sched.c 228.1 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
#include <linux/highmem.h>
#include <asm/mmu_context.h>
#include <linux/interrupt.h>
37
#include <linux/capability.h>
L
Linus Torvalds 已提交
38 39
#include <linux/completion.h>
#include <linux/kernel_stat.h>
40
#include <linux/debug_locks.h>
41
#include <linux/perf_event.h>
L
Linus Torvalds 已提交
42 43 44
#include <linux/security.h>
#include <linux/notifier.h>
#include <linux/profile.h>
45
#include <linux/freezer.h>
46
#include <linux/vmalloc.h>
L
Linus Torvalds 已提交
47 48
#include <linux/blkdev.h>
#include <linux/delay.h>
49
#include <linux/pid_namespace.h>
L
Linus Torvalds 已提交
50 51 52 53 54 55 56
#include <linux/smp.h>
#include <linux/threads.h>
#include <linux/timer.h>
#include <linux/rcupdate.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/percpu.h>
57
#include <linux/proc_fs.h>
L
Linus Torvalds 已提交
58
#include <linux/seq_file.h>
59
#include <linux/stop_machine.h>
60
#include <linux/sysctl.h>
L
Linus Torvalds 已提交
61 62
#include <linux/syscalls.h>
#include <linux/times.h>
63
#include <linux/tsacct_kern.h>
64
#include <linux/kprobes.h>
65
#include <linux/delayacct.h>
66
#include <linux/unistd.h>
J
Jens Axboe 已提交
67
#include <linux/pagemap.h>
P
Peter Zijlstra 已提交
68
#include <linux/hrtimer.h>
R
Reynes Philippe 已提交
69
#include <linux/tick.h>
P
Peter Zijlstra 已提交
70 71
#include <linux/debugfs.h>
#include <linux/ctype.h>
72
#include <linux/ftrace.h>
73
#include <linux/slab.h>
L
Linus Torvalds 已提交
74

75
#include <asm/tlb.h>
76
#include <asm/irq_regs.h>
77
#include <asm/mutex.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
#include "sched_autogroup.h"
82

83
#define CREATE_TRACE_POINTS
84
#include <trace/events/sched.h>
85

L
Linus Torvalds 已提交
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104
/*
 * 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))

/*
105
 * Helpers for converting nanosecond timing to jiffy resolution
L
Linus Torvalds 已提交
106
 */
107
#define NS_TO_JIFFIES(TIME)	((unsigned long)(TIME) / (NSEC_PER_SEC / HZ))
L
Linus Torvalds 已提交
108

I
Ingo Molnar 已提交
109 110 111
#define NICE_0_LOAD		SCHED_LOAD_SCALE
#define NICE_0_SHIFT		SCHED_LOAD_SHIFT

L
Linus Torvalds 已提交
112 113 114
/*
 * These are the 'tuning knobs' of the scheduler:
 *
D
Dmitry Adamushko 已提交
115
 * default timeslice is 100 msecs (used only for SCHED_RR tasks).
L
Linus Torvalds 已提交
116 117 118
 * Timeslices get refilled after they expire.
 */
#define DEF_TIMESLICE		(100 * HZ / 1000)
119

120 121 122 123 124
/*
 * single value that denotes runtime == period, ie unlimited time.
 */
#define RUNTIME_INF	((u64)~0ULL)

125 126
static inline int rt_policy(int policy)
{
127
	if (unlikely(policy == SCHED_FIFO || policy == SCHED_RR))
128 129 130 131 132 133 134 135 136
		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
137
/*
I
Ingo Molnar 已提交
138
 * This is the priority-queue data structure of the RT scheduling class:
L
Linus Torvalds 已提交
139
 */
I
Ingo Molnar 已提交
140 141 142 143 144
struct rt_prio_array {
	DECLARE_BITMAP(bitmap, MAX_RT_PRIO+1); /* include 1 bit for delimiter */
	struct list_head queue[MAX_RT_PRIO];
};

145
struct rt_bandwidth {
I
Ingo Molnar 已提交
146
	/* nests inside the rq lock: */
147
	raw_spinlock_t		rt_runtime_lock;
I
Ingo Molnar 已提交
148 149 150
	ktime_t			rt_period;
	u64			rt_runtime;
	struct hrtimer		rt_period_timer;
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
};

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;

184
	raw_spin_lock_init(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
185

186 187 188 189 190
	hrtimer_init(&rt_b->rt_period_timer,
			CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rt_b->rt_period_timer.function = sched_rt_period_timer;
}

191 192 193
static inline int rt_bandwidth_enabled(void)
{
	return sysctl_sched_rt_runtime >= 0;
194 195 196 197 198 199
}

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

200
	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
201 202 203 204 205
		return;

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

206
	raw_spin_lock(&rt_b->rt_runtime_lock);
207
	for (;;) {
208 209 210
		unsigned long delta;
		ktime_t soft, hard;

211 212 213 214 215
		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);
216 217 218 219 220

		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,
221
				HRTIMER_MODE_ABS_PINNED, 0);
222
	}
223
	raw_spin_unlock(&rt_b->rt_runtime_lock);
224 225 226 227 228 229 230 231 232
}

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

233 234 235 236 237 238
/*
 * 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
239
#ifdef CONFIG_CGROUP_SCHED
S
Srivatsa Vaddagiri 已提交
240

241 242
#include <linux/cgroup.h>

S
Srivatsa Vaddagiri 已提交
243 244
struct cfs_rq;

P
Peter Zijlstra 已提交
245 246
static LIST_HEAD(task_groups);

S
Srivatsa Vaddagiri 已提交
247
/* task group related information */
248
struct task_group {
249
	struct cgroup_subsys_state css;
250

251
#ifdef CONFIG_FAIR_GROUP_SCHED
S
Srivatsa Vaddagiri 已提交
252 253 254 255 256
	/* 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;
P
Peter Zijlstra 已提交
257 258

	atomic_t load_weight;
259 260 261 262 263 264
#endif

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

265
	struct rt_bandwidth rt_bandwidth;
266
#endif
267

268
	struct rcu_head rcu;
P
Peter Zijlstra 已提交
269
	struct list_head list;
P
Peter Zijlstra 已提交
270 271 272 273

	struct task_group *parent;
	struct list_head siblings;
	struct list_head children;
274 275 276 277

#ifdef CONFIG_SCHED_AUTOGROUP
	struct autogroup *autogroup;
#endif
S
Srivatsa Vaddagiri 已提交
278 279
};

280
/* task_group_lock serializes the addition/removal of task groups */
281
static DEFINE_SPINLOCK(task_group_lock);
282

283 284
#ifdef CONFIG_FAIR_GROUP_SCHED

285
# define ROOT_TASK_GROUP_LOAD	NICE_0_LOAD
286

287
/*
288 289 290 291
 * 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.
292 293 294
 * (The default weight is 1024 - so there's no practical
 *  limitation from this.)
 */
295
#define MIN_SHARES	2
296
#define MAX_SHARES	(1UL << 18)
297

298
static int root_task_group_load = ROOT_TASK_GROUP_LOAD;
299 300
#endif

S
Srivatsa Vaddagiri 已提交
301
/* Default task group.
I
Ingo Molnar 已提交
302
 *	Every task in system belong to this group at bootup.
S
Srivatsa Vaddagiri 已提交
303
 */
304
struct task_group root_task_group;
S
Srivatsa Vaddagiri 已提交
305

D
Dhaval Giani 已提交
306
#endif	/* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
307

I
Ingo Molnar 已提交
308 309 310 311 312 313
/* 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
314
	u64 min_vruntime;
315 316 317
#ifndef CONFIG_64BIT
	u64 min_vruntime_copy;
#endif
I
Ingo Molnar 已提交
318 319 320

	struct rb_root tasks_timeline;
	struct rb_node *rb_leftmost;
321 322 323 324 325 326

	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
327 328
	 * It is set to NULL otherwise (i.e when none are currently running).
	 */
329
	struct sched_entity *curr, *next, *last, *skip;
P
Peter Zijlstra 已提交
330

P
Peter Zijlstra 已提交
331
	unsigned int nr_spread_over;
P
Peter Zijlstra 已提交
332

333
#ifdef CONFIG_FAIR_GROUP_SCHED
I
Ingo Molnar 已提交
334 335
	struct rq *rq;	/* cpu runqueue to which this cfs_rq is attached */

I
Ingo Molnar 已提交
336 337
	/*
	 * leaf cfs_rqs are those that hold tasks (lowest schedulable entity in
I
Ingo Molnar 已提交
338 339 340 341 342 343
	 * 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.
	 */
344
	int on_list;
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
	/*
363 364 365 366 367
	 * Maintaining per-cpu shares distribution for group scheduling
	 *
	 * load_stamp is the last time we updated the load average
	 * load_last is the last time we updated the load average and saw load
	 * load_unacc_exec_time is currently unaccounted execution time
368
	 */
P
Peter Zijlstra 已提交
369 370
	u64 load_avg;
	u64 load_period;
371
	u64 load_stamp, load_last, load_unacc_exec_time;
372

P
Peter Zijlstra 已提交
373
	unsigned long load_contribution;
374
#endif
I
Ingo Molnar 已提交
375 376
#endif
};
L
Linus Torvalds 已提交
377

I
Ingo Molnar 已提交
378 379 380
/* Real-Time classes' related field in a runqueue: */
struct rt_rq {
	struct rt_prio_array active;
381
	unsigned long rt_nr_running;
382
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
383 384
	struct {
		int curr; /* highest queued rt task prio */
385
#ifdef CONFIG_SMP
386
		int next; /* next highest */
387
#endif
388
	} highest_prio;
P
Peter Zijlstra 已提交
389
#endif
P
Peter Zijlstra 已提交
390
#ifdef CONFIG_SMP
391
	unsigned long rt_nr_migratory;
392
	unsigned long rt_nr_total;
G
Gregory Haskins 已提交
393
	int overloaded;
394
	struct plist_head pushable_tasks;
P
Peter Zijlstra 已提交
395
#endif
P
Peter Zijlstra 已提交
396
	int rt_throttled;
P
Peter Zijlstra 已提交
397
	u64 rt_time;
P
Peter Zijlstra 已提交
398
	u64 rt_runtime;
I
Ingo Molnar 已提交
399
	/* Nests inside the rq lock: */
400
	raw_spinlock_t rt_runtime_lock;
P
Peter Zijlstra 已提交
401

402
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
403 404
	unsigned long rt_nr_boosted;

P
Peter Zijlstra 已提交
405 406 407 408
	struct rq *rq;
	struct list_head leaf_rt_rq_list;
	struct task_group *tg;
#endif
I
Ingo Molnar 已提交
409 410
};

G
Gregory Haskins 已提交
411 412 413 414
#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
415 416
 * 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
417 418 419 420 421 422
 * exclusive cpuset is created, we also create and attach a new root-domain
 * object.
 *
 */
struct root_domain {
	atomic_t refcount;
423 424
	cpumask_var_t span;
	cpumask_var_t online;
425

I
Ingo Molnar 已提交
426
	/*
427 428 429
	 * The "RT overload" flag: it gets set if a CPU has more than
	 * one runnable RT task.
	 */
430
	cpumask_var_t rto_mask;
I
Ingo Molnar 已提交
431
	atomic_t rto_count;
432
	struct cpupri cpupri;
G
Gregory Haskins 已提交
433 434
};

435 436 437 438
/*
 * 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
439 440
static struct root_domain def_root_domain;

441
#endif /* CONFIG_SMP */
G
Gregory Haskins 已提交
442

L
Linus Torvalds 已提交
443 444 445 446 447 448 449
/*
 * 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.
 */
450
struct rq {
451
	/* runqueue lock: */
452
	raw_spinlock_t lock;
L
Linus Torvalds 已提交
453 454 455 456 457 458

	/*
	 * 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
459 460
	#define CPU_LOAD_IDX_MAX 5
	unsigned long cpu_load[CPU_LOAD_IDX_MAX];
461
	unsigned long last_load_update_tick;
462
#ifdef CONFIG_NO_HZ
M
Mike Galbraith 已提交
463
	u64 nohz_stamp;
464
	unsigned char nohz_balance_kick;
465
#endif
466 467
	unsigned int skip_clock_update;

468 469
	/* capture load from *all* tasks on this cpu: */
	struct load_weight load;
I
Ingo Molnar 已提交
470 471 472 473
	unsigned long nr_load_updates;
	u64 nr_switches;

	struct cfs_rq cfs;
P
Peter Zijlstra 已提交
474 475
	struct rt_rq rt;

I
Ingo Molnar 已提交
476
#ifdef CONFIG_FAIR_GROUP_SCHED
477 478
	/* list of leaf cfs_rq on this cpu: */
	struct list_head leaf_cfs_rq_list;
479 480
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
481
	struct list_head leaf_rt_rq_list;
L
Linus Torvalds 已提交
482 483 484 485 486 487 488 489 490 491
#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;

492
	struct task_struct *curr, *idle, *stop;
493
	unsigned long next_balance;
L
Linus Torvalds 已提交
494
	struct mm_struct *prev_mm;
I
Ingo Molnar 已提交
495

496
	u64 clock;
497
	u64 clock_task;
I
Ingo Molnar 已提交
498

L
Linus Torvalds 已提交
499 500 501
	atomic_t nr_iowait;

#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
502
	struct root_domain *rd;
L
Linus Torvalds 已提交
503 504
	struct sched_domain *sd;

505 506
	unsigned long cpu_power;

507
	unsigned char idle_at_tick;
L
Linus Torvalds 已提交
508
	/* For active balancing */
509
	int post_schedule;
L
Linus Torvalds 已提交
510 511
	int active_balance;
	int push_cpu;
512
	struct cpu_stop_work active_balance_work;
513 514
	/* cpu of this runqueue: */
	int cpu;
515
	int online;
L
Linus Torvalds 已提交
516

517
	unsigned long avg_load_per_task;
L
Linus Torvalds 已提交
518

519 520
	u64 rt_avg;
	u64 age_stamp;
M
Mike Galbraith 已提交
521 522
	u64 idle_stamp;
	u64 avg_idle;
L
Linus Torvalds 已提交
523 524
#endif

525 526 527 528
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
	u64 prev_irq_time;
#endif

529 530 531 532
	/* calc_load related fields */
	unsigned long calc_load_update;
	long calc_load_active;

P
Peter Zijlstra 已提交
533
#ifdef CONFIG_SCHED_HRTICK
534 535 536 537
#ifdef CONFIG_SMP
	int hrtick_csd_pending;
	struct call_single_data hrtick_csd;
#endif
P
Peter Zijlstra 已提交
538 539 540
	struct hrtimer hrtick_timer;
#endif

L
Linus Torvalds 已提交
541 542 543
#ifdef CONFIG_SCHEDSTATS
	/* latency stats */
	struct sched_info rq_sched_info;
544 545
	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
546 547

	/* sys_sched_yield() stats */
548
	unsigned int yld_count;
L
Linus Torvalds 已提交
549 550

	/* schedule() stats */
551 552 553
	unsigned int sched_switch;
	unsigned int sched_count;
	unsigned int sched_goidle;
L
Linus Torvalds 已提交
554 555

	/* try_to_wake_up() stats */
556 557
	unsigned int ttwu_count;
	unsigned int ttwu_local;
L
Linus Torvalds 已提交
558 559 560
#endif
};

561
static DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
L
Linus Torvalds 已提交
562

563

564
static void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags);
I
Ingo Molnar 已提交
565

566 567 568 569 570 571 572 573 574
static inline int cpu_of(struct rq *rq)
{
#ifdef CONFIG_SMP
	return rq->cpu;
#else
	return 0;
#endif
}

575
#define rcu_dereference_check_sched_domain(p) \
576 577 578 579
	rcu_dereference_check((p), \
			      rcu_read_lock_sched_held() || \
			      lockdep_is_held(&sched_domains_mutex))

N
Nick Piggin 已提交
580 581
/*
 * The domain tree (rq->sd) is protected by RCU's quiescent state transition.
582
 * See detach_destroy_domains: synchronize_sched for details.
N
Nick Piggin 已提交
583 584 585 586
 *
 * The domain tree of any CPU may only be accessed from within
 * preempt-disabled sections.
 */
587
#define for_each_domain(cpu, __sd) \
588
	for (__sd = rcu_dereference_check_sched_domain(cpu_rq(cpu)->sd); __sd; __sd = __sd->parent)
L
Linus Torvalds 已提交
589 590 591 592 593

#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)
594
#define raw_rq()		(&__raw_get_cpu_var(runqueues))
L
Linus Torvalds 已提交
595

596 597 598 599 600 601
#ifdef CONFIG_CGROUP_SCHED

/*
 * Return the group to which this tasks belongs.
 *
 * We use task_subsys_state_check() and extend the RCU verification
602
 * with lockdep_is_held(&p->pi_lock) because cpu_cgroup_attach()
603 604 605 606 607
 * 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)
{
608
	struct task_group *tg;
609 610 611
	struct cgroup_subsys_state *css;

	css = task_subsys_state_check(p, cpu_cgroup_subsys_id,
612
			lockdep_is_held(&p->pi_lock));
613 614 615
	tg = container_of(css, struct task_group, css);

	return autogroup_task_group(p, tg);
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
}

/* 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 */

642
static void update_rq_clock_task(struct rq *rq, s64 delta);
643

644
static void update_rq_clock(struct rq *rq)
645
{
646
	s64 delta;
647

648 649
	if (rq->skip_clock_update)
		return;
650

651 652 653
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
654 655
}

I
Ingo Molnar 已提交
656 657 658 659 660 661 662 663 664
/*
 * 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
665
/**
666
 * runqueue_is_locked - Returns true if the current cpu runqueue is locked
667
 * @cpu: the processor in question.
I
Ingo Molnar 已提交
668 669 670 671
 *
 * 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.
 */
672
int runqueue_is_locked(int cpu)
I
Ingo Molnar 已提交
673
{
674
	return raw_spin_is_locked(&cpu_rq(cpu)->lock);
I
Ingo Molnar 已提交
675 676
}

I
Ingo Molnar 已提交
677 678 679
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
680 681 682 683

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

I
Ingo Molnar 已提交
684
enum {
P
Peter Zijlstra 已提交
685
#include "sched_features.h"
I
Ingo Molnar 已提交
686 687
};

P
Peter Zijlstra 已提交
688 689 690 691 692
#undef SCHED_FEAT

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

I
Ingo Molnar 已提交
693
const_debug unsigned int sysctl_sched_features =
P
Peter Zijlstra 已提交
694 695 696 697 698 699 700 701 702
#include "sched_features.h"
	0;

#undef SCHED_FEAT

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

703
static __read_mostly char *sched_feat_names[] = {
P
Peter Zijlstra 已提交
704 705 706 707 708 709
#include "sched_features.h"
	NULL
};

#undef SCHED_FEAT

L
Li Zefan 已提交
710
static int sched_feat_show(struct seq_file *m, void *v)
P
Peter Zijlstra 已提交
711 712 713 714
{
	int i;

	for (i = 0; sched_feat_names[i]; i++) {
L
Li Zefan 已提交
715 716 717
		if (!(sysctl_sched_features & (1UL << i)))
			seq_puts(m, "NO_");
		seq_printf(m, "%s ", sched_feat_names[i]);
P
Peter Zijlstra 已提交
718
	}
L
Li Zefan 已提交
719
	seq_puts(m, "\n");
P
Peter Zijlstra 已提交
720

L
Li Zefan 已提交
721
	return 0;
P
Peter Zijlstra 已提交
722 723 724 725 726 727 728
}

static ssize_t
sched_feat_write(struct file *filp, const char __user *ubuf,
		size_t cnt, loff_t *ppos)
{
	char buf[64];
729
	char *cmp;
P
Peter Zijlstra 已提交
730 731 732 733 734 735 736 737 738 739
	int neg = 0;
	int i;

	if (cnt > 63)
		cnt = 63;

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

	buf[cnt] = 0;
740
	cmp = strstrip(buf);
P
Peter Zijlstra 已提交
741

H
Hillf Danton 已提交
742
	if (strncmp(cmp, "NO_", 3) == 0) {
P
Peter Zijlstra 已提交
743 744 745 746 747
		neg = 1;
		cmp += 3;
	}

	for (i = 0; sched_feat_names[i]; i++) {
748
		if (strcmp(cmp, sched_feat_names[i]) == 0) {
P
Peter Zijlstra 已提交
749 750 751 752 753 754 755 756 757 758 759
			if (neg)
				sysctl_sched_features &= ~(1UL << i);
			else
				sysctl_sched_features |= (1UL << i);
			break;
		}
	}

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

760
	*ppos += cnt;
P
Peter Zijlstra 已提交
761 762 763 764

	return cnt;
}

L
Li Zefan 已提交
765 766 767 768 769
static int sched_feat_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, sched_feat_show, NULL);
}

770
static const struct file_operations sched_feat_fops = {
L
Li Zefan 已提交
771 772 773 774 775
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
P
Peter Zijlstra 已提交
776 777 778 779 780 781 782 783 784 785 786 787 788 789
};

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
790

791 792 793 794 795 796
/*
 * 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;

797 798 799 800 801 802 803 804
/*
 * 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
805
/*
P
Peter Zijlstra 已提交
806
 * period over which we measure -rt task cpu usage in us.
P
Peter Zijlstra 已提交
807 808
 * default: 1s
 */
P
Peter Zijlstra 已提交
809
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
810

811 812
static __read_mostly int scheduler_running;

P
Peter Zijlstra 已提交
813 814 815 816 817
/*
 * 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
818

819 820 821 822 823 824 825
static inline u64 global_rt_period(void)
{
	return (u64)sysctl_sched_rt_period * NSEC_PER_USEC;
}

static inline u64 global_rt_runtime(void)
{
826
	if (sysctl_sched_rt_runtime < 0)
827 828 829 830
		return RUNTIME_INF;

	return (u64)sysctl_sched_rt_runtime * NSEC_PER_USEC;
}
P
Peter Zijlstra 已提交
831

L
Linus Torvalds 已提交
832
#ifndef prepare_arch_switch
833 834 835 836 837 838
# define prepare_arch_switch(next)	do { } while (0)
#endif
#ifndef finish_arch_switch
# define finish_arch_switch(prev)	do { } while (0)
#endif

839 840 841 842 843
static inline int task_current(struct rq *rq, struct task_struct *p)
{
	return rq->curr == p;
}

844
static inline int task_running(struct rq *rq, struct task_struct *p)
845
{
P
Peter Zijlstra 已提交
846 847 848
#ifdef CONFIG_SMP
	return p->on_cpu;
#else
849
	return task_current(rq, p);
P
Peter Zijlstra 已提交
850
#endif
851 852
}

P
Peter Zijlstra 已提交
853
#ifndef __ARCH_WANT_UNLOCKED_CTXSW
854
static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
855
{
P
Peter Zijlstra 已提交
856 857 858 859 860 861 862 863
#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->on_cpu = 1;
#endif
864 865
}

866
static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
867
{
P
Peter Zijlstra 已提交
868 869 870 871 872 873 874 875 876
#ifdef CONFIG_SMP
	/*
	 * After ->on_cpu 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->on_cpu = 0;
#endif
877 878 879 880
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
	rq->lock.owner = current;
#endif
881 882 883 884 885 886 887
	/*
	 * 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_);

888
	raw_spin_unlock_irq(&rq->lock);
889 890 891
}

#else /* __ARCH_WANT_UNLOCKED_CTXSW */
892
static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
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.
	 */
P
Peter Zijlstra 已提交
900
	next->on_cpu = 1;
901 902
#endif
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
903
	raw_spin_unlock_irq(&rq->lock);
904
#else
905
	raw_spin_unlock(&rq->lock);
906 907 908
#endif
}

909
static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
910 911 912
{
#ifdef CONFIG_SMP
	/*
P
Peter Zijlstra 已提交
913
	 * After ->on_cpu is cleared, the task can be moved to a different CPU.
914 915 916 917
	 * We must ensure this doesn't happen until the switch is completely
	 * finished.
	 */
	smp_wmb();
P
Peter Zijlstra 已提交
918
	prev->on_cpu = 0;
919 920 921
#endif
#ifndef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
L
Linus Torvalds 已提交
922
#endif
923 924
}
#endif /* __ARCH_WANT_UNLOCKED_CTXSW */
L
Linus Torvalds 已提交
925

926
/*
927
 * __task_rq_lock - lock the rq @p resides on.
928
 */
929
static inline struct rq *__task_rq_lock(struct task_struct *p)
930 931
	__acquires(rq->lock)
{
932 933
	struct rq *rq;

934 935
	lockdep_assert_held(&p->pi_lock);

936
	for (;;) {
937
		rq = task_rq(p);
938
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
939
		if (likely(rq == task_rq(p)))
940
			return rq;
941
		raw_spin_unlock(&rq->lock);
942 943 944
	}
}

L
Linus Torvalds 已提交
945
/*
946
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
L
Linus Torvalds 已提交
947
 */
948
static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
949
	__acquires(p->pi_lock)
L
Linus Torvalds 已提交
950 951
	__acquires(rq->lock)
{
952
	struct rq *rq;
L
Linus Torvalds 已提交
953

954
	for (;;) {
955
		raw_spin_lock_irqsave(&p->pi_lock, *flags);
956
		rq = task_rq(p);
957
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
958
		if (likely(rq == task_rq(p)))
959
			return rq;
960 961
		raw_spin_unlock(&rq->lock);
		raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
962 963 964
	}
}

A
Alexey Dobriyan 已提交
965
static void __task_rq_unlock(struct rq *rq)
966 967
	__releases(rq->lock)
{
968
	raw_spin_unlock(&rq->lock);
969 970
}

971 972
static inline void
task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
L
Linus Torvalds 已提交
973
	__releases(rq->lock)
974
	__releases(p->pi_lock)
L
Linus Torvalds 已提交
975
{
976 977
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
978 979 980
}

/*
981
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
982
 */
A
Alexey Dobriyan 已提交
983
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
984 985
	__acquires(rq->lock)
{
986
	struct rq *rq;
L
Linus Torvalds 已提交
987 988 989

	local_irq_disable();
	rq = this_rq();
990
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
991 992 993 994

	return rq;
}

P
Peter Zijlstra 已提交
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
#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;
1016
	if (!cpu_active(cpu_of(rq)))
1017
		return 0;
P
Peter Zijlstra 已提交
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
	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());

1037
	raw_spin_lock(&rq->lock);
1038
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
1039
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
1040
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
1041 1042 1043 1044

	return HRTIMER_NORESTART;
}

1045
#ifdef CONFIG_SMP
1046 1047 1048 1049
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
1050
{
1051
	struct rq *rq = arg;
1052

1053
	raw_spin_lock(&rq->lock);
1054 1055
	hrtimer_restart(&rq->hrtick_timer);
	rq->hrtick_csd_pending = 0;
1056
	raw_spin_unlock(&rq->lock);
1057 1058
}

1059 1060 1061 1062 1063 1064
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
static void hrtick_start(struct rq *rq, u64 delay)
1065
{
1066 1067
	struct hrtimer *timer = &rq->hrtick_timer;
	ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
1068

1069
	hrtimer_set_expires(timer, time);
1070 1071 1072 1073

	if (rq == this_rq()) {
		hrtimer_restart(timer);
	} else if (!rq->hrtick_csd_pending) {
1074
		__smp_call_function_single(cpu_of(rq), &rq->hrtick_csd, 0);
1075 1076
		rq->hrtick_csd_pending = 1;
	}
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090
}

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:
1091
		hrtick_clear(cpu_rq(cpu));
1092 1093 1094 1095 1096 1097
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

1098
static __init void init_hrtick(void)
1099 1100 1101
{
	hotcpu_notifier(hotplug_hrtick, 0);
}
1102 1103 1104 1105 1106 1107 1108 1109
#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)
{
1110
	__hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
1111
			HRTIMER_MODE_REL_PINNED, 0);
1112
}
1113

A
Andrew Morton 已提交
1114
static inline void init_hrtick(void)
P
Peter Zijlstra 已提交
1115 1116
{
}
1117
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
1118

1119
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
1120
{
1121 1122
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
1123

1124 1125 1126 1127
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
1128

1129 1130
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
1131
}
A
Andrew Morton 已提交
1132
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
1133 1134 1135 1136 1137 1138 1139 1140
static inline void hrtick_clear(struct rq *rq)
{
}

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

1141 1142 1143
static inline void init_hrtick(void)
{
}
A
Andrew Morton 已提交
1144
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
1145

I
Ingo Molnar 已提交
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
/*
 * 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

1159
static void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
1160 1161 1162
{
	int cpu;

1163
	assert_raw_spin_locked(&task_rq(p)->lock);
I
Ingo Molnar 已提交
1164

1165
	if (test_tsk_need_resched(p))
I
Ingo Molnar 已提交
1166 1167
		return;

1168
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184

	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;

1185
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
1186 1187
		return;
	resched_task(cpu_curr(cpu));
1188
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
1189
}
1190 1191

#ifdef CONFIG_NO_HZ
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
/*
 * 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;
}
1213 1214 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
/*
 * 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()
	 */
1245
	set_tsk_need_resched(rq->idle);
1246 1247 1248 1249 1250 1251

	/* 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
1252

1253
#endif /* CONFIG_NO_HZ */
1254

1255 1256 1257 1258 1259 1260 1261 1262 1263 1264
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) {
1265 1266 1267 1268 1269 1270
		/*
		 * 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));
1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
		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);
}

1282
#else /* !CONFIG_SMP */
1283
static void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
1284
{
1285
	assert_raw_spin_locked(&task_rq(p)->lock);
1286
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
1287
}
1288 1289 1290 1291

static void sched_rt_avg_update(struct rq *rq, u64 rt_delta)
{
}
1292 1293 1294 1295

static void sched_avg_update(struct rq *rq)
{
}
1296
#endif /* CONFIG_SMP */
I
Ingo Molnar 已提交
1297

1298 1299 1300 1301 1302 1303 1304 1305
#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
1306 1307 1308
/*
 * Shift right and round:
 */
I
Ingo Molnar 已提交
1309
#define SRR(x, y) (((x) + (1UL << ((y) - 1))) >> (y))
I
Ingo Molnar 已提交
1310

1311 1312 1313
/*
 * delta *= weight / lw
 */
1314
static unsigned long
1315 1316 1317 1318 1319
calc_delta_mine(unsigned long delta_exec, unsigned long weight,
		struct load_weight *lw)
{
	u64 tmp;

1320 1321 1322 1323 1324 1325 1326
	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);
	}
1327 1328 1329 1330 1331

	tmp = (u64)delta_exec * weight;
	/*
	 * Check whether we'd overflow the 64-bit multiplication:
	 */
I
Ingo Molnar 已提交
1332
	if (unlikely(tmp > WMULT_CONST))
I
Ingo Molnar 已提交
1333
		tmp = SRR(SRR(tmp, WMULT_SHIFT/2) * lw->inv_weight,
I
Ingo Molnar 已提交
1334 1335
			WMULT_SHIFT/2);
	else
I
Ingo Molnar 已提交
1336
		tmp = SRR(tmp * lw->inv_weight, WMULT_SHIFT);
1337

1338
	return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX);
1339 1340
}

1341
static inline void update_load_add(struct load_weight *lw, unsigned long inc)
1342 1343
{
	lw->weight += inc;
I
Ingo Molnar 已提交
1344
	lw->inv_weight = 0;
1345 1346
}

1347
static inline void update_load_sub(struct load_weight *lw, unsigned long dec)
1348 1349
{
	lw->weight -= dec;
I
Ingo Molnar 已提交
1350
	lw->inv_weight = 0;
1351 1352
}

P
Peter Zijlstra 已提交
1353 1354 1355 1356 1357 1358
static inline void update_load_set(struct load_weight *lw, unsigned long w)
{
	lw->weight = w;
	lw->inv_weight = 0;
}

1359 1360 1361 1362
/*
 * 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
1363
 * scheduling class and "nice" value. For SCHED_NORMAL tasks this is just a
1364 1365 1366 1367
 * scaled version of the new time slice allocation that they receive on time
 * slice expiry etc.
 */

P
Peter Zijlstra 已提交
1368 1369
#define WEIGHT_IDLEPRIO                3
#define WMULT_IDLEPRIO         1431655765
I
Ingo Molnar 已提交
1370 1371 1372 1373 1374 1375 1376 1377 1378

/*
 * 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
1379 1380 1381
 * 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
1382 1383
 */
static const int prio_to_weight[40] = {
1384 1385 1386 1387 1388 1389 1390 1391
 /* -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
1392 1393
};

1394 1395 1396 1397 1398 1399 1400
/*
 * 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
1401
static const u32 prio_to_wmult[40] = {
1402 1403 1404 1405 1406 1407 1408 1409
 /* -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
1410
};
1411

1412 1413 1414 1415 1416 1417 1418 1419
/* 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,
};

1420 1421
#ifdef CONFIG_CGROUP_CPUACCT
static void cpuacct_charge(struct task_struct *tsk, u64 cputime);
1422 1423
static void cpuacct_update_stats(struct task_struct *tsk,
		enum cpuacct_stat_index idx, cputime_t val);
1424 1425
#else
static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
1426 1427
static inline void cpuacct_update_stats(struct task_struct *tsk,
		enum cpuacct_stat_index idx, cputime_t val) {}
1428 1429
#endif

1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
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
1440
#if (defined(CONFIG_SMP) && defined(CONFIG_FAIR_GROUP_SCHED)) || defined(CONFIG_RT_GROUP_SCHED)
P
Peter Zijlstra 已提交
1441
typedef int (*tg_visitor)(struct task_group *, void *);
1442 1443 1444 1445 1446

/*
 * 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
1447
static int walk_tg_tree(tg_visitor down, tg_visitor up, void *data)
1448 1449
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
1450
	int ret;
1451 1452 1453 1454

	rcu_read_lock();
	parent = &root_task_group;
down:
P
Peter Zijlstra 已提交
1455 1456 1457
	ret = (*down)(parent, data);
	if (ret)
		goto out_unlock;
1458 1459 1460 1461 1462 1463 1464
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
1465 1466 1467
	ret = (*up)(parent, data);
	if (ret)
		goto out_unlock;
1468 1469 1470 1471 1472

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
P
Peter Zijlstra 已提交
1473
out_unlock:
1474
	rcu_read_unlock();
P
Peter Zijlstra 已提交
1475 1476

	return ret;
1477 1478
}

P
Peter Zijlstra 已提交
1479 1480 1481
static int tg_nop(struct task_group *tg, void *data)
{
	return 0;
1482
}
P
Peter Zijlstra 已提交
1483 1484 1485
#endif

#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
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 1517 1518 1519 1520 1521 1522 1523 1524
/* 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);
}

1525 1526
static unsigned long power_of(int cpu)
{
1527
	return cpu_rq(cpu)->cpu_power;
1528 1529
}

P
Peter Zijlstra 已提交
1530 1531 1532 1533 1534
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);
1535
	unsigned long nr_running = ACCESS_ONCE(rq->nr_running);
P
Peter Zijlstra 已提交
1536

1537 1538
	if (nr_running)
		rq->avg_load_per_task = rq->load.weight / nr_running;
1539 1540
	else
		rq->avg_load_per_task = 0;
P
Peter Zijlstra 已提交
1541 1542 1543 1544 1545

	return rq->avg_load_per_task;
}

#ifdef CONFIG_FAIR_GROUP_SCHED
1546 1547

/*
1548 1549 1550
 * 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.
1551
 */
P
Peter Zijlstra 已提交
1552
static int tg_load_down(struct task_group *tg, void *data)
1553
{
1554
	unsigned long load;
P
Peter Zijlstra 已提交
1555
	long cpu = (long)data;
1556

1557 1558 1559 1560
	if (!tg->parent) {
		load = cpu_rq(cpu)->load.weight;
	} else {
		load = tg->parent->cfs_rq[cpu]->h_load;
P
Peter Zijlstra 已提交
1561
		load *= tg->se[cpu]->load.weight;
1562 1563
		load /= tg->parent->cfs_rq[cpu]->load.weight + 1;
	}
1564

1565
	tg->cfs_rq[cpu]->h_load = load;
1566

P
Peter Zijlstra 已提交
1567
	return 0;
1568 1569
}

P
Peter Zijlstra 已提交
1570
static void update_h_load(long cpu)
1571
{
P
Peter Zijlstra 已提交
1572
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
1573 1574
}

1575 1576
#endif

1577 1578
#ifdef CONFIG_PREEMPT

1579 1580
static void double_rq_lock(struct rq *rq1, struct rq *rq2);

1581
/*
1582 1583 1584 1585 1586 1587
 * 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.
1588
 */
1589 1590 1591 1592 1593
static inline int _double_lock_balance(struct rq *this_rq, struct rq *busiest)
	__releases(this_rq->lock)
	__acquires(busiest->lock)
	__acquires(this_rq->lock)
{
1594
	raw_spin_unlock(&this_rq->lock);
1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
	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)
1609 1610 1611 1612 1613 1614
	__releases(this_rq->lock)
	__acquires(busiest->lock)
	__acquires(this_rq->lock)
{
	int ret = 0;

1615
	if (unlikely(!raw_spin_trylock(&busiest->lock))) {
1616
		if (busiest < this_rq) {
1617 1618 1619 1620
			raw_spin_unlock(&this_rq->lock);
			raw_spin_lock(&busiest->lock);
			raw_spin_lock_nested(&this_rq->lock,
					      SINGLE_DEPTH_NESTING);
1621 1622
			ret = 1;
		} else
1623 1624
			raw_spin_lock_nested(&busiest->lock,
					      SINGLE_DEPTH_NESTING);
1625 1626 1627 1628
	}
	return ret;
}

1629 1630 1631 1632 1633 1634 1635 1636 1637
#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 */
1638
		raw_spin_unlock(&this_rq->lock);
1639 1640 1641 1642 1643 1644
		BUG_ON(1);
	}

	return _double_lock_balance(this_rq, busiest);
}

1645 1646 1647
static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest)
	__releases(busiest->lock)
{
1648
	raw_spin_unlock(&busiest->lock);
1649 1650
	lock_set_subclass(&this_rq->lock.dep_map, 0, _RET_IP_);
}
1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693

/*
 * 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);
}

1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
#else /* CONFIG_SMP */

/*
 * 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());
	BUG_ON(rq1 != rq2);
	raw_spin_lock(&rq1->lock);
	__acquire(rq2->lock);	/* Fake it out ;) */
}

/*
 * 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)
{
	BUG_ON(rq1 != rq2);
	raw_spin_unlock(&rq1->lock);
	__release(rq2->lock);
}

1727 1728
#endif

1729
static void calc_load_account_idle(struct rq *this_rq);
1730
static void update_sysctl(void);
1731
static int get_update_sysctl_factor(void);
1732
static void update_cpu_load(struct rq *this_rq);
1733

P
Peter Zijlstra 已提交
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
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
}
1747

1748
static const struct sched_class rt_sched_class;
I
Ingo Molnar 已提交
1749

1750
#define sched_class_highest (&stop_sched_class)
1751 1752
#define for_each_class(class) \
   for (class = sched_class_highest; class; class = class->next)
I
Ingo Molnar 已提交
1753

1754 1755
#include "sched_stats.h"

1756
static void inc_nr_running(struct rq *rq)
1757 1758 1759 1760
{
	rq->nr_running++;
}

1761
static void dec_nr_running(struct rq *rq)
1762 1763 1764 1765
{
	rq->nr_running--;
}

1766 1767
static void set_load_weight(struct task_struct *p)
{
I
Ingo Molnar 已提交
1768 1769 1770 1771 1772 1773 1774 1775
	/*
	 * 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;
	}
1776

I
Ingo Molnar 已提交
1777 1778
	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];
1779 1780
}

1781
static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
1782
{
1783
	update_rq_clock(rq);
I
Ingo Molnar 已提交
1784
	sched_info_queued(p);
1785
	p->sched_class->enqueue_task(rq, p, flags);
1786 1787
}

1788
static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
1789
{
1790
	update_rq_clock(rq);
1791
	sched_info_dequeued(p);
1792
	p->sched_class->dequeue_task(rq, p, flags);
1793 1794
}

1795 1796 1797
/*
 * activate_task - move a task to the runqueue.
 */
1798
static void activate_task(struct rq *rq, struct task_struct *p, int flags)
1799 1800 1801 1802
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

1803
	enqueue_task(rq, p, flags);
1804 1805 1806 1807 1808 1809
	inc_nr_running(rq);
}

/*
 * deactivate_task - remove a task from the runqueue.
 */
1810
static void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
1811 1812 1813 1814
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

1815
	dequeue_task(rq, p, flags);
1816 1817 1818
	dec_nr_running(rq);
}

1819 1820
#ifdef CONFIG_IRQ_TIME_ACCOUNTING

1821 1822 1823 1824 1825 1826 1827
/*
 * There are no locks covering percpu hardirq/softirq time.
 * They are only modified in account_system_vtime, on corresponding CPU
 * with interrupts disabled. So, writes are safe.
 * They are read and saved off onto struct rq in update_rq_clock().
 * This may result in other CPU reading this CPU's irq time and can
 * race with irq/account_system_vtime on this CPU. We would either get old
1828 1829 1830
 * or new value with a side effect of accounting a slice of irq time to wrong
 * task when irq is in progress while we read rq->clock. That is a worthy
 * compromise in place of having locks on each irq in account_system_time.
1831
 */
1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847
static DEFINE_PER_CPU(u64, cpu_hardirq_time);
static DEFINE_PER_CPU(u64, cpu_softirq_time);

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

void enable_sched_clock_irqtime(void)
{
	sched_clock_irqtime = 1;
}

void disable_sched_clock_irqtime(void)
{
	sched_clock_irqtime = 0;
}

1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
#ifndef CONFIG_64BIT
static DEFINE_PER_CPU(seqcount_t, irq_time_seq);

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

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

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

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

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

static inline void irq_time_write_end(void)
{
}

static inline u64 irq_time_read(int cpu)
1886 1887 1888
{
	return per_cpu(cpu_softirq_time, cpu) + per_cpu(cpu_hardirq_time, cpu);
}
1889
#endif /* CONFIG_64BIT */
1890

1891 1892 1893 1894
/*
 * Called before incrementing preempt_count on {soft,}irq_enter
 * and before decrementing preempt_count on {soft,}irq_exit.
 */
1895 1896 1897
void account_system_vtime(struct task_struct *curr)
{
	unsigned long flags;
1898
	s64 delta;
1899 1900 1901 1902 1903 1904 1905 1906
	int cpu;

	if (!sched_clock_irqtime)
		return;

	local_irq_save(flags);

	cpu = smp_processor_id();
1907 1908 1909
	delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
	__this_cpu_add(irq_start_time, delta);

1910
	irq_time_write_begin();
1911 1912 1913 1914 1915 1916 1917
	/*
	 * We do not account for softirq time from ksoftirqd here.
	 * We want to continue accounting softirq time to ksoftirqd thread
	 * in that case, so as not to confuse scheduler with a special task
	 * that do not consume any time, but still wants to run.
	 */
	if (hardirq_count())
1918
		__this_cpu_add(cpu_hardirq_time, delta);
1919
	else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
1920
		__this_cpu_add(cpu_softirq_time, delta);
1921

1922
	irq_time_write_end();
1923 1924
	local_irq_restore(flags);
}
I
Ingo Molnar 已提交
1925
EXPORT_SYMBOL_GPL(account_system_vtime);
1926

1927
static void update_rq_clock_task(struct rq *rq, s64 delta)
1928
{
1929 1930
	s64 irq_delta;

1931
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956

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

	rq->prev_irq_time += irq_delta;
	delta -= irq_delta;
	rq->clock_task += delta;

	if (irq_delta && sched_feat(NONIRQ_POWER))
		sched_rt_avg_update(rq, irq_delta);
1957 1958
}

1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
static int irqtime_account_hi_update(void)
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_hardirq_time);
	if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat->irq))
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

static int irqtime_account_si_update(void)
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_softirq_time);
	if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat->softirq))
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

1989
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
1990

1991 1992
#define sched_clock_irqtime	(0)

1993
static void update_rq_clock_task(struct rq *rq, s64 delta)
1994
{
1995
	rq->clock_task += delta;
1996 1997
}

1998
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
1999

2000 2001 2002
#include "sched_idletask.c"
#include "sched_fair.c"
#include "sched_rt.c"
2003
#include "sched_autogroup.c"
2004
#include "sched_stoptask.c"
2005 2006 2007 2008
#ifdef CONFIG_SCHED_DEBUG
# include "sched_debug.c"
#endif

2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038
void sched_set_stop_task(int cpu, struct task_struct *stop)
{
	struct sched_param param = { .sched_priority = MAX_RT_PRIO - 1 };
	struct task_struct *old_stop = cpu_rq(cpu)->stop;

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

		stop->sched_class = &stop_sched_class;
	}

	cpu_rq(cpu)->stop = stop;

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

2039
/*
I
Ingo Molnar 已提交
2040
 * __normal_prio - return the priority that is based on the static prio
2041 2042 2043
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
2044
	return p->static_prio;
2045 2046
}

2047 2048 2049 2050 2051 2052 2053
/*
 * 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.
 */
2054
static inline int normal_prio(struct task_struct *p)
2055 2056 2057
{
	int prio;

2058
	if (task_has_rt_policy(p))
2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
		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.
 */
2072
static int effective_prio(struct task_struct *p)
2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
{
	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
2085 2086 2087 2088
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
2089
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
2090 2091 2092 2093
{
	return cpu_curr(task_cpu(p)) == p;
}

2094 2095
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
P
Peter Zijlstra 已提交
2096
				       int oldprio)
2097 2098 2099
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
2100 2101 2102 2103
			prev_class->switched_from(rq, p);
		p->sched_class->switched_to(rq, p);
	} else if (oldprio != p->prio)
		p->sched_class->prio_changed(rq, p, oldprio);
2104 2105
}

2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
static void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
{
	const struct sched_class *class;

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

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
P
Peter Zijlstra 已提交
2127
	if (rq->curr->on_rq && test_tsk_need_resched(rq->curr))
2128 2129 2130
		rq->skip_clock_update = 1;
}

L
Linus Torvalds 已提交
2131
#ifdef CONFIG_SMP
2132 2133 2134
/*
 * Is this task likely cache-hot:
 */
2135
static int
2136 2137 2138 2139
task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
{
	s64 delta;

P
Peter Zijlstra 已提交
2140 2141 2142
	if (p->sched_class != &fair_sched_class)
		return 0;

2143 2144 2145
	if (unlikely(p->policy == SCHED_IDLE))
		return 0;

2146 2147 2148
	/*
	 * Buddy candidates are cache hot:
	 */
2149
	if (sched_feat(CACHE_HOT_BUDDY) && this_rq()->nr_running &&
P
Peter Zijlstra 已提交
2150 2151
			(&p->se == cfs_rq_of(&p->se)->next ||
			 &p->se == cfs_rq_of(&p->se)->last))
2152 2153
		return 1;

2154 2155 2156 2157 2158
	if (sysctl_sched_migration_cost == -1)
		return 1;
	if (sysctl_sched_migration_cost == 0)
		return 0;

2159 2160 2161 2162 2163
	delta = now - p->se.exec_start;

	return delta < (s64)sysctl_sched_migration_cost;
}

I
Ingo Molnar 已提交
2164
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
2165
{
2166 2167 2168 2169 2170
#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
2171 2172
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
2173 2174 2175 2176 2177

#ifdef CONFIG_LOCKDEP
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
2178 2179
#endif

2180
	trace_sched_migrate_task(p, new_cpu);
2181

2182 2183 2184 2185
	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
2186 2187

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
2188 2189
}

2190
struct migration_arg {
2191
	struct task_struct *task;
L
Linus Torvalds 已提交
2192
	int dest_cpu;
2193
};
L
Linus Torvalds 已提交
2194

2195 2196
static int migration_cpu_stop(void *data);

L
Linus Torvalds 已提交
2197 2198 2199 2200
/*
 * The task's runqueue lock must be held.
 * Returns true if you have to wait for migration thread.
 */
2201
static bool need_migrate_task(struct task_struct *p)
L
Linus Torvalds 已提交
2202 2203 2204
{
	/*
	 * If the task is not on a runqueue (and not running), then
2205
	 * the next wake-up will properly place the task.
L
Linus Torvalds 已提交
2206
	 */
2207 2208 2209
	bool running = p->on_rq || p->on_cpu;
	smp_rmb(); /* finish_lock_switch() */
	return running;
L
Linus Torvalds 已提交
2210 2211 2212 2213 2214
}

/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
2215 2216 2217 2218 2219 2220 2221
 * 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
2222 2223 2224 2225 2226 2227
 * 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
2228
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
2229 2230
{
	unsigned long flags;
I
Ingo Molnar 已提交
2231
	int running, on_rq;
R
Roland McGrath 已提交
2232
	unsigned long ncsw;
2233
	struct rq *rq;
L
Linus Torvalds 已提交
2234

2235 2236 2237 2238 2239 2240 2241 2242
	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);
2243

2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
		/*
		 * 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
2255 2256 2257
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
2258
			cpu_relax();
R
Roland McGrath 已提交
2259
		}
2260

2261 2262 2263 2264 2265 2266
		/*
		 * 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);
2267
		trace_sched_wait_task(p);
2268
		running = task_running(rq, p);
P
Peter Zijlstra 已提交
2269
		on_rq = p->on_rq;
R
Roland McGrath 已提交
2270
		ncsw = 0;
2271
		if (!match_state || p->state == match_state)
2272
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
2273
		task_rq_unlock(rq, p, &flags);
2274

R
Roland McGrath 已提交
2275 2276 2277 2278 2279 2280
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
		/*
		 * 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;
		}
2291

2292 2293 2294 2295 2296
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
2297
		 * So if it was still runnable (but just not actively
2298 2299 2300 2301
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
		if (unlikely(on_rq)) {
2302 2303 2304 2305
			ktime_t to = ktime_set(0, NSEC_PER_SEC/HZ);

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
2306 2307
			continue;
		}
2308

2309 2310 2311 2312 2313 2314 2315
		/*
		 * 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
2316 2317

	return ncsw;
L
Linus Torvalds 已提交
2318 2319 2320 2321 2322 2323 2324 2325 2326
}

/***
 * kick_process - kick a running thread to enter/exit the kernel
 * @p: the to-be-kicked thread
 *
 * Cause a process which is running on another CPU to enter
 * kernel-mode, without any delay. (to get signals handled.)
 *
L
Lucas De Marchi 已提交
2327
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
2328 2329 2330 2331 2332
 * 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.
 */
2333
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
2334 2335 2336 2337 2338 2339 2340 2341 2342
{
	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
2343
EXPORT_SYMBOL_GPL(kick_process);
N
Nick Piggin 已提交
2344
#endif /* CONFIG_SMP */
L
Linus Torvalds 已提交
2345

2346
#ifdef CONFIG_SMP
2347
/*
2348
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
2349
 */
2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365
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. */
2366 2367 2368 2369 2370 2371 2372 2373 2374
	dest_cpu = cpuset_cpus_allowed_fallback(p);
	/*
	 * Don't tell them about moving exiting tasks or
	 * kernel threads (both mm NULL), since they never
	 * leave kernel.
	 */
	if (p->mm && printk_ratelimit()) {
		printk(KERN_INFO "process %d (%s) no longer affine to cpu%d\n",
				task_pid_nr(p), p->comm, cpu);
2375 2376 2377 2378 2379
	}

	return dest_cpu;
}

2380
/*
2381
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
2382
 */
2383
static inline
2384
int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
2385
{
2386
	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398

	/*
	 * 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
2399
		     !cpu_online(cpu)))
2400
		cpu = select_fallback_rq(task_cpu(p), p);
2401 2402

	return cpu;
2403
}
2404 2405 2406 2407 2408 2409

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
2410 2411
#endif

P
Peter Zijlstra 已提交
2412 2413
static void
ttwu_stat(struct rq *rq, struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
2414
{
P
Peter Zijlstra 已提交
2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435
#ifdef CONFIG_SCHEDSTATS
#ifdef CONFIG_SMP
	int this_cpu = smp_processor_id();

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

		schedstat_inc(p, se.statistics.nr_wakeups_remote);
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
	}
#endif /* CONFIG_SMP */

	schedstat_inc(rq, ttwu_count);
T
Tejun Heo 已提交
2436
	schedstat_inc(p, se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
2437 2438

	if (wake_flags & WF_SYNC)
T
Tejun Heo 已提交
2439
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
2440 2441

	if (cpu != task_cpu(p))
T
Tejun Heo 已提交
2442 2443
		schedstat_inc(p, se.statistics.nr_wakeups_migrate);

P
Peter Zijlstra 已提交
2444 2445 2446 2447 2448
#endif /* CONFIG_SCHEDSTATS */
}

static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
{
T
Tejun Heo 已提交
2449
	activate_task(rq, p, en_flags);
P
Peter Zijlstra 已提交
2450
	p->on_rq = 1;
2451 2452 2453 2454

	/* if a worker is waking up, notify workqueue */
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
2455 2456
}

2457 2458
static void
ttwu_post_activation(struct task_struct *p, struct rq *rq, int wake_flags)
T
Tejun Heo 已提交
2459
{
2460
	trace_sched_wakeup(p, true);
T
Tejun Heo 已提交
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481
	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
}

/**
L
Linus Torvalds 已提交
2482
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
2483
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
2484
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
2485
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
2486 2487 2488 2489 2490 2491 2492
 *
 * 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
2493 2494
 * 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
2495
 */
P
Peter Zijlstra 已提交
2496 2497
static int try_to_wake_up(struct task_struct *p, unsigned int state,
			  int wake_flags)
L
Linus Torvalds 已提交
2498
{
2499
	int cpu, orig_cpu, this_cpu, success = 0;
L
Linus Torvalds 已提交
2500
	unsigned long flags;
2501
	unsigned long en_flags = ENQUEUE_WAKEUP;
2502
	struct rq *rq;
L
Linus Torvalds 已提交
2503

P
Peter Zijlstra 已提交
2504
	this_cpu = get_cpu();
P
Peter Zijlstra 已提交
2505

2506
	smp_wmb();
2507 2508
	raw_spin_lock_irqsave(&p->pi_lock, flags);
	rq = __task_rq_lock(p);
P
Peter Zijlstra 已提交
2509
	if (!(p->state & state))
L
Linus Torvalds 已提交
2510 2511
		goto out;

P
Peter Zijlstra 已提交
2512 2513
	cpu = task_cpu(p);

P
Peter Zijlstra 已提交
2514
	if (p->on_rq)
L
Linus Torvalds 已提交
2515 2516
		goto out_running;

2517
	orig_cpu = cpu;
L
Linus Torvalds 已提交
2518 2519 2520 2521
#ifdef CONFIG_SMP
	if (unlikely(task_running(rq, p)))
		goto out_activate;

2522
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
2523
	p->state = TASK_WAKING;
2524

2525
	if (p->sched_class->task_waking) {
2526
		p->sched_class->task_waking(p);
2527 2528
		en_flags |= ENQUEUE_WAKING;
	}
2529

2530
	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
2531
	if (cpu != orig_cpu)
2532
		set_task_cpu(p, cpu);
2533
	__task_rq_unlock(rq);
P
Peter Zijlstra 已提交
2534

2535 2536
	rq = cpu_rq(cpu);
	raw_spin_lock(&rq->lock);
2537

2538 2539 2540 2541 2542 2543 2544
	/*
	 * 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
2545
	WARN_ON(p->state != TASK_WAKING);
L
Linus Torvalds 已提交
2546

2547 2548 2549
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;

L
Linus Torvalds 已提交
2550 2551
out_activate:
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
2552
	ttwu_activate(rq, p, en_flags);
L
Linus Torvalds 已提交
2553
out_running:
2554
	ttwu_post_activation(p, rq, wake_flags);
P
Peter Zijlstra 已提交
2555
	ttwu_stat(rq, p, cpu, wake_flags);
2556
	success = 1;
L
Linus Torvalds 已提交
2557
out:
2558 2559
	__task_rq_unlock(rq);
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
2560
	put_cpu();
L
Linus Torvalds 已提交
2561 2562 2563 2564

	return success;
}

T
Tejun Heo 已提交
2565 2566 2567 2568
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
2569
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
2570
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
2571
 * the current task.
T
Tejun Heo 已提交
2572 2573 2574 2575 2576 2577 2578 2579 2580
 */
static void try_to_wake_up_local(struct task_struct *p)
{
	struct rq *rq = task_rq(p);

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

2581 2582 2583 2584 2585 2586
	if (!raw_spin_trylock(&p->pi_lock)) {
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
	}

T
Tejun Heo 已提交
2587
	if (!(p->state & TASK_NORMAL))
2588
		goto out;
T
Tejun Heo 已提交
2589

P
Peter Zijlstra 已提交
2590
	if (!p->on_rq)
P
Peter Zijlstra 已提交
2591 2592
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);

2593
	ttwu_post_activation(p, rq, 0);
P
Peter Zijlstra 已提交
2594
	ttwu_stat(rq, p, smp_processor_id(), 0);
2595 2596
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2597 2598
}

2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609
/**
 * 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.
 */
2610
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2611
{
2612
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
2613 2614 2615
}
EXPORT_SYMBOL(wake_up_process);

2616
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2617 2618 2619 2620 2621 2622 2623
{
	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
2624 2625 2626 2627 2628
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
P
Peter Zijlstra 已提交
2629 2630 2631
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2632 2633
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2634
	p->se.prev_sum_exec_runtime	= 0;
2635
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2636
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2637
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2638 2639

#ifdef CONFIG_SCHEDSTATS
2640
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2641
#endif
N
Nick Piggin 已提交
2642

P
Peter Zijlstra 已提交
2643
	INIT_LIST_HEAD(&p->rt.run_list);
N
Nick Piggin 已提交
2644

2645 2646 2647
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
I
Ingo Molnar 已提交
2648 2649 2650 2651 2652 2653 2654
}

/*
 * fork()/clone()-time setup:
 */
void sched_fork(struct task_struct *p, int clone_flags)
{
2655
	unsigned long flags;
I
Ingo Molnar 已提交
2656 2657 2658
	int cpu = get_cpu();

	__sched_fork(p);
2659
	/*
2660
	 * We mark the process as running here. This guarantees that
2661 2662 2663
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2664
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
2665

2666 2667 2668 2669
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2670
		if (p->policy == SCHED_FIFO || p->policy == SCHED_RR) {
2671
			p->policy = SCHED_NORMAL;
2672 2673
			p->normal_prio = p->static_prio;
		}
2674

2675 2676
		if (PRIO_TO_NICE(p->static_prio) < 0) {
			p->static_prio = NICE_TO_PRIO(0);
2677
			p->normal_prio = p->static_prio;
2678 2679 2680
			set_load_weight(p);
		}

2681 2682 2683 2684 2685 2686
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2687

2688 2689 2690 2691 2692
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

H
Hiroshi Shimamoto 已提交
2693 2694
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
2695

P
Peter Zijlstra 已提交
2696 2697 2698
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

2699 2700 2701 2702 2703 2704 2705
	/*
	 * 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.
	 */
2706
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2707
	set_task_cpu(p, cpu);
2708
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2709

2710
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
I
Ingo Molnar 已提交
2711
	if (likely(sched_info_on()))
2712
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2713
#endif
P
Peter Zijlstra 已提交
2714 2715
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2716
#endif
L
Linus Torvalds 已提交
2717
#ifdef CONFIG_PREEMPT
2718
	/* Want to start with kernel preemption disabled. */
A
Al Viro 已提交
2719
	task_thread_info(p)->preempt_count = 1;
L
Linus Torvalds 已提交
2720
#endif
2721
#ifdef CONFIG_SMP
2722
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2723
#endif
2724

N
Nick Piggin 已提交
2725
	put_cpu();
L
Linus Torvalds 已提交
2726 2727 2728 2729 2730 2731 2732 2733 2734
}

/*
 * 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.
 */
2735
void wake_up_new_task(struct task_struct *p, unsigned long clone_flags)
L
Linus Torvalds 已提交
2736 2737
{
	unsigned long flags;
I
Ingo Molnar 已提交
2738
	struct rq *rq;
2739

2740
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2741 2742 2743 2744 2745 2746
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
	 *  - cpus_allowed can change in the fork path
	 *  - any previously selected cpu might disappear through hotplug
	 */
2747
	set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
2748 2749
#endif

2750
	rq = __task_rq_lock(p);
P
Peter Zijlstra 已提交
2751
	activate_task(rq, p, 0);
P
Peter Zijlstra 已提交
2752
	p->on_rq = 1;
2753
	trace_sched_wakeup_new(p, true);
P
Peter Zijlstra 已提交
2754
	check_preempt_curr(rq, p, WF_FORK);
2755
#ifdef CONFIG_SMP
2756 2757
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
2758
#endif
2759
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
2760 2761
}

2762 2763 2764
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
2765
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2766
 * @notifier: notifier struct to register
2767 2768 2769 2770 2771 2772 2773 2774 2775
 */
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
2776
 * @notifier: notifier struct to unregister
2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805
 *
 * 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);
}

2806
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817

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)
{
}

2818
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2819

2820 2821 2822
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2823
 * @prev: the current task that is being switched out
2824 2825 2826 2827 2828 2829 2830 2831 2832
 * @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.
 */
2833 2834 2835
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2836
{
2837 2838
	sched_info_switch(prev, next);
	perf_event_task_sched_out(prev, next);
2839
	fire_sched_out_preempt_notifiers(prev, next);
2840 2841
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
2842
	trace_sched_switch(prev, next);
2843 2844
}

L
Linus Torvalds 已提交
2845 2846
/**
 * finish_task_switch - clean up after a task-switch
2847
 * @rq: runqueue associated with task-switch
L
Linus Torvalds 已提交
2848 2849
 * @prev: the thread we just switched away from.
 *
2850 2851 2852 2853
 * 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
2854 2855
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2856
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2857 2858 2859
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
A
Alexey Dobriyan 已提交
2860
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
L
Linus Torvalds 已提交
2861 2862 2863
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2864
	long prev_state;
L
Linus Torvalds 已提交
2865 2866 2867 2868 2869

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2870
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2871 2872
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2873
	 * The test for TASK_DEAD must occur while the runqueue locks are
L
Linus Torvalds 已提交
2874 2875 2876 2877 2878
	 * 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
2879
	prev_state = prev->state;
2880
	finish_arch_switch(prev);
2881 2882 2883
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_disable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
2884
	perf_event_task_sched_in(current);
2885 2886 2887
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
2888
	finish_lock_switch(rq, prev);
S
Steven Rostedt 已提交
2889

2890
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2891 2892
	if (mm)
		mmdrop(mm);
2893
	if (unlikely(prev_state == TASK_DEAD)) {
2894 2895 2896
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2897
		 */
2898
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
2899
		put_task_struct(prev);
2900
	}
L
Linus Torvalds 已提交
2901 2902
}

2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917
#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;

2918
		raw_spin_lock_irqsave(&rq->lock, flags);
2919 2920
		if (rq->curr->sched_class->post_schedule)
			rq->curr->sched_class->post_schedule(rq);
2921
		raw_spin_unlock_irqrestore(&rq->lock, flags);
2922 2923 2924 2925 2926 2927

		rq->post_schedule = 0;
	}
}

#else
2928

2929 2930 2931 2932 2933 2934
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
2935 2936
}

2937 2938
#endif

L
Linus Torvalds 已提交
2939 2940 2941 2942
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2943
asmlinkage void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2944 2945
	__releases(rq->lock)
{
2946 2947
	struct rq *rq = this_rq();

2948
	finish_task_switch(rq, prev);
2949

2950 2951 2952 2953 2954
	/*
	 * FIXME: do we need to worry about rq being invalidated by the
	 * task_switch?
	 */
	post_schedule(rq);
2955

2956 2957 2958 2959
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
L
Linus Torvalds 已提交
2960
	if (current->set_child_tid)
2961
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2962 2963 2964 2965 2966 2967
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
I
Ingo Molnar 已提交
2968
static inline void
2969
context_switch(struct rq *rq, struct task_struct *prev,
2970
	       struct task_struct *next)
L
Linus Torvalds 已提交
2971
{
I
Ingo Molnar 已提交
2972
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2973

2974
	prepare_task_switch(rq, prev, next);
2975

I
Ingo Molnar 已提交
2976 2977
	mm = next->mm;
	oldmm = prev->active_mm;
2978 2979 2980 2981 2982
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2983
	arch_start_context_switch(prev);
2984

2985
	if (!mm) {
L
Linus Torvalds 已提交
2986 2987 2988 2989 2990 2991
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2992
	if (!prev->mm) {
L
Linus Torvalds 已提交
2993 2994 2995
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2996 2997 2998 2999 3000 3001 3002
	/*
	 * 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
3003
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
3004
#endif
L
Linus Torvalds 已提交
3005 3006 3007 3008

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

I
Ingo Molnar 已提交
3009 3010 3011 3012 3013 3014 3015
	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
3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032
}

/*
 * 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;
3033
}
L
Linus Torvalds 已提交
3034 3035

unsigned long nr_uninterruptible(void)
3036
{
L
Linus Torvalds 已提交
3037
	unsigned long i, sum = 0;
3038

3039
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
3040
		sum += cpu_rq(i)->nr_uninterruptible;
3041 3042

	/*
L
Linus Torvalds 已提交
3043 3044
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
3045
	 */
L
Linus Torvalds 已提交
3046 3047
	if (unlikely((long)sum < 0))
		sum = 0;
3048

L
Linus Torvalds 已提交
3049
	return sum;
3050 3051
}

L
Linus Torvalds 已提交
3052
unsigned long long nr_context_switches(void)
3053
{
3054 3055
	int i;
	unsigned long long sum = 0;
3056

3057
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
3058
		sum += cpu_rq(i)->nr_switches;
3059

L
Linus Torvalds 已提交
3060 3061
	return sum;
}
3062

L
Linus Torvalds 已提交
3063 3064 3065
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
3066

3067
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
3068
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
3069

L
Linus Torvalds 已提交
3070 3071
	return sum;
}
3072

3073
unsigned long nr_iowait_cpu(int cpu)
3074
{
3075
	struct rq *this = cpu_rq(cpu);
3076 3077
	return atomic_read(&this->nr_iowait);
}
3078

3079 3080 3081 3082 3083
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
3084

3085

3086 3087 3088 3089 3090
/* 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);
3091

3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106
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;
}

3107 3108 3109 3110 3111 3112 3113 3114 3115
static unsigned long
calc_load(unsigned long load, unsigned long exp, unsigned long active)
{
	load *= exp;
	load += active * (FIXED_1 - exp);
	load += 1UL << (FSHIFT - 1);
	return load >> FSHIFT;
}

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 3141 3142 3143 3144
#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;
}
3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266

/**
 * fixed_power_int - compute: x^n, in O(log n) time
 *
 * @x:         base of the power
 * @frac_bits: fractional bits of @x
 * @n:         power to raise @x to.
 *
 * By exploiting the relation between the definition of the natural power
 * function: x^n := x*x*...*x (x multiplied by itself for n times), and
 * the binary encoding of numbers used by computers: n := \Sum n_i * 2^i,
 * (where: n_i \elem {0, 1}, the binary vector representing n),
 * we find: x^n := x^(\Sum n_i * 2^i) := \Prod x^(n_i * 2^i), which is
 * of course trivially computable in O(log_2 n), the length of our binary
 * vector.
 */
static unsigned long
fixed_power_int(unsigned long x, unsigned int frac_bits, unsigned int n)
{
	unsigned long result = 1UL << frac_bits;

	if (n) for (;;) {
		if (n & 1) {
			result *= x;
			result += 1UL << (frac_bits - 1);
			result >>= frac_bits;
		}
		n >>= 1;
		if (!n)
			break;
		x *= x;
		x += 1UL << (frac_bits - 1);
		x >>= frac_bits;
	}

	return result;
}

/*
 * a1 = a0 * e + a * (1 - e)
 *
 * a2 = a1 * e + a * (1 - e)
 *    = (a0 * e + a * (1 - e)) * e + a * (1 - e)
 *    = a0 * e^2 + a * (1 - e) * (1 + e)
 *
 * a3 = a2 * e + a * (1 - e)
 *    = (a0 * e^2 + a * (1 - e) * (1 + e)) * e + a * (1 - e)
 *    = a0 * e^3 + a * (1 - e) * (1 + e + e^2)
 *
 *  ...
 *
 * an = a0 * e^n + a * (1 - e) * (1 + e + ... + e^n-1) [1]
 *    = a0 * e^n + a * (1 - e) * (1 - e^n)/(1 - e)
 *    = a0 * e^n + a * (1 - e^n)
 *
 * [1] application of the geometric series:
 *
 *              n         1 - x^(n+1)
 *     S_n := \Sum x^i = -------------
 *             i=0          1 - x
 */
static unsigned long
calc_load_n(unsigned long load, unsigned long exp,
	    unsigned long active, unsigned int n)
{

	return calc_load(load, fixed_power_int(exp, FSHIFT, n), active);
}

/*
 * NO_HZ can leave us missing all per-cpu ticks calling
 * calc_load_account_active(), but since an idle CPU folds its delta into
 * calc_load_tasks_idle per calc_load_account_idle(), all we need to do is fold
 * in the pending idle delta if our idle period crossed a load cycle boundary.
 *
 * Once we've updated the global active value, we need to apply the exponential
 * weights adjusted to the number of cycles missed.
 */
static void calc_global_nohz(unsigned long ticks)
{
	long delta, active, n;

	if (time_before(jiffies, calc_load_update))
		return;

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

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

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

		avenrun[0] = calc_load_n(avenrun[0], EXP_1, active, n);
		avenrun[1] = calc_load_n(avenrun[1], EXP_5, active, n);
		avenrun[2] = calc_load_n(avenrun[2], EXP_15, active, n);

		calc_load_update += n * LOAD_FREQ;
	}

	/*
	 * Its possible the remainder of the above division also crosses
	 * a LOAD_FREQ period, the regular check in calc_global_load()
	 * which comes after this will take care of that.
	 *
	 * Consider us being 11 ticks before a cycle completion, and us
	 * sleeping for 4*LOAD_FREQ + 22 ticks, then the above code will
	 * age us 4 cycles, and the test in calc_global_load() will
	 * pick up the final one.
	 */
}
3267 3268 3269 3270 3271 3272 3273 3274 3275
#else
static void calc_load_account_idle(struct rq *this_rq)
{
}

static inline long calc_load_fold_idle(void)
{
	return 0;
}
3276 3277 3278 3279

static void calc_global_nohz(unsigned long ticks)
{
}
3280 3281
#endif

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294
/**
 * 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;
3295 3296 3297
}

/*
3298 3299
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
3300
 */
3301
void calc_global_load(unsigned long ticks)
3302
{
3303
	long active;
L
Linus Torvalds 已提交
3304

3305 3306 3307
	calc_global_nohz(ticks);

	if (time_before(jiffies, calc_load_update + 10))
3308
		return;
L
Linus Torvalds 已提交
3309

3310 3311
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
L
Linus Torvalds 已提交
3312

3313 3314 3315
	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
3316

3317 3318
	calc_load_update += LOAD_FREQ;
}
L
Linus Torvalds 已提交
3319

3320
/*
3321 3322
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
3323 3324 3325
 */
static void calc_load_account_active(struct rq *this_rq)
{
3326
	long delta;
3327

3328 3329
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
3330

3331 3332 3333
	delta  = calc_load_fold_active(this_rq);
	delta += calc_load_fold_idle();
	if (delta)
3334
		atomic_long_add(delta, &calc_load_tasks);
3335 3336

	this_rq->calc_load_update += LOAD_FREQ;
3337 3338
}

3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405
/*
 * 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;
}

3406
/*
I
Ingo Molnar 已提交
3407
 * Update rq->cpu_load[] statistics. This function is usually called every
3408 3409
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
3410
 */
I
Ingo Molnar 已提交
3411
static void update_cpu_load(struct rq *this_rq)
3412
{
3413
	unsigned long this_load = this_rq->load.weight;
3414 3415
	unsigned long curr_jiffies = jiffies;
	unsigned long pending_updates;
I
Ingo Molnar 已提交
3416
	int i, scale;
3417

I
Ingo Molnar 已提交
3418
	this_rq->nr_load_updates++;
3419

3420 3421 3422 3423 3424 3425 3426
	/* 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
3427
	/* Update our load: */
3428 3429
	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
3430
		unsigned long old_load, new_load;
3431

I
Ingo Molnar 已提交
3432
		/* scale is effectively 1 << i now, and >> i divides by scale */
3433

I
Ingo Molnar 已提交
3434
		old_load = this_rq->cpu_load[i];
3435
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
I
Ingo Molnar 已提交
3436
		new_load = this_load;
I
Ingo Molnar 已提交
3437 3438 3439 3440 3441 3442
		/*
		 * 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)
3443 3444 3445
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
I
Ingo Molnar 已提交
3446
	}
3447 3448

	sched_avg_update(this_rq);
3449 3450 3451 3452 3453
}

static void update_cpu_load_active(struct rq *this_rq)
{
	update_cpu_load(this_rq);
3454

3455
	calc_load_account_active(this_rq);
3456 3457
}

I
Ingo Molnar 已提交
3458
#ifdef CONFIG_SMP
3459

3460
/*
P
Peter Zijlstra 已提交
3461 3462
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
3463
 */
P
Peter Zijlstra 已提交
3464
void sched_exec(void)
3465
{
P
Peter Zijlstra 已提交
3466
	struct task_struct *p = current;
L
Linus Torvalds 已提交
3467
	unsigned long flags;
3468
	struct rq *rq;
3469
	int dest_cpu;
3470

L
Linus Torvalds 已提交
3471
	rq = task_rq_lock(p, &flags);
3472
	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
3473 3474
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
3475

3476
	/*
P
Peter Zijlstra 已提交
3477
	 * select_task_rq() can race against ->cpus_allowed
3478
	 */
3479
	if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed) &&
3480
	    likely(cpu_active(dest_cpu)) && need_migrate_task(p)) {
3481
		struct migration_arg arg = { p, dest_cpu };
3482

3483
		task_rq_unlock(rq, p, &flags);
3484
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3485 3486
		return;
	}
3487
unlock:
3488
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
3489
}
I
Ingo Molnar 已提交
3490

L
Linus Torvalds 已提交
3491 3492 3493 3494 3495 3496 3497
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);

EXPORT_PER_CPU_SYMBOL(kstat);

/*
3498
 * Return any ns on the sched_clock that have not yet been accounted in
3499
 * @p in case that task is currently running.
3500 3501
 *
 * Called with task_rq_lock() held on @rq.
L
Linus Torvalds 已提交
3502
 */
3503 3504 3505 3506 3507 3508
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);
3509
		ns = rq->clock_task - p->se.exec_start;
3510 3511 3512 3513 3514 3515 3516
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

3517
unsigned long long task_delta_exec(struct task_struct *p)
L
Linus Torvalds 已提交
3518 3519
{
	unsigned long flags;
3520
	struct rq *rq;
3521
	u64 ns = 0;
3522

3523
	rq = task_rq_lock(p, &flags);
3524
	ns = do_task_delta_exec(p, rq);
3525
	task_rq_unlock(rq, p, &flags);
3526

3527 3528
	return ns;
}
3529

3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542
/*
 * 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);
3543
	task_rq_unlock(rq, p, &flags);
3544 3545 3546

	return ns;
}
3547

3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566
/*
 * 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);
3567
	task_rq_unlock(rq, p, &flags);
3568

L
Linus Torvalds 已提交
3569 3570 3571 3572 3573 3574 3575
	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
3576
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
3577
 */
3578 3579
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
L
Linus Torvalds 已提交
3580 3581 3582 3583
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t tmp;

3584
	/* Add user time to process. */
L
Linus Torvalds 已提交
3585
	p->utime = cputime_add(p->utime, cputime);
3586
	p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
3587
	account_group_user_time(p, cputime);
L
Linus Torvalds 已提交
3588 3589 3590 3591 3592 3593 3594

	/* 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);
3595 3596

	cpuacct_update_stats(p, CPUACCT_STAT_USER, cputime);
3597 3598
	/* Account for user time used */
	acct_update_integrals(p);
L
Linus Torvalds 已提交
3599 3600
}

3601 3602 3603 3604
/*
 * 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
3605
 * @cputime_scaled: cputime scaled by cpu frequency
3606
 */
3607 3608
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
3609 3610 3611 3612 3613 3614
{
	cputime64_t tmp;
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;

	tmp = cputime_to_cputime64(cputime);

3615
	/* Add guest time to process. */
3616
	p->utime = cputime_add(p->utime, cputime);
3617
	p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
3618
	account_group_user_time(p, cputime);
3619 3620
	p->gtime = cputime_add(p->gtime, cputime);

3621
	/* Add guest time to cpustat. */
3622 3623 3624 3625 3626 3627 3628
	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);
	}
3629 3630
}

3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656
/*
 * Account system cpu time to a process and desired cpustat field
 * @p: the process that the cpu time gets accounted to
 * @cputime: the cpu time spent in kernel space since the last update
 * @cputime_scaled: cputime scaled by cpu frequency
 * @target_cputime64: pointer to cpustat field that has to be updated
 */
static inline
void __account_system_time(struct task_struct *p, cputime_t cputime,
			cputime_t cputime_scaled, cputime64_t *target_cputime64)
{
	cputime64_t tmp = cputime_to_cputime64(cputime);

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

	/* Add system time to cpustat. */
	*target_cputime64 = cputime64_add(*target_cputime64, tmp);
	cpuacct_update_stats(p, CPUACCT_STAT_SYSTEM, cputime);

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

L
Linus Torvalds 已提交
3657 3658 3659 3660 3661
/*
 * 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
3662
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
3663 3664
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
3665
			 cputime_t cputime, cputime_t cputime_scaled)
L
Linus Torvalds 已提交
3666 3667
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
3668
	cputime64_t *target_cputime64;
L
Linus Torvalds 已提交
3669

3670
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
3671
		account_guest_time(p, cputime, cputime_scaled);
3672 3673
		return;
	}
3674

L
Linus Torvalds 已提交
3675
	if (hardirq_count() - hardirq_offset)
3676
		target_cputime64 = &cpustat->irq;
3677
	else if (in_serving_softirq())
3678
		target_cputime64 = &cpustat->softirq;
L
Linus Torvalds 已提交
3679
	else
3680
		target_cputime64 = &cpustat->system;
3681

3682
	__account_system_time(p, cputime, cputime_scaled, target_cputime64);
L
Linus Torvalds 已提交
3683 3684
}

3685
/*
L
Linus Torvalds 已提交
3686
 * Account for involuntary wait time.
3687
 * @cputime: the cpu time spent in involuntary wait
3688
 */
3689
void account_steal_time(cputime_t cputime)
3690
{
3691 3692 3693 3694
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t cputime64 = cputime_to_cputime64(cputime);

	cpustat->steal = cputime64_add(cpustat->steal, cputime64);
3695 3696
}

L
Linus Torvalds 已提交
3697
/*
3698 3699
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
L
Linus Torvalds 已提交
3700
 */
3701
void account_idle_time(cputime_t cputime)
L
Linus Torvalds 已提交
3702 3703
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
3704
	cputime64_t cputime64 = cputime_to_cputime64(cputime);
3705
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
3706

3707 3708 3709 3710
	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
3711 3712
}

3713 3714
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
/*
 * Account a tick to a process and cpustat
 * @p: the process that the cpu time gets accounted to
 * @user_tick: is the tick from userspace
 * @rq: the pointer to rq
 *
 * Tick demultiplexing follows the order
 * - pending hardirq update
 * - pending softirq update
 * - user_time
 * - idle_time
 * - system time
 *   - check for guest_time
 *   - else account as system_time
 *
 * Check for hardirq is done both for system and user time as there is
 * no timer going off while we are on hardirq and hence we may never get an
 * opportunity to update it solely in system time.
 * p->stime and friends are only updated on system time and not on irq
 * softirq as those do not count in task exec_runtime any more.
 */
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq)
{
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
	cputime64_t tmp = cputime_to_cputime64(cputime_one_jiffy);
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;

	if (irqtime_account_hi_update()) {
		cpustat->irq = cputime64_add(cpustat->irq, tmp);
	} else if (irqtime_account_si_update()) {
		cpustat->softirq = cputime64_add(cpustat->softirq, tmp);
3748 3749 3750 3751 3752 3753 3754 3755
	} else if (this_cpu_ksoftirqd() == p) {
		/*
		 * ksoftirqd time do not get accounted in cpu_softirq_time.
		 * So, we have to handle it separately here.
		 * Also, p->stime needs to be updated for ksoftirqd.
		 */
		__account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
					&cpustat->softirq);
3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775
	} else if (user_tick) {
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
	} else if (p == rq->idle) {
		account_idle_time(cputime_one_jiffy);
	} else if (p->flags & PF_VCPU) { /* System time or guest time */
		account_guest_time(p, cputime_one_jiffy, one_jiffy_scaled);
	} else {
		__account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
					&cpustat->system);
	}
}

static void irqtime_account_idle_ticks(int ticks)
{
	int i;
	struct rq *rq = this_rq();

	for (i = 0; i < ticks; i++)
		irqtime_account_process_tick(current, 0, rq);
}
3776
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
3777 3778 3779
static void irqtime_account_idle_ticks(int ticks) {}
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq) {}
3780
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
3781 3782 3783 3784 3785 3786 3787 3788

/*
 * 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)
{
3789
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
3790 3791
	struct rq *rq = this_rq();

3792 3793 3794 3795 3796
	if (sched_clock_irqtime) {
		irqtime_account_process_tick(p, user_tick, rq);
		return;
	}

3797
	if (user_tick)
3798
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
3799
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
3800
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
3801 3802
				    one_jiffy_scaled);
	else
3803
		account_idle_time(cputime_one_jiffy);
3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821
}

/*
 * 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)
{
3822 3823 3824 3825 3826 3827

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

3828
	account_idle_time(jiffies_to_cputime(ticks));
L
Linus Torvalds 已提交
3829 3830
}

3831 3832
#endif

3833 3834 3835 3836
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
3837
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3838
{
3839 3840
	*ut = p->utime;
	*st = p->stime;
3841 3842
}

3843
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3844
{
3845 3846 3847 3848 3849 3850
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
3851 3852
}
#else
3853 3854

#ifndef nsecs_to_cputime
3855
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
3856 3857
#endif

3858
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3859
{
3860
	cputime_t rtime, utime = p->utime, total = cputime_add(utime, p->stime);
3861 3862 3863 3864

	/*
	 * Use CFS's precise accounting:
	 */
3865
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
3866 3867

	if (total) {
3868
		u64 temp = rtime;
3869

3870
		temp *= utime;
3871
		do_div(temp, total);
3872 3873 3874
		utime = (cputime_t)temp;
	} else
		utime = rtime;
3875

3876 3877 3878
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
3879
	p->prev_utime = max(p->prev_utime, utime);
3880
	p->prev_stime = max(p->prev_stime, cputime_sub(rtime, p->prev_utime));
3881

3882 3883
	*ut = p->prev_utime;
	*st = p->prev_stime;
3884 3885
}

3886 3887 3888 3889
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3890
{
3891 3892 3893
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
3894

3895
	thread_group_cputime(p, &cputime);
3896

3897 3898
	total = cputime_add(cputime.utime, cputime.stime);
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
3899

3900
	if (total) {
3901
		u64 temp = rtime;
3902

3903
		temp *= cputime.utime;
3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914
		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;
3915 3916 3917
}
#endif

3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928
/*
 * 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
3929
	struct task_struct *curr = rq->curr;
3930 3931

	sched_clock_tick();
I
Ingo Molnar 已提交
3932

3933
	raw_spin_lock(&rq->lock);
3934
	update_rq_clock(rq);
3935
	update_cpu_load_active(rq);
P
Peter Zijlstra 已提交
3936
	curr->sched_class->task_tick(rq, curr, 0);
3937
	raw_spin_unlock(&rq->lock);
3938

3939
	perf_event_task_tick();
3940

3941
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
3942 3943
	rq->idle_at_tick = idle_cpu(cpu);
	trigger_load_balance(rq, cpu);
3944
#endif
L
Linus Torvalds 已提交
3945 3946
}

3947
notrace unsigned long get_parent_ip(unsigned long addr)
3948 3949 3950 3951 3952 3953 3954 3955
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
3956

3957 3958 3959
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

3960
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
3961
{
3962
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3963 3964 3965
	/*
	 * Underflow?
	 */
3966 3967
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3968
#endif
L
Linus Torvalds 已提交
3969
	preempt_count() += val;
3970
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3971 3972 3973
	/*
	 * Spinlock count overflowing soon?
	 */
3974 3975
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3976 3977 3978
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3979 3980 3981
}
EXPORT_SYMBOL(add_preempt_count);

3982
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
3983
{
3984
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3985 3986 3987
	/*
	 * Underflow?
	 */
3988
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3989
		return;
L
Linus Torvalds 已提交
3990 3991 3992
	/*
	 * Is the spinlock portion underflowing?
	 */
3993 3994 3995
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3996
#endif
3997

3998 3999
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
4000 4001 4002 4003 4004 4005 4006
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
4007
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
4008
 */
I
Ingo Molnar 已提交
4009
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
4010
{
4011 4012
	struct pt_regs *regs = get_irq_regs();

P
Peter Zijlstra 已提交
4013 4014
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
4015

I
Ingo Molnar 已提交
4016
	debug_show_held_locks(prev);
4017
	print_modules();
I
Ingo Molnar 已提交
4018 4019
	if (irqs_disabled())
		print_irqtrace_events(prev);
4020 4021 4022 4023 4024

	if (regs)
		show_regs(regs);
	else
		dump_stack();
I
Ingo Molnar 已提交
4025
}
L
Linus Torvalds 已提交
4026

I
Ingo Molnar 已提交
4027 4028 4029 4030 4031
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
4032
	/*
I
Ingo Molnar 已提交
4033
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
4034 4035 4036
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
4037
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
I
Ingo Molnar 已提交
4038 4039
		__schedule_bug(prev);

L
Linus Torvalds 已提交
4040 4041
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

4042
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
4043 4044
#ifdef CONFIG_SCHEDSTATS
	if (unlikely(prev->lock_depth >= 0)) {
4045
		schedstat_inc(this_rq(), rq_sched_info.bkl_count);
4046
		schedstat_inc(prev, sched_info.bkl_count);
I
Ingo Molnar 已提交
4047 4048
	}
#endif
I
Ingo Molnar 已提交
4049 4050
}

P
Peter Zijlstra 已提交
4051
static void put_prev_task(struct rq *rq, struct task_struct *prev)
M
Mike Galbraith 已提交
4052
{
P
Peter Zijlstra 已提交
4053
	if (prev->on_rq)
4054
		update_rq_clock(rq);
P
Peter Zijlstra 已提交
4055
	prev->sched_class->put_prev_task(rq, prev);
M
Mike Galbraith 已提交
4056 4057
}

I
Ingo Molnar 已提交
4058 4059 4060 4061
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
4062
pick_next_task(struct rq *rq)
I
Ingo Molnar 已提交
4063
{
4064
	const struct sched_class *class;
I
Ingo Molnar 已提交
4065
	struct task_struct *p;
L
Linus Torvalds 已提交
4066 4067

	/*
I
Ingo Molnar 已提交
4068 4069
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
4070
	 */
I
Ingo Molnar 已提交
4071
	if (likely(rq->nr_running == rq->cfs.nr_running)) {
4072
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
4073 4074
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
4075 4076
	}

4077
	for_each_class(class) {
4078
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
4079 4080 4081
		if (p)
			return p;
	}
4082 4083

	BUG(); /* the idle class will always have a runnable task */
I
Ingo Molnar 已提交
4084
}
L
Linus Torvalds 已提交
4085

I
Ingo Molnar 已提交
4086 4087 4088
/*
 * schedule() is the main scheduler function.
 */
4089
asmlinkage void __sched schedule(void)
I
Ingo Molnar 已提交
4090 4091
{
	struct task_struct *prev, *next;
4092
	unsigned long *switch_count;
I
Ingo Molnar 已提交
4093
	struct rq *rq;
4094
	int cpu;
I
Ingo Molnar 已提交
4095

4096 4097
need_resched:
	preempt_disable();
I
Ingo Molnar 已提交
4098 4099
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
4100
	rcu_note_context_switch(cpu);
I
Ingo Molnar 已提交
4101 4102 4103
	prev = rq->curr;

	schedule_debug(prev);
L
Linus Torvalds 已提交
4104

4105
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
4106
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
4107

4108
	raw_spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
4109

4110
	switch_count = &prev->nivcsw;
L
Linus Torvalds 已提交
4111
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
T
Tejun Heo 已提交
4112
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
4113
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
4114
		} else {
4115 4116 4117
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

T
Tejun Heo 已提交
4118
			/*
4119 4120 4121
			 * If a worker went to sleep, notify and ask workqueue
			 * whether it wants to wake up a task to maintain
			 * concurrency.
T
Tejun Heo 已提交
4122 4123 4124 4125 4126 4127 4128 4129
			 */
			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);
			}
P
Peter Zijlstra 已提交
4130

4131
			/*
4132 4133
			 * If we are going to sleep and we have plugged IO
			 * queued, make sure to submit it to avoid deadlocks.
4134 4135 4136 4137 4138 4139
			 */
			if (blk_needs_flush_plug(prev)) {
				raw_spin_unlock(&rq->lock);
				blk_flush_plug(prev);
				raw_spin_lock(&rq->lock);
			}
T
Tejun Heo 已提交
4140
		}
I
Ingo Molnar 已提交
4141
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
4142 4143
	}

4144
	pre_schedule(rq, prev);
4145

I
Ingo Molnar 已提交
4146
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
4147 4148
		idle_balance(cpu, rq);

M
Mike Galbraith 已提交
4149
	put_prev_task(rq, prev);
4150
	next = pick_next_task(rq);
4151 4152
	clear_tsk_need_resched(prev);
	rq->skip_clock_update = 0;
L
Linus Torvalds 已提交
4153 4154 4155 4156 4157 4158

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

I
Ingo Molnar 已提交
4159
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
4160
		/*
4161 4162 4163 4164
		 * 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
4165 4166 4167
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
4168
	} else
4169
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
4170

4171
	post_schedule(rq);
L
Linus Torvalds 已提交
4172 4173

	preempt_enable_no_resched();
4174
	if (need_resched())
L
Linus Torvalds 已提交
4175 4176 4177 4178
		goto need_resched;
}
EXPORT_SYMBOL(schedule);

4179
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
4180

4181 4182 4183
static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
{
	bool ret = false;
4184

4185 4186 4187
	rcu_read_lock();
	if (lock->owner != owner)
		goto fail;
4188 4189

	/*
4190 4191 4192 4193
	 * Ensure we emit the owner->on_cpu, dereference _after_ checking
	 * lock->owner still matches owner, if that fails, owner might
	 * point to free()d memory, if it still matches, the rcu_read_lock()
	 * ensures the memory stays valid.
4194
	 */
4195
	barrier();
4196

4197 4198 4199
	ret = owner->on_cpu;
fail:
	rcu_read_unlock();
4200

4201 4202
	return ret;
}
4203

4204 4205 4206 4207 4208 4209 4210 4211
/*
 * Look out! "owner" is an entirely speculative pointer
 * access and not reliable.
 */
int mutex_spin_on_owner(struct mutex *lock, struct task_struct *owner)
{
	if (!sched_feat(OWNER_SPIN))
		return 0;
4212

4213 4214
	while (owner_running(lock, owner)) {
		if (need_resched())
4215 4216
			return 0;

4217
		arch_mutex_cpu_relax();
4218
	}
4219

4220 4221 4222 4223 4224 4225 4226
	/*
	 * If the owner changed to another task there is likely
	 * heavy contention, stop spinning.
	 */
	if (lock->owner)
		return 0;

4227 4228 4229 4230
	return 1;
}
#endif

L
Linus Torvalds 已提交
4231 4232
#ifdef CONFIG_PREEMPT
/*
4233
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
4234
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
4235 4236
 * occur there and call schedule directly.
 */
4237
asmlinkage void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
4238 4239
{
	struct thread_info *ti = current_thread_info();
4240

L
Linus Torvalds 已提交
4241 4242
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
4243
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
4244
	 */
N
Nick Piggin 已提交
4245
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
4246 4247
		return;

4248
	do {
4249
		add_preempt_count_notrace(PREEMPT_ACTIVE);
4250
		schedule();
4251
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4252

4253 4254 4255 4256 4257
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
4258
	} while (need_resched());
L
Linus Torvalds 已提交
4259 4260 4261 4262
}
EXPORT_SYMBOL(preempt_schedule);

/*
4263
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
4264 4265 4266 4267 4268 4269 4270
 * 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();
4271

4272
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
4273 4274
	BUG_ON(ti->preempt_count || !irqs_disabled());

4275 4276 4277 4278 4279 4280
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
		schedule();
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4281

4282 4283 4284 4285 4286
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
4287
	} while (need_resched());
L
Linus Torvalds 已提交
4288 4289 4290 4291
}

#endif /* CONFIG_PREEMPT */

P
Peter Zijlstra 已提交
4292
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
4293
			  void *key)
L
Linus Torvalds 已提交
4294
{
P
Peter Zijlstra 已提交
4295
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
4296 4297 4298 4299
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
4300 4301
 * 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
4302 4303 4304
 * 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
4305
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
4306 4307
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
4308
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
P
Peter Zijlstra 已提交
4309
			int nr_exclusive, int wake_flags, void *key)
L
Linus Torvalds 已提交
4310
{
4311
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
4312

4313
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
4314 4315
		unsigned flags = curr->flags;

P
Peter Zijlstra 已提交
4316
		if (curr->func(curr, mode, wake_flags, key) &&
4317
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
4318 4319 4320 4321 4322 4323 4324 4325 4326
			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
4327
 * @key: is directly passed to the wakeup function
4328 4329 4330
 *
 * 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
4331
 */
4332
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
4333
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345
{
	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.
 */
4346
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode)
L
Linus Torvalds 已提交
4347 4348 4349
{
	__wake_up_common(q, mode, 1, 0, NULL);
}
4350
EXPORT_SYMBOL_GPL(__wake_up_locked);
L
Linus Torvalds 已提交
4351

4352 4353 4354 4355
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}
4356
EXPORT_SYMBOL_GPL(__wake_up_locked_key);
4357

L
Linus Torvalds 已提交
4358
/**
4359
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
4360 4361 4362
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
4363
 * @key: opaque value to be passed to wakeup targets
L
Linus Torvalds 已提交
4364 4365 4366 4367 4368 4369 4370
 *
 * 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.
4371 4372 4373
 *
 * 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
4374
 */
4375 4376
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
4377 4378
{
	unsigned long flags;
P
Peter Zijlstra 已提交
4379
	int wake_flags = WF_SYNC;
L
Linus Torvalds 已提交
4380 4381 4382 4383 4384

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
P
Peter Zijlstra 已提交
4385
		wake_flags = 0;
L
Linus Torvalds 已提交
4386 4387

	spin_lock_irqsave(&q->lock, flags);
P
Peter Zijlstra 已提交
4388
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
L
Linus Torvalds 已提交
4389 4390
	spin_unlock_irqrestore(&q->lock, flags);
}
4391 4392 4393 4394 4395 4396 4397 4398 4399
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
4400 4401
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

4402 4403 4404 4405 4406 4407 4408 4409
/**
 * 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.
4410 4411 4412
 *
 * 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.
4413
 */
4414
void complete(struct completion *x)
L
Linus Torvalds 已提交
4415 4416 4417 4418 4419
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
4420
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
4421 4422 4423 4424
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

4425 4426 4427 4428 4429
/**
 * 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.
4430 4431 4432
 *
 * 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.
4433
 */
4434
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
4435 4436 4437 4438 4439
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
4440
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
4441 4442 4443 4444
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

4445 4446
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
4447 4448 4449 4450
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

C
Changli Gao 已提交
4451
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
L
Linus Torvalds 已提交
4452
		do {
4453
			if (signal_pending_state(state, current)) {
4454 4455
				timeout = -ERESTARTSYS;
				break;
4456 4457
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
4458 4459 4460
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
4461
		} while (!x->done && timeout);
L
Linus Torvalds 已提交
4462
		__remove_wait_queue(&x->wait, &wait);
4463 4464
		if (!x->done)
			return timeout;
L
Linus Torvalds 已提交
4465 4466
	}
	x->done--;
4467
	return timeout ?: 1;
L
Linus Torvalds 已提交
4468 4469
}

4470 4471
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
4472 4473 4474 4475
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
4476
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
4477
	spin_unlock_irq(&x->wait.lock);
4478 4479
	return timeout;
}
L
Linus Torvalds 已提交
4480

4481 4482 4483 4484 4485 4486 4487 4488 4489 4490
/**
 * 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().
 */
4491
void __sched wait_for_completion(struct completion *x)
4492 4493
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4494
}
4495
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
4496

4497 4498 4499 4500 4501 4502 4503 4504 4505
/**
 * 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.
 */
4506
unsigned long __sched
4507
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
4508
{
4509
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4510
}
4511
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
4512

4513 4514 4515 4516 4517 4518 4519
/**
 * 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.
 */
4520
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
4521
{
4522 4523 4524 4525
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
4526
}
4527
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
4528

4529 4530 4531 4532 4533 4534 4535 4536
/**
 * 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.
 */
4537
long __sched
4538 4539
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
4540
{
4541
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
4542
}
4543
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
4544

4545 4546 4547 4548 4549 4550 4551
/**
 * 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
4552 4553 4554 4555 4556 4557 4558 4559 4560
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);

4561 4562 4563 4564 4565 4566 4567 4568 4569
/**
 * 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.
 */
4570
long __sched
4571 4572 4573 4574 4575 4576 4577
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);

4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591
/**
 *	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)
{
4592
	unsigned long flags;
4593 4594
	int ret = 1;

4595
	spin_lock_irqsave(&x->wait.lock, flags);
4596 4597 4598 4599
	if (!x->done)
		ret = 0;
	else
		x->done--;
4600
	spin_unlock_irqrestore(&x->wait.lock, flags);
4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614
	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)
{
4615
	unsigned long flags;
4616 4617
	int ret = 1;

4618
	spin_lock_irqsave(&x->wait.lock, flags);
4619 4620
	if (!x->done)
		ret = 0;
4621
	spin_unlock_irqrestore(&x->wait.lock, flags);
4622 4623 4624 4625
	return ret;
}
EXPORT_SYMBOL(completion_done);

4626 4627
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
4628
{
I
Ingo Molnar 已提交
4629 4630 4631 4632
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
4633

4634
	__set_current_state(state);
L
Linus Torvalds 已提交
4635

4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649
	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
4650 4651 4652
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
4653
long __sched
I
Ingo Molnar 已提交
4654
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
4655
{
4656
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
4657 4658 4659
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
4660
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
4661
{
4662
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
4663 4664 4665
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
4666
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
4667
{
4668
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
4669 4670 4671
}
EXPORT_SYMBOL(sleep_on_timeout);

4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683
#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.
 */
4684
void rt_mutex_setprio(struct task_struct *p, int prio)
4685
{
4686
	int oldprio, on_rq, running;
4687
	struct rq *rq;
4688
	const struct sched_class *prev_class;
4689 4690 4691

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

4692
	rq = __task_rq_lock(p);
4693

4694
	trace_sched_pi_setprio(p, prio);
4695
	oldprio = p->prio;
4696
	prev_class = p->sched_class;
P
Peter Zijlstra 已提交
4697
	on_rq = p->on_rq;
4698
	running = task_current(rq, p);
4699
	if (on_rq)
4700
		dequeue_task(rq, p, 0);
4701 4702
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
4703 4704 4705 4706 4707 4708

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

4709 4710
	p->prio = prio;

4711 4712
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4713
	if (on_rq)
4714
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
4715

P
Peter Zijlstra 已提交
4716
	check_class_changed(rq, p, prev_class, oldprio);
4717
	__task_rq_unlock(rq);
4718 4719 4720 4721
}

#endif

4722
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
4723
{
I
Ingo Molnar 已提交
4724
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
4725
	unsigned long flags;
4726
	struct rq *rq;
L
Linus Torvalds 已提交
4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738

	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
4739
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
4740
	 */
4741
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
4742 4743 4744
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
P
Peter Zijlstra 已提交
4745
	on_rq = p->on_rq;
4746
	if (on_rq)
4747
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
4748 4749

	p->static_prio = NICE_TO_PRIO(nice);
4750
	set_load_weight(p);
4751 4752 4753
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
4754

I
Ingo Molnar 已提交
4755
	if (on_rq) {
4756
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
4757
		/*
4758 4759
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
4760
		 */
4761
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
4762 4763 4764
			resched_task(rq->curr);
	}
out_unlock:
4765
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4766 4767 4768
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
4769 4770 4771 4772 4773
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
4774
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
4775
{
4776 4777
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
4778

4779
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
4780 4781 4782
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
4783 4784 4785 4786 4787 4788 4789 4790 4791
#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.
 */
4792
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
4793
{
4794
	long nice, retval;
L
Linus Torvalds 已提交
4795 4796 4797 4798 4799 4800

	/*
	 * 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
4801 4802
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
4803 4804 4805
	if (increment > 40)
		increment = 40;

4806
	nice = TASK_NICE(current) + increment;
L
Linus Torvalds 已提交
4807 4808 4809 4810 4811
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
4812 4813 4814
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832
	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.
 */
4833
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
4834 4835 4836 4837 4838 4839 4840 4841
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
4842
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
4843 4844 4845
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
4846
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860

/**
 * 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.
 */
4861
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4862 4863 4864 4865 4866 4867 4868 4869
{
	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
4870
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4871
{
4872
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4873 4874 4875
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
4876 4877
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
4878 4879 4880
{
	p->policy = policy;
	p->rt_priority = prio;
4881 4882 4883
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4884 4885 4886 4887
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
4888
	set_load_weight(p);
L
Linus Torvalds 已提交
4889 4890
}

4891 4892 4893 4894 4895 4896 4897 4898 4899 4900
/*
 * 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);
4901 4902 4903 4904 4905
	if (cred->user->user_ns == pcred->user->user_ns)
		match = (cred->euid == pcred->euid ||
			 cred->euid == pcred->uid);
	else
		match = false;
4906 4907 4908 4909
	rcu_read_unlock();
	return match;
}

4910
static int __sched_setscheduler(struct task_struct *p, int policy,
4911
				const struct sched_param *param, bool user)
L
Linus Torvalds 已提交
4912
{
4913
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
4914
	unsigned long flags;
4915
	const struct sched_class *prev_class;
4916
	struct rq *rq;
4917
	int reset_on_fork;
L
Linus Torvalds 已提交
4918

4919 4920
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4921 4922
recheck:
	/* double check policy once rq lock held */
4923 4924
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4925
		policy = oldpolicy = p->policy;
4926 4927 4928 4929 4930 4931 4932 4933 4934 4935
	} 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
4936 4937
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4938 4939
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4940 4941
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
4942
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4943
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4944
		return -EINVAL;
4945
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
4946 4947
		return -EINVAL;

4948 4949 4950
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4951
	if (user && !capable(CAP_SYS_NICE)) {
4952
		if (rt_policy(policy)) {
4953 4954
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4955 4956 4957 4958 4959 4960 4961 4962 4963 4964

			/* 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;
		}
4965

I
Ingo Molnar 已提交
4966
		/*
4967 4968
		 * Treat SCHED_IDLE as nice 20. Only allow a switch to
		 * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
I
Ingo Molnar 已提交
4969
		 */
4970 4971 4972 4973
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
			if (!can_nice(p, TASK_NICE(p)))
				return -EPERM;
		}
4974

4975
		/* can't change other user's priorities */
4976
		if (!check_same_owner(p))
4977
			return -EPERM;
4978 4979 4980 4981

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4982
	}
L
Linus Torvalds 已提交
4983

4984
	if (user) {
4985
		retval = security_task_setscheduler(p);
4986 4987 4988 4989
		if (retval)
			return retval;
	}

4990 4991 4992
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
4993
	 *
L
Lucas De Marchi 已提交
4994
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4995 4996
	 * runqueue lock must be held.
	 */
4997
	rq = task_rq_lock(p, &flags);
4998

4999 5000 5001 5002
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
5003
		task_rq_unlock(rq, p, &flags);
5004 5005 5006
		return -EINVAL;
	}

5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017
	/*
	 * If not changing anything there's no need to proceed further:
	 */
	if (unlikely(policy == p->policy && (!rt_policy(policy) ||
			param->sched_priority == p->rt_priority))) {

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

5018 5019 5020 5021 5022 5023 5024
#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) &&
5025 5026
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
5027
			task_rq_unlock(rq, p, &flags);
5028 5029 5030 5031 5032
			return -EPERM;
		}
	}
#endif

L
Linus Torvalds 已提交
5033 5034 5035
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
5036
		task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
5037 5038
		goto recheck;
	}
P
Peter Zijlstra 已提交
5039
	on_rq = p->on_rq;
5040
	running = task_current(rq, p);
5041
	if (on_rq)
5042
		deactivate_task(rq, p, 0);
5043 5044
	if (running)
		p->sched_class->put_prev_task(rq, p);
5045

5046 5047
	p->sched_reset_on_fork = reset_on_fork;

L
Linus Torvalds 已提交
5048
	oldprio = p->prio;
5049
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
5050
	__setscheduler(rq, p, policy, param->sched_priority);
5051

5052 5053
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
5054
	if (on_rq)
I
Ingo Molnar 已提交
5055
		activate_task(rq, p, 0);
5056

P
Peter Zijlstra 已提交
5057
	check_class_changed(rq, p, prev_class, oldprio);
5058
	task_rq_unlock(rq, p, &flags);
5059

5060 5061
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
5062 5063
	return 0;
}
5064 5065 5066 5067 5068 5069 5070 5071 5072 5073

/**
 * 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,
5074
		       const struct sched_param *param)
5075 5076 5077
{
	return __sched_setscheduler(p, policy, param, true);
}
L
Linus Torvalds 已提交
5078 5079
EXPORT_SYMBOL_GPL(sched_setscheduler);

5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091
/**
 * 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,
5092
			       const struct sched_param *param)
5093 5094 5095 5096
{
	return __sched_setscheduler(p, policy, param, false);
}

I
Ingo Molnar 已提交
5097 5098
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
5099 5100 5101
{
	struct sched_param lparam;
	struct task_struct *p;
5102
	int retval;
L
Linus Torvalds 已提交
5103 5104 5105 5106 5107

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
5108 5109 5110

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
5111
	p = find_process_by_pid(pid);
5112 5113 5114
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
5115

L
Linus Torvalds 已提交
5116 5117 5118 5119 5120 5121 5122 5123 5124
	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.
 */
5125 5126
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
5127
{
5128 5129 5130 5131
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
5132 5133 5134 5135 5136 5137 5138 5139
	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.
 */
5140
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
5141 5142 5143 5144 5145 5146 5147 5148
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
5149
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
5150
{
5151
	struct task_struct *p;
5152
	int retval;
L
Linus Torvalds 已提交
5153 5154

	if (pid < 0)
5155
		return -EINVAL;
L
Linus Torvalds 已提交
5156 5157

	retval = -ESRCH;
5158
	rcu_read_lock();
L
Linus Torvalds 已提交
5159 5160 5161 5162
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
5163 5164
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
5165
	}
5166
	rcu_read_unlock();
L
Linus Torvalds 已提交
5167 5168 5169 5170
	return retval;
}

/**
5171
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
5172 5173 5174
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
5175
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
5176 5177
{
	struct sched_param lp;
5178
	struct task_struct *p;
5179
	int retval;
L
Linus Torvalds 已提交
5180 5181

	if (!param || pid < 0)
5182
		return -EINVAL;
L
Linus Torvalds 已提交
5183

5184
	rcu_read_lock();
L
Linus Torvalds 已提交
5185 5186 5187 5188 5189 5190 5191 5192 5193 5194
	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;
5195
	rcu_read_unlock();
L
Linus Torvalds 已提交
5196 5197 5198 5199 5200 5201 5202 5203 5204

	/*
	 * 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:
5205
	rcu_read_unlock();
L
Linus Torvalds 已提交
5206 5207 5208
	return retval;
}

5209
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
5210
{
5211
	cpumask_var_t cpus_allowed, new_mask;
5212 5213
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
5214

5215
	get_online_cpus();
5216
	rcu_read_lock();
L
Linus Torvalds 已提交
5217 5218 5219

	p = find_process_by_pid(pid);
	if (!p) {
5220
		rcu_read_unlock();
5221
		put_online_cpus();
L
Linus Torvalds 已提交
5222 5223 5224
		return -ESRCH;
	}

5225
	/* Prevent p going away */
L
Linus Torvalds 已提交
5226
	get_task_struct(p);
5227
	rcu_read_unlock();
L
Linus Torvalds 已提交
5228

5229 5230 5231 5232 5233 5234 5235 5236
	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
5237
	retval = -EPERM;
5238
	if (!check_same_owner(p) && !task_ns_capable(p, CAP_SYS_NICE))
L
Linus Torvalds 已提交
5239 5240
		goto out_unlock;

5241
	retval = security_task_setscheduler(p);
5242 5243 5244
	if (retval)
		goto out_unlock;

5245 5246
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
P
Peter Zijlstra 已提交
5247
again:
5248
	retval = set_cpus_allowed_ptr(p, new_mask);
L
Linus Torvalds 已提交
5249

P
Paul Menage 已提交
5250
	if (!retval) {
5251 5252
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
5253 5254 5255 5256 5257
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
5258
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
5259 5260 5261
			goto again;
		}
	}
L
Linus Torvalds 已提交
5262
out_unlock:
5263 5264 5265 5266
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
5267
	put_task_struct(p);
5268
	put_online_cpus();
L
Linus Torvalds 已提交
5269 5270 5271 5272
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
5273
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
5274
{
5275 5276 5277 5278 5279
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
5280 5281 5282 5283 5284 5285 5286 5287 5288
	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
 */
5289 5290
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
5291
{
5292
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
5293 5294
	int retval;

5295 5296
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
5297

5298 5299 5300 5301 5302
	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
5303 5304
}

5305
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
5306
{
5307
	struct task_struct *p;
5308
	unsigned long flags;
L
Linus Torvalds 已提交
5309 5310
	int retval;

5311
	get_online_cpus();
5312
	rcu_read_lock();
L
Linus Torvalds 已提交
5313 5314 5315 5316 5317 5318

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

5319 5320 5321 5322
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

5323
	raw_spin_lock_irqsave(&p->pi_lock, flags);
5324
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
5325
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
5326 5327

out_unlock:
5328
	rcu_read_unlock();
5329
	put_online_cpus();
L
Linus Torvalds 已提交
5330

5331
	return retval;
L
Linus Torvalds 已提交
5332 5333 5334 5335 5336 5337 5338 5339
}

/**
 * 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
 */
5340 5341
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
5342 5343
{
	int ret;
5344
	cpumask_var_t mask;
L
Linus Torvalds 已提交
5345

A
Anton Blanchard 已提交
5346
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
5347 5348
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
5349 5350
		return -EINVAL;

5351 5352
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
5353

5354 5355
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
5356
		size_t retlen = min_t(size_t, len, cpumask_size());
5357 5358

		if (copy_to_user(user_mask_ptr, mask, retlen))
5359 5360
			ret = -EFAULT;
		else
5361
			ret = retlen;
5362 5363
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
5364

5365
	return ret;
L
Linus Torvalds 已提交
5366 5367 5368 5369 5370
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
5371 5372
 * 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
5373
 */
5374
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
5375
{
5376
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
5377

5378
	schedstat_inc(rq, yld_count);
5379
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
5380 5381 5382 5383 5384 5385

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
5386
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
5387
	do_raw_spin_unlock(&rq->lock);
L
Linus Torvalds 已提交
5388 5389 5390 5391 5392 5393 5394
	preempt_enable_no_resched();

	schedule();

	return 0;
}

P
Peter Zijlstra 已提交
5395 5396 5397 5398 5399
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

A
Andrew Morton 已提交
5400
static void __cond_resched(void)
L
Linus Torvalds 已提交
5401
{
5402 5403 5404
	add_preempt_count(PREEMPT_ACTIVE);
	schedule();
	sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
5405 5406
}

5407
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
5408
{
P
Peter Zijlstra 已提交
5409
	if (should_resched()) {
L
Linus Torvalds 已提交
5410 5411 5412 5413 5414
		__cond_resched();
		return 1;
	}
	return 0;
}
5415
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
5416 5417

/*
5418
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
5419 5420
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
5421
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
5422 5423 5424
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
5425
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
5426
{
P
Peter Zijlstra 已提交
5427
	int resched = should_resched();
J
Jan Kara 已提交
5428 5429
	int ret = 0;

5430 5431
	lockdep_assert_held(lock);

N
Nick Piggin 已提交
5432
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
5433
		spin_unlock(lock);
P
Peter Zijlstra 已提交
5434
		if (resched)
N
Nick Piggin 已提交
5435 5436 5437
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
5438
		ret = 1;
L
Linus Torvalds 已提交
5439 5440
		spin_lock(lock);
	}
J
Jan Kara 已提交
5441
	return ret;
L
Linus Torvalds 已提交
5442
}
5443
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
5444

5445
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
5446 5447 5448
{
	BUG_ON(!in_softirq());

P
Peter Zijlstra 已提交
5449
	if (should_resched()) {
5450
		local_bh_enable();
L
Linus Torvalds 已提交
5451 5452 5453 5454 5455 5456
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
5457
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
5458 5459 5460 5461

/**
 * yield - yield the current processor to other threads.
 *
5462
 * This is a shortcut for kernel-space yielding - it marks the
L
Linus Torvalds 已提交
5463 5464 5465 5466 5467 5468 5469 5470 5471
 * thread runnable and calls sys_sched_yield().
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

5472 5473 5474 5475
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
R
Randy Dunlap 已提交
5476 5477
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511
 *
 * It's the caller's job to ensure that the target task struct
 * can't go away on us before we can do any checks.
 *
 * Returns true if we indeed boosted the target task.
 */
bool __sched yield_to(struct task_struct *p, bool preempt)
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
	bool yielded = 0;

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
	double_rq_lock(rq, p_rq);
	while (task_rq(p) != p_rq) {
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
		goto out;

	if (curr->sched_class != p->sched_class)
		goto out;

	if (task_running(p_rq, p) || p->state)
		goto out;

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5512
	if (yielded) {
5513
		schedstat_inc(rq, yld_count);
5514 5515 5516 5517 5518 5519 5520
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
			resched_task(p_rq->curr);
	}
5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532

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

	if (yielded)
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

L
Linus Torvalds 已提交
5533
/*
I
Ingo Molnar 已提交
5534
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5535 5536 5537 5538
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
5539
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
5540

5541
	delayacct_blkio_start();
L
Linus Torvalds 已提交
5542
	atomic_inc(&rq->nr_iowait);
5543
	blk_flush_plug(current);
5544
	current->in_iowait = 1;
L
Linus Torvalds 已提交
5545
	schedule();
5546
	current->in_iowait = 0;
L
Linus Torvalds 已提交
5547
	atomic_dec(&rq->nr_iowait);
5548
	delayacct_blkio_end();
L
Linus Torvalds 已提交
5549 5550 5551 5552 5553
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
5554
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
5555 5556
	long ret;

5557
	delayacct_blkio_start();
L
Linus Torvalds 已提交
5558
	atomic_inc(&rq->nr_iowait);
5559
	blk_flush_plug(current);
5560
	current->in_iowait = 1;
L
Linus Torvalds 已提交
5561
	ret = schedule_timeout(timeout);
5562
	current->in_iowait = 0;
L
Linus Torvalds 已提交
5563
	atomic_dec(&rq->nr_iowait);
5564
	delayacct_blkio_end();
L
Linus Torvalds 已提交
5565 5566 5567 5568 5569 5570 5571 5572 5573 5574
	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.
 */
5575
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5576 5577 5578 5579 5580 5581 5582 5583 5584
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
5585
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5586
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599
		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.
 */
5600
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5601 5602 5603 5604 5605 5606 5607 5608 5609
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
5610
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5611
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624
		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.
 */
5625
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5626
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5627
{
5628
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5629
	unsigned int time_slice;
5630 5631
	unsigned long flags;
	struct rq *rq;
5632
	int retval;
L
Linus Torvalds 已提交
5633 5634 5635
	struct timespec t;

	if (pid < 0)
5636
		return -EINVAL;
L
Linus Torvalds 已提交
5637 5638

	retval = -ESRCH;
5639
	rcu_read_lock();
L
Linus Torvalds 已提交
5640 5641 5642 5643 5644 5645 5646 5647
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5648 5649
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
5650
	task_rq_unlock(rq, p, &flags);
D
Dmitry Adamushko 已提交
5651

5652
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5653
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5654 5655
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5656

L
Linus Torvalds 已提交
5657
out_unlock:
5658
	rcu_read_unlock();
L
Linus Torvalds 已提交
5659 5660 5661
	return retval;
}

5662
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5663

5664
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5665 5666
{
	unsigned long free = 0;
5667
	unsigned state;
L
Linus Torvalds 已提交
5668 5669

	state = p->state ? __ffs(p->state) + 1 : 0;
5670
	printk(KERN_INFO "%-15.15s %c", p->comm,
5671
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
5672
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
5673
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5674
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
5675
	else
P
Peter Zijlstra 已提交
5676
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5677 5678
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5679
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5680
	else
P
Peter Zijlstra 已提交
5681
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5682 5683
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
5684
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5685
#endif
P
Peter Zijlstra 已提交
5686
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5687 5688
		task_pid_nr(p), task_pid_nr(p->real_parent),
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5689

5690
	show_stack(p, NULL);
L
Linus Torvalds 已提交
5691 5692
}

I
Ingo Molnar 已提交
5693
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5694
{
5695
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5696

5697
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5698 5699
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5700
#else
P
Peter Zijlstra 已提交
5701 5702
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5703 5704 5705 5706 5707
#endif
	read_lock(&tasklist_lock);
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5708
		 * console might take a lot of time:
L
Linus Torvalds 已提交
5709 5710
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
5711
		if (!state_filter || (p->state & state_filter))
5712
			sched_show_task(p);
L
Linus Torvalds 已提交
5713 5714
	} while_each_thread(g, p);

5715 5716
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
5717 5718 5719
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
L
Linus Torvalds 已提交
5720
	read_unlock(&tasklist_lock);
I
Ingo Molnar 已提交
5721 5722 5723
	/*
	 * Only show locks if all tasks are dumped:
	 */
5724
	if (!state_filter)
I
Ingo Molnar 已提交
5725
		debug_show_all_locks();
L
Linus Torvalds 已提交
5726 5727
}

I
Ingo Molnar 已提交
5728 5729
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
5730
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5731 5732
}

5733 5734 5735 5736 5737 5738 5739 5740
/**
 * 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.
 */
5741
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5742
{
5743
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5744 5745
	unsigned long flags;

5746
	raw_spin_lock_irqsave(&rq->lock, flags);
5747

I
Ingo Molnar 已提交
5748
	__sched_fork(idle);
5749
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5750 5751
	idle->se.exec_start = sched_clock();

5752
	cpumask_copy(&idle->cpus_allowed, cpumask_of(cpu));
5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763
	/*
	 * We're having a chicken and egg problem, even though we are
	 * holding rq->lock, the cpu isn't yet set to this cpu so the
	 * lockdep check in task_group() will fail.
	 *
	 * Similar case to sched_fork(). / Alternatively we could
	 * use task_rq_lock() here and obtain the other rq->lock.
	 *
	 * Silence PROVE_RCU
	 */
	rcu_read_lock();
I
Ingo Molnar 已提交
5764
	__set_task_cpu(idle, cpu);
5765
	rcu_read_unlock();
L
Linus Torvalds 已提交
5766 5767

	rq->curr = rq->idle = idle;
P
Peter Zijlstra 已提交
5768 5769
#if defined(CONFIG_SMP)
	idle->on_cpu = 1;
5770
#endif
5771
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5772 5773

	/* Set the preempt count _outside_ the spinlocks! */
5774 5775 5776
#if defined(CONFIG_PREEMPT)
	task_thread_info(idle)->preempt_count = (idle->lock_depth >= 0);
#else
A
Al Viro 已提交
5777
	task_thread_info(idle)->preempt_count = 0;
5778
#endif
I
Ingo Molnar 已提交
5779 5780 5781 5782
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5783
	ftrace_graph_init_idle_task(idle, cpu);
L
Linus Torvalds 已提交
5784 5785 5786 5787 5788 5789 5790
}

/*
 * 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
5791
 * always be CPU_BITS_NONE.
L
Linus Torvalds 已提交
5792
 */
5793
cpumask_var_t nohz_cpu_mask;
L
Linus Torvalds 已提交
5794

I
Ingo Molnar 已提交
5795 5796 5797 5798 5799 5800 5801 5802 5803
/*
 * 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:
 */
5804
static int get_update_sysctl_factor(void)
I
Ingo Molnar 已提交
5805
{
5806
	unsigned int cpus = min_t(int, num_online_cpus(), 8);
5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820
	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
5821

5822 5823
	return factor;
}
I
Ingo Molnar 已提交
5824

5825 5826 5827
static void update_sysctl(void)
{
	unsigned int factor = get_update_sysctl_factor();
I
Ingo Molnar 已提交
5828

5829 5830 5831 5832 5833 5834 5835
#define SET_SYSCTL(name) \
	(sysctl_##name = (factor) * normalized_sysctl_##name)
	SET_SYSCTL(sched_min_granularity);
	SET_SYSCTL(sched_latency);
	SET_SYSCTL(sched_wakeup_granularity);
#undef SET_SYSCTL
}
5836

5837 5838 5839
static inline void sched_init_granularity(void)
{
	update_sysctl();
I
Ingo Molnar 已提交
5840 5841
}

L
Linus Torvalds 已提交
5842 5843 5844 5845
#ifdef CONFIG_SMP
/*
 * This is how migration works:
 *
5846 5847 5848 5849 5850 5851
 * 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
5852
 *    it and puts it into the right queue.
5853 5854
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
L
Linus Torvalds 已提交
5855 5856 5857 5858 5859 5860 5861 5862
 */

/*
 * 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
5863
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
5864 5865
 * call is not atomic; no spinlocks may be held.
 */
5866
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
L
Linus Torvalds 已提交
5867 5868
{
	unsigned long flags;
5869
	struct rq *rq;
5870
	unsigned int dest_cpu;
5871
	int ret = 0;
L
Linus Torvalds 已提交
5872

5873
	rq = task_rq_lock(p, &flags);
5874

5875
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
L
Linus Torvalds 已提交
5876 5877 5878 5879
		ret = -EINVAL;
		goto out;
	}

5880
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current &&
5881
		     !cpumask_equal(&p->cpus_allowed, new_mask))) {
5882 5883 5884 5885
		ret = -EINVAL;
		goto out;
	}

5886
	if (p->sched_class->set_cpus_allowed)
5887
		p->sched_class->set_cpus_allowed(p, new_mask);
5888
	else {
5889 5890
		cpumask_copy(&p->cpus_allowed, new_mask);
		p->rt.nr_cpus_allowed = cpumask_weight(new_mask);
5891 5892
	}

L
Linus Torvalds 已提交
5893
	/* Can the task run on the task's current CPU? If so, we're done */
5894
	if (cpumask_test_cpu(task_cpu(p), new_mask))
L
Linus Torvalds 已提交
5895 5896
		goto out;

5897
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
5898
	if (need_migrate_task(p)) {
5899
		struct migration_arg arg = { p, dest_cpu };
L
Linus Torvalds 已提交
5900
		/* Need help from migration thread: drop lock and wait. */
5901
		task_rq_unlock(rq, p, &flags);
5902
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
5903 5904 5905 5906
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
5907
	task_rq_unlock(rq, p, &flags);
5908

L
Linus Torvalds 已提交
5909 5910
	return ret;
}
5911
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
5912 5913

/*
I
Ingo Molnar 已提交
5914
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
5915 5916 5917 5918 5919 5920
 * 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.
5921 5922
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
5923
 */
5924
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
5925
{
5926
	struct rq *rq_dest, *rq_src;
5927
	int ret = 0;
L
Linus Torvalds 已提交
5928

5929
	if (unlikely(!cpu_active(dest_cpu)))
5930
		return ret;
L
Linus Torvalds 已提交
5931 5932 5933 5934

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

5935
	raw_spin_lock(&p->pi_lock);
L
Linus Torvalds 已提交
5936 5937 5938
	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
L
Linus Torvalds 已提交
5939
		goto done;
L
Linus Torvalds 已提交
5940
	/* Affinity changed (again). */
5941
	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
L
Linus Torvalds 已提交
5942
		goto fail;
L
Linus Torvalds 已提交
5943

5944 5945 5946 5947
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
P
Peter Zijlstra 已提交
5948
	if (p->on_rq) {
5949
		deactivate_task(rq_src, p, 0);
5950
		set_task_cpu(p, dest_cpu);
I
Ingo Molnar 已提交
5951
		activate_task(rq_dest, p, 0);
5952
		check_preempt_curr(rq_dest, p, 0);
L
Linus Torvalds 已提交
5953
	}
L
Linus Torvalds 已提交
5954
done:
5955
	ret = 1;
L
Linus Torvalds 已提交
5956
fail:
L
Linus Torvalds 已提交
5957
	double_rq_unlock(rq_src, rq_dest);
5958
	raw_spin_unlock(&p->pi_lock);
5959
	return ret;
L
Linus Torvalds 已提交
5960 5961 5962
}

/*
5963 5964 5965
 * 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
5966
 */
5967
static int migration_cpu_stop(void *data)
L
Linus Torvalds 已提交
5968
{
5969
	struct migration_arg *arg = data;
5970

5971 5972 5973 5974
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
5975
	local_irq_disable();
5976
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
5977
	local_irq_enable();
L
Linus Torvalds 已提交
5978
	return 0;
5979 5980
}

L
Linus Torvalds 已提交
5981
#ifdef CONFIG_HOTPLUG_CPU
5982

5983
/*
5984 5985
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
5986
 */
5987
void idle_task_exit(void)
L
Linus Torvalds 已提交
5988
{
5989
	struct mm_struct *mm = current->active_mm;
5990

5991
	BUG_ON(cpu_online(smp_processor_id()));
5992

5993 5994 5995
	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
L
Linus Torvalds 已提交
5996 5997 5998 5999 6000 6001 6002 6003 6004
}

/*
 * 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:
 */
6005
static void migrate_nr_uninterruptible(struct rq *rq_src)
L
Linus Torvalds 已提交
6006
{
6007
	struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
L
Linus Torvalds 已提交
6008 6009 6010 6011 6012

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

I
Ingo Molnar 已提交
6013
/*
6014
 * remove the tasks which were accounted by rq from calc_load_tasks.
L
Linus Torvalds 已提交
6015
 */
6016
static void calc_global_load_remove(struct rq *rq)
L
Linus Torvalds 已提交
6017
{
6018 6019
	atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
	rq->calc_load_active = 0;
L
Linus Torvalds 已提交
6020 6021
}

6022
/*
6023 6024 6025 6026 6027 6028
 * Migrate all tasks from the rq, sleeping tasks will be migrated by
 * try_to_wake_up()->select_task_rq().
 *
 * Called with rq->lock held even though we'er in stop_machine() and
 * there's no concurrency possible, we hold the required locks anyway
 * because of lock validation efforts.
L
Linus Torvalds 已提交
6029
 */
6030
static void migrate_tasks(unsigned int dead_cpu)
L
Linus Torvalds 已提交
6031
{
6032
	struct rq *rq = cpu_rq(dead_cpu);
6033 6034
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
L
Linus Torvalds 已提交
6035 6036

	/*
6037 6038 6039 6040 6041 6042 6043
	 * Fudge the rq selection such that the below task selection loop
	 * doesn't get stuck on the currently eligible stop task.
	 *
	 * We're currently inside stop_machine() and the rq is either stuck
	 * in the stop_machine_cpu_stop() loop, or we're executing this code,
	 * either way we should never end up calling schedule() until we're
	 * done here.
L
Linus Torvalds 已提交
6044
	 */
6045
	rq->stop = NULL;
6046

I
Ingo Molnar 已提交
6047
	for ( ; ; ) {
6048 6049 6050 6051 6052
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
6053
			break;
6054

6055
		next = pick_next_task(rq);
6056
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
6057
		next->sched_class->put_prev_task(rq, next);
6058

6059 6060 6061 6062 6063 6064 6065
		/* Find suitable destination for @next, with force if needed. */
		dest_cpu = select_fallback_rq(dead_cpu, next);
		raw_spin_unlock(&rq->lock);

		__migrate_task(next, dead_cpu, dest_cpu);

		raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
6066
	}
6067

6068
	rq->stop = stop;
6069
}
6070

L
Linus Torvalds 已提交
6071 6072
#endif /* CONFIG_HOTPLUG_CPU */

6073 6074 6075
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
6076 6077
	{
		.procname	= "sched_domain",
6078
		.mode		= 0555,
6079
	},
6080
	{}
6081 6082 6083
};

static struct ctl_table sd_ctl_root[] = {
6084 6085
	{
		.procname	= "kernel",
6086
		.mode		= 0555,
6087 6088
		.child		= sd_ctl_dir,
	},
6089
	{}
6090 6091 6092 6093 6094
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
6095
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
6096 6097 6098 6099

	return entry;
}

6100 6101
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
6102
	struct ctl_table *entry;
6103

6104 6105 6106
	/*
	 * 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
6107
	 * will always be set. In the lowest directory the names are
6108 6109 6110
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
6111 6112
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
6113 6114 6115
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
6116 6117 6118 6119 6120

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

6121
static void
6122
set_table_entry(struct ctl_table *entry,
6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135
		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)
{
6136
	struct ctl_table *table = sd_alloc_ctl_entry(13);
6137

6138 6139 6140
	if (table == NULL)
		return NULL;

6141
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
6142
		sizeof(long), 0644, proc_doulongvec_minmax);
6143
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
6144
		sizeof(long), 0644, proc_doulongvec_minmax);
6145
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
6146
		sizeof(int), 0644, proc_dointvec_minmax);
6147
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
6148
		sizeof(int), 0644, proc_dointvec_minmax);
6149
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
6150
		sizeof(int), 0644, proc_dointvec_minmax);
6151
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
6152
		sizeof(int), 0644, proc_dointvec_minmax);
6153
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
6154
		sizeof(int), 0644, proc_dointvec_minmax);
6155
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
6156
		sizeof(int), 0644, proc_dointvec_minmax);
6157
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
6158
		sizeof(int), 0644, proc_dointvec_minmax);
6159
	set_table_entry(&table[9], "cache_nice_tries",
6160 6161
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
6162
	set_table_entry(&table[10], "flags", &sd->flags,
6163
		sizeof(int), 0644, proc_dointvec_minmax);
6164 6165 6166
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
6167 6168 6169 6170

	return table;
}

6171
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
6172 6173 6174 6175 6176 6177 6178 6179 6180
{
	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);
6181 6182
	if (table == NULL)
		return NULL;
6183 6184 6185 6186 6187

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
6188
		entry->mode = 0555;
6189 6190 6191 6192 6193 6194 6195 6196
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
6197
static void register_sched_domain_sysctl(void)
6198
{
6199
	int i, cpu_num = num_possible_cpus();
6200 6201 6202
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

6203 6204 6205
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

6206 6207 6208
	if (entry == NULL)
		return;

6209
	for_each_possible_cpu(i) {
6210 6211
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
6212
		entry->mode = 0555;
6213
		entry->child = sd_alloc_ctl_cpu_table(i);
6214
		entry++;
6215
	}
6216 6217

	WARN_ON(sd_sysctl_header);
6218 6219
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
6220

6221
/* may be called multiple times per register */
6222 6223
static void unregister_sched_domain_sysctl(void)
{
6224 6225
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
6226
	sd_sysctl_header = NULL;
6227 6228
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
6229
}
6230
#else
6231 6232 6233 6234
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
6235 6236 6237 6238
{
}
#endif

6239 6240 6241 6242 6243
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

6244
		cpumask_set_cpu(rq->cpu, rq->rd->online);
6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263
		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);
		}

6264
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
6265 6266 6267 6268
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
6269 6270 6271 6272
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
6273 6274
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
6275
{
6276
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
6277
	unsigned long flags;
6278
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
6279

6280
	switch (action & ~CPU_TASKS_FROZEN) {
6281

L
Linus Torvalds 已提交
6282
	case CPU_UP_PREPARE:
6283
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
6284
		break;
6285

L
Linus Torvalds 已提交
6286
	case CPU_ONLINE:
6287
		/* Update our root-domain */
6288
		raw_spin_lock_irqsave(&rq->lock, flags);
6289
		if (rq->rd) {
6290
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
6291 6292

			set_rq_online(rq);
6293
		}
6294
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
6295
		break;
6296

L
Linus Torvalds 已提交
6297
#ifdef CONFIG_HOTPLUG_CPU
6298
	case CPU_DYING:
G
Gregory Haskins 已提交
6299
		/* Update our root-domain */
6300
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
6301
		if (rq->rd) {
6302
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
6303
			set_rq_offline(rq);
G
Gregory Haskins 已提交
6304
		}
6305 6306
		migrate_tasks(cpu);
		BUG_ON(rq->nr_running != 1); /* the migration thread */
6307
		raw_spin_unlock_irqrestore(&rq->lock, flags);
6308 6309 6310

		migrate_nr_uninterruptible(rq);
		calc_global_load_remove(rq);
G
Gregory Haskins 已提交
6311
		break;
L
Linus Torvalds 已提交
6312 6313
#endif
	}
6314 6315 6316

	update_max_interval();

L
Linus Torvalds 已提交
6317 6318 6319
	return NOTIFY_OK;
}

6320 6321 6322
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
6323
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
6324
 */
6325
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
6326
	.notifier_call = migration_call,
6327
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
6328 6329
};

6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343 6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354
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;
	}
}

6355
static int __init migration_init(void)
L
Linus Torvalds 已提交
6356 6357
{
	void *cpu = (void *)(long)smp_processor_id();
6358
	int err;
6359

6360
	/* Initialize migration for the boot CPU */
6361 6362
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
6363 6364
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
6365

6366 6367 6368 6369
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

6370
	return 0;
L
Linus Torvalds 已提交
6371
}
6372
early_initcall(migration_init);
L
Linus Torvalds 已提交
6373 6374 6375
#endif

#ifdef CONFIG_SMP
6376

6377
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
6378

6379 6380 6381 6382 6383 6384 6385 6386 6387 6388
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);

6389
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
6390
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
6391
{
I
Ingo Molnar 已提交
6392
	struct sched_group *group = sd->groups;
6393
	char str[256];
L
Linus Torvalds 已提交
6394

R
Rusty Russell 已提交
6395
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
6396
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
6397 6398 6399 6400

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
6401
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
6402
		if (sd->parent)
P
Peter Zijlstra 已提交
6403 6404
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
6405
		return -1;
N
Nick Piggin 已提交
6406 6407
	}

P
Peter Zijlstra 已提交
6408
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
I
Ingo Molnar 已提交
6409

6410
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
6411 6412
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
6413
	}
6414
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
6415 6416
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
6417
	}
L
Linus Torvalds 已提交
6418

I
Ingo Molnar 已提交
6419
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
6420
	do {
I
Ingo Molnar 已提交
6421
		if (!group) {
P
Peter Zijlstra 已提交
6422 6423
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
6424 6425 6426
			break;
		}

6427
		if (!group->cpu_power) {
P
Peter Zijlstra 已提交
6428 6429 6430
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
I
Ingo Molnar 已提交
6431 6432
			break;
		}
L
Linus Torvalds 已提交
6433

6434
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
6435 6436
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
6437 6438
			break;
		}
L
Linus Torvalds 已提交
6439

6440
		if (cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
6441 6442
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
6443 6444
			break;
		}
L
Linus Torvalds 已提交
6445

6446
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
6447

R
Rusty Russell 已提交
6448
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
6449

P
Peter Zijlstra 已提交
6450
		printk(KERN_CONT " %s", str);
6451
		if (group->cpu_power != SCHED_LOAD_SCALE) {
P
Peter Zijlstra 已提交
6452 6453
			printk(KERN_CONT " (cpu_power = %d)",
				group->cpu_power);
6454
		}
L
Linus Torvalds 已提交
6455

I
Ingo Molnar 已提交
6456 6457
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
6458
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
6459

6460
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
6461
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
6462

6463 6464
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
6465 6466
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
6467 6468
	return 0;
}
L
Linus Torvalds 已提交
6469

I
Ingo Molnar 已提交
6470 6471
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
6472
	cpumask_var_t groupmask;
I
Ingo Molnar 已提交
6473
	int level = 0;
L
Linus Torvalds 已提交
6474

6475 6476 6477
	if (!sched_domain_debug_enabled)
		return;

I
Ingo Molnar 已提交
6478 6479 6480 6481
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
6482

I
Ingo Molnar 已提交
6483 6484
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

6485
	if (!alloc_cpumask_var(&groupmask, GFP_KERNEL)) {
6486 6487 6488 6489
		printk(KERN_DEBUG "Cannot load-balance (out of memory)\n");
		return;
	}

I
Ingo Molnar 已提交
6490
	for (;;) {
6491
		if (sched_domain_debug_one(sd, cpu, level, groupmask))
I
Ingo Molnar 已提交
6492
			break;
L
Linus Torvalds 已提交
6493 6494
		level++;
		sd = sd->parent;
6495
		if (!sd)
I
Ingo Molnar 已提交
6496 6497
			break;
	}
6498
	free_cpumask_var(groupmask);
L
Linus Torvalds 已提交
6499
}
6500
#else /* !CONFIG_SCHED_DEBUG */
6501
# define sched_domain_debug(sd, cpu) do { } while (0)
6502
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
6503

6504
static int sd_degenerate(struct sched_domain *sd)
6505
{
6506
	if (cpumask_weight(sched_domain_span(sd)) == 1)
6507 6508 6509 6510 6511 6512
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
6513 6514 6515
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
6516 6517 6518 6519 6520
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
6521
	if (sd->flags & (SD_WAKE_AFFINE))
6522 6523 6524 6525 6526
		return 0;

	return 1;
}

6527 6528
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
6529 6530 6531 6532 6533 6534
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

6535
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
6536 6537 6538 6539 6540 6541 6542
		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 |
6543 6544 6545
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
6546 6547
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
6548 6549 6550 6551 6552 6553 6554
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

6555 6556
static void free_rootdomain(struct root_domain *rd)
{
6557 6558
	synchronize_sched();

6559 6560
	cpupri_cleanup(&rd->cpupri);

6561 6562 6563 6564 6565 6566
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
6567 6568
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
6569
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
6570 6571
	unsigned long flags;

6572
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
6573 6574

	if (rq->rd) {
I
Ingo Molnar 已提交
6575
		old_rd = rq->rd;
G
Gregory Haskins 已提交
6576

6577
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
6578
			set_rq_offline(rq);
G
Gregory Haskins 已提交
6579

6580
		cpumask_clear_cpu(rq->cpu, old_rd->span);
6581

I
Ingo Molnar 已提交
6582 6583 6584 6585 6586 6587 6588
		/*
		 * 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
6589 6590 6591 6592 6593
	}

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

6594
	cpumask_set_cpu(rq->cpu, rd->span);
6595
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
6596
		set_rq_online(rq);
G
Gregory Haskins 已提交
6597

6598
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
6599 6600 6601

	if (old_rd)
		free_rootdomain(old_rd);
G
Gregory Haskins 已提交
6602 6603
}

6604
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
6605 6606 6607
{
	memset(rd, 0, sizeof(*rd));

6608
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
6609
		goto out;
6610
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
6611
		goto free_span;
6612
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
6613
		goto free_online;
6614

6615
	if (cpupri_init(&rd->cpupri) != 0)
6616
		goto free_rto_mask;
6617
	return 0;
6618

6619 6620
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
6621 6622 6623 6624
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
6625
out:
6626
	return -ENOMEM;
G
Gregory Haskins 已提交
6627 6628 6629 6630
}

static void init_defrootdomain(void)
{
6631
	init_rootdomain(&def_root_domain);
6632

G
Gregory Haskins 已提交
6633 6634 6635
	atomic_set(&def_root_domain.refcount, 1);
}

6636
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
6637 6638 6639 6640 6641 6642 6643
{
	struct root_domain *rd;

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

6644
	if (init_rootdomain(rd) != 0) {
6645 6646 6647
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
6648 6649 6650 6651

	return rd;
}

L
Linus Torvalds 已提交
6652
/*
I
Ingo Molnar 已提交
6653
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
6654 6655
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
6656 6657
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
6658
{
6659
	struct rq *rq = cpu_rq(cpu);
6660 6661
	struct sched_domain *tmp;

6662 6663 6664
	for (tmp = sd; tmp; tmp = tmp->parent)
		tmp->span_weight = cpumask_weight(sched_domain_span(tmp));

6665
	/* Remove the sched domains which do not contribute to scheduling. */
6666
	for (tmp = sd; tmp; ) {
6667 6668 6669
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
6670

6671
		if (sd_parent_degenerate(tmp, parent)) {
6672
			tmp->parent = parent->parent;
6673 6674
			if (parent->parent)
				parent->parent->child = tmp;
6675 6676
		} else
			tmp = tmp->parent;
6677 6678
	}

6679
	if (sd && sd_degenerate(sd)) {
6680
		sd = sd->parent;
6681 6682 6683
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
6684 6685 6686

	sched_domain_debug(sd, cpu);

G
Gregory Haskins 已提交
6687
	rq_attach_root(rq, rd);
N
Nick Piggin 已提交
6688
	rcu_assign_pointer(rq->sd, sd);
L
Linus Torvalds 已提交
6689 6690 6691
}

/* cpus with isolated domains */
6692
static cpumask_var_t cpu_isolated_map;
L
Linus Torvalds 已提交
6693 6694 6695 6696

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
6697
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
6698
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
6699 6700 6701
	return 1;
}

I
Ingo Molnar 已提交
6702
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
6703 6704

/*
6705 6706
 * 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
6707 6708
 * 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
6709 6710 6711 6712 6713
 *
 * 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.
 */
6714
static void
6715 6716 6717
init_sched_build_groups(const struct cpumask *span,
			const struct cpumask *cpu_map,
			int (*group_fn)(int cpu, const struct cpumask *cpu_map,
6718
					struct sched_group **sg,
6719 6720
					struct cpumask *tmpmask),
			struct cpumask *covered, struct cpumask *tmpmask)
L
Linus Torvalds 已提交
6721 6722 6723 6724
{
	struct sched_group *first = NULL, *last = NULL;
	int i;

6725
	cpumask_clear(covered);
6726

6727
	for_each_cpu(i, span) {
6728
		struct sched_group *sg;
6729
		int group = group_fn(i, cpu_map, &sg, tmpmask);
L
Linus Torvalds 已提交
6730 6731
		int j;

6732
		if (cpumask_test_cpu(i, covered))
L
Linus Torvalds 已提交
6733 6734
			continue;

6735
		cpumask_clear(sched_group_cpus(sg));
6736
		sg->cpu_power = 0;
L
Linus Torvalds 已提交
6737

6738
		for_each_cpu(j, span) {
6739
			if (group_fn(j, cpu_map, NULL, tmpmask) != group)
L
Linus Torvalds 已提交
6740 6741
				continue;

6742
			cpumask_set_cpu(j, covered);
6743
			cpumask_set_cpu(j, sched_group_cpus(sg));
L
Linus Torvalds 已提交
6744 6745 6746 6747 6748 6749 6750 6751 6752 6753
		}
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
}

6754
#define SD_NODES_PER_DOMAIN 16
L
Linus Torvalds 已提交
6755

6756
#ifdef CONFIG_NUMA
6757

6758 6759 6760 6761 6762
/**
 * 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
6763
 * Find the next node to include in a given scheduling domain. Simply
6764 6765 6766 6767
 * finds the closest node not already in the @used_nodes map.
 *
 * Should use nodemask_t.
 */
6768
static int find_next_best_node(int node, nodemask_t *used_nodes)
6769 6770 6771 6772 6773
{
	int i, n, val, min_val, best_node = 0;

	min_val = INT_MAX;

6774
	for (i = 0; i < nr_node_ids; i++) {
6775
		/* Start at @node */
6776
		n = (node + i) % nr_node_ids;
6777 6778 6779 6780 6781

		if (!nr_cpus_node(n))
			continue;

		/* Skip already used nodes */
6782
		if (node_isset(n, *used_nodes))
6783 6784 6785 6786 6787 6788 6789 6790 6791 6792 6793
			continue;

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

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

6794
	node_set(best_node, *used_nodes);
6795 6796 6797 6798 6799 6800
	return best_node;
}

/**
 * sched_domain_node_span - get a cpumask for a node's sched_domain
 * @node: node whose cpumask we're constructing
6801
 * @span: resulting cpumask
6802
 *
I
Ingo Molnar 已提交
6803
 * Given a node, construct a good cpumask for its sched_domain to span. It
6804 6805 6806
 * should be one that prevents unnecessary balancing, but also spreads tasks
 * out optimally.
 */
6807
static void sched_domain_node_span(int node, struct cpumask *span)
6808
{
6809
	nodemask_t used_nodes;
6810
	int i;
6811

6812
	cpumask_clear(span);
6813
	nodes_clear(used_nodes);
6814

6815
	cpumask_or(span, span, cpumask_of_node(node));
6816
	node_set(node, used_nodes);
6817 6818

	for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
6819
		int next_node = find_next_best_node(node, &used_nodes);
6820

6821
		cpumask_or(span, span, cpumask_of_node(next_node));
6822 6823
	}
}
6824
#endif /* CONFIG_NUMA */
6825

6826
int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
6827

6828 6829
/*
 * The cpus mask in sched_group and sched_domain hangs off the end.
6830 6831 6832
 *
 * ( See the the comments in include/linux/sched.h:struct sched_group
 *   and struct sched_domain. )
6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843
 */
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);
};

6844 6845 6846 6847 6848 6849 6850 6851 6852 6853
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;
6854
	cpumask_var_t		this_book_map;
6855 6856 6857 6858 6859 6860
	cpumask_var_t		send_covered;
	cpumask_var_t		tmpmask;
	struct sched_group	**sched_group_nodes;
	struct root_domain	*rd;
};

6861 6862 6863 6864 6865
enum s_alloc {
	sa_sched_groups = 0,
	sa_rootdomain,
	sa_tmpmask,
	sa_send_covered,
6866
	sa_this_book_map,
6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878
	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,
};

6879
/*
6880
 * SMT sched-domains:
6881
 */
L
Linus Torvalds 已提交
6882
#ifdef CONFIG_SCHED_SMT
6883
static DEFINE_PER_CPU(struct static_sched_domain, cpu_domains);
6884
static DEFINE_PER_CPU(struct static_sched_group, sched_groups);
6885

I
Ingo Molnar 已提交
6886
static int
6887 6888
cpu_to_cpu_group(int cpu, const struct cpumask *cpu_map,
		 struct sched_group **sg, struct cpumask *unused)
L
Linus Torvalds 已提交
6889
{
6890
	if (sg)
6891
		*sg = &per_cpu(sched_groups, cpu).sg;
L
Linus Torvalds 已提交
6892 6893
	return cpu;
}
6894
#endif /* CONFIG_SCHED_SMT */
L
Linus Torvalds 已提交
6895

6896 6897 6898
/*
 * multi-core sched-domains:
 */
6899
#ifdef CONFIG_SCHED_MC
6900 6901
static DEFINE_PER_CPU(struct static_sched_domain, core_domains);
static DEFINE_PER_CPU(struct static_sched_group, sched_group_core);
6902

I
Ingo Molnar 已提交
6903
static int
6904 6905
cpu_to_core_group(int cpu, const struct cpumask *cpu_map,
		  struct sched_group **sg, struct cpumask *mask)
6906
{
6907
	int group;
6908
#ifdef CONFIG_SCHED_SMT
6909
	cpumask_and(mask, topology_thread_cpumask(cpu), cpu_map);
6910
	group = cpumask_first(mask);
6911 6912 6913
#else
	group = cpu;
#endif
6914
	if (sg)
6915
		*sg = &per_cpu(sched_group_core, group).sg;
6916
	return group;
6917
}
6918
#endif /* CONFIG_SCHED_MC */
6919

6920 6921 6922 6923 6924 6925 6926
/*
 * 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);

I
Ingo Molnar 已提交
6927
static int
6928 6929
cpu_to_book_group(int cpu, const struct cpumask *cpu_map,
		  struct sched_group **sg, struct cpumask *mask)
6930
{
6931 6932 6933 6934 6935 6936 6937 6938
	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
6939
	if (sg)
6940 6941
		*sg = &per_cpu(sched_group_book, group).sg;
	return group;
6942
}
6943
#endif /* CONFIG_SCHED_BOOK */
6944

6945 6946
static DEFINE_PER_CPU(struct static_sched_domain, phys_domains);
static DEFINE_PER_CPU(struct static_sched_group, sched_group_phys);
6947

I
Ingo Molnar 已提交
6948
static int
6949 6950
cpu_to_phys_group(int cpu, const struct cpumask *cpu_map,
		  struct sched_group **sg, struct cpumask *mask)
L
Linus Torvalds 已提交
6951
{
6952
	int group;
6953 6954 6955 6956
#ifdef CONFIG_SCHED_BOOK
	cpumask_and(mask, cpu_book_mask(cpu), cpu_map);
	group = cpumask_first(mask);
#elif defined(CONFIG_SCHED_MC)
6957
	cpumask_and(mask, cpu_coregroup_mask(cpu), cpu_map);
6958
	group = cpumask_first(mask);
6959
#elif defined(CONFIG_SCHED_SMT)
6960
	cpumask_and(mask, topology_thread_cpumask(cpu), cpu_map);
6961
	group = cpumask_first(mask);
L
Linus Torvalds 已提交
6962
#else
6963
	group = cpu;
L
Linus Torvalds 已提交
6964
#endif
6965
	if (sg)
6966
		*sg = &per_cpu(sched_group_phys, group).sg;
6967
	return group;
L
Linus Torvalds 已提交
6968 6969 6970 6971
}

#ifdef CONFIG_NUMA
/*
6972 6973 6974
 * 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
6975
 */
6976
static DEFINE_PER_CPU(struct static_sched_domain, node_domains);
6977
static struct sched_group ***sched_group_nodes_bycpu;
L
Linus Torvalds 已提交
6978

6979
static DEFINE_PER_CPU(struct static_sched_domain, allnodes_domains);
6980
static DEFINE_PER_CPU(struct static_sched_group, sched_group_allnodes);
6981

6982 6983 6984
static int cpu_to_allnodes_group(int cpu, const struct cpumask *cpu_map,
				 struct sched_group **sg,
				 struct cpumask *nodemask)
6985
{
6986 6987
	int group;

6988
	cpumask_and(nodemask, cpumask_of_node(cpu_to_node(cpu)), cpu_map);
6989
	group = cpumask_first(nodemask);
6990 6991

	if (sg)
6992
		*sg = &per_cpu(sched_group_allnodes, group).sg;
6993
	return group;
L
Linus Torvalds 已提交
6994
}
6995

6996 6997 6998 6999 7000 7001 7002
static void init_numa_sched_groups_power(struct sched_group *group_head)
{
	struct sched_group *sg = group_head;
	int j;

	if (!sg)
		return;
7003
	do {
7004
		for_each_cpu(j, sched_group_cpus(sg)) {
7005
			struct sched_domain *sd;
7006

7007
			sd = &per_cpu(phys_domains, j).sd;
7008
			if (j != group_first_cpu(sd->groups)) {
7009 7010 7011 7012 7013 7014
				/*
				 * Only add "power" once for each
				 * physical package.
				 */
				continue;
			}
7015

7016
			sg->cpu_power += sd->groups->cpu_power;
7017 7018 7019
		}
		sg = sg->next;
	} while (sg != group_head);
7020
}
7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031 7032 7033 7034 7035 7036 7037 7038 7039 7040 7041

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
7042 7043
		printk(KERN_WARNING "Can not alloc domain group for node %d\n",
		       num);
7044 7045 7046 7047 7048 7049 7050 7051 7052
		return -ENOMEM;
	}
	d->sched_group_nodes[num] = sg;

	for_each_cpu(j, d->nodemask) {
		sd = &per_cpu(node_domains, j).sd;
		sd->groups = sg;
	}

7053
	sg->cpu_power = 0;
7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069 7070 7071
	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
7072 7073
			printk(KERN_WARNING
			       "Can not alloc domain group for node %d\n", j);
7074 7075
			return -ENOMEM;
		}
7076
		sg->cpu_power = 0;
7077 7078 7079 7080 7081 7082 7083 7084 7085
		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;
}
7086
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
7087

7088
#ifdef CONFIG_NUMA
7089
/* Free memory allocated for various sched_group structures */
7090 7091
static void free_sched_groups(const struct cpumask *cpu_map,
			      struct cpumask *nodemask)
7092
{
7093
	int cpu, i;
7094

7095
	for_each_cpu(cpu, cpu_map) {
7096 7097 7098 7099 7100 7101
		struct sched_group **sched_group_nodes
			= sched_group_nodes_bycpu[cpu];

		if (!sched_group_nodes)
			continue;

7102
		for (i = 0; i < nr_node_ids; i++) {
7103 7104
			struct sched_group *oldsg, *sg = sched_group_nodes[i];

7105
			cpumask_and(nodemask, cpumask_of_node(i), cpu_map);
7106
			if (cpumask_empty(nodemask))
7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122
				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;
	}
}
7123
#else /* !CONFIG_NUMA */
7124 7125
static void free_sched_groups(const struct cpumask *cpu_map,
			      struct cpumask *nodemask)
7126 7127
{
}
7128
#endif /* CONFIG_NUMA */
7129

7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143
/*
 * 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;
7144 7145
	long power;
	int weight;
7146 7147 7148

	WARN_ON(!sd || !sd->groups);

7149
	if (cpu != group_first_cpu(sd->groups))
7150 7151
		return;

7152 7153
	sd->groups->group_weight = cpumask_weight(sched_group_cpus(sd->groups));

7154 7155
	child = sd->child;

7156
	sd->groups->cpu_power = 0;
7157

7158 7159 7160 7161 7162
	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
7163 7164 7165
		 * Usually multiple threads get a better yield out of
		 * that one core than a single thread would have,
		 * reflect that in sd->smt_gain.
7166
		 */
P
Peter Zijlstra 已提交
7167 7168
		if ((sd->flags & SD_SHARE_CPUPOWER) && weight > 1) {
			power *= sd->smt_gain;
7169
			power /= weight;
P
Peter Zijlstra 已提交
7170 7171
			power >>= SCHED_LOAD_SHIFT;
		}
7172
		sd->groups->cpu_power += power;
7173 7174 7175 7176
		return;
	}

	/*
7177
	 * Add cpu_power of each child group to this groups cpu_power.
7178 7179 7180
	 */
	group = child->groups;
	do {
7181
		sd->groups->cpu_power += group->cpu_power;
7182 7183 7184 7185
		group = group->next;
	} while (group != child->groups);
}

7186 7187 7188 7189 7190
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

7191 7192 7193 7194 7195 7196
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

7197
#define	SD_INIT(sd, type)	sd_init_##type(sd)
7198

7199 7200 7201 7202 7203
#define SD_INIT_FUNC(type)	\
static noinline void sd_init_##type(struct sched_domain *sd)	\
{								\
	memset(sd, 0, sizeof(*sd));				\
	*sd = SD_##type##_INIT;					\
7204
	sd->level = SD_LV_##type;				\
7205
	SD_INIT_NAME(sd, type);					\
7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218
}

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
7219 7220 7221
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
7222

7223 7224 7225 7226
static int default_relax_domain_level = -1;

static int __init setup_relax_domain_level(char *str)
{
7227 7228 7229 7230 7231 7232
	unsigned long val;

	val = simple_strtoul(str, NULL, 0);
	if (val < SD_LV_MAX)
		default_relax_domain_level = val;

7233 7234 7235 7236 7237 7238 7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250
	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 */
7251
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
7252 7253
	} else {
		/* turn on idle balance on this domain */
7254
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
7255 7256 7257
	}
}

7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270
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 */
7271 7272
	case sa_this_book_map:
		free_cpumask_var(d->this_book_map); /* fall through */
7273 7274 7275 7276 7277 7278 7279
	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:
7280
#ifdef CONFIG_NUMA
7281 7282 7283 7284 7285 7286 7287
		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 */
7288
#endif
7289 7290 7291 7292
	case sa_none:
		break;
	}
}
7293

7294 7295 7296
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
7297
#ifdef CONFIG_NUMA
7298 7299 7300 7301 7302 7303 7304 7305 7306 7307
	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
7308
		printk(KERN_WARNING "Can not alloc sched group node list\n");
7309
		return sa_notcovered;
7310
	}
7311
	sched_group_nodes_bycpu[cpumask_first(cpu_map)] = d->sched_group_nodes;
7312
#endif
7313 7314 7315 7316 7317 7318
	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;
7319
	if (!alloc_cpumask_var(&d->this_book_map, GFP_KERNEL))
7320
		return sa_this_core_map;
7321 7322
	if (!alloc_cpumask_var(&d->send_covered, GFP_KERNEL))
		return sa_this_book_map;
7323 7324 7325 7326
	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
7327
		printk(KERN_WARNING "Cannot alloc root domain\n");
7328
		return sa_tmpmask;
G
Gregory Haskins 已提交
7329
	}
7330 7331
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
7332

7333 7334 7335 7336
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;
7337
#ifdef CONFIG_NUMA
7338
	struct sched_domain *parent;
L
Linus Torvalds 已提交
7339

7340 7341 7342 7343 7344
	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);
7345
		set_domain_attribute(sd, attr);
7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359
		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
7360
#endif
7361 7362
	return sd;
}
L
Linus Torvalds 已提交
7363

7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378
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
7379

7380 7381 7382 7383 7384 7385 7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396
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;
}

7397 7398 7399 7400 7401
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;
7402
#ifdef CONFIG_SCHED_MC
7403 7404 7405 7406 7407 7408 7409
	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);
7410
#endif
7411 7412
	return sd;
}
7413

7414 7415 7416 7417 7418
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
7419
#ifdef CONFIG_SCHED_SMT
7420 7421 7422 7423 7424 7425 7426
	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
7427
#endif
7428 7429
	return sd;
}
L
Linus Torvalds 已提交
7430

7431 7432 7433 7434
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
7435
#ifdef CONFIG_SCHED_SMT
7436 7437 7438 7439 7440 7441 7442 7443
	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
7444
#endif
7445
#ifdef CONFIG_SCHED_MC
7446 7447 7448 7449 7450 7451 7452
	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;
7453 7454 7455 7456 7457 7458 7459 7460 7461
#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;
7462
#endif
7463 7464 7465 7466 7467 7468 7469
	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
7470
#ifdef CONFIG_NUMA
7471 7472 7473 7474 7475
	case SD_LV_ALLNODES:
		init_sched_build_groups(cpu_map, cpu_map, &cpu_to_allnodes_group,
					d->send_covered, d->tmpmask);
		break;
#endif
7476 7477
	default:
		break;
7478
	}
7479
}
7480

7481 7482 7483 7484 7485 7486 7487 7488 7489
/*
 * 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;
7490
	struct sched_domain *sd;
7491
	int i;
7492
#ifdef CONFIG_NUMA
7493
	d.sd_allnodes = 0;
7494
#endif
7495

7496 7497 7498 7499
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
	alloc_state = sa_sched_groups;
7500

L
Linus Torvalds 已提交
7501
	/*
7502
	 * Set up domains for cpus specified by the cpu_map.
L
Linus Torvalds 已提交
7503
	 */
7504
	for_each_cpu(i, cpu_map) {
7505 7506
		cpumask_and(d.nodemask, cpumask_of_node(cpu_to_node(i)),
			    cpu_map);
I
Ingo Molnar 已提交
7507

7508
		sd = __build_numa_sched_domains(&d, cpu_map, attr, i);
7509
		sd = __build_cpu_sched_domain(&d, cpu_map, attr, sd, i);
7510
		sd = __build_book_sched_domain(&d, cpu_map, attr, sd, i);
7511
		sd = __build_mc_sched_domain(&d, cpu_map, attr, sd, i);
7512
		sd = __build_smt_sched_domain(&d, cpu_map, attr, sd, i);
L
Linus Torvalds 已提交
7513
	}
7514

7515
	for_each_cpu(i, cpu_map) {
7516
		build_sched_groups(&d, SD_LV_SIBLING, cpu_map, i);
7517
		build_sched_groups(&d, SD_LV_BOOK, cpu_map, i);
7518
		build_sched_groups(&d, SD_LV_MC, cpu_map, i);
L
Linus Torvalds 已提交
7519
	}
7520

L
Linus Torvalds 已提交
7521
	/* Set up physical groups */
7522 7523
	for (i = 0; i < nr_node_ids; i++)
		build_sched_groups(&d, SD_LV_CPU, cpu_map, i);
7524

L
Linus Torvalds 已提交
7525 7526
#ifdef CONFIG_NUMA
	/* Set up node groups */
7527 7528
	if (d.sd_allnodes)
		build_sched_groups(&d, SD_LV_ALLNODES, cpu_map, 0);
7529

7530 7531
	for (i = 0; i < nr_node_ids; i++)
		if (build_numa_sched_groups(&d, cpu_map, i))
7532
			goto error;
L
Linus Torvalds 已提交
7533 7534 7535
#endif

	/* Calculate CPU power for physical packages and nodes */
7536
#ifdef CONFIG_SCHED_SMT
7537
	for_each_cpu(i, cpu_map) {
7538
		sd = &per_cpu(cpu_domains, i).sd;
7539
		init_sched_groups_power(i, sd);
7540
	}
L
Linus Torvalds 已提交
7541
#endif
7542
#ifdef CONFIG_SCHED_MC
7543
	for_each_cpu(i, cpu_map) {
7544
		sd = &per_cpu(core_domains, i).sd;
7545
		init_sched_groups_power(i, sd);
7546 7547
	}
#endif
7548 7549 7550 7551 7552 7553
#ifdef CONFIG_SCHED_BOOK
	for_each_cpu(i, cpu_map) {
		sd = &per_cpu(book_domains, i).sd;
		init_sched_groups_power(i, sd);
	}
#endif
7554

7555
	for_each_cpu(i, cpu_map) {
7556
		sd = &per_cpu(phys_domains, i).sd;
7557
		init_sched_groups_power(i, sd);
L
Linus Torvalds 已提交
7558 7559
	}

7560
#ifdef CONFIG_NUMA
7561
	for (i = 0; i < nr_node_ids; i++)
7562
		init_numa_sched_groups_power(d.sched_group_nodes[i]);
7563

7564
	if (d.sd_allnodes) {
7565
		struct sched_group *sg;
7566

7567
		cpu_to_allnodes_group(cpumask_first(cpu_map), cpu_map, &sg,
7568
								d.tmpmask);
7569 7570
		init_numa_sched_groups_power(sg);
	}
7571 7572
#endif

L
Linus Torvalds 已提交
7573
	/* Attach the domains */
7574
	for_each_cpu(i, cpu_map) {
L
Linus Torvalds 已提交
7575
#ifdef CONFIG_SCHED_SMT
7576
		sd = &per_cpu(cpu_domains, i).sd;
7577
#elif defined(CONFIG_SCHED_MC)
7578
		sd = &per_cpu(core_domains, i).sd;
7579 7580
#elif defined(CONFIG_SCHED_BOOK)
		sd = &per_cpu(book_domains, i).sd;
L
Linus Torvalds 已提交
7581
#else
7582
		sd = &per_cpu(phys_domains, i).sd;
L
Linus Torvalds 已提交
7583
#endif
7584
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
7585
	}
7586

7587 7588 7589
	d.sched_group_nodes = NULL; /* don't free this we still need it */
	__free_domain_allocs(&d, sa_tmpmask, cpu_map);
	return 0;
7590 7591

error:
7592 7593
	__free_domain_allocs(&d, alloc_state, cpu_map);
	return -ENOMEM;
L
Linus Torvalds 已提交
7594
}
P
Paul Jackson 已提交
7595

7596
static int build_sched_domains(const struct cpumask *cpu_map)
7597 7598 7599 7600
{
	return __build_sched_domains(cpu_map, NULL);
}

7601
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
7602
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
7603 7604
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
7605 7606 7607

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
7608 7609
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
7610
 */
7611
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
7612

7613 7614 7615 7616 7617 7618
/*
 * 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)
7619
{
7620
	return 0;
7621 7622
}

7623 7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642 7643 7644 7645 7646 7647
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);
}

7648
/*
I
Ingo Molnar 已提交
7649
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
7650 7651
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
7652
 */
7653
static int arch_init_sched_domains(const struct cpumask *cpu_map)
7654
{
7655 7656
	int err;

7657
	arch_update_cpu_topology();
P
Paul Jackson 已提交
7658
	ndoms_cur = 1;
7659
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
7660
	if (!doms_cur)
7661 7662
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
7663
	dattr_cur = NULL;
7664
	err = build_sched_domains(doms_cur[0]);
7665
	register_sched_domain_sysctl();
7666 7667

	return err;
7668 7669
}

7670 7671
static void arch_destroy_sched_domains(const struct cpumask *cpu_map,
				       struct cpumask *tmpmask)
L
Linus Torvalds 已提交
7672
{
7673
	free_sched_groups(cpu_map, tmpmask);
7674
}
L
Linus Torvalds 已提交
7675

7676 7677 7678 7679
/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
7680
static void detach_destroy_domains(const struct cpumask *cpu_map)
7681
{
7682 7683
	/* Save because hotplug lock held. */
	static DECLARE_BITMAP(tmpmask, CONFIG_NR_CPUS);
7684 7685
	int i;

7686
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
7687
		cpu_attach_domain(NULL, &def_root_domain, i);
7688
	synchronize_sched();
7689
	arch_destroy_sched_domains(cpu_map, to_cpumask(tmpmask));
7690 7691
}

7692 7693 7694 7695 7696 7697 7698 7699 7700 7701 7702 7703 7704 7705 7706 7707
/* 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
7708 7709
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
7710
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
7711 7712 7713
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
7714
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
7715 7716 7717
 * 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
7718 7719 7720
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
7721 7722 7723 7724 7725 7726
 * 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
7727
 *
7728
 * If doms_new == NULL it will be replaced with cpu_online_mask.
7729 7730
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
7731
 *
P
Paul Jackson 已提交
7732 7733
 * Call with hotplug lock held
 */
7734
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
7735
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
7736
{
7737
	int i, j, n;
7738
	int new_topology;
P
Paul Jackson 已提交
7739

7740
	mutex_lock(&sched_domains_mutex);
7741

7742 7743 7744
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

7745 7746 7747
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

7748
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
7749 7750 7751

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
7752
		for (j = 0; j < n && !new_topology; j++) {
7753
			if (cpumask_equal(doms_cur[i], doms_new[j])
7754
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
7755 7756 7757
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
7758
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
7759 7760 7761 7762
match1:
		;
	}

7763 7764
	if (doms_new == NULL) {
		ndoms_cur = 0;
7765
		doms_new = &fallback_doms;
7766
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
7767
		WARN_ON_ONCE(dattr_new);
7768 7769
	}

P
Paul Jackson 已提交
7770 7771
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
7772
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
7773
			if (cpumask_equal(doms_new[i], doms_cur[j])
7774
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
7775 7776 7777
				goto match2;
		}
		/* no match - add a new doms_new */
7778
		__build_sched_domains(doms_new[i],
7779
					dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
7780 7781 7782 7783 7784
match2:
		;
	}

	/* Remember the new sched domains */
7785 7786
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
7787
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
7788
	doms_cur = doms_new;
7789
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
7790
	ndoms_cur = ndoms_new;
7791 7792

	register_sched_domain_sysctl();
7793

7794
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
7795 7796
}

7797
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
7798
static void arch_reinit_sched_domains(void)
7799
{
7800
	get_online_cpus();
7801 7802 7803 7804

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

7805
	rebuild_sched_domains();
7806
	put_online_cpus();
7807 7808 7809 7810
}

static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
{
7811
	unsigned int level = 0;
7812

7813 7814 7815 7816 7817 7818 7819 7820 7821 7822 7823
	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)
7824 7825 7826
		return -EINVAL;

	if (smt)
7827
		sched_smt_power_savings = level;
7828
	else
7829
		sched_mc_power_savings = level;
7830

7831
	arch_reinit_sched_domains();
7832

7833
	return count;
7834 7835 7836
}

#ifdef CONFIG_SCHED_MC
7837
static ssize_t sched_mc_power_savings_show(struct sysdev_class *class,
7838
					   struct sysdev_class_attribute *attr,
7839
					   char *page)
7840 7841 7842
{
	return sprintf(page, "%u\n", sched_mc_power_savings);
}
7843
static ssize_t sched_mc_power_savings_store(struct sysdev_class *class,
7844
					    struct sysdev_class_attribute *attr,
7845
					    const char *buf, size_t count)
7846 7847 7848
{
	return sched_power_savings_store(buf, count, 0);
}
7849 7850 7851
static SYSDEV_CLASS_ATTR(sched_mc_power_savings, 0644,
			 sched_mc_power_savings_show,
			 sched_mc_power_savings_store);
7852 7853 7854
#endif

#ifdef CONFIG_SCHED_SMT
7855
static ssize_t sched_smt_power_savings_show(struct sysdev_class *dev,
7856
					    struct sysdev_class_attribute *attr,
7857
					    char *page)
7858 7859 7860
{
	return sprintf(page, "%u\n", sched_smt_power_savings);
}
7861
static ssize_t sched_smt_power_savings_store(struct sysdev_class *dev,
7862
					     struct sysdev_class_attribute *attr,
7863
					     const char *buf, size_t count)
7864 7865 7866
{
	return sched_power_savings_store(buf, count, 1);
}
7867 7868
static SYSDEV_CLASS_ATTR(sched_smt_power_savings, 0644,
		   sched_smt_power_savings_show,
A
Adrian Bunk 已提交
7869 7870 7871
		   sched_smt_power_savings_store);
#endif

7872
int __init sched_create_sysfs_power_savings_entries(struct sysdev_class *cls)
A
Adrian Bunk 已提交
7873 7874 7875 7876 7877 7878 7879 7880 7881 7882 7883 7884 7885 7886 7887
{
	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;
}
7888
#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
7889

L
Linus Torvalds 已提交
7890
/*
7891 7892 7893
 * 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
7894
 */
7895 7896
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
7897
{
7898
	switch (action & ~CPU_TASKS_FROZEN) {
7899
	case CPU_ONLINE:
7900
	case CPU_DOWN_FAILED:
7901
		cpuset_update_active_cpus();
7902
		return NOTIFY_OK;
7903 7904 7905 7906
	default:
		return NOTIFY_DONE;
	}
}
7907

7908 7909
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
7910 7911 7912 7913 7914
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
		cpuset_update_active_cpus();
		return NOTIFY_OK;
7915 7916 7917 7918 7919 7920 7921
	default:
		return NOTIFY_DONE;
	}
}

static int update_runtime(struct notifier_block *nfb,
				unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
7922
{
P
Peter Zijlstra 已提交
7923 7924
	int cpu = (int)(long)hcpu;

L
Linus Torvalds 已提交
7925 7926
	switch (action) {
	case CPU_DOWN_PREPARE:
7927
	case CPU_DOWN_PREPARE_FROZEN:
P
Peter Zijlstra 已提交
7928
		disable_runtime(cpu_rq(cpu));
L
Linus Torvalds 已提交
7929 7930 7931
		return NOTIFY_OK;

	case CPU_DOWN_FAILED:
7932
	case CPU_DOWN_FAILED_FROZEN:
L
Linus Torvalds 已提交
7933
	case CPU_ONLINE:
7934
	case CPU_ONLINE_FROZEN:
P
Peter Zijlstra 已提交
7935
		enable_runtime(cpu_rq(cpu));
7936 7937
		return NOTIFY_OK;

L
Linus Torvalds 已提交
7938 7939 7940 7941 7942 7943 7944
	default:
		return NOTIFY_DONE;
	}
}

void __init sched_init_smp(void)
{
7945 7946 7947
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
7948
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
7949

7950 7951 7952 7953 7954
#if defined(CONFIG_NUMA)
	sched_group_nodes_bycpu = kzalloc(nr_cpu_ids * sizeof(void **),
								GFP_KERNEL);
	BUG_ON(sched_group_nodes_bycpu == NULL);
#endif
7955
	get_online_cpus();
7956
	mutex_lock(&sched_domains_mutex);
7957
	arch_init_sched_domains(cpu_active_mask);
7958 7959 7960
	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);
7961
	mutex_unlock(&sched_domains_mutex);
7962
	put_online_cpus();
7963

7964 7965
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
7966 7967 7968 7969

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

7970
	init_hrtick();
7971 7972

	/* Move init over to a non-isolated CPU */
7973
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
7974
		BUG();
I
Ingo Molnar 已提交
7975
	sched_init_granularity();
7976
	free_cpumask_var(non_isolated_cpus);
7977

7978
	init_sched_rt_class();
L
Linus Torvalds 已提交
7979 7980 7981 7982
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
7983
	sched_init_granularity();
L
Linus Torvalds 已提交
7984 7985 7986
}
#endif /* CONFIG_SMP */

7987 7988
const_debug unsigned int sysctl_timer_migration = 1;

L
Linus Torvalds 已提交
7989 7990 7991 7992 7993 7994 7995
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
7996
static void init_cfs_rq(struct cfs_rq *cfs_rq, struct rq *rq)
I
Ingo Molnar 已提交
7997 7998
{
	cfs_rq->tasks_timeline = RB_ROOT;
7999
	INIT_LIST_HEAD(&cfs_rq->tasks);
I
Ingo Molnar 已提交
8000 8001
#ifdef CONFIG_FAIR_GROUP_SCHED
	cfs_rq->rq = rq;
8002
	/* allow initial update_cfs_load() to truncate */
8003
#ifdef CONFIG_SMP
8004
	cfs_rq->load_stamp = 1;
8005
#endif
I
Ingo Molnar 已提交
8006
#endif
P
Peter Zijlstra 已提交
8007
	cfs_rq->min_vruntime = (u64)(-(1LL << 20));
I
Ingo Molnar 已提交
8008 8009
}

P
Peter Zijlstra 已提交
8010 8011 8012 8013 8014 8015 8016 8017 8018 8019 8020 8021 8022
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);

8023
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
8024
	rt_rq->highest_prio.curr = MAX_RT_PRIO;
8025
#ifdef CONFIG_SMP
8026
	rt_rq->highest_prio.next = MAX_RT_PRIO;
P
Peter Zijlstra 已提交
8027 8028
#endif
#endif
P
Peter Zijlstra 已提交
8029 8030 8031
#ifdef CONFIG_SMP
	rt_rq->rt_nr_migratory = 0;
	rt_rq->overloaded = 0;
8032
	plist_head_init_raw(&rt_rq->pushable_tasks, &rq->lock);
P
Peter Zijlstra 已提交
8033 8034 8035 8036
#endif

	rt_rq->rt_time = 0;
	rt_rq->rt_throttled = 0;
P
Peter Zijlstra 已提交
8037
	rt_rq->rt_runtime = 0;
8038
	raw_spin_lock_init(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8039

8040
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8041
	rt_rq->rt_nr_boosted = 0;
P
Peter Zijlstra 已提交
8042 8043
	rt_rq->rq = rq;
#endif
P
Peter Zijlstra 已提交
8044 8045
}

P
Peter Zijlstra 已提交
8046
#ifdef CONFIG_FAIR_GROUP_SCHED
8047
static void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq,
8048
				struct sched_entity *se, int cpu,
8049
				struct sched_entity *parent)
P
Peter Zijlstra 已提交
8050
{
8051
	struct rq *rq = cpu_rq(cpu);
P
Peter Zijlstra 已提交
8052 8053 8054 8055 8056
	tg->cfs_rq[cpu] = cfs_rq;
	init_cfs_rq(cfs_rq, rq);
	cfs_rq->tg = tg;

	tg->se[cpu] = se;
8057
	/* se could be NULL for root_task_group */
D
Dhaval Giani 已提交
8058 8059 8060
	if (!se)
		return;

8061 8062 8063 8064 8065
	if (!parent)
		se->cfs_rq = &rq->cfs;
	else
		se->cfs_rq = parent->my_q;

P
Peter Zijlstra 已提交
8066
	se->my_q = cfs_rq;
8067
	update_load_set(&se->load, 0);
8068
	se->parent = parent;
P
Peter Zijlstra 已提交
8069
}
8070
#endif
P
Peter Zijlstra 已提交
8071

8072
#ifdef CONFIG_RT_GROUP_SCHED
8073
static void init_tg_rt_entry(struct task_group *tg, struct rt_rq *rt_rq,
8074
		struct sched_rt_entity *rt_se, int cpu,
8075
		struct sched_rt_entity *parent)
P
Peter Zijlstra 已提交
8076
{
8077 8078
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
8079 8080 8081
	tg->rt_rq[cpu] = rt_rq;
	init_rt_rq(rt_rq, rq);
	rt_rq->tg = tg;
P
Peter Zijlstra 已提交
8082
	rt_rq->rt_runtime = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
8083 8084

	tg->rt_se[cpu] = rt_se;
D
Dhaval Giani 已提交
8085 8086 8087
	if (!rt_se)
		return;

8088 8089 8090 8091 8092
	if (!parent)
		rt_se->rt_rq = &rq->rt;
	else
		rt_se->rt_rq = parent->my_q;

P
Peter Zijlstra 已提交
8093
	rt_se->my_q = rt_rq;
8094
	rt_se->parent = parent;
P
Peter Zijlstra 已提交
8095 8096 8097 8098
	INIT_LIST_HEAD(&rt_se->run_list);
}
#endif

L
Linus Torvalds 已提交
8099 8100
void __init sched_init(void)
{
I
Ingo Molnar 已提交
8101
	int i, j;
8102 8103 8104 8105 8106 8107 8108
	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 **);
8109
#endif
8110
#ifdef CONFIG_CPUMASK_OFFSTACK
8111
	alloc_size += num_possible_cpus() * cpumask_size();
8112 8113
#endif
	if (alloc_size) {
8114
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
8115 8116

#ifdef CONFIG_FAIR_GROUP_SCHED
8117
		root_task_group.se = (struct sched_entity **)ptr;
8118 8119
		ptr += nr_cpu_ids * sizeof(void **);

8120
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
8121
		ptr += nr_cpu_ids * sizeof(void **);
8122

8123
#endif /* CONFIG_FAIR_GROUP_SCHED */
8124
#ifdef CONFIG_RT_GROUP_SCHED
8125
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
8126 8127
		ptr += nr_cpu_ids * sizeof(void **);

8128
		root_task_group.rt_rq = (struct rt_rq **)ptr;
8129 8130
		ptr += nr_cpu_ids * sizeof(void **);

8131
#endif /* CONFIG_RT_GROUP_SCHED */
8132 8133 8134 8135 8136 8137
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
8138
	}
I
Ingo Molnar 已提交
8139

G
Gregory Haskins 已提交
8140 8141 8142 8143
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

8144 8145 8146 8147
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
8148
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
8149
			global_rt_period(), global_rt_runtime());
8150
#endif /* CONFIG_RT_GROUP_SCHED */
8151

D
Dhaval Giani 已提交
8152
#ifdef CONFIG_CGROUP_SCHED
8153 8154
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
8155
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
8156
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
8157

8158
	for_each_possible_cpu(i) {
8159
		struct rq *rq;
L
Linus Torvalds 已提交
8160 8161

		rq = cpu_rq(i);
8162
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
8163
		rq->nr_running = 0;
8164 8165
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
I
Ingo Molnar 已提交
8166
		init_cfs_rq(&rq->cfs, rq);
P
Peter Zijlstra 已提交
8167
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
8168
#ifdef CONFIG_FAIR_GROUP_SCHED
8169
		root_task_group.shares = root_task_group_load;
P
Peter Zijlstra 已提交
8170
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
8171
		/*
8172
		 * How much cpu bandwidth does root_task_group get?
D
Dhaval Giani 已提交
8173 8174 8175 8176
		 *
		 * 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
8177
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
8178 8179 8180
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
8181
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
8182 8183 8184
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
8185
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
8186
		 *
8187 8188
		 * We achieve this by letting root_task_group's tasks sit
		 * directly in rq->cfs (i.e root_task_group->se[] = NULL).
D
Dhaval Giani 已提交
8189
		 */
8190
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
8191 8192 8193
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
8194
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8195
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
8196
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
8197
#endif
L
Linus Torvalds 已提交
8198

I
Ingo Molnar 已提交
8199 8200
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
8201 8202 8203

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
8204
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
8205
		rq->sd = NULL;
G
Gregory Haskins 已提交
8206
		rq->rd = NULL;
8207
		rq->cpu_power = SCHED_LOAD_SCALE;
8208
		rq->post_schedule = 0;
L
Linus Torvalds 已提交
8209
		rq->active_balance = 0;
I
Ingo Molnar 已提交
8210
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
8211
		rq->push_cpu = 0;
8212
		rq->cpu = i;
8213
		rq->online = 0;
8214 8215
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
8216
		rq_attach_root(rq, &def_root_domain);
8217 8218 8219 8220
#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
8221
#endif
P
Peter Zijlstra 已提交
8222
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
8223 8224 8225
		atomic_set(&rq->nr_iowait, 0);
	}

8226
	set_load_weight(&init_task);
8227

8228 8229 8230 8231
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

8232
#ifdef CONFIG_SMP
8233
	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);
8234 8235
#endif

8236
#ifdef CONFIG_RT_MUTEXES
8237
	plist_head_init_raw(&init_task.pi_waiters, &init_task.pi_lock);
8238 8239
#endif

L
Linus Torvalds 已提交
8240 8241 8242 8243 8244 8245 8246 8247 8248 8249 8250 8251 8252
	/*
	 * 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());
8253 8254 8255

	calc_load_update = jiffies + LOAD_FREQ;

I
Ingo Molnar 已提交
8256 8257 8258 8259
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
8260

8261
	/* Allocate the nohz_cpu_mask if CONFIG_CPUMASK_OFFSTACK */
8262
	zalloc_cpumask_var(&nohz_cpu_mask, GFP_NOWAIT);
8263
#ifdef CONFIG_SMP
8264
#ifdef CONFIG_NO_HZ
8265 8266 8267 8268 8269
	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);
8270
#endif
R
Rusty Russell 已提交
8271 8272 8273
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
8274
#endif /* SMP */
8275

8276
	scheduler_running = 1;
L
Linus Torvalds 已提交
8277 8278 8279
}

#ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
8280 8281
static inline int preempt_count_equals(int preempt_offset)
{
8282
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
8283

A
Arnd Bergmann 已提交
8284
	return (nested == preempt_offset);
8285 8286
}

8287
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
8288
{
8289
#ifdef in_atomic
L
Linus Torvalds 已提交
8290 8291
	static unsigned long prev_jiffy;	/* ratelimiting */

8292 8293
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
8294 8295 8296 8297 8298
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
8299 8300 8301 8302 8303 8304 8305
	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
8306 8307 8308 8309 8310

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
L
Linus Torvalds 已提交
8311 8312 8313 8314 8315 8316
#endif
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
8317 8318
static void normalize_task(struct rq *rq, struct task_struct *p)
{
P
Peter Zijlstra 已提交
8319 8320
	const struct sched_class *prev_class = p->sched_class;
	int old_prio = p->prio;
8321
	int on_rq;
8322

P
Peter Zijlstra 已提交
8323
	on_rq = p->on_rq;
8324 8325 8326 8327 8328 8329 8330
	if (on_rq)
		deactivate_task(rq, p, 0);
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
		activate_task(rq, p, 0);
		resched_task(rq->curr);
	}
P
Peter Zijlstra 已提交
8331 8332

	check_class_changed(rq, p, prev_class, old_prio);
8333 8334
}

L
Linus Torvalds 已提交
8335 8336
void normalize_rt_tasks(void)
{
8337
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
8338
	unsigned long flags;
8339
	struct rq *rq;
L
Linus Torvalds 已提交
8340

8341
	read_lock_irqsave(&tasklist_lock, flags);
8342
	do_each_thread(g, p) {
8343 8344 8345 8346 8347 8348
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
8349 8350
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
8351 8352 8353
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
8354
#endif
I
Ingo Molnar 已提交
8355 8356 8357 8358 8359 8360 8361 8362

		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
8363
			continue;
I
Ingo Molnar 已提交
8364
		}
L
Linus Torvalds 已提交
8365

8366
		raw_spin_lock(&p->pi_lock);
8367
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
8368

8369
		normalize_task(rq, p);
8370

8371
		__task_rq_unlock(rq);
8372
		raw_spin_unlock(&p->pi_lock);
8373 8374
	} while_each_thread(g, p);

8375
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
8376 8377 8378
}

#endif /* CONFIG_MAGIC_SYSRQ */
8379

8380
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
8381
/*
8382
 * These functions are only useful for the IA64 MCA handling, or kdb.
8383 8384 8385 8386 8387 8388 8389 8390 8391 8392 8393 8394 8395 8396
 *
 * 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!
 */
8397
struct task_struct *curr_task(int cpu)
8398 8399 8400 8401
{
	return cpu_curr(cpu);
}

8402 8403 8404
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
8405 8406 8407 8408 8409 8410
/**
 * 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
8411 8412
 * 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
8413 8414 8415 8416 8417 8418 8419
 * 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!
 */
8420
void set_curr_task(int cpu, struct task_struct *p)
8421 8422 8423 8424 8425
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
8426

8427 8428
#ifdef CONFIG_FAIR_GROUP_SCHED
static void free_fair_sched_group(struct task_group *tg)
P
Peter Zijlstra 已提交
8429 8430 8431 8432 8433 8434 8435 8436 8437 8438 8439 8440 8441 8442
{
	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);
}

8443 8444
static
int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
S
Srivatsa Vaddagiri 已提交
8445 8446
{
	struct cfs_rq *cfs_rq;
8447
	struct sched_entity *se;
S
Srivatsa Vaddagiri 已提交
8448 8449
	int i;

8450
	tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL);
S
Srivatsa Vaddagiri 已提交
8451 8452
	if (!tg->cfs_rq)
		goto err;
8453
	tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL);
S
Srivatsa Vaddagiri 已提交
8454 8455
	if (!tg->se)
		goto err;
8456 8457

	tg->shares = NICE_0_LOAD;
S
Srivatsa Vaddagiri 已提交
8458 8459

	for_each_possible_cpu(i) {
8460 8461
		cfs_rq = kzalloc_node(sizeof(struct cfs_rq),
				      GFP_KERNEL, cpu_to_node(i));
S
Srivatsa Vaddagiri 已提交
8462 8463 8464
		if (!cfs_rq)
			goto err;

8465 8466
		se = kzalloc_node(sizeof(struct sched_entity),
				  GFP_KERNEL, cpu_to_node(i));
S
Srivatsa Vaddagiri 已提交
8467
		if (!se)
8468
			goto err_free_rq;
S
Srivatsa Vaddagiri 已提交
8469

8470
		init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]);
8471 8472 8473 8474
	}

	return 1;

P
Peter Zijlstra 已提交
8475
err_free_rq:
8476
	kfree(cfs_rq);
P
Peter Zijlstra 已提交
8477
err:
8478 8479 8480 8481 8482
	return 0;
}

static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
8483 8484 8485 8486 8487 8488 8489 8490 8491 8492 8493
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

	/*
	* Only empty task groups can be destroyed; so we can speculatively
	* check on_list without danger of it being re-added.
	*/
	if (!tg->cfs_rq[cpu]->on_list)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
8494
	list_del_leaf_cfs_rq(tg->cfs_rq[cpu]);
8495
	raw_spin_unlock_irqrestore(&rq->lock, flags);
8496
}
8497
#else /* !CONFG_FAIR_GROUP_SCHED */
8498 8499 8500 8501
static inline void free_fair_sched_group(struct task_group *tg)
{
}

8502 8503
static inline
int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
8504 8505 8506 8507 8508 8509 8510
{
	return 1;
}

static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
}
8511
#endif /* CONFIG_FAIR_GROUP_SCHED */
8512 8513

#ifdef CONFIG_RT_GROUP_SCHED
8514 8515 8516 8517
static void free_rt_sched_group(struct task_group *tg)
{
	int i;

8518 8519
	destroy_rt_bandwidth(&tg->rt_bandwidth);

8520 8521 8522 8523 8524 8525 8526 8527 8528 8529 8530
	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);
}

8531 8532
static
int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
8533 8534
{
	struct rt_rq *rt_rq;
8535
	struct sched_rt_entity *rt_se;
8536 8537 8538
	struct rq *rq;
	int i;

8539
	tg->rt_rq = kzalloc(sizeof(rt_rq) * nr_cpu_ids, GFP_KERNEL);
8540 8541
	if (!tg->rt_rq)
		goto err;
8542
	tg->rt_se = kzalloc(sizeof(rt_se) * nr_cpu_ids, GFP_KERNEL);
8543 8544 8545
	if (!tg->rt_se)
		goto err;

8546 8547
	init_rt_bandwidth(&tg->rt_bandwidth,
			ktime_to_ns(def_rt_bandwidth.rt_period), 0);
8548 8549 8550 8551

	for_each_possible_cpu(i) {
		rq = cpu_rq(i);

8552 8553
		rt_rq = kzalloc_node(sizeof(struct rt_rq),
				     GFP_KERNEL, cpu_to_node(i));
P
Peter Zijlstra 已提交
8554 8555
		if (!rt_rq)
			goto err;
S
Srivatsa Vaddagiri 已提交
8556

8557 8558
		rt_se = kzalloc_node(sizeof(struct sched_rt_entity),
				     GFP_KERNEL, cpu_to_node(i));
P
Peter Zijlstra 已提交
8559
		if (!rt_se)
8560
			goto err_free_rq;
S
Srivatsa Vaddagiri 已提交
8561

8562
		init_tg_rt_entry(tg, rt_rq, rt_se, i, parent->rt_se[i]);
S
Srivatsa Vaddagiri 已提交
8563 8564
	}

8565 8566
	return 1;

P
Peter Zijlstra 已提交
8567
err_free_rq:
8568
	kfree(rt_rq);
P
Peter Zijlstra 已提交
8569
err:
8570 8571
	return 0;
}
8572
#else /* !CONFIG_RT_GROUP_SCHED */
8573 8574 8575 8576
static inline void free_rt_sched_group(struct task_group *tg)
{
}

8577 8578
static inline
int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
8579 8580 8581
{
	return 1;
}
8582
#endif /* CONFIG_RT_GROUP_SCHED */
8583

D
Dhaval Giani 已提交
8584
#ifdef CONFIG_CGROUP_SCHED
8585 8586 8587 8588
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
8589
	autogroup_free(tg);
8590 8591 8592 8593
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
8594
struct task_group *sched_create_group(struct task_group *parent)
8595 8596 8597 8598 8599 8600 8601 8602
{
	struct task_group *tg;
	unsigned long flags;

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

8603
	if (!alloc_fair_sched_group(tg, parent))
8604 8605
		goto err;

8606
	if (!alloc_rt_sched_group(tg, parent))
8607 8608
		goto err;

8609
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
8610
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
8611 8612 8613 8614 8615

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

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
8616
	list_add_rcu(&tg->siblings, &parent->children);
8617
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
8618

8619
	return tg;
S
Srivatsa Vaddagiri 已提交
8620 8621

err:
P
Peter Zijlstra 已提交
8622
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
8623 8624 8625
	return ERR_PTR(-ENOMEM);
}

8626
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
8627
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
8628 8629
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
8630
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
8631 8632
}

8633
/* Destroy runqueue etc associated with a task group */
8634
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
8635
{
8636
	unsigned long flags;
8637
	int i;
S
Srivatsa Vaddagiri 已提交
8638

8639 8640
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
8641
		unregister_fair_sched_group(tg, i);
8642 8643

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
8644
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
8645
	list_del_rcu(&tg->siblings);
8646
	spin_unlock_irqrestore(&task_group_lock, flags);
8647 8648

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
8649
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
8650 8651
}

8652
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
8653 8654 8655
 *	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.
8656 8657
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
8658 8659 8660 8661 8662 8663 8664
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

8665
	running = task_current(rq, tsk);
P
Peter Zijlstra 已提交
8666
	on_rq = tsk->on_rq;
S
Srivatsa Vaddagiri 已提交
8667

8668
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
8669
		dequeue_task(rq, tsk, 0);
8670 8671
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
8672

P
Peter Zijlstra 已提交
8673
#ifdef CONFIG_FAIR_GROUP_SCHED
8674 8675 8676
	if (tsk->sched_class->task_move_group)
		tsk->sched_class->task_move_group(tsk, on_rq);
	else
P
Peter Zijlstra 已提交
8677
#endif
8678
		set_task_rq(tsk, task_cpu(tsk));
P
Peter Zijlstra 已提交
8679

8680 8681 8682
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
8683
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
8684

8685
	task_rq_unlock(rq, tsk, &flags);
S
Srivatsa Vaddagiri 已提交
8686
}
D
Dhaval Giani 已提交
8687
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
8688

8689
#ifdef CONFIG_FAIR_GROUP_SCHED
8690 8691
static DEFINE_MUTEX(shares_mutex);

8692
int sched_group_set_shares(struct task_group *tg, unsigned long shares)
S
Srivatsa Vaddagiri 已提交
8693 8694
{
	int i;
8695
	unsigned long flags;
8696

8697 8698 8699 8700 8701 8702
	/*
	 * We can't change the weight of the root cgroup.
	 */
	if (!tg->se[0])
		return -EINVAL;

8703 8704
	if (shares < MIN_SHARES)
		shares = MIN_SHARES;
8705 8706
	else if (shares > MAX_SHARES)
		shares = MAX_SHARES;
8707

8708
	mutex_lock(&shares_mutex);
8709
	if (tg->shares == shares)
8710
		goto done;
S
Srivatsa Vaddagiri 已提交
8711

8712
	tg->shares = shares;
8713
	for_each_possible_cpu(i) {
8714 8715 8716 8717 8718 8719 8720
		struct rq *rq = cpu_rq(i);
		struct sched_entity *se;

		se = tg->se[i];
		/* Propagate contribution to hierarchy */
		raw_spin_lock_irqsave(&rq->lock, flags);
		for_each_sched_entity(se)
8721
			update_cfs_shares(group_cfs_rq(se));
8722
		raw_spin_unlock_irqrestore(&rq->lock, flags);
8723
	}
S
Srivatsa Vaddagiri 已提交
8724

8725
done:
8726
	mutex_unlock(&shares_mutex);
8727
	return 0;
S
Srivatsa Vaddagiri 已提交
8728 8729
}

8730 8731 8732 8733
unsigned long sched_group_shares(struct task_group *tg)
{
	return tg->shares;
}
8734
#endif
8735

8736
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8737
/*
P
Peter Zijlstra 已提交
8738
 * Ensure that the real time constraints are schedulable.
P
Peter Zijlstra 已提交
8739
 */
P
Peter Zijlstra 已提交
8740 8741 8742 8743 8744
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
8745
		return 1ULL << 20;
P
Peter Zijlstra 已提交
8746

P
Peter Zijlstra 已提交
8747
	return div64_u64(runtime << 20, period);
P
Peter Zijlstra 已提交
8748 8749
}

P
Peter Zijlstra 已提交
8750 8751
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
8752
{
P
Peter Zijlstra 已提交
8753
	struct task_struct *g, *p;
8754

P
Peter Zijlstra 已提交
8755 8756 8757 8758
	do_each_thread(g, p) {
		if (rt_task(p) && rt_rq_of_se(&p->rt)->tg == tg)
			return 1;
	} while_each_thread(g, p);
8759

P
Peter Zijlstra 已提交
8760 8761
	return 0;
}
8762

P
Peter Zijlstra 已提交
8763 8764 8765 8766 8767
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
8768

P
Peter Zijlstra 已提交
8769 8770 8771 8772 8773 8774
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;
8775

P
Peter Zijlstra 已提交
8776 8777
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
8778

P
Peter Zijlstra 已提交
8779 8780 8781
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
8782 8783
	}

8784 8785 8786 8787 8788
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
8789

8790 8791 8792
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
8793 8794
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
8795

P
Peter Zijlstra 已提交
8796
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
8797

8798 8799 8800 8801 8802
	/*
	 * 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
8803

8804 8805 8806
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
8807 8808 8809
	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
8810

P
Peter Zijlstra 已提交
8811 8812 8813 8814
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
8815

P
Peter Zijlstra 已提交
8816
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
8817
	}
P
Peter Zijlstra 已提交
8818

P
Peter Zijlstra 已提交
8819 8820 8821 8822
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
8823 8824
}

P
Peter Zijlstra 已提交
8825
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
8826
{
P
Peter Zijlstra 已提交
8827 8828 8829 8830 8831 8832 8833
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

	return walk_tg_tree(tg_schedulable, tg_nop, &data);
8834 8835
}

8836 8837
static int tg_set_bandwidth(struct task_group *tg,
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
8838
{
P
Peter Zijlstra 已提交
8839
	int i, err = 0;
P
Peter Zijlstra 已提交
8840 8841

	mutex_lock(&rt_constraints_mutex);
8842
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
8843 8844
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
8845
		goto unlock;
P
Peter Zijlstra 已提交
8846

8847
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
8848 8849
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
8850 8851 8852 8853

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

8854
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8855
		rt_rq->rt_runtime = rt_runtime;
8856
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8857
	}
8858
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
8859
unlock:
8860
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
8861 8862 8863
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
8864 8865
}

8866 8867 8868 8869 8870 8871 8872 8873 8874 8875 8876 8877
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
8878 8879 8880 8881
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

8882
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
8883 8884
		return -1;

8885
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
8886 8887 8888
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
8889 8890 8891 8892 8893 8894 8895 8896

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;

8897 8898 8899
	if (rt_period == 0)
		return -EINVAL;

8900 8901 8902 8903 8904 8905 8906 8907 8908 8909 8910 8911 8912 8913
	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)
{
8914
	u64 runtime, period;
8915 8916
	int ret = 0;

8917 8918 8919
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

8920 8921 8922 8923 8924 8925 8926 8927
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
8928

8929
	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
8930
	read_lock(&tasklist_lock);
8931
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
8932
	read_unlock(&tasklist_lock);
8933 8934 8935 8936
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
8937 8938 8939 8940 8941 8942 8943 8944 8945 8946

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;
}

8947
#else /* !CONFIG_RT_GROUP_SCHED */
8948 8949
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
8950 8951 8952
	unsigned long flags;
	int i;

8953 8954 8955
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

8956 8957 8958 8959 8960 8961 8962
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

8963
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
8964 8965 8966
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

8967
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8968
		rt_rq->rt_runtime = global_rt_runtime();
8969
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
8970
	}
8971
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
8972

8973 8974
	return 0;
}
8975
#endif /* CONFIG_RT_GROUP_SCHED */
8976 8977

int sched_rt_handler(struct ctl_table *table, int write,
8978
		void __user *buffer, size_t *lenp,
8979 8980 8981 8982 8983 8984 8985 8986 8987 8988
		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;

8989
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
8990 8991 8992 8993 8994 8995 8996 8997 8998 8999 9000 9001 9002 9003 9004 9005

	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;
}
9006

9007
#ifdef CONFIG_CGROUP_SCHED
9008 9009

/* return corresponding task_group object of a cgroup */
9010
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
9011
{
9012 9013
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
9014 9015 9016
}

static struct cgroup_subsys_state *
9017
cpu_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
9018
{
9019
	struct task_group *tg, *parent;
9020

9021
	if (!cgrp->parent) {
9022
		/* This is early initialization for the top cgroup */
9023
		return &root_task_group.css;
9024 9025
	}

9026 9027
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
9028 9029 9030 9031 9032 9033
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

I
Ingo Molnar 已提交
9034 9035
static void
cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
9036
{
9037
	struct task_group *tg = cgroup_tg(cgrp);
9038 9039 9040 9041

	sched_destroy_group(tg);
}

I
Ingo Molnar 已提交
9042
static int
9043
cpu_cgroup_can_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
9044
{
9045
#ifdef CONFIG_RT_GROUP_SCHED
9046
	if (!sched_rt_can_attach(cgroup_tg(cgrp), tsk))
9047 9048
		return -EINVAL;
#else
9049 9050 9051
	/* We don't support RT-tasks being in separate groups */
	if (tsk->sched_class != &fair_sched_class)
		return -EINVAL;
9052
#endif
9053 9054
	return 0;
}
9055

9056 9057 9058 9059 9060 9061 9062 9063 9064 9065 9066 9067 9068 9069 9070 9071 9072 9073 9074
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();
	}
9075 9076 9077 9078
	return 0;
}

static void
9079
cpu_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
9080 9081
		  struct cgroup *old_cont, struct task_struct *tsk,
		  bool threadgroup)
9082 9083
{
	sched_move_task(tsk);
9084 9085 9086 9087 9088 9089 9090 9091
	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();
	}
9092 9093
}

9094
static void
9095 9096
cpu_cgroup_exit(struct cgroup_subsys *ss, struct cgroup *cgrp,
		struct cgroup *old_cgrp, struct task_struct *task)
9097 9098 9099 9100 9101 9102 9103 9104 9105 9106 9107 9108
{
	/*
	 * cgroup_exit() is called in the copy_process() failure path.
	 * Ignore this case since the task hasn't ran yet, this avoids
	 * trying to poke a half freed task state from generic code.
	 */
	if (!(task->flags & PF_EXITING))
		return;

	sched_move_task(task);
}

9109
#ifdef CONFIG_FAIR_GROUP_SCHED
9110
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
9111
				u64 shareval)
9112
{
9113
	return sched_group_set_shares(cgroup_tg(cgrp), shareval);
9114 9115
}

9116
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
9117
{
9118
	struct task_group *tg = cgroup_tg(cgrp);
9119 9120 9121

	return (u64) tg->shares;
}
9122
#endif /* CONFIG_FAIR_GROUP_SCHED */
9123

9124
#ifdef CONFIG_RT_GROUP_SCHED
M
Mirco Tischler 已提交
9125
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
9126
				s64 val)
P
Peter Zijlstra 已提交
9127
{
9128
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
P
Peter Zijlstra 已提交
9129 9130
}

9131
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
P
Peter Zijlstra 已提交
9132
{
9133
	return sched_group_rt_runtime(cgroup_tg(cgrp));
P
Peter Zijlstra 已提交
9134
}
9135 9136 9137 9138 9139 9140 9141 9142 9143 9144 9145

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));
}
9146
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
9147

9148
static struct cftype cpu_files[] = {
9149
#ifdef CONFIG_FAIR_GROUP_SCHED
9150 9151
	{
		.name = "shares",
9152 9153
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
9154
	},
9155 9156
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
9157
	{
P
Peter Zijlstra 已提交
9158
		.name = "rt_runtime_us",
9159 9160
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
9161
	},
9162 9163
	{
		.name = "rt_period_us",
9164 9165
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
9166
	},
9167
#endif
9168 9169 9170 9171
};

static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
{
9172
	return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
9173 9174 9175
}

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
9176 9177 9178 9179 9180
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
9181
	.exit		= cpu_cgroup_exit,
I
Ingo Molnar 已提交
9182 9183
	.populate	= cpu_cgroup_populate,
	.subsys_id	= cpu_cgroup_subsys_id,
9184 9185 9186
	.early_init	= 1,
};

9187
#endif	/* CONFIG_CGROUP_SCHED */
9188 9189 9190 9191 9192 9193 9194 9195 9196 9197

#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).
 */

9198
/* track cpu usage of a group of tasks and its child groups */
9199 9200 9201
struct cpuacct {
	struct cgroup_subsys_state css;
	/* cpuusage holds pointer to a u64-type object on every cpu */
9202
	u64 __percpu *cpuusage;
9203
	struct percpu_counter cpustat[CPUACCT_STAT_NSTATS];
9204
	struct cpuacct *parent;
9205 9206 9207 9208 9209
};

struct cgroup_subsys cpuacct_subsys;

/* return cpu accounting group corresponding to this container */
9210
static inline struct cpuacct *cgroup_ca(struct cgroup *cgrp)
9211
{
9212
	return container_of(cgroup_subsys_state(cgrp, cpuacct_subsys_id),
9213 9214 9215 9216 9217 9218 9219 9220 9221 9222 9223 9224
			    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(
9225
	struct cgroup_subsys *ss, struct cgroup *cgrp)
9226 9227
{
	struct cpuacct *ca = kzalloc(sizeof(*ca), GFP_KERNEL);
9228
	int i;
9229 9230

	if (!ca)
9231
		goto out;
9232 9233

	ca->cpuusage = alloc_percpu(u64);
9234 9235 9236 9237 9238 9239
	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;
9240

9241 9242 9243
	if (cgrp->parent)
		ca->parent = cgroup_ca(cgrp->parent);

9244
	return &ca->css;
9245 9246 9247 9248 9249 9250 9251 9252 9253

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);
9254 9255 9256
}

/* destroy an existing cpu accounting group */
I
Ingo Molnar 已提交
9257
static void
9258
cpuacct_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
9259
{
9260
	struct cpuacct *ca = cgroup_ca(cgrp);
9261
	int i;
9262

9263 9264
	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
		percpu_counter_destroy(&ca->cpustat[i]);
9265 9266 9267 9268
	free_percpu(ca->cpuusage);
	kfree(ca);
}

9269 9270
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
9271
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
9272 9273 9274 9275 9276 9277
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
9278
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
9279
	data = *cpuusage;
9280
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
9281 9282 9283 9284 9285 9286 9287 9288 9289
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
9290
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
9291 9292 9293 9294 9295

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
9296
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
9297
	*cpuusage = val;
9298
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
9299 9300 9301 9302 9303
#else
	*cpuusage = val;
#endif
}

9304
/* return total cpu usage (in nanoseconds) of a group */
9305
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
9306
{
9307
	struct cpuacct *ca = cgroup_ca(cgrp);
9308 9309 9310
	u64 totalcpuusage = 0;
	int i;

9311 9312
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
9313 9314 9315 9316

	return totalcpuusage;
}

9317 9318 9319 9320 9321 9322 9323 9324 9325 9326 9327 9328
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;
	}

9329 9330
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
9331 9332 9333 9334 9335

out:
	return err;
}

9336 9337 9338 9339 9340 9341 9342 9343 9344 9345 9346 9347 9348 9349 9350
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;
}

9351 9352 9353 9354 9355 9356 9357 9358 9359 9360 9361 9362 9363 9364 9365 9366 9367 9368 9369
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;
}

9370 9371 9372
static struct cftype files[] = {
	{
		.name = "usage",
9373 9374
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
9375
	},
9376 9377 9378 9379
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
9380 9381 9382 9383
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
9384 9385
};

9386
static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
9387
{
9388
	return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
9389 9390 9391 9392 9393 9394 9395 9396 9397 9398
}

/*
 * 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;
9399
	int cpu;
9400

L
Li Zefan 已提交
9401
	if (unlikely(!cpuacct_subsys.active))
9402 9403
		return;

9404
	cpu = task_cpu(tsk);
9405 9406 9407

	rcu_read_lock();

9408 9409
	ca = task_ca(tsk);

9410
	for (; ca; ca = ca->parent) {
9411
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
9412 9413
		*cpuusage += cputime;
	}
9414 9415

	rcu_read_unlock();
9416 9417
}

9418 9419 9420 9421 9422 9423 9424 9425 9426 9427 9428 9429 9430 9431 9432 9433 9434
/*
 * 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

9435 9436 9437 9438 9439 9440 9441
/*
 * 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;
9442
	int batch = CPUACCT_BATCH;
9443 9444 9445 9446 9447 9448 9449 9450

	if (unlikely(!cpuacct_subsys.active))
		return;

	rcu_read_lock();
	ca = task_ca(tsk);

	do {
9451
		__percpu_counter_add(&ca->cpustat[idx], val, batch);
9452 9453 9454 9455 9456
		ca = ca->parent;
	} while (ca);
	rcu_read_unlock();
}

9457 9458 9459 9460 9461 9462 9463 9464
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 */
9465