->locks in a fair
 * way at the expense of forcing extra atomic operations in all
 * invocations.  This assures that the double_lock is acquired using the
 * same underlying policy as the spinlock_t on this architecture, which
 * reduces latency compared to the unfair variant below.  However, it
 * also adds more overhead and therefore may reduce throughput.
1698
 */
1699 1700 1701 1702 1703
static inline int _double_lock_balance(struct rq *this_rq, struct rq *busiest)
	__releases(this_rq->lock)
	__acquires(busiest->lock)
	__acquires(this_rq->lock)
{
1704
	raw_spin_unlock(&this_rq->lock);
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
	double_rq_lock(this_rq, busiest);

	return 1;
}

#else
/*
 * Unfair double_lock_balance: Optimizes throughput at the expense of
 * latency by eliminating extra atomic operations when the locks are
 * already in proper order on entry.  This favors lower cpu-ids and will
 * grant the double lock to lower cpus over higher ids under contention,
 * regardless of entry order into the function.
 */
static int _double_lock_balance(struct rq *this_rq, struct rq *busiest)
1719 1720 1721 1722 1723 1724
	__releases(this_rq->lock)
	__acquires(busiest->lock)
	__acquires(this_rq->lock)
{
	int ret = 0;

1725
	if (unlikely(!raw_spin_trylock(&busiest->lock))) {
1726
		if (busiest < this_rq) {
1727 1728 1729 1730
			raw_spin_unlock(&this_rq->lock);
			raw_spin_lock(&busiest->lock);
			raw_spin_lock_nested(&this_rq->lock,
					      SINGLE_DEPTH_NESTING);
1731 1732
			ret = 1;
		} else
1733 1734
			raw_spin_lock_nested(&busiest->lock,
					      SINGLE_DEPTH_NESTING);
1735 1736 1737 1738
	}
	return ret;
}

1739 1740 1741 1742 1743 1744 1745 1746 1747
#endif /* CONFIG_PREEMPT */

/*
 * double_lock_balance - lock the busiest runqueue, this_rq is locked already.
 */
static int double_lock_balance(struct rq *this_rq, struct rq *busiest)
{
	if (unlikely(!irqs_disabled())) {
		/* printk() doesn't work good under rq->lock */
1748
		raw_spin_unlock(&this_rq->lock);
1749 1750 1751 1752 1753 1754
		BUG_ON(1);
	}

	return _double_lock_balance(this_rq, busiest);
}

1755 1756 1757
static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest)
	__releases(busiest->lock)
{
1758
	raw_spin_unlock(&busiest->lock);
1759 1760
	lock_set_subclass(&this_rq->lock.dep_map, 0, _RET_IP_);
}
1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803

/*
 * double_rq_lock - safely lock two runqueues
 *
 * Note this does not disable interrupts like task_rq_lock,
 * you need to do so manually before calling.
 */
static void double_rq_lock(struct rq *rq1, struct rq *rq2)
	__acquires(rq1->lock)
	__acquires(rq2->lock)
{
	BUG_ON(!irqs_disabled());
	if (rq1 == rq2) {
		raw_spin_lock(&rq1->lock);
		__acquire(rq2->lock);	/* Fake it out ;) */
	} else {
		if (rq1 < rq2) {
			raw_spin_lock(&rq1->lock);
			raw_spin_lock_nested(&rq2->lock, SINGLE_DEPTH_NESTING);
		} else {
			raw_spin_lock(&rq2->lock);
			raw_spin_lock_nested(&rq1->lock, SINGLE_DEPTH_NESTING);
		}
	}
}

/*
 * double_rq_unlock - safely unlock two runqueues
 *
 * Note this does not restore interrupts like task_rq_unlock,
 * you need to do so manually after calling.
 */
static void double_rq_unlock(struct rq *rq1, struct rq *rq2)
	__releases(rq1->lock)
	__releases(rq2->lock)
{
	raw_spin_unlock(&rq1->lock);
	if (rq1 != rq2)
		raw_spin_unlock(&rq2->lock);
	else
		__release(rq2->lock);
}

1804 1805
#endif

V
Vegard Nossum 已提交
1806
#ifdef CONFIG_FAIR_GROUP_SCHED
I
Ingo Molnar 已提交
1807 1808
static void cfs_rq_set_shares(struct cfs_rq *cfs_rq, unsigned long shares)
{
V
Vegard Nossum 已提交
1809
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
1810 1811 1812
	cfs_rq->shares = shares;
#endif
}
V
Vegard Nossum 已提交
1813
#endif
1814

1815
static void calc_load_account_idle(struct rq *this_rq);
1816
static void update_sysctl(void);
1817
static int get_update_sysctl_factor(void);
1818
static void update_cpu_load(struct rq *this_rq);
1819

P
Peter Zijlstra 已提交
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832
static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
{
	set_task_rq(p, cpu);
#ifdef CONFIG_SMP
	/*
	 * After ->cpu is set up to a new value, task_rq_lock(p, ...) can be
	 * successfuly executed on another CPU. We must ensure that updates of
	 * per-task data have been completed by this moment.
	 */
	smp_wmb();
	task_thread_info(p)->cpu = cpu;
#endif
}
1833

1834
static const struct sched_class rt_sched_class;
I
Ingo Molnar 已提交
1835 1836

#define sched_class_highest (&rt_sched_class)
1837 1838
#define for_each_class(class) \
   for (class = sched_class_highest; class; class = class->next)
I
Ingo Molnar 已提交
1839

1840 1841
#include "sched_stats.h"

1842
static void inc_nr_running(struct rq *rq)
1843 1844 1845 1846
{
	rq->nr_running++;
}

1847
static void dec_nr_running(struct rq *rq)
1848 1849 1850 1851
{
	rq->nr_running--;
}

1852 1853 1854
static void set_load_weight(struct task_struct *p)
{
	if (task_has_rt_policy(p)) {
1855 1856
		p->se.load.weight = 0;
		p->se.load.inv_weight = WMULT_CONST;
I
Ingo Molnar 已提交
1857 1858
		return;
	}
1859

I
Ingo Molnar 已提交
1860 1861 1862 1863 1864 1865 1866 1867
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
	if (p->policy == SCHED_IDLE) {
		p->se.load.weight = WEIGHT_IDLEPRIO;
		p->se.load.inv_weight = WMULT_IDLEPRIO;
		return;
	}
1868

I
Ingo Molnar 已提交
1869 1870
	p->se.load.weight = prio_to_weight[p->static_prio - MAX_RT_PRIO];
	p->se.load.inv_weight = prio_to_wmult[p->static_prio - MAX_RT_PRIO];
1871 1872
}

1873
static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
1874
{
1875
	update_rq_clock(rq);
I
Ingo Molnar 已提交
1876
	sched_info_queued(p);
1877
	p->sched_class->enqueue_task(rq, p, flags);
I
Ingo Molnar 已提交
1878
	p->se.on_rq = 1;
1879 1880
}

1881
static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
1882
{
1883
	update_rq_clock(rq);
1884
	sched_info_dequeued(p);
1885
	p->sched_class->dequeue_task(rq, p, flags);
I
Ingo Molnar 已提交
1886
	p->se.on_rq = 0;
1887 1888
}

1889 1890 1891
/*
 * activate_task - move a task to the runqueue.
 */
1892
static void activate_task(struct rq *rq, struct task_struct *p, int flags)
1893 1894 1895 1896
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

1897
	enqueue_task(rq, p, flags);
1898 1899 1900 1901 1902 1903
	inc_nr_running(rq);
}

/*
 * deactivate_task - remove a task from the runqueue.
 */
1904
static void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
1905 1906 1907 1908
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

1909
	dequeue_task(rq, p, flags);
1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
	dec_nr_running(rq);
}

#include "sched_idletask.c"
#include "sched_fair.c"
#include "sched_rt.c"
#ifdef CONFIG_SCHED_DEBUG
# include "sched_debug.c"
#endif

1920
/*
I
Ingo Molnar 已提交
1921
 * __normal_prio - return the priority that is based on the static prio
1922 1923 1924
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
1925
	return p->static_prio;
1926 1927
}

1928 1929 1930 1931 1932 1933 1934
/*
 * 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.
 */
1935
static inline int normal_prio(struct task_struct *p)
1936 1937 1938
{
	int prio;

1939
	if (task_has_rt_policy(p))
1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952
		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.
 */
1953
static int effective_prio(struct task_struct *p)
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
{
	p->normal_prio = normal_prio(p);
	/*
	 * If we are RT tasks or we were boosted to RT priority,
	 * keep the priority unchanged. Otherwise, update priority
	 * to the normal priority:
	 */
	if (!rt_prio(p->prio))
		return p->normal_prio;
	return p->prio;
}

L
Linus Torvalds 已提交
1966 1967 1968 1969
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
1970
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
1971 1972 1973 1974
{
	return cpu_curr(task_cpu(p)) == p;
}

1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
				       int oldprio, int running)
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
			prev_class->switched_from(rq, p, running);
		p->sched_class->switched_to(rq, p, running);
	} else
		p->sched_class->prio_changed(rq, p, oldprio, running);
}

L
Linus Torvalds 已提交
1987
#ifdef CONFIG_SMP
1988 1989 1990
/*
 * Is this task likely cache-hot:
 */
1991
static int
1992 1993 1994 1995
task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
{
	s64 delta;

P
Peter Zijlstra 已提交
1996 1997 1998
	if (p->sched_class != &fair_sched_class)
		return 0;

1999 2000 2001
	/*
	 * Buddy candidates are cache hot:
	 */
2002
	if (sched_feat(CACHE_HOT_BUDDY) && this_rq()->nr_running &&
P
Peter Zijlstra 已提交
2003 2004
			(&p->se == cfs_rq_of(&p->se)->next ||
			 &p->se == cfs_rq_of(&p->se)->last))
2005 2006
		return 1;

2007 2008 2009 2010 2011
	if (sysctl_sched_migration_cost == -1)
		return 1;
	if (sysctl_sched_migration_cost == 0)
		return 0;

2012 2013 2014 2015 2016
	delta = now - p->se.exec_start;

	return delta < (s64)sysctl_sched_migration_cost;
}

I
Ingo Molnar 已提交
2017
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
2018
{
2019 2020 2021 2022 2023
#ifdef CONFIG_SCHED_DEBUG
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
P
Peter Zijlstra 已提交
2024 2025
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
2026 2027
#endif

2028
	trace_sched_migrate_task(p, new_cpu);
2029

2030 2031 2032 2033
	if (task_cpu(p) != new_cpu) {
		p->se.nr_migrations++;
		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, 1, NULL, 0);
	}
I
Ingo Molnar 已提交
2034 2035

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
2036 2037
}

2038
struct migration_arg {
2039
	struct task_struct *task;
L
Linus Torvalds 已提交
2040
	int dest_cpu;
2041
};
L
Linus Torvalds 已提交
2042

2043 2044
static int migration_cpu_stop(void *data);

L
Linus Torvalds 已提交
2045 2046 2047 2048
/*
 * The task's runqueue lock must be held.
 * Returns true if you have to wait for migration thread.
 */
2049
static bool migrate_task(struct task_struct *p, int dest_cpu)
L
Linus Torvalds 已提交
2050
{
2051
	struct rq *rq = task_rq(p);
L
Linus Torvalds 已提交
2052 2053 2054

	/*
	 * If the task is not on a runqueue (and not running), then
2055
	 * the next wake-up will properly place the task.
L
Linus Torvalds 已提交
2056
	 */
2057
	return p->se.on_rq || task_running(rq, p);
L
Linus Torvalds 已提交
2058 2059 2060 2061 2062
}

/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
2063 2064 2065 2066 2067 2068 2069
 * If @match_state is nonzero, it's the @p->state value just checked and
 * not expected to change.  If it changes, i.e. @p might have woken up,
 * then return zero.  When we succeed in waiting for @p to be off its CPU,
 * we return a positive number (its total switch count).  If a second call
 * a short while later returns the same number, the caller can be sure that
 * @p has remained unscheduled the whole time.
 *
L
Linus Torvalds 已提交
2070 2071 2072 2073 2074 2075
 * The caller must ensure that the task *will* unschedule sometime soon,
 * else this function might spin for a *long* time. This function can't
 * be called with interrupts off, or it may introduce deadlock with
 * smp_call_function() if an IPI is sent by the same process we are
 * waiting to become inactive.
 */
R
Roland McGrath 已提交
2076
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
2077 2078
{
	unsigned long flags;
I
Ingo Molnar 已提交
2079
	int running, on_rq;
R
Roland McGrath 已提交
2080
	unsigned long ncsw;
2081
	struct rq *rq;
L
Linus Torvalds 已提交
2082

2083 2084 2085 2086 2087 2088 2089 2090
	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);
2091

2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
		/*
		 * If the task is actively running on another CPU
		 * still, just relax and busy-wait without holding
		 * any locks.
		 *
		 * NOTE! Since we don't hold any locks, it's not
		 * even sure that "rq" stays as the right runqueue!
		 * But we don't care, since "task_running()" will
		 * return false if the runqueue has changed and p
		 * is actually now running somewhere else!
		 */
R
Roland McGrath 已提交
2103 2104 2105
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
2106
			cpu_relax();
R
Roland McGrath 已提交
2107
		}
2108

2109 2110 2111 2112 2113 2114
		/*
		 * 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);
2115
		trace_sched_wait_task(p);
2116 2117
		running = task_running(rq, p);
		on_rq = p->se.on_rq;
R
Roland McGrath 已提交
2118
		ncsw = 0;
2119
		if (!match_state || p->state == match_state)
2120
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
2121
		task_rq_unlock(rq, &flags);
2122

R
Roland McGrath 已提交
2123 2124 2125 2126 2127 2128
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

2129 2130 2131 2132 2133 2134 2135 2136 2137 2138
		/*
		 * 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;
		}
2139

2140 2141 2142 2143 2144
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
2145
		 * So if it was still runnable (but just not actively
2146 2147 2148 2149 2150 2151 2152
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
		if (unlikely(on_rq)) {
			schedule_timeout_uninterruptible(1);
			continue;
		}
2153

2154 2155 2156 2157 2158 2159 2160
		/*
		 * Ahh, all good. It wasn't running, and it wasn't
		 * runnable, which means that it will never become
		 * running in the future either. We're all done!
		 */
		break;
	}
R
Roland McGrath 已提交
2161 2162

	return ncsw;
L
Linus Torvalds 已提交
2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177
}

/***
 * 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.)
 *
 * NOTE: this function doesnt have to take the runqueue lock,
 * 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.
 */
2178
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
2179 2180 2181 2182 2183 2184 2185 2186 2187
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}
R
Rusty Russell 已提交
2188
EXPORT_SYMBOL_GPL(kick_process);
N
Nick Piggin 已提交
2189
#endif /* CONFIG_SMP */
L
Linus Torvalds 已提交
2190

T
Thomas Gleixner 已提交
2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211
/**
 * task_oncpu_function_call - call a function on the cpu on which a task runs
 * @p:		the task to evaluate
 * @func:	the function to be called
 * @info:	the function call argument
 *
 * Calls the function @func when the task is currently running. This might
 * be on the current CPU, which just calls the function directly
 */
void task_oncpu_function_call(struct task_struct *p,
			      void (*func) (void *info), void *info)
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if (task_curr(p))
		smp_call_function_single(cpu, func, info, 1);
	preempt_enable();
}

2212
#ifdef CONFIG_SMP
2213 2214 2215
/*
 * ->cpus_allowed is protected by either TASK_WAKING or rq->lock held.
 */
2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
static int select_fallback_rq(int cpu, struct task_struct *p)
{
	int dest_cpu;
	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(cpu));

	/* Look for allowed, online CPU in same node. */
	for_each_cpu_and(dest_cpu, nodemask, cpu_active_mask)
		if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
			return dest_cpu;

	/* Any allowed, online CPU? */
	dest_cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
	if (dest_cpu < nr_cpu_ids)
		return dest_cpu;

	/* No more Mr. Nice Guy. */
2232
	if (unlikely(dest_cpu >= nr_cpu_ids)) {
2233
		dest_cpu = cpuset_cpus_allowed_fallback(p);
2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
		/*
		 * Don't tell them about moving exiting tasks or
		 * kernel threads (both mm NULL), since they never
		 * leave kernel.
		 */
		if (p->mm && printk_ratelimit()) {
			printk(KERN_INFO "process %d (%s) no "
			       "longer affine to cpu%d\n",
			       task_pid_nr(p), p->comm, cpu);
		}
	}

	return dest_cpu;
}

2249
/*
2250
 * The caller (fork, wakeup) owns TASK_WAKING, ->cpus_allowed is stable.
2251
 */
2252
static inline
2253
int select_task_rq(struct rq *rq, struct task_struct *p, int sd_flags, int wake_flags)
2254
{
2255
	int cpu = p->sched_class->select_task_rq(rq, p, sd_flags, wake_flags);
2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267

	/*
	 * 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 ]
	 */
	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
P
Peter Zijlstra 已提交
2268
		     !cpu_online(cpu)))
2269
		cpu = select_fallback_rq(task_cpu(p), p);
2270 2271

	return cpu;
2272
}
2273 2274 2275 2276 2277 2278

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
2279 2280
#endif

T
Tejun Heo 已提交
2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
static inline void ttwu_activate(struct task_struct *p, struct rq *rq,
				 bool is_sync, bool is_migrate, bool is_local,
				 unsigned long en_flags)
{
	schedstat_inc(p, se.statistics.nr_wakeups);
	if (is_sync)
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
	if (is_migrate)
		schedstat_inc(p, se.statistics.nr_wakeups_migrate);
	if (is_local)
		schedstat_inc(p, se.statistics.nr_wakeups_local);
	else
		schedstat_inc(p, se.statistics.nr_wakeups_remote);

	activate_task(rq, p, en_flags);
}

static inline void ttwu_post_activation(struct task_struct *p, struct rq *rq,
					int wake_flags, bool success)
{
	trace_sched_wakeup(p, success);
	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);

	if (unlikely(rq->idle_stamp)) {
		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
T
Tejun Heo 已提交
2320 2321 2322
	/* if a worker is waking up, notify workqueue */
	if ((p->flags & PF_WQ_WORKER) && success)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
2323 2324 2325
}

/**
L
Linus Torvalds 已提交
2326
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
2327
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
2328
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
2329
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
2330 2331 2332 2333 2334 2335 2336
 *
 * Put it on the run-queue if it's not already there. The "current"
 * thread is always on the run-queue (except when the actual
 * re-schedule is in progress), and as such you're allowed to do
 * the simpler "current->state = TASK_RUNNING" to mark yourself
 * runnable without the overhead of this.
 *
T
Tejun Heo 已提交
2337 2338
 * Returns %true if @p was woken up, %false if it was already running
 * or @state didn't match @p's state.
L
Linus Torvalds 已提交
2339
 */
P
Peter Zijlstra 已提交
2340 2341
static int try_to_wake_up(struct task_struct *p, unsigned int state,
			  int wake_flags)
L
Linus Torvalds 已提交
2342
{
2343
	int cpu, orig_cpu, this_cpu, success = 0;
L
Linus Torvalds 已提交
2344
	unsigned long flags;
2345
	unsigned long en_flags = ENQUEUE_WAKEUP;
2346
	struct rq *rq;
L
Linus Torvalds 已提交
2347

P
Peter Zijlstra 已提交
2348
	this_cpu = get_cpu();
P
Peter Zijlstra 已提交
2349

2350
	smp_wmb();
2351
	rq = task_rq_lock(p, &flags);
P
Peter Zijlstra 已提交
2352
	if (!(p->state & state))
L
Linus Torvalds 已提交
2353 2354
		goto out;

I
Ingo Molnar 已提交
2355
	if (p->se.on_rq)
L
Linus Torvalds 已提交
2356 2357 2358
		goto out_running;

	cpu = task_cpu(p);
2359
	orig_cpu = cpu;
L
Linus Torvalds 已提交
2360 2361 2362 2363 2364

#ifdef CONFIG_SMP
	if (unlikely(task_running(rq, p)))
		goto out_activate;

P
Peter Zijlstra 已提交
2365 2366 2367
	/*
	 * In order to handle concurrent wakeups and release the rq->lock
	 * we put the task in TASK_WAKING state.
2368 2369
	 *
	 * First fix up the nr_uninterruptible count:
P
Peter Zijlstra 已提交
2370
	 */
2371 2372 2373 2374 2375 2376
	if (task_contributes_to_load(p)) {
		if (likely(cpu_online(orig_cpu)))
			rq->nr_uninterruptible--;
		else
			this_rq()->nr_uninterruptible--;
	}
P
Peter Zijlstra 已提交
2377
	p->state = TASK_WAKING;
2378

2379
	if (p->sched_class->task_waking) {
2380
		p->sched_class->task_waking(rq, p);
2381 2382
		en_flags |= ENQUEUE_WAKING;
	}
2383

2384 2385
	cpu = select_task_rq(rq, p, SD_BALANCE_WAKE, wake_flags);
	if (cpu != orig_cpu)
2386
		set_task_cpu(p, cpu);
2387
	__task_rq_unlock(rq);
P
Peter Zijlstra 已提交
2388

2389 2390
	rq = cpu_rq(cpu);
	raw_spin_lock(&rq->lock);
2391

2392 2393 2394 2395 2396 2397 2398
	/*
	 * We migrated the task without holding either rq->lock, however
	 * since the task is not on the task list itself, nobody else
	 * will try and migrate the task, hence the rq should match the
	 * cpu we just moved it to.
	 */
	WARN_ON(task_cpu(p) != cpu);
P
Peter Zijlstra 已提交
2399
	WARN_ON(p->state != TASK_WAKING);
L
Linus Torvalds 已提交
2400

2401 2402 2403 2404 2405 2406 2407
#ifdef CONFIG_SCHEDSTATS
	schedstat_inc(rq, ttwu_count);
	if (cpu == this_cpu)
		schedstat_inc(rq, ttwu_local);
	else {
		struct sched_domain *sd;
		for_each_domain(this_cpu, sd) {
2408
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
2409 2410 2411 2412 2413
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
	}
2414
#endif /* CONFIG_SCHEDSTATS */
2415

L
Linus Torvalds 已提交
2416 2417
out_activate:
#endif /* CONFIG_SMP */
T
Tejun Heo 已提交
2418 2419
	ttwu_activate(p, rq, wake_flags & WF_SYNC, orig_cpu != cpu,
		      cpu == this_cpu, en_flags);
L
Linus Torvalds 已提交
2420 2421
	success = 1;
out_running:
T
Tejun Heo 已提交
2422
	ttwu_post_activation(p, rq, wake_flags, success);
L
Linus Torvalds 已提交
2423 2424
out:
	task_rq_unlock(rq, &flags);
P
Peter Zijlstra 已提交
2425
	put_cpu();
L
Linus Torvalds 已提交
2426 2427 2428 2429

	return success;
}

T
Tejun Heo 已提交
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
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
 * Put @p on the run-queue if it's not alredy there.  The caller must
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
 * the current task.  this_rq() stays locked over invocation.
 */
static void try_to_wake_up_local(struct task_struct *p)
{
	struct rq *rq = task_rq(p);
	bool success = false;

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

	if (!(p->state & TASK_NORMAL))
		return;

	if (!p->se.on_rq) {
		if (likely(!task_running(rq, p))) {
			schedstat_inc(rq, ttwu_count);
			schedstat_inc(rq, ttwu_local);
		}
		ttwu_activate(p, rq, false, false, true, ENQUEUE_WAKEUP);
		success = true;
	}
	ttwu_post_activation(p, rq, 0, success);
}

2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471
/**
 * 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.
 */
2472
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2473
{
2474
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
2475 2476 2477
}
EXPORT_SYMBOL(wake_up_process);

2478
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2479 2480 2481 2482 2483 2484 2485
{
	return try_to_wake_up(p, state, 0);
}

/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
2486 2487 2488 2489 2490 2491 2492
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2493
	p->se.prev_sum_exec_runtime	= 0;
2494
	p->se.nr_migrations		= 0;
I
Ingo Molnar 已提交
2495 2496

#ifdef CONFIG_SCHEDSTATS
2497
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2498
#endif
N
Nick Piggin 已提交
2499

P
Peter Zijlstra 已提交
2500
	INIT_LIST_HEAD(&p->rt.run_list);
I
Ingo Molnar 已提交
2501
	p->se.on_rq = 0;
2502
	INIT_LIST_HEAD(&p->se.group_node);
N
Nick Piggin 已提交
2503

2504 2505 2506
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
I
Ingo Molnar 已提交
2507 2508 2509 2510 2511 2512 2513 2514 2515 2516
}

/*
 * fork()/clone()-time setup:
 */
void sched_fork(struct task_struct *p, int clone_flags)
{
	int cpu = get_cpu();

	__sched_fork(p);
2517
	/*
2518
	 * We mark the process as running here. This guarantees that
2519 2520 2521
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2522
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
2523

2524 2525 2526 2527
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2528
		if (p->policy == SCHED_FIFO || p->policy == SCHED_RR) {
2529
			p->policy = SCHED_NORMAL;
2530 2531
			p->normal_prio = p->static_prio;
		}
2532

2533 2534
		if (PRIO_TO_NICE(p->static_prio) < 0) {
			p->static_prio = NICE_TO_PRIO(0);
2535
			p->normal_prio = p->static_prio;
2536 2537 2538
			set_load_weight(p);
		}

2539 2540 2541 2542 2543 2544
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2545

2546 2547 2548 2549 2550
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

H
Hiroshi Shimamoto 已提交
2551 2552
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
2553

P
Peter Zijlstra 已提交
2554 2555 2556
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

2557 2558 2559 2560 2561 2562 2563 2564
	/*
	 * 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.
	 */
	rcu_read_lock();
2565
	set_task_cpu(p, cpu);
2566
	rcu_read_unlock();
2567

2568
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
I
Ingo Molnar 已提交
2569
	if (likely(sched_info_on()))
2570
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2571
#endif
2572
#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
2573 2574
	p->oncpu = 0;
#endif
L
Linus Torvalds 已提交
2575
#ifdef CONFIG_PREEMPT
2576
	/* Want to start with kernel preemption disabled. */
A
Al Viro 已提交
2577
	task_thread_info(p)->preempt_count = 1;
L
Linus Torvalds 已提交
2578
#endif
2579 2580
	plist_node_init(&p->pushable_tasks, MAX_PRIO);

N
Nick Piggin 已提交
2581
	put_cpu();
L
Linus Torvalds 已提交
2582 2583 2584 2585 2586 2587 2588 2589 2590
}

/*
 * 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.
 */
2591
void wake_up_new_task(struct task_struct *p, unsigned long clone_flags)
L
Linus Torvalds 已提交
2592 2593
{
	unsigned long flags;
I
Ingo Molnar 已提交
2594
	struct rq *rq;
2595
	int cpu __maybe_unused = get_cpu();
2596 2597

#ifdef CONFIG_SMP
2598 2599 2600
	rq = task_rq_lock(p, &flags);
	p->state = TASK_WAKING;

2601 2602 2603 2604 2605
	/*
	 * 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
	 *
2606 2607
	 * We set TASK_WAKING so that select_task_rq() can drop rq->lock
	 * without people poking at ->cpus_allowed.
2608
	 */
2609
	cpu = select_task_rq(rq, p, SD_BALANCE_FORK, 0);
2610
	set_task_cpu(p, cpu);
L
Linus Torvalds 已提交
2611

2612
	p->state = TASK_RUNNING;
2613 2614 2615 2616
	task_rq_unlock(rq, &flags);
#endif

	rq = task_rq_lock(p, &flags);
P
Peter Zijlstra 已提交
2617
	activate_task(rq, p, 0);
2618
	trace_sched_wakeup_new(p, 1);
P
Peter Zijlstra 已提交
2619
	check_preempt_curr(rq, p, WF_FORK);
2620
#ifdef CONFIG_SMP
2621 2622
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
2623
#endif
I
Ingo Molnar 已提交
2624
	task_rq_unlock(rq, &flags);
2625
	put_cpu();
L
Linus Torvalds 已提交
2626 2627
}

2628 2629 2630
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
2631
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2632
 * @notifier: notifier struct to register
2633 2634 2635 2636 2637 2638 2639 2640 2641
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
R
Randy Dunlap 已提交
2642
 * @notifier: notifier struct to unregister
2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671
 *
 * 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);
}

2672
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683

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)
{
}

2684
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2685

2686 2687 2688
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2689
 * @prev: the current task that is being switched out
2690 2691 2692 2693 2694 2695 2696 2697 2698
 * @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.
 */
2699 2700 2701
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2702
{
2703
	fire_sched_out_preempt_notifiers(prev, next);
2704 2705 2706 2707
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2708 2709
/**
 * finish_task_switch - clean up after a task-switch
2710
 * @rq: runqueue associated with task-switch
L
Linus Torvalds 已提交
2711 2712
 * @prev: the thread we just switched away from.
 *
2713 2714 2715 2716
 * finish_task_switch must be called after the context switch, paired
 * with a prepare_task_switch call before the context switch.
 * finish_task_switch will reconcile locking set up by prepare_task_switch,
 * and do any other architecture-specific cleanup actions.
L
Linus Torvalds 已提交
2717 2718
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2719
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2720 2721 2722
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
A
Alexey Dobriyan 已提交
2723
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
L
Linus Torvalds 已提交
2724 2725 2726
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2727
	long prev_state;
L
Linus Torvalds 已提交
2728 2729 2730 2731 2732

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2733
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2734 2735
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2736
	 * The test for TASK_DEAD must occur while the runqueue locks are
L
Linus Torvalds 已提交
2737 2738 2739 2740 2741
	 * still held, otherwise prev could be scheduled on another cpu, die
	 * there before we look at prev->state, and then the reference would
	 * be dropped twice.
	 *		Manfred Spraul <manfred@colorfullife.com>
	 */
O
Oleg Nesterov 已提交
2742
	prev_state = prev->state;
2743
	finish_arch_switch(prev);
2744 2745 2746
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_disable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
2747
	perf_event_task_sched_in(current);
2748 2749 2750
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
2751
	finish_lock_switch(rq, prev);
S
Steven Rostedt 已提交
2752

2753
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2754 2755
	if (mm)
		mmdrop(mm);
2756
	if (unlikely(prev_state == TASK_DEAD)) {
2757 2758 2759
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2760
		 */
2761
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
2762
		put_task_struct(prev);
2763
	}
L
Linus Torvalds 已提交
2764 2765
}

2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
#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;

2781
		raw_spin_lock_irqsave(&rq->lock, flags);
2782 2783
		if (rq->curr->sched_class->post_schedule)
			rq->curr->sched_class->post_schedule(rq);
2784
		raw_spin_unlock_irqrestore(&rq->lock, flags);
2785 2786 2787 2788 2789 2790

		rq->post_schedule = 0;
	}
}

#else
2791

2792 2793 2794 2795 2796 2797
static inline void pre_schedule(struct rq *rq, struct task_struct *p)
{
}

static inline void post_schedule(struct rq *rq)
{
L
Linus Torvalds 已提交
2798 2799
}

2800 2801
#endif

L
Linus Torvalds 已提交
2802 2803 2804 2805
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2806
asmlinkage void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2807 2808
	__releases(rq->lock)
{
2809 2810
	struct rq *rq = this_rq();

2811
	finish_task_switch(rq, prev);
2812

2813 2814 2815 2816 2817
	/*
	 * FIXME: do we need to worry about rq being invalidated by the
	 * task_switch?
	 */
	post_schedule(rq);
2818

2819 2820 2821 2822
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
L
Linus Torvalds 已提交
2823
	if (current->set_child_tid)
2824
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2825 2826 2827 2828 2829 2830
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
I
Ingo Molnar 已提交
2831
static inline void
2832
context_switch(struct rq *rq, struct task_struct *prev,
2833
	       struct task_struct *next)
L
Linus Torvalds 已提交
2834
{
I
Ingo Molnar 已提交
2835
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2836

2837
	prepare_task_switch(rq, prev, next);
2838
	trace_sched_switch(prev, next);
I
Ingo Molnar 已提交
2839 2840
	mm = next->mm;
	oldmm = prev->active_mm;
2841 2842 2843 2844 2845
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2846
	arch_start_context_switch(prev);
2847

2848
	if (likely(!mm)) {
L
Linus Torvalds 已提交
2849 2850 2851 2852 2853 2854
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2855
	if (likely(!prev->mm)) {
L
Linus Torvalds 已提交
2856 2857 2858
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2859 2860 2861 2862 2863 2864 2865
	/*
	 * 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
2866
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2867
#endif
L
Linus Torvalds 已提交
2868 2869 2870 2871

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

I
Ingo Molnar 已提交
2872 2873 2874 2875 2876 2877 2878
	barrier();
	/*
	 * this_rq must be evaluated again because prev may have moved
	 * CPUs since it called schedule(), thus the 'rq' on its stack
	 * frame will be invalid.
	 */
	finish_task_switch(this_rq(), prev);
L
Linus Torvalds 已提交
2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895
}

/*
 * 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;
2896
}
L
Linus Torvalds 已提交
2897 2898

unsigned long nr_uninterruptible(void)
2899
{
L
Linus Torvalds 已提交
2900
	unsigned long i, sum = 0;
2901

2902
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2903
		sum += cpu_rq(i)->nr_uninterruptible;
2904 2905

	/*
L
Linus Torvalds 已提交
2906 2907
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
2908
	 */
L
Linus Torvalds 已提交
2909 2910
	if (unlikely((long)sum < 0))
		sum = 0;
2911

L
Linus Torvalds 已提交
2912
	return sum;
2913 2914
}

L
Linus Torvalds 已提交
2915
unsigned long long nr_context_switches(void)
2916
{
2917 2918
	int i;
	unsigned long long sum = 0;
2919

2920
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2921
		sum += cpu_rq(i)->nr_switches;
2922

L
Linus Torvalds 已提交
2923 2924
	return sum;
}
2925

L
Linus Torvalds 已提交
2926 2927 2928
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2929

2930
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2931
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2932

L
Linus Torvalds 已提交
2933 2934
	return sum;
}
2935

2936
unsigned long nr_iowait_cpu(int cpu)
2937
{
2938
	struct rq *this = cpu_rq(cpu);
2939 2940
	return atomic_read(&this->nr_iowait);
}
2941

2942 2943 2944 2945 2946
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
2947

2948

2949 2950 2951 2952 2953
/* 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);
2954

2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009
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;
}

#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;

static void calc_load_account_idle(struct rq *this_rq)
{
	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;
}
#else
static void calc_load_account_idle(struct rq *this_rq)
{
}

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

3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022
/**
 * 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;
3023 3024
}

3025 3026
static unsigned long
calc_load(unsigned long load, unsigned long exp, unsigned long active)
3027
{
3028 3029 3030 3031
	load *= exp;
	load += active * (FIXED_1 - exp);
	return load >> FSHIFT;
}
3032 3033

/*
3034 3035
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
3036
 */
3037
void calc_global_load(void)
3038
{
3039 3040
	unsigned long upd = calc_load_update + 10;
	long active;
L
Linus Torvalds 已提交
3041

3042 3043
	if (time_before(jiffies, upd))
		return;
L
Linus Torvalds 已提交
3044

3045 3046
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
L
Linus Torvalds 已提交
3047

3048 3049 3050
	avenrun[0] = calc_load(avenrun[0], EXP_1, active);
	avenrun[1] = calc_load(avenrun[1], EXP_5, active);
	avenrun[2] = calc_load(avenrun[2], EXP_15, active);
I
Ingo Molnar 已提交
3051

3052 3053
	calc_load_update += LOAD_FREQ;
}
L
Linus Torvalds 已提交
3054

3055
/*
3056 3057
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
3058 3059 3060
 */
static void calc_load_account_active(struct rq *this_rq)
{
3061
	long delta;
3062

3063 3064
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
3065

3066 3067 3068
	delta  = calc_load_fold_active(this_rq);
	delta += calc_load_fold_idle();
	if (delta)
3069
		atomic_long_add(delta, &calc_load_tasks);
3070 3071

	this_rq->calc_load_update += LOAD_FREQ;
3072 3073
}

3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140
/*
 * 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;
}

3141
/*
I
Ingo Molnar 已提交
3142
 * Update rq->cpu_load[] statistics. This function is usually called every
3143 3144
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
3145
 */
I
Ingo Molnar 已提交
3146
static void update_cpu_load(struct rq *this_rq)
3147
{
3148
	unsigned long this_load = this_rq->load.weight;
3149 3150
	unsigned long curr_jiffies = jiffies;
	unsigned long pending_updates;
I
Ingo Molnar 已提交
3151
	int i, scale;
3152

I
Ingo Molnar 已提交
3153
	this_rq->nr_load_updates++;
3154

3155 3156 3157 3158 3159 3160 3161
	/* Avoid repeated calls on same jiffy, when moving in and out of idle */
	if (curr_jiffies == this_rq->last_load_update_tick)
		return;

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

I
Ingo Molnar 已提交
3162
	/* Update our load: */
3163 3164
	this_rq->cpu_load[0] = this_load; /* Fasttrack for idx 0 */
	for (i = 1, scale = 2; i < CPU_LOAD_IDX_MAX; i++, scale += scale) {
I
Ingo Molnar 已提交
3165
		unsigned long old_load, new_load;
3166

I
Ingo Molnar 已提交
3167
		/* scale is effectively 1 << i now, and >> i divides by scale */
3168

I
Ingo Molnar 已提交
3169
		old_load = this_rq->cpu_load[i];
3170
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
I
Ingo Molnar 已提交
3171
		new_load = this_load;
I
Ingo Molnar 已提交
3172 3173 3174 3175 3176 3177
		/*
		 * 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)
3178 3179 3180
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
I
Ingo Molnar 已提交
3181
	}
3182 3183 3184 3185 3186
}

static void update_cpu_load_active(struct rq *this_rq)
{
	update_cpu_load(this_rq);
3187

3188
	calc_load_account_active(this_rq);
3189 3190
}

I
Ingo Molnar 已提交
3191
#ifdef CONFIG_SMP
3192

3193
/*
P
Peter Zijlstra 已提交
3194 3195
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
3196
 */
P
Peter Zijlstra 已提交
3197
void sched_exec(void)
3198
{
P
Peter Zijlstra 已提交
3199
	struct task_struct *p = current;
L
Linus Torvalds 已提交
3200
	unsigned long flags;
3201
	struct rq *rq;
3202
	int dest_cpu;
3203

L
Linus Torvalds 已提交
3204
	rq = task_rq_lock(p, &flags);
3205 3206 3207
	dest_cpu = p->sched_class->select_task_rq(rq, p, SD_BALANCE_EXEC, 0);
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
3208

3209
	/*
P
Peter Zijlstra 已提交
3210
	 * select_task_rq() can race against ->cpus_allowed
3211
	 */
3212
	if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed) &&
3213 3214
	    likely(cpu_active(dest_cpu)) && migrate_task(p, dest_cpu)) {
		struct migration_arg arg = { p, dest_cpu };
3215

L
Linus Torvalds 已提交
3216
		task_rq_unlock(rq, &flags);
3217
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3218 3219
		return;
	}
3220
unlock:
L
Linus Torvalds 已提交
3221 3222
	task_rq_unlock(rq, &flags);
}
I
Ingo Molnar 已提交
3223

L
Linus Torvalds 已提交
3224 3225 3226 3227 3228 3229 3230
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);

EXPORT_PER_CPU_SYMBOL(kstat);

/*
3231
 * Return any ns on the sched_clock that have not yet been accounted in
3232
 * @p in case that task is currently running.
3233 3234
 *
 * Called with task_rq_lock() held on @rq.
L
Linus Torvalds 已提交
3235
 */
3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249
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);
		ns = rq->clock - p->se.exec_start;
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

3250
unsigned long long task_delta_exec(struct task_struct *p)
L
Linus Torvalds 已提交
3251 3252
{
	unsigned long flags;
3253
	struct rq *rq;
3254
	u64 ns = 0;
3255

3256
	rq = task_rq_lock(p, &flags);
3257 3258
	ns = do_task_delta_exec(p, rq);
	task_rq_unlock(rq, &flags);
3259

3260 3261
	return ns;
}
3262

3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279
/*
 * 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);
	task_rq_unlock(rq, &flags);

	return ns;
}
3280

3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299
/*
 * Return sum_exec_runtime for the thread group.
 * In case the task is currently running, return the sum plus current's
 * pending runtime that have not been accounted yet.
 *
 * Note that the thread group might have other running tasks as well,
 * so the return value not includes other pending runtime that other
 * running tasks might have.
 */
unsigned long long thread_group_sched_runtime(struct task_struct *p)
{
	struct task_cputime totals;
	unsigned long flags;
	struct rq *rq;
	u64 ns;

	rq = task_rq_lock(p, &flags);
	thread_group_cputime(p, &totals);
	ns = totals.sum_exec_runtime + do_task_delta_exec(p, rq);
3300
	task_rq_unlock(rq, &flags);
3301

L
Linus Torvalds 已提交
3302 3303 3304 3305 3306 3307 3308
	return ns;
}

/*
 * 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
3309
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
3310
 */
3311 3312
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
L
Linus Torvalds 已提交
3313 3314 3315 3316
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t tmp;

3317
	/* Add user time to process. */
L
Linus Torvalds 已提交
3318
	p->utime = cputime_add(p->utime, cputime);
3319
	p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
3320
	account_group_user_time(p, cputime);
L
Linus Torvalds 已提交
3321 3322 3323 3324 3325 3326 3327

	/* Add user time to cpustat. */
	tmp = cputime_to_cputime64(cputime);
	if (TASK_NICE(p) > 0)
		cpustat->nice = cputime64_add(cpustat->nice, tmp);
	else
		cpustat->user = cputime64_add(cpustat->user, tmp);
3328 3329

	cpuacct_update_stats(p, CPUACCT_STAT_USER, cputime);
3330 3331
	/* Account for user time used */
	acct_update_integrals(p);
L
Linus Torvalds 已提交
3332 3333
}

3334 3335 3336 3337
/*
 * 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
3338
 * @cputime_scaled: cputime scaled by cpu frequency
3339
 */
3340 3341
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
3342 3343 3344 3345 3346 3347
{
	cputime64_t tmp;
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;

	tmp = cputime_to_cputime64(cputime);

3348
	/* Add guest time to process. */
3349
	p->utime = cputime_add(p->utime, cputime);
3350
	p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
3351
	account_group_user_time(p, cputime);
3352 3353
	p->gtime = cputime_add(p->gtime, cputime);

3354
	/* Add guest time to cpustat. */
3355 3356 3357 3358 3359 3360 3361
	if (TASK_NICE(p) > 0) {
		cpustat->nice = cputime64_add(cpustat->nice, tmp);
		cpustat->guest_nice = cputime64_add(cpustat->guest_nice, tmp);
	} else {
		cpustat->user = cputime64_add(cpustat->user, tmp);
		cpustat->guest = cputime64_add(cpustat->guest, tmp);
	}
3362 3363
}

L
Linus Torvalds 已提交
3364 3365 3366 3367 3368
/*
 * 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
3369
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
3370 3371
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
3372
			 cputime_t cputime, cputime_t cputime_scaled)
L
Linus Torvalds 已提交
3373 3374 3375 3376
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t tmp;

3377
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
3378
		account_guest_time(p, cputime, cputime_scaled);
3379 3380
		return;
	}
3381

3382
	/* Add system time to process. */
L
Linus Torvalds 已提交
3383
	p->stime = cputime_add(p->stime, cputime);
3384
	p->stimescaled = cputime_add(p->stimescaled, cputime_scaled);
3385
	account_group_system_time(p, cputime);
L
Linus Torvalds 已提交
3386 3387 3388 3389 3390 3391 3392 3393

	/* Add system time to cpustat. */
	tmp = cputime_to_cputime64(cputime);
	if (hardirq_count() - hardirq_offset)
		cpustat->irq = cputime64_add(cpustat->irq, tmp);
	else if (softirq_count())
		cpustat->softirq = cputime64_add(cpustat->softirq, tmp);
	else
3394 3395
		cpustat->system = cputime64_add(cpustat->system, tmp);

3396 3397
	cpuacct_update_stats(p, CPUACCT_STAT_SYSTEM, cputime);

L
Linus Torvalds 已提交
3398 3399 3400 3401
	/* Account for system time used */
	acct_update_integrals(p);
}

3402
/*
L
Linus Torvalds 已提交
3403 3404
 * Account for involuntary wait time.
 * @steal: the cpu time spent in involuntary wait
3405
 */
3406
void account_steal_time(cputime_t cputime)
3407
{
3408 3409 3410 3411
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t cputime64 = cputime_to_cputime64(cputime);

	cpustat->steal = cputime64_add(cpustat->steal, cputime64);
3412 3413
}

L
Linus Torvalds 已提交
3414
/*
3415 3416
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
L
Linus Torvalds 已提交
3417
 */
3418
void account_idle_time(cputime_t cputime)
L
Linus Torvalds 已提交
3419 3420
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
3421
	cputime64_t cputime64 = cputime_to_cputime64(cputime);
3422
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
3423

3424 3425 3426 3427
	if (atomic_read(&rq->nr_iowait) > 0)
		cpustat->iowait = cputime64_add(cpustat->iowait, cputime64);
	else
		cpustat->idle = cputime64_add(cpustat->idle, cputime64);
L
Linus Torvalds 已提交
3428 3429
}

3430 3431 3432 3433 3434 3435 3436 3437 3438
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

/*
 * 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)
{
3439
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
3440 3441 3442
	struct rq *rq = this_rq();

	if (user_tick)
3443
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
3444
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
3445
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
3446 3447
				    one_jiffy_scaled);
	else
3448
		account_idle_time(cputime_one_jiffy);
3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467
}

/*
 * 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)
{
	account_idle_time(jiffies_to_cputime(ticks));
L
Linus Torvalds 已提交
3468 3469
}

3470 3471
#endif

3472 3473 3474 3475
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
3476
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3477
{
3478 3479
	*ut = p->utime;
	*st = p->stime;
3480 3481
}

3482
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3483
{
3484 3485 3486 3487 3488 3489
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
3490 3491
}
#else
3492 3493

#ifndef nsecs_to_cputime
3494
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
3495 3496
#endif

3497
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3498
{
3499
	cputime_t rtime, utime = p->utime, total = cputime_add(utime, p->stime);
3500 3501 3502 3503

	/*
	 * Use CFS's precise accounting:
	 */
3504
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
3505 3506

	if (total) {
3507 3508 3509
		u64 temp;

		temp = (u64)(rtime * utime);
3510
		do_div(temp, total);
3511 3512 3513
		utime = (cputime_t)temp;
	} else
		utime = rtime;
3514

3515 3516 3517
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
3518
	p->prev_utime = max(p->prev_utime, utime);
3519
	p->prev_stime = max(p->prev_stime, cputime_sub(rtime, p->prev_utime));
3520

3521 3522
	*ut = p->prev_utime;
	*st = p->prev_stime;
3523 3524
}

3525 3526 3527 3528
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3529
{
3530 3531 3532
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
3533

3534
	thread_group_cputime(p, &cputime);
3535

3536 3537
	total = cputime_add(cputime.utime, cputime.stime);
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
3538

3539 3540
	if (total) {
		u64 temp;
3541

3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553
		temp = (u64)(rtime * cputime.utime);
		do_div(temp, total);
		utime = (cputime_t)temp;
	} else
		utime = rtime;

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

	*ut = sig->prev_utime;
	*st = sig->prev_stime;
3554 3555 3556
}
#endif

3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567
/*
 * This function gets called by the timer code, with HZ frequency.
 * We call it with interrupts disabled.
 *
 * It also gets called by the fork code, when changing the parent's
 * timeslices.
 */
void scheduler_tick(void)
{
	int cpu = smp_processor_id();
	struct rq *rq = cpu_rq(cpu);
I
Ingo Molnar 已提交
3568
	struct task_struct *curr = rq->curr;
3569 3570

	sched_clock_tick();
I
Ingo Molnar 已提交
3571

3572
	raw_spin_lock(&rq->lock);
3573
	update_rq_clock(rq);
3574
	update_cpu_load_active(rq);
P
Peter Zijlstra 已提交
3575
	curr->sched_class->task_tick(rq, curr, 0);
3576
	raw_spin_unlock(&rq->lock);
3577

3578
	perf_event_task_tick(curr);
3579

3580
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
3581 3582
	rq->idle_at_tick = idle_cpu(cpu);
	trigger_load_balance(rq, cpu);
3583
#endif
L
Linus Torvalds 已提交
3584 3585
}

3586
notrace unsigned long get_parent_ip(unsigned long addr)
3587 3588 3589 3590 3591 3592 3593 3594
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
3595

3596 3597 3598
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

3599
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
3600
{
3601
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3602 3603 3604
	/*
	 * Underflow?
	 */
3605 3606
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3607
#endif
L
Linus Torvalds 已提交
3608
	preempt_count() += val;
3609
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3610 3611 3612
	/*
	 * Spinlock count overflowing soon?
	 */
3613 3614
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3615 3616 3617
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3618 3619 3620
}
EXPORT_SYMBOL(add_preempt_count);

3621
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
3622
{
3623
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3624 3625 3626
	/*
	 * Underflow?
	 */
3627
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3628
		return;
L
Linus Torvalds 已提交
3629 3630 3631
	/*
	 * Is the spinlock portion underflowing?
	 */
3632 3633 3634
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3635
#endif
3636

3637 3638
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3639 3640 3641 3642 3643 3644 3645
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
3646
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3647
 */
I
Ingo Molnar 已提交
3648
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3649
{
3650 3651
	struct pt_regs *regs = get_irq_regs();

P
Peter Zijlstra 已提交
3652 3653
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3654

I
Ingo Molnar 已提交
3655
	debug_show_held_locks(prev);
3656
	print_modules();
I
Ingo Molnar 已提交
3657 3658
	if (irqs_disabled())
		print_irqtrace_events(prev);
3659 3660 3661 3662 3663

	if (regs)
		show_regs(regs);
	else
		dump_stack();
I
Ingo Molnar 已提交
3664
}
L
Linus Torvalds 已提交
3665

I
Ingo Molnar 已提交
3666 3667 3668 3669 3670
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
3671
	/*
I
Ingo Molnar 已提交
3672
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
3673 3674 3675
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
3676
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
I
Ingo Molnar 已提交
3677 3678
		__schedule_bug(prev);

L
Linus Torvalds 已提交
3679 3680
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3681
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
3682 3683
#ifdef CONFIG_SCHEDSTATS
	if (unlikely(prev->lock_depth >= 0)) {
3684 3685
		schedstat_inc(this_rq(), bkl_count);
		schedstat_inc(prev, sched_info.bkl_count);
I
Ingo Molnar 已提交
3686 3687
	}
#endif
I
Ingo Molnar 已提交
3688 3689
}

P
Peter Zijlstra 已提交
3690
static void put_prev_task(struct rq *rq, struct task_struct *prev)
M
Mike Galbraith 已提交
3691
{
3692 3693 3694
	if (prev->se.on_rq)
		update_rq_clock(rq);
	rq->skip_clock_update = 0;
P
Peter Zijlstra 已提交
3695
	prev->sched_class->put_prev_task(rq, prev);
M
Mike Galbraith 已提交
3696 3697
}

I
Ingo Molnar 已提交
3698 3699 3700 3701
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3702
pick_next_task(struct rq *rq)
I
Ingo Molnar 已提交
3703
{
3704
	const struct sched_class *class;
I
Ingo Molnar 已提交
3705
	struct task_struct *p;
L
Linus Torvalds 已提交
3706 3707

	/*
I
Ingo Molnar 已提交
3708 3709
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3710
	 */
I
Ingo Molnar 已提交
3711
	if (likely(rq->nr_running == rq->cfs.nr_running)) {
3712
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
3713 3714
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
3715 3716
	}

I
Ingo Molnar 已提交
3717 3718
	class = sched_class_highest;
	for ( ; ; ) {
3719
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
3720 3721 3722 3723 3724 3725 3726 3727 3728
		if (p)
			return p;
		/*
		 * Will never be NULL as the idle class always
		 * returns a non-NULL p:
		 */
		class = class->next;
	}
}
L
Linus Torvalds 已提交
3729

I
Ingo Molnar 已提交
3730 3731 3732
/*
 * schedule() is the main scheduler function.
 */
3733
asmlinkage void __sched schedule(void)
I
Ingo Molnar 已提交
3734 3735
{
	struct task_struct *prev, *next;
3736
	unsigned long *switch_count;
I
Ingo Molnar 已提交
3737
	struct rq *rq;
3738
	int cpu;
I
Ingo Molnar 已提交
3739

3740 3741
need_resched:
	preempt_disable();
I
Ingo Molnar 已提交
3742 3743
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3744
	rcu_note_context_switch(cpu);
I
Ingo Molnar 已提交
3745 3746 3747 3748 3749 3750
	prev = rq->curr;

	release_kernel_lock(prev);
need_resched_nonpreemptible:

	schedule_debug(prev);
L
Linus Torvalds 已提交
3751

3752
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3753
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3754

3755
	raw_spin_lock_irq(&rq->lock);
3756
	clear_tsk_need_resched(prev);
L
Linus Torvalds 已提交
3757

3758
	switch_count = &prev->nivcsw;
L
Linus Torvalds 已提交
3759
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
T
Tejun Heo 已提交
3760
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3761
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775
		} else {
			/*
			 * If a worker is going to sleep, notify and
			 * ask workqueue whether it wants to wake up a
			 * task to maintain concurrency.  If so, wake
			 * up the task.
			 */
			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);
			}
3776
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
T
Tejun Heo 已提交
3777
		}
I
Ingo Molnar 已提交
3778
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3779 3780
	}

3781
	pre_schedule(rq, prev);
3782

I
Ingo Molnar 已提交
3783
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
3784 3785
		idle_balance(cpu, rq);

M
Mike Galbraith 已提交
3786
	put_prev_task(rq, prev);
3787
	next = pick_next_task(rq);
L
Linus Torvalds 已提交
3788 3789

	if (likely(prev != next)) {
3790
		sched_info_switch(prev, next);
3791
		perf_event_task_sched_out(prev, next);
3792

L
Linus Torvalds 已提交
3793 3794 3795 3796
		rq->nr_switches++;
		rq->curr = next;
		++*switch_count;

I
Ingo Molnar 已提交
3797
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
3798
		/*
3799 3800 3801 3802
		 * The context switch have flipped the stack from under us
		 * and restored the local variables which were saved when
		 * this task called schedule() in the past. prev == current
		 * is still correct, but it can be moved to another cpu/rq.
P
Peter Zijlstra 已提交
3803 3804 3805
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
3806
	} else
3807
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
3808

3809
	post_schedule(rq);
L
Linus Torvalds 已提交
3810

3811
	if (unlikely(reacquire_kernel_lock(prev)))
L
Linus Torvalds 已提交
3812
		goto need_resched_nonpreemptible;
P
Peter Zijlstra 已提交
3813

L
Linus Torvalds 已提交
3814
	preempt_enable_no_resched();
3815
	if (need_resched())
L
Linus Torvalds 已提交
3816 3817 3818 3819
		goto need_resched;
}
EXPORT_SYMBOL(schedule);

3820
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839
/*
 * Look out! "owner" is an entirely speculative pointer
 * access and not reliable.
 */
int mutex_spin_on_owner(struct mutex *lock, struct thread_info *owner)
{
	unsigned int cpu;
	struct rq *rq;

	if (!sched_feat(OWNER_SPIN))
		return 0;

#ifdef CONFIG_DEBUG_PAGEALLOC
	/*
	 * Need to access the cpu field knowing that
	 * DEBUG_PAGEALLOC could have unmapped it if
	 * the mutex owner just released it and exited.
	 */
	if (probe_kernel_address(&owner->cpu, cpu))
3840
		return 0;
3841 3842 3843 3844 3845 3846 3847 3848 3849
#else
	cpu = owner->cpu;
#endif

	/*
	 * Even if the access succeeded (likely case),
	 * the cpu field may no longer be valid.
	 */
	if (cpu >= nr_cpumask_bits)
3850
		return 0;
3851 3852 3853 3854 3855 3856

	/*
	 * We need to validate that we can do a
	 * get_cpu() and that we have the percpu area.
	 */
	if (!cpu_online(cpu))
3857
		return 0;
3858 3859 3860 3861 3862 3863 3864

	rq = cpu_rq(cpu);

	for (;;) {
		/*
		 * Owner changed, break to re-assess state.
		 */
3865 3866 3867 3868 3869 3870 3871 3872
		if (lock->owner != owner) {
			/*
			 * If the lock has switched to a different owner,
			 * we likely have heavy contention. Return 0 to quit
			 * optimistic spinning and not contend further:
			 */
			if (lock->owner)
				return 0;
3873
			break;
3874
		}
3875 3876 3877 3878 3879 3880 3881 3882 3883

		/*
		 * Is that owner really running on that cpu?
		 */
		if (task_thread_info(rq->curr) != owner || need_resched())
			return 0;

		cpu_relax();
	}
3884

3885 3886 3887 3888
	return 1;
}
#endif

L
Linus Torvalds 已提交
3889 3890
#ifdef CONFIG_PREEMPT
/*
3891
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3892
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3893 3894
 * occur there and call schedule directly.
 */
3895
asmlinkage void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3896 3897
{
	struct thread_info *ti = current_thread_info();
3898

L
Linus Torvalds 已提交
3899 3900
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3901
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3902
	 */
N
Nick Piggin 已提交
3903
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
3904 3905
		return;

3906
	do {
3907
		add_preempt_count_notrace(PREEMPT_ACTIVE);
3908
		schedule();
3909
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3910

3911 3912 3913 3914 3915
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3916
	} while (need_resched());
L
Linus Torvalds 已提交
3917 3918 3919 3920
}
EXPORT_SYMBOL(preempt_schedule);

/*
3921
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3922 3923 3924 3925 3926 3927 3928
 * 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();
3929

3930
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
3931 3932
	BUG_ON(ti->preempt_count || !irqs_disabled());

3933 3934 3935 3936 3937 3938
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
		schedule();
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3939

3940 3941 3942 3943 3944
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3945
	} while (need_resched());
L
Linus Torvalds 已提交
3946 3947 3948 3949
}

#endif /* CONFIG_PREEMPT */

P
Peter Zijlstra 已提交
3950
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3951
			  void *key)
L
Linus Torvalds 已提交
3952
{
P
Peter Zijlstra 已提交
3953
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3954 3955 3956 3957
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
3958 3959
 * The core wakeup function. Non-exclusive wakeups (nr_exclusive == 0) just
 * wake everything up. If it's an exclusive wakeup (nr_exclusive == small +ve
L
Linus Torvalds 已提交
3960 3961 3962
 * number) then we wake all the non-exclusive tasks and one exclusive task.
 *
 * There are circumstances in which we can try to wake a task which has already
I
Ingo Molnar 已提交
3963
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
3964 3965
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
3966
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
P
Peter Zijlstra 已提交
3967
			int nr_exclusive, int wake_flags, void *key)
L
Linus Torvalds 已提交
3968
{
3969
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
3970

3971
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
3972 3973
		unsigned flags = curr->flags;

P
Peter Zijlstra 已提交
3974
		if (curr->func(curr, mode, wake_flags, key) &&
3975
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
3976 3977 3978 3979 3980 3981 3982 3983 3984
			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
3985
 * @key: is directly passed to the wakeup function
3986 3987 3988
 *
 * It may be assumed that this function implies a write memory barrier before
 * changing the task state if and only if any tasks are woken up.
L
Linus Torvalds 已提交
3989
 */
3990
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
3991
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003
{
	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.
 */
4004
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode)
L
Linus Torvalds 已提交
4005 4006 4007
{
	__wake_up_common(q, mode, 1, 0, NULL);
}
4008
EXPORT_SYMBOL_GPL(__wake_up_locked);
L
Linus Torvalds 已提交
4009

4010 4011 4012 4013 4014
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}

L
Linus Torvalds 已提交
4015
/**
4016
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
4017 4018 4019
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
4020
 * @key: opaque value to be passed to wakeup targets
L
Linus Torvalds 已提交
4021 4022 4023 4024 4025 4026 4027
 *
 * 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.
4028 4029 4030
 *
 * It may be assumed that this function implies a write memory barrier before
 * changing the task state if and only if any tasks are woken up.
L
Linus Torvalds 已提交
4031
 */
4032 4033
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
4034 4035
{
	unsigned long flags;
P
Peter Zijlstra 已提交
4036
	int wake_flags = WF_SYNC;
L
Linus Torvalds 已提交
4037 4038 4039 4040 4041

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
P
Peter Zijlstra 已提交
4042
		wake_flags = 0;
L
Linus Torvalds 已提交
4043 4044

	spin_lock_irqsave(&q->lock, flags);
P
Peter Zijlstra 已提交
4045
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
L
Linus Torvalds 已提交
4046 4047
	spin_unlock_irqrestore(&q->lock, flags);
}
4048 4049 4050 4051 4052 4053 4054 4055 4056
EXPORT_SYMBOL_GPL(__wake_up_sync_key);

/*
 * __wake_up_sync - see __wake_up_sync_key()
 */
void __wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr_exclusive)
{
	__wake_up_sync_key(q, mode, nr_exclusive, NULL);
}
L
Linus Torvalds 已提交
4057 4058
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

4059 4060 4061 4062 4063 4064 4065 4066
/**
 * 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.
4067 4068 4069
 *
 * 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.
4070
 */
4071
void complete(struct completion *x)
L
Linus Torvalds 已提交
4072 4073 4074 4075 4076
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
4077
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
4078 4079 4080 4081
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

4082 4083 4084 4085 4086
/**
 * 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.
4087 4088 4089
 *
 * 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.
4090
 */
4091
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
4092 4093 4094 4095 4096
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
4097
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
4098 4099 4100 4101
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

4102 4103
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
4104 4105 4106 4107
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

C
Changli Gao 已提交
4108
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
L
Linus Torvalds 已提交
4109
		do {
4110
			if (signal_pending_state(state, current)) {
4111 4112
				timeout = -ERESTARTSYS;
				break;
4113 4114
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
4115 4116 4117
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
4118
		} while (!x->done && timeout);
L
Linus Torvalds 已提交
4119
		__remove_wait_queue(&x->wait, &wait);
4120 4121
		if (!x->done)
			return timeout;
L
Linus Torvalds 已提交
4122 4123
	}
	x->done--;
4124
	return timeout ?: 1;
L
Linus Torvalds 已提交
4125 4126
}

4127 4128
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
4129 4130 4131 4132
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
4133
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
4134
	spin_unlock_irq(&x->wait.lock);
4135 4136
	return timeout;
}
L
Linus Torvalds 已提交
4137

4138 4139 4140 4141 4142 4143 4144 4145 4146 4147
/**
 * 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().
 */
4148
void __sched wait_for_completion(struct completion *x)
4149 4150
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4151
}
4152
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
4153

4154 4155 4156 4157 4158 4159 4160 4161 4162
/**
 * 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.
 */
4163
unsigned long __sched
4164
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
4165
{
4166
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4167
}
4168
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
4169

4170 4171 4172 4173 4174 4175 4176
/**
 * 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.
 */
4177
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
4178
{
4179 4180 4181 4182
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
4183
}
4184
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
4185

4186 4187 4188 4189 4190 4191 4192 4193
/**
 * 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.
 */
4194
unsigned long __sched
4195 4196
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
4197
{
4198
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
4199
}
4200
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
4201

4202 4203 4204 4205 4206 4207 4208
/**
 * 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.
 */
M
Matthew Wilcox 已提交
4209 4210 4211 4212 4213 4214 4215 4216 4217
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);

4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
/**
 * 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.
 */
unsigned long __sched
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);

4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248
/**
 *	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)
{
4249
	unsigned long flags;
4250 4251
	int ret = 1;

4252
	spin_lock_irqsave(&x->wait.lock, flags);
4253 4254 4255 4256
	if (!x->done)
		ret = 0;
	else
		x->done--;
4257
	spin_unlock_irqrestore(&x->wait.lock, flags);
4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271
	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)
{
4272
	unsigned long flags;
4273 4274
	int ret = 1;

4275
	spin_lock_irqsave(&x->wait.lock, flags);
4276 4277
	if (!x->done)
		ret = 0;
4278
	spin_unlock_irqrestore(&x->wait.lock, flags);
4279 4280 4281 4282
	return ret;
}
EXPORT_SYMBOL(completion_done);

4283 4284
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
4285
{
I
Ingo Molnar 已提交
4286 4287 4288 4289
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
4290

4291
	__set_current_state(state);
L
Linus Torvalds 已提交
4292

4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306
	spin_lock_irqsave(&q->lock, flags);
	__add_wait_queue(q, &wait);
	spin_unlock(&q->lock);
	timeout = schedule_timeout(timeout);
	spin_lock_irq(&q->lock);
	__remove_wait_queue(q, &wait);
	spin_unlock_irqrestore(&q->lock, flags);

	return timeout;
}

void __sched interruptible_sleep_on(wait_queue_head_t *q)
{
	sleep_on_common(q, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
4307 4308 4309
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
4310
long __sched
I
Ingo Molnar 已提交
4311
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
4312
{
4313
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
4314 4315 4316
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
4317
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
4318
{
4319
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
4320 4321 4322
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
4323
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
4324
{
4325
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
4326 4327 4328
}
EXPORT_SYMBOL(sleep_on_timeout);

4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340
#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.
 */
4341
void rt_mutex_setprio(struct task_struct *p, int prio)
4342 4343
{
	unsigned long flags;
4344
	int oldprio, on_rq, running;
4345
	struct rq *rq;
4346
	const struct sched_class *prev_class;
4347 4348 4349 4350 4351

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

	rq = task_rq_lock(p, &flags);

4352
	oldprio = p->prio;
4353
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
4354
	on_rq = p->se.on_rq;
4355
	running = task_current(rq, p);
4356
	if (on_rq)
4357
		dequeue_task(rq, p, 0);
4358 4359
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
4360 4361 4362 4363 4364 4365

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

4366 4367
	p->prio = prio;

4368 4369
	if (running)
		p->sched_class->set_curr_task(rq);
I
Ingo Molnar 已提交
4370
	if (on_rq) {
4371
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
4372 4373

		check_class_changed(rq, p, prev_class, oldprio, running);
4374 4375 4376 4377 4378 4379
	}
	task_rq_unlock(rq, &flags);
}

#endif

4380
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
4381
{
I
Ingo Molnar 已提交
4382
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
4383
	unsigned long flags;
4384
	struct rq *rq;
L
Linus Torvalds 已提交
4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396

	if (TASK_NICE(p) == nice || nice < -20 || nice > 19)
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
	rq = task_rq_lock(p, &flags);
	/*
	 * The RT priorities are set via sched_setscheduler(), but we still
	 * allow the 'normal' nice value to be set - but as expected
	 * it wont have any effect on scheduling until the task is
I
Ingo Molnar 已提交
4397
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
4398
	 */
4399
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
4400 4401 4402
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
I
Ingo Molnar 已提交
4403
	on_rq = p->se.on_rq;
4404
	if (on_rq)
4405
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
4406 4407

	p->static_prio = NICE_TO_PRIO(nice);
4408
	set_load_weight(p);
4409 4410 4411
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
4412

I
Ingo Molnar 已提交
4413
	if (on_rq) {
4414
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
4415
		/*
4416 4417
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
4418
		 */
4419
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
4420 4421 4422 4423 4424 4425 4426
			resched_task(rq->curr);
	}
out_unlock:
	task_rq_unlock(rq, &flags);
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
4427 4428 4429 4430 4431
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
4432
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
4433
{
4434 4435
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
4436

4437
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
4438 4439 4440
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
4441 4442 4443 4444 4445 4446 4447 4448 4449
#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.
 */
4450
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
4451
{
4452
	long nice, retval;
L
Linus Torvalds 已提交
4453 4454 4455 4456 4457 4458

	/*
	 * Setpriority might change our priority at the same moment.
	 * We don't have to worry. Conceptually one call occurs first
	 * and we have a single winner.
	 */
M
Matt Mackall 已提交
4459 4460
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
4461 4462 4463
	if (increment > 40)
		increment = 40;

4464
	nice = TASK_NICE(current) + increment;
L
Linus Torvalds 已提交
4465 4466 4467 4468 4469
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
4470 4471 4472
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490
	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.
 */
4491
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
4492 4493 4494 4495 4496 4497 4498 4499
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
4500
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
4501 4502 4503
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
4504
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
 */
int idle_cpu(int cpu)
{
	return cpu_curr(cpu) == cpu_rq(cpu)->idle;
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
 */
4519
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4520 4521 4522 4523 4524 4525 4526 4527
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
 */
A
Alexey Dobriyan 已提交
4528
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4529
{
4530
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4531 4532 4533
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
4534 4535
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
4536
{
I
Ingo Molnar 已提交
4537
	BUG_ON(p->se.on_rq);
4538

L
Linus Torvalds 已提交
4539 4540
	p->policy = policy;
	p->rt_priority = prio;
4541 4542 4543
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4544 4545 4546 4547
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
4548
	set_load_weight(p);
L
Linus Torvalds 已提交
4549 4550
}

4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566
/*
 * 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);
	match = (cred->euid == pcred->euid ||
		 cred->euid == pcred->uid);
	rcu_read_unlock();
	return match;
}

4567 4568
static int __sched_setscheduler(struct task_struct *p, int policy,
				struct sched_param *param, bool user)
L
Linus Torvalds 已提交
4569
{
4570
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
4571
	unsigned long flags;
4572
	const struct sched_class *prev_class;
4573
	struct rq *rq;
4574
	int reset_on_fork;
L
Linus Torvalds 已提交
4575

4576 4577
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4578 4579
recheck:
	/* double check policy once rq lock held */
4580 4581
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4582
		policy = oldpolicy = p->policy;
4583 4584 4585 4586 4587 4588 4589 4590 4591 4592
	} else {
		reset_on_fork = !!(policy & SCHED_RESET_ON_FORK);
		policy &= ~SCHED_RESET_ON_FORK;

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

L
Linus Torvalds 已提交
4593 4594
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4595 4596
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4597 4598
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
4599
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4600
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4601
		return -EINVAL;
4602
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
4603 4604
		return -EINVAL;

4605 4606 4607
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4608
	if (user && !capable(CAP_SYS_NICE)) {
4609
		if (rt_policy(policy)) {
4610 4611
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4612 4613 4614 4615 4616 4617 4618 4619 4620 4621

			/* 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;
		}
I
Ingo Molnar 已提交
4622 4623 4624 4625 4626 4627
		/*
		 * Like positive nice levels, dont allow tasks to
		 * move out of SCHED_IDLE either:
		 */
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE)
			return -EPERM;
4628

4629
		/* can't change other user's priorities */
4630
		if (!check_same_owner(p))
4631
			return -EPERM;
4632 4633 4634 4635

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4636
	}
L
Linus Torvalds 已提交
4637

4638 4639 4640 4641 4642 4643
	if (user) {
		retval = security_task_setscheduler(p, policy, param);
		if (retval)
			return retval;
	}

4644 4645 4646 4647
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
	 */
4648
	raw_spin_lock_irqsave(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4649 4650 4651 4652
	/*
	 * To be able to change p->policy safely, the apropriate
	 * runqueue lock must be held.
	 */
4653
	rq = __task_rq_lock(p);
4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669

#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) &&
				task_group(p)->rt_bandwidth.rt_runtime == 0) {
			__task_rq_unlock(rq);
			raw_spin_unlock_irqrestore(&p->pi_lock, flags);
			return -EPERM;
		}
	}
#endif

L
Linus Torvalds 已提交
4670 4671 4672
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4673
		__task_rq_unlock(rq);
4674
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4675 4676
		goto recheck;
	}
I
Ingo Molnar 已提交
4677
	on_rq = p->se.on_rq;
4678
	running = task_current(rq, p);
4679
	if (on_rq)
4680
		deactivate_task(rq, p, 0);
4681 4682
	if (running)
		p->sched_class->put_prev_task(rq, p);
4683

4684 4685
	p->sched_reset_on_fork = reset_on_fork;

L
Linus Torvalds 已提交
4686
	oldprio = p->prio;
4687
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
4688
	__setscheduler(rq, p, policy, param->sched_priority);
4689

4690 4691
	if (running)
		p->sched_class->set_curr_task(rq);
I
Ingo Molnar 已提交
4692 4693
	if (on_rq) {
		activate_task(rq, p, 0);
4694 4695

		check_class_changed(rq, p, prev_class, oldprio, running);
L
Linus Torvalds 已提交
4696
	}
4697
	__task_rq_unlock(rq);
4698
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
4699

4700 4701
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
4702 4703
	return 0;
}
4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717

/**
 * 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,
		       struct sched_param *param)
{
	return __sched_setscheduler(p, policy, param, true);
}
L
Linus Torvalds 已提交
4718 4719
EXPORT_SYMBOL_GPL(sched_setscheduler);

4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736
/**
 * 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,
			       struct sched_param *param)
{
	return __sched_setscheduler(p, policy, param, false);
}

I
Ingo Molnar 已提交
4737 4738
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4739 4740 4741
{
	struct sched_param lparam;
	struct task_struct *p;
4742
	int retval;
L
Linus Torvalds 已提交
4743 4744 4745 4746 4747

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4748 4749 4750

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4751
	p = find_process_by_pid(pid);
4752 4753 4754
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4755

L
Linus Torvalds 已提交
4756 4757 4758 4759 4760 4761 4762 4763 4764
	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.
 */
4765 4766
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
4767
{
4768 4769 4770 4771
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4772 4773 4774 4775 4776 4777 4778 4779
	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.
 */
4780
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4781 4782 4783 4784 4785 4786 4787 4788
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
4789
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4790
{
4791
	struct task_struct *p;
4792
	int retval;
L
Linus Torvalds 已提交
4793 4794

	if (pid < 0)
4795
		return -EINVAL;
L
Linus Torvalds 已提交
4796 4797

	retval = -ESRCH;
4798
	rcu_read_lock();
L
Linus Torvalds 已提交
4799 4800 4801 4802
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4803 4804
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4805
	}
4806
	rcu_read_unlock();
L
Linus Torvalds 已提交
4807 4808 4809 4810
	return retval;
}

/**
4811
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4812 4813 4814
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
4815
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4816 4817
{
	struct sched_param lp;
4818
	struct task_struct *p;
4819
	int retval;
L
Linus Torvalds 已提交
4820 4821

	if (!param || pid < 0)
4822
		return -EINVAL;
L
Linus Torvalds 已提交
4823

4824
	rcu_read_lock();
L
Linus Torvalds 已提交
4825 4826 4827 4828 4829 4830 4831 4832 4833 4834
	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;
4835
	rcu_read_unlock();
L
Linus Torvalds 已提交
4836 4837 4838 4839 4840 4841 4842 4843 4844

	/*
	 * 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:
4845
	rcu_read_unlock();
L
Linus Torvalds 已提交
4846 4847 4848
	return retval;
}

4849
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4850
{
4851
	cpumask_var_t cpus_allowed, new_mask;
4852 4853
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4854

4855
	get_online_cpus();
4856
	rcu_read_lock();
L
Linus Torvalds 已提交
4857 4858 4859

	p = find_process_by_pid(pid);
	if (!p) {
4860
		rcu_read_unlock();
4861
		put_online_cpus();
L
Linus Torvalds 已提交
4862 4863 4864
		return -ESRCH;
	}

4865
	/* Prevent p going away */
L
Linus Torvalds 已提交
4866
	get_task_struct(p);
4867
	rcu_read_unlock();
L
Linus Torvalds 已提交
4868

4869 4870 4871 4872 4873 4874 4875 4876
	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_put_task;
	}
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_free_cpus_allowed;
	}
L
Linus Torvalds 已提交
4877
	retval = -EPERM;
4878
	if (!check_same_owner(p) && !capable(CAP_SYS_NICE))
L
Linus Torvalds 已提交
4879 4880
		goto out_unlock;

4881 4882 4883 4884
	retval = security_task_setscheduler(p, 0, NULL);
	if (retval)
		goto out_unlock;

4885 4886
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
P
Paul Menage 已提交
4887
 again:
4888
	retval = set_cpus_allowed_ptr(p, new_mask);
L
Linus Torvalds 已提交
4889

P
Paul Menage 已提交
4890
	if (!retval) {
4891 4892
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4893 4894 4895 4896 4897
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4898
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4899 4900 4901
			goto again;
		}
	}
L
Linus Torvalds 已提交
4902
out_unlock:
4903 4904 4905 4906
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4907
	put_task_struct(p);
4908
	put_online_cpus();
L
Linus Torvalds 已提交
4909 4910 4911 4912
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4913
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4914
{
4915 4916 4917 4918 4919
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4920 4921 4922 4923 4924 4925 4926 4927 4928
	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
 */
4929 4930
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4931
{
4932
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4933 4934
	int retval;

4935 4936
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4937

4938 4939 4940 4941 4942
	retval = get_user_cpu_mask(user_mask_ptr, len, new_mask);
	if (retval == 0)
		retval = sched_setaffinity(pid, new_mask);
	free_cpumask_var(new_mask);
	return retval;
L
Linus Torvalds 已提交
4943 4944
}

4945
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4946
{
4947
	struct task_struct *p;
4948 4949
	unsigned long flags;
	struct rq *rq;
L
Linus Torvalds 已提交
4950 4951
	int retval;

4952
	get_online_cpus();
4953
	rcu_read_lock();
L
Linus Torvalds 已提交
4954 4955 4956 4957 4958 4959

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

4960 4961 4962 4963
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4964
	rq = task_rq_lock(p, &flags);
4965
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
4966
	task_rq_unlock(rq, &flags);
L
Linus Torvalds 已提交
4967 4968

out_unlock:
4969
	rcu_read_unlock();
4970
	put_online_cpus();
L
Linus Torvalds 已提交
4971

4972
	return retval;
L
Linus Torvalds 已提交
4973 4974 4975 4976 4977 4978 4979 4980
}

/**
 * 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
 */
4981 4982
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4983 4984
{
	int ret;
4985
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4986

A
Anton Blanchard 已提交
4987
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4988 4989
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4990 4991
		return -EINVAL;

4992 4993
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4994

4995 4996
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4997
		size_t retlen = min_t(size_t, len, cpumask_size());
4998 4999

		if (copy_to_user(user_mask_ptr, mask, retlen))
5000 5001
			ret = -EFAULT;
		else
5002
			ret = retlen;
5003 5004
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
5005

5006
	return ret;
L
Linus Torvalds 已提交
5007 5008 5009 5010 5011
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
5012 5013
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
L
Linus Torvalds 已提交
5014
 */
5015
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
5016
{
5017
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
5018

5019
	schedstat_inc(rq, yld_count);
5020
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
5021 5022 5023 5024 5025 5026

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
5027
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
5028
	do_raw_spin_unlock(&rq->lock);
L
Linus Torvalds 已提交
5029 5030 5031 5032 5033 5034 5035
	preempt_enable_no_resched();

	schedule();

	return 0;
}

P
Peter Zijlstra 已提交
5036 5037 5038 5039 5040
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

A
Andrew Morton 已提交
5041
static void __cond_resched(void)
L
Linus Torvalds 已提交
5042
{
5043 5044 5045
	add_preempt_count(PREEMPT_ACTIVE);
	schedule();
	sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
5046 5047
}

5048
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
5049
{
P
Peter Zijlstra 已提交
5050
	if (should_resched()) {
L
Linus Torvalds 已提交
5051 5052 5053 5054 5055
		__cond_resched();
		return 1;
	}
	return 0;
}
5056
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
5057 5058

/*
5059
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
5060 5061
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
5062
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
5063 5064 5065
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
5066
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
5067
{
P
Peter Zijlstra 已提交
5068
	int resched = should_resched();
J
Jan Kara 已提交
5069 5070
	int ret = 0;

5071 5072
	lockdep_assert_held(lock);

N
Nick Piggin 已提交
5073
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
5074
		spin_unlock(lock);
P
Peter Zijlstra 已提交
5075
		if (resched)
N
Nick Piggin 已提交
5076 5077 5078
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
5079
		ret = 1;
L
Linus Torvalds 已提交
5080 5081
		spin_lock(lock);
	}
J
Jan Kara 已提交
5082
	return ret;
L
Linus Torvalds 已提交
5083
}
5084
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
5085

5086
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
5087 5088 5089
{
	BUG_ON(!in_softirq());

P
Peter Zijlstra 已提交
5090
	if (should_resched()) {
5091
		local_bh_enable();
L
Linus Torvalds 已提交
5092 5093 5094 5095 5096 5097
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
5098
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
5099 5100 5101 5102

/**
 * yield - yield the current processor to other threads.
 *
5103
 * This is a shortcut for kernel-space yielding - it marks the
L
Linus Torvalds 已提交
5104 5105 5106 5107 5108 5109 5110 5111 5112 5113
 * thread runnable and calls sys_sched_yield().
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

/*
I
Ingo Molnar 已提交
5114
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5115 5116 5117 5118
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
5119
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
5120

5121
	delayacct_blkio_start();
L
Linus Torvalds 已提交
5122
	atomic_inc(&rq->nr_iowait);
5123
	current->in_iowait = 1;
L
Linus Torvalds 已提交
5124
	schedule();
5125
	current->in_iowait = 0;
L
Linus Torvalds 已提交
5126
	atomic_dec(&rq->nr_iowait);
5127
	delayacct_blkio_end();
L
Linus Torvalds 已提交
5128 5129 5130 5131 5132
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
5133
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
5134 5135
	long ret;

5136
	delayacct_blkio_start();
L
Linus Torvalds 已提交
5137
	atomic_inc(&rq->nr_iowait);
5138
	current->in_iowait = 1;
L
Linus Torvalds 已提交
5139
	ret = schedule_timeout(timeout);
5140
	current->in_iowait = 0;
L
Linus Torvalds 已提交
5141
	atomic_dec(&rq->nr_iowait);
5142
	delayacct_blkio_end();
L
Linus Torvalds 已提交
5143 5144 5145 5146 5147 5148 5149 5150 5151 5152
	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.
 */
5153
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5154 5155 5156 5157 5158 5159 5160 5161 5162
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
5163
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5164
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177
		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.
 */
5178
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5179 5180 5181 5182 5183 5184 5185 5186 5187
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
5188
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5189
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202
		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.
 */
5203
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5204
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5205
{
5206
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5207
	unsigned int time_slice;
5208 5209
	unsigned long flags;
	struct rq *rq;
5210
	int retval;
L
Linus Torvalds 已提交
5211 5212 5213
	struct timespec t;

	if (pid < 0)
5214
		return -EINVAL;
L
Linus Torvalds 已提交
5215 5216

	retval = -ESRCH;
5217
	rcu_read_lock();
L
Linus Torvalds 已提交
5218 5219 5220 5221 5222 5223 5224 5225
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5226 5227 5228
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
	task_rq_unlock(rq, &flags);
D
Dmitry Adamushko 已提交
5229

5230
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5231
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5232 5233
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5234

L
Linus Torvalds 已提交
5235
out_unlock:
5236
	rcu_read_unlock();
L
Linus Torvalds 已提交
5237 5238 5239
	return retval;
}

5240
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5241

5242
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5243 5244
{
	unsigned long free = 0;
5245
	unsigned state;
L
Linus Torvalds 已提交
5246 5247

	state = p->state ? __ffs(p->state) + 1 : 0;
P
Peter Zijlstra 已提交
5248
	printk(KERN_INFO "%-13.13s %c", p->comm,
5249
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
5250
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
5251
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5252
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
5253
	else
P
Peter Zijlstra 已提交
5254
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5255 5256
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5257
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5258
	else
P
Peter Zijlstra 已提交
5259
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5260 5261
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
5262
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5263
#endif
P
Peter Zijlstra 已提交
5264
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5265 5266
		task_pid_nr(p), task_pid_nr(p->real_parent),
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5267

5268
	show_stack(p, NULL);
L
Linus Torvalds 已提交
5269 5270
}

I
Ingo Molnar 已提交
5271
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5272
{
5273
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5274

5275
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5276 5277
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5278
#else
P
Peter Zijlstra 已提交
5279 5280
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5281 5282 5283 5284 5285 5286 5287 5288
#endif
	read_lock(&tasklist_lock);
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
		 * console might take alot of time:
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
5289
		if (!state_filter || (p->state & state_filter))
5290
			sched_show_task(p);
L
Linus Torvalds 已提交
5291 5292
	} while_each_thread(g, p);

5293 5294
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
5295 5296 5297
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
L
Linus Torvalds 已提交
5298
	read_unlock(&tasklist_lock);
I
Ingo Molnar 已提交
5299 5300 5301
	/*
	 * Only show locks if all tasks are dumped:
	 */
5302
	if (!state_filter)
I
Ingo Molnar 已提交
5303
		debug_show_all_locks();
L
Linus Torvalds 已提交
5304 5305
}

I
Ingo Molnar 已提交
5306 5307
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
5308
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5309 5310
}

5311 5312 5313 5314 5315 5316 5317 5318
/**
 * 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.
 */
5319
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5320
{
5321
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5322 5323
	unsigned long flags;

5324
	raw_spin_lock_irqsave(&rq->lock, flags);
5325

I
Ingo Molnar 已提交
5326
	__sched_fork(idle);
5327
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5328 5329
	idle->se.exec_start = sched_clock();

5330
	cpumask_copy(&idle->cpus_allowed, cpumask_of(cpu));
I
Ingo Molnar 已提交
5331
	__set_task_cpu(idle, cpu);
L
Linus Torvalds 已提交
5332 5333

	rq->curr = rq->idle = idle;
5334 5335 5336
#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
	idle->oncpu = 1;
#endif
5337
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5338 5339

	/* Set the preempt count _outside_ the spinlocks! */
5340 5341 5342
#if defined(CONFIG_PREEMPT)
	task_thread_info(idle)->preempt_count = (idle->lock_depth >= 0);
#else
A
Al Viro 已提交
5343
	task_thread_info(idle)->preempt_count = 0;
5344
#endif
I
Ingo Molnar 已提交
5345 5346 5347 5348
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5349
	ftrace_graph_init_task(idle);
L
Linus Torvalds 已提交
5350 5351 5352 5353 5354 5355 5356
}

/*
 * In a system that switches off the HZ timer nohz_cpu_mask
 * indicates which cpus entered this state. This is used
 * in the rcu update to wait only for active cpus. For system
 * which do not switch off the HZ timer nohz_cpu_mask should
5357
 * always be CPU_BITS_NONE.
L
Linus Torvalds 已提交
5358
 */
5359
cpumask_var_t nohz_cpu_mask;
L
Linus Torvalds 已提交
5360

I
Ingo Molnar 已提交
5361 5362 5363 5364 5365 5366 5367 5368 5369
/*
 * Increase the granularity value when there are more CPUs,
 * because with more CPUs the 'effective latency' as visible
 * to users decreases. But the relationship is not linear,
 * so pick a second-best guess by going with the log2 of the
 * number of CPUs.
 *
 * This idea comes from the SD scheduler of Con Kolivas:
 */
5370
static int get_update_sysctl_factor(void)
I
Ingo Molnar 已提交
5371
{
5372
	unsigned int cpus = min_t(int, num_online_cpus(), 8);
5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386
	unsigned int factor;

	switch (sysctl_sched_tunable_scaling) {
	case SCHED_TUNABLESCALING_NONE:
		factor = 1;
		break;
	case SCHED_TUNABLESCALING_LINEAR:
		factor = cpus;
		break;
	case SCHED_TUNABLESCALING_LOG:
	default:
		factor = 1 + ilog2(cpus);
		break;
	}
I
Ingo Molnar 已提交
5387

5388 5389
	return factor;
}
I
Ingo Molnar 已提交
5390

5391 5392 5393
static void update_sysctl(void)
{
	unsigned int factor = get_update_sysctl_factor();
I
Ingo Molnar 已提交
5394

5395 5396 5397 5398 5399 5400 5401 5402
#define SET_SYSCTL(name) \
	(sysctl_##name = (factor) * normalized_sysctl_##name)
	SET_SYSCTL(sched_min_granularity);
	SET_SYSCTL(sched_latency);
	SET_SYSCTL(sched_wakeup_granularity);
	SET_SYSCTL(sched_shares_ratelimit);
#undef SET_SYSCTL
}
5403

5404 5405 5406
static inline void sched_init_granularity(void)
{
	update_sysctl();
I
Ingo Molnar 已提交
5407 5408
}

L
Linus Torvalds 已提交
5409 5410 5411 5412
#ifdef CONFIG_SMP
/*
 * This is how migration works:
 *
5413 5414 5415 5416 5417 5418
 * 1) we invoke migration_cpu_stop() on the target CPU using
 *    stop_one_cpu().
 * 2) stopper starts to run (implicitly forcing the migrated thread
 *    off the CPU)
 * 3) it checks whether the migrated task is still in the wrong runqueue.
 * 4) if it's in the wrong runqueue then the migration thread removes
L
Linus Torvalds 已提交
5419
 *    it and puts it into the right queue.
5420 5421
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
L
Linus Torvalds 已提交
5422 5423 5424 5425 5426 5427 5428 5429
 */

/*
 * Change a given task's CPU affinity. Migrate the thread to a
 * proper CPU and schedule it away if the CPU it's executing on
 * is removed from the allowed bitmask.
 *
 * NOTE: the caller must have a valid reference to the task, the
I
Ingo Molnar 已提交
5430
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
5431 5432
 * call is not atomic; no spinlocks may be held.
 */
5433
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
L
Linus Torvalds 已提交
5434 5435
{
	unsigned long flags;
5436
	struct rq *rq;
5437
	unsigned int dest_cpu;
5438
	int ret = 0;
L
Linus Torvalds 已提交
5439

P
Peter Zijlstra 已提交
5440 5441 5442 5443 5444 5445 5446
	/*
	 * Serialize against TASK_WAKING so that ttwu() and wunt() can
	 * drop the rq->lock and still rely on ->cpus_allowed.
	 */
again:
	while (task_is_waking(p))
		cpu_relax();
L
Linus Torvalds 已提交
5447
	rq = task_rq_lock(p, &flags);
P
Peter Zijlstra 已提交
5448 5449 5450 5451
	if (task_is_waking(p)) {
		task_rq_unlock(rq, &flags);
		goto again;
	}
5452

5453
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
L
Linus Torvalds 已提交
5454 5455 5456 5457
		ret = -EINVAL;
		goto out;
	}

5458
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current &&
5459
		     !cpumask_equal(&p->cpus_allowed, new_mask))) {
5460 5461 5462 5463
		ret = -EINVAL;
		goto out;
	}

5464
	if (p->sched_class->set_cpus_allowed)
5465
		p->sched_class->set_cpus_allowed(p, new_mask);
5466
	else {
5467 5468
		cpumask_copy(&p->cpus_allowed, new_mask);
		p->rt.nr_cpus_allowed = cpumask_weight(new_mask);
5469 5470
	}

L
Linus Torvalds 已提交
5471
	/* Can the task run on the task's current CPU? If so, we're done */
5472
	if (cpumask_test_cpu(task_cpu(p), new_mask))
L
Linus Torvalds 已提交
5473 5474
		goto out;

5475 5476 5477
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
	if (migrate_task(p, dest_cpu)) {
		struct migration_arg arg = { p, dest_cpu };
L
Linus Torvalds 已提交
5478 5479
		/* Need help from migration thread: drop lock and wait. */
		task_rq_unlock(rq, &flags);
5480
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
5481 5482 5483 5484 5485
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
	task_rq_unlock(rq, &flags);
5486

L
Linus Torvalds 已提交
5487 5488
	return ret;
}
5489
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
5490 5491

/*
I
Ingo Molnar 已提交
5492
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
5493 5494 5495 5496 5497 5498
 * 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.
5499 5500
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
5501
 */
5502
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
5503
{
5504
	struct rq *rq_dest, *rq_src;
5505
	int ret = 0;
L
Linus Torvalds 已提交
5506

5507
	if (unlikely(!cpu_active(dest_cpu)))
5508
		return ret;
L
Linus Torvalds 已提交
5509 5510 5511 5512 5513 5514 5515

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

	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
L
Linus Torvalds 已提交
5516
		goto done;
L
Linus Torvalds 已提交
5517
	/* Affinity changed (again). */
5518
	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
L
Linus Torvalds 已提交
5519
		goto fail;
L
Linus Torvalds 已提交
5520

5521 5522 5523 5524 5525
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
	if (p->se.on_rq) {
5526
		deactivate_task(rq_src, p, 0);
5527
		set_task_cpu(p, dest_cpu);
I
Ingo Molnar 已提交
5528
		activate_task(rq_dest, p, 0);
5529
		check_preempt_curr(rq_dest, p, 0);
L
Linus Torvalds 已提交
5530
	}
L
Linus Torvalds 已提交
5531
done:
5532
	ret = 1;
L
Linus Torvalds 已提交
5533
fail:
L
Linus Torvalds 已提交
5534
	double_rq_unlock(rq_src, rq_dest);
5535
	return ret;
L
Linus Torvalds 已提交
5536 5537 5538
}

/*
5539 5540 5541
 * migration_cpu_stop - this will be executed by a highprio stopper thread
 * and performs thread migration by bumping thread off CPU then
 * 'pushing' onto another runqueue.
L
Linus Torvalds 已提交
5542
 */
5543
static int migration_cpu_stop(void *data)
L
Linus Torvalds 已提交
5544
{
5545
	struct migration_arg *arg = data;
5546

5547 5548 5549 5550
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
5551
	local_irq_disable();
5552
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
5553
	local_irq_enable();
L
Linus Torvalds 已提交
5554
	return 0;
5555 5556
}

L
Linus Torvalds 已提交
5557
#ifdef CONFIG_HOTPLUG_CPU
5558
/*
5559
 * Figure out where task on dead CPU should go, use force if necessary.
5560
 */
5561
void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p)
L
Linus Torvalds 已提交
5562
{
5563 5564 5565
	struct rq *rq = cpu_rq(dead_cpu);
	int needs_cpu, uninitialized_var(dest_cpu);
	unsigned long flags;
5566

5567
	local_irq_save(flags);
5568

5569 5570 5571 5572 5573
	raw_spin_lock(&rq->lock);
	needs_cpu = (task_cpu(p) == dead_cpu) && (p->state != TASK_WAKING);
	if (needs_cpu)
		dest_cpu = select_fallback_rq(dead_cpu, p);
	raw_spin_unlock(&rq->lock);
5574 5575 5576 5577
	/*
	 * It can only fail if we race with set_cpus_allowed(),
	 * in the racer should migrate the task anyway.
	 */
5578
	if (needs_cpu)
5579
		__migrate_task(p, dead_cpu, dest_cpu);
5580
	local_irq_restore(flags);
L
Linus Torvalds 已提交
5581 5582 5583 5584 5585 5586 5587 5588 5589
}

/*
 * 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:
 */
5590
static void migrate_nr_uninterruptible(struct rq *rq_src)
L
Linus Torvalds 已提交
5591
{
5592
	struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
L
Linus Torvalds 已提交
5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605
	unsigned long flags;

	local_irq_save(flags);
	double_rq_lock(rq_src, rq_dest);
	rq_dest->nr_uninterruptible += rq_src->nr_uninterruptible;
	rq_src->nr_uninterruptible = 0;
	double_rq_unlock(rq_src, rq_dest);
	local_irq_restore(flags);
}

/* Run through task list and migrate tasks from the dead cpu. */
static void migrate_live_tasks(int src_cpu)
{
5606
	struct task_struct *p, *t;
L
Linus Torvalds 已提交
5607

5608
	read_lock(&tasklist_lock);
L
Linus Torvalds 已提交
5609

5610 5611
	do_each_thread(t, p) {
		if (p == current)
L
Linus Torvalds 已提交
5612 5613
			continue;

5614 5615 5616
		if (task_cpu(p) == src_cpu)
			move_task_off_dead_cpu(src_cpu, p);
	} while_each_thread(t, p);
L
Linus Torvalds 已提交
5617

5618
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
5619 5620
}

I
Ingo Molnar 已提交
5621 5622
/*
 * Schedules idle task to be the next runnable task on current CPU.
5623 5624
 * It does so by boosting its priority to highest possible.
 * Used by CPU offline code.
L
Linus Torvalds 已提交
5625 5626 5627
 */
void sched_idle_next(void)
{
5628
	int this_cpu = smp_processor_id();
5629
	struct rq *rq = cpu_rq(this_cpu);
L
Linus Torvalds 已提交
5630 5631 5632 5633
	struct task_struct *p = rq->idle;
	unsigned long flags;

	/* cpu has to be offline */
5634
	BUG_ON(cpu_online(this_cpu));
L
Linus Torvalds 已提交
5635

5636 5637 5638
	/*
	 * Strictly not necessary since rest of the CPUs are stopped by now
	 * and interrupts disabled on the current cpu.
L
Linus Torvalds 已提交
5639
	 */
5640
	raw_spin_lock_irqsave(&rq->lock, flags);
L
Linus Torvalds 已提交
5641

I
Ingo Molnar 已提交
5642
	__setscheduler(rq, p, SCHED_FIFO, MAX_RT_PRIO-1);
5643

5644
	activate_task(rq, p, 0);
L
Linus Torvalds 已提交
5645

5646
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5647 5648
}

5649 5650
/*
 * Ensures that the idle task is using init_mm right before its cpu goes
L
Linus Torvalds 已提交
5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663
 * offline.
 */
void idle_task_exit(void)
{
	struct mm_struct *mm = current->active_mm;

	BUG_ON(cpu_online(smp_processor_id()));

	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
}

5664
/* called under rq->lock with disabled interrupts */
5665
static void migrate_dead(unsigned int dead_cpu, struct task_struct *p)
L
Linus Torvalds 已提交
5666
{
5667
	struct rq *rq = cpu_rq(dead_cpu);
L
Linus Torvalds 已提交
5668 5669

	/* Must be exiting, otherwise would be on tasklist. */
E
Eugene Teo 已提交
5670
	BUG_ON(!p->exit_state);
L
Linus Torvalds 已提交
5671 5672

	/* Cannot have done final schedule yet: would have vanished. */
5673
	BUG_ON(p->state == TASK_DEAD);
L
Linus Torvalds 已提交
5674

5675
	get_task_struct(p);
L
Linus Torvalds 已提交
5676 5677 5678

	/*
	 * Drop lock around migration; if someone else moves it,
I
Ingo Molnar 已提交
5679
	 * that's OK. No task can be added to this CPU, so iteration is
L
Linus Torvalds 已提交
5680 5681
	 * fine.
	 */
5682
	raw_spin_unlock_irq(&rq->lock);
5683
	move_task_off_dead_cpu(dead_cpu, p);
5684
	raw_spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
5685

5686
	put_task_struct(p);
L
Linus Torvalds 已提交
5687 5688 5689 5690 5691
}

/* release_task() removes task from tasklist, so we won't find dead tasks. */
static void migrate_dead_tasks(unsigned int dead_cpu)
{
5692
	struct rq *rq = cpu_rq(dead_cpu);
I
Ingo Molnar 已提交
5693
	struct task_struct *next;
5694

I
Ingo Molnar 已提交
5695 5696 5697
	for ( ; ; ) {
		if (!rq->nr_running)
			break;
5698
		next = pick_next_task(rq);
I
Ingo Molnar 已提交
5699 5700
		if (!next)
			break;
D
Dmitry Adamushko 已提交
5701
		next->sched_class->put_prev_task(rq, next);
I
Ingo Molnar 已提交
5702
		migrate_dead(dead_cpu, next);
5703

L
Linus Torvalds 已提交
5704 5705
	}
}
5706 5707 5708 5709 5710 5711 5712

/*
 * remove the tasks which were accounted by rq from calc_load_tasks.
 */
static void calc_global_load_remove(struct rq *rq)
{
	atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
5713
	rq->calc_load_active = 0;
5714
}
L
Linus Torvalds 已提交
5715 5716
#endif /* CONFIG_HOTPLUG_CPU */

5717 5718 5719
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5720 5721
	{
		.procname	= "sched_domain",
5722
		.mode		= 0555,
5723
	},
5724
	{}
5725 5726 5727
};

static struct ctl_table sd_ctl_root[] = {
5728 5729
	{
		.procname	= "kernel",
5730
		.mode		= 0555,
5731 5732
		.child		= sd_ctl_dir,
	},
5733
	{}
5734 5735 5736 5737 5738
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5739
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5740 5741 5742 5743

	return entry;
}

5744 5745
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5746
	struct ctl_table *entry;
5747

5748 5749 5750
	/*
	 * In the intermediate directories, both the child directory and
	 * procname are dynamically allocated and could fail but the mode
I
Ingo Molnar 已提交
5751
	 * will always be set. In the lowest directory the names are
5752 5753 5754
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5755 5756
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5757 5758 5759
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5760 5761 5762 5763 5764

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

5765
static void
5766
set_table_entry(struct ctl_table *entry,
5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779
		const char *procname, void *data, int maxlen,
		mode_t mode, proc_handler *proc_handler)
{
	entry->procname = procname;
	entry->data = data;
	entry->maxlen = maxlen;
	entry->mode = mode;
	entry->proc_handler = proc_handler;
}

static struct ctl_table *
sd_alloc_ctl_domain_table(struct sched_domain *sd)
{
5780
	struct ctl_table *table = sd_alloc_ctl_entry(13);
5781

5782 5783 5784
	if (table == NULL)
		return NULL;

5785
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5786
		sizeof(long), 0644, proc_doulongvec_minmax);
5787
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5788
		sizeof(long), 0644, proc_doulongvec_minmax);
5789
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5790
		sizeof(int), 0644, proc_dointvec_minmax);
5791
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5792
		sizeof(int), 0644, proc_dointvec_minmax);
5793
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5794
		sizeof(int), 0644, proc_dointvec_minmax);
5795
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5796
		sizeof(int), 0644, proc_dointvec_minmax);
5797
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5798
		sizeof(int), 0644, proc_dointvec_minmax);
5799
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5800
		sizeof(int), 0644, proc_dointvec_minmax);
5801
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5802
		sizeof(int), 0644, proc_dointvec_minmax);
5803
	set_table_entry(&table[9], "cache_nice_tries",
5804 5805
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5806
	set_table_entry(&table[10], "flags", &sd->flags,
5807
		sizeof(int), 0644, proc_dointvec_minmax);
5808 5809 5810
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
5811 5812 5813 5814

	return table;
}

5815
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5816 5817 5818 5819 5820 5821 5822 5823 5824
{
	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);
5825 5826
	if (table == NULL)
		return NULL;
5827 5828 5829 5830 5831

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5832
		entry->mode = 0555;
5833 5834 5835 5836 5837 5838 5839 5840
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5841
static void register_sched_domain_sysctl(void)
5842
{
5843
	int i, cpu_num = num_possible_cpus();
5844 5845 5846
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5847 5848 5849
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5850 5851 5852
	if (entry == NULL)
		return;

5853
	for_each_possible_cpu(i) {
5854 5855
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5856
		entry->mode = 0555;
5857
		entry->child = sd_alloc_ctl_cpu_table(i);
5858
		entry++;
5859
	}
5860 5861

	WARN_ON(sd_sysctl_header);
5862 5863
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5864

5865
/* may be called multiple times per register */
5866 5867
static void unregister_sched_domain_sysctl(void)
{
5868 5869
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5870
	sd_sysctl_header = NULL;
5871 5872
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5873
}
5874
#else
5875 5876 5877 5878
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5879 5880 5881 5882
{
}
#endif

5883 5884 5885 5886 5887
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5888
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907
		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);
		}

5908
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5909 5910 5911 5912
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
5913 5914 5915 5916
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5917 5918
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5919
{
5920
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
5921
	unsigned long flags;
5922
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5923 5924

	switch (action) {
5925

L
Linus Torvalds 已提交
5926
	case CPU_UP_PREPARE:
5927
	case CPU_UP_PREPARE_FROZEN:
5928
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
5929
		break;
5930

L
Linus Torvalds 已提交
5931
	case CPU_ONLINE:
5932
	case CPU_ONLINE_FROZEN:
5933
		/* Update our root-domain */
5934
		raw_spin_lock_irqsave(&rq->lock, flags);
5935
		if (rq->rd) {
5936
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5937 5938

			set_rq_online(rq);
5939
		}
5940
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5941
		break;
5942

L
Linus Torvalds 已提交
5943 5944
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DEAD:
5945
	case CPU_DEAD_FROZEN:
L
Linus Torvalds 已提交
5946 5947
		migrate_live_tasks(cpu);
		/* Idle task back to normal (off runqueue, low prio) */
5948
		raw_spin_lock_irq(&rq->lock);
5949
		deactivate_task(rq, rq->idle, 0);
I
Ingo Molnar 已提交
5950 5951
		__setscheduler(rq, rq->idle, SCHED_NORMAL, 0);
		rq->idle->sched_class = &idle_sched_class;
L
Linus Torvalds 已提交
5952
		migrate_dead_tasks(cpu);
5953
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
5954 5955
		migrate_nr_uninterruptible(rq);
		BUG_ON(rq->nr_running != 0);
5956
		calc_global_load_remove(rq);
L
Linus Torvalds 已提交
5957
		break;
G
Gregory Haskins 已提交
5958

5959 5960
	case CPU_DYING:
	case CPU_DYING_FROZEN:
G
Gregory Haskins 已提交
5961
		/* Update our root-domain */
5962
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5963
		if (rq->rd) {
5964
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5965
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5966
		}
5967
		raw_spin_unlock_irqrestore(&rq->lock, flags);
G
Gregory Haskins 已提交
5968
		break;
L
Linus Torvalds 已提交
5969 5970 5971 5972 5973
#endif
	}
	return NOTIFY_OK;
}

5974 5975 5976
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5977
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
5978
 */
5979
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
5980
	.notifier_call = migration_call,
5981
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
5982 5983
};

5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008
static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_ONLINE:
	case CPU_DOWN_FAILED:
		set_cpu_active((long)hcpu, true);
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}

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

6009
static int __init migration_init(void)
L
Linus Torvalds 已提交
6010 6011
{
	void *cpu = (void *)(long)smp_processor_id();
6012
	int err;
6013

6014
	/* Initialize migration for the boot CPU */
6015 6016
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
6017 6018
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
6019

6020 6021 6022 6023
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

6024
	return 0;
L
Linus Torvalds 已提交
6025
}
6026
early_initcall(migration_init);
L
Linus Torvalds 已提交
6027 6028 6029
#endif

#ifdef CONFIG_SMP
6030

6031
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
6032

6033 6034 6035 6036 6037 6038 6039 6040 6041 6042
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);

6043
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
6044
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
6045
{
I
Ingo Molnar 已提交
6046
	struct sched_group *group = sd->groups;
6047
	char str[256];
L
Linus Torvalds 已提交
6048

R
Rusty Russell 已提交
6049
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
6050
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
6051 6052 6053 6054

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
6055
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
6056
		if (sd->parent)
P
Peter Zijlstra 已提交
6057 6058
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
6059
		return -1;
N
Nick Piggin 已提交
6060 6061
	}

P
Peter Zijlstra 已提交
6062
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
I
Ingo Molnar 已提交
6063

6064
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
6065 6066
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
6067
	}
6068
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
6069 6070
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
6071
	}
L
Linus Torvalds 已提交
6072

I
Ingo Molnar 已提交
6073
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
6074
	do {
I
Ingo Molnar 已提交
6075
		if (!group) {
P
Peter Zijlstra 已提交
6076 6077
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
6078 6079 6080
			break;
		}

6081
		if (!group->cpu_power) {
P
Peter Zijlstra 已提交
6082 6083 6084
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
I
Ingo Molnar 已提交
6085 6086
			break;
		}
L
Linus Torvalds 已提交
6087

6088
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
6089 6090
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
6091 6092
			break;
		}
L
Linus Torvalds 已提交
6093

6094
		if (cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
6095 6096
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
6097 6098
			break;
		}
L
Linus Torvalds 已提交
6099

6100
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
6101

R
Rusty Russell 已提交
6102
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
6103

P
Peter Zijlstra 已提交
6104
		printk(KERN_CONT " %s", str);
6105
		if (group->cpu_power != SCHED_LOAD_SCALE) {
P
Peter Zijlstra 已提交
6106 6107
			printk(KERN_CONT " (cpu_power = %d)",
				group->cpu_power);
6108
		}
L
Linus Torvalds 已提交
6109

I
Ingo Molnar 已提交
6110 6111
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
6112
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
6113

6114
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
6115
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
6116

6117 6118
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
6119 6120
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
6121 6122
	return 0;
}
L
Linus Torvalds 已提交
6123

I
Ingo Molnar 已提交
6124 6125
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
6126
	cpumask_var_t groupmask;
I
Ingo Molnar 已提交
6127
	int level = 0;
L
Linus Torvalds 已提交
6128

6129 6130 6131
	if (!sched_domain_debug_enabled)
		return;

I
Ingo Molnar 已提交
6132 6133 6134 6135
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
6136

I
Ingo Molnar 已提交
6137 6138
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

6139
	if (!alloc_cpumask_var(&groupmask, GFP_KERNEL)) {
6140 6141 6142 6143
		printk(KERN_DEBUG "Cannot load-balance (out of memory)\n");
		return;
	}

I
Ingo Molnar 已提交
6144
	for (;;) {
6145
		if (sched_domain_debug_one(sd, cpu, level, groupmask))
I
Ingo Molnar 已提交
6146
			break;
L
Linus Torvalds 已提交
6147 6148
		level++;
		sd = sd->parent;
6149
		if (!sd)
I
Ingo Molnar 已提交
6150 6151
			break;
	}
6152
	free_cpumask_var(groupmask);
L
Linus Torvalds 已提交
6153
}
6154
#else /* !CONFIG_SCHED_DEBUG */
6155
# define sched_domain_debug(sd, cpu) do { } while (0)
6156
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
6157

6158
static int sd_degenerate(struct sched_domain *sd)
6159
{
6160
	if (cpumask_weight(sched_domain_span(sd)) == 1)
6161 6162 6163 6164 6165 6166
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
6167 6168 6169
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
6170 6171 6172 6173 6174
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
6175
	if (sd->flags & (SD_WAKE_AFFINE))
6176 6177 6178 6179 6180
		return 0;

	return 1;
}

6181 6182
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
6183 6184 6185 6186 6187 6188
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

6189
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
6190 6191 6192 6193 6194 6195 6196
		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 |
6197 6198 6199
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
6200 6201
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
6202 6203 6204 6205 6206 6207 6208
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

6209 6210
static void free_rootdomain(struct root_domain *rd)
{
6211 6212
	synchronize_sched();

6213 6214
	cpupri_cleanup(&rd->cpupri);

6215 6216 6217 6218 6219 6220
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
6221 6222
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
6223
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
6224 6225
	unsigned long flags;

6226
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
6227 6228

	if (rq->rd) {
I
Ingo Molnar 已提交
6229
		old_rd = rq->rd;
G
Gregory Haskins 已提交
6230

6231
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
6232
			set_rq_offline(rq);
G
Gregory Haskins 已提交
6233

6234
		cpumask_clear_cpu(rq->cpu, old_rd->span);
6235

I
Ingo Molnar 已提交
6236 6237 6238 6239 6240 6241 6242
		/*
		 * If we dont want to free the old_rt yet then
		 * set old_rd to NULL to skip the freeing later
		 * in this function:
		 */
		if (!atomic_dec_and_test(&old_rd->refcount))
			old_rd = NULL;
G
Gregory Haskins 已提交
6243 6244 6245 6246 6247
	}

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

6248
	cpumask_set_cpu(rq->cpu, rd->span);
6249
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
6250
		set_rq_online(rq);
G
Gregory Haskins 已提交
6251

6252
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
6253 6254 6255

	if (old_rd)
		free_rootdomain(old_rd);
G
Gregory Haskins 已提交
6256 6257
}

6258
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
6259 6260 6261
{
	memset(rd, 0, sizeof(*rd));

6262
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
6263
		goto out;
6264
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
6265
		goto free_span;
6266
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
6267
		goto free_online;
6268

6269
	if (cpupri_init(&rd->cpupri) != 0)
6270
		goto free_rto_mask;
6271
	return 0;
6272

6273 6274
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
6275 6276 6277 6278
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
6279
out:
6280
	return -ENOMEM;
G
Gregory Haskins 已提交
6281 6282 6283 6284
}

static void init_defrootdomain(void)
{
6285
	init_rootdomain(&def_root_domain);
6286

G
Gregory Haskins 已提交
6287 6288 6289
	atomic_set(&def_root_domain.refcount, 1);
}

6290
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
6291 6292 6293 6294 6295 6296 6297
{
	struct root_domain *rd;

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

6298
	if (init_rootdomain(rd) != 0) {
6299 6300 6301
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
6302 6303 6304 6305

	return rd;
}

L
Linus Torvalds 已提交
6306
/*
I
Ingo Molnar 已提交
6307
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
6308 6309
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
6310 6311
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
6312
{
6313
	struct rq *rq = cpu_rq(cpu);
6314 6315
	struct sched_domain *tmp;

6316 6317 6318
	for (tmp = sd; tmp; tmp = tmp->parent)
		tmp->span_weight = cpumask_weight(sched_domain_span(tmp));

6319
	/* Remove the sched domains which do not contribute to scheduling. */
6320
	for (tmp = sd; tmp; ) {
6321 6322 6323
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
6324

6325
		if (sd_parent_degenerate(tmp, parent)) {
6326
			tmp->parent = parent->parent;
6327 6328
			if (parent->parent)
				parent->parent->child = tmp;
6329 6330
		} else
			tmp = tmp->parent;
6331 6332
	}

6333
	if (sd && sd_degenerate(sd)) {
6334
		sd = sd->parent;
6335 6336 6337
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
6338 6339 6340

	sched_domain_debug(sd, cpu);

G
Gregory Haskins 已提交
6341
	rq_attach_root(rq, rd);
N
Nick Piggin 已提交
6342
	rcu_assign_pointer(rq->sd, sd);
L
Linus Torvalds 已提交
6343 6344 6345
}

/* cpus with isolated domains */
6346
static cpumask_var_t cpu_isolated_map;
L
Linus Torvalds 已提交
6347 6348 6349 6350

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
6351
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
6352
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
6353 6354 6355
	return 1;
}

I
Ingo Molnar 已提交
6356
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
6357 6358

/*
6359 6360
 * init_sched_build_groups takes the cpumask we wish to span, and a pointer
 * to a function which identifies what group(along with sched group) a CPU
6361 6362
 * belongs to. The return value of group_fn must be a >= 0 and < nr_cpu_ids
 * (due to the fact that we keep track of groups covered with a struct cpumask).
L
Linus Torvalds 已提交
6363 6364 6365 6366 6367
 *
 * init_sched_build_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.
 */
6368
static void
6369 6370 6371
init_sched_build_groups(const struct cpumask *span,
			const struct cpumask *cpu_map,
			int (*group_fn)(int cpu, const struct cpumask *cpu_map,
6372
					struct sched_group **sg,
6373 6374
					struct cpumask *tmpmask),
			struct cpumask *covered, struct cpumask *tmpmask)
L
Linus Torvalds 已提交
6375 6376 6377 6378
{
	struct sched_group *first = NULL, *last = NULL;
	int i;

6379
	cpumask_clear(covered);
6380

6381
	for_each_cpu(i, span) {
6382
		struct sched_group *sg;
6383
		int group = group_fn(i, cpu_map, &sg, tmpmask);
L
Linus Torvalds 已提交
6384 6385
		int j;

6386
		if (cpumask_test_cpu(i, covered))
L
Linus Torvalds 已提交
6387 6388
			continue;

6389
		cpumask_clear(sched_group_cpus(sg));
6390
		sg->cpu_power = 0;
L
Linus Torvalds 已提交
6391

6392
		for_each_cpu(j, span) {
6393
			if (group_fn(j, cpu_map, NULL, tmpmask) != group)
L
Linus Torvalds 已提交
6394 6395
				continue;

6396
			cpumask_set_cpu(j, covered);
6397
			cpumask_set_cpu(j, sched_group_cpus(sg));
L
Linus Torvalds 已提交
6398 6399 6400 6401 6402 6403 6404 6405 6406 6407
		}
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
}

6408
#define SD_NODES_PER_DOMAIN 16
L
Linus Torvalds 已提交
6409

6410
#ifdef CONFIG_NUMA
6411

6412 6413 6414 6415 6416
/**
 * find_next_best_node - find the next node to include in a sched_domain
 * @node: node whose sched_domain we're building
 * @used_nodes: nodes already in the sched_domain
 *
I
Ingo Molnar 已提交
6417
 * Find the next node to include in a given scheduling domain. Simply
6418 6419 6420 6421
 * finds the closest node not already in the @used_nodes map.
 *
 * Should use nodemask_t.
 */
6422
static int find_next_best_node(int node, nodemask_t *used_nodes)
6423 6424 6425 6426 6427
{
	int i, n, val, min_val, best_node = 0;

	min_val = INT_MAX;

6428
	for (i = 0; i < nr_node_ids; i++) {
6429
		/* Start at @node */
6430
		n = (node + i) % nr_node_ids;
6431 6432 6433 6434 6435

		if (!nr_cpus_node(n))
			continue;

		/* Skip already used nodes */
6436
		if (node_isset(n, *used_nodes))
6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447
			continue;

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

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

6448
	node_set(best_node, *used_nodes);
6449 6450 6451 6452 6453 6454
	return best_node;
}

/**
 * sched_domain_node_span - get a cpumask for a node's sched_domain
 * @node: node whose cpumask we're constructing
6455
 * @span: resulting cpumask
6456
 *
I
Ingo Molnar 已提交
6457
 * Given a node, construct a good cpumask for its sched_domain to span. It
6458 6459 6460
 * should be one that prevents unnecessary balancing, but also spreads tasks
 * out optimally.
 */
6461
static void sched_domain_node_span(int node, struct cpumask *span)
6462
{
6463
	nodemask_t used_nodes;
6464
	int i;
6465

6466
	cpumask_clear(span);
6467
	nodes_clear(used_nodes);
6468

6469
	cpumask_or(span, span, cpumask_of_node(node));
6470
	node_set(node, used_nodes);
6471 6472

	for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
6473
		int next_node = find_next_best_node(node, &used_nodes);
6474

6475
		cpumask_or(span, span, cpumask_of_node(next_node));
6476 6477
	}
}
6478
#endif /* CONFIG_NUMA */
6479

6480
int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
6481

6482 6483
/*
 * The cpus mask in sched_group and sched_domain hangs off the end.
6484 6485 6486
 *
 * ( See the the comments in include/linux/sched.h:struct sched_group
 *   and struct sched_domain. )
6487 6488 6489 6490 6491 6492 6493 6494 6495 6496 6497
 */
struct static_sched_group {
	struct sched_group sg;
	DECLARE_BITMAP(cpus, CONFIG_NR_CPUS);
};

struct static_sched_domain {
	struct sched_domain sd;
	DECLARE_BITMAP(span, CONFIG_NR_CPUS);
};

6498 6499 6500 6501 6502 6503 6504 6505 6506 6507
struct s_data {
#ifdef CONFIG_NUMA
	int			sd_allnodes;
	cpumask_var_t		domainspan;
	cpumask_var_t		covered;
	cpumask_var_t		notcovered;
#endif
	cpumask_var_t		nodemask;
	cpumask_var_t		this_sibling_map;
	cpumask_var_t		this_core_map;
6508
	cpumask_var_t		this_book_map;
6509 6510 6511 6512 6513 6514
	cpumask_var_t		send_covered;
	cpumask_var_t		tmpmask;
	struct sched_group	**sched_group_nodes;
	struct root_domain	*rd;
};

6515 6516 6517 6518 6519
enum s_alloc {
	sa_sched_groups = 0,
	sa_rootdomain,
	sa_tmpmask,
	sa_send_covered,
6520
	sa_this_book_map,
6521 6522 6523 6524 6525 6526 6527 6528 6529 6530 6531 6532
	sa_this_core_map,
	sa_this_sibling_map,
	sa_nodemask,
	sa_sched_group_nodes,
#ifdef CONFIG_NUMA
	sa_notcovered,
	sa_covered,
	sa_domainspan,
#endif
	sa_none,
};

6533
/*
6534
 * SMT sched-domains:
6535
 */
L
Linus Torvalds 已提交
6536
#ifdef CONFIG_SCHED_SMT
6537
static DEFINE_PER_CPU(struct static_sched_domain, cpu_domains);
6538
static DEFINE_PER_CPU(struct static_sched_group, sched_groups);
6539

I
Ingo Molnar 已提交
6540
static int
6541 6542
cpu_to_cpu_group(int cpu, const struct cpumask *cpu_map,
		 struct sched_group **sg, struct cpumask *unused)
L
Linus Torvalds 已提交
6543
{
6544
	if (sg)
6545
		*sg = &per_cpu(sched_groups, cpu).sg;
L
Linus Torvalds 已提交
6546 6547
	return cpu;
}
6548
#endif /* CONFIG_SCHED_SMT */
L
Linus Torvalds 已提交
6549

6550 6551 6552
/*
 * multi-core sched-domains:
 */
6553
#ifdef CONFIG_SCHED_MC
6554 6555
static DEFINE_PER_CPU(struct static_sched_domain, core_domains);
static DEFINE_PER_CPU(struct static_sched_group, sched_group_core);
6556

I
Ingo Molnar 已提交
6557
static int
6558 6559
cpu_to_core_group(int cpu, const struct cpumask *cpu_map,
		  struct sched_group **sg, struct cpumask *mask)
6560
{
6561
	int group;
6562
#ifdef CONFIG_SCHED_SMT
6563
	cpumask_and(mask, topology_thread_cpumask(cpu), cpu_map);
6564
	group = cpumask_first(mask);
6565 6566 6567
#else
	group = cpu;
#endif
6568
	if (sg)
6569
		*sg = &per_cpu(sched_group_core, group).sg;
6570
	return group;
6571
}
6572
#endif /* CONFIG_SCHED_MC */
6573

6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598
/*
 * book sched-domains:
 */
#ifdef CONFIG_SCHED_BOOK
static DEFINE_PER_CPU(struct static_sched_domain, book_domains);
static DEFINE_PER_CPU(struct static_sched_group, sched_group_book);

static int
cpu_to_book_group(int cpu, const struct cpumask *cpu_map,
		  struct sched_group **sg, struct cpumask *mask)
{
	int group = cpu;
#ifdef CONFIG_SCHED_MC
	cpumask_and(mask, cpu_coregroup_mask(cpu), cpu_map);
	group = cpumask_first(mask);
#elif defined(CONFIG_SCHED_SMT)
	cpumask_and(mask, topology_thread_cpumask(cpu), cpu_map);
	group = cpumask_first(mask);
#endif
	if (sg)
		*sg = &per_cpu(sched_group_book, group).sg;
	return group;
}
#endif /* CONFIG_SCHED_BOOK */

6599 6600
static DEFINE_PER_CPU(struct static_sched_domain, phys_domains);
static DEFINE_PER_CPU(struct static_sched_group, sched_group_phys);
6601

I
Ingo Molnar 已提交
6602
static int
6603 6604
cpu_to_phys_group(int cpu, const struct cpumask *cpu_map,
		  struct sched_group **sg, struct cpumask *mask)
L
Linus Torvalds 已提交
6605
{
6606
	int group;
6607 6608 6609 6610
#ifdef CONFIG_SCHED_BOOK
	cpumask_and(mask, cpu_book_mask(cpu), cpu_map);
	group = cpumask_first(mask);
#elif defined(CONFIG_SCHED_MC)
6611
	cpumask_and(mask, cpu_coregroup_mask(cpu), cpu_map);
6612
	group = cpumask_first(mask);
6613
#elif defined(CONFIG_SCHED_SMT)
6614
	cpumask_and(mask, topology_thread_cpumask(cpu), cpu_map);
6615
	group = cpumask_first(mask);
L
Linus Torvalds 已提交
6616
#else
6617
	group = cpu;
L
Linus Torvalds 已提交
6618
#endif
6619
	if (sg)
6620
		*sg = &per_cpu(sched_group_phys, group).sg;
6621
	return group;
L
Linus Torvalds 已提交
6622 6623 6624 6625
}

#ifdef CONFIG_NUMA
/*
6626 6627 6628
 * The init_sched_build_groups can't handle what we want to do with node
 * groups, so roll our own. Now each node has its own list of groups which
 * gets dynamically allocated.
L
Linus Torvalds 已提交
6629
 */
6630
static DEFINE_PER_CPU(struct static_sched_domain, node_domains);
6631
static struct sched_group ***sched_group_nodes_bycpu;
L
Linus Torvalds 已提交
6632

6633
static DEFINE_PER_CPU(struct static_sched_domain, allnodes_domains);
6634
static DEFINE_PER_CPU(struct static_sched_group, sched_group_allnodes);
6635

6636 6637 6638
static int cpu_to_allnodes_group(int cpu, const struct cpumask *cpu_map,
				 struct sched_group **sg,
				 struct cpumask *nodemask)
6639
{
6640 6641
	int group;

6642
	cpumask_and(nodemask, cpumask_of_node(cpu_to_node(cpu)), cpu_map);
6643
	group = cpumask_first(nodemask);
6644 6645

	if (sg)
6646
		*sg = &per_cpu(sched_group_allnodes, group).sg;
6647
	return group;
L
Linus Torvalds 已提交
6648
}
6649

6650 6651 6652 6653 6654 6655 6656
static void init_numa_sched_groups_power(struct sched_group *group_head)
{
	struct sched_group *sg = group_head;
	int j;

	if (!sg)
		return;
6657
	do {
6658
		for_each_cpu(j, sched_group_cpus(sg)) {
6659
			struct sched_domain *sd;
6660

6661
			sd = &per_cpu(phys_domains, j).sd;
6662
			if (j != group_first_cpu(sd->groups)) {
6663 6664 6665 6666 6667 6668
				/*
				 * Only add "power" once for each
				 * physical package.
				 */
				continue;
			}
6669

6670
			sg->cpu_power += sd->groups->cpu_power;
6671 6672 6673
		}
		sg = sg->next;
	} while (sg != group_head);
6674
}
6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695

static int build_numa_sched_groups(struct s_data *d,
				   const struct cpumask *cpu_map, int num)
{
	struct sched_domain *sd;
	struct sched_group *sg, *prev;
	int n, j;

	cpumask_clear(d->covered);
	cpumask_and(d->nodemask, cpumask_of_node(num), cpu_map);
	if (cpumask_empty(d->nodemask)) {
		d->sched_group_nodes[num] = NULL;
		goto out;
	}

	sched_domain_node_span(num, d->domainspan);
	cpumask_and(d->domainspan, d->domainspan, cpu_map);

	sg = kmalloc_node(sizeof(struct sched_group) + cpumask_size(),
			  GFP_KERNEL, num);
	if (!sg) {
P
Peter Zijlstra 已提交
6696 6697
		printk(KERN_WARNING "Can not alloc domain group for node %d\n",
		       num);
6698 6699 6700 6701 6702 6703 6704 6705 6706
		return -ENOMEM;
	}
	d->sched_group_nodes[num] = sg;

	for_each_cpu(j, d->nodemask) {
		sd = &per_cpu(node_domains, j).sd;
		sd->groups = sg;
	}

6707
	sg->cpu_power = 0;
6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721 6722 6723 6724 6725
	cpumask_copy(sched_group_cpus(sg), d->nodemask);
	sg->next = sg;
	cpumask_or(d->covered, d->covered, d->nodemask);

	prev = sg;
	for (j = 0; j < nr_node_ids; j++) {
		n = (num + j) % nr_node_ids;
		cpumask_complement(d->notcovered, d->covered);
		cpumask_and(d->tmpmask, d->notcovered, cpu_map);
		cpumask_and(d->tmpmask, d->tmpmask, d->domainspan);
		if (cpumask_empty(d->tmpmask))
			break;
		cpumask_and(d->tmpmask, d->tmpmask, cpumask_of_node(n));
		if (cpumask_empty(d->tmpmask))
			continue;
		sg = kmalloc_node(sizeof(struct sched_group) + cpumask_size(),
				  GFP_KERNEL, num);
		if (!sg) {
P
Peter Zijlstra 已提交
6726 6727
			printk(KERN_WARNING
			       "Can not alloc domain group for node %d\n", j);
6728 6729
			return -ENOMEM;
		}
6730
		sg->cpu_power = 0;
6731 6732 6733 6734 6735 6736 6737 6738 6739
		cpumask_copy(sched_group_cpus(sg), d->tmpmask);
		sg->next = prev->next;
		cpumask_or(d->covered, d->covered, d->tmpmask);
		prev->next = sg;
		prev = sg;
	}
out:
	return 0;
}
6740
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
6741

6742
#ifdef CONFIG_NUMA
6743
/* Free memory allocated for various sched_group structures */
6744 6745
static void free_sched_groups(const struct cpumask *cpu_map,
			      struct cpumask *nodemask)
6746
{
6747
	int cpu, i;
6748

6749
	for_each_cpu(cpu, cpu_map) {
6750 6751 6752 6753 6754 6755
		struct sched_group **sched_group_nodes
			= sched_group_nodes_bycpu[cpu];

		if (!sched_group_nodes)
			continue;

6756
		for (i = 0; i < nr_node_ids; i++) {
6757 6758
			struct sched_group *oldsg, *sg = sched_group_nodes[i];

6759
			cpumask_and(nodemask, cpumask_of_node(i), cpu_map);
6760
			if (cpumask_empty(nodemask))
6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776
				continue;

			if (sg == NULL)
				continue;
			sg = sg->next;
next_sg:
			oldsg = sg;
			sg = sg->next;
			kfree(oldsg);
			if (oldsg != sched_group_nodes[i])
				goto next_sg;
		}
		kfree(sched_group_nodes);
		sched_group_nodes_bycpu[cpu] = NULL;
	}
}
6777
#else /* !CONFIG_NUMA */
6778 6779
static void free_sched_groups(const struct cpumask *cpu_map,
			      struct cpumask *nodemask)
6780 6781
{
}
6782
#endif /* CONFIG_NUMA */
6783

6784 6785 6786 6787 6788 6789 6790 6791 6792 6793 6794 6795 6796 6797
/*
 * 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)
{
	struct sched_domain *child;
	struct sched_group *group;
6798 6799
	long power;
	int weight;
6800 6801 6802

	WARN_ON(!sd || !sd->groups);

6803
	if (cpu != group_first_cpu(sd->groups))
6804 6805 6806 6807
		return;

	child = sd->child;

6808
	sd->groups->cpu_power = 0;
6809

6810 6811 6812 6813 6814
	if (!child) {
		power = SCHED_LOAD_SCALE;
		weight = cpumask_weight(sched_domain_span(sd));
		/*
		 * SMT siblings share the power of a single core.
P
Peter Zijlstra 已提交
6815 6816 6817
		 * Usually multiple threads get a better yield out of
		 * that one core than a single thread would have,
		 * reflect that in sd->smt_gain.
6818
		 */
P
Peter Zijlstra 已提交
6819 6820
		if ((sd->flags & SD_SHARE_CPUPOWER) && weight > 1) {
			power *= sd->smt_gain;
6821
			power /= weight;
P
Peter Zijlstra 已提交
6822 6823
			power >>= SCHED_LOAD_SHIFT;
		}
6824
		sd->groups->cpu_power += power;
6825 6826 6827 6828
		return;
	}

	/*
6829
	 * Add cpu_power of each child group to this groups cpu_power.
6830 6831 6832
	 */
	group = child->groups;
	do {
6833
		sd->groups->cpu_power += group->cpu_power;
6834 6835 6836 6837
		group = group->next;
	} while (group != child->groups);
}

6838 6839 6840 6841 6842
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6843 6844 6845 6846 6847 6848
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

6849
#define	SD_INIT(sd, type)	sd_init_##type(sd)
6850

6851 6852 6853 6854 6855
#define SD_INIT_FUNC(type)	\
static noinline void sd_init_##type(struct sched_domain *sd)	\
{								\
	memset(sd, 0, sizeof(*sd));				\
	*sd = SD_##type##_INIT;					\
6856
	sd->level = SD_LV_##type;				\
6857
	SD_INIT_NAME(sd, type);					\
6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870
}

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
6871 6872 6873
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
6874

6875 6876 6877 6878
static int default_relax_domain_level = -1;

static int __init setup_relax_domain_level(char *str)
{
6879 6880 6881 6882 6883 6884
	unsigned long val;

	val = simple_strtoul(str, NULL, 0);
	if (val < SD_LV_MAX)
		default_relax_domain_level = val;

6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902
	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 */
6903
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6904 6905
	} else {
		/* turn on idle balance on this domain */
6906
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6907 6908 6909
	}
}

6910 6911 6912 6913 6914 6915 6916 6917 6918 6919 6920 6921 6922
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_sched_groups:
		free_sched_groups(cpu_map, d->tmpmask); /* fall through */
		d->sched_group_nodes = NULL;
	case sa_rootdomain:
		free_rootdomain(d->rd); /* fall through */
	case sa_tmpmask:
		free_cpumask_var(d->tmpmask); /* fall through */
	case sa_send_covered:
		free_cpumask_var(d->send_covered); /* fall through */
6923 6924
	case sa_this_book_map:
		free_cpumask_var(d->this_book_map); /* fall through */
6925 6926 6927 6928 6929 6930 6931
	case sa_this_core_map:
		free_cpumask_var(d->this_core_map); /* fall through */
	case sa_this_sibling_map:
		free_cpumask_var(d->this_sibling_map); /* fall through */
	case sa_nodemask:
		free_cpumask_var(d->nodemask); /* fall through */
	case sa_sched_group_nodes:
6932
#ifdef CONFIG_NUMA
6933 6934 6935 6936 6937 6938 6939
		kfree(d->sched_group_nodes); /* fall through */
	case sa_notcovered:
		free_cpumask_var(d->notcovered); /* fall through */
	case sa_covered:
		free_cpumask_var(d->covered); /* fall through */
	case sa_domainspan:
		free_cpumask_var(d->domainspan); /* fall through */
6940
#endif
6941 6942 6943 6944
	case sa_none:
		break;
	}
}
6945

6946 6947 6948
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6949
#ifdef CONFIG_NUMA
6950 6951 6952 6953 6954 6955 6956 6957 6958 6959
	if (!alloc_cpumask_var(&d->domainspan, GFP_KERNEL))
		return sa_none;
	if (!alloc_cpumask_var(&d->covered, GFP_KERNEL))
		return sa_domainspan;
	if (!alloc_cpumask_var(&d->notcovered, GFP_KERNEL))
		return sa_covered;
	/* Allocate the per-node list of sched groups */
	d->sched_group_nodes = kcalloc(nr_node_ids,
				      sizeof(struct sched_group *), GFP_KERNEL);
	if (!d->sched_group_nodes) {
P
Peter Zijlstra 已提交
6960
		printk(KERN_WARNING "Can not alloc sched group node list\n");
6961
		return sa_notcovered;
6962
	}
6963
	sched_group_nodes_bycpu[cpumask_first(cpu_map)] = d->sched_group_nodes;
6964
#endif
6965 6966 6967 6968 6969 6970
	if (!alloc_cpumask_var(&d->nodemask, GFP_KERNEL))
		return sa_sched_group_nodes;
	if (!alloc_cpumask_var(&d->this_sibling_map, GFP_KERNEL))
		return sa_nodemask;
	if (!alloc_cpumask_var(&d->this_core_map, GFP_KERNEL))
		return sa_this_sibling_map;
6971
	if (!alloc_cpumask_var(&d->this_book_map, GFP_KERNEL))
6972
		return sa_this_core_map;
6973 6974
	if (!alloc_cpumask_var(&d->send_covered, GFP_KERNEL))
		return sa_this_book_map;
6975 6976 6977 6978
	if (!alloc_cpumask_var(&d->tmpmask, GFP_KERNEL))
		return sa_send_covered;
	d->rd = alloc_rootdomain();
	if (!d->rd) {
P
Peter Zijlstra 已提交
6979
		printk(KERN_WARNING "Cannot alloc root domain\n");
6980
		return sa_tmpmask;
G
Gregory Haskins 已提交
6981
	}
6982 6983
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
6984

6985 6986 6987 6988
static struct sched_domain *__build_numa_sched_domains(struct s_data *d,
	const struct cpumask *cpu_map, struct sched_domain_attr *attr, int i)
{
	struct sched_domain *sd = NULL;
6989
#ifdef CONFIG_NUMA
6990
	struct sched_domain *parent;
L
Linus Torvalds 已提交
6991

6992 6993 6994 6995 6996
	d->sd_allnodes = 0;
	if (cpumask_weight(cpu_map) >
	    SD_NODES_PER_DOMAIN * cpumask_weight(d->nodemask)) {
		sd = &per_cpu(allnodes_domains, i).sd;
		SD_INIT(sd, ALLNODES);
6997
		set_domain_attribute(sd, attr);
6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011
		cpumask_copy(sched_domain_span(sd), cpu_map);
		cpu_to_allnodes_group(i, cpu_map, &sd->groups, d->tmpmask);
		d->sd_allnodes = 1;
	}
	parent = sd;

	sd = &per_cpu(node_domains, i).sd;
	SD_INIT(sd, NODE);
	set_domain_attribute(sd, attr);
	sched_domain_node_span(cpu_to_node(i), sched_domain_span(sd));
	sd->parent = parent;
	if (parent)
		parent->child = sd;
	cpumask_and(sched_domain_span(sd), sched_domain_span(sd), cpu_map);
L
Linus Torvalds 已提交
7012
#endif
7013 7014
	return sd;
}
L
Linus Torvalds 已提交
7015

7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030
static struct sched_domain *__build_cpu_sched_domain(struct s_data *d,
	const struct cpumask *cpu_map, struct sched_domain_attr *attr,
	struct sched_domain *parent, int i)
{
	struct sched_domain *sd;
	sd = &per_cpu(phys_domains, i).sd;
	SD_INIT(sd, CPU);
	set_domain_attribute(sd, attr);
	cpumask_copy(sched_domain_span(sd), d->nodemask);
	sd->parent = parent;
	if (parent)
		parent->child = sd;
	cpu_to_phys_group(i, cpu_map, &sd->groups, d->tmpmask);
	return sd;
}
L
Linus Torvalds 已提交
7031

7032 7033 7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046 7047 7048
static struct sched_domain *__build_book_sched_domain(struct s_data *d,
	const struct cpumask *cpu_map, struct sched_domain_attr *attr,
	struct sched_domain *parent, int i)
{
	struct sched_domain *sd = parent;
#ifdef CONFIG_SCHED_BOOK
	sd = &per_cpu(book_domains, i).sd;
	SD_INIT(sd, BOOK);
	set_domain_attribute(sd, attr);
	cpumask_and(sched_domain_span(sd), cpu_map, cpu_book_mask(i));
	sd->parent = parent;
	parent->child = sd;
	cpu_to_book_group(i, cpu_map, &sd->groups, d->tmpmask);
#endif
	return sd;
}

7049 7050 7051 7052 7053
static struct sched_domain *__build_mc_sched_domain(struct s_data *d,
	const struct cpumask *cpu_map, struct sched_domain_attr *attr,
	struct sched_domain *parent, int i)
{
	struct sched_domain *sd = parent;
7054
#ifdef CONFIG_SCHED_MC
7055 7056 7057 7058 7059 7060 7061
	sd = &per_cpu(core_domains, i).sd;
	SD_INIT(sd, MC);
	set_domain_attribute(sd, attr);
	cpumask_and(sched_domain_span(sd), cpu_map, cpu_coregroup_mask(i));
	sd->parent = parent;
	parent->child = sd;
	cpu_to_core_group(i, cpu_map, &sd->groups, d->tmpmask);
7062
#endif
7063 7064
	return sd;
}
7065

7066 7067 7068 7069 7070
static struct sched_domain *__build_smt_sched_domain(struct s_data *d,
	const struct cpumask *cpu_map, struct sched_domain_attr *attr,
	struct sched_domain *parent, int i)
{
	struct sched_domain *sd = parent;
L
Linus Torvalds 已提交
7071
#ifdef CONFIG_SCHED_SMT
7072 7073 7074 7075 7076 7077 7078
	sd = &per_cpu(cpu_domains, i).sd;
	SD_INIT(sd, SIBLING);
	set_domain_attribute(sd, attr);
	cpumask_and(sched_domain_span(sd), cpu_map, topology_thread_cpumask(i));
	sd->parent = parent;
	parent->child = sd;
	cpu_to_cpu_group(i, cpu_map, &sd->groups, d->tmpmask);
L
Linus Torvalds 已提交
7079
#endif
7080 7081
	return sd;
}
L
Linus Torvalds 已提交
7082

7083 7084 7085 7086
static void build_sched_groups(struct s_data *d, enum sched_domain_level l,
			       const struct cpumask *cpu_map, int cpu)
{
	switch (l) {
L
Linus Torvalds 已提交
7087
#ifdef CONFIG_SCHED_SMT
7088 7089 7090 7091 7092 7093 7094 7095
	case SD_LV_SIBLING: /* set up CPU (sibling) groups */
		cpumask_and(d->this_sibling_map, cpu_map,
			    topology_thread_cpumask(cpu));
		if (cpu == cpumask_first(d->this_sibling_map))
			init_sched_build_groups(d->this_sibling_map, cpu_map,
						&cpu_to_cpu_group,
						d->send_covered, d->tmpmask);
		break;
L
Linus Torvalds 已提交
7096
#endif
7097
#ifdef CONFIG_SCHED_MC
7098 7099 7100 7101 7102 7103 7104
	case SD_LV_MC: /* set up multi-core groups */
		cpumask_and(d->this_core_map, cpu_map, cpu_coregroup_mask(cpu));
		if (cpu == cpumask_first(d->this_core_map))
			init_sched_build_groups(d->this_core_map, cpu_map,
						&cpu_to_core_group,
						d->send_covered, d->tmpmask);
		break;
7105 7106 7107 7108 7109 7110 7111 7112 7113
#endif
#ifdef CONFIG_SCHED_BOOK
	case SD_LV_BOOK: /* set up book groups */
		cpumask_and(d->this_book_map, cpu_map, cpu_book_mask(cpu));
		if (cpu == cpumask_first(d->this_book_map))
			init_sched_build_groups(d->this_book_map, cpu_map,
						&cpu_to_book_group,
						d->send_covered, d->tmpmask);
		break;
7114
#endif
7115 7116 7117 7118 7119 7120 7121
	case SD_LV_CPU: /* set up physical groups */
		cpumask_and(d->nodemask, cpumask_of_node(cpu), cpu_map);
		if (!cpumask_empty(d->nodemask))
			init_sched_build_groups(d->nodemask, cpu_map,
						&cpu_to_phys_group,
						d->send_covered, d->tmpmask);
		break;
L
Linus Torvalds 已提交
7122
#ifdef CONFIG_NUMA
7123 7124 7125 7126 7127
	case SD_LV_ALLNODES:
		init_sched_build_groups(cpu_map, cpu_map, &cpu_to_allnodes_group,
					d->send_covered, d->tmpmask);
		break;
#endif
7128 7129
	default:
		break;
7130
	}
7131
}
7132

7133 7134 7135 7136 7137 7138 7139 7140 7141
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
static int __build_sched_domains(const struct cpumask *cpu_map,
				 struct sched_domain_attr *attr)
{
	enum s_alloc alloc_state = sa_none;
	struct s_data d;
7142
	struct sched_domain *sd;
7143
	int i;
7144
#ifdef CONFIG_NUMA
7145
	d.sd_allnodes = 0;
7146
#endif
7147

7148 7149 7150 7151
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
	alloc_state = sa_sched_groups;
7152

L
Linus Torvalds 已提交
7153
	/*
7154
	 * Set up domains for cpus specified by the cpu_map.
L
Linus Torvalds 已提交
7155
	 */
7156
	for_each_cpu(i, cpu_map) {
7157 7158
		cpumask_and(d.nodemask, cpumask_of_node(cpu_to_node(i)),
			    cpu_map);
I
Ingo Molnar 已提交
7159

7160
		sd = __build_numa_sched_domains(&d, cpu_map, attr, i);
7161
		sd = __build_cpu_sched_domain(&d, cpu_map, attr, sd, i);
7162
		sd = __build_book_sched_domain(&d, cpu_map, attr, sd, i);
7163
		sd = __build_mc_sched_domain(&d, cpu_map, attr, sd, i);
7164
		sd = __build_smt_sched_domain(&d, cpu_map, attr, sd, i);
L
Linus Torvalds 已提交
7165
	}
7166

7167
	for_each_cpu(i, cpu_map) {
7168
		build_sched_groups(&d, SD_LV_SIBLING, cpu_map, i);
7169
		build_sched_groups(&d, SD_LV_BOOK, cpu_map, i);
7170
		build_sched_groups(&d, SD_LV_MC, cpu_map, i);
L
Linus Torvalds 已提交
7171
	}
7172

L
Linus Torvalds 已提交
7173
	/* Set up physical groups */
7174 7175
	for (i = 0; i < nr_node_ids; i++)
		build_sched_groups(&d, SD_LV_CPU, cpu_map, i);
7176

L
Linus Torvalds 已提交
7177 7178
#ifdef CONFIG_NUMA
	/* Set up node groups */
7179 7180
	if (d.sd_allnodes)
		build_sched_groups(&d, SD_LV_ALLNODES, cpu_map, 0);
7181

7182 7183
	for (i = 0; i < nr_node_ids; i++)
		if (build_numa_sched_groups(&d, cpu_map, i))
7184
			goto error;
L
Linus Torvalds 已提交
7185 7186 7187
#endif

	/* Calculate CPU power for physical packages and nodes */
7188
#ifdef CONFIG_SCHED_SMT
7189
	for_each_cpu(i, cpu_map) {
7190
		sd = &per_cpu(cpu_domains, i).sd;
7191
		init_sched_groups_power(i, sd);
7192
	}
L
Linus Torvalds 已提交
7193
#endif
7194
#ifdef CONFIG_SCHED_MC
7195
	for_each_cpu(i, cpu_map) {
7196
		sd = &per_cpu(core_domains, i).sd;
7197
		init_sched_groups_power(i, sd);
7198 7199
	}
#endif
7200 7201 7202 7203 7204 7205
#ifdef CONFIG_SCHED_BOOK
	for_each_cpu(i, cpu_map) {
		sd = &per_cpu(book_domains, i).sd;
		init_sched_groups_power(i, sd);
	}
#endif
7206

7207
	for_each_cpu(i, cpu_map) {
7208
		sd = &per_cpu(phys_domains, i).sd;
7209
		init_sched_groups_power(i, sd);
L
Linus Torvalds 已提交
7210 7211
	}

7212
#ifdef CONFIG_NUMA
7213
	for (i = 0; i < nr_node_ids; i++)
7214
		init_numa_sched_groups_power(d.sched_group_nodes[i]);
7215

7216
	if (d.sd_allnodes) {
7217
		struct sched_group *sg;
7218

7219
		cpu_to_allnodes_group(cpumask_first(cpu_map), cpu_map, &sg,
7220
								d.tmpmask);
7221 7222
		init_numa_sched_groups_power(sg);
	}
7223 7224
#endif

L
Linus Torvalds 已提交
7225
	/* Attach the domains */
7226
	for_each_cpu(i, cpu_map) {
L
Linus Torvalds 已提交
7227
#ifdef CONFIG_SCHED_SMT
7228
		sd = &per_cpu(cpu_domains, i).sd;
7229
#elif defined(CONFIG_SCHED_MC)
7230
		sd = &per_cpu(core_domains, i).sd;
7231 7232
#elif defined(CONFIG_SCHED_BOOK)
		sd = &per_cpu(book_domains, i).sd;
L
Linus Torvalds 已提交
7233
#else
7234
		sd = &per_cpu(phys_domains, i).sd;
L
Linus Torvalds 已提交
7235
#endif
7236
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
7237
	}
7238

7239 7240 7241
	d.sched_group_nodes = NULL; /* don't free this we still need it */
	__free_domain_allocs(&d, sa_tmpmask, cpu_map);
	return 0;
7242 7243

error:
7244 7245
	__free_domain_allocs(&d, alloc_state, cpu_map);
	return -ENOMEM;
L
Linus Torvalds 已提交
7246
}
P
Paul Jackson 已提交
7247

7248
static int build_sched_domains(const struct cpumask *cpu_map)
7249 7250 7251 7252
{
	return __build_sched_domains(cpu_map, NULL);
}

7253
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
7254
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
7255 7256
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
7257 7258 7259

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
7260 7261
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
7262
 */
7263
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
7264

7265 7266 7267 7268 7269 7270
/*
 * 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)
7271
{
7272
	return 0;
7273 7274
}

7275 7276 7277 7278 7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294 7295 7296 7297 7298 7299
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);
}

7300
/*
I
Ingo Molnar 已提交
7301
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
7302 7303
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
7304
 */
7305
static int arch_init_sched_domains(const struct cpumask *cpu_map)
7306
{
7307 7308
	int err;

7309
	arch_update_cpu_topology();
P
Paul Jackson 已提交
7310
	ndoms_cur = 1;
7311
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
7312
	if (!doms_cur)
7313 7314
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
7315
	dattr_cur = NULL;
7316
	err = build_sched_domains(doms_cur[0]);
7317
	register_sched_domain_sysctl();
7318 7319

	return err;
7320 7321
}

7322 7323
static void arch_destroy_sched_domains(const struct cpumask *cpu_map,
				       struct cpumask *tmpmask)
L
Linus Torvalds 已提交
7324
{
7325
	free_sched_groups(cpu_map, tmpmask);
7326
}
L
Linus Torvalds 已提交
7327

7328 7329 7330 7331
/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
7332
static void detach_destroy_domains(const struct cpumask *cpu_map)
7333
{
7334 7335
	/* Save because hotplug lock held. */
	static DECLARE_BITMAP(tmpmask, CONFIG_NR_CPUS);
7336 7337
	int i;

7338
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
7339
		cpu_attach_domain(NULL, &def_root_domain, i);
7340
	synchronize_sched();
7341
	arch_destroy_sched_domains(cpu_map, to_cpumask(tmpmask));
7342 7343
}

7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359
/* handle null as "default" */
static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur,
			struct sched_domain_attr *new, int idx_new)
{
	struct sched_domain_attr tmp;

	/* fast path */
	if (!new && !cur)
		return 1;

	tmp = SD_ATTR_INIT;
	return !memcmp(cur ? (cur + idx_cur) : &tmp,
			new ? (new + idx_new) : &tmp,
			sizeof(struct sched_domain_attr));
}

P
Paul Jackson 已提交
7360 7361
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
7362
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
7363 7364 7365
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
7366
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
7367 7368 7369
 * The masks don't intersect (don't overlap.) We should setup one
 * sched domain for each mask. CPUs not in any of the cpumasks will
 * not be load balanced. If the same cpumask appears both in the
P
Paul Jackson 已提交
7370 7371 7372
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
7373 7374 7375 7376 7377 7378
 * The passed in 'doms_new' should be allocated using
 * alloc_sched_domains.  This routine takes ownership of it and will
 * free_sched_domains it when done with it. If the caller failed the
 * alloc call, then it can pass in doms_new == NULL && ndoms_new == 1,
 * and partition_sched_domains() will fallback to the single partition
 * 'fallback_doms', it also forces the domains to be rebuilt.
P
Paul Jackson 已提交
7379
 *
7380
 * If doms_new == NULL it will be replaced with cpu_online_mask.
7381 7382
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
7383
 *
P
Paul Jackson 已提交
7384 7385
 * Call with hotplug lock held
 */
7386
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
7387
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
7388
{
7389
	int i, j, n;
7390
	int new_topology;
P
Paul Jackson 已提交
7391

7392
	mutex_lock(&sched_domains_mutex);
7393

7394 7395 7396
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

7397 7398 7399
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

7400
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
7401 7402 7403

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
7404
		for (j = 0; j < n && !new_topology; j++) {
7405
			if (cpumask_equal(doms_cur[i], doms_new[j])
7406
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
7407 7408 7409
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
7410
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
7411 7412 7413 7414
match1:
		;
	}

7415 7416
	if (doms_new == NULL) {
		ndoms_cur = 0;
7417
		doms_new = &fallback_doms;
7418
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
7419
		WARN_ON_ONCE(dattr_new);
7420 7421
	}

P
Paul Jackson 已提交
7422 7423
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
7424
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
7425
			if (cpumask_equal(doms_new[i], doms_cur[j])
7426
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
7427 7428 7429
				goto match2;
		}
		/* no match - add a new doms_new */
7430
		__build_sched_domains(doms_new[i],
7431
					dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
7432 7433 7434 7435 7436
match2:
		;
	}

	/* Remember the new sched domains */
7437 7438
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
7439
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
7440
	doms_cur = doms_new;
7441
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
7442
	ndoms_cur = ndoms_new;
7443 7444

	register_sched_domain_sysctl();
7445

7446
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
7447 7448
}

7449
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
7450
static void arch_reinit_sched_domains(void)
7451
{
7452
	get_online_cpus();
7453 7454 7455 7456

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

7457
	rebuild_sched_domains();
7458
	put_online_cpus();
7459 7460 7461 7462
}

static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
{
7463
	unsigned int level = 0;
7464

7465 7466 7467 7468 7469 7470 7471 7472 7473 7474 7475
	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)
7476 7477 7478
		return -EINVAL;

	if (smt)
7479
		sched_smt_power_savings = level;
7480
	else
7481
		sched_mc_power_savings = level;
7482

7483
	arch_reinit_sched_domains();
7484

7485
	return count;
7486 7487 7488
}

#ifdef CONFIG_SCHED_MC
7489
static ssize_t sched_mc_power_savings_show(struct sysdev_class *class,
7490
					   struct sysdev_class_attribute *attr,
7491
					   char *page)
7492 7493 7494
{
	return sprintf(page, "%u\n", sched_mc_power_savings);
}
7495
static ssize_t sched_mc_power_savings_store(struct sysdev_class *class,
7496
					    struct sysdev_class_attribute *attr,
7497
					    const char *buf, size_t count)
7498 7499 7500
{
	return sched_power_savings_store(buf, count, 0);
}
7501 7502 7503
static SYSDEV_CLASS_ATTR(sched_mc_power_savings, 0644,
			 sched_mc_power_savings_show,
			 sched_mc_power_savings_store);
7504 7505 7506
#endif

#ifdef CONFIG_SCHED_SMT
7507
static ssize_t sched_smt_power_savings_show(struct sysdev_class *dev,
7508
					    struct sysdev_class_attribute *attr,
7509
					    char *page)
7510 7511 7512
{
	return sprintf(page, "%u\n", sched_smt_power_savings);
}
7513
static ssize_t sched_smt_power_savings_store(struct sysdev_class *dev,
7514
					     struct sysdev_class_attribute *attr,
7515
					     const char *buf, size_t count)
7516 7517 7518
{
	return sched_power_savings_store(buf, count, 1);
}
7519 7520
static SYSDEV_CLASS_ATTR(sched_smt_power_savings, 0644,
		   sched_smt_power_savings_show,
A
Adrian Bunk 已提交
7521 7522 7523
		   sched_smt_power_savings_store);
#endif

7524
int __init sched_create_sysfs_power_savings_entries(struct sysdev_class *cls)
A
Adrian Bunk 已提交
7525 7526 7527 7528 7529 7530 7531 7532 7533 7534 7535 7536 7537 7538 7539
{
	int err = 0;

#ifdef CONFIG_SCHED_SMT
	if (smt_capable())
		err = sysfs_create_file(&cls->kset.kobj,
					&attr_sched_smt_power_savings.attr);
#endif
#ifdef CONFIG_SCHED_MC
	if (!err && mc_capable())
		err = sysfs_create_file(&cls->kset.kobj,
					&attr_sched_mc_power_savings.attr);
#endif
	return err;
}
7540
#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
7541

L
Linus Torvalds 已提交
7542
/*
7543 7544 7545
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
L
Linus Torvalds 已提交
7546
 */
7547 7548
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
7549
{
7550
	switch (action & ~CPU_TASKS_FROZEN) {
7551
	case CPU_ONLINE:
7552
	case CPU_DOWN_FAILED:
7553
		cpuset_update_active_cpus();
7554
		return NOTIFY_OK;
7555 7556 7557 7558
	default:
		return NOTIFY_DONE;
	}
}
7559

7560 7561
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
7562 7563 7564 7565 7566
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
		cpuset_update_active_cpus();
		return NOTIFY_OK;
7567 7568 7569 7570 7571 7572 7573
	default:
		return NOTIFY_DONE;
	}
}

static int update_runtime(struct notifier_block *nfb,
				unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
7574
{
P
Peter Zijlstra 已提交
7575 7576
	int cpu = (int)(long)hcpu;

L
Linus Torvalds 已提交
7577 7578
	switch (action) {
	case CPU_DOWN_PREPARE:
7579
	case CPU_DOWN_PREPARE_FROZEN:
P
Peter Zijlstra 已提交
7580
		disable_runtime(cpu_rq(cpu));
L
Linus Torvalds 已提交
7581 7582 7583
		return NOTIFY_OK;

	case CPU_DOWN_FAILED:
7584
	case CPU_DOWN_FAILED_FROZEN:
L
Linus Torvalds 已提交
7585
	case CPU_ONLINE:
7586
	case CPU_ONLINE_FROZEN:
P
Peter Zijlstra 已提交
7587
		enable_runtime(cpu_rq(cpu));
7588 7589
		return NOTIFY_OK;

L
Linus Torvalds 已提交
7590 7591 7592 7593 7594 7595 7596
	default:
		return NOTIFY_DONE;
	}
}

void __init sched_init_smp(void)
{
7597 7598 7599
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
7600
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
7601

7602 7603 7604 7605 7606
#if defined(CONFIG_NUMA)
	sched_group_nodes_bycpu = kzalloc(nr_cpu_ids * sizeof(void **),
								GFP_KERNEL);
	BUG_ON(sched_group_nodes_bycpu == NULL);
#endif
7607
	get_online_cpus();
7608
	mutex_lock(&sched_domains_mutex);
7609
	arch_init_sched_domains(cpu_active_mask);
7610 7611 7612
	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);
7613
	mutex_unlock(&sched_domains_mutex);
7614
	put_online_cpus();
7615

7616 7617
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
7618 7619 7620 7621

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

7622
	init_hrtick();
7623 7624

	/* Move init over to a non-isolated CPU */
7625
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
7626
		BUG();
I
Ingo Molnar 已提交
7627
	sched_init_granularity();
7628
	free_cpumask_var(non_isolated_cpus);
7629

7630
	init_sched_rt_class();
L
Linus Torvalds 已提交
7631 7632 7633 7634
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
7635
	sched_init_granularity();
L
Linus Torvalds 已提交
7636 7637 7638
}
#endif /* CONFIG_SMP */

7639 7640
const_debug unsigned int sysctl_timer_migration = 1;

L
Linus Torvalds 已提交
7641 7642 7643 7644 7645 7646 7647
int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

A
Alexey Dobriyan 已提交
7648
static void init_cfs_rq(struct cfs_rq *cfs_rq, struct rq *rq)
I
Ingo Molnar 已提交
7649 7650
{
	cfs_rq->tasks_timeline = RB_ROOT;
7651
	INIT_LIST_HEAD(&cfs_rq->tasks);
I
Ingo Molnar 已提交
7652 7653 7654
#ifdef CONFIG_FAIR_GROUP_SCHED
	cfs_rq->rq = rq;
#endif
P
Peter Zijlstra 已提交
7655
	cfs_rq->min_vruntime = (u64)(-(1LL << 20));
I
Ingo Molnar 已提交
7656 7657
}

P
Peter Zijlstra 已提交
7658 7659 7660 7661 7662 7663 7664 7665 7666 7667 7668 7669 7670
static void init_rt_rq(struct rt_rq *rt_rq, struct rq *rq)
{
	struct rt_prio_array *array;
	int i;

	array = &rt_rq->active;
	for (i = 0; i < MAX_RT_PRIO; i++) {
		INIT_LIST_HEAD(array->queue + i);
		__clear_bit(i, array->bitmap);
	}
	/* delimiter for bitsearch: */
	__set_bit(MAX_RT_PRIO, array->bitmap);

7671
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
7672
	rt_rq->highest_prio.curr = MAX_RT_PRIO;
7673
#ifdef CONFIG_SMP
7674
	rt_rq->highest_prio.next = MAX_RT_PRIO;
P
Peter Zijlstra 已提交
7675 7676
#endif
#endif
P
Peter Zijlstra 已提交
7677 7678 7679
#ifdef CONFIG_SMP
	rt_rq->rt_nr_migratory = 0;
	rt_rq->overloaded = 0;
7680
	plist_head_init_raw(&rt_rq->pushable_tasks, &rq->lock);
P
Peter Zijlstra 已提交
7681 7682 7683 7684
#endif

	rt_rq->rt_time = 0;
	rt_rq->rt_throttled = 0;
P
Peter Zijlstra 已提交
7685
	rt_rq->rt_runtime = 0;
7686
	raw_spin_lock_init(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7687

7688
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7689
	rt_rq->rt_nr_boosted = 0;
P
Peter Zijlstra 已提交
7690 7691
	rt_rq->rq = rq;
#endif
P
Peter Zijlstra 已提交
7692 7693
}

P
Peter Zijlstra 已提交
7694
#ifdef CONFIG_FAIR_GROUP_SCHED
7695 7696 7697
static void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq,
				struct sched_entity *se, int cpu, int add,
				struct sched_entity *parent)
P
Peter Zijlstra 已提交
7698
{
7699
	struct rq *rq = cpu_rq(cpu);
P
Peter Zijlstra 已提交
7700 7701 7702 7703 7704 7705 7706
	tg->cfs_rq[cpu] = cfs_rq;
	init_cfs_rq(cfs_rq, rq);
	cfs_rq->tg = tg;
	if (add)
		list_add(&cfs_rq->leaf_cfs_rq_list, &rq->leaf_cfs_rq_list);

	tg->se[cpu] = se;
D
Dhaval Giani 已提交
7707 7708 7709 7710
	/* se could be NULL for init_task_group */
	if (!se)
		return;

7711 7712 7713 7714 7715
	if (!parent)
		se->cfs_rq = &rq->cfs;
	else
		se->cfs_rq = parent->my_q;

P
Peter Zijlstra 已提交
7716 7717
	se->my_q = cfs_rq;
	se->load.weight = tg->shares;
7718
	se->load.inv_weight = 0;
7719
	se->parent = parent;
P
Peter Zijlstra 已提交
7720
}
7721
#endif
P
Peter Zijlstra 已提交
7722

7723
#ifdef CONFIG_RT_GROUP_SCHED
7724 7725 7726
static void init_tg_rt_entry(struct task_group *tg, struct rt_rq *rt_rq,
		struct sched_rt_entity *rt_se, int cpu, int add,
		struct sched_rt_entity *parent)
P
Peter Zijlstra 已提交
7727
{
7728 7729
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
7730 7731 7732
	tg->rt_rq[cpu] = rt_rq;
	init_rt_rq(rt_rq, rq);
	rt_rq->tg = tg;
P
Peter Zijlstra 已提交
7733
	rt_rq->rt_runtime = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7734 7735 7736 7737
	if (add)
		list_add(&rt_rq->leaf_rt_rq_list, &rq->leaf_rt_rq_list);

	tg->rt_se[cpu] = rt_se;
D
Dhaval Giani 已提交
7738 7739 7740
	if (!rt_se)
		return;

7741 7742 7743 7744 7745
	if (!parent)
		rt_se->rt_rq = &rq->rt;
	else
		rt_se->rt_rq = parent->my_q;

P
Peter Zijlstra 已提交
7746
	rt_se->my_q = rt_rq;
7747
	rt_se->parent = parent;
P
Peter Zijlstra 已提交
7748 7749 7750 7751
	INIT_LIST_HEAD(&rt_se->run_list);
}
#endif

L
Linus Torvalds 已提交
7752 7753
void __init sched_init(void)
{
I
Ingo Molnar 已提交
7754
	int i, j;
7755 7756 7757 7758 7759 7760 7761
	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 **);
7762
#endif
7763
#ifdef CONFIG_CPUMASK_OFFSTACK
7764
	alloc_size += num_possible_cpus() * cpumask_size();
7765 7766
#endif
	if (alloc_size) {
7767
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
7768 7769 7770 7771 7772 7773 7774

#ifdef CONFIG_FAIR_GROUP_SCHED
		init_task_group.se = (struct sched_entity **)ptr;
		ptr += nr_cpu_ids * sizeof(void **);

		init_task_group.cfs_rq = (struct cfs_rq **)ptr;
		ptr += nr_cpu_ids * sizeof(void **);
7775

7776
#endif /* CONFIG_FAIR_GROUP_SCHED */
7777 7778 7779 7780 7781
#ifdef CONFIG_RT_GROUP_SCHED
		init_task_group.rt_se = (struct sched_rt_entity **)ptr;
		ptr += nr_cpu_ids * sizeof(void **);

		init_task_group.rt_rq = (struct rt_rq **)ptr;
7782 7783
		ptr += nr_cpu_ids * sizeof(void **);

7784
#endif /* CONFIG_RT_GROUP_SCHED */
7785 7786 7787 7788 7789 7790
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
7791
	}
I
Ingo Molnar 已提交
7792

G
Gregory Haskins 已提交
7793 7794 7795 7796
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

7797 7798 7799 7800 7801 7802
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
	init_rt_bandwidth(&init_task_group.rt_bandwidth,
			global_rt_period(), global_rt_runtime());
7803
#endif /* CONFIG_RT_GROUP_SCHED */
7804

D
Dhaval Giani 已提交
7805
#ifdef CONFIG_CGROUP_SCHED
P
Peter Zijlstra 已提交
7806
	list_add(&init_task_group.list, &task_groups);
P
Peter Zijlstra 已提交
7807 7808
	INIT_LIST_HEAD(&init_task_group.children);

D
Dhaval Giani 已提交
7809
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
7810

7811 7812 7813 7814
#if defined CONFIG_FAIR_GROUP_SCHED && defined CONFIG_SMP
	update_shares_data = __alloc_percpu(nr_cpu_ids * sizeof(unsigned long),
					    __alignof__(unsigned long));
#endif
7815
	for_each_possible_cpu(i) {
7816
		struct rq *rq;
L
Linus Torvalds 已提交
7817 7818

		rq = cpu_rq(i);
7819
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
7820
		rq->nr_running = 0;
7821 7822
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
I
Ingo Molnar 已提交
7823
		init_cfs_rq(&rq->cfs, rq);
P
Peter Zijlstra 已提交
7824
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
7825
#ifdef CONFIG_FAIR_GROUP_SCHED
7826
		init_task_group.shares = init_task_group_load;
P
Peter Zijlstra 已提交
7827
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
7828 7829 7830 7831 7832 7833 7834 7835 7836 7837 7838 7839 7840 7841 7842
#ifdef CONFIG_CGROUP_SCHED
		/*
		 * How much cpu bandwidth does init_task_group get?
		 *
		 * 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
		 * init_task_group and its child task-groups in a fair manner,
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
		 * In other words, if init_task_group has 10 tasks of weight
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
7843
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
7844 7845 7846 7847
		 *
		 * We achieve this by letting init_task_group's tasks sit
		 * directly in rq->cfs (i.e init_task_group->se[] = NULL).
		 */
7848
		init_tg_cfs_entry(&init_task_group, &rq->cfs, NULL, i, 1, NULL);
7849
#endif
D
Dhaval Giani 已提交
7850 7851 7852
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
7853
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7854
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
D
Dhaval Giani 已提交
7855
#ifdef CONFIG_CGROUP_SCHED
7856
		init_tg_rt_entry(&init_task_group, &rq->rt, NULL, i, 1, NULL);
D
Dhaval Giani 已提交
7857
#endif
I
Ingo Molnar 已提交
7858
#endif
L
Linus Torvalds 已提交
7859

I
Ingo Molnar 已提交
7860 7861
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
7862 7863 7864

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
7865
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
7866
		rq->sd = NULL;
G
Gregory Haskins 已提交
7867
		rq->rd = NULL;
7868
		rq->cpu_power = SCHED_LOAD_SCALE;
7869
		rq->post_schedule = 0;
L
Linus Torvalds 已提交
7870
		rq->active_balance = 0;
I
Ingo Molnar 已提交
7871
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
7872
		rq->push_cpu = 0;
7873
		rq->cpu = i;
7874
		rq->online = 0;
7875 7876
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
7877
		rq_attach_root(rq, &def_root_domain);
7878 7879 7880 7881
#ifdef CONFIG_NO_HZ
		rq->nohz_balance_kick = 0;
		init_sched_softirq_csd(&per_cpu(remote_sched_softirq_cb, i));
#endif
L
Linus Torvalds 已提交
7882
#endif
P
Peter Zijlstra 已提交
7883
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
7884 7885 7886
		atomic_set(&rq->nr_iowait, 0);
	}

7887
	set_load_weight(&init_task);
7888

7889 7890 7891 7892
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

7893
#ifdef CONFIG_SMP
7894
	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);
7895 7896
#endif

7897
#ifdef CONFIG_RT_MUTEXES
7898
	plist_head_init_raw(&init_task.pi_waiters, &init_task.pi_lock);
7899 7900
#endif

L
Linus Torvalds 已提交
7901 7902 7903 7904 7905 7906 7907 7908 7909 7910 7911 7912 7913
	/*
	 * 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());
7914 7915 7916

	calc_load_update = jiffies + LOAD_FREQ;

I
Ingo Molnar 已提交
7917 7918 7919 7920
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
7921

7922
	/* Allocate the nohz_cpu_mask if CONFIG_CPUMASK_OFFSTACK */
7923
	zalloc_cpumask_var(&nohz_cpu_mask, GFP_NOWAIT);
7924
#ifdef CONFIG_SMP
7925
#ifdef CONFIG_NO_HZ
7926 7927 7928 7929 7930
	zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
	alloc_cpumask_var(&nohz.grp_idle_mask, GFP_NOWAIT);
	atomic_set(&nohz.load_balancer, nr_cpu_ids);
	atomic_set(&nohz.first_pick_cpu, nr_cpu_ids);
	atomic_set(&nohz.second_pick_cpu, nr_cpu_ids);
7931
#endif
R
Rusty Russell 已提交
7932 7933 7934
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
7935
#endif /* SMP */
7936

7937
	perf_event_init();
7938

7939
	scheduler_running = 1;
L
Linus Torvalds 已提交
7940 7941 7942
}

#ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
7943 7944
static inline int preempt_count_equals(int preempt_offset)
{
7945
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
7946 7947 7948 7949

	return (nested == PREEMPT_INATOMIC_BASE + preempt_offset);
}

7950
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
7951
{
7952
#ifdef in_atomic
L
Linus Torvalds 已提交
7953 7954
	static unsigned long prev_jiffy;	/* ratelimiting */

7955 7956
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
7957 7958 7959 7960 7961
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
7962 7963 7964 7965 7966 7967 7968
	printk(KERN_ERR
		"BUG: sleeping function called from invalid context at %s:%d\n",
			file, line);
	printk(KERN_ERR
		"in_atomic(): %d, irqs_disabled(): %d, pid: %d, name: %s\n",
			in_atomic(), irqs_disabled(),
			current->pid, current->comm);
I
Ingo Molnar 已提交
7969 7970 7971 7972 7973

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
L
Linus Torvalds 已提交
7974 7975 7976 7977 7978 7979
#endif
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7980 7981 7982
static void normalize_task(struct rq *rq, struct task_struct *p)
{
	int on_rq;
7983

7984 7985 7986 7987 7988 7989 7990 7991 7992 7993
	on_rq = p->se.on_rq;
	if (on_rq)
		deactivate_task(rq, p, 0);
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
		activate_task(rq, p, 0);
		resched_task(rq->curr);
	}
}

L
Linus Torvalds 已提交
7994 7995
void normalize_rt_tasks(void)
{
7996
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
7997
	unsigned long flags;
7998
	struct rq *rq;
L
Linus Torvalds 已提交
7999

8000
	read_lock_irqsave(&tasklist_lock, flags);
8001
	do_each_thread(g, p) {
8002 8003 8004 8005 8006 8007
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
8008 8009
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
8010 8011 8012
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
8013
#endif
I
Ingo Molnar 已提交
8014 8015 8016 8017 8018 8019 8020 8021

		if (!rt_task(p)) {
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
			if (TASK_NICE(p) < 0 && p->mm)
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
8022
			continue;
I
Ingo Molnar 已提交
8023
		}
L
Linus Torvalds 已提交
8024

8025
		raw_spin_lock(&p->pi_lock);
8026
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
8027

8028
		normalize_task(rq, p);
8029

8030
		__task_rq_unlock(rq);
8031
		raw_spin_unlock(&p->pi_lock);
8032 8033
	} while_each_thread(g, p);

8034
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
8035 8036 8037
}

#endif /* CONFIG_MAGIC_SYSRQ */
8038

8039
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
8040
/*
8041
 * These functions are only useful for the IA64 MCA handling, or kdb.
8042 8043 8044 8045 8046 8047 8048 8049 8050 8051 8052 8053 8054 8055
 *
 * 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!
 */
8056
struct task_struct *curr_task(int cpu)
8057 8058 8059 8060
{
	return cpu_curr(cpu);
}

8061 8062 8063
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
8064 8065 8066 8067 8068 8069
/**
 * set_curr_task - set the current task for a given cpu.
 * @cpu: the processor in question.
 * @p: the task pointer to set.
 *
 * Description: This function must only be used when non-maskable interrupts
I
Ingo Molnar 已提交
8070 8071
 * 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
8072 8073 8074 8075 8076 8077 8078
 * 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!
 */
8079
void set_curr_task(int cpu, struct task_struct *p)
8080 8081 8082 8083 8084
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
8085

8086 8087
#ifdef CONFIG_FAIR_GROUP_SCHED
static void free_fair_sched_group(struct task_group *tg)
P
Peter Zijlstra 已提交
8088 8089 8090 8091 8092 8093 8094 8095 8096 8097 8098 8099 8100 8101
{
	int i;

	for_each_possible_cpu(i) {
		if (tg->cfs_rq)
			kfree(tg->cfs_rq[i]);
		if (tg->se)
			kfree(tg->se[i]);
	}

	kfree(tg->cfs_rq);
	kfree(tg->se);
}

8102 8103
static
int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
S
Srivatsa Vaddagiri 已提交
8104 8105
{
	struct cfs_rq *cfs_rq;
8106
	struct sched_entity *se;
8107
	struct rq *rq;
S
Srivatsa Vaddagiri 已提交
8108 8109
	int i;

8110
	tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL);
S
Srivatsa Vaddagiri 已提交
8111 8112
	if (!tg->cfs_rq)
		goto err;
8113
	tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL);
S
Srivatsa Vaddagiri 已提交
8114 8115
	if (!tg->se)
		goto err;
8116 8117

	tg->shares = NICE_0_LOAD;
S
Srivatsa Vaddagiri 已提交
8118 8119

	for_each_possible_cpu(i) {
8120
		rq = cpu_rq(i);
S
Srivatsa Vaddagiri 已提交
8121

8122 8123
		cfs_rq = kzalloc_node(sizeof(struct cfs_rq),
				      GFP_KERNEL, cpu_to_node(i));
S
Srivatsa Vaddagiri 已提交
8124 8125 8126
		if (!cfs_rq)
			goto err;

8127 8128
		se = kzalloc_node(sizeof(struct sched_entity),
				  GFP_KERNEL, cpu_to_node(i));
S
Srivatsa Vaddagiri 已提交
8129
		if (!se)
8130
			goto err_free_rq;
S
Srivatsa Vaddagiri 已提交
8131

8132
		init_tg_cfs_entry(tg, cfs_rq, se, i, 0, parent->se[i]);
8133 8134 8135 8136
	}

	return 1;

8137 8138
 err_free_rq:
	kfree(cfs_rq);
8139 8140 8141 8142 8143 8144 8145 8146 8147 8148 8149 8150 8151 8152
 err:
	return 0;
}

static inline void register_fair_sched_group(struct task_group *tg, int cpu)
{
	list_add_rcu(&tg->cfs_rq[cpu]->leaf_cfs_rq_list,
			&cpu_rq(cpu)->leaf_cfs_rq_list);
}

static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
	list_del_rcu(&tg->cfs_rq[cpu]->leaf_cfs_rq_list);
}
8153
#else /* !CONFG_FAIR_GROUP_SCHED */
8154 8155 8156 8157
static inline void free_fair_sched_group(struct task_group *tg)
{
}

8158 8159
static inline
int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
8160 8161 8162 8163 8164 8165 8166 8167 8168 8169 8170
{
	return 1;
}

static inline void register_fair_sched_group(struct task_group *tg, int cpu)
{
}

static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
}
8171
#endif /* CONFIG_FAIR_GROUP_SCHED */
8172 8173

#ifdef CONFIG_RT_GROUP_SCHED
8174 8175 8176 8177
static void free_rt_sched_group(struct task_group *tg)
{
	int i;

8178 8179
	destroy_rt_bandwidth(&tg->rt_bandwidth);

8180 8181 8182 8183 8184 8185 8186 8187 8188 8189 8190
	for_each_possible_cpu(i) {
		if (tg->rt_rq)
			kfree(tg->rt_rq[i]);
		if (tg->rt_se)
			kfree(tg->rt_se[i]);
	}

	kfree(tg->rt_rq);
	kfree(tg->rt_se);
}

8191 8192
static
int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
8193 8194
{
	struct rt_rq *rt_rq;
8195
	struct sched_rt_entity *rt_se;
8196 8197 8198
	struct rq *rq;
	int i;

8199
	tg->rt_rq = kzalloc(sizeof(rt_rq) * nr_cpu_ids, GFP_KERNEL);
8200 8201
	if (!tg->rt_rq)
		goto err;
8202
	tg->rt_se = kzalloc(sizeof(rt_se) * nr_cpu_ids, GFP_KERNEL);
8203 8204 8205
	if (!tg->rt_se)
		goto err;

8206 8207
	init_rt_bandwidth(&tg->rt_bandwidth,
			ktime_to_ns(def_rt_bandwidth.rt_period), 0);
8208 8209 8210 8211

	for_each_possible_cpu(i) {
		rq = cpu_rq(i);

8212 8213
		rt_rq = kzalloc_node(sizeof(struct rt_rq),
				     GFP_KERNEL, cpu_to_node(i));
P
Peter Zijlstra 已提交
8214 8215
		if (!rt_rq)
			goto err;
S
Srivatsa Vaddagiri 已提交
8216

8217 8218
		rt_se = kzalloc_node(sizeof(struct sched_rt_entity),
				     GFP_KERNEL, cpu_to_node(i));
P
Peter Zijlstra 已提交
8219
		if (!rt_se)
8220
			goto err_free_rq;
S
Srivatsa Vaddagiri 已提交
8221

8222
		init_tg_rt_entry(tg, rt_rq, rt_se, i, 0, parent->rt_se[i]);
S
Srivatsa Vaddagiri 已提交
8223 8224
	}

8225 8226
	return 1;

8227 8228
 err_free_rq:
	kfree(rt_rq);
8229 8230 8231 8232 8233 8234 8235 8236 8237 8238 8239 8240 8241 8242
 err:
	return 0;
}

static inline void register_rt_sched_group(struct task_group *tg, int cpu)
{
	list_add_rcu(&tg->rt_rq[cpu]->leaf_rt_rq_list,
			&cpu_rq(cpu)->leaf_rt_rq_list);
}

static inline void unregister_rt_sched_group(struct task_group *tg, int cpu)
{
	list_del_rcu(&tg->rt_rq[cpu]->leaf_rt_rq_list);
}
8243
#else /* !CONFIG_RT_GROUP_SCHED */
8244 8245 8246 8247
static inline void free_rt_sched_group(struct task_group *tg)
{
}

8248 8249
static inline
int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
8250 8251 8252 8253 8254 8255 8256 8257 8258 8259 8260
{
	return 1;
}

static inline void register_rt_sched_group(struct task_group *tg, int cpu)
{
}

static inline void unregister_rt_sched_group(struct task_group *tg, int cpu)
{
}
8261
#endif /* CONFIG_RT_GROUP_SCHED */
8262

D
Dhaval Giani 已提交
8263
#ifdef CONFIG_CGROUP_SCHED
8264 8265 8266 8267 8268 8269 8270 8271
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
8272
struct task_group *sched_create_group(struct task_group *parent)
8273 8274 8275 8276 8277 8278 8279 8280 8281
{
	struct task_group *tg;
	unsigned long flags;
	int i;

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

8282
	if (!alloc_fair_sched_group(tg, parent))
8283 8284
		goto err;

8285
	if (!alloc_rt_sched_group(tg, parent))
8286 8287
		goto err;

8288
	spin_lock_irqsave(&task_group_lock, flags);
8289
	for_each_possible_cpu(i) {
8290 8291
		register_fair_sched_group(tg, i);
		register_rt_sched_group(tg, i);
8292
	}
P
Peter Zijlstra 已提交
8293
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
8294 8295 8296 8297 8298

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

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
8299
	list_add_rcu(&tg->siblings, &parent->children);
8300
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
8301

8302
	return tg;
S
Srivatsa Vaddagiri 已提交
8303 8304

err:
P
Peter Zijlstra 已提交
8305
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
8306 8307 8308
	return ERR_PTR(-ENOMEM);
}

8309
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
8310
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
8311 8312
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
8313
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
8314 8315
}

8316
/* Destroy runqueue etc associated with a task group */
8317
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
8318
{
8319
	unsigned long flags;
8320
	int i;
S
Srivatsa Vaddagiri 已提交
8321

8322
	spin_lock_irqsave(&task_group_lock, flags);
8323
	for_each_possible_cpu(i) {
8324 8325
		unregister_fair_sched_group(tg, i);
		unregister_rt_sched_group(tg, i);
8326
	}
P
Peter Zijlstra 已提交
8327
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
8328
	list_del_rcu(&tg->siblings);
8329
	spin_unlock_irqrestore(&task_group_lock, flags);
8330 8331

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
8332
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
8333 8334
}

8335
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
8336 8337 8338
 *	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.
8339 8340
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
8341 8342 8343 8344 8345 8346 8347
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

8348
	running = task_current(rq, tsk);
S
Srivatsa Vaddagiri 已提交
8349 8350
	on_rq = tsk->se.on_rq;

8351
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
8352
		dequeue_task(rq, tsk, 0);
8353 8354
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
8355

P
Peter Zijlstra 已提交
8356
	set_task_rq(tsk, task_cpu(tsk));
S
Srivatsa Vaddagiri 已提交
8357

P
Peter Zijlstra 已提交
8358 8359
#ifdef CONFIG_FAIR_GROUP_SCHED
	if (tsk->sched_class->moved_group)
8360
		tsk->sched_class->moved_group(tsk, on_rq);
P
Peter Zijlstra 已提交
8361 8362
#endif

8363 8364 8365
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
8366
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
8367 8368 8369

	task_rq_unlock(rq, &flags);
}
D
Dhaval Giani 已提交
8370
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
8371

8372
#ifdef CONFIG_FAIR_GROUP_SCHED
8373
static void __set_se_shares(struct sched_entity *se, unsigned long shares)
S
Srivatsa Vaddagiri 已提交
8374 8375 8376 8377 8378
{
	struct cfs_rq *cfs_rq = se->cfs_rq;
	int on_rq;

	on_rq = se->on_rq;
8379
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
8380 8381 8382
		dequeue_entity(cfs_rq, se, 0);

	se->load.weight = shares;
8383
	se->load.inv_weight = 0;
S
Srivatsa Vaddagiri 已提交
8384

8385
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
8386
		enqueue_entity(cfs_rq, se, 0);
8387
}
8388

8389 8390 8391 8392 8393 8394
static void set_se_shares(struct sched_entity *se, unsigned long shares)
{
	struct cfs_rq *cfs_rq = se->cfs_rq;
	struct rq *rq = cfs_rq->rq;
	unsigned long flags;

8395
	raw_spin_lock_irqsave(&rq->lock, flags);
8396
	__set_se_shares(se, shares);
8397
	raw_spin_unlock_irqrestore(&rq->lock, flags);
S
Srivatsa Vaddagiri 已提交
8398 8399
}

8400 8401
static DEFINE_MUTEX(shares_mutex);

8402
int sched_group_set_shares(struct task_group *tg, unsigned long shares)
S
Srivatsa Vaddagiri 已提交
8403 8404
{
	int i;
8405
	unsigned long flags;
8406

8407 8408 8409 8410 8411 8412
	/*
	 * We can't change the weight of the root cgroup.
	 */
	if (!tg->se[0])
		return -EINVAL;

8413 8414
	if (shares < MIN_SHARES)
		shares = MIN_SHARES;
8415 8416
	else if (shares > MAX_SHARES)
		shares = MAX_SHARES;
8417

8418
	mutex_lock(&shares_mutex);
8419
	if (tg->shares == shares)
8420
		goto done;
S
Srivatsa Vaddagiri 已提交
8421

8422
	spin_lock_irqsave(&task_group_lock, flags);
8423 8424
	for_each_possible_cpu(i)
		unregister_fair_sched_group(tg, i);
P
Peter Zijlstra 已提交
8425
	list_del_rcu(&tg->siblings);
8426
	spin_unlock_irqrestore(&task_group_lock, flags);
8427 8428 8429 8430 8431 8432 8433 8434

	/* wait for any ongoing reference to this group to finish */
	synchronize_sched();

	/*
	 * Now we are free to modify the group's share on each cpu
	 * w/o tripping rebalance_share or load_balance_fair.
	 */
8435
	tg->shares = shares;
8436 8437 8438 8439 8440
	for_each_possible_cpu(i) {
		/*
		 * force a rebalance
		 */
		cfs_rq_set_shares(tg->cfs_rq[i], 0);
8441
		set_se_shares(tg->se[i], shares);
8442
	}
S
Srivatsa Vaddagiri 已提交
8443

8444 8445 8446 8447
	/*
	 * Enable load balance activity on this group, by inserting it back on
	 * each cpu's rq->leaf_cfs_rq_list.
	 */
8448
	spin_lock_irqsave(&task_group_lock, flags);
8449 8450
	for_each_possible_cpu(i)
		register_fair_sched_group(tg, i);
P
Peter Zijlstra 已提交
8451
	list_add_rcu(&tg->siblings, &tg->parent->children);
8452
	spin_unlock_irqrestore(&task_group_lock, flags);
8453
done:
8454
	mutex_unlock(&shares_mutex);
8455
	return 0;
S
Srivatsa Vaddagiri 已提交
8456 8457
}

8458 8459 8460 8461
unsigned long sched_group_shares(struct task_group *tg)
{
	return tg->shares;
}
8462
#endif
8463

8464
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8465
/*
P
Peter Zijlstra 已提交
8466
 * Ensure that the real time constraints are schedulable.
P
Peter Zijlstra 已提交
8467
 */
P
Peter Zijlstra 已提交
8468 8469 8470 8471 8472
static DEFINE_MUTEX(rt_constraints_mutex);

static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
8473
		return 1ULL << 20;
P
Peter Zijlstra 已提交
8474

P
Peter Zijlstra 已提交
8475
	return div64_u64(runtime << 20, period);
P
Peter Zijlstra 已提交
8476 8477
}

P
Peter Zijlstra 已提交
8478 8479
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
8480
{
P
Peter Zijlstra 已提交
8481
	struct task_struct *g, *p;
8482

P
Peter Zijlstra 已提交
8483 8484 8485 8486
	do_each_thread(g, p) {
		if (rt_task(p) && rt_rq_of_se(&p->rt)->tg == tg)
			return 1;
	} while_each_thread(g, p);
8487

P
Peter Zijlstra 已提交
8488 8489
	return 0;
}
8490

P
Peter Zijlstra 已提交
8491 8492 8493 8494 8495
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
8496

P
Peter Zijlstra 已提交
8497 8498 8499 8500 8501 8502
static int tg_schedulable(struct task_group *tg, void *data)
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
8503

P
Peter Zijlstra 已提交
8504 8505
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
8506

P
Peter Zijlstra 已提交
8507 8508 8509
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
8510 8511
	}

8512 8513 8514 8515 8516
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
8517

8518 8519 8520
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
8521 8522
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
8523

P
Peter Zijlstra 已提交
8524
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
8525

8526 8527 8528 8529 8530
	/*
	 * Nobody can have more than the global setting allows.
	 */
	if (total > to_ratio(global_rt_period(), global_rt_runtime()))
		return -EINVAL;
P
Peter Zijlstra 已提交
8531

8532 8533 8534
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
8535 8536 8537
	list_for_each_entry_rcu(child, &tg->children, siblings) {
		period = ktime_to_ns(child->rt_bandwidth.rt_period);
		runtime = child->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
8538

P
Peter Zijlstra 已提交
8539 8540 8541 8542
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
8543

P
Peter Zijlstra 已提交
8544
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
8545
	}
P
Peter Zijlstra 已提交
8546

P
Peter Zijlstra 已提交
8547 8548 8549 8550
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
8551 8552
}

P
Peter Zijlstra 已提交
8553
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
8554
{
P
Peter Zijlstra 已提交
8555 8556 8557 8558 8559 8560 8561
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

	return walk_tg_tree(tg_schedulable, tg_nop, &data);
8562 8563
}

8564 8565
static int tg_set_bandwidth(struct task_group *tg,
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
8566
{
P
Peter Zijlstra 已提交
8567
	int i, err = 0;
P
Peter Zijlstra 已提交
8568 8569

	mutex_lock(&rt_constraints_mutex);
8570
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
8571 8572
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
8573
		goto unlock;
P
Peter Zijlstra 已提交
8574

8575
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
8576 8577
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
8578 8579 8580 8581

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

8582
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8583
		rt_rq->rt_runtime = rt_runtime;
8584
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8585
	}
8586
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
8587
 unlock:
8588
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
8589 8590 8591
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
8592 8593
}

8594 8595 8596 8597 8598 8599 8600 8601 8602 8603 8604 8605
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;

	return tg_set_bandwidth(tg, rt_period, rt_runtime);
}

P
Peter Zijlstra 已提交
8606 8607 8608 8609
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

8610
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
8611 8612
		return -1;

8613
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
8614 8615 8616
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
8617 8618 8619 8620 8621 8622 8623 8624

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;

8625 8626 8627
	if (rt_period == 0)
		return -EINVAL;

8628 8629 8630 8631 8632 8633 8634 8635 8636 8637 8638 8639 8640 8641
	return tg_set_bandwidth(tg, rt_period, rt_runtime);
}

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)
{
8642
	u64 runtime, period;
8643 8644
	int ret = 0;

8645 8646 8647
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

8648 8649 8650 8651 8652 8653 8654 8655
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
8656

8657
	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
8658
	read_lock(&tasklist_lock);
8659
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
8660
	read_unlock(&tasklist_lock);
8661 8662 8663 8664
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
8665 8666 8667 8668 8669 8670 8671 8672 8673 8674

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;
}

8675
#else /* !CONFIG_RT_GROUP_SCHED */
8676 8677
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
8678 8679 8680
	unsigned long flags;
	int i;

8681 8682 8683
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

8684 8685 8686 8687 8688 8689 8690
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

8691
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
8692 8693 8694
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

8695
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8696
		rt_rq->rt_runtime = global_rt_runtime();
8697
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8698
	}
8699
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
8700

8701 8702
	return 0;
}
8703
#endif /* CONFIG_RT_GROUP_SCHED */
8704 8705

int sched_rt_handler(struct ctl_table *table, int write,
8706
		void __user *buffer, size_t *lenp,
8707 8708 8709 8710 8711 8712 8713 8714 8715 8716
		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;

8717
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
8718 8719 8720 8721 8722 8723 8724 8725 8726 8727 8728 8729 8730 8731 8732 8733

	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;
}
8734

8735
#ifdef CONFIG_CGROUP_SCHED
8736 8737

/* return corresponding task_group object of a cgroup */
8738
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
8739
{
8740 8741
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
8742 8743 8744
}

static struct cgroup_subsys_state *
8745
cpu_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
8746
{
8747
	struct task_group *tg, *parent;
8748

8749
	if (!cgrp->parent) {
8750 8751 8752 8753
		/* This is early initialization for the top cgroup */
		return &init_task_group.css;
	}

8754 8755
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
8756 8757 8758 8759 8760 8761
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

I
Ingo Molnar 已提交
8762 8763
static void
cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
8764
{
8765
	struct task_group *tg = cgroup_tg(cgrp);
8766 8767 8768 8769

	sched_destroy_group(tg);
}

I
Ingo Molnar 已提交
8770
static int
8771
cpu_cgroup_can_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
8772
{
8773
#ifdef CONFIG_RT_GROUP_SCHED
8774
	if (!sched_rt_can_attach(cgroup_tg(cgrp), tsk))
8775 8776
		return -EINVAL;
#else
8777 8778 8779
	/* We don't support RT-tasks being in separate groups */
	if (tsk->sched_class != &fair_sched_class)
		return -EINVAL;
8780
#endif
8781 8782
	return 0;
}
8783

8784 8785 8786 8787 8788 8789 8790 8791 8792 8793 8794 8795 8796 8797 8798 8799 8800 8801 8802
static int
cpu_cgroup_can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
		      struct task_struct *tsk, bool threadgroup)
{
	int retval = cpu_cgroup_can_attach_task(cgrp, tsk);
	if (retval)
		return retval;
	if (threadgroup) {
		struct task_struct *c;
		rcu_read_lock();
		list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
			retval = cpu_cgroup_can_attach_task(cgrp, c);
			if (retval) {
				rcu_read_unlock();
				return retval;
			}
		}
		rcu_read_unlock();
	}
8803 8804 8805 8806
	return 0;
}

static void
8807
cpu_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
8808 8809
		  struct cgroup *old_cont, struct task_struct *tsk,
		  bool threadgroup)
8810 8811
{
	sched_move_task(tsk);
8812 8813 8814 8815 8816 8817 8818 8819
	if (threadgroup) {
		struct task_struct *c;
		rcu_read_lock();
		list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
			sched_move_task(c);
		}
		rcu_read_unlock();
	}
8820 8821
}

8822
#ifdef CONFIG_FAIR_GROUP_SCHED
8823
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
8824
				u64 shareval)
8825
{
8826
	return sched_group_set_shares(cgroup_tg(cgrp), shareval);
8827 8828
}

8829
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
8830
{
8831
	struct task_group *tg = cgroup_tg(cgrp);
8832 8833 8834

	return (u64) tg->shares;
}
8835
#endif /* CONFIG_FAIR_GROUP_SCHED */
8836

8837
#ifdef CONFIG_RT_GROUP_SCHED
M
Mirco Tischler 已提交
8838
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
8839
				s64 val)
P
Peter Zijlstra 已提交
8840
{
8841
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
P
Peter Zijlstra 已提交
8842 8843
}

8844
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
P
Peter Zijlstra 已提交
8845
{
8846
	return sched_group_rt_runtime(cgroup_tg(cgrp));
P
Peter Zijlstra 已提交
8847
}
8848 8849 8850 8851 8852 8853 8854 8855 8856 8857 8858

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));
}
8859
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
8860

8861
static struct cftype cpu_files[] = {
8862
#ifdef CONFIG_FAIR_GROUP_SCHED
8863 8864
	{
		.name = "shares",
8865 8866
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
8867
	},
8868 8869
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8870
	{
P
Peter Zijlstra 已提交
8871
		.name = "rt_runtime_us",
8872 8873
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
8874
	},
8875 8876
	{
		.name = "rt_period_us",
8877 8878
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
8879
	},
8880
#endif
8881 8882 8883 8884
};

static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
{
8885
	return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
8886 8887 8888
}

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
8889 8890 8891 8892 8893 8894 8895
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
	.populate	= cpu_cgroup_populate,
	.subsys_id	= cpu_cgroup_subsys_id,
8896 8897 8898
	.early_init	= 1,
};

8899
#endif	/* CONFIG_CGROUP_SCHED */
8900 8901 8902 8903 8904 8905 8906 8907 8908 8909

#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).
 */

8910
/* track cpu usage of a group of tasks and its child groups */
8911 8912 8913
struct cpuacct {
	struct cgroup_subsys_state css;
	/* cpuusage holds pointer to a u64-type object on every cpu */
8914
	u64 __percpu *cpuusage;
8915
	struct percpu_counter cpustat[CPUACCT_STAT_NSTATS];
8916
	struct cpuacct *parent;
8917 8918 8919 8920 8921
};

struct cgroup_subsys cpuacct_subsys;

/* return cpu accounting group corresponding to this container */
8922
static inline struct cpuacct *cgroup_ca(struct cgroup *cgrp)
8923
{
8924
	return container_of(cgroup_subsys_state(cgrp, cpuacct_subsys_id),
8925 8926 8927 8928 8929 8930 8931 8932 8933 8934 8935 8936
			    struct cpuacct, css);
}

/* return cpu accounting group to which this task belongs */
static inline struct cpuacct *task_ca(struct task_struct *tsk)
{
	return container_of(task_subsys_state(tsk, cpuacct_subsys_id),
			    struct cpuacct, css);
}

/* create a new cpu accounting group */
static struct cgroup_subsys_state *cpuacct_create(
8937
	struct cgroup_subsys *ss, struct cgroup *cgrp)
8938 8939
{
	struct cpuacct *ca = kzalloc(sizeof(*ca), GFP_KERNEL);
8940
	int i;
8941 8942

	if (!ca)
8943
		goto out;
8944 8945

	ca->cpuusage = alloc_percpu(u64);
8946 8947 8948 8949 8950 8951
	if (!ca->cpuusage)
		goto out_free_ca;

	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
		if (percpu_counter_init(&ca->cpustat[i], 0))
			goto out_free_counters;
8952

8953 8954 8955
	if (cgrp->parent)
		ca->parent = cgroup_ca(cgrp->parent);

8956
	return &ca->css;
8957 8958 8959 8960 8961 8962 8963 8964 8965

out_free_counters:
	while (--i >= 0)
		percpu_counter_destroy(&ca->cpustat[i]);
	free_percpu(ca->cpuusage);
out_free_ca:
	kfree(ca);
out:
	return ERR_PTR(-ENOMEM);
8966 8967 8968
}

/* destroy an existing cpu accounting group */
I
Ingo Molnar 已提交
8969
static void
8970
cpuacct_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
8971
{
8972
	struct cpuacct *ca = cgroup_ca(cgrp);
8973
	int i;
8974

8975 8976
	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
		percpu_counter_destroy(&ca->cpustat[i]);
8977 8978 8979 8980
	free_percpu(ca->cpuusage);
	kfree(ca);
}

8981 8982
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
8983
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8984 8985 8986 8987 8988 8989
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
8990
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8991
	data = *cpuusage;
8992
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8993 8994 8995 8996 8997 8998 8999 9000 9001
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
9002
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
9003 9004 9005 9006 9007

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
9008
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
9009
	*cpuusage = val;
9010
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
9011 9012 9013 9014 9015
#else
	*cpuusage = val;
#endif
}

9016
/* return total cpu usage (in nanoseconds) of a group */
9017
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
9018
{
9019
	struct cpuacct *ca = cgroup_ca(cgrp);
9020 9021 9022
	u64 totalcpuusage = 0;
	int i;

9023 9024
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
9025 9026 9027 9028

	return totalcpuusage;
}

9029 9030 9031 9032 9033 9034 9035 9036 9037 9038 9039 9040
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;
	}

9041 9042
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
9043 9044 9045 9046 9047

out:
	return err;
}

9048 9049 9050 9051 9052 9053 9054 9055 9056 9057 9058 9059 9060 9061 9062
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;
}

9063 9064 9065 9066 9067 9068 9069 9070 9071 9072 9073 9074 9075 9076 9077 9078 9079 9080 9081
static const char *cpuacct_stat_desc[] = {
	[CPUACCT_STAT_USER] = "user",
	[CPUACCT_STAT_SYSTEM] = "system",
};

static int cpuacct_stats_show(struct cgroup *cgrp, struct cftype *cft,
		struct cgroup_map_cb *cb)
{
	struct cpuacct *ca = cgroup_ca(cgrp);
	int i;

	for (i = 0; i < CPUACCT_STAT_NSTATS; i++) {
		s64 val = percpu_counter_read(&ca->cpustat[i]);
		val = cputime64_to_clock_t(val);
		cb->fill(cb, cpuacct_stat_desc[i], val);
	}
	return 0;
}

9082 9083 9084
static struct cftype files[] = {
	{
		.name = "usage",
9085 9086
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
9087
	},
9088 9089 9090 9091
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
9092 9093 9094 9095
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
9096 9097
};

9098
static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
9099
{
9100
	return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
9101 9102 9103 9104 9105 9106 9107 9108 9109 9110
}

/*
 * charge this task's execution time to its accounting group.
 *
 * called with rq->lock held.
 */
static void cpuacct_charge(struct task_struct *tsk, u64 cputime)
{
	struct cpuacct *ca;
9111
	int cpu;
9112

L
Li Zefan 已提交
9113
	if (unlikely(!cpuacct_subsys.active))
9114 9115
		return;

9116
	cpu = task_cpu(tsk);
9117 9118 9119

	rcu_read_lock();

9120 9121
	ca = task_ca(tsk);

9122
	for (; ca; ca = ca->parent) {
9123
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
9124 9125
		*cpuusage += cputime;
	}
9126 9127

	rcu_read_unlock();
9128 9129
}

9130 9131 9132 9133 9134 9135 9136 9137 9138 9139 9140 9141 9142 9143 9144 9145 9146
/*
 * When CONFIG_VIRT_CPU_ACCOUNTING is enabled one jiffy can be very large
 * in cputime_t units. As a result, cpuacct_update_stats calls
 * percpu_counter_add with values large enough to always overflow the
 * per cpu batch limit causing bad SMP scalability.
 *
 * To fix this we scale percpu_counter_batch by cputime_one_jiffy so we
 * batch the same amount of time with CONFIG_VIRT_CPU_ACCOUNTING disabled
 * and enabled. We cap it at INT_MAX which is the largest allowed batch value.
 */
#ifdef CONFIG_SMP
#define CPUACCT_BATCH	\
	min_t(long, percpu_counter_batch * cputime_one_jiffy, INT_MAX)
#else
#define CPUACCT_BATCH	0
#endif

9147 9148 9149 9150 9151 9152 9153
/*
 * Charge the system/user time to the task's accounting group.
 */
static void cpuacct_update_stats(struct task_struct *tsk,
		enum cpuacct_stat_index idx, cputime_t val)
{
	struct cpuacct *ca;
9154
	int batch = CPUACCT_BATCH;
9155 9156 9157 9158 9159 9160 9161 9162

	if (unlikely(!cpuacct_subsys.active))
		return;

	rcu_read_lock();
	ca = task_ca(tsk);

	do {
9163
		__percpu_counter_add(&ca->cpustat[idx], val, batch);
9164 9165 9166 9167 9168
		ca = ca->parent;
	} while (ca);
	rcu_read_unlock();
}

9169 9170 9171 9172 9173 9174 9175 9176
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 */
9177 9178 9179 9180 9181

#ifndef CONFIG_SMP

void synchronize_sched_expedited(void)
{
9182
	barrier();
9183 9184 9185 9186 9187
}
EXPORT_SYMBOL_GPL(synchronize_sched_expedited);

#else /* #ifndef CONFIG_SMP */

9188
static atomic_t synchronize_sched_expedited_count = ATOMIC_INIT(0);
9189

9190
static int synchronize_sched_expedited_cpu_stop(void *data)
9191
{
9192 9193 9194 9195 9196 9197 9198 9199 9200 9201 9202
	/*
	 * There must be a full memory barrier on each affected CPU
	 * between the time that try_stop_cpus() is called and the
	 * time that it returns.
	 *
	 * In the current initial implementation of cpu_stop, the
	 * above condition is already met when the control reaches
	 * this point and the following smp_mb() is not strictly
	 * necessary.  Do smp_mb() anyway for documentation and
	 * robustness against future implementation changes.
	 */
9203
	smp_mb(); /* See above comment block. */
9204
	return 0;
9205 9206 9207 9208 9209 9210 9211 9212 9213 9214 9215 9216 9217 9218
}

/*
 * Wait for an rcu-sched grace period to elapse, but use "big hammer"
 * approach to force grace period to end quickly.  This consumes
 * significant time on all CPUs, and is thus not recommended for
 * any sort of common-case code.
 *
 * Note that it is illegal to call this function while holding any
 * lock that is acquired by a CPU-hotplug notifier.  Failing to
 * observe this restriction will result in deadlock.
 */
void synchronize_sched_expedited(void)
{
9219
	int snap, trycount = 0;
9220 9221

	smp_mb();  /* ensure prior mod happens before capturing snap. */
9222
	snap = atomic_read(&synchronize_sched_expedited_count) + 1;
9223
	get_online_cpus();
9224 9225
	while (try_stop_cpus(cpu_online_mask,
			     synchronize_sched_expedited_cpu_stop,
9226
			     NULL) == -EAGAIN) {
9227 9228 9229 9230 9231 9232 9233
		put_online_cpus();
		if (trycount++ < 10)
			udelay(trycount * num_online_cpus());
		else {
			synchronize_sched();
			return;
		}
9234
		if (atomic_read(&synchronize_sched_expedited_count) - snap > 0) {
9235 9236 9237 9238 9239
			smp_mb(); /* ensure test happens before caller kfree */
			return;
		}
		get_online_cpus();
	}
9240
	atomic_inc(&synchronize_sched_expedited_count);
9241
	smp_mb__after_atomic_inc(); /* ensure post-GP actions seen after GP. */
9242 9243 9244 9245 9246
	put_online_cpus();
}
EXPORT_SYMBOL_GPL(synchronize_sched_expedited);

#endif /* #else #ifndef CONFIG_SMP */