sched.c 206.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 *  kernel/sched.c
 *
 *  Kernel scheduler and related syscalls
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 *
 *  1996-12-23  Modified by Dave Grothe to fix bugs in semaphores and
 *		make semaphores SMP safe
 *  1998-11-19	Implemented schedule_timeout() and related stuff
 *		by Andrea Arcangeli
 *  2002-01-04	New ultra-scalable O(1) scheduler by Ingo Molnar:
 *		hybrid priority-list and round-robin design with
 *		an array-switch method of distributing timeslices
 *		and per-CPU runqueues.  Cleanups and useful suggestions
 *		by Davide Libenzi, preemptible kernel bits by Robert Love.
 *  2003-09-03	Interactivity tuning by Con Kolivas.
 *  2004-04-02	Scheduler domains code by Nick Piggin
I
Ingo Molnar 已提交
19 20 21 22 23 24
 *  2007-04-15  Work begun on replacing all interactivity tuning with a
 *              fair scheduling design by Con Kolivas.
 *  2007-05-05  Load balancing (smp-nice) and other improvements
 *              by Peter Williams
 *  2007-05-06  Interactivity improvements to CFS by Mike Galbraith
 *  2007-07-01  Group scheduling enhancements by Srivatsa Vaddagiri
25 26
 *  2007-11-29  RT balancing improvements by Steven Rostedt, Gregory Haskins,
 *              Thomas Gleixner, Mike Kravetz
L
Linus Torvalds 已提交
27 28 29 30 31 32
 */

#include <linux/mm.h>
#include <linux/module.h>
#include <linux/nmi.h>
#include <linux/init.h>
33
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
34 35 36 37
#include <linux/highmem.h>
#include <linux/smp_lock.h>
#include <asm/mmu_context.h>
#include <linux/interrupt.h>
38
#include <linux/capability.h>
L
Linus Torvalds 已提交
39 40
#include <linux/completion.h>
#include <linux/kernel_stat.h>
41
#include <linux/debug_locks.h>
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 57 58
#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>
#include <linux/kthread.h>
#include <linux/seq_file.h>
59
#include <linux/sysctl.h>
L
Linus Torvalds 已提交
60 61
#include <linux/syscalls.h>
#include <linux/times.h>
62
#include <linux/tsacct_kern.h>
63
#include <linux/kprobes.h>
64
#include <linux/delayacct.h>
65
#include <linux/reciprocal_div.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>
70
#include <linux/bootmem.h>
L
Linus Torvalds 已提交
71

72
#include <asm/tlb.h>
73
#include <asm/irq_regs.h>
L
Linus Torvalds 已提交
74

75 76 77 78 79 80 81
/*
 * Scheduler clock - returns current time in nanosec units.
 * This is default implementation.
 * Architectures and sub-architectures can override this.
 */
unsigned long long __attribute__((weak)) sched_clock(void)
{
82
	return (unsigned long long)jiffies * (NSEC_PER_SEC / HZ);
83 84
}

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

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

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

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

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

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

124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
#ifdef CONFIG_SMP
/*
 * Divide a load by a sched group cpu_power : (load / sg->__cpu_power)
 * Since cpu_power is a 'constant', we can use a reciprocal divide.
 */
static inline u32 sg_div_cpu_power(const struct sched_group *sg, u32 load)
{
	return reciprocal_divide(load, sg->reciprocal_cpu_power);
}

/*
 * Each time a sched group cpu_power is changed,
 * we must compute its reciprocal value
 */
static inline void sg_inc_cpu_power(struct sched_group *sg, u32 val)
{
	sg->__cpu_power += val;
	sg->reciprocal_cpu_power = reciprocal_value(sg->__cpu_power);
}
#endif

145 146 147 148 149 150 151 152 153 154 155 156
static inline int rt_policy(int policy)
{
	if (unlikely(policy == SCHED_FIFO) || unlikely(policy == SCHED_RR))
		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
157
/*
I
Ingo Molnar 已提交
158
 * This is the priority-queue data structure of the RT scheduling class:
L
Linus Torvalds 已提交
159
 */
I
Ingo Molnar 已提交
160 161 162 163 164
struct rt_prio_array {
	DECLARE_BITMAP(bitmap, MAX_RT_PRIO+1); /* include 1 bit for delimiter */
	struct list_head queue[MAX_RT_PRIO];
};

165 166 167
struct rt_bandwidth {
	ktime_t rt_period;
	u64 rt_runtime;
P
Peter Zijlstra 已提交
168
	spinlock_t rt_runtime_lock;
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
	struct hrtimer rt_period_timer;
};

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;

P
Peter Zijlstra 已提交
203 204
	spin_lock_init(&rt_b->rt_runtime_lock);

205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
	hrtimer_init(&rt_b->rt_period_timer,
			CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rt_b->rt_period_timer.function = sched_rt_period_timer;
	rt_b->rt_period_timer.cb_mode = HRTIMER_CB_IRQSAFE_NO_SOFTIRQ;
}

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

	if (rt_b->rt_runtime == RUNTIME_INF)
		return;

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

	spin_lock(&rt_b->rt_runtime_lock);
	for (;;) {
		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);
		hrtimer_start(&rt_b->rt_period_timer,
			      rt_b->rt_period_timer.expires,
			      HRTIMER_MODE_ABS);
	}
	spin_unlock(&rt_b->rt_runtime_lock);
}

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

242
#ifdef CONFIG_GROUP_SCHED
S
Srivatsa Vaddagiri 已提交
243

244 245
#include <linux/cgroup.h>

S
Srivatsa Vaddagiri 已提交
246 247
struct cfs_rq;

P
Peter Zijlstra 已提交
248 249
static LIST_HEAD(task_groups);

S
Srivatsa Vaddagiri 已提交
250
/* task group related information */
251
struct task_group {
252
#ifdef CONFIG_CGROUP_SCHED
253 254
	struct cgroup_subsys_state css;
#endif
255 256

#ifdef CONFIG_FAIR_GROUP_SCHED
S
Srivatsa Vaddagiri 已提交
257 258 259 260 261
	/* 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;
262 263 264 265 266 267
#endif

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

268
	struct rt_bandwidth rt_bandwidth;
269
#endif
270

271
	struct rcu_head rcu;
P
Peter Zijlstra 已提交
272
	struct list_head list;
S
Srivatsa Vaddagiri 已提交
273 274
};

275
#ifdef CONFIG_FAIR_GROUP_SCHED
S
Srivatsa Vaddagiri 已提交
276 277 278 279
/* Default task group's sched entity on each cpu */
static DEFINE_PER_CPU(struct sched_entity, init_sched_entity);
/* Default task group's cfs_rq on each cpu */
static DEFINE_PER_CPU(struct cfs_rq, init_cfs_rq) ____cacheline_aligned_in_smp;
280 281 282 283 284 285
#endif

#ifdef CONFIG_RT_GROUP_SCHED
static DEFINE_PER_CPU(struct sched_rt_entity, init_sched_rt_entity);
static DEFINE_PER_CPU(struct rt_rq, init_rt_rq) ____cacheline_aligned_in_smp;
#endif
P
Peter Zijlstra 已提交
286

287
/* task_group_lock serializes add/remove of task groups and also changes to
288 289
 * a task group's cpu shares.
 */
290
static DEFINE_SPINLOCK(task_group_lock);
291

292 293 294
/* doms_cur_mutex serializes access to doms_cur[] array */
static DEFINE_MUTEX(doms_cur_mutex);

295 296 297 298 299 300 301 302 303 304
#ifdef CONFIG_FAIR_GROUP_SCHED
#ifdef CONFIG_USER_SCHED
# define INIT_TASK_GROUP_LOAD	(2*NICE_0_LOAD)
#else
# define INIT_TASK_GROUP_LOAD	NICE_0_LOAD
#endif

static int init_task_group_load = INIT_TASK_GROUP_LOAD;
#endif

S
Srivatsa Vaddagiri 已提交
305
/* Default task group.
I
Ingo Molnar 已提交
306
 *	Every task in system belong to this group at bootup.
S
Srivatsa Vaddagiri 已提交
307
 */
308
struct task_group init_task_group;
S
Srivatsa Vaddagiri 已提交
309 310

/* return group to which a task belongs */
311
static inline struct task_group *task_group(struct task_struct *p)
S
Srivatsa Vaddagiri 已提交
312
{
313
	struct task_group *tg;
314

315
#ifdef CONFIG_USER_SCHED
316
	tg = p->user->tg;
317
#elif defined(CONFIG_CGROUP_SCHED)
318 319
	tg = container_of(task_subsys_state(p, cpu_cgroup_subsys_id),
				struct task_group, css);
320
#else
I
Ingo Molnar 已提交
321
	tg = &init_task_group;
322
#endif
323
	return tg;
S
Srivatsa Vaddagiri 已提交
324 325 326
}

/* Change a task's cfs_rq and parent entity if it moves across CPUs/groups */
P
Peter Zijlstra 已提交
327
static inline void set_task_rq(struct task_struct *p, unsigned int cpu)
S
Srivatsa Vaddagiri 已提交
328
{
329
#ifdef CONFIG_FAIR_GROUP_SCHED
330 331
	p->se.cfs_rq = task_group(p)->cfs_rq[cpu];
	p->se.parent = task_group(p)->se[cpu];
332
#endif
P
Peter Zijlstra 已提交
333

334
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
335 336
	p->rt.rt_rq  = task_group(p)->rt_rq[cpu];
	p->rt.parent = task_group(p)->rt_se[cpu];
337
#endif
S
Srivatsa Vaddagiri 已提交
338 339
}

340 341 342 343 344 345 346 347 348 349
static inline void lock_doms_cur(void)
{
	mutex_lock(&doms_cur_mutex);
}

static inline void unlock_doms_cur(void)
{
	mutex_unlock(&doms_cur_mutex);
}

S
Srivatsa Vaddagiri 已提交
350 351
#else

P
Peter Zijlstra 已提交
352
static inline void set_task_rq(struct task_struct *p, unsigned int cpu) { }
353 354
static inline void lock_doms_cur(void) { }
static inline void unlock_doms_cur(void) { }
S
Srivatsa Vaddagiri 已提交
355

356
#endif	/* CONFIG_GROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
357

I
Ingo Molnar 已提交
358 359 360 361 362 363
/* 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
364
	u64 min_vruntime;
I
Ingo Molnar 已提交
365 366 367 368 369 370 371

	struct rb_root tasks_timeline;
	struct rb_node *rb_leftmost;
	struct rb_node *rb_load_balance_curr;
	/* 'curr' points to currently running entity on this cfs_rq.
	 * It is set to NULL otherwise (i.e when none are currently running).
	 */
372
	struct sched_entity *curr, *next;
P
Peter Zijlstra 已提交
373 374 375

	unsigned long nr_spread_over;

376
#ifdef CONFIG_FAIR_GROUP_SCHED
I
Ingo Molnar 已提交
377 378
	struct rq *rq;	/* cpu runqueue to which this cfs_rq is attached */

I
Ingo Molnar 已提交
379 380
	/*
	 * leaf cfs_rqs are those that hold tasks (lowest schedulable entity in
I
Ingo Molnar 已提交
381 382 383 384 385 386
	 * a hierarchy). Non-leaf lrqs hold other higher schedulable entities
	 * (like users, containers etc.)
	 *
	 * leaf_cfs_rq_list ties together list of leaf cfs_rq's in a cpu. This
	 * list is used during load balance.
	 */
I
Ingo Molnar 已提交
387 388
	struct list_head leaf_cfs_rq_list;
	struct task_group *tg;	/* group that "owns" this runqueue */
I
Ingo Molnar 已提交
389 390
#endif
};
L
Linus Torvalds 已提交
391

I
Ingo Molnar 已提交
392 393 394
/* Real-Time classes' related field in a runqueue: */
struct rt_rq {
	struct rt_prio_array active;
395
	unsigned long rt_nr_running;
396
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
397 398
	int highest_prio; /* highest queued rt task prio */
#endif
P
Peter Zijlstra 已提交
399
#ifdef CONFIG_SMP
400
	unsigned long rt_nr_migratory;
G
Gregory Haskins 已提交
401
	int overloaded;
P
Peter Zijlstra 已提交
402
#endif
P
Peter Zijlstra 已提交
403
	int rt_throttled;
P
Peter Zijlstra 已提交
404
	u64 rt_time;
P
Peter Zijlstra 已提交
405 406
	u64 rt_runtime;
	spinlock_t rt_runtime_lock;
P
Peter Zijlstra 已提交
407

408
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
409 410
	unsigned long rt_nr_boosted;

P
Peter Zijlstra 已提交
411 412 413 414 415
	struct rq *rq;
	struct list_head leaf_rt_rq_list;
	struct task_group *tg;
	struct sched_rt_entity *rt_se;
#endif
I
Ingo Molnar 已提交
416 417
};

G
Gregory Haskins 已提交
418 419 420 421
#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
422 423
 * 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
424 425 426 427 428 429 430 431
 * exclusive cpuset is created, we also create and attach a new root-domain
 * object.
 *
 */
struct root_domain {
	atomic_t refcount;
	cpumask_t span;
	cpumask_t online;
432

I
Ingo Molnar 已提交
433
	/*
434 435 436 437
	 * The "RT overload" flag: it gets set if a CPU has more than
	 * one runnable RT task.
	 */
	cpumask_t rto_mask;
I
Ingo Molnar 已提交
438
	atomic_t rto_count;
G
Gregory Haskins 已提交
439 440
};

441 442 443 444
/*
 * 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
445 446 447 448
static struct root_domain def_root_domain;

#endif

L
Linus Torvalds 已提交
449 450 451 452 453 454 455
/*
 * 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.
 */
456
struct rq {
457 458
	/* runqueue lock: */
	spinlock_t lock;
L
Linus Torvalds 已提交
459 460 461 462 463 464

	/*
	 * 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
465 466
	#define CPU_LOAD_IDX_MAX 5
	unsigned long cpu_load[CPU_LOAD_IDX_MAX];
467
	unsigned char idle_at_tick;
468
#ifdef CONFIG_NO_HZ
469
	unsigned long last_tick_seen;
470 471
	unsigned char in_nohz_recently;
#endif
472 473
	/* capture load from *all* tasks on this cpu: */
	struct load_weight load;
I
Ingo Molnar 已提交
474 475 476 477
	unsigned long nr_load_updates;
	u64 nr_switches;

	struct cfs_rq cfs;
P
Peter Zijlstra 已提交
478 479
	struct rt_rq rt;

I
Ingo Molnar 已提交
480
#ifdef CONFIG_FAIR_GROUP_SCHED
481 482
	/* list of leaf cfs_rq on this cpu: */
	struct list_head leaf_cfs_rq_list;
483 484
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
485
	struct list_head leaf_rt_rq_list;
L
Linus Torvalds 已提交
486 487 488 489 490 491 492 493 494 495
#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;

496
	struct task_struct *curr, *idle;
497
	unsigned long next_balance;
L
Linus Torvalds 已提交
498
	struct mm_struct *prev_mm;
I
Ingo Molnar 已提交
499 500 501 502

	u64 clock, prev_clock_raw;
	s64 clock_max_delta;

503
	unsigned int clock_warps, clock_overflows, clock_underflows;
504 505
	u64 idle_clock;
	unsigned int clock_deep_idle_events;
506
	u64 tick_timestamp;
I
Ingo Molnar 已提交
507

L
Linus Torvalds 已提交
508 509 510
	atomic_t nr_iowait;

#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
511
	struct root_domain *rd;
L
Linus Torvalds 已提交
512 513 514 515 516
	struct sched_domain *sd;

	/* For active balancing */
	int active_balance;
	int push_cpu;
517 518
	/* cpu of this runqueue: */
	int cpu;
L
Linus Torvalds 已提交
519

520
	struct task_struct *migration_thread;
L
Linus Torvalds 已提交
521 522 523
	struct list_head migration_queue;
#endif

P
Peter Zijlstra 已提交
524 525 526 527 528 529
#ifdef CONFIG_SCHED_HRTICK
	unsigned long hrtick_flags;
	ktime_t hrtick_expire;
	struct hrtimer hrtick_timer;
#endif

L
Linus Torvalds 已提交
530 531 532 533 534
#ifdef CONFIG_SCHEDSTATS
	/* latency stats */
	struct sched_info rq_sched_info;

	/* sys_sched_yield() stats */
535 536 537 538
	unsigned int yld_exp_empty;
	unsigned int yld_act_empty;
	unsigned int yld_both_empty;
	unsigned int yld_count;
L
Linus Torvalds 已提交
539 540

	/* schedule() stats */
541 542 543
	unsigned int sched_switch;
	unsigned int sched_count;
	unsigned int sched_goidle;
L
Linus Torvalds 已提交
544 545

	/* try_to_wake_up() stats */
546 547
	unsigned int ttwu_count;
	unsigned int ttwu_local;
I
Ingo Molnar 已提交
548 549

	/* BKL stats */
550
	unsigned int bkl_count;
L
Linus Torvalds 已提交
551
#endif
552
	struct lock_class_key rq_lock_key;
L
Linus Torvalds 已提交
553 554
};

555
static DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
L
Linus Torvalds 已提交
556

I
Ingo Molnar 已提交
557 558 559 560 561
static inline void check_preempt_curr(struct rq *rq, struct task_struct *p)
{
	rq->curr->sched_class->check_preempt_curr(rq, p);
}

562 563 564 565 566 567 568 569 570
static inline int cpu_of(struct rq *rq)
{
#ifdef CONFIG_SMP
	return rq->cpu;
#else
	return 0;
#endif
}

571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
#ifdef CONFIG_NO_HZ
static inline bool nohz_on(int cpu)
{
	return tick_get_tick_sched(cpu)->nohz_mode != NOHZ_MODE_INACTIVE;
}

static inline u64 max_skipped_ticks(struct rq *rq)
{
	return nohz_on(cpu_of(rq)) ? jiffies - rq->last_tick_seen + 2 : 1;
}

static inline void update_last_tick_seen(struct rq *rq)
{
	rq->last_tick_seen = jiffies;
}
#else
static inline u64 max_skipped_ticks(struct rq *rq)
{
	return 1;
}

static inline void update_last_tick_seen(struct rq *rq)
{
}
#endif

I
Ingo Molnar 已提交
597
/*
I
Ingo Molnar 已提交
598 599
 * Update the per-runqueue clock, as finegrained as the platform can give
 * us, but without assuming monotonicity, etc.:
I
Ingo Molnar 已提交
600
 */
I
Ingo Molnar 已提交
601
static void __update_rq_clock(struct rq *rq)
I
Ingo Molnar 已提交
602 603 604 605 606 607
{
	u64 prev_raw = rq->prev_clock_raw;
	u64 now = sched_clock();
	s64 delta = now - prev_raw;
	u64 clock = rq->clock;

I
Ingo Molnar 已提交
608 609 610
#ifdef CONFIG_SCHED_DEBUG
	WARN_ON_ONCE(cpu_of(rq) != smp_processor_id());
#endif
I
Ingo Molnar 已提交
611 612 613 614 615 616 617 618 619 620
	/*
	 * Protect against sched_clock() occasionally going backwards:
	 */
	if (unlikely(delta < 0)) {
		clock++;
		rq->clock_warps++;
	} else {
		/*
		 * Catch too large forward jumps too:
		 */
621 622 623 624 625 626
		u64 max_jump = max_skipped_ticks(rq) * TICK_NSEC;
		u64 max_time = rq->tick_timestamp + max_jump;

		if (unlikely(clock + delta > max_time)) {
			if (clock < max_time)
				clock = max_time;
627 628
			else
				clock++;
I
Ingo Molnar 已提交
629 630 631 632 633 634 635 636 637 638
			rq->clock_overflows++;
		} else {
			if (unlikely(delta > rq->clock_max_delta))
				rq->clock_max_delta = delta;
			clock += delta;
		}
	}

	rq->prev_clock_raw = now;
	rq->clock = clock;
I
Ingo Molnar 已提交
639
}
I
Ingo Molnar 已提交
640

I
Ingo Molnar 已提交
641 642 643 644
static void update_rq_clock(struct rq *rq)
{
	if (likely(smp_processor_id() == cpu_of(rq)))
		__update_rq_clock(rq);
I
Ingo Molnar 已提交
645 646
}

N
Nick Piggin 已提交
647 648
/*
 * The domain tree (rq->sd) is protected by RCU's quiescent state transition.
649
 * See detach_destroy_domains: synchronize_sched for details.
N
Nick Piggin 已提交
650 651 652 653
 *
 * The domain tree of any CPU may only be accessed from within
 * preempt-disabled sections.
 */
654 655
#define for_each_domain(cpu, __sd) \
	for (__sd = rcu_dereference(cpu_rq(cpu)->sd); __sd; __sd = __sd->parent)
L
Linus Torvalds 已提交
656 657 658 659 660 661

#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)

I
Ingo Molnar 已提交
662 663 664 665 666 667 668 669 670 671 672 673 674
/*
 * 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

/*
 * Debugging: various feature bits
 */
enum {
675
	SCHED_FEAT_NEW_FAIR_SLEEPERS	= 1,
I
Ingo Molnar 已提交
676 677
	SCHED_FEAT_WAKEUP_PREEMPT	= 2,
	SCHED_FEAT_START_DEBIT		= 4,
I
Ingo Molnar 已提交
678 679
	SCHED_FEAT_AFFINE_WAKEUPS	= 8,
	SCHED_FEAT_CACHE_HOT_BUDDY	= 16,
I
Ingo Molnar 已提交
680 681 682
	SCHED_FEAT_SYNC_WAKEUPS		= 32,
	SCHED_FEAT_HRTICK		= 64,
	SCHED_FEAT_DOUBLE_TICK		= 128,
P
Peter Zijlstra 已提交
683
	SCHED_FEAT_NORMALIZED_SLEEPER	= 256,
I
Ingo Molnar 已提交
684 685 686
};

const_debug unsigned int sysctl_sched_features =
I
Ingo Molnar 已提交
687
		SCHED_FEAT_NEW_FAIR_SLEEPERS	* 1 |
I
Ingo Molnar 已提交
688
		SCHED_FEAT_WAKEUP_PREEMPT	* 1 |
I
Ingo Molnar 已提交
689
		SCHED_FEAT_START_DEBIT		* 1 |
I
Ingo Molnar 已提交
690 691
		SCHED_FEAT_AFFINE_WAKEUPS	* 1 |
		SCHED_FEAT_CACHE_HOT_BUDDY	* 1 |
I
Ingo Molnar 已提交
692
		SCHED_FEAT_SYNC_WAKEUPS		* 1 |
P
Peter Zijlstra 已提交
693
		SCHED_FEAT_HRTICK		* 1 |
P
Peter Zijlstra 已提交
694 695
		SCHED_FEAT_DOUBLE_TICK		* 0 |
		SCHED_FEAT_NORMALIZED_SLEEPER	* 1;
I
Ingo Molnar 已提交
696 697 698

#define sched_feat(x) (sysctl_sched_features & SCHED_FEAT_##x)

699 700 701 702 703 704
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
const_debug unsigned int sysctl_sched_nr_migrate = 32;

P
Peter Zijlstra 已提交
705
/*
P
Peter Zijlstra 已提交
706
 * period over which we measure -rt task cpu usage in us.
P
Peter Zijlstra 已提交
707 708
 * default: 1s
 */
P
Peter Zijlstra 已提交
709
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
710

711 712
static __read_mostly int scheduler_running;

P
Peter Zijlstra 已提交
713 714 715 716 717
/*
 * 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
718

719 720 721 722 723 724 725 726 727 728 729 730
static inline u64 global_rt_period(void)
{
	return (u64)sysctl_sched_rt_period * NSEC_PER_USEC;
}

static inline u64 global_rt_runtime(void)
{
	if (sysctl_sched_rt_period < 0)
		return RUNTIME_INF;

	return (u64)sysctl_sched_rt_runtime * NSEC_PER_USEC;
}
P
Peter Zijlstra 已提交
731

732 733 734 735 736
static const unsigned long long time_sync_thresh = 100000;

static DEFINE_PER_CPU(unsigned long long, time_offset);
static DEFINE_PER_CPU(unsigned long long, prev_cpu_time);

737
/*
738 739 740 741
 * Global lock which we take every now and then to synchronize
 * the CPUs time. This method is not warp-safe, but it's good
 * enough to synchronize slowly diverging time sources and thus
 * it's good enough for tracing:
742
 */
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764
static DEFINE_SPINLOCK(time_sync_lock);
static unsigned long long prev_global_time;

static unsigned long long __sync_cpu_clock(cycles_t time, int cpu)
{
	unsigned long flags;

	spin_lock_irqsave(&time_sync_lock, flags);

	if (time < prev_global_time) {
		per_cpu(time_offset, cpu) += prev_global_time - time;
		time = prev_global_time;
	} else {
		prev_global_time = time;
	}

	spin_unlock_irqrestore(&time_sync_lock, flags);

	return time;
}

static unsigned long long __cpu_clock(int cpu)
765 766 767
{
	unsigned long long now;
	unsigned long flags;
I
Ingo Molnar 已提交
768
	struct rq *rq;
769

770 771 772 773
	/*
	 * Only call sched_clock() if the scheduler has already been
	 * initialized (some code might call cpu_clock() very early):
	 */
774 775 776 777 778 779
	if (unlikely(!scheduler_running))
		return 0;

	local_irq_save(flags);
	rq = cpu_rq(cpu);
	update_rq_clock(rq);
I
Ingo Molnar 已提交
780
	now = rq->clock;
781
	local_irq_restore(flags);
782 783 784

	return now;
}
785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802

/*
 * For kernel-internal use: high-speed (but slightly incorrect) per-cpu
 * clock constructed from sched_clock():
 */
unsigned long long cpu_clock(int cpu)
{
	unsigned long long prev_cpu_time, time, delta_time;

	prev_cpu_time = per_cpu(prev_cpu_time, cpu);
	time = __cpu_clock(cpu) + per_cpu(time_offset, cpu);
	delta_time = time-prev_cpu_time;

	if (unlikely(delta_time > time_sync_thresh))
		time = __sync_cpu_clock(time, cpu);

	return time;
}
P
Paul E. McKenney 已提交
803
EXPORT_SYMBOL_GPL(cpu_clock);
804

L
Linus Torvalds 已提交
805
#ifndef prepare_arch_switch
806 807 808 809 810 811
# define prepare_arch_switch(next)	do { } while (0)
#endif
#ifndef finish_arch_switch
# define finish_arch_switch(prev)	do { } while (0)
#endif

812 813 814 815 816
static inline int task_current(struct rq *rq, struct task_struct *p)
{
	return rq->curr == p;
}

817
#ifndef __ARCH_WANT_UNLOCKED_CTXSW
818
static inline int task_running(struct rq *rq, struct task_struct *p)
819
{
820
	return task_current(rq, p);
821 822
}

823
static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
824 825 826
{
}

827
static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
828
{
829 830 831 832
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
	rq->lock.owner = current;
#endif
833 834 835 836 837 838 839
	/*
	 * 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_);

840 841 842 843
	spin_unlock_irq(&rq->lock);
}

#else /* __ARCH_WANT_UNLOCKED_CTXSW */
844
static inline int task_running(struct rq *rq, struct task_struct *p)
845 846 847 848
{
#ifdef CONFIG_SMP
	return p->oncpu;
#else
849
	return task_current(rq, p);
850 851 852
#endif
}

853
static inline void prepare_lock_switch(struct rq *rq, struct task_struct *next)
854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
{
#ifdef CONFIG_SMP
	/*
	 * We can optimise this out completely for !SMP, because the
	 * SMP rebalancing from interrupt is the only thing that cares
	 * here.
	 */
	next->oncpu = 1;
#endif
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	spin_unlock_irq(&rq->lock);
#else
	spin_unlock(&rq->lock);
#endif
}

870
static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
871 872 873 874 875 876 877 878 879 880 881 882
{
#ifdef CONFIG_SMP
	/*
	 * After ->oncpu is cleared, the task can be moved to a different CPU.
	 * We must ensure this doesn't happen until the switch is completely
	 * finished.
	 */
	smp_wmb();
	prev->oncpu = 0;
#endif
#ifndef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
L
Linus Torvalds 已提交
883
#endif
884 885
}
#endif /* __ARCH_WANT_UNLOCKED_CTXSW */
L
Linus Torvalds 已提交
886

887 888 889 890
/*
 * __task_rq_lock - lock the runqueue a given task resides on.
 * Must be called interrupts disabled.
 */
891
static inline struct rq *__task_rq_lock(struct task_struct *p)
892 893
	__acquires(rq->lock)
{
894 895 896 897 898
	for (;;) {
		struct rq *rq = task_rq(p);
		spin_lock(&rq->lock);
		if (likely(rq == task_rq(p)))
			return rq;
899 900 901 902
		spin_unlock(&rq->lock);
	}
}

L
Linus Torvalds 已提交
903 904
/*
 * task_rq_lock - lock the runqueue a given task resides on and disable
I
Ingo Molnar 已提交
905
 * interrupts. Note the ordering: we can safely lookup the task_rq without
L
Linus Torvalds 已提交
906 907
 * explicitly disabling preemption.
 */
908
static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
L
Linus Torvalds 已提交
909 910
	__acquires(rq->lock)
{
911
	struct rq *rq;
L
Linus Torvalds 已提交
912

913 914 915 916 917 918
	for (;;) {
		local_irq_save(*flags);
		rq = task_rq(p);
		spin_lock(&rq->lock);
		if (likely(rq == task_rq(p)))
			return rq;
L
Linus Torvalds 已提交
919 920 921 922
		spin_unlock_irqrestore(&rq->lock, *flags);
	}
}

A
Alexey Dobriyan 已提交
923
static void __task_rq_unlock(struct rq *rq)
924 925 926 927 928
	__releases(rq->lock)
{
	spin_unlock(&rq->lock);
}

929
static inline void task_rq_unlock(struct rq *rq, unsigned long *flags)
L
Linus Torvalds 已提交
930 931 932 933 934 935
	__releases(rq->lock)
{
	spin_unlock_irqrestore(&rq->lock, *flags);
}

/*
936
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
937
 */
A
Alexey Dobriyan 已提交
938
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
939 940
	__acquires(rq->lock)
{
941
	struct rq *rq;
L
Linus Torvalds 已提交
942 943 944 945 946 947 948 949

	local_irq_disable();
	rq = this_rq();
	spin_lock(&rq->lock);

	return rq;
}

950
/*
951
 * We are going deep-idle (irqs are disabled):
952
 */
953
void sched_clock_idle_sleep_event(void)
954
{
955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970
	struct rq *rq = cpu_rq(smp_processor_id());

	spin_lock(&rq->lock);
	__update_rq_clock(rq);
	spin_unlock(&rq->lock);
	rq->clock_deep_idle_events++;
}
EXPORT_SYMBOL_GPL(sched_clock_idle_sleep_event);

/*
 * We just idled delta nanoseconds (called with irqs disabled):
 */
void sched_clock_idle_wakeup_event(u64 delta_ns)
{
	struct rq *rq = cpu_rq(smp_processor_id());
	u64 now = sched_clock();
971

972 973 974 975 976 977 978 979 980 981 982
	rq->idle_clock += delta_ns;
	/*
	 * Override the previous timestamp and ignore all
	 * sched_clock() deltas that occured while we idled,
	 * and use the PM-provided delta_ns to advance the
	 * rq clock:
	 */
	spin_lock(&rq->lock);
	rq->prev_clock_raw = now;
	rq->clock += delta_ns;
	spin_unlock(&rq->lock);
983
	touch_softlockup_watchdog();
984
}
985
EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event);
986

P
Peter Zijlstra 已提交
987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
static void __resched_task(struct task_struct *p, int tif_bit);

static inline void resched_task(struct task_struct *p)
{
	__resched_task(p, TIF_NEED_RESCHED);
}

#ifdef CONFIG_SCHED_HRTICK
/*
 * Use HR-timers to deliver accurate preemption points.
 *
 * Its all a bit involved since we cannot program an hrt while holding the
 * rq->lock. So what we do is store a state in in rq->hrtick_* and ask for a
 * reschedule event.
 *
 * When we get rescheduled we reprogram the hrtick_timer outside of the
 * rq->lock.
 */
static inline void resched_hrt(struct task_struct *p)
{
	__resched_task(p, TIF_HRTICK_RESCHED);
}

static inline void resched_rq(struct rq *rq)
{
	unsigned long flags;

	spin_lock_irqsave(&rq->lock, flags);
	resched_task(rq->curr);
	spin_unlock_irqrestore(&rq->lock, flags);
}

enum {
	HRTICK_SET,		/* re-programm hrtick_timer */
	HRTICK_RESET,		/* not a new slice */
};

/*
 * 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;
	return hrtimer_is_hres_active(&rq->hrtick_timer);
}

/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
static void hrtick_start(struct rq *rq, u64 delay, int reset)
{
	assert_spin_locked(&rq->lock);

	/*
	 * preempt at: now + delay
	 */
	rq->hrtick_expire =
		ktime_add_ns(rq->hrtick_timer.base->get_time(), delay);
	/*
	 * indicate we need to program the timer
	 */
	__set_bit(HRTICK_SET, &rq->hrtick_flags);
	if (reset)
		__set_bit(HRTICK_RESET, &rq->hrtick_flags);

	/*
	 * New slices are called from the schedule path and don't need a
	 * forced reschedule.
	 */
	if (reset)
		resched_hrt(rq->curr);
}

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

/*
 * Update the timer from the possible pending state.
 */
static void hrtick_set(struct rq *rq)
{
	ktime_t time;
	int set, reset;
	unsigned long flags;

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

	spin_lock_irqsave(&rq->lock, flags);
	set = __test_and_clear_bit(HRTICK_SET, &rq->hrtick_flags);
	reset = __test_and_clear_bit(HRTICK_RESET, &rq->hrtick_flags);
	time = rq->hrtick_expire;
	clear_thread_flag(TIF_HRTICK_RESCHED);
	spin_unlock_irqrestore(&rq->lock, flags);

	if (set) {
		hrtimer_start(&rq->hrtick_timer, time, HRTIMER_MODE_ABS);
		if (reset && !hrtimer_active(&rq->hrtick_timer))
			resched_rq(rq);
	} else
		hrtick_clear(rq);
}

/*
 * 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());

	spin_lock(&rq->lock);
	__update_rq_clock(rq);
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
	spin_unlock(&rq->lock);

	return HRTIMER_NORESTART;
}

static inline void init_rq_hrtick(struct rq *rq)
{
	rq->hrtick_flags = 0;
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
	rq->hrtick_timer.cb_mode = HRTIMER_CB_IRQSAFE_NO_SOFTIRQ;
}

void hrtick_resched(void)
{
	struct rq *rq;
	unsigned long flags;

	if (!test_thread_flag(TIF_HRTICK_RESCHED))
		return;

	local_irq_save(flags);
	rq = cpu_rq(smp_processor_id());
	hrtick_set(rq);
	local_irq_restore(flags);
}
#else
static inline void hrtick_clear(struct rq *rq)
{
}

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

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

void hrtick_resched(void)
{
}
#endif

I
Ingo Molnar 已提交
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
/*
 * 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

P
Peter Zijlstra 已提交
1167
static void __resched_task(struct task_struct *p, int tif_bit)
I
Ingo Molnar 已提交
1168 1169 1170 1171 1172
{
	int cpu;

	assert_spin_locked(&task_rq(p)->lock);

P
Peter Zijlstra 已提交
1173
	if (unlikely(test_tsk_thread_flag(p, tif_bit)))
I
Ingo Molnar 已提交
1174 1175
		return;

P
Peter Zijlstra 已提交
1176
	set_tsk_thread_flag(p, tif_bit);
I
Ingo Molnar 已提交
1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197

	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;

	if (!spin_trylock_irqsave(&rq->lock, flags))
		return;
	resched_task(cpu_curr(cpu));
	spin_unlock_irqrestore(&rq->lock, flags);
}
1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 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

#ifdef CONFIG_NO_HZ
/*
 * 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()
	 */
	set_tsk_thread_flag(rq->idle, TIF_NEED_RESCHED);

	/* NEED_RESCHED must be visible before we test polling */
	smp_mb();
	if (!tsk_is_polling(rq->idle))
		smp_send_reschedule(cpu);
}
#endif

I
Ingo Molnar 已提交
1241
#else
P
Peter Zijlstra 已提交
1242
static void __resched_task(struct task_struct *p, int tif_bit)
I
Ingo Molnar 已提交
1243 1244
{
	assert_spin_locked(&task_rq(p)->lock);
P
Peter Zijlstra 已提交
1245
	set_tsk_thread_flag(p, tif_bit);
I
Ingo Molnar 已提交
1246 1247 1248
}
#endif

1249 1250 1251 1252 1253 1254 1255 1256
#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
1257 1258 1259
/*
 * Shift right and round:
 */
I
Ingo Molnar 已提交
1260
#define SRR(x, y) (((x) + (1UL << ((y) - 1))) >> (y))
I
Ingo Molnar 已提交
1261

1262
static unsigned long
1263 1264 1265 1266 1267 1268
calc_delta_mine(unsigned long delta_exec, unsigned long weight,
		struct load_weight *lw)
{
	u64 tmp;

	if (unlikely(!lw->inv_weight))
I
Ingo Molnar 已提交
1269
		lw->inv_weight = (WMULT_CONST-lw->weight/2) / (lw->weight+1);
1270 1271 1272 1273 1274

	tmp = (u64)delta_exec * weight;
	/*
	 * Check whether we'd overflow the 64-bit multiplication:
	 */
I
Ingo Molnar 已提交
1275
	if (unlikely(tmp > WMULT_CONST))
I
Ingo Molnar 已提交
1276
		tmp = SRR(SRR(tmp, WMULT_SHIFT/2) * lw->inv_weight,
I
Ingo Molnar 已提交
1277 1278
			WMULT_SHIFT/2);
	else
I
Ingo Molnar 已提交
1279
		tmp = SRR(tmp * lw->inv_weight, WMULT_SHIFT);
1280

1281
	return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX);
1282 1283 1284 1285 1286 1287 1288 1289
}

static inline unsigned long
calc_delta_fair(unsigned long delta_exec, struct load_weight *lw)
{
	return calc_delta_mine(delta_exec, NICE_0_LOAD, lw);
}

1290
static inline void update_load_add(struct load_weight *lw, unsigned long inc)
1291 1292
{
	lw->weight += inc;
I
Ingo Molnar 已提交
1293
	lw->inv_weight = 0;
1294 1295
}

1296
static inline void update_load_sub(struct load_weight *lw, unsigned long dec)
1297 1298
{
	lw->weight -= dec;
I
Ingo Molnar 已提交
1299
	lw->inv_weight = 0;
1300 1301
}

1302 1303 1304 1305
/*
 * 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
1306
 * scheduling class and "nice" value. For SCHED_NORMAL tasks this is just a
1307 1308 1309 1310
 * scaled version of the new time slice allocation that they receive on time
 * slice expiry etc.
 */

I
Ingo Molnar 已提交
1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321
#define WEIGHT_IDLEPRIO		2
#define WMULT_IDLEPRIO		(1 << 31)

/*
 * 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
1322 1323 1324
 * 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
1325 1326
 */
static const int prio_to_weight[40] = {
1327 1328 1329 1330 1331 1332 1333 1334
 /* -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
1335 1336
};

1337 1338 1339 1340 1341 1342 1343
/*
 * 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
1344
static const u32 prio_to_wmult[40] = {
1345 1346 1347 1348 1349 1350 1351 1352
 /* -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
1353
};
1354

I
Ingo Molnar 已提交
1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367
static void activate_task(struct rq *rq, struct task_struct *p, int wakeup);

/*
 * runqueue iterator, to support SMP load-balancing between different
 * scheduling classes, without having to expose their internal data
 * structures to the load-balancing proper:
 */
struct rq_iterator {
	void *arg;
	struct task_struct *(*start)(void *);
	struct task_struct *(*next)(void *);
};

1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
#ifdef CONFIG_SMP
static unsigned long
balance_tasks(struct rq *this_rq, int this_cpu, struct rq *busiest,
	      unsigned long max_load_move, struct sched_domain *sd,
	      enum cpu_idle_type idle, int *all_pinned,
	      int *this_best_prio, struct rq_iterator *iterator);

static int
iter_move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest,
		   struct sched_domain *sd, enum cpu_idle_type idle,
		   struct rq_iterator *iterator);
#endif
I
Ingo Molnar 已提交
1380

1381 1382 1383 1384 1385 1386
#ifdef CONFIG_CGROUP_CPUACCT
static void cpuacct_charge(struct task_struct *tsk, u64 cputime);
#else
static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
#endif

1387 1388 1389 1390 1391 1392 1393
#ifdef CONFIG_SMP
static unsigned long source_load(int cpu, int type);
static unsigned long target_load(int cpu, int type);
static unsigned long cpu_avg_load_per_task(int cpu);
static int task_hot(struct task_struct *p, u64 now, struct sched_domain *sd);
#endif /* CONFIG_SMP */

I
Ingo Molnar 已提交
1394 1395
#include "sched_stats.h"
#include "sched_idletask.c"
1396 1397
#include "sched_fair.c"
#include "sched_rt.c"
I
Ingo Molnar 已提交
1398 1399 1400 1401 1402 1403
#ifdef CONFIG_SCHED_DEBUG
# include "sched_debug.c"
#endif

#define sched_class_highest (&rt_sched_class)

1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414
static inline void inc_load(struct rq *rq, const struct task_struct *p)
{
	update_load_add(&rq->load, p->se.load.weight);
}

static inline void dec_load(struct rq *rq, const struct task_struct *p)
{
	update_load_sub(&rq->load, p->se.load.weight);
}

static void inc_nr_running(struct task_struct *p, struct rq *rq)
1415 1416
{
	rq->nr_running++;
1417
	inc_load(rq, p);
1418 1419
}

1420
static void dec_nr_running(struct task_struct *p, struct rq *rq)
1421 1422
{
	rq->nr_running--;
1423
	dec_load(rq, p);
1424 1425
}

1426 1427 1428
static void set_load_weight(struct task_struct *p)
{
	if (task_has_rt_policy(p)) {
I
Ingo Molnar 已提交
1429 1430 1431 1432
		p->se.load.weight = prio_to_weight[0] * 2;
		p->se.load.inv_weight = prio_to_wmult[0] >> 1;
		return;
	}
1433

I
Ingo Molnar 已提交
1434 1435 1436 1437 1438 1439 1440 1441
	/*
	 * 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;
	}
1442

I
Ingo Molnar 已提交
1443 1444
	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];
1445 1446
}

1447
static void enqueue_task(struct rq *rq, struct task_struct *p, int wakeup)
1448
{
I
Ingo Molnar 已提交
1449
	sched_info_queued(p);
1450
	p->sched_class->enqueue_task(rq, p, wakeup);
I
Ingo Molnar 已提交
1451
	p->se.on_rq = 1;
1452 1453
}

1454
static void dequeue_task(struct rq *rq, struct task_struct *p, int sleep)
1455
{
1456
	p->sched_class->dequeue_task(rq, p, sleep);
I
Ingo Molnar 已提交
1457
	p->se.on_rq = 0;
1458 1459
}

1460
/*
I
Ingo Molnar 已提交
1461
 * __normal_prio - return the priority that is based on the static prio
1462 1463 1464
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
1465
	return p->static_prio;
1466 1467
}

1468 1469 1470 1471 1472 1473 1474
/*
 * 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.
 */
1475
static inline int normal_prio(struct task_struct *p)
1476 1477 1478
{
	int prio;

1479
	if (task_has_rt_policy(p))
1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
		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.
 */
1493
static int effective_prio(struct task_struct *p)
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
{
	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
1506
/*
I
Ingo Molnar 已提交
1507
 * activate_task - move a task to the runqueue.
L
Linus Torvalds 已提交
1508
 */
I
Ingo Molnar 已提交
1509
static void activate_task(struct rq *rq, struct task_struct *p, int wakeup)
L
Linus Torvalds 已提交
1510
{
1511
	if (task_contributes_to_load(p))
I
Ingo Molnar 已提交
1512
		rq->nr_uninterruptible--;
L
Linus Torvalds 已提交
1513

1514
	enqueue_task(rq, p, wakeup);
1515
	inc_nr_running(p, rq);
L
Linus Torvalds 已提交
1516 1517 1518 1519 1520
}

/*
 * deactivate_task - remove a task from the runqueue.
 */
1521
static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep)
L
Linus Torvalds 已提交
1522
{
1523
	if (task_contributes_to_load(p))
I
Ingo Molnar 已提交
1524 1525
		rq->nr_uninterruptible++;

1526
	dequeue_task(rq, p, sleep);
1527
	dec_nr_running(p, rq);
L
Linus Torvalds 已提交
1528 1529 1530 1531 1532 1533
}

/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
1534
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
1535 1536 1537 1538
{
	return cpu_curr(task_cpu(p)) == p;
}

1539 1540 1541
/* Used instead of source_load when we know the type == 0 */
unsigned long weighted_cpuload(const int cpu)
{
1542
	return cpu_rq(cpu)->load.weight;
I
Ingo Molnar 已提交
1543 1544 1545 1546
}

static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
{
P
Peter Zijlstra 已提交
1547
	set_task_rq(p, cpu);
I
Ingo Molnar 已提交
1548
#ifdef CONFIG_SMP
1549 1550 1551 1552 1553 1554
	/*
	 * 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();
I
Ingo Molnar 已提交
1555 1556
	task_thread_info(p)->cpu = cpu;
#endif
1557 1558
}

1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
				       int oldprio, int running)
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
			prev_class->switched_from(rq, p, running);
		p->sched_class->switched_to(rq, p, running);
	} else
		p->sched_class->prio_changed(rq, p, oldprio, running);
}

L
Linus Torvalds 已提交
1571
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
1572

1573 1574 1575
/*
 * Is this task likely cache-hot:
 */
1576
static int
1577 1578 1579 1580
task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
{
	s64 delta;

1581 1582 1583
	/*
	 * Buddy candidates are cache hot:
	 */
I
Ingo Molnar 已提交
1584
	if (sched_feat(CACHE_HOT_BUDDY) && (&p->se == cfs_rq_of(&p->se)->next))
1585 1586
		return 1;

1587 1588 1589
	if (p->sched_class != &fair_sched_class)
		return 0;

1590 1591 1592 1593 1594
	if (sysctl_sched_migration_cost == -1)
		return 1;
	if (sysctl_sched_migration_cost == 0)
		return 0;

1595 1596 1597 1598 1599 1600
	delta = now - p->se.exec_start;

	return delta < (s64)sysctl_sched_migration_cost;
}


I
Ingo Molnar 已提交
1601
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1602
{
I
Ingo Molnar 已提交
1603 1604
	int old_cpu = task_cpu(p);
	struct rq *old_rq = cpu_rq(old_cpu), *new_rq = cpu_rq(new_cpu);
1605 1606
	struct cfs_rq *old_cfsrq = task_cfs_rq(p),
		      *new_cfsrq = cpu_cfs_rq(old_cfsrq, new_cpu);
1607
	u64 clock_offset;
I
Ingo Molnar 已提交
1608 1609

	clock_offset = old_rq->clock - new_rq->clock;
I
Ingo Molnar 已提交
1610 1611 1612 1613

#ifdef CONFIG_SCHEDSTATS
	if (p->se.wait_start)
		p->se.wait_start -= clock_offset;
I
Ingo Molnar 已提交
1614 1615 1616 1617
	if (p->se.sleep_start)
		p->se.sleep_start -= clock_offset;
	if (p->se.block_start)
		p->se.block_start -= clock_offset;
1618 1619 1620 1621 1622
	if (old_cpu != new_cpu) {
		schedstat_inc(p, se.nr_migrations);
		if (task_hot(p, old_rq->clock, NULL))
			schedstat_inc(p, se.nr_forced2_migrations);
	}
I
Ingo Molnar 已提交
1623
#endif
1624 1625
	p->se.vruntime -= old_cfsrq->min_vruntime -
					 new_cfsrq->min_vruntime;
I
Ingo Molnar 已提交
1626 1627

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1628 1629
}

1630
struct migration_req {
L
Linus Torvalds 已提交
1631 1632
	struct list_head list;

1633
	struct task_struct *task;
L
Linus Torvalds 已提交
1634 1635 1636
	int dest_cpu;

	struct completion done;
1637
};
L
Linus Torvalds 已提交
1638 1639 1640 1641 1642

/*
 * The task's runqueue lock must be held.
 * Returns true if you have to wait for migration thread.
 */
1643
static int
1644
migrate_task(struct task_struct *p, int dest_cpu, struct migration_req *req)
L
Linus Torvalds 已提交
1645
{
1646
	struct rq *rq = task_rq(p);
L
Linus Torvalds 已提交
1647 1648 1649 1650 1651

	/*
	 * If the task is not on a runqueue (and not running), then
	 * it is sufficient to simply update the task's cpu field.
	 */
I
Ingo Molnar 已提交
1652
	if (!p->se.on_rq && !task_running(rq, p)) {
L
Linus Torvalds 已提交
1653 1654 1655 1656 1657 1658 1659 1660
		set_task_cpu(p, dest_cpu);
		return 0;
	}

	init_completion(&req->done);
	req->task = p;
	req->dest_cpu = dest_cpu;
	list_add(&req->list, &rq->migration_queue);
1661

L
Linus Torvalds 已提交
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673
	return 1;
}

/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
 * 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.
 */
1674
void wait_task_inactive(struct task_struct *p)
L
Linus Torvalds 已提交
1675 1676
{
	unsigned long flags;
I
Ingo Molnar 已提交
1677
	int running, on_rq;
1678
	struct rq *rq;
L
Linus Torvalds 已提交
1679

1680 1681 1682 1683 1684 1685 1686 1687
	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);
1688

1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
		/*
		 * 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!
		 */
		while (task_running(rq, p))
			cpu_relax();
1702

1703 1704 1705 1706 1707 1708 1709 1710 1711
		/*
		 * 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);
		running = task_running(rq, p);
		on_rq = p->se.on_rq;
		task_rq_unlock(rq, &flags);
1712

1713 1714 1715 1716 1717 1718 1719 1720 1721 1722
		/*
		 * 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;
		}
1723

1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
		 * So if it wa still runnable (but just not actively
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
		if (unlikely(on_rq)) {
			schedule_timeout_uninterruptible(1);
			continue;
		}
1737

1738 1739 1740 1741 1742 1743 1744
		/*
		 * 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;
	}
L
Linus Torvalds 已提交
1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759
}

/***
 * kick_process - kick a running thread to enter/exit the kernel
 * @p: the to-be-kicked thread
 *
 * Cause a process which is running on another CPU to enter
 * kernel-mode, without any delay. (to get signals handled.)
 *
 * NOTE: this function doesnt have to take the runqueue lock,
 * because all it wants to ensure is that the remote task enters
 * the kernel. If the IPI races and the task has been migrated
 * to another CPU then no harm is done and the purpose has been
 * achieved as well.
 */
1760
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}

/*
1772 1773
 * Return a low guess at the load of a migration-source cpu weighted
 * according to the scheduling class and "nice" value.
L
Linus Torvalds 已提交
1774 1775 1776 1777
 *
 * We want to under-estimate the load of migration sources, to
 * balance conservatively.
 */
A
Alexey Dobriyan 已提交
1778
static unsigned long source_load(int cpu, int type)
L
Linus Torvalds 已提交
1779
{
1780
	struct rq *rq = cpu_rq(cpu);
I
Ingo Molnar 已提交
1781
	unsigned long total = weighted_cpuload(cpu);
1782

1783
	if (type == 0)
I
Ingo Molnar 已提交
1784
		return total;
1785

I
Ingo Molnar 已提交
1786
	return min(rq->cpu_load[type-1], total);
L
Linus Torvalds 已提交
1787 1788 1789
}

/*
1790 1791
 * Return a high guess at the load of a migration-target cpu weighted
 * according to the scheduling class and "nice" value.
L
Linus Torvalds 已提交
1792
 */
A
Alexey Dobriyan 已提交
1793
static unsigned long target_load(int cpu, int type)
L
Linus Torvalds 已提交
1794
{
1795
	struct rq *rq = cpu_rq(cpu);
I
Ingo Molnar 已提交
1796
	unsigned long total = weighted_cpuload(cpu);
1797

N
Nick Piggin 已提交
1798
	if (type == 0)
I
Ingo Molnar 已提交
1799
		return total;
1800

I
Ingo Molnar 已提交
1801
	return max(rq->cpu_load[type-1], total);
1802 1803 1804 1805 1806
}

/*
 * Return the average load per task on the cpu's run queue
 */
1807
static unsigned long cpu_avg_load_per_task(int cpu)
1808
{
1809
	struct rq *rq = cpu_rq(cpu);
I
Ingo Molnar 已提交
1810
	unsigned long total = weighted_cpuload(cpu);
1811 1812
	unsigned long n = rq->nr_running;

I
Ingo Molnar 已提交
1813
	return n ? total / n : SCHED_LOAD_SCALE;
L
Linus Torvalds 已提交
1814 1815
}

N
Nick Piggin 已提交
1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832
/*
 * find_idlest_group finds and returns the least busy CPU group within the
 * domain.
 */
static struct sched_group *
find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu)
{
	struct sched_group *idlest = NULL, *this = NULL, *group = sd->groups;
	unsigned long min_load = ULONG_MAX, this_load = 0;
	int load_idx = sd->forkexec_idx;
	int imbalance = 100 + (sd->imbalance_pct-100)/2;

	do {
		unsigned long load, avg_load;
		int local_group;
		int i;

1833 1834
		/* Skip over this group if it has no CPUs allowed */
		if (!cpus_intersects(group->cpumask, p->cpus_allowed))
1835
			continue;
1836

N
Nick Piggin 已提交
1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
		local_group = cpu_isset(this_cpu, group->cpumask);

		/* Tally up the load of all CPUs in the group */
		avg_load = 0;

		for_each_cpu_mask(i, group->cpumask) {
			/* Bias balancing toward cpus of our domain */
			if (local_group)
				load = source_load(i, load_idx);
			else
				load = target_load(i, load_idx);

			avg_load += load;
		}

		/* Adjust by relative CPU power of the group */
1853 1854
		avg_load = sg_div_cpu_power(group,
				avg_load * SCHED_LOAD_SCALE);
N
Nick Piggin 已提交
1855 1856 1857 1858 1859 1860 1861 1862

		if (local_group) {
			this_load = avg_load;
			this = group;
		} else if (avg_load < min_load) {
			min_load = avg_load;
			idlest = group;
		}
1863
	} while (group = group->next, group != sd->groups);
N
Nick Piggin 已提交
1864 1865 1866 1867 1868 1869 1870

	if (!idlest || 100*this_load < imbalance*min_load)
		return NULL;
	return idlest;
}

/*
1871
 * find_idlest_cpu - find the idlest cpu among the cpus in group.
N
Nick Piggin 已提交
1872
 */
I
Ingo Molnar 已提交
1873
static int
1874 1875
find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu,
		cpumask_t *tmp)
N
Nick Piggin 已提交
1876 1877 1878 1879 1880
{
	unsigned long load, min_load = ULONG_MAX;
	int idlest = -1;
	int i;

1881
	/* Traverse only the allowed CPUs */
1882
	cpus_and(*tmp, group->cpumask, p->cpus_allowed);
1883

1884
	for_each_cpu_mask(i, *tmp) {
1885
		load = weighted_cpuload(i);
N
Nick Piggin 已提交
1886 1887 1888 1889 1890 1891 1892 1893 1894 1895

		if (load < min_load || (load == min_load && i == this_cpu)) {
			min_load = load;
			idlest = i;
		}
	}

	return idlest;
}

N
Nick Piggin 已提交
1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
/*
 * sched_balance_self: balance the current task (running on cpu) in domains
 * that have the 'flag' flag set. In practice, this is SD_BALANCE_FORK and
 * SD_BALANCE_EXEC.
 *
 * Balance, ie. select the least loaded group.
 *
 * Returns the target CPU number, or the same CPU if no balancing is needed.
 *
 * preempt must be disabled.
 */
static int sched_balance_self(int cpu, int flag)
{
	struct task_struct *t = current;
	struct sched_domain *tmp, *sd = NULL;
N
Nick Piggin 已提交
1911

1912
	for_each_domain(cpu, tmp) {
I
Ingo Molnar 已提交
1913 1914 1915
		/*
		 * If power savings logic is enabled for a domain, stop there.
		 */
1916 1917
		if (tmp->flags & SD_POWERSAVINGS_BALANCE)
			break;
N
Nick Piggin 已提交
1918 1919
		if (tmp->flags & flag)
			sd = tmp;
1920
	}
N
Nick Piggin 已提交
1921 1922

	while (sd) {
1923
		cpumask_t span, tmpmask;
N
Nick Piggin 已提交
1924
		struct sched_group *group;
1925 1926 1927 1928 1929 1930
		int new_cpu, weight;

		if (!(sd->flags & flag)) {
			sd = sd->child;
			continue;
		}
N
Nick Piggin 已提交
1931 1932 1933

		span = sd->span;
		group = find_idlest_group(sd, t, cpu);
1934 1935 1936 1937
		if (!group) {
			sd = sd->child;
			continue;
		}
N
Nick Piggin 已提交
1938

1939
		new_cpu = find_idlest_cpu(group, t, cpu, &tmpmask);
1940 1941 1942 1943 1944
		if (new_cpu == -1 || new_cpu == cpu) {
			/* Now try balancing at a lower domain level of cpu */
			sd = sd->child;
			continue;
		}
N
Nick Piggin 已提交
1945

1946
		/* Now try balancing at a lower domain level of new_cpu */
N
Nick Piggin 已提交
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
		cpu = new_cpu;
		sd = NULL;
		weight = cpus_weight(span);
		for_each_domain(cpu, tmp) {
			if (weight <= cpus_weight(tmp->span))
				break;
			if (tmp->flags & flag)
				sd = tmp;
		}
		/* while loop will break here if sd == NULL */
	}

	return cpu;
}

#endif /* CONFIG_SMP */
L
Linus Torvalds 已提交
1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977

/***
 * try_to_wake_up - wake up a thread
 * @p: the to-be-woken-up thread
 * @state: the mask of task states that can be woken
 * @sync: do a synchronous wakeup?
 *
 * 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.
 *
 * returns failure only if the task is already active.
 */
1978
static int try_to_wake_up(struct task_struct *p, unsigned int state, int sync)
L
Linus Torvalds 已提交
1979
{
1980
	int cpu, orig_cpu, this_cpu, success = 0;
L
Linus Torvalds 已提交
1981 1982
	unsigned long flags;
	long old_state;
1983
	struct rq *rq;
L
Linus Torvalds 已提交
1984

1985 1986 1987
	if (!sched_feat(SYNC_WAKEUPS))
		sync = 0;

1988
	smp_wmb();
L
Linus Torvalds 已提交
1989 1990 1991 1992 1993
	rq = task_rq_lock(p, &flags);
	old_state = p->state;
	if (!(old_state & state))
		goto out;

I
Ingo Molnar 已提交
1994
	if (p->se.on_rq)
L
Linus Torvalds 已提交
1995 1996 1997
		goto out_running;

	cpu = task_cpu(p);
1998
	orig_cpu = cpu;
L
Linus Torvalds 已提交
1999 2000 2001 2002 2003 2004
	this_cpu = smp_processor_id();

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

2005 2006 2007
	cpu = p->sched_class->select_task_rq(p, sync);
	if (cpu != orig_cpu) {
		set_task_cpu(p, cpu);
L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013
		task_rq_unlock(rq, &flags);
		/* might preempt at this point */
		rq = task_rq_lock(p, &flags);
		old_state = p->state;
		if (!(old_state & state))
			goto out;
I
Ingo Molnar 已提交
2014
		if (p->se.on_rq)
L
Linus Torvalds 已提交
2015 2016 2017 2018 2019 2020
			goto out_running;

		this_cpu = smp_processor_id();
		cpu = task_cpu(p);
	}

2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
#ifdef CONFIG_SCHEDSTATS
	schedstat_inc(rq, ttwu_count);
	if (cpu == this_cpu)
		schedstat_inc(rq, ttwu_local);
	else {
		struct sched_domain *sd;
		for_each_domain(this_cpu, sd) {
			if (cpu_isset(cpu, sd->span)) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
	}
#endif

L
Linus Torvalds 已提交
2036 2037
out_activate:
#endif /* CONFIG_SMP */
2038 2039 2040 2041 2042 2043 2044 2045 2046
	schedstat_inc(p, se.nr_wakeups);
	if (sync)
		schedstat_inc(p, se.nr_wakeups_sync);
	if (orig_cpu != cpu)
		schedstat_inc(p, se.nr_wakeups_migrate);
	if (cpu == this_cpu)
		schedstat_inc(p, se.nr_wakeups_local);
	else
		schedstat_inc(p, se.nr_wakeups_remote);
I
Ingo Molnar 已提交
2047
	update_rq_clock(rq);
I
Ingo Molnar 已提交
2048
	activate_task(rq, p, 1);
L
Linus Torvalds 已提交
2049 2050 2051
	success = 1;

out_running:
I
Ingo Molnar 已提交
2052 2053
	check_preempt_curr(rq, p);

L
Linus Torvalds 已提交
2054
	p->state = TASK_RUNNING;
2055 2056 2057 2058
#ifdef CONFIG_SMP
	if (p->sched_class->task_wake_up)
		p->sched_class->task_wake_up(rq, p);
#endif
L
Linus Torvalds 已提交
2059 2060 2061 2062 2063 2064
out:
	task_rq_unlock(rq, &flags);

	return success;
}

2065
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2066
{
2067
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
2068 2069 2070
}
EXPORT_SYMBOL(wake_up_process);

2071
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2072 2073 2074 2075 2076 2077 2078
{
	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
2079 2080 2081 2082 2083 2084 2085
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2086
	p->se.prev_sum_exec_runtime	= 0;
I
Ingo Molnar 已提交
2087 2088
	p->se.last_wakeup		= 0;
	p->se.avg_overlap		= 0;
I
Ingo Molnar 已提交
2089 2090 2091

#ifdef CONFIG_SCHEDSTATS
	p->se.wait_start		= 0;
I
Ingo Molnar 已提交
2092 2093 2094 2095 2096 2097
	p->se.sum_sleep_runtime		= 0;
	p->se.sleep_start		= 0;
	p->se.block_start		= 0;
	p->se.sleep_max			= 0;
	p->se.block_max			= 0;
	p->se.exec_max			= 0;
I
Ingo Molnar 已提交
2098
	p->se.slice_max			= 0;
I
Ingo Molnar 已提交
2099
	p->se.wait_max			= 0;
I
Ingo Molnar 已提交
2100
#endif
N
Nick Piggin 已提交
2101

P
Peter Zijlstra 已提交
2102
	INIT_LIST_HEAD(&p->rt.run_list);
I
Ingo Molnar 已提交
2103
	p->se.on_rq = 0;
N
Nick Piggin 已提交
2104

2105 2106 2107 2108
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif

L
Linus Torvalds 已提交
2109 2110 2111 2112 2113 2114 2115
	/*
	 * We mark the process as running here, but have not actually
	 * inserted it onto the runqueue yet. This guarantees that
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129
}

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

	__sched_fork(p);

#ifdef CONFIG_SMP
	cpu = sched_balance_self(cpu, SD_BALANCE_FORK);
#endif
I
Ingo Molnar 已提交
2130
	set_task_cpu(p, cpu);
2131 2132 2133 2134 2135

	/*
	 * Make sure we do not leak PI boosting priority to the child:
	 */
	p->prio = current->normal_prio;
H
Hiroshi Shimamoto 已提交
2136 2137
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
2138

2139
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
I
Ingo Molnar 已提交
2140
	if (likely(sched_info_on()))
2141
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2142
#endif
2143
#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
2144 2145
	p->oncpu = 0;
#endif
L
Linus Torvalds 已提交
2146
#ifdef CONFIG_PREEMPT
2147
	/* Want to start with kernel preemption disabled. */
A
Al Viro 已提交
2148
	task_thread_info(p)->preempt_count = 1;
L
Linus Torvalds 已提交
2149
#endif
N
Nick Piggin 已提交
2150
	put_cpu();
L
Linus Torvalds 已提交
2151 2152 2153 2154 2155 2156 2157 2158 2159
}

/*
 * 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.
 */
2160
void wake_up_new_task(struct task_struct *p, unsigned long clone_flags)
L
Linus Torvalds 已提交
2161 2162
{
	unsigned long flags;
I
Ingo Molnar 已提交
2163
	struct rq *rq;
L
Linus Torvalds 已提交
2164 2165

	rq = task_rq_lock(p, &flags);
N
Nick Piggin 已提交
2166
	BUG_ON(p->state != TASK_RUNNING);
I
Ingo Molnar 已提交
2167
	update_rq_clock(rq);
L
Linus Torvalds 已提交
2168 2169 2170

	p->prio = effective_prio(p);

2171
	if (!p->sched_class->task_new || !current->se.on_rq) {
I
Ingo Molnar 已提交
2172
		activate_task(rq, p, 0);
L
Linus Torvalds 已提交
2173 2174
	} else {
		/*
I
Ingo Molnar 已提交
2175 2176
		 * Let the scheduling class do new task startup
		 * management (if any):
L
Linus Torvalds 已提交
2177
		 */
2178
		p->sched_class->task_new(rq, p);
2179
		inc_nr_running(p, rq);
L
Linus Torvalds 已提交
2180
	}
I
Ingo Molnar 已提交
2181
	check_preempt_curr(rq, p);
2182 2183 2184 2185
#ifdef CONFIG_SMP
	if (p->sched_class->task_wake_up)
		p->sched_class->task_wake_up(rq, p);
#endif
I
Ingo Molnar 已提交
2186
	task_rq_unlock(rq, &flags);
L
Linus Torvalds 已提交
2187 2188
}

2189 2190 2191
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
R
Randy Dunlap 已提交
2192 2193
 * preempt_notifier_register - tell me when current is being being preempted & rescheduled
 * @notifier: notifier struct to register
2194 2195 2196 2197 2198 2199 2200 2201 2202
 */
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
2203
 * @notifier: notifier struct to unregister
2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246
 *
 * 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);
}

#else

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)
{
}

#endif

2247 2248 2249
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2250
 * @prev: the current task that is being switched out
2251 2252 2253 2254 2255 2256 2257 2258 2259
 * @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.
 */
2260 2261 2262
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2263
{
2264
	fire_sched_out_preempt_notifiers(prev, next);
2265 2266 2267 2268
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2269 2270
/**
 * finish_task_switch - clean up after a task-switch
2271
 * @rq: runqueue associated with task-switch
L
Linus Torvalds 已提交
2272 2273
 * @prev: the thread we just switched away from.
 *
2274 2275 2276 2277
 * 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
2278 2279
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2280
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2281 2282 2283
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
A
Alexey Dobriyan 已提交
2284
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
L
Linus Torvalds 已提交
2285 2286 2287
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2288
	long prev_state;
L
Linus Torvalds 已提交
2289 2290 2291 2292 2293

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2294
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2295 2296
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2297
	 * The test for TASK_DEAD must occur while the runqueue locks are
L
Linus Torvalds 已提交
2298 2299 2300 2301 2302
	 * 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
2303
	prev_state = prev->state;
2304 2305
	finish_arch_switch(prev);
	finish_lock_switch(rq, prev);
2306 2307 2308 2309
#ifdef CONFIG_SMP
	if (current->sched_class->post_schedule)
		current->sched_class->post_schedule(rq);
#endif
S
Steven Rostedt 已提交
2310

2311
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2312 2313
	if (mm)
		mmdrop(mm);
2314
	if (unlikely(prev_state == TASK_DEAD)) {
2315 2316 2317
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2318
		 */
2319
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
2320
		put_task_struct(prev);
2321
	}
L
Linus Torvalds 已提交
2322 2323 2324 2325 2326 2327
}

/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2328
asmlinkage void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2329 2330
	__releases(rq->lock)
{
2331 2332
	struct rq *rq = this_rq();

2333 2334 2335 2336 2337
	finish_task_switch(rq, prev);
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
L
Linus Torvalds 已提交
2338
	if (current->set_child_tid)
2339
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2340 2341 2342 2343 2344 2345
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
I
Ingo Molnar 已提交
2346
static inline void
2347
context_switch(struct rq *rq, struct task_struct *prev,
2348
	       struct task_struct *next)
L
Linus Torvalds 已提交
2349
{
I
Ingo Molnar 已提交
2350
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2351

2352
	prepare_task_switch(rq, prev, next);
I
Ingo Molnar 已提交
2353 2354
	mm = next->mm;
	oldmm = prev->active_mm;
2355 2356 2357 2358 2359 2360 2361
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
	arch_enter_lazy_cpu_mode();

I
Ingo Molnar 已提交
2362
	if (unlikely(!mm)) {
L
Linus Torvalds 已提交
2363 2364 2365 2366 2367 2368
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

I
Ingo Molnar 已提交
2369
	if (unlikely(!prev->mm)) {
L
Linus Torvalds 已提交
2370 2371 2372
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2373 2374 2375 2376 2377 2378 2379
	/*
	 * 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
2380
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2381
#endif
L
Linus Torvalds 已提交
2382 2383 2384 2385

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

I
Ingo Molnar 已提交
2386 2387 2388 2389 2390 2391 2392
	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
2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415
}

/*
 * 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;
}

unsigned long nr_uninterruptible(void)
{
	unsigned long i, sum = 0;

2416
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430
		sum += cpu_rq(i)->nr_uninterruptible;

	/*
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
	 */
	if (unlikely((long)sum < 0))
		sum = 0;

	return sum;
}

unsigned long long nr_context_switches(void)
{
2431 2432
	int i;
	unsigned long long sum = 0;
L
Linus Torvalds 已提交
2433

2434
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2435 2436 2437 2438 2439 2440 2441 2442 2443
		sum += cpu_rq(i)->nr_switches;

	return sum;
}

unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;

2444
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2445 2446 2447 2448 2449
		sum += atomic_read(&cpu_rq(i)->nr_iowait);

	return sum;
}

2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464
unsigned long nr_active(void)
{
	unsigned long i, running = 0, uninterruptible = 0;

	for_each_online_cpu(i) {
		running += cpu_rq(i)->nr_running;
		uninterruptible += cpu_rq(i)->nr_uninterruptible;
	}

	if (unlikely((long)uninterruptible < 0))
		uninterruptible = 0;

	return running + uninterruptible;
}

2465
/*
I
Ingo Molnar 已提交
2466 2467
 * Update rq->cpu_load[] statistics. This function is usually called every
 * scheduler tick (TICK_NSEC).
2468
 */
I
Ingo Molnar 已提交
2469
static void update_cpu_load(struct rq *this_rq)
2470
{
2471
	unsigned long this_load = this_rq->load.weight;
I
Ingo Molnar 已提交
2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
	int i, scale;

	this_rq->nr_load_updates++;

	/* Update our load: */
	for (i = 0, scale = 1; i < CPU_LOAD_IDX_MAX; i++, scale += scale) {
		unsigned long old_load, new_load;

		/* scale is effectively 1 << i now, and >> i divides by scale */

		old_load = this_rq->cpu_load[i];
		new_load = this_load;
I
Ingo Molnar 已提交
2484 2485 2486 2487 2488 2489 2490
		/*
		 * 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)
			new_load += scale-1;
I
Ingo Molnar 已提交
2491 2492
		this_rq->cpu_load[i] = (old_load*(scale-1) + new_load) >> i;
	}
2493 2494
}

I
Ingo Molnar 已提交
2495 2496
#ifdef CONFIG_SMP

L
Linus Torvalds 已提交
2497 2498 2499 2500 2501 2502
/*
 * 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.
 */
2503
static void double_rq_lock(struct rq *rq1, struct rq *rq2)
L
Linus Torvalds 已提交
2504 2505 2506
	__acquires(rq1->lock)
	__acquires(rq2->lock)
{
2507
	BUG_ON(!irqs_disabled());
L
Linus Torvalds 已提交
2508 2509 2510 2511
	if (rq1 == rq2) {
		spin_lock(&rq1->lock);
		__acquire(rq2->lock);	/* Fake it out ;) */
	} else {
2512
		if (rq1 < rq2) {
L
Linus Torvalds 已提交
2513 2514 2515 2516 2517 2518 2519
			spin_lock(&rq1->lock);
			spin_lock(&rq2->lock);
		} else {
			spin_lock(&rq2->lock);
			spin_lock(&rq1->lock);
		}
	}
2520 2521
	update_rq_clock(rq1);
	update_rq_clock(rq2);
L
Linus Torvalds 已提交
2522 2523 2524 2525 2526 2527 2528 2529
}

/*
 * 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.
 */
2530
static void double_rq_unlock(struct rq *rq1, struct rq *rq2)
L
Linus Torvalds 已提交
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
	__releases(rq1->lock)
	__releases(rq2->lock)
{
	spin_unlock(&rq1->lock);
	if (rq1 != rq2)
		spin_unlock(&rq2->lock);
	else
		__release(rq2->lock);
}

/*
 * double_lock_balance - lock the busiest runqueue, this_rq is locked already.
 */
S
Steven Rostedt 已提交
2544
static int double_lock_balance(struct rq *this_rq, struct rq *busiest)
L
Linus Torvalds 已提交
2545 2546 2547 2548
	__releases(this_rq->lock)
	__acquires(busiest->lock)
	__acquires(this_rq->lock)
{
S
Steven Rostedt 已提交
2549 2550
	int ret = 0;

2551 2552 2553 2554 2555
	if (unlikely(!irqs_disabled())) {
		/* printk() doesn't work good under rq->lock */
		spin_unlock(&this_rq->lock);
		BUG_ON(1);
	}
L
Linus Torvalds 已提交
2556
	if (unlikely(!spin_trylock(&busiest->lock))) {
2557
		if (busiest < this_rq) {
L
Linus Torvalds 已提交
2558 2559 2560
			spin_unlock(&this_rq->lock);
			spin_lock(&busiest->lock);
			spin_lock(&this_rq->lock);
S
Steven Rostedt 已提交
2561
			ret = 1;
L
Linus Torvalds 已提交
2562 2563 2564
		} else
			spin_lock(&busiest->lock);
	}
S
Steven Rostedt 已提交
2565
	return ret;
L
Linus Torvalds 已提交
2566 2567 2568 2569 2570
}

/*
 * If dest_cpu is allowed for this process, migrate the task to it.
 * This is accomplished by forcing the cpu_allowed mask to only
I
Ingo Molnar 已提交
2571
 * allow dest_cpu, which will force the cpu onto dest_cpu. Then
L
Linus Torvalds 已提交
2572 2573
 * the cpu_allowed mask is restored.
 */
2574
static void sched_migrate_task(struct task_struct *p, int dest_cpu)
L
Linus Torvalds 已提交
2575
{
2576
	struct migration_req req;
L
Linus Torvalds 已提交
2577
	unsigned long flags;
2578
	struct rq *rq;
L
Linus Torvalds 已提交
2579 2580 2581 2582 2583 2584 2585 2586 2587 2588

	rq = task_rq_lock(p, &flags);
	if (!cpu_isset(dest_cpu, p->cpus_allowed)
	    || unlikely(cpu_is_offline(dest_cpu)))
		goto out;

	/* force the process onto the specified CPU */
	if (migrate_task(p, dest_cpu, &req)) {
		/* Need to wait for migration thread (might exit: take ref). */
		struct task_struct *mt = rq->migration_thread;
2589

L
Linus Torvalds 已提交
2590 2591 2592 2593 2594
		get_task_struct(mt);
		task_rq_unlock(rq, &flags);
		wake_up_process(mt);
		put_task_struct(mt);
		wait_for_completion(&req.done);
2595

L
Linus Torvalds 已提交
2596 2597 2598 2599 2600 2601 2602
		return;
	}
out:
	task_rq_unlock(rq, &flags);
}

/*
N
Nick Piggin 已提交
2603 2604
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
L
Linus Torvalds 已提交
2605 2606 2607 2608
 */
void sched_exec(void)
{
	int new_cpu, this_cpu = get_cpu();
N
Nick Piggin 已提交
2609
	new_cpu = sched_balance_self(this_cpu, SD_BALANCE_EXEC);
L
Linus Torvalds 已提交
2610
	put_cpu();
N
Nick Piggin 已提交
2611 2612
	if (new_cpu != this_cpu)
		sched_migrate_task(current, new_cpu);
L
Linus Torvalds 已提交
2613 2614 2615 2616 2617 2618
}

/*
 * pull_task - move a task from a remote runqueue to the local runqueue.
 * Both runqueues must be locked.
 */
I
Ingo Molnar 已提交
2619 2620
static void pull_task(struct rq *src_rq, struct task_struct *p,
		      struct rq *this_rq, int this_cpu)
L
Linus Torvalds 已提交
2621
{
2622
	deactivate_task(src_rq, p, 0);
L
Linus Torvalds 已提交
2623
	set_task_cpu(p, this_cpu);
I
Ingo Molnar 已提交
2624
	activate_task(this_rq, p, 0);
L
Linus Torvalds 已提交
2625 2626 2627 2628
	/*
	 * Note that idle threads have a prio of MAX_PRIO, for this test
	 * to be always true for them.
	 */
I
Ingo Molnar 已提交
2629
	check_preempt_curr(this_rq, p);
L
Linus Torvalds 已提交
2630 2631 2632 2633 2634
}

/*
 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
 */
2635
static
2636
int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu,
I
Ingo Molnar 已提交
2637
		     struct sched_domain *sd, enum cpu_idle_type idle,
I
Ingo Molnar 已提交
2638
		     int *all_pinned)
L
Linus Torvalds 已提交
2639 2640 2641 2642 2643 2644 2645
{
	/*
	 * We do not migrate tasks that are:
	 * 1) running (obviously), or
	 * 2) cannot be migrated to this CPU due to cpus_allowed, or
	 * 3) are cache-hot on their current CPU.
	 */
2646 2647
	if (!cpu_isset(this_cpu, p->cpus_allowed)) {
		schedstat_inc(p, se.nr_failed_migrations_affine);
L
Linus Torvalds 已提交
2648
		return 0;
2649
	}
2650 2651
	*all_pinned = 0;

2652 2653
	if (task_running(rq, p)) {
		schedstat_inc(p, se.nr_failed_migrations_running);
2654
		return 0;
2655
	}
L
Linus Torvalds 已提交
2656

2657 2658 2659 2660 2661 2662
	/*
	 * Aggressive migration if:
	 * 1) task is cache cold, or
	 * 2) too many balance attempts have failed.
	 */

2663 2664
	if (!task_hot(p, rq->clock, sd) ||
			sd->nr_balance_failed > sd->cache_nice_tries) {
2665
#ifdef CONFIG_SCHEDSTATS
2666
		if (task_hot(p, rq->clock, sd)) {
2667
			schedstat_inc(sd, lb_hot_gained[idle]);
2668 2669
			schedstat_inc(p, se.nr_forced_migrations);
		}
2670 2671 2672 2673
#endif
		return 1;
	}

2674 2675
	if (task_hot(p, rq->clock, sd)) {
		schedstat_inc(p, se.nr_failed_migrations_hot);
2676
		return 0;
2677
	}
L
Linus Torvalds 已提交
2678 2679 2680
	return 1;
}

2681 2682 2683 2684 2685
static unsigned long
balance_tasks(struct rq *this_rq, int this_cpu, struct rq *busiest,
	      unsigned long max_load_move, struct sched_domain *sd,
	      enum cpu_idle_type idle, int *all_pinned,
	      int *this_best_prio, struct rq_iterator *iterator)
L
Linus Torvalds 已提交
2686
{
2687
	int loops = 0, pulled = 0, pinned = 0, skip_for_load;
I
Ingo Molnar 已提交
2688 2689
	struct task_struct *p;
	long rem_load_move = max_load_move;
L
Linus Torvalds 已提交
2690

2691
	if (max_load_move == 0)
L
Linus Torvalds 已提交
2692 2693
		goto out;

2694 2695
	pinned = 1;

L
Linus Torvalds 已提交
2696
	/*
I
Ingo Molnar 已提交
2697
	 * Start the load-balancing iterator:
L
Linus Torvalds 已提交
2698
	 */
I
Ingo Molnar 已提交
2699 2700
	p = iterator->start(iterator->arg);
next:
2701
	if (!p || loops++ > sysctl_sched_nr_migrate)
L
Linus Torvalds 已提交
2702
		goto out;
2703
	/*
2704
	 * To help distribute high priority tasks across CPUs we don't
2705 2706 2707
	 * skip a task if it will be the highest priority task (i.e. smallest
	 * prio value) on its new queue regardless of its load weight
	 */
I
Ingo Molnar 已提交
2708 2709
	skip_for_load = (p->se.load.weight >> 1) > rem_load_move +
							 SCHED_LOAD_SCALE_FUZZ;
2710
	if ((skip_for_load && p->prio >= *this_best_prio) ||
I
Ingo Molnar 已提交
2711 2712 2713
	    !can_migrate_task(p, busiest, this_cpu, sd, idle, &pinned)) {
		p = iterator->next(iterator->arg);
		goto next;
L
Linus Torvalds 已提交
2714 2715
	}

I
Ingo Molnar 已提交
2716
	pull_task(busiest, p, this_rq, this_cpu);
L
Linus Torvalds 已提交
2717
	pulled++;
I
Ingo Molnar 已提交
2718
	rem_load_move -= p->se.load.weight;
L
Linus Torvalds 已提交
2719

2720
	/*
2721
	 * We only want to steal up to the prescribed amount of weighted load.
2722
	 */
2723
	if (rem_load_move > 0) {
2724 2725
		if (p->prio < *this_best_prio)
			*this_best_prio = p->prio;
I
Ingo Molnar 已提交
2726 2727
		p = iterator->next(iterator->arg);
		goto next;
L
Linus Torvalds 已提交
2728 2729 2730
	}
out:
	/*
2731
	 * Right now, this is one of only two places pull_task() is called,
L
Linus Torvalds 已提交
2732 2733 2734 2735
	 * so we can safely collect pull_task() stats here rather than
	 * inside pull_task().
	 */
	schedstat_add(sd, lb_gained[idle], pulled);
2736 2737 2738

	if (all_pinned)
		*all_pinned = pinned;
2739 2740

	return max_load_move - rem_load_move;
L
Linus Torvalds 已提交
2741 2742
}

I
Ingo Molnar 已提交
2743
/*
P
Peter Williams 已提交
2744 2745 2746
 * move_tasks tries to move up to max_load_move weighted load from busiest to
 * this_rq, as part of a balancing operation within domain "sd".
 * Returns 1 if successful and 0 otherwise.
I
Ingo Molnar 已提交
2747 2748 2749 2750
 *
 * Called with both runqueues locked.
 */
static int move_tasks(struct rq *this_rq, int this_cpu, struct rq *busiest,
P
Peter Williams 已提交
2751
		      unsigned long max_load_move,
I
Ingo Molnar 已提交
2752 2753 2754
		      struct sched_domain *sd, enum cpu_idle_type idle,
		      int *all_pinned)
{
2755
	const struct sched_class *class = sched_class_highest;
P
Peter Williams 已提交
2756
	unsigned long total_load_moved = 0;
2757
	int this_best_prio = this_rq->curr->prio;
I
Ingo Molnar 已提交
2758 2759

	do {
P
Peter Williams 已提交
2760 2761
		total_load_moved +=
			class->load_balance(this_rq, this_cpu, busiest,
2762
				max_load_move - total_load_moved,
2763
				sd, idle, all_pinned, &this_best_prio);
I
Ingo Molnar 已提交
2764
		class = class->next;
P
Peter Williams 已提交
2765
	} while (class && max_load_move > total_load_moved);
I
Ingo Molnar 已提交
2766

P
Peter Williams 已提交
2767 2768 2769
	return total_load_moved > 0;
}

2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795
static int
iter_move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest,
		   struct sched_domain *sd, enum cpu_idle_type idle,
		   struct rq_iterator *iterator)
{
	struct task_struct *p = iterator->start(iterator->arg);
	int pinned = 0;

	while (p) {
		if (can_migrate_task(p, busiest, this_cpu, sd, idle, &pinned)) {
			pull_task(busiest, p, this_rq, this_cpu);
			/*
			 * Right now, this is only the second place pull_task()
			 * is called, so we can safely collect pull_task()
			 * stats here rather than inside pull_task().
			 */
			schedstat_inc(sd, lb_gained[idle]);

			return 1;
		}
		p = iterator->next(iterator->arg);
	}

	return 0;
}

P
Peter Williams 已提交
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805
/*
 * move_one_task tries to move exactly one task from busiest to this_rq, as
 * part of active balancing operations within "domain".
 * Returns 1 if successful and 0 otherwise.
 *
 * Called with both runqueues locked.
 */
static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest,
			 struct sched_domain *sd, enum cpu_idle_type idle)
{
2806
	const struct sched_class *class;
P
Peter Williams 已提交
2807 2808

	for (class = sched_class_highest; class; class = class->next)
2809
		if (class->move_one_task(this_rq, this_cpu, busiest, sd, idle))
P
Peter Williams 已提交
2810 2811 2812
			return 1;

	return 0;
I
Ingo Molnar 已提交
2813 2814
}

L
Linus Torvalds 已提交
2815 2816
/*
 * find_busiest_group finds and returns the busiest CPU group within the
2817 2818
 * domain. It calculates and returns the amount of weighted load which
 * should be moved to restore balance via the imbalance parameter.
L
Linus Torvalds 已提交
2819 2820 2821
 */
static struct sched_group *
find_busiest_group(struct sched_domain *sd, int this_cpu,
I
Ingo Molnar 已提交
2822
		   unsigned long *imbalance, enum cpu_idle_type idle,
2823
		   int *sd_idle, const cpumask_t *cpus, int *balance)
L
Linus Torvalds 已提交
2824 2825 2826
{
	struct sched_group *busiest = NULL, *this = NULL, *group = sd->groups;
	unsigned long max_load, avg_load, total_load, this_load, total_pwr;
2827
	unsigned long max_pull;
2828 2829
	unsigned long busiest_load_per_task, busiest_nr_running;
	unsigned long this_load_per_task, this_nr_running;
2830
	int load_idx, group_imb = 0;
2831 2832 2833 2834 2835 2836
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
	int power_savings_balance = 1;
	unsigned long leader_nr_running = 0, min_load_per_task = 0;
	unsigned long min_nr_running = ULONG_MAX;
	struct sched_group *group_min = NULL, *group_leader = NULL;
#endif
L
Linus Torvalds 已提交
2837 2838

	max_load = this_load = total_load = total_pwr = 0;
2839 2840
	busiest_load_per_task = busiest_nr_running = 0;
	this_load_per_task = this_nr_running = 0;
I
Ingo Molnar 已提交
2841
	if (idle == CPU_NOT_IDLE)
N
Nick Piggin 已提交
2842
		load_idx = sd->busy_idx;
I
Ingo Molnar 已提交
2843
	else if (idle == CPU_NEWLY_IDLE)
N
Nick Piggin 已提交
2844 2845 2846
		load_idx = sd->newidle_idx;
	else
		load_idx = sd->idle_idx;
L
Linus Torvalds 已提交
2847 2848

	do {
2849
		unsigned long load, group_capacity, max_cpu_load, min_cpu_load;
L
Linus Torvalds 已提交
2850 2851
		int local_group;
		int i;
2852
		int __group_imb = 0;
2853
		unsigned int balance_cpu = -1, first_idle_cpu = 0;
2854
		unsigned long sum_nr_running, sum_weighted_load;
L
Linus Torvalds 已提交
2855 2856 2857

		local_group = cpu_isset(this_cpu, group->cpumask);

2858 2859 2860
		if (local_group)
			balance_cpu = first_cpu(group->cpumask);

L
Linus Torvalds 已提交
2861
		/* Tally up the load of all CPUs in the group */
2862
		sum_weighted_load = sum_nr_running = avg_load = 0;
2863 2864
		max_cpu_load = 0;
		min_cpu_load = ~0UL;
L
Linus Torvalds 已提交
2865 2866

		for_each_cpu_mask(i, group->cpumask) {
2867 2868 2869 2870 2871 2872
			struct rq *rq;

			if (!cpu_isset(i, *cpus))
				continue;

			rq = cpu_rq(i);
2873

2874
			if (*sd_idle && rq->nr_running)
N
Nick Piggin 已提交
2875 2876
				*sd_idle = 0;

L
Linus Torvalds 已提交
2877
			/* Bias balancing toward cpus of our domain */
2878 2879 2880 2881 2882 2883
			if (local_group) {
				if (idle_cpu(i) && !first_idle_cpu) {
					first_idle_cpu = 1;
					balance_cpu = i;
				}

N
Nick Piggin 已提交
2884
				load = target_load(i, load_idx);
2885
			} else {
N
Nick Piggin 已提交
2886
				load = source_load(i, load_idx);
2887 2888 2889 2890 2891
				if (load > max_cpu_load)
					max_cpu_load = load;
				if (min_cpu_load > load)
					min_cpu_load = load;
			}
L
Linus Torvalds 已提交
2892 2893

			avg_load += load;
2894
			sum_nr_running += rq->nr_running;
I
Ingo Molnar 已提交
2895
			sum_weighted_load += weighted_cpuload(i);
L
Linus Torvalds 已提交
2896 2897
		}

2898 2899 2900
		/*
		 * First idle cpu or the first cpu(busiest) in this sched group
		 * is eligible for doing load balancing at this and above
2901 2902
		 * domains. In the newly idle case, we will allow all the cpu's
		 * to do the newly idle load balance.
2903
		 */
2904 2905
		if (idle != CPU_NEWLY_IDLE && local_group &&
		    balance_cpu != this_cpu && balance) {
2906 2907 2908 2909
			*balance = 0;
			goto ret;
		}

L
Linus Torvalds 已提交
2910
		total_load += avg_load;
2911
		total_pwr += group->__cpu_power;
L
Linus Torvalds 已提交
2912 2913

		/* Adjust by relative CPU power of the group */
2914 2915
		avg_load = sg_div_cpu_power(group,
				avg_load * SCHED_LOAD_SCALE);
L
Linus Torvalds 已提交
2916

2917 2918 2919
		if ((max_cpu_load - min_cpu_load) > SCHED_LOAD_SCALE)
			__group_imb = 1;

2920
		group_capacity = group->__cpu_power / SCHED_LOAD_SCALE;
2921

L
Linus Torvalds 已提交
2922 2923 2924
		if (local_group) {
			this_load = avg_load;
			this = group;
2925 2926 2927
			this_nr_running = sum_nr_running;
			this_load_per_task = sum_weighted_load;
		} else if (avg_load > max_load &&
2928
			   (sum_nr_running > group_capacity || __group_imb)) {
L
Linus Torvalds 已提交
2929 2930
			max_load = avg_load;
			busiest = group;
2931 2932
			busiest_nr_running = sum_nr_running;
			busiest_load_per_task = sum_weighted_load;
2933
			group_imb = __group_imb;
L
Linus Torvalds 已提交
2934
		}
2935 2936 2937 2938 2939 2940

#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
		/*
		 * Busy processors will not participate in power savings
		 * balance.
		 */
I
Ingo Molnar 已提交
2941 2942 2943
		if (idle == CPU_NOT_IDLE ||
				!(sd->flags & SD_POWERSAVINGS_BALANCE))
			goto group_next;
2944 2945 2946 2947 2948 2949 2950 2951 2952

		/*
		 * If the local group is idle or completely loaded
		 * no need to do power savings balance at this domain
		 */
		if (local_group && (this_nr_running >= group_capacity ||
				    !this_nr_running))
			power_savings_balance = 0;

I
Ingo Molnar 已提交
2953
		/*
2954 2955
		 * If a group is already running at full capacity or idle,
		 * don't include that group in power savings calculations
I
Ingo Molnar 已提交
2956 2957
		 */
		if (!power_savings_balance || sum_nr_running >= group_capacity
2958
		    || !sum_nr_running)
I
Ingo Molnar 已提交
2959
			goto group_next;
2960

I
Ingo Molnar 已提交
2961
		/*
2962
		 * Calculate the group which has the least non-idle load.
I
Ingo Molnar 已提交
2963 2964 2965 2966 2967
		 * This is the group from where we need to pick up the load
		 * for saving power
		 */
		if ((sum_nr_running < min_nr_running) ||
		    (sum_nr_running == min_nr_running &&
2968 2969
		     first_cpu(group->cpumask) <
		     first_cpu(group_min->cpumask))) {
I
Ingo Molnar 已提交
2970 2971
			group_min = group;
			min_nr_running = sum_nr_running;
2972 2973
			min_load_per_task = sum_weighted_load /
						sum_nr_running;
I
Ingo Molnar 已提交
2974
		}
2975

I
Ingo Molnar 已提交
2976
		/*
2977
		 * Calculate the group which is almost near its
I
Ingo Molnar 已提交
2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988
		 * capacity but still has some space to pick up some load
		 * from other group and save more power
		 */
		if (sum_nr_running <= group_capacity - 1) {
			if (sum_nr_running > leader_nr_running ||
			    (sum_nr_running == leader_nr_running &&
			     first_cpu(group->cpumask) >
			      first_cpu(group_leader->cpumask))) {
				group_leader = group;
				leader_nr_running = sum_nr_running;
			}
2989
		}
2990 2991
group_next:
#endif
L
Linus Torvalds 已提交
2992 2993 2994
		group = group->next;
	} while (group != sd->groups);

2995
	if (!busiest || this_load >= max_load || busiest_nr_running == 0)
L
Linus Torvalds 已提交
2996 2997 2998 2999 3000 3001 3002 3003
		goto out_balanced;

	avg_load = (SCHED_LOAD_SCALE * total_load) / total_pwr;

	if (this_load >= avg_load ||
			100*max_load <= sd->imbalance_pct*this_load)
		goto out_balanced;

3004
	busiest_load_per_task /= busiest_nr_running;
3005 3006 3007
	if (group_imb)
		busiest_load_per_task = min(busiest_load_per_task, avg_load);

L
Linus Torvalds 已提交
3008 3009 3010 3011 3012 3013 3014 3015
	/*
	 * We're trying to get all the cpus to the average_load, so we don't
	 * want to push ourselves above the average load, nor do we wish to
	 * reduce the max loaded cpu below the average load, as either of these
	 * actions would just result in more rebalancing later, and ping-pong
	 * tasks around. Thus we look for the minimum possible imbalance.
	 * Negative imbalances (*we* are more loaded than anyone else) will
	 * be counted as no imbalance for these purposes -- we can't fix that
I
Ingo Molnar 已提交
3016
	 * by pulling tasks to us. Be careful of negative numbers as they'll
L
Linus Torvalds 已提交
3017 3018
	 * appear as very large values with unsigned longs.
	 */
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030
	if (max_load <= busiest_load_per_task)
		goto out_balanced;

	/*
	 * In the presence of smp nice balancing, certain scenarios can have
	 * max load less than avg load(as we skip the groups at or below
	 * its cpu_power, while calculating max_load..)
	 */
	if (max_load < avg_load) {
		*imbalance = 0;
		goto small_imbalance;
	}
3031 3032

	/* Don't want to pull so many tasks that a group would go idle */
3033
	max_pull = min(max_load - avg_load, max_load - busiest_load_per_task);
3034

L
Linus Torvalds 已提交
3035
	/* How much load to actually move to equalise the imbalance */
3036 3037
	*imbalance = min(max_pull * busiest->__cpu_power,
				(avg_load - this_load) * this->__cpu_power)
L
Linus Torvalds 已提交
3038 3039
			/ SCHED_LOAD_SCALE;

3040 3041 3042 3043 3044 3045
	/*
	 * if *imbalance is less than the average load per runnable task
	 * there is no gaurantee that any tasks will be moved so we'll have
	 * a think about bumping its value to force at least one task to be
	 * moved
	 */
3046
	if (*imbalance < busiest_load_per_task) {
3047
		unsigned long tmp, pwr_now, pwr_move;
3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058
		unsigned int imbn;

small_imbalance:
		pwr_move = pwr_now = 0;
		imbn = 2;
		if (this_nr_running) {
			this_load_per_task /= this_nr_running;
			if (busiest_load_per_task > this_load_per_task)
				imbn = 1;
		} else
			this_load_per_task = SCHED_LOAD_SCALE;
L
Linus Torvalds 已提交
3059

I
Ingo Molnar 已提交
3060 3061
		if (max_load - this_load + SCHED_LOAD_SCALE_FUZZ >=
					busiest_load_per_task * imbn) {
3062
			*imbalance = busiest_load_per_task;
L
Linus Torvalds 已提交
3063 3064 3065 3066 3067 3068 3069 3070 3071
			return busiest;
		}

		/*
		 * OK, we don't have enough imbalance to justify moving tasks,
		 * however we may be able to increase total CPU power used by
		 * moving them.
		 */

3072 3073 3074 3075
		pwr_now += busiest->__cpu_power *
				min(busiest_load_per_task, max_load);
		pwr_now += this->__cpu_power *
				min(this_load_per_task, this_load);
L
Linus Torvalds 已提交
3076 3077 3078
		pwr_now /= SCHED_LOAD_SCALE;

		/* Amount of load we'd subtract */
3079 3080
		tmp = sg_div_cpu_power(busiest,
				busiest_load_per_task * SCHED_LOAD_SCALE);
L
Linus Torvalds 已提交
3081
		if (max_load > tmp)
3082
			pwr_move += busiest->__cpu_power *
3083
				min(busiest_load_per_task, max_load - tmp);
L
Linus Torvalds 已提交
3084 3085

		/* Amount of load we'd add */
3086
		if (max_load * busiest->__cpu_power <
3087
				busiest_load_per_task * SCHED_LOAD_SCALE)
3088 3089
			tmp = sg_div_cpu_power(this,
					max_load * busiest->__cpu_power);
L
Linus Torvalds 已提交
3090
		else
3091 3092 3093 3094
			tmp = sg_div_cpu_power(this,
				busiest_load_per_task * SCHED_LOAD_SCALE);
		pwr_move += this->__cpu_power *
				min(this_load_per_task, this_load + tmp);
L
Linus Torvalds 已提交
3095 3096 3097
		pwr_move /= SCHED_LOAD_SCALE;

		/* Move if we gain throughput */
3098 3099
		if (pwr_move > pwr_now)
			*imbalance = busiest_load_per_task;
L
Linus Torvalds 已提交
3100 3101 3102 3103 3104
	}

	return busiest;

out_balanced:
3105
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
I
Ingo Molnar 已提交
3106
	if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE))
3107
		goto ret;
L
Linus Torvalds 已提交
3108

3109 3110 3111 3112 3113
	if (this == group_leader && group_leader != group_min) {
		*imbalance = min_load_per_task;
		return group_min;
	}
#endif
3114
ret:
L
Linus Torvalds 已提交
3115 3116 3117 3118 3119 3120 3121
	*imbalance = 0;
	return NULL;
}

/*
 * find_busiest_queue - find the busiest runqueue among the cpus in group.
 */
3122
static struct rq *
I
Ingo Molnar 已提交
3123
find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle,
3124
		   unsigned long imbalance, const cpumask_t *cpus)
L
Linus Torvalds 已提交
3125
{
3126
	struct rq *busiest = NULL, *rq;
3127
	unsigned long max_load = 0;
L
Linus Torvalds 已提交
3128 3129 3130
	int i;

	for_each_cpu_mask(i, group->cpumask) {
I
Ingo Molnar 已提交
3131
		unsigned long wl;
3132 3133 3134 3135

		if (!cpu_isset(i, *cpus))
			continue;

3136
		rq = cpu_rq(i);
I
Ingo Molnar 已提交
3137
		wl = weighted_cpuload(i);
3138

I
Ingo Molnar 已提交
3139
		if (rq->nr_running == 1 && wl > imbalance)
3140
			continue;
L
Linus Torvalds 已提交
3141

I
Ingo Molnar 已提交
3142 3143
		if (wl > max_load) {
			max_load = wl;
3144
			busiest = rq;
L
Linus Torvalds 已提交
3145 3146 3147 3148 3149 3150
		}
	}

	return busiest;
}

3151 3152 3153 3154 3155 3156
/*
 * Max backoff if we encounter pinned tasks. Pretty arbitrary value, but
 * so long as it is large enough.
 */
#define MAX_PINNED_INTERVAL	512

L
Linus Torvalds 已提交
3157 3158 3159 3160
/*
 * Check this_cpu to ensure it is balanced within domain. Attempt to move
 * tasks if there is an imbalance.
 */
3161
static int load_balance(int this_cpu, struct rq *this_rq,
I
Ingo Molnar 已提交
3162
			struct sched_domain *sd, enum cpu_idle_type idle,
3163
			int *balance, cpumask_t *cpus)
L
Linus Torvalds 已提交
3164
{
P
Peter Williams 已提交
3165
	int ld_moved, all_pinned = 0, active_balance = 0, sd_idle = 0;
L
Linus Torvalds 已提交
3166 3167
	struct sched_group *group;
	unsigned long imbalance;
3168
	struct rq *busiest;
3169
	unsigned long flags;
N
Nick Piggin 已提交
3170

3171 3172
	cpus_setall(*cpus);

3173 3174 3175
	/*
	 * When power savings policy is enabled for the parent domain, idle
	 * sibling can pick up load irrespective of busy siblings. In this case,
I
Ingo Molnar 已提交
3176
	 * let the state of idle sibling percolate up as CPU_IDLE, instead of
I
Ingo Molnar 已提交
3177
	 * portraying it as CPU_NOT_IDLE.
3178
	 */
I
Ingo Molnar 已提交
3179
	if (idle != CPU_NOT_IDLE && sd->flags & SD_SHARE_CPUPOWER &&
3180
	    !test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
N
Nick Piggin 已提交
3181
		sd_idle = 1;
L
Linus Torvalds 已提交
3182

3183
	schedstat_inc(sd, lb_count[idle]);
L
Linus Torvalds 已提交
3184

3185 3186
redo:
	group = find_busiest_group(sd, this_cpu, &imbalance, idle, &sd_idle,
3187
				   cpus, balance);
3188

3189
	if (*balance == 0)
3190 3191
		goto out_balanced;

L
Linus Torvalds 已提交
3192 3193 3194 3195 3196
	if (!group) {
		schedstat_inc(sd, lb_nobusyg[idle]);
		goto out_balanced;
	}

3197
	busiest = find_busiest_queue(group, idle, imbalance, cpus);
L
Linus Torvalds 已提交
3198 3199 3200 3201 3202
	if (!busiest) {
		schedstat_inc(sd, lb_nobusyq[idle]);
		goto out_balanced;
	}

N
Nick Piggin 已提交
3203
	BUG_ON(busiest == this_rq);
L
Linus Torvalds 已提交
3204 3205 3206

	schedstat_add(sd, lb_imbalance[idle], imbalance);

P
Peter Williams 已提交
3207
	ld_moved = 0;
L
Linus Torvalds 已提交
3208 3209 3210 3211
	if (busiest->nr_running > 1) {
		/*
		 * Attempt to move tasks. If find_busiest_group has found
		 * an imbalance but busiest->nr_running <= 1, the group is
P
Peter Williams 已提交
3212
		 * still unbalanced. ld_moved simply stays zero, so it is
L
Linus Torvalds 已提交
3213 3214
		 * correctly treated as an imbalance.
		 */
3215
		local_irq_save(flags);
N
Nick Piggin 已提交
3216
		double_rq_lock(this_rq, busiest);
P
Peter Williams 已提交
3217
		ld_moved = move_tasks(this_rq, this_cpu, busiest,
3218
				      imbalance, sd, idle, &all_pinned);
N
Nick Piggin 已提交
3219
		double_rq_unlock(this_rq, busiest);
3220
		local_irq_restore(flags);
3221

3222 3223 3224
		/*
		 * some other cpu did the load balance for us.
		 */
P
Peter Williams 已提交
3225
		if (ld_moved && this_cpu != smp_processor_id())
3226 3227
			resched_cpu(this_cpu);

3228
		/* All tasks on this runqueue were pinned by CPU affinity */
3229
		if (unlikely(all_pinned)) {
3230 3231
			cpu_clear(cpu_of(busiest), *cpus);
			if (!cpus_empty(*cpus))
3232
				goto redo;
3233
			goto out_balanced;
3234
		}
L
Linus Torvalds 已提交
3235
	}
3236

P
Peter Williams 已提交
3237
	if (!ld_moved) {
L
Linus Torvalds 已提交
3238 3239 3240 3241 3242
		schedstat_inc(sd, lb_failed[idle]);
		sd->nr_balance_failed++;

		if (unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2)) {

3243
			spin_lock_irqsave(&busiest->lock, flags);
3244 3245 3246 3247 3248

			/* don't kick the migration_thread, if the curr
			 * task on busiest cpu can't be moved to this_cpu
			 */
			if (!cpu_isset(this_cpu, busiest->curr->cpus_allowed)) {
3249
				spin_unlock_irqrestore(&busiest->lock, flags);
3250 3251 3252 3253
				all_pinned = 1;
				goto out_one_pinned;
			}

L
Linus Torvalds 已提交
3254 3255 3256
			if (!busiest->active_balance) {
				busiest->active_balance = 1;
				busiest->push_cpu = this_cpu;
3257
				active_balance = 1;
L
Linus Torvalds 已提交
3258
			}
3259
			spin_unlock_irqrestore(&busiest->lock, flags);
3260
			if (active_balance)
L
Linus Torvalds 已提交
3261 3262 3263 3264 3265 3266
				wake_up_process(busiest->migration_thread);

			/*
			 * We've kicked active balancing, reset the failure
			 * counter.
			 */
3267
			sd->nr_balance_failed = sd->cache_nice_tries+1;
L
Linus Torvalds 已提交
3268
		}
3269
	} else
L
Linus Torvalds 已提交
3270 3271
		sd->nr_balance_failed = 0;

3272
	if (likely(!active_balance)) {
L
Linus Torvalds 已提交
3273 3274
		/* We were unbalanced, so reset the balancing interval */
		sd->balance_interval = sd->min_interval;
3275 3276 3277 3278 3279 3280 3281 3282 3283
	} else {
		/*
		 * If we've begun active balancing, start to back off. This
		 * case may not be covered by the all_pinned logic if there
		 * is only 1 task on the busy runqueue (because we don't call
		 * move_tasks).
		 */
		if (sd->balance_interval < sd->max_interval)
			sd->balance_interval *= 2;
L
Linus Torvalds 已提交
3284 3285
	}

P
Peter Williams 已提交
3286
	if (!ld_moved && !sd_idle && sd->flags & SD_SHARE_CPUPOWER &&
3287
	    !test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
N
Nick Piggin 已提交
3288
		return -1;
P
Peter Williams 已提交
3289
	return ld_moved;
L
Linus Torvalds 已提交
3290 3291 3292 3293

out_balanced:
	schedstat_inc(sd, lb_balanced[idle]);

3294
	sd->nr_balance_failed = 0;
3295 3296

out_one_pinned:
L
Linus Torvalds 已提交
3297
	/* tune up the balancing interval */
3298 3299
	if ((all_pinned && sd->balance_interval < MAX_PINNED_INTERVAL) ||
			(sd->balance_interval < sd->max_interval))
L
Linus Torvalds 已提交
3300 3301
		sd->balance_interval *= 2;

3302
	if (!sd_idle && sd->flags & SD_SHARE_CPUPOWER &&
3303
	    !test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
N
Nick Piggin 已提交
3304
		return -1;
L
Linus Torvalds 已提交
3305 3306 3307 3308 3309 3310 3311
	return 0;
}

/*
 * Check this_cpu to ensure it is balanced within domain. Attempt to move
 * tasks if there is an imbalance.
 *
I
Ingo Molnar 已提交
3312
 * Called from schedule when this_rq is about to become idle (CPU_NEWLY_IDLE).
L
Linus Torvalds 已提交
3313 3314
 * this_rq is locked.
 */
3315
static int
3316 3317
load_balance_newidle(int this_cpu, struct rq *this_rq, struct sched_domain *sd,
			cpumask_t *cpus)
L
Linus Torvalds 已提交
3318 3319
{
	struct sched_group *group;
3320
	struct rq *busiest = NULL;
L
Linus Torvalds 已提交
3321
	unsigned long imbalance;
P
Peter Williams 已提交
3322
	int ld_moved = 0;
N
Nick Piggin 已提交
3323
	int sd_idle = 0;
3324
	int all_pinned = 0;
3325 3326

	cpus_setall(*cpus);
N
Nick Piggin 已提交
3327

3328 3329 3330 3331
	/*
	 * When power savings policy is enabled for the parent domain, idle
	 * sibling can pick up load irrespective of busy siblings. In this case,
	 * let the state of idle sibling percolate up as IDLE, instead of
I
Ingo Molnar 已提交
3332
	 * portraying it as CPU_NOT_IDLE.
3333 3334 3335
	 */
	if (sd->flags & SD_SHARE_CPUPOWER &&
	    !test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
N
Nick Piggin 已提交
3336
		sd_idle = 1;
L
Linus Torvalds 已提交
3337

3338
	schedstat_inc(sd, lb_count[CPU_NEWLY_IDLE]);
3339
redo:
I
Ingo Molnar 已提交
3340
	group = find_busiest_group(sd, this_cpu, &imbalance, CPU_NEWLY_IDLE,
3341
				   &sd_idle, cpus, NULL);
L
Linus Torvalds 已提交
3342
	if (!group) {
I
Ingo Molnar 已提交
3343
		schedstat_inc(sd, lb_nobusyg[CPU_NEWLY_IDLE]);
3344
		goto out_balanced;
L
Linus Torvalds 已提交
3345 3346
	}

3347
	busiest = find_busiest_queue(group, CPU_NEWLY_IDLE, imbalance, cpus);
N
Nick Piggin 已提交
3348
	if (!busiest) {
I
Ingo Molnar 已提交
3349
		schedstat_inc(sd, lb_nobusyq[CPU_NEWLY_IDLE]);
3350
		goto out_balanced;
L
Linus Torvalds 已提交
3351 3352
	}

N
Nick Piggin 已提交
3353 3354
	BUG_ON(busiest == this_rq);

I
Ingo Molnar 已提交
3355
	schedstat_add(sd, lb_imbalance[CPU_NEWLY_IDLE], imbalance);
3356

P
Peter Williams 已提交
3357
	ld_moved = 0;
3358 3359 3360
	if (busiest->nr_running > 1) {
		/* Attempt to move tasks */
		double_lock_balance(this_rq, busiest);
3361 3362
		/* this_rq->clock is already updated */
		update_rq_clock(busiest);
P
Peter Williams 已提交
3363
		ld_moved = move_tasks(this_rq, this_cpu, busiest,
3364 3365
					imbalance, sd, CPU_NEWLY_IDLE,
					&all_pinned);
3366
		spin_unlock(&busiest->lock);
3367

3368
		if (unlikely(all_pinned)) {
3369 3370
			cpu_clear(cpu_of(busiest), *cpus);
			if (!cpus_empty(*cpus))
3371 3372
				goto redo;
		}
3373 3374
	}

P
Peter Williams 已提交
3375
	if (!ld_moved) {
I
Ingo Molnar 已提交
3376
		schedstat_inc(sd, lb_failed[CPU_NEWLY_IDLE]);
3377 3378
		if (!sd_idle && sd->flags & SD_SHARE_CPUPOWER &&
		    !test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
N
Nick Piggin 已提交
3379 3380
			return -1;
	} else
3381
		sd->nr_balance_failed = 0;
L
Linus Torvalds 已提交
3382

P
Peter Williams 已提交
3383
	return ld_moved;
3384 3385

out_balanced:
I
Ingo Molnar 已提交
3386
	schedstat_inc(sd, lb_balanced[CPU_NEWLY_IDLE]);
3387
	if (!sd_idle && sd->flags & SD_SHARE_CPUPOWER &&
3388
	    !test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
N
Nick Piggin 已提交
3389
		return -1;
3390
	sd->nr_balance_failed = 0;
3391

3392
	return 0;
L
Linus Torvalds 已提交
3393 3394 3395 3396 3397 3398
}

/*
 * idle_balance is called by schedule() if this_cpu is about to become
 * idle. Attempts to pull tasks from other CPUs.
 */
3399
static void idle_balance(int this_cpu, struct rq *this_rq)
L
Linus Torvalds 已提交
3400 3401
{
	struct sched_domain *sd;
I
Ingo Molnar 已提交
3402 3403
	int pulled_task = -1;
	unsigned long next_balance = jiffies + HZ;
3404
	cpumask_t tmpmask;
L
Linus Torvalds 已提交
3405 3406

	for_each_domain(this_cpu, sd) {
3407 3408 3409 3410 3411 3412
		unsigned long interval;

		if (!(sd->flags & SD_LOAD_BALANCE))
			continue;

		if (sd->flags & SD_BALANCE_NEWIDLE)
3413
			/* If we've pulled tasks over stop searching: */
3414 3415
			pulled_task = load_balance_newidle(this_cpu, this_rq,
							   sd, &tmpmask);
3416 3417 3418 3419 3420 3421

		interval = msecs_to_jiffies(sd->balance_interval);
		if (time_after(next_balance, sd->last_balance + interval))
			next_balance = sd->last_balance + interval;
		if (pulled_task)
			break;
L
Linus Torvalds 已提交
3422
	}
I
Ingo Molnar 已提交
3423
	if (pulled_task || time_after(jiffies, this_rq->next_balance)) {
3424 3425 3426 3427 3428
		/*
		 * We are going idle. next_balance may be set based on
		 * a busy processor. So reset next_balance.
		 */
		this_rq->next_balance = next_balance;
I
Ingo Molnar 已提交
3429
	}
L
Linus Torvalds 已提交
3430 3431 3432 3433 3434 3435 3436 3437 3438 3439
}

/*
 * active_load_balance is run by migration threads. It pushes running tasks
 * off the busiest CPU onto idle CPUs. It requires at least 1 task to be
 * running on each physical CPU where possible, and avoids physical /
 * logical imbalances.
 *
 * Called with busiest_rq locked.
 */
3440
static void active_load_balance(struct rq *busiest_rq, int busiest_cpu)
L
Linus Torvalds 已提交
3441
{
3442
	int target_cpu = busiest_rq->push_cpu;
3443 3444
	struct sched_domain *sd;
	struct rq *target_rq;
3445

3446
	/* Is there any task to move? */
3447 3448 3449 3450
	if (busiest_rq->nr_running <= 1)
		return;

	target_rq = cpu_rq(target_cpu);
L
Linus Torvalds 已提交
3451 3452

	/*
3453
	 * This condition is "impossible", if it occurs
I
Ingo Molnar 已提交
3454
	 * we need to fix it. Originally reported by
3455
	 * Bjorn Helgaas on a 128-cpu setup.
L
Linus Torvalds 已提交
3456
	 */
3457
	BUG_ON(busiest_rq == target_rq);
L
Linus Torvalds 已提交
3458

3459 3460
	/* move a task from busiest_rq to target_rq */
	double_lock_balance(busiest_rq, target_rq);
3461 3462
	update_rq_clock(busiest_rq);
	update_rq_clock(target_rq);
3463 3464

	/* Search for an sd spanning us and the target CPU. */
3465
	for_each_domain(target_cpu, sd) {
3466
		if ((sd->flags & SD_LOAD_BALANCE) &&
3467
		    cpu_isset(busiest_cpu, sd->span))
3468
				break;
3469
	}
3470

3471
	if (likely(sd)) {
3472
		schedstat_inc(sd, alb_count);
3473

P
Peter Williams 已提交
3474 3475
		if (move_one_task(target_rq, target_cpu, busiest_rq,
				  sd, CPU_IDLE))
3476 3477 3478 3479
			schedstat_inc(sd, alb_pushed);
		else
			schedstat_inc(sd, alb_failed);
	}
3480
	spin_unlock(&target_rq->lock);
L
Linus Torvalds 已提交
3481 3482
}

3483 3484 3485
#ifdef CONFIG_NO_HZ
static struct {
	atomic_t load_balancer;
I
Ingo Molnar 已提交
3486
	cpumask_t cpu_mask;
3487 3488 3489 3490 3491
} nohz ____cacheline_aligned = {
	.load_balancer = ATOMIC_INIT(-1),
	.cpu_mask = CPU_MASK_NONE,
};

3492
/*
3493 3494 3495 3496 3497 3498 3499 3500 3501 3502
 * This routine will try to nominate the ilb (idle load balancing)
 * owner among the cpus whose ticks are stopped. ilb owner will do the idle
 * load balancing on behalf of all those cpus. If all the cpus in the system
 * go into this tickless mode, then there will be no ilb owner (as there is
 * no need for one) and all the cpus will sleep till the next wakeup event
 * arrives...
 *
 * For the ilb owner, tick is not stopped. And this tick will be used
 * for idle load balancing. ilb owner will still be part of
 * nohz.cpu_mask..
3503
 *
3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559
 * While stopping the tick, this cpu will become the ilb owner if there
 * is no other owner. And will be the owner till that cpu becomes busy
 * or if all cpus in the system stop their ticks at which point
 * there is no need for ilb owner.
 *
 * When the ilb owner becomes busy, it nominates another owner, during the
 * next busy scheduler_tick()
 */
int select_nohz_load_balancer(int stop_tick)
{
	int cpu = smp_processor_id();

	if (stop_tick) {
		cpu_set(cpu, nohz.cpu_mask);
		cpu_rq(cpu)->in_nohz_recently = 1;

		/*
		 * If we are going offline and still the leader, give up!
		 */
		if (cpu_is_offline(cpu) &&
		    atomic_read(&nohz.load_balancer) == cpu) {
			if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu)
				BUG();
			return 0;
		}

		/* time for ilb owner also to sleep */
		if (cpus_weight(nohz.cpu_mask) == num_online_cpus()) {
			if (atomic_read(&nohz.load_balancer) == cpu)
				atomic_set(&nohz.load_balancer, -1);
			return 0;
		}

		if (atomic_read(&nohz.load_balancer) == -1) {
			/* make me the ilb owner */
			if (atomic_cmpxchg(&nohz.load_balancer, -1, cpu) == -1)
				return 1;
		} else if (atomic_read(&nohz.load_balancer) == cpu)
			return 1;
	} else {
		if (!cpu_isset(cpu, nohz.cpu_mask))
			return 0;

		cpu_clear(cpu, nohz.cpu_mask);

		if (atomic_read(&nohz.load_balancer) == cpu)
			if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu)
				BUG();
	}
	return 0;
}
#endif

static DEFINE_SPINLOCK(balancing);

/*
3560 3561 3562 3563 3564
 * It checks each scheduling domain to see if it is due to be balanced,
 * and initiates a balancing operation if so.
 *
 * Balancing parameters are set up in arch_init_sched_domains.
 */
A
Alexey Dobriyan 已提交
3565
static void rebalance_domains(int cpu, enum cpu_idle_type idle)
3566
{
3567 3568
	int balance = 1;
	struct rq *rq = cpu_rq(cpu);
3569 3570
	unsigned long interval;
	struct sched_domain *sd;
3571
	/* Earliest time when we have to do rebalance again */
3572
	unsigned long next_balance = jiffies + 60*HZ;
3573
	int update_next_balance = 0;
3574
	cpumask_t tmp;
L
Linus Torvalds 已提交
3575

3576
	for_each_domain(cpu, sd) {
L
Linus Torvalds 已提交
3577 3578 3579 3580
		if (!(sd->flags & SD_LOAD_BALANCE))
			continue;

		interval = sd->balance_interval;
I
Ingo Molnar 已提交
3581
		if (idle != CPU_IDLE)
L
Linus Torvalds 已提交
3582 3583 3584 3585 3586 3587
			interval *= sd->busy_factor;

		/* scale ms to jiffies */
		interval = msecs_to_jiffies(interval);
		if (unlikely(!interval))
			interval = 1;
I
Ingo Molnar 已提交
3588 3589 3590
		if (interval > HZ*NR_CPUS/10)
			interval = HZ*NR_CPUS/10;

L
Linus Torvalds 已提交
3591

3592 3593 3594 3595 3596
		if (sd->flags & SD_SERIALIZE) {
			if (!spin_trylock(&balancing))
				goto out;
		}

3597
		if (time_after_eq(jiffies, sd->last_balance + interval)) {
3598
			if (load_balance(cpu, rq, sd, idle, &balance, &tmp)) {
3599 3600
				/*
				 * We've pulled tasks over so either we're no
N
Nick Piggin 已提交
3601 3602 3603
				 * longer idle, or one of our SMT siblings is
				 * not idle.
				 */
I
Ingo Molnar 已提交
3604
				idle = CPU_NOT_IDLE;
L
Linus Torvalds 已提交
3605
			}
3606
			sd->last_balance = jiffies;
L
Linus Torvalds 已提交
3607
		}
3608 3609 3610
		if (sd->flags & SD_SERIALIZE)
			spin_unlock(&balancing);
out:
3611
		if (time_after(next_balance, sd->last_balance + interval)) {
3612
			next_balance = sd->last_balance + interval;
3613 3614
			update_next_balance = 1;
		}
3615 3616 3617 3618 3619 3620 3621 3622

		/*
		 * Stop the load balance at this level. There is another
		 * CPU in our sched group which is doing load balancing more
		 * actively.
		 */
		if (!balance)
			break;
L
Linus Torvalds 已提交
3623
	}
3624 3625 3626 3627 3628 3629 3630 3631

	/*
	 * next_balance will be updated only when there is a need.
	 * When the cpu is attached to null domain for ex, it will not be
	 * updated.
	 */
	if (likely(update_next_balance))
		rq->next_balance = next_balance;
3632 3633 3634 3635 3636 3637 3638 3639 3640
}

/*
 * run_rebalance_domains is triggered when needed from the scheduler tick.
 * In CONFIG_NO_HZ case, the idle load balance owner will do the
 * rebalancing for all the cpus for whom scheduler ticks are stopped.
 */
static void run_rebalance_domains(struct softirq_action *h)
{
I
Ingo Molnar 已提交
3641 3642 3643 3644
	int this_cpu = smp_processor_id();
	struct rq *this_rq = cpu_rq(this_cpu);
	enum cpu_idle_type idle = this_rq->idle_at_tick ?
						CPU_IDLE : CPU_NOT_IDLE;
3645

I
Ingo Molnar 已提交
3646
	rebalance_domains(this_cpu, idle);
3647 3648 3649 3650 3651 3652 3653

#ifdef CONFIG_NO_HZ
	/*
	 * If this cpu is the owner for idle load balancing, then do the
	 * balancing on behalf of the other idle cpus whose ticks are
	 * stopped.
	 */
I
Ingo Molnar 已提交
3654 3655
	if (this_rq->idle_at_tick &&
	    atomic_read(&nohz.load_balancer) == this_cpu) {
3656 3657 3658 3659
		cpumask_t cpus = nohz.cpu_mask;
		struct rq *rq;
		int balance_cpu;

I
Ingo Molnar 已提交
3660
		cpu_clear(this_cpu, cpus);
3661 3662 3663 3664 3665 3666 3667 3668 3669
		for_each_cpu_mask(balance_cpu, cpus) {
			/*
			 * If this cpu gets work to do, stop the load balancing
			 * work being done for other cpus. Next load
			 * balancing owner will pick it up.
			 */
			if (need_resched())
				break;

3670
			rebalance_domains(balance_cpu, CPU_IDLE);
3671 3672

			rq = cpu_rq(balance_cpu);
I
Ingo Molnar 已提交
3673 3674
			if (time_after(this_rq->next_balance, rq->next_balance))
				this_rq->next_balance = rq->next_balance;
3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686
		}
	}
#endif
}

/*
 * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing.
 *
 * In case of CONFIG_NO_HZ, this is the place where we nominate a new
 * idle load balancing owner or decide to stop the periodic load balancing,
 * if the whole system is idle.
 */
I
Ingo Molnar 已提交
3687
static inline void trigger_load_balance(struct rq *rq, int cpu)
3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713
{
#ifdef CONFIG_NO_HZ
	/*
	 * If we were in the nohz mode recently and busy at the current
	 * scheduler tick, then check if we need to nominate new idle
	 * load balancer.
	 */
	if (rq->in_nohz_recently && !rq->idle_at_tick) {
		rq->in_nohz_recently = 0;

		if (atomic_read(&nohz.load_balancer) == cpu) {
			cpu_clear(cpu, nohz.cpu_mask);
			atomic_set(&nohz.load_balancer, -1);
		}

		if (atomic_read(&nohz.load_balancer) == -1) {
			/*
			 * simple selection for now: Nominate the
			 * first cpu in the nohz list to be the next
			 * ilb owner.
			 *
			 * TBD: Traverse the sched domains and nominate
			 * the nearest cpu in the nohz.cpu_mask.
			 */
			int ilb = first_cpu(nohz.cpu_mask);

3714
			if (ilb < nr_cpu_ids)
3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738
				resched_cpu(ilb);
		}
	}

	/*
	 * If this cpu is idle and doing idle load balancing for all the
	 * cpus with ticks stopped, is it time for that to stop?
	 */
	if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) == cpu &&
	    cpus_weight(nohz.cpu_mask) == num_online_cpus()) {
		resched_cpu(cpu);
		return;
	}

	/*
	 * If this cpu is idle and the idle load balancing is done by
	 * someone else, then no need raise the SCHED_SOFTIRQ
	 */
	if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) != cpu &&
	    cpu_isset(cpu, nohz.cpu_mask))
		return;
#endif
	if (time_after_eq(jiffies, rq->next_balance))
		raise_softirq(SCHED_SOFTIRQ);
L
Linus Torvalds 已提交
3739
}
I
Ingo Molnar 已提交
3740 3741 3742

#else	/* CONFIG_SMP */

L
Linus Torvalds 已提交
3743 3744 3745
/*
 * on UP we do not need to balance between CPUs:
 */
3746
static inline void idle_balance(int cpu, struct rq *rq)
L
Linus Torvalds 已提交
3747 3748
{
}
I
Ingo Molnar 已提交
3749

L
Linus Torvalds 已提交
3750 3751 3752 3753 3754 3755 3756
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);

EXPORT_PER_CPU_SYMBOL(kstat);

/*
3757 3758
 * Return p->sum_exec_runtime plus any more ns on the sched_clock
 * that have not yet been banked in case the task is currently running.
L
Linus Torvalds 已提交
3759
 */
3760
unsigned long long task_sched_runtime(struct task_struct *p)
L
Linus Torvalds 已提交
3761 3762
{
	unsigned long flags;
3763 3764
	u64 ns, delta_exec;
	struct rq *rq;
3765

3766 3767
	rq = task_rq_lock(p, &flags);
	ns = p->se.sum_exec_runtime;
3768
	if (task_current(rq, p)) {
I
Ingo Molnar 已提交
3769 3770
		update_rq_clock(rq);
		delta_exec = rq->clock - p->se.exec_start;
3771 3772 3773 3774
		if ((s64)delta_exec > 0)
			ns += delta_exec;
	}
	task_rq_unlock(rq, &flags);
3775

L
Linus Torvalds 已提交
3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798
	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
 */
void account_user_time(struct task_struct *p, cputime_t cputime)
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t tmp;

	p->utime = cputime_add(p->utime, cputime);

	/* 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);
}

3799 3800 3801 3802 3803
/*
 * 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
 */
3804
static void account_guest_time(struct task_struct *p, cputime_t cputime)
3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817
{
	cputime64_t tmp;
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;

	tmp = cputime_to_cputime64(cputime);

	p->utime = cputime_add(p->utime, cputime);
	p->gtime = cputime_add(p->gtime, cputime);

	cpustat->user = cputime64_add(cpustat->user, tmp);
	cpustat->guest = cputime64_add(cpustat->guest, tmp);
}

3818 3819 3820 3821 3822 3823 3824 3825 3826 3827
/*
 * Account scaled 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
 */
void account_user_time_scaled(struct task_struct *p, cputime_t cputime)
{
	p->utimescaled = cputime_add(p->utimescaled, cputime);
}

L
Linus Torvalds 已提交
3828 3829 3830 3831 3832 3833 3834 3835 3836 3837
/*
 * 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
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
			 cputime_t cputime)
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
3838
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
3839 3840
	cputime64_t tmp;

3841 3842
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0))
		return account_guest_time(p, cputime);
3843

L
Linus Torvalds 已提交
3844 3845 3846 3847 3848 3849 3850 3851
	p->stime = cputime_add(p->stime, cputime);

	/* Add system time to cpustat. */
	tmp = cputime_to_cputime64(cputime);
	if (hardirq_count() - hardirq_offset)
		cpustat->irq = cputime64_add(cpustat->irq, tmp);
	else if (softirq_count())
		cpustat->softirq = cputime64_add(cpustat->softirq, tmp);
3852
	else if (p != rq->idle)
L
Linus Torvalds 已提交
3853
		cpustat->system = cputime64_add(cpustat->system, tmp);
3854
	else if (atomic_read(&rq->nr_iowait) > 0)
L
Linus Torvalds 已提交
3855 3856 3857 3858 3859 3860 3861
		cpustat->iowait = cputime64_add(cpustat->iowait, tmp);
	else
		cpustat->idle = cputime64_add(cpustat->idle, tmp);
	/* Account for system time used */
	acct_update_integrals(p);
}

3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872
/*
 * Account scaled 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
 */
void account_system_time_scaled(struct task_struct *p, cputime_t cputime)
{
	p->stimescaled = cputime_add(p->stimescaled, cputime);
}

L
Linus Torvalds 已提交
3873 3874 3875 3876 3877 3878 3879 3880 3881
/*
 * Account for involuntary wait time.
 * @p: the process from which the cpu time has been stolen
 * @steal: the cpu time spent in involuntary wait
 */
void account_steal_time(struct task_struct *p, cputime_t steal)
{
	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
	cputime64_t tmp = cputime_to_cputime64(steal);
3882
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
3883 3884 3885 3886 3887 3888 3889

	if (p == rq->idle) {
		p->stime = cputime_add(p->stime, steal);
		if (atomic_read(&rq->nr_iowait) > 0)
			cpustat->iowait = cputime64_add(cpustat->iowait, tmp);
		else
			cpustat->idle = cputime64_add(cpustat->idle, tmp);
3890
	} else
L
Linus Torvalds 已提交
3891 3892 3893
		cpustat->steal = cputime64_add(cpustat->steal, tmp);
}

3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904
/*
 * 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
3905
	struct task_struct *curr = rq->curr;
3906
	u64 next_tick = rq->tick_timestamp + TICK_NSEC;
I
Ingo Molnar 已提交
3907 3908

	spin_lock(&rq->lock);
3909
	__update_rq_clock(rq);
3910 3911 3912
	/*
	 * Let rq->clock advance by at least TICK_NSEC:
	 */
3913
	if (unlikely(rq->clock < next_tick)) {
3914
		rq->clock = next_tick;
3915 3916
		rq->clock_underflows++;
	}
3917
	rq->tick_timestamp = rq->clock;
3918
	update_last_tick_seen(rq);
3919
	update_cpu_load(rq);
P
Peter Zijlstra 已提交
3920
	curr->sched_class->task_tick(rq, curr, 0);
I
Ingo Molnar 已提交
3921
	spin_unlock(&rq->lock);
3922

3923
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
3924 3925
	rq->idle_at_tick = idle_cpu(cpu);
	trigger_load_balance(rq, cpu);
3926
#endif
L
Linus Torvalds 已提交
3927 3928 3929 3930
}

#if defined(CONFIG_PREEMPT) && defined(CONFIG_DEBUG_PREEMPT)

3931
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
3932 3933 3934 3935
{
	/*
	 * Underflow?
	 */
3936 3937
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
L
Linus Torvalds 已提交
3938 3939 3940 3941
	preempt_count() += val;
	/*
	 * Spinlock count overflowing soon?
	 */
3942 3943
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
L
Linus Torvalds 已提交
3944 3945 3946
}
EXPORT_SYMBOL(add_preempt_count);

3947
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
3948 3949 3950 3951
{
	/*
	 * Underflow?
	 */
3952 3953
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
		return;
L
Linus Torvalds 已提交
3954 3955 3956
	/*
	 * Is the spinlock portion underflowing?
	 */
3957 3958 3959 3960
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;

L
Linus Torvalds 已提交
3961 3962 3963 3964 3965 3966 3967
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
3968
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3969
 */
I
Ingo Molnar 已提交
3970
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3971
{
3972 3973 3974 3975 3976
	struct pt_regs *regs = get_irq_regs();

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

I
Ingo Molnar 已提交
3977 3978 3979
	debug_show_held_locks(prev);
	if (irqs_disabled())
		print_irqtrace_events(prev);
3980 3981 3982 3983 3984

	if (regs)
		show_regs(regs);
	else
		dump_stack();
I
Ingo Molnar 已提交
3985
}
L
Linus Torvalds 已提交
3986

I
Ingo Molnar 已提交
3987 3988 3989 3990 3991
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
3992
	/*
I
Ingo Molnar 已提交
3993
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
3994 3995 3996
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
I
Ingo Molnar 已提交
3997 3998 3999
	if (unlikely(in_atomic_preempt_off()) && unlikely(!prev->exit_state))
		__schedule_bug(prev);

L
Linus Torvalds 已提交
4000 4001
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

4002
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
4003 4004
#ifdef CONFIG_SCHEDSTATS
	if (unlikely(prev->lock_depth >= 0)) {
4005 4006
		schedstat_inc(this_rq(), bkl_count);
		schedstat_inc(prev, sched_info.bkl_count);
I
Ingo Molnar 已提交
4007 4008
	}
#endif
I
Ingo Molnar 已提交
4009 4010 4011 4012 4013 4014
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
4015
pick_next_task(struct rq *rq, struct task_struct *prev)
I
Ingo Molnar 已提交
4016
{
4017
	const struct sched_class *class;
I
Ingo Molnar 已提交
4018
	struct task_struct *p;
L
Linus Torvalds 已提交
4019 4020

	/*
I
Ingo Molnar 已提交
4021 4022
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
4023
	 */
I
Ingo Molnar 已提交
4024
	if (likely(rq->nr_running == rq->cfs.nr_running)) {
4025
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
4026 4027
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
4028 4029
	}

I
Ingo Molnar 已提交
4030 4031
	class = sched_class_highest;
	for ( ; ; ) {
4032
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
4033 4034 4035 4036 4037 4038 4039 4040 4041
		if (p)
			return p;
		/*
		 * Will never be NULL as the idle class always
		 * returns a non-NULL p:
		 */
		class = class->next;
	}
}
L
Linus Torvalds 已提交
4042

I
Ingo Molnar 已提交
4043 4044 4045 4046 4047 4048
/*
 * schedule() is the main scheduler function.
 */
asmlinkage void __sched schedule(void)
{
	struct task_struct *prev, *next;
4049
	unsigned long *switch_count;
I
Ingo Molnar 已提交
4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064
	struct rq *rq;
	int cpu;

need_resched:
	preempt_disable();
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
	rcu_qsctr_inc(cpu);
	prev = rq->curr;
	switch_count = &prev->nivcsw;

	release_kernel_lock(prev);
need_resched_nonpreemptible:

	schedule_debug(prev);
L
Linus Torvalds 已提交
4065

P
Peter Zijlstra 已提交
4066 4067
	hrtick_clear(rq);

4068 4069 4070 4071
	/*
	 * Do the rq-clock update outside the rq lock:
	 */
	local_irq_disable();
I
Ingo Molnar 已提交
4072
	__update_rq_clock(rq);
4073 4074
	spin_lock(&rq->lock);
	clear_tsk_need_resched(prev);
L
Linus Torvalds 已提交
4075 4076 4077

	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
		if (unlikely((prev->state & TASK_INTERRUPTIBLE) &&
4078
				signal_pending(prev))) {
L
Linus Torvalds 已提交
4079
			prev->state = TASK_RUNNING;
I
Ingo Molnar 已提交
4080
		} else {
4081
			deactivate_task(rq, prev, 1);
L
Linus Torvalds 已提交
4082
		}
I
Ingo Molnar 已提交
4083
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
4084 4085
	}

4086 4087 4088 4089
#ifdef CONFIG_SMP
	if (prev->sched_class->pre_schedule)
		prev->sched_class->pre_schedule(rq, prev);
#endif
4090

I
Ingo Molnar 已提交
4091
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
4092 4093
		idle_balance(cpu, rq);

4094
	prev->sched_class->put_prev_task(rq, prev);
4095
	next = pick_next_task(rq, prev);
L
Linus Torvalds 已提交
4096 4097

	sched_info_switch(prev, next);
I
Ingo Molnar 已提交
4098

L
Linus Torvalds 已提交
4099 4100 4101 4102 4103
	if (likely(prev != next)) {
		rq->nr_switches++;
		rq->curr = next;
		++*switch_count;

I
Ingo Molnar 已提交
4104
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
4105 4106 4107 4108 4109 4110
		/*
		 * the context switch might have flipped the stack from under
		 * us, hence refresh the local variables.
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
4111 4112 4113
	} else
		spin_unlock_irq(&rq->lock);

P
Peter Zijlstra 已提交
4114 4115 4116
	hrtick_set(rq);

	if (unlikely(reacquire_kernel_lock(current) < 0))
L
Linus Torvalds 已提交
4117
		goto need_resched_nonpreemptible;
P
Peter Zijlstra 已提交
4118

L
Linus Torvalds 已提交
4119 4120 4121 4122 4123 4124 4125 4126
	preempt_enable_no_resched();
	if (unlikely(test_thread_flag(TIF_NEED_RESCHED)))
		goto need_resched;
}
EXPORT_SYMBOL(schedule);

#ifdef CONFIG_PREEMPT
/*
4127
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
4128
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
4129 4130 4131 4132 4133 4134 4135
 * occur there and call schedule directly.
 */
asmlinkage void __sched preempt_schedule(void)
{
	struct thread_info *ti = current_thread_info();
	struct task_struct *task = current;
	int saved_lock_depth;
4136

L
Linus Torvalds 已提交
4137 4138
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
4139
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
4140
	 */
N
Nick Piggin 已提交
4141
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
4142 4143
		return;

4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156
	do {
		add_preempt_count(PREEMPT_ACTIVE);

		/*
		 * We keep the big kernel semaphore locked, but we
		 * clear ->lock_depth so that schedule() doesnt
		 * auto-release the semaphore:
		 */
		saved_lock_depth = task->lock_depth;
		task->lock_depth = -1;
		schedule();
		task->lock_depth = saved_lock_depth;
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4157

4158 4159 4160 4161 4162 4163
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
	} while (unlikely(test_thread_flag(TIF_NEED_RESCHED)));
L
Linus Torvalds 已提交
4164 4165 4166 4167
}
EXPORT_SYMBOL(preempt_schedule);

/*
4168
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
4169 4170 4171 4172 4173 4174 4175 4176 4177
 * 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();
	struct task_struct *task = current;
	int saved_lock_depth;
4178

4179
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
4180 4181
	BUG_ON(ti->preempt_count || !irqs_disabled());

4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196
	do {
		add_preempt_count(PREEMPT_ACTIVE);

		/*
		 * We keep the big kernel semaphore locked, but we
		 * clear ->lock_depth so that schedule() doesnt
		 * auto-release the semaphore:
		 */
		saved_lock_depth = task->lock_depth;
		task->lock_depth = -1;
		local_irq_enable();
		schedule();
		local_irq_disable();
		task->lock_depth = saved_lock_depth;
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4197

4198 4199 4200 4201 4202 4203
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
	} while (unlikely(test_thread_flag(TIF_NEED_RESCHED)));
L
Linus Torvalds 已提交
4204 4205 4206 4207
}

#endif /* CONFIG_PREEMPT */

I
Ingo Molnar 已提交
4208 4209
int default_wake_function(wait_queue_t *curr, unsigned mode, int sync,
			  void *key)
L
Linus Torvalds 已提交
4210
{
4211
	return try_to_wake_up(curr->private, mode, sync);
L
Linus Torvalds 已提交
4212 4213 4214 4215
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
4216 4217
 * 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
4218 4219 4220
 * 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
4221
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
4222 4223 4224 4225 4226
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
			     int nr_exclusive, int sync, void *key)
{
4227
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
4228

4229
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
4230 4231
		unsigned flags = curr->flags;

L
Linus Torvalds 已提交
4232
		if (curr->func(curr, mode, sync, key) &&
4233
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
4234 4235 4236 4237 4238 4239 4240 4241 4242
			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
4243
 * @key: is directly passed to the wakeup function
L
Linus Torvalds 已提交
4244
 */
4245
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
4246
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258
{
	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.
 */
4259
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode)
L
Linus Torvalds 已提交
4260 4261 4262 4263 4264
{
	__wake_up_common(q, mode, 1, 0, NULL);
}

/**
4265
 * __wake_up_sync - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
 *
 * 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.
 */
4277
void
I
Ingo Molnar 已提交
4278
__wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr_exclusive)
L
Linus Torvalds 已提交
4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294
{
	unsigned long flags;
	int sync = 1;

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
		sync = 0;

	spin_lock_irqsave(&q->lock, flags);
	__wake_up_common(q, mode, nr_exclusive, sync, NULL);
	spin_unlock_irqrestore(&q->lock, flags);
}
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

4295
void complete(struct completion *x)
L
Linus Torvalds 已提交
4296 4297 4298 4299 4300
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
4301
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
4302 4303 4304 4305
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

4306
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
4307 4308 4309 4310 4311
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
4312
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
4313 4314 4315 4316
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

4317 4318
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
4319 4320 4321 4322 4323 4324 4325
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

		wait.flags |= WQ_FLAG_EXCLUSIVE;
		__add_wait_queue_tail(&x->wait, &wait);
		do {
M
Matthew Wilcox 已提交
4326 4327 4328 4329
			if ((state == TASK_INTERRUPTIBLE &&
			     signal_pending(current)) ||
			    (state == TASK_KILLABLE &&
			     fatal_signal_pending(current))) {
4330 4331 4332 4333
				__remove_wait_queue(&x->wait, &wait);
				return -ERESTARTSYS;
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
4334 4335 4336 4337 4338
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
			if (!timeout) {
				__remove_wait_queue(&x->wait, &wait);
4339
				return timeout;
L
Linus Torvalds 已提交
4340 4341 4342 4343 4344 4345 4346 4347
			}
		} while (!x->done);
		__remove_wait_queue(&x->wait, &wait);
	}
	x->done--;
	return timeout;
}

4348 4349
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
4350 4351 4352 4353
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
4354
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
4355
	spin_unlock_irq(&x->wait.lock);
4356 4357
	return timeout;
}
L
Linus Torvalds 已提交
4358

4359
void __sched wait_for_completion(struct completion *x)
4360 4361
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4362
}
4363
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
4364

4365
unsigned long __sched
4366
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
4367
{
4368
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4369
}
4370
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
4371

4372
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
4373
{
4374 4375 4376 4377
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
4378
}
4379
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
4380

4381
unsigned long __sched
4382 4383
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
4384
{
4385
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
4386
}
4387
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
4388

M
Matthew Wilcox 已提交
4389 4390 4391 4392 4393 4394 4395 4396 4397
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);

4398 4399
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
4400
{
I
Ingo Molnar 已提交
4401 4402 4403 4404
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
4405

4406
	__set_current_state(state);
L
Linus Torvalds 已提交
4407

4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421
	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
4422 4423 4424
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
4425
long __sched
I
Ingo Molnar 已提交
4426
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
4427
{
4428
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
4429 4430 4431
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
4432
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
4433
{
4434
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
4435 4436 4437
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
4438
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
4439
{
4440
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
4441 4442 4443
}
EXPORT_SYMBOL(sleep_on_timeout);

4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455
#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.
 */
4456
void rt_mutex_setprio(struct task_struct *p, int prio)
4457 4458
{
	unsigned long flags;
4459
	int oldprio, on_rq, running;
4460
	struct rq *rq;
4461
	const struct sched_class *prev_class = p->sched_class;
4462 4463 4464 4465

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

	rq = task_rq_lock(p, &flags);
I
Ingo Molnar 已提交
4466
	update_rq_clock(rq);
4467

4468
	oldprio = p->prio;
I
Ingo Molnar 已提交
4469
	on_rq = p->se.on_rq;
4470
	running = task_current(rq, p);
4471
	if (on_rq)
4472
		dequeue_task(rq, p, 0);
4473 4474
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
4475 4476 4477 4478 4479 4480

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

4481 4482
	p->prio = prio;

4483 4484
	if (running)
		p->sched_class->set_curr_task(rq);
I
Ingo Molnar 已提交
4485
	if (on_rq) {
4486
		enqueue_task(rq, p, 0);
4487 4488

		check_class_changed(rq, p, prev_class, oldprio, running);
4489 4490 4491 4492 4493 4494
	}
	task_rq_unlock(rq, &flags);
}

#endif

4495
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
4496
{
I
Ingo Molnar 已提交
4497
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
4498
	unsigned long flags;
4499
	struct rq *rq;
L
Linus Torvalds 已提交
4500 4501 4502 4503 4504 4505 4506 4507

	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);
I
Ingo Molnar 已提交
4508
	update_rq_clock(rq);
L
Linus Torvalds 已提交
4509 4510 4511 4512
	/*
	 * 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
4513
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
4514
	 */
4515
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
4516 4517 4518
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
I
Ingo Molnar 已提交
4519
	on_rq = p->se.on_rq;
4520
	if (on_rq) {
4521
		dequeue_task(rq, p, 0);
4522 4523
		dec_load(rq, p);
	}
L
Linus Torvalds 已提交
4524 4525

	p->static_prio = NICE_TO_PRIO(nice);
4526
	set_load_weight(p);
4527 4528 4529
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
4530

I
Ingo Molnar 已提交
4531
	if (on_rq) {
4532
		enqueue_task(rq, p, 0);
4533
		inc_load(rq, p);
L
Linus Torvalds 已提交
4534
		/*
4535 4536
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
4537
		 */
4538
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
4539 4540 4541 4542 4543 4544 4545
			resched_task(rq->curr);
	}
out_unlock:
	task_rq_unlock(rq, &flags);
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
4546 4547 4548 4549 4550
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
4551
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
4552
{
4553 4554
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
4555

M
Matt Mackall 已提交
4556 4557 4558 4559
	return (nice_rlim <= p->signal->rlim[RLIMIT_NICE].rlim_cur ||
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570
#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.
 */
asmlinkage long sys_nice(int increment)
{
4571
	long nice, retval;
L
Linus Torvalds 已提交
4572 4573 4574 4575 4576 4577

	/*
	 * 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
4578 4579
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
4580 4581 4582 4583 4584 4585 4586 4587 4588
	if (increment > 40)
		increment = 40;

	nice = PRIO_TO_NICE(current->static_prio) + increment;
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
4589 4590 4591
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609
	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.
 */
4610
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
4611 4612 4613 4614 4615 4616 4617 4618
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
4619
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
4620 4621 4622
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
4623
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637

/**
 * 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.
 */
4638
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4639 4640 4641 4642 4643 4644 4645 4646
{
	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
4647
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4648
{
4649
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4650 4651 4652
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
4653 4654
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
4655
{
I
Ingo Molnar 已提交
4656
	BUG_ON(p->se.on_rq);
4657

L
Linus Torvalds 已提交
4658
	p->policy = policy;
I
Ingo Molnar 已提交
4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670
	switch (p->policy) {
	case SCHED_NORMAL:
	case SCHED_BATCH:
	case SCHED_IDLE:
		p->sched_class = &fair_sched_class;
		break;
	case SCHED_FIFO:
	case SCHED_RR:
		p->sched_class = &rt_sched_class;
		break;
	}

L
Linus Torvalds 已提交
4671
	p->rt_priority = prio;
4672 4673 4674
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4675
	set_load_weight(p);
L
Linus Torvalds 已提交
4676 4677 4678
}

/**
4679
 * sched_setscheduler - change the scheduling policy and/or RT priority of a thread.
L
Linus Torvalds 已提交
4680 4681 4682
 * @p: the task in question.
 * @policy: new policy.
 * @param: structure containing the new RT priority.
4683
 *
4684
 * NOTE that the task may be already dead.
L
Linus Torvalds 已提交
4685
 */
I
Ingo Molnar 已提交
4686 4687
int sched_setscheduler(struct task_struct *p, int policy,
		       struct sched_param *param)
L
Linus Torvalds 已提交
4688
{
4689
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
4690
	unsigned long flags;
4691
	const struct sched_class *prev_class = p->sched_class;
4692
	struct rq *rq;
L
Linus Torvalds 已提交
4693

4694 4695
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4696 4697 4698 4699 4700
recheck:
	/* double check policy once rq lock held */
	if (policy < 0)
		policy = oldpolicy = p->policy;
	else if (policy != SCHED_FIFO && policy != SCHED_RR &&
I
Ingo Molnar 已提交
4701 4702
			policy != SCHED_NORMAL && policy != SCHED_BATCH &&
			policy != SCHED_IDLE)
4703
		return -EINVAL;
L
Linus Torvalds 已提交
4704 4705
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4706 4707
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4708 4709
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
4710
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4711
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4712
		return -EINVAL;
4713
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
4714 4715
		return -EINVAL;

4716 4717 4718 4719
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
	if (!capable(CAP_SYS_NICE)) {
4720
		if (rt_policy(policy)) {
4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736
			unsigned long rlim_rtprio;

			if (!lock_task_sighand(p, &flags))
				return -ESRCH;
			rlim_rtprio = p->signal->rlim[RLIMIT_RTPRIO].rlim_cur;
			unlock_task_sighand(p, &flags);

			/* can't set/change the rt policy */
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

			/* can't increase priority */
			if (param->sched_priority > p->rt_priority &&
			    param->sched_priority > rlim_rtprio)
				return -EPERM;
		}
I
Ingo Molnar 已提交
4737 4738 4739 4740 4741 4742
		/*
		 * Like positive nice levels, dont allow tasks to
		 * move out of SCHED_IDLE either:
		 */
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE)
			return -EPERM;
4743

4744 4745 4746 4747 4748
		/* can't change other user's priorities */
		if ((current->euid != p->euid) &&
		    (current->euid != p->uid))
			return -EPERM;
	}
L
Linus Torvalds 已提交
4749

4750 4751 4752 4753 4754
#ifdef CONFIG_RT_GROUP_SCHED
	/*
	 * Do not allow realtime tasks into groups that have no runtime
	 * assigned.
	 */
4755
	if (rt_policy(policy) && task_group(p)->rt_bandwidth.rt_runtime == 0)
4756 4757 4758
		return -EPERM;
#endif

L
Linus Torvalds 已提交
4759 4760 4761
	retval = security_task_setscheduler(p, policy, param);
	if (retval)
		return retval;
4762 4763 4764 4765 4766
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
	 */
	spin_lock_irqsave(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4767 4768 4769 4770
	/*
	 * To be able to change p->policy safely, the apropriate
	 * runqueue lock must be held.
	 */
4771
	rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
4772 4773 4774
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4775 4776
		__task_rq_unlock(rq);
		spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4777 4778
		goto recheck;
	}
I
Ingo Molnar 已提交
4779
	update_rq_clock(rq);
I
Ingo Molnar 已提交
4780
	on_rq = p->se.on_rq;
4781
	running = task_current(rq, p);
4782
	if (on_rq)
4783
		deactivate_task(rq, p, 0);
4784 4785
	if (running)
		p->sched_class->put_prev_task(rq, p);
4786

L
Linus Torvalds 已提交
4787
	oldprio = p->prio;
I
Ingo Molnar 已提交
4788
	__setscheduler(rq, p, policy, param->sched_priority);
4789

4790 4791
	if (running)
		p->sched_class->set_curr_task(rq);
I
Ingo Molnar 已提交
4792 4793
	if (on_rq) {
		activate_task(rq, p, 0);
4794 4795

		check_class_changed(rq, p, prev_class, oldprio, running);
L
Linus Torvalds 已提交
4796
	}
4797 4798 4799
	__task_rq_unlock(rq);
	spin_unlock_irqrestore(&p->pi_lock, flags);

4800 4801
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
4802 4803 4804 4805
	return 0;
}
EXPORT_SYMBOL_GPL(sched_setscheduler);

I
Ingo Molnar 已提交
4806 4807
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4808 4809 4810
{
	struct sched_param lparam;
	struct task_struct *p;
4811
	int retval;
L
Linus Torvalds 已提交
4812 4813 4814 4815 4816

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4817 4818 4819

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4820
	p = find_process_by_pid(pid);
4821 4822 4823
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4824

L
Linus Torvalds 已提交
4825 4826 4827 4828 4829 4830 4831 4832 4833
	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.
 */
I
Ingo Molnar 已提交
4834 4835
asmlinkage long
sys_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4836
{
4837 4838 4839 4840
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859
	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.
 */
asmlinkage long sys_sched_setparam(pid_t pid, struct sched_param __user *param)
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
asmlinkage long sys_sched_getscheduler(pid_t pid)
{
4860
	struct task_struct *p;
4861
	int retval;
L
Linus Torvalds 已提交
4862 4863

	if (pid < 0)
4864
		return -EINVAL;
L
Linus Torvalds 已提交
4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885

	retval = -ESRCH;
	read_lock(&tasklist_lock);
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
			retval = p->policy;
	}
	read_unlock(&tasklist_lock);
	return retval;
}

/**
 * sys_sched_getscheduler - get the RT priority of a thread
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
asmlinkage long sys_sched_getparam(pid_t pid, struct sched_param __user *param)
{
	struct sched_param lp;
4886
	struct task_struct *p;
4887
	int retval;
L
Linus Torvalds 已提交
4888 4889

	if (!param || pid < 0)
4890
		return -EINVAL;
L
Linus Torvalds 已提交
4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916

	read_lock(&tasklist_lock);
	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;
	read_unlock(&tasklist_lock);

	/*
	 * 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:
	read_unlock(&tasklist_lock);
	return retval;
}

4917
long sched_setaffinity(pid_t pid, const cpumask_t *in_mask)
L
Linus Torvalds 已提交
4918 4919
{
	cpumask_t cpus_allowed;
4920
	cpumask_t new_mask = *in_mask;
4921 4922
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4923

4924
	get_online_cpus();
L
Linus Torvalds 已提交
4925 4926 4927 4928 4929
	read_lock(&tasklist_lock);

	p = find_process_by_pid(pid);
	if (!p) {
		read_unlock(&tasklist_lock);
4930
		put_online_cpus();
L
Linus Torvalds 已提交
4931 4932 4933 4934 4935
		return -ESRCH;
	}

	/*
	 * It is not safe to call set_cpus_allowed with the
I
Ingo Molnar 已提交
4936
	 * tasklist_lock held. We will bump the task_struct's
L
Linus Torvalds 已提交
4937 4938 4939 4940 4941 4942 4943 4944 4945 4946
	 * usage count and then drop tasklist_lock.
	 */
	get_task_struct(p);
	read_unlock(&tasklist_lock);

	retval = -EPERM;
	if ((current->euid != p->euid) && (current->euid != p->uid) &&
			!capable(CAP_SYS_NICE))
		goto out_unlock;

4947 4948 4949 4950
	retval = security_task_setscheduler(p, 0, NULL);
	if (retval)
		goto out_unlock;

4951
	cpuset_cpus_allowed(p, &cpus_allowed);
L
Linus Torvalds 已提交
4952
	cpus_and(new_mask, new_mask, cpus_allowed);
P
Paul Menage 已提交
4953
 again:
4954
	retval = set_cpus_allowed_ptr(p, &new_mask);
L
Linus Torvalds 已提交
4955

P
Paul Menage 已提交
4956
	if (!retval) {
4957
		cpuset_cpus_allowed(p, &cpus_allowed);
P
Paul Menage 已提交
4958 4959 4960 4961 4962 4963 4964 4965 4966 4967
		if (!cpus_subset(new_mask, cpus_allowed)) {
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
			new_mask = cpus_allowed;
			goto again;
		}
	}
L
Linus Torvalds 已提交
4968 4969
out_unlock:
	put_task_struct(p);
4970
	put_online_cpus();
L
Linus Torvalds 已提交
4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
			     cpumask_t *new_mask)
{
	if (len < sizeof(cpumask_t)) {
		memset(new_mask, 0, sizeof(cpumask_t));
	} else if (len > sizeof(cpumask_t)) {
		len = sizeof(cpumask_t);
	}
	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
 */
asmlinkage long sys_sched_setaffinity(pid_t pid, unsigned int len,
				      unsigned long __user *user_mask_ptr)
{
	cpumask_t new_mask;
	int retval;

	retval = get_user_cpu_mask(user_mask_ptr, len, &new_mask);
	if (retval)
		return retval;

5001
	return sched_setaffinity(pid, &new_mask);
L
Linus Torvalds 已提交
5002 5003 5004 5005 5006 5007 5008 5009 5010
}

/*
 * Represents all cpu's present in the system
 * In systems capable of hotplug, this map could dynamically grow
 * as new cpu's are detected in the system via any platform specific
 * method, such as ACPI for e.g.
 */

5011
cpumask_t cpu_present_map __read_mostly;
L
Linus Torvalds 已提交
5012 5013 5014
EXPORT_SYMBOL(cpu_present_map);

#ifndef CONFIG_SMP
5015
cpumask_t cpu_online_map __read_mostly = CPU_MASK_ALL;
5016 5017
EXPORT_SYMBOL(cpu_online_map);

5018
cpumask_t cpu_possible_map __read_mostly = CPU_MASK_ALL;
5019
EXPORT_SYMBOL(cpu_possible_map);
L
Linus Torvalds 已提交
5020 5021 5022 5023
#endif

long sched_getaffinity(pid_t pid, cpumask_t *mask)
{
5024
	struct task_struct *p;
L
Linus Torvalds 已提交
5025 5026
	int retval;

5027
	get_online_cpus();
L
Linus Torvalds 已提交
5028 5029 5030 5031 5032 5033 5034
	read_lock(&tasklist_lock);

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

5035 5036 5037 5038
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

5039
	cpus_and(*mask, p->cpus_allowed, cpu_online_map);
L
Linus Torvalds 已提交
5040 5041 5042

out_unlock:
	read_unlock(&tasklist_lock);
5043
	put_online_cpus();
L
Linus Torvalds 已提交
5044

5045
	return retval;
L
Linus Torvalds 已提交
5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075
}

/**
 * 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
 */
asmlinkage long sys_sched_getaffinity(pid_t pid, unsigned int len,
				      unsigned long __user *user_mask_ptr)
{
	int ret;
	cpumask_t mask;

	if (len < sizeof(cpumask_t))
		return -EINVAL;

	ret = sched_getaffinity(pid, &mask);
	if (ret < 0)
		return ret;

	if (copy_to_user(user_mask_ptr, &mask, sizeof(cpumask_t)))
		return -EFAULT;

	return sizeof(cpumask_t);
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
5076 5077
 * 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
5078 5079 5080
 */
asmlinkage long sys_sched_yield(void)
{
5081
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
5082

5083
	schedstat_inc(rq, yld_count);
5084
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
5085 5086 5087 5088 5089 5090

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
5091
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
5092 5093 5094 5095 5096 5097 5098 5099
	_raw_spin_unlock(&rq->lock);
	preempt_enable_no_resched();

	schedule();

	return 0;
}

A
Andrew Morton 已提交
5100
static void __cond_resched(void)
L
Linus Torvalds 已提交
5101
{
5102 5103 5104
#ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
	__might_sleep(__FILE__, __LINE__);
#endif
5105 5106 5107 5108 5109
	/*
	 * The BKS might be reacquired before we have dropped
	 * PREEMPT_ACTIVE, which could trigger a second
	 * cond_resched() call.
	 */
L
Linus Torvalds 已提交
5110 5111 5112 5113 5114 5115 5116
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		schedule();
		sub_preempt_count(PREEMPT_ACTIVE);
	} while (need_resched());
}

5117 5118
#if !defined(CONFIG_PREEMPT) || defined(CONFIG_PREEMPT_VOLUNTARY)
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
5119
{
5120 5121
	if (need_resched() && !(preempt_count() & PREEMPT_ACTIVE) &&
					system_state == SYSTEM_RUNNING) {
L
Linus Torvalds 已提交
5122 5123 5124 5125 5126
		__cond_resched();
		return 1;
	}
	return 0;
}
5127 5128
EXPORT_SYMBOL(_cond_resched);
#endif
L
Linus Torvalds 已提交
5129 5130 5131 5132 5133

/*
 * cond_resched_lock() - if a reschedule is pending, drop the given lock,
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
5134
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
5135 5136 5137
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
I
Ingo Molnar 已提交
5138
int cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
5139
{
N
Nick Piggin 已提交
5140
	int resched = need_resched() && system_state == SYSTEM_RUNNING;
J
Jan Kara 已提交
5141 5142
	int ret = 0;

N
Nick Piggin 已提交
5143
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
5144
		spin_unlock(lock);
N
Nick Piggin 已提交
5145 5146 5147 5148
		if (resched && need_resched())
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
5149
		ret = 1;
L
Linus Torvalds 已提交
5150 5151
		spin_lock(lock);
	}
J
Jan Kara 已提交
5152
	return ret;
L
Linus Torvalds 已提交
5153 5154 5155 5156 5157 5158 5159
}
EXPORT_SYMBOL(cond_resched_lock);

int __sched cond_resched_softirq(void)
{
	BUG_ON(!in_softirq());

5160
	if (need_resched() && system_state == SYSTEM_RUNNING) {
5161
		local_bh_enable();
L
Linus Torvalds 已提交
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
EXPORT_SYMBOL(cond_resched_softirq);

/**
 * yield - yield the current processor to other threads.
 *
5173
 * This is a shortcut for kernel-space yielding - it marks the
L
Linus Torvalds 已提交
5174 5175 5176 5177 5178 5179 5180 5181 5182 5183
 * thread runnable and calls sys_sched_yield().
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

/*
I
Ingo Molnar 已提交
5184
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5185 5186 5187 5188 5189 5190 5191
 * that process accounting knows that this is a task in IO wait state.
 *
 * But don't do that if it is a deliberate, throttling IO wait (this task
 * has set its backing_dev_info: the queue against which it should throttle)
 */
void __sched io_schedule(void)
{
5192
	struct rq *rq = &__raw_get_cpu_var(runqueues);
L
Linus Torvalds 已提交
5193

5194
	delayacct_blkio_start();
L
Linus Torvalds 已提交
5195 5196 5197
	atomic_inc(&rq->nr_iowait);
	schedule();
	atomic_dec(&rq->nr_iowait);
5198
	delayacct_blkio_end();
L
Linus Torvalds 已提交
5199 5200 5201 5202 5203
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
5204
	struct rq *rq = &__raw_get_cpu_var(runqueues);
L
Linus Torvalds 已提交
5205 5206
	long ret;

5207
	delayacct_blkio_start();
L
Linus Torvalds 已提交
5208 5209 5210
	atomic_inc(&rq->nr_iowait);
	ret = schedule_timeout(timeout);
	atomic_dec(&rq->nr_iowait);
5211
	delayacct_blkio_end();
L
Linus Torvalds 已提交
5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231
	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.
 */
asmlinkage long sys_sched_get_priority_max(int policy)
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
5232
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5233
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256
		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.
 */
asmlinkage long sys_sched_get_priority_min(int policy)
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
5257
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5258
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274
		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.
 */
asmlinkage
long sys_sched_rr_get_interval(pid_t pid, struct timespec __user *interval)
{
5275
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5276
	unsigned int time_slice;
5277
	int retval;
L
Linus Torvalds 已提交
5278 5279 5280
	struct timespec t;

	if (pid < 0)
5281
		return -EINVAL;
L
Linus Torvalds 已提交
5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292

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

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

5293 5294 5295 5296 5297 5298
	/*
	 * Time slice is 0 for SCHED_FIFO tasks and for SCHED_OTHER
	 * tasks that are on an otherwise idle runqueue:
	 */
	time_slice = 0;
	if (p->policy == SCHED_RR) {
D
Dmitry Adamushko 已提交
5299
		time_slice = DEF_TIMESLICE;
5300
	} else if (p->policy != SCHED_FIFO) {
D
Dmitry Adamushko 已提交
5301 5302 5303 5304 5305
		struct sched_entity *se = &p->se;
		unsigned long flags;
		struct rq *rq;

		rq = task_rq_lock(p, &flags);
5306 5307
		if (rq->cfs.load.weight)
			time_slice = NS_TO_JIFFIES(sched_slice(&rq->cfs, se));
D
Dmitry Adamushko 已提交
5308 5309
		task_rq_unlock(rq, &flags);
	}
L
Linus Torvalds 已提交
5310
	read_unlock(&tasklist_lock);
D
Dmitry Adamushko 已提交
5311
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5312 5313
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5314

L
Linus Torvalds 已提交
5315 5316 5317 5318 5319
out_unlock:
	read_unlock(&tasklist_lock);
	return retval;
}

5320
static const char stat_nam[] = "RSDTtZX";
5321

5322
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5323 5324
{
	unsigned long free = 0;
5325
	unsigned state;
L
Linus Torvalds 已提交
5326 5327

	state = p->state ? __ffs(p->state) + 1 : 0;
I
Ingo Molnar 已提交
5328
	printk(KERN_INFO "%-13.13s %c", p->comm,
5329
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
5330
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
5331
	if (state == TASK_RUNNING)
I
Ingo Molnar 已提交
5332
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
5333
	else
I
Ingo Molnar 已提交
5334
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5335 5336
#else
	if (state == TASK_RUNNING)
I
Ingo Molnar 已提交
5337
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5338
	else
I
Ingo Molnar 已提交
5339
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5340 5341 5342
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
	{
5343
		unsigned long *n = end_of_stack(p);
L
Linus Torvalds 已提交
5344 5345
		while (!*n)
			n++;
5346
		free = (unsigned long)n - (unsigned long)end_of_stack(p);
L
Linus Torvalds 已提交
5347 5348
	}
#endif
5349
	printk(KERN_CONT "%5lu %5d %6d\n", free,
R
Roland McGrath 已提交
5350
		task_pid_nr(p), task_pid_nr(p->real_parent));
L
Linus Torvalds 已提交
5351

5352
	show_stack(p, NULL);
L
Linus Torvalds 已提交
5353 5354
}

I
Ingo Molnar 已提交
5355
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5356
{
5357
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5358

5359 5360 5361
#if BITS_PER_LONG == 32
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5362
#else
5363 5364
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5365 5366 5367 5368 5369 5370 5371 5372
#endif
	read_lock(&tasklist_lock);
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
		 * console might take alot of time:
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
5373
		if (!state_filter || (p->state & state_filter))
5374
			sched_show_task(p);
L
Linus Torvalds 已提交
5375 5376
	} while_each_thread(g, p);

5377 5378
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
5379 5380 5381
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
L
Linus Torvalds 已提交
5382
	read_unlock(&tasklist_lock);
I
Ingo Molnar 已提交
5383 5384 5385 5386 5387
	/*
	 * Only show locks if all tasks are dumped:
	 */
	if (state_filter == -1)
		debug_show_all_locks();
L
Linus Torvalds 已提交
5388 5389
}

I
Ingo Molnar 已提交
5390 5391
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
5392
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5393 5394
}

5395 5396 5397 5398 5399 5400 5401 5402
/**
 * 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.
 */
5403
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5404
{
5405
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5406 5407
	unsigned long flags;

I
Ingo Molnar 已提交
5408 5409 5410
	__sched_fork(idle);
	idle->se.exec_start = sched_clock();

5411
	idle->prio = idle->normal_prio = MAX_PRIO;
L
Linus Torvalds 已提交
5412
	idle->cpus_allowed = cpumask_of_cpu(cpu);
I
Ingo Molnar 已提交
5413
	__set_task_cpu(idle, cpu);
L
Linus Torvalds 已提交
5414 5415 5416

	spin_lock_irqsave(&rq->lock, flags);
	rq->curr = rq->idle = idle;
5417 5418 5419
#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
	idle->oncpu = 1;
#endif
L
Linus Torvalds 已提交
5420 5421 5422
	spin_unlock_irqrestore(&rq->lock, flags);

	/* Set the preempt count _outside_ the spinlocks! */
A
Al Viro 已提交
5423
	task_thread_info(idle)->preempt_count = 0;
5424

I
Ingo Molnar 已提交
5425 5426 5427 5428
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
L
Linus Torvalds 已提交
5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439
}

/*
 * 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
 * always be CPU_MASK_NONE.
 */
cpumask_t nohz_cpu_mask = CPU_MASK_NONE;

I
Ingo Molnar 已提交
5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464
/*
 * 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:
 */
static inline void sched_init_granularity(void)
{
	unsigned int factor = 1 + ilog2(num_online_cpus());
	const unsigned long limit = 200000000;

	sysctl_sched_min_granularity *= factor;
	if (sysctl_sched_min_granularity > limit)
		sysctl_sched_min_granularity = limit;

	sysctl_sched_latency *= factor;
	if (sysctl_sched_latency > limit)
		sysctl_sched_latency = limit;

	sysctl_sched_wakeup_granularity *= factor;
}

L
Linus Torvalds 已提交
5465 5466 5467 5468
#ifdef CONFIG_SMP
/*
 * This is how migration works:
 *
5469
 * 1) we queue a struct migration_req structure in the source CPU's
L
Linus Torvalds 已提交
5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487
 *    runqueue and wake up that CPU's migration thread.
 * 2) we down() the locked semaphore => thread blocks.
 * 3) migration thread wakes up (implicitly it forces the migrated
 *    thread off the CPU)
 * 4) it gets the migration request and checks whether the migrated
 *    task is still in the wrong runqueue.
 * 5) if it's in the wrong runqueue then the migration thread removes
 *    it and puts it into the right queue.
 * 6) migration thread up()s the semaphore.
 * 7) we wake up and the migration is done.
 */

/*
 * 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
5488
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
5489 5490
 * call is not atomic; no spinlocks may be held.
 */
5491
int set_cpus_allowed_ptr(struct task_struct *p, const cpumask_t *new_mask)
L
Linus Torvalds 已提交
5492
{
5493
	struct migration_req req;
L
Linus Torvalds 已提交
5494
	unsigned long flags;
5495
	struct rq *rq;
5496
	int ret = 0;
L
Linus Torvalds 已提交
5497 5498

	rq = task_rq_lock(p, &flags);
5499
	if (!cpus_intersects(*new_mask, cpu_online_map)) {
L
Linus Torvalds 已提交
5500 5501 5502 5503
		ret = -EINVAL;
		goto out;
	}

5504
	if (p->sched_class->set_cpus_allowed)
5505
		p->sched_class->set_cpus_allowed(p, new_mask);
5506
	else {
5507 5508
		p->cpus_allowed = *new_mask;
		p->rt.nr_cpus_allowed = cpus_weight(*new_mask);
5509 5510
	}

L
Linus Torvalds 已提交
5511
	/* Can the task run on the task's current CPU? If so, we're done */
5512
	if (cpu_isset(task_cpu(p), *new_mask))
L
Linus Torvalds 已提交
5513 5514
		goto out;

5515
	if (migrate_task(p, any_online_cpu(*new_mask), &req)) {
L
Linus Torvalds 已提交
5516 5517 5518 5519 5520 5521 5522 5523 5524
		/* Need help from migration thread: drop lock and wait. */
		task_rq_unlock(rq, &flags);
		wake_up_process(rq->migration_thread);
		wait_for_completion(&req.done);
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
	task_rq_unlock(rq, &flags);
5525

L
Linus Torvalds 已提交
5526 5527
	return ret;
}
5528
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
5529 5530

/*
I
Ingo Molnar 已提交
5531
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
5532 5533 5534 5535 5536 5537
 * 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.
5538 5539
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
5540
 */
5541
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
5542
{
5543
	struct rq *rq_dest, *rq_src;
I
Ingo Molnar 已提交
5544
	int ret = 0, on_rq;
L
Linus Torvalds 已提交
5545 5546

	if (unlikely(cpu_is_offline(dest_cpu)))
5547
		return ret;
L
Linus Torvalds 已提交
5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559

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

	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
		goto out;
	/* Affinity changed (again). */
	if (!cpu_isset(dest_cpu, p->cpus_allowed))
		goto out;

I
Ingo Molnar 已提交
5560
	on_rq = p->se.on_rq;
5561
	if (on_rq)
5562
		deactivate_task(rq_src, p, 0);
5563

L
Linus Torvalds 已提交
5564
	set_task_cpu(p, dest_cpu);
I
Ingo Molnar 已提交
5565 5566 5567
	if (on_rq) {
		activate_task(rq_dest, p, 0);
		check_preempt_curr(rq_dest, p);
L
Linus Torvalds 已提交
5568
	}
5569
	ret = 1;
L
Linus Torvalds 已提交
5570 5571
out:
	double_rq_unlock(rq_src, rq_dest);
5572
	return ret;
L
Linus Torvalds 已提交
5573 5574 5575 5576 5577 5578 5579
}

/*
 * migration_thread - this is a highprio system thread that performs
 * thread migration by bumping thread off CPU then 'pushing' onto
 * another runqueue.
 */
I
Ingo Molnar 已提交
5580
static int migration_thread(void *data)
L
Linus Torvalds 已提交
5581 5582
{
	int cpu = (long)data;
5583
	struct rq *rq;
L
Linus Torvalds 已提交
5584 5585 5586 5587 5588 5589

	rq = cpu_rq(cpu);
	BUG_ON(rq->migration_thread != current);

	set_current_state(TASK_INTERRUPTIBLE);
	while (!kthread_should_stop()) {
5590
		struct migration_req *req;
L
Linus Torvalds 已提交
5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612
		struct list_head *head;

		spin_lock_irq(&rq->lock);

		if (cpu_is_offline(cpu)) {
			spin_unlock_irq(&rq->lock);
			goto wait_to_die;
		}

		if (rq->active_balance) {
			active_load_balance(rq, cpu);
			rq->active_balance = 0;
		}

		head = &rq->migration_queue;

		if (list_empty(head)) {
			spin_unlock_irq(&rq->lock);
			schedule();
			set_current_state(TASK_INTERRUPTIBLE);
			continue;
		}
5613
		req = list_entry(head->next, struct migration_req, list);
L
Linus Torvalds 已提交
5614 5615
		list_del_init(head->next);

N
Nick Piggin 已提交
5616 5617 5618
		spin_unlock(&rq->lock);
		__migrate_task(req->task, cpu, req->dest_cpu);
		local_irq_enable();
L
Linus Torvalds 已提交
5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636

		complete(&req->done);
	}
	__set_current_state(TASK_RUNNING);
	return 0;

wait_to_die:
	/* Wait for kthread_stop */
	set_current_state(TASK_INTERRUPTIBLE);
	while (!kthread_should_stop()) {
		schedule();
		set_current_state(TASK_INTERRUPTIBLE);
	}
	__set_current_state(TASK_RUNNING);
	return 0;
}

#ifdef CONFIG_HOTPLUG_CPU
5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647

static int __migrate_task_irq(struct task_struct *p, int src_cpu, int dest_cpu)
{
	int ret;

	local_irq_disable();
	ret = __migrate_task(p, src_cpu, dest_cpu);
	local_irq_enable();
	return ret;
}

5648
/*
5649
 * Figure out where task on dead CPU should go, use force if necessary.
5650 5651
 * NOTE: interrupts should be disabled by the caller
 */
5652
static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p)
L
Linus Torvalds 已提交
5653
{
5654
	unsigned long flags;
L
Linus Torvalds 已提交
5655
	cpumask_t mask;
5656 5657
	struct rq *rq;
	int dest_cpu;
L
Linus Torvalds 已提交
5658

5659 5660 5661 5662 5663 5664 5665
	do {
		/* On same node? */
		mask = node_to_cpumask(cpu_to_node(dead_cpu));
		cpus_and(mask, mask, p->cpus_allowed);
		dest_cpu = any_online_cpu(mask);

		/* On any allowed CPU? */
5666
		if (dest_cpu >= nr_cpu_ids)
5667 5668 5669
			dest_cpu = any_online_cpu(p->cpus_allowed);

		/* No more Mr. Nice Guy. */
5670
		if (dest_cpu >= nr_cpu_ids) {
5671 5672 5673
			cpumask_t cpus_allowed;

			cpuset_cpus_allowed_locked(p, &cpus_allowed);
5674 5675 5676 5677
			/*
			 * Try to stay on the same cpuset, where the
			 * current cpuset may be a subset of all cpus.
			 * The cpuset_cpus_allowed_locked() variant of
I
Ingo Molnar 已提交
5678
			 * cpuset_cpus_allowed() will not block. It must be
5679 5680
			 * called within calls to cpuset_lock/cpuset_unlock.
			 */
5681
			rq = task_rq_lock(p, &flags);
5682
			p->cpus_allowed = cpus_allowed;
5683 5684
			dest_cpu = any_online_cpu(p->cpus_allowed);
			task_rq_unlock(rq, &flags);
L
Linus Torvalds 已提交
5685

5686 5687 5688 5689 5690
			/*
			 * Don't tell them about moving exiting tasks or
			 * kernel threads (both mm NULL), since they never
			 * leave kernel.
			 */
I
Ingo Molnar 已提交
5691
			if (p->mm && printk_ratelimit()) {
5692 5693
				printk(KERN_INFO "process %d (%s) no "
				       "longer affine to cpu%d\n",
I
Ingo Molnar 已提交
5694 5695
					task_pid_nr(p), p->comm, dead_cpu);
			}
5696
		}
5697
	} while (!__migrate_task_irq(p, dead_cpu, dest_cpu));
L
Linus Torvalds 已提交
5698 5699 5700 5701 5702 5703 5704 5705 5706
}

/*
 * 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:
 */
5707
static void migrate_nr_uninterruptible(struct rq *rq_src)
L
Linus Torvalds 已提交
5708
{
5709
	struct rq *rq_dest = cpu_rq(any_online_cpu(*CPU_MASK_ALL_PTR));
L
Linus Torvalds 已提交
5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722
	unsigned long flags;

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

/* Run through task list and migrate tasks from the dead cpu. */
static void migrate_live_tasks(int src_cpu)
{
5723
	struct task_struct *p, *t;
L
Linus Torvalds 已提交
5724

5725
	read_lock(&tasklist_lock);
L
Linus Torvalds 已提交
5726

5727 5728
	do_each_thread(t, p) {
		if (p == current)
L
Linus Torvalds 已提交
5729 5730
			continue;

5731 5732 5733
		if (task_cpu(p) == src_cpu)
			move_task_off_dead_cpu(src_cpu, p);
	} while_each_thread(t, p);
L
Linus Torvalds 已提交
5734

5735
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
5736 5737
}

I
Ingo Molnar 已提交
5738 5739
/*
 * Schedules idle task to be the next runnable task on current CPU.
5740 5741
 * It does so by boosting its priority to highest possible.
 * Used by CPU offline code.
L
Linus Torvalds 已提交
5742 5743 5744
 */
void sched_idle_next(void)
{
5745
	int this_cpu = smp_processor_id();
5746
	struct rq *rq = cpu_rq(this_cpu);
L
Linus Torvalds 已提交
5747 5748 5749 5750
	struct task_struct *p = rq->idle;
	unsigned long flags;

	/* cpu has to be offline */
5751
	BUG_ON(cpu_online(this_cpu));
L
Linus Torvalds 已提交
5752

5753 5754 5755
	/*
	 * Strictly not necessary since rest of the CPUs are stopped by now
	 * and interrupts disabled on the current cpu.
L
Linus Torvalds 已提交
5756 5757 5758
	 */
	spin_lock_irqsave(&rq->lock, flags);

I
Ingo Molnar 已提交
5759
	__setscheduler(rq, p, SCHED_FIFO, MAX_RT_PRIO-1);
5760

5761 5762
	update_rq_clock(rq);
	activate_task(rq, p, 0);
L
Linus Torvalds 已提交
5763 5764 5765 5766

	spin_unlock_irqrestore(&rq->lock, flags);
}

5767 5768
/*
 * Ensures that the idle task is using init_mm right before its cpu goes
L
Linus Torvalds 已提交
5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781
 * offline.
 */
void idle_task_exit(void)
{
	struct mm_struct *mm = current->active_mm;

	BUG_ON(cpu_online(smp_processor_id()));

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

5782
/* called under rq->lock with disabled interrupts */
5783
static void migrate_dead(unsigned int dead_cpu, struct task_struct *p)
L
Linus Torvalds 已提交
5784
{
5785
	struct rq *rq = cpu_rq(dead_cpu);
L
Linus Torvalds 已提交
5786 5787

	/* Must be exiting, otherwise would be on tasklist. */
E
Eugene Teo 已提交
5788
	BUG_ON(!p->exit_state);
L
Linus Torvalds 已提交
5789 5790

	/* Cannot have done final schedule yet: would have vanished. */
5791
	BUG_ON(p->state == TASK_DEAD);
L
Linus Torvalds 已提交
5792

5793
	get_task_struct(p);
L
Linus Torvalds 已提交
5794 5795 5796

	/*
	 * Drop lock around migration; if someone else moves it,
I
Ingo Molnar 已提交
5797
	 * that's OK. No task can be added to this CPU, so iteration is
L
Linus Torvalds 已提交
5798 5799
	 * fine.
	 */
5800
	spin_unlock_irq(&rq->lock);
5801
	move_task_off_dead_cpu(dead_cpu, p);
5802
	spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
5803

5804
	put_task_struct(p);
L
Linus Torvalds 已提交
5805 5806 5807 5808 5809
}

/* release_task() removes task from tasklist, so we won't find dead tasks. */
static void migrate_dead_tasks(unsigned int dead_cpu)
{
5810
	struct rq *rq = cpu_rq(dead_cpu);
I
Ingo Molnar 已提交
5811
	struct task_struct *next;
5812

I
Ingo Molnar 已提交
5813 5814 5815
	for ( ; ; ) {
		if (!rq->nr_running)
			break;
I
Ingo Molnar 已提交
5816
		update_rq_clock(rq);
5817
		next = pick_next_task(rq, rq->curr);
I
Ingo Molnar 已提交
5818 5819 5820
		if (!next)
			break;
		migrate_dead(dead_cpu, next);
5821

L
Linus Torvalds 已提交
5822 5823 5824 5825
	}
}
#endif /* CONFIG_HOTPLUG_CPU */

5826 5827 5828
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5829 5830
	{
		.procname	= "sched_domain",
5831
		.mode		= 0555,
5832
	},
I
Ingo Molnar 已提交
5833
	{0, },
5834 5835 5836
};

static struct ctl_table sd_ctl_root[] = {
5837
	{
5838
		.ctl_name	= CTL_KERN,
5839
		.procname	= "kernel",
5840
		.mode		= 0555,
5841 5842
		.child		= sd_ctl_dir,
	},
I
Ingo Molnar 已提交
5843
	{0, },
5844 5845 5846 5847 5848
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5849
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5850 5851 5852 5853

	return entry;
}

5854 5855
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5856
	struct ctl_table *entry;
5857

5858 5859 5860
	/*
	 * 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
5861
	 * will always be set. In the lowest directory the names are
5862 5863 5864
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5865 5866
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5867 5868 5869
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5870 5871 5872 5873 5874

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

5875
static void
5876
set_table_entry(struct ctl_table *entry,
5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889
		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)
{
5890
	struct ctl_table *table = sd_alloc_ctl_entry(12);
5891

5892 5893 5894
	if (table == NULL)
		return NULL;

5895
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5896
		sizeof(long), 0644, proc_doulongvec_minmax);
5897
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5898
		sizeof(long), 0644, proc_doulongvec_minmax);
5899
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5900
		sizeof(int), 0644, proc_dointvec_minmax);
5901
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5902
		sizeof(int), 0644, proc_dointvec_minmax);
5903
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5904
		sizeof(int), 0644, proc_dointvec_minmax);
5905
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5906
		sizeof(int), 0644, proc_dointvec_minmax);
5907
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5908
		sizeof(int), 0644, proc_dointvec_minmax);
5909
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5910
		sizeof(int), 0644, proc_dointvec_minmax);
5911
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5912
		sizeof(int), 0644, proc_dointvec_minmax);
5913
	set_table_entry(&table[9], "cache_nice_tries",
5914 5915
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5916
	set_table_entry(&table[10], "flags", &sd->flags,
5917
		sizeof(int), 0644, proc_dointvec_minmax);
5918
	/* &table[11] is terminator */
5919 5920 5921 5922

	return table;
}

5923
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5924 5925 5926 5927 5928 5929 5930 5931 5932
{
	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);
5933 5934
	if (table == NULL)
		return NULL;
5935 5936 5937 5938 5939

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5940
		entry->mode = 0555;
5941 5942 5943 5944 5945 5946 5947 5948
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5949
static void register_sched_domain_sysctl(void)
5950 5951 5952 5953 5954
{
	int i, cpu_num = num_online_cpus();
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5955 5956 5957
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5958 5959 5960
	if (entry == NULL)
		return;

5961
	for_each_online_cpu(i) {
5962 5963
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5964
		entry->mode = 0555;
5965
		entry->child = sd_alloc_ctl_cpu_table(i);
5966
		entry++;
5967
	}
5968 5969

	WARN_ON(sd_sysctl_header);
5970 5971
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5972

5973
/* may be called multiple times per register */
5974 5975
static void unregister_sched_domain_sysctl(void)
{
5976 5977
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5978
	sd_sysctl_header = NULL;
5979 5980
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5981
}
5982
#else
5983 5984 5985 5986
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5987 5988 5989 5990
{
}
#endif

L
Linus Torvalds 已提交
5991 5992 5993 5994
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5995 5996
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5997 5998
{
	struct task_struct *p;
5999
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
6000
	unsigned long flags;
6001
	struct rq *rq;
L
Linus Torvalds 已提交
6002 6003

	switch (action) {
6004

L
Linus Torvalds 已提交
6005
	case CPU_UP_PREPARE:
6006
	case CPU_UP_PREPARE_FROZEN:
I
Ingo Molnar 已提交
6007
		p = kthread_create(migration_thread, hcpu, "migration/%d", cpu);
L
Linus Torvalds 已提交
6008 6009 6010 6011 6012
		if (IS_ERR(p))
			return NOTIFY_BAD;
		kthread_bind(p, cpu);
		/* Must be high prio: stop_machine expects to yield to it. */
		rq = task_rq_lock(p, &flags);
I
Ingo Molnar 已提交
6013
		__setscheduler(rq, p, SCHED_FIFO, MAX_RT_PRIO-1);
L
Linus Torvalds 已提交
6014 6015 6016
		task_rq_unlock(rq, &flags);
		cpu_rq(cpu)->migration_thread = p;
		break;
6017

L
Linus Torvalds 已提交
6018
	case CPU_ONLINE:
6019
	case CPU_ONLINE_FROZEN:
6020
		/* Strictly unnecessary, as first user will wake it. */
L
Linus Torvalds 已提交
6021
		wake_up_process(cpu_rq(cpu)->migration_thread);
6022 6023 6024 6025 6026 6027 6028 6029 6030

		/* Update our root-domain */
		rq = cpu_rq(cpu);
		spin_lock_irqsave(&rq->lock, flags);
		if (rq->rd) {
			BUG_ON(!cpu_isset(cpu, rq->rd->span));
			cpu_set(cpu, rq->rd->online);
		}
		spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
6031
		break;
6032

L
Linus Torvalds 已提交
6033 6034
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_UP_CANCELED:
6035
	case CPU_UP_CANCELED_FROZEN:
6036 6037
		if (!cpu_rq(cpu)->migration_thread)
			break;
I
Ingo Molnar 已提交
6038
		/* Unbind it from offline cpu so it can run. Fall thru. */
6039 6040
		kthread_bind(cpu_rq(cpu)->migration_thread,
			     any_online_cpu(cpu_online_map));
L
Linus Torvalds 已提交
6041 6042 6043
		kthread_stop(cpu_rq(cpu)->migration_thread);
		cpu_rq(cpu)->migration_thread = NULL;
		break;
6044

L
Linus Torvalds 已提交
6045
	case CPU_DEAD:
6046
	case CPU_DEAD_FROZEN:
6047
		cpuset_lock(); /* around calls to cpuset_cpus_allowed_lock() */
L
Linus Torvalds 已提交
6048 6049 6050 6051 6052
		migrate_live_tasks(cpu);
		rq = cpu_rq(cpu);
		kthread_stop(rq->migration_thread);
		rq->migration_thread = NULL;
		/* Idle task back to normal (off runqueue, low prio) */
6053
		spin_lock_irq(&rq->lock);
I
Ingo Molnar 已提交
6054
		update_rq_clock(rq);
6055
		deactivate_task(rq, rq->idle, 0);
L
Linus Torvalds 已提交
6056
		rq->idle->static_prio = MAX_PRIO;
I
Ingo Molnar 已提交
6057 6058
		__setscheduler(rq, rq->idle, SCHED_NORMAL, 0);
		rq->idle->sched_class = &idle_sched_class;
L
Linus Torvalds 已提交
6059
		migrate_dead_tasks(cpu);
6060
		spin_unlock_irq(&rq->lock);
6061
		cpuset_unlock();
L
Linus Torvalds 已提交
6062 6063 6064
		migrate_nr_uninterruptible(rq);
		BUG_ON(rq->nr_running != 0);

I
Ingo Molnar 已提交
6065 6066 6067 6068 6069
		/*
		 * No need to migrate the tasks: it was best-effort if
		 * they didn't take sched_hotcpu_mutex. Just wake up
		 * the requestors.
		 */
L
Linus Torvalds 已提交
6070 6071
		spin_lock_irq(&rq->lock);
		while (!list_empty(&rq->migration_queue)) {
6072 6073
			struct migration_req *req;

L
Linus Torvalds 已提交
6074
			req = list_entry(rq->migration_queue.next,
6075
					 struct migration_req, list);
L
Linus Torvalds 已提交
6076 6077 6078 6079 6080
			list_del_init(&req->list);
			complete(&req->done);
		}
		spin_unlock_irq(&rq->lock);
		break;
G
Gregory Haskins 已提交
6081

6082 6083
	case CPU_DYING:
	case CPU_DYING_FROZEN:
G
Gregory Haskins 已提交
6084 6085 6086 6087 6088 6089 6090 6091 6092
		/* Update our root-domain */
		rq = cpu_rq(cpu);
		spin_lock_irqsave(&rq->lock, flags);
		if (rq->rd) {
			BUG_ON(!cpu_isset(cpu, rq->rd->span));
			cpu_clear(cpu, rq->rd->online);
		}
		spin_unlock_irqrestore(&rq->lock, flags);
		break;
L
Linus Torvalds 已提交
6093 6094 6095 6096 6097 6098 6099 6100
#endif
	}
	return NOTIFY_OK;
}

/* Register at highest priority so that task migration (migrate_all_tasks)
 * happens before everything else.
 */
6101
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
6102 6103 6104 6105
	.notifier_call = migration_call,
	.priority = 10
};

6106
void __init migration_init(void)
L
Linus Torvalds 已提交
6107 6108
{
	void *cpu = (void *)(long)smp_processor_id();
6109
	int err;
6110 6111

	/* Start one for the boot CPU: */
6112 6113
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
6114 6115 6116 6117 6118 6119
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
}
#endif

#ifdef CONFIG_SMP
6120

6121
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
6122

6123 6124
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
				  cpumask_t *groupmask)
L
Linus Torvalds 已提交
6125
{
I
Ingo Molnar 已提交
6126
	struct sched_group *group = sd->groups;
6127
	char str[256];
L
Linus Torvalds 已提交
6128

6129
	cpulist_scnprintf(str, sizeof(str), sd->span);
6130
	cpus_clear(*groupmask);
I
Ingo Molnar 已提交
6131 6132 6133 6134 6135 6136 6137 6138 6139

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
		printk("does not load-balance\n");
		if (sd->parent)
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
		return -1;
N
Nick Piggin 已提交
6140 6141
	}

I
Ingo Molnar 已提交
6142 6143 6144 6145 6146 6147 6148 6149 6150 6151
	printk(KERN_CONT "span %s\n", str);

	if (!cpu_isset(cpu, sd->span)) {
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
	}
	if (!cpu_isset(cpu, group->cpumask)) {
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
	}
L
Linus Torvalds 已提交
6152

I
Ingo Molnar 已提交
6153
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
6154
	do {
I
Ingo Molnar 已提交
6155 6156 6157
		if (!group) {
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
6158 6159 6160
			break;
		}

I
Ingo Molnar 已提交
6161 6162 6163 6164 6165 6166
		if (!group->__cpu_power) {
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
			break;
		}
L
Linus Torvalds 已提交
6167

I
Ingo Molnar 已提交
6168 6169 6170 6171 6172
		if (!cpus_weight(group->cpumask)) {
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
			break;
		}
L
Linus Torvalds 已提交
6173

6174
		if (cpus_intersects(*groupmask, group->cpumask)) {
I
Ingo Molnar 已提交
6175 6176 6177 6178
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
			break;
		}
L
Linus Torvalds 已提交
6179

6180
		cpus_or(*groupmask, *groupmask, group->cpumask);
L
Linus Torvalds 已提交
6181

6182
		cpulist_scnprintf(str, sizeof(str), group->cpumask);
I
Ingo Molnar 已提交
6183
		printk(KERN_CONT " %s", str);
L
Linus Torvalds 已提交
6184

I
Ingo Molnar 已提交
6185 6186 6187
		group = group->next;
	} while (group != sd->groups);
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
6188

6189
	if (!cpus_equal(sd->span, *groupmask))
I
Ingo Molnar 已提交
6190
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
6191

6192
	if (sd->parent && !cpus_subset(*groupmask, sd->parent->span))
I
Ingo Molnar 已提交
6193 6194 6195 6196
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
	return 0;
}
L
Linus Torvalds 已提交
6197

I
Ingo Molnar 已提交
6198 6199
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
6200
	cpumask_t *groupmask;
I
Ingo Molnar 已提交
6201
	int level = 0;
L
Linus Torvalds 已提交
6202

I
Ingo Molnar 已提交
6203 6204 6205 6206
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
6207

I
Ingo Molnar 已提交
6208 6209
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

6210 6211 6212 6213 6214 6215
	groupmask = kmalloc(sizeof(cpumask_t), GFP_KERNEL);
	if (!groupmask) {
		printk(KERN_DEBUG "Cannot load-balance (out of memory)\n");
		return;
	}

I
Ingo Molnar 已提交
6216
	for (;;) {
6217
		if (sched_domain_debug_one(sd, cpu, level, groupmask))
I
Ingo Molnar 已提交
6218
			break;
L
Linus Torvalds 已提交
6219 6220
		level++;
		sd = sd->parent;
6221
		if (!sd)
I
Ingo Molnar 已提交
6222 6223
			break;
	}
6224
	kfree(groupmask);
L
Linus Torvalds 已提交
6225 6226
}
#else
6227
# define sched_domain_debug(sd, cpu) do { } while (0)
L
Linus Torvalds 已提交
6228 6229
#endif

6230
static int sd_degenerate(struct sched_domain *sd)
6231 6232 6233 6234 6235 6236 6237 6238
{
	if (cpus_weight(sd->span) == 1)
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
6239 6240 6241
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
	if (sd->flags & (SD_WAKE_IDLE |
			 SD_WAKE_AFFINE |
			 SD_WAKE_BALANCE))
		return 0;

	return 1;
}

6255 6256
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

	if (!cpus_equal(sd->span, parent->span))
		return 0;

	/* Does parent contain flags not in child? */
	/* WAKE_BALANCE is a subset of WAKE_AFFINE */
	if (cflags & SD_WAKE_AFFINE)
		pflags &= ~SD_WAKE_BALANCE;
	/* 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 |
6275 6276 6277
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
6278 6279 6280 6281 6282 6283 6284
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

G
Gregory Haskins 已提交
6285 6286 6287 6288 6289 6290 6291 6292 6293 6294
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
	unsigned long flags;
	const struct sched_class *class;

	spin_lock_irqsave(&rq->lock, flags);

	if (rq->rd) {
		struct root_domain *old_rd = rq->rd;

I
Ingo Molnar 已提交
6295
		for (class = sched_class_highest; class; class = class->next) {
G
Gregory Haskins 已提交
6296 6297
			if (class->leave_domain)
				class->leave_domain(rq);
I
Ingo Molnar 已提交
6298
		}
G
Gregory Haskins 已提交
6299

6300 6301 6302
		cpu_clear(rq->cpu, old_rd->span);
		cpu_clear(rq->cpu, old_rd->online);

G
Gregory Haskins 已提交
6303 6304 6305 6306 6307 6308 6309
		if (atomic_dec_and_test(&old_rd->refcount))
			kfree(old_rd);
	}

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

6310
	cpu_set(rq->cpu, rd->span);
6311 6312
	if (cpu_isset(rq->cpu, cpu_online_map))
		cpu_set(rq->cpu, rd->online);
6313

I
Ingo Molnar 已提交
6314
	for (class = sched_class_highest; class; class = class->next) {
G
Gregory Haskins 已提交
6315 6316
		if (class->join_domain)
			class->join_domain(rq);
I
Ingo Molnar 已提交
6317
	}
G
Gregory Haskins 已提交
6318 6319 6320 6321

	spin_unlock_irqrestore(&rq->lock, flags);
}

6322
static void init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
6323 6324 6325
{
	memset(rd, 0, sizeof(*rd));

6326 6327
	cpus_clear(rd->span);
	cpus_clear(rd->online);
G
Gregory Haskins 已提交
6328 6329 6330 6331
}

static void init_defrootdomain(void)
{
6332
	init_rootdomain(&def_root_domain);
G
Gregory Haskins 已提交
6333 6334 6335
	atomic_set(&def_root_domain.refcount, 1);
}

6336
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
6337 6338 6339 6340 6341 6342 6343
{
	struct root_domain *rd;

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

6344
	init_rootdomain(rd);
G
Gregory Haskins 已提交
6345 6346 6347 6348

	return rd;
}

L
Linus Torvalds 已提交
6349
/*
I
Ingo Molnar 已提交
6350
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
6351 6352
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
6353 6354
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
6355
{
6356
	struct rq *rq = cpu_rq(cpu);
6357 6358 6359 6360 6361 6362 6363
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
	for (tmp = sd; tmp; tmp = tmp->parent) {
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
6364
		if (sd_parent_degenerate(tmp, parent)) {
6365
			tmp->parent = parent->parent;
6366 6367 6368
			if (parent->parent)
				parent->parent->child = tmp;
		}
6369 6370
	}

6371
	if (sd && sd_degenerate(sd)) {
6372
		sd = sd->parent;
6373 6374 6375
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
6376 6377 6378

	sched_domain_debug(sd, cpu);

G
Gregory Haskins 已提交
6379
	rq_attach_root(rq, rd);
N
Nick Piggin 已提交
6380
	rcu_assign_pointer(rq->sd, sd);
L
Linus Torvalds 已提交
6381 6382 6383
}

/* cpus with isolated domains */
6384
static cpumask_t cpu_isolated_map = CPU_MASK_NONE;
L
Linus Torvalds 已提交
6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
	int ints[NR_CPUS], i;

	str = get_options(str, ARRAY_SIZE(ints), ints);
	cpus_clear(cpu_isolated_map);
	for (i = 1; i <= ints[0]; i++)
		if (ints[i] < NR_CPUS)
			cpu_set(ints[i], cpu_isolated_map);
	return 1;
}

I
Ingo Molnar 已提交
6399
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
6400 6401

/*
6402 6403 6404 6405
 * 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
 * belongs to. The return value of group_fn must be a >= 0 and < NR_CPUS
 * (due to the fact that we keep track of groups covered with a cpumask_t).
L
Linus Torvalds 已提交
6406 6407 6408 6409 6410
 *
 * 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.
 */
6411
static void
6412
init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map,
6413
			int (*group_fn)(int cpu, const cpumask_t *cpu_map,
6414 6415 6416
					struct sched_group **sg,
					cpumask_t *tmpmask),
			cpumask_t *covered, cpumask_t *tmpmask)
L
Linus Torvalds 已提交
6417 6418 6419 6420
{
	struct sched_group *first = NULL, *last = NULL;
	int i;

6421 6422 6423
	cpus_clear(*covered);

	for_each_cpu_mask(i, *span) {
6424
		struct sched_group *sg;
6425
		int group = group_fn(i, cpu_map, &sg, tmpmask);
L
Linus Torvalds 已提交
6426 6427
		int j;

6428
		if (cpu_isset(i, *covered))
L
Linus Torvalds 已提交
6429 6430
			continue;

6431
		cpus_clear(sg->cpumask);
6432
		sg->__cpu_power = 0;
L
Linus Torvalds 已提交
6433

6434 6435
		for_each_cpu_mask(j, *span) {
			if (group_fn(j, cpu_map, NULL, tmpmask) != group)
L
Linus Torvalds 已提交
6436 6437
				continue;

6438
			cpu_set(j, *covered);
L
Linus Torvalds 已提交
6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449
			cpu_set(j, sg->cpumask);
		}
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
}

6450
#define SD_NODES_PER_DOMAIN 16
L
Linus Torvalds 已提交
6451

6452
#ifdef CONFIG_NUMA
6453

6454 6455 6456 6457 6458
/**
 * 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
6459
 * Find the next node to include in a given scheduling domain. Simply
6460 6461 6462 6463
 * finds the closest node not already in the @used_nodes map.
 *
 * Should use nodemask_t.
 */
6464
static int find_next_best_node(int node, nodemask_t *used_nodes)
6465 6466 6467 6468 6469 6470 6471 6472 6473 6474 6475 6476 6477
{
	int i, n, val, min_val, best_node = 0;

	min_val = INT_MAX;

	for (i = 0; i < MAX_NUMNODES; i++) {
		/* Start at @node */
		n = (node + i) % MAX_NUMNODES;

		if (!nr_cpus_node(n))
			continue;

		/* Skip already used nodes */
6478
		if (node_isset(n, *used_nodes))
6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489
			continue;

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

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

6490
	node_set(best_node, *used_nodes);
6491 6492 6493 6494 6495 6496 6497
	return best_node;
}

/**
 * sched_domain_node_span - get a cpumask for a node's sched_domain
 * @node: node whose cpumask we're constructing
 *
I
Ingo Molnar 已提交
6498
 * Given a node, construct a good cpumask for its sched_domain to span. It
6499 6500 6501
 * should be one that prevents unnecessary balancing, but also spreads tasks
 * out optimally.
 */
6502
static void sched_domain_node_span(int node, cpumask_t *span)
6503
{
6504 6505
	nodemask_t used_nodes;
	node_to_cpumask_ptr(nodemask, node);
6506
	int i;
6507

6508
	cpus_clear(*span);
6509
	nodes_clear(used_nodes);
6510

6511
	cpus_or(*span, *span, *nodemask);
6512
	node_set(node, used_nodes);
6513 6514

	for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
6515
		int next_node = find_next_best_node(node, &used_nodes);
6516

6517
		node_to_cpumask_ptr_next(nodemask, next_node);
6518
		cpus_or(*span, *span, *nodemask);
6519 6520 6521 6522
	}
}
#endif

6523
int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
6524

6525
/*
6526
 * SMT sched-domains:
6527
 */
L
Linus Torvalds 已提交
6528 6529
#ifdef CONFIG_SCHED_SMT
static DEFINE_PER_CPU(struct sched_domain, cpu_domains);
6530
static DEFINE_PER_CPU(struct sched_group, sched_group_cpus);
6531

I
Ingo Molnar 已提交
6532
static int
6533 6534
cpu_to_cpu_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg,
		 cpumask_t *unused)
L
Linus Torvalds 已提交
6535
{
6536 6537
	if (sg)
		*sg = &per_cpu(sched_group_cpus, cpu);
L
Linus Torvalds 已提交
6538 6539 6540 6541
	return cpu;
}
#endif

6542 6543 6544
/*
 * multi-core sched-domains:
 */
6545 6546
#ifdef CONFIG_SCHED_MC
static DEFINE_PER_CPU(struct sched_domain, core_domains);
6547
static DEFINE_PER_CPU(struct sched_group, sched_group_core);
6548 6549 6550
#endif

#if defined(CONFIG_SCHED_MC) && defined(CONFIG_SCHED_SMT)
I
Ingo Molnar 已提交
6551
static int
6552 6553
cpu_to_core_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg,
		  cpumask_t *mask)
6554
{
6555
	int group;
6556 6557 6558 6559

	*mask = per_cpu(cpu_sibling_map, cpu);
	cpus_and(*mask, *mask, *cpu_map);
	group = first_cpu(*mask);
6560 6561 6562
	if (sg)
		*sg = &per_cpu(sched_group_core, group);
	return group;
6563 6564
}
#elif defined(CONFIG_SCHED_MC)
I
Ingo Molnar 已提交
6565
static int
6566 6567
cpu_to_core_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg,
		  cpumask_t *unused)
6568
{
6569 6570
	if (sg)
		*sg = &per_cpu(sched_group_core, cpu);
6571 6572 6573 6574
	return cpu;
}
#endif

L
Linus Torvalds 已提交
6575
static DEFINE_PER_CPU(struct sched_domain, phys_domains);
6576
static DEFINE_PER_CPU(struct sched_group, sched_group_phys);
6577

I
Ingo Molnar 已提交
6578
static int
6579 6580
cpu_to_phys_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg,
		  cpumask_t *mask)
L
Linus Torvalds 已提交
6581
{
6582
	int group;
6583
#ifdef CONFIG_SCHED_MC
6584 6585 6586
	*mask = cpu_coregroup_map(cpu);
	cpus_and(*mask, *mask, *cpu_map);
	group = first_cpu(*mask);
6587
#elif defined(CONFIG_SCHED_SMT)
6588 6589 6590
	*mask = per_cpu(cpu_sibling_map, cpu);
	cpus_and(*mask, *mask, *cpu_map);
	group = first_cpu(*mask);
L
Linus Torvalds 已提交
6591
#else
6592
	group = cpu;
L
Linus Torvalds 已提交
6593
#endif
6594 6595 6596
	if (sg)
		*sg = &per_cpu(sched_group_phys, group);
	return group;
L
Linus Torvalds 已提交
6597 6598 6599 6600
}

#ifdef CONFIG_NUMA
/*
6601 6602 6603
 * 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
6604
 */
6605
static DEFINE_PER_CPU(struct sched_domain, node_domains);
6606
static struct sched_group ***sched_group_nodes_bycpu;
L
Linus Torvalds 已提交
6607

6608
static DEFINE_PER_CPU(struct sched_domain, allnodes_domains);
6609
static DEFINE_PER_CPU(struct sched_group, sched_group_allnodes);
6610

6611
static int cpu_to_allnodes_group(int cpu, const cpumask_t *cpu_map,
6612
				 struct sched_group **sg, cpumask_t *nodemask)
6613
{
6614 6615
	int group;

6616 6617 6618
	*nodemask = node_to_cpumask(cpu_to_node(cpu));
	cpus_and(*nodemask, *nodemask, *cpu_map);
	group = first_cpu(*nodemask);
6619 6620 6621 6622

	if (sg)
		*sg = &per_cpu(sched_group_allnodes, group);
	return group;
L
Linus Torvalds 已提交
6623
}
6624

6625 6626 6627 6628 6629 6630 6631
static void init_numa_sched_groups_power(struct sched_group *group_head)
{
	struct sched_group *sg = group_head;
	int j;

	if (!sg)
		return;
6632 6633 6634
	do {
		for_each_cpu_mask(j, sg->cpumask) {
			struct sched_domain *sd;
6635

6636 6637 6638 6639 6640 6641 6642 6643
			sd = &per_cpu(phys_domains, j);
			if (j != first_cpu(sd->groups->cpumask)) {
				/*
				 * Only add "power" once for each
				 * physical package.
				 */
				continue;
			}
6644

6645 6646 6647 6648
			sg_inc_cpu_power(sg, sd->groups->__cpu_power);
		}
		sg = sg->next;
	} while (sg != group_head);
6649
}
L
Linus Torvalds 已提交
6650 6651
#endif

6652
#ifdef CONFIG_NUMA
6653
/* Free memory allocated for various sched_group structures */
6654
static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask)
6655
{
6656
	int cpu, i;
6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667

	for_each_cpu_mask(cpu, *cpu_map) {
		struct sched_group **sched_group_nodes
			= sched_group_nodes_bycpu[cpu];

		if (!sched_group_nodes)
			continue;

		for (i = 0; i < MAX_NUMNODES; i++) {
			struct sched_group *oldsg, *sg = sched_group_nodes[i];

6668 6669 6670
			*nodemask = node_to_cpumask(i);
			cpus_and(*nodemask, *nodemask, *cpu_map);
			if (cpus_empty(*nodemask))
6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686
				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;
	}
}
6687
#else
6688
static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask)
6689 6690 6691
{
}
#endif
6692

6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705 6706 6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718
/*
 * 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.
 *
 * cpu_power will be a multiple of SCHED_LOAD_SCALE. This multiple represents
 * the maximum number of tasks a group can handle in the presence of other idle
 * or lightly loaded groups in the same sched domain.
 */
static void init_sched_groups_power(int cpu, struct sched_domain *sd)
{
	struct sched_domain *child;
	struct sched_group *group;

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

	if (cpu != first_cpu(sd->groups->cpumask))
		return;

	child = sd->child;

6719 6720
	sd->groups->__cpu_power = 0;

6721 6722 6723 6724 6725 6726 6727 6728 6729 6730
	/*
	 * For perf policy, if the groups in child domain share resources
	 * (for example cores sharing some portions of the cache hierarchy
	 * or SMT), then set this domain groups cpu_power such that each group
	 * can handle only one task, when there are other idle groups in the
	 * same sched domain.
	 */
	if (!child || (!(sd->flags & SD_POWERSAVINGS_BALANCE) &&
		       (child->flags &
			(SD_SHARE_CPUPOWER | SD_SHARE_PKG_RESOURCES)))) {
6731
		sg_inc_cpu_power(sd->groups, SCHED_LOAD_SCALE);
6732 6733 6734 6735 6736 6737 6738 6739
		return;
	}

	/*
	 * add cpu_power of each child group to this groups cpu_power
	 */
	group = child->groups;
	do {
6740
		sg_inc_cpu_power(sd->groups, group->__cpu_power);
6741 6742 6743 6744
		group = group->next;
	} while (group != child->groups);
}

6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784 6785 6786 6787 6788 6789 6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

#define	SD_INIT(sd, type)	sd_init_##type(sd)
#define SD_INIT_FUNC(type)	\
static noinline void sd_init_##type(struct sched_domain *sd)	\
{								\
	memset(sd, 0, sizeof(*sd));				\
	*sd = SD_##type##_INIT;					\
}

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

/*
 * To minimize stack usage kmalloc room for cpumasks and share the
 * space as the usage in build_sched_domains() dictates.  Used only
 * if the amount of space is significant.
 */
struct allmasks {
	cpumask_t tmpmask;			/* make this one first */
	union {
		cpumask_t nodemask;
		cpumask_t this_sibling_map;
		cpumask_t this_core_map;
	};
	cpumask_t send_covered;

#ifdef CONFIG_NUMA
	cpumask_t domainspan;
	cpumask_t covered;
	cpumask_t notcovered;
#endif
};

#if	NR_CPUS > 128
#define	SCHED_CPUMASK_ALLOC		1
#define	SCHED_CPUMASK_FREE(v)		kfree(v)
#define	SCHED_CPUMASK_DECLARE(v)	struct allmasks *v
#else
#define	SCHED_CPUMASK_ALLOC		0
#define	SCHED_CPUMASK_FREE(v)
#define	SCHED_CPUMASK_DECLARE(v)	struct allmasks _v, *v = &_v
#endif

#define	SCHED_CPUMASK_VAR(v, a) 	cpumask_t *v = (cpumask_t *) \
			((unsigned long)(a) + offsetof(struct allmasks, v))

L
Linus Torvalds 已提交
6804
/*
6805 6806
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
L
Linus Torvalds 已提交
6807
 */
6808
static int build_sched_domains(const cpumask_t *cpu_map)
L
Linus Torvalds 已提交
6809 6810
{
	int i;
G
Gregory Haskins 已提交
6811
	struct root_domain *rd;
6812 6813
	SCHED_CPUMASK_DECLARE(allmasks);
	cpumask_t *tmpmask;
6814 6815
#ifdef CONFIG_NUMA
	struct sched_group **sched_group_nodes = NULL;
6816
	int sd_allnodes = 0;
6817 6818 6819 6820

	/*
	 * Allocate the per-node list of sched groups
	 */
6821
	sched_group_nodes = kcalloc(MAX_NUMNODES, sizeof(struct sched_group *),
I
Ingo Molnar 已提交
6822
				    GFP_KERNEL);
6823 6824
	if (!sched_group_nodes) {
		printk(KERN_WARNING "Can not alloc sched group node list\n");
6825
		return -ENOMEM;
6826 6827
	}
#endif
L
Linus Torvalds 已提交
6828

6829
	rd = alloc_rootdomain();
G
Gregory Haskins 已提交
6830 6831
	if (!rd) {
		printk(KERN_WARNING "Cannot alloc root domain\n");
6832 6833 6834
#ifdef CONFIG_NUMA
		kfree(sched_group_nodes);
#endif
G
Gregory Haskins 已提交
6835 6836 6837
		return -ENOMEM;
	}

6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856
#if SCHED_CPUMASK_ALLOC
	/* get space for all scratch cpumask variables */
	allmasks = kmalloc(sizeof(*allmasks), GFP_KERNEL);
	if (!allmasks) {
		printk(KERN_WARNING "Cannot alloc cpumask array\n");
		kfree(rd);
#ifdef CONFIG_NUMA
		kfree(sched_group_nodes);
#endif
		return -ENOMEM;
	}
#endif
	tmpmask = (cpumask_t *)allmasks;


#ifdef CONFIG_NUMA
	sched_group_nodes_bycpu[first_cpu(*cpu_map)] = sched_group_nodes;
#endif

L
Linus Torvalds 已提交
6857
	/*
6858
	 * Set up domains for cpus specified by the cpu_map.
L
Linus Torvalds 已提交
6859
	 */
6860
	for_each_cpu_mask(i, *cpu_map) {
L
Linus Torvalds 已提交
6861
		struct sched_domain *sd = NULL, *p;
6862
		SCHED_CPUMASK_VAR(nodemask, allmasks);
L
Linus Torvalds 已提交
6863

6864 6865
		*nodemask = node_to_cpumask(cpu_to_node(i));
		cpus_and(*nodemask, *nodemask, *cpu_map);
L
Linus Torvalds 已提交
6866 6867

#ifdef CONFIG_NUMA
I
Ingo Molnar 已提交
6868
		if (cpus_weight(*cpu_map) >
6869
				SD_NODES_PER_DOMAIN*cpus_weight(*nodemask)) {
6870
			sd = &per_cpu(allnodes_domains, i);
6871
			SD_INIT(sd, ALLNODES);
6872
			sd->span = *cpu_map;
6873
			cpu_to_allnodes_group(i, cpu_map, &sd->groups, tmpmask);
6874
			p = sd;
6875
			sd_allnodes = 1;
6876 6877 6878
		} else
			p = NULL;

L
Linus Torvalds 已提交
6879
		sd = &per_cpu(node_domains, i);
6880
		SD_INIT(sd, NODE);
6881
		sched_domain_node_span(cpu_to_node(i), &sd->span);
6882
		sd->parent = p;
6883 6884
		if (p)
			p->child = sd;
6885
		cpus_and(sd->span, sd->span, *cpu_map);
L
Linus Torvalds 已提交
6886 6887 6888 6889
#endif

		p = sd;
		sd = &per_cpu(phys_domains, i);
6890 6891
		SD_INIT(sd, CPU);
		sd->span = *nodemask;
L
Linus Torvalds 已提交
6892
		sd->parent = p;
6893 6894
		if (p)
			p->child = sd;
6895
		cpu_to_phys_group(i, cpu_map, &sd->groups, tmpmask);
L
Linus Torvalds 已提交
6896

6897 6898 6899
#ifdef CONFIG_SCHED_MC
		p = sd;
		sd = &per_cpu(core_domains, i);
6900
		SD_INIT(sd, MC);
6901 6902 6903
		sd->span = cpu_coregroup_map(i);
		cpus_and(sd->span, sd->span, *cpu_map);
		sd->parent = p;
6904
		p->child = sd;
6905
		cpu_to_core_group(i, cpu_map, &sd->groups, tmpmask);
6906 6907
#endif

L
Linus Torvalds 已提交
6908 6909 6910
#ifdef CONFIG_SCHED_SMT
		p = sd;
		sd = &per_cpu(cpu_domains, i);
6911
		SD_INIT(sd, SIBLING);
6912
		sd->span = per_cpu(cpu_sibling_map, i);
6913
		cpus_and(sd->span, sd->span, *cpu_map);
L
Linus Torvalds 已提交
6914
		sd->parent = p;
6915
		p->child = sd;
6916
		cpu_to_cpu_group(i, cpu_map, &sd->groups, tmpmask);
L
Linus Torvalds 已提交
6917 6918 6919 6920 6921
#endif
	}

#ifdef CONFIG_SCHED_SMT
	/* Set up CPU (sibling) groups */
6922
	for_each_cpu_mask(i, *cpu_map) {
6923 6924 6925 6926 6927 6928
		SCHED_CPUMASK_VAR(this_sibling_map, allmasks);
		SCHED_CPUMASK_VAR(send_covered, allmasks);

		*this_sibling_map = per_cpu(cpu_sibling_map, i);
		cpus_and(*this_sibling_map, *this_sibling_map, *cpu_map);
		if (i != first_cpu(*this_sibling_map))
L
Linus Torvalds 已提交
6929 6930
			continue;

I
Ingo Molnar 已提交
6931
		init_sched_build_groups(this_sibling_map, cpu_map,
6932 6933
					&cpu_to_cpu_group,
					send_covered, tmpmask);
L
Linus Torvalds 已提交
6934 6935 6936
	}
#endif

6937 6938 6939
#ifdef CONFIG_SCHED_MC
	/* Set up multi-core groups */
	for_each_cpu_mask(i, *cpu_map) {
6940 6941 6942 6943 6944 6945
		SCHED_CPUMASK_VAR(this_core_map, allmasks);
		SCHED_CPUMASK_VAR(send_covered, allmasks);

		*this_core_map = cpu_coregroup_map(i);
		cpus_and(*this_core_map, *this_core_map, *cpu_map);
		if (i != first_cpu(*this_core_map))
6946
			continue;
6947

I
Ingo Molnar 已提交
6948
		init_sched_build_groups(this_core_map, cpu_map,
6949 6950
					&cpu_to_core_group,
					send_covered, tmpmask);
6951 6952 6953
	}
#endif

L
Linus Torvalds 已提交
6954 6955
	/* Set up physical groups */
	for (i = 0; i < MAX_NUMNODES; i++) {
6956 6957
		SCHED_CPUMASK_VAR(nodemask, allmasks);
		SCHED_CPUMASK_VAR(send_covered, allmasks);
L
Linus Torvalds 已提交
6958

6959 6960 6961
		*nodemask = node_to_cpumask(i);
		cpus_and(*nodemask, *nodemask, *cpu_map);
		if (cpus_empty(*nodemask))
L
Linus Torvalds 已提交
6962 6963
			continue;

6964 6965 6966
		init_sched_build_groups(nodemask, cpu_map,
					&cpu_to_phys_group,
					send_covered, tmpmask);
L
Linus Torvalds 已提交
6967 6968 6969 6970
	}

#ifdef CONFIG_NUMA
	/* Set up node groups */
6971 6972 6973 6974 6975 6976 6977
	if (sd_allnodes) {
		SCHED_CPUMASK_VAR(send_covered, allmasks);

		init_sched_build_groups(cpu_map, cpu_map,
					&cpu_to_allnodes_group,
					send_covered, tmpmask);
	}
6978 6979 6980 6981

	for (i = 0; i < MAX_NUMNODES; i++) {
		/* Set up node groups */
		struct sched_group *sg, *prev;
6982 6983 6984
		SCHED_CPUMASK_VAR(nodemask, allmasks);
		SCHED_CPUMASK_VAR(domainspan, allmasks);
		SCHED_CPUMASK_VAR(covered, allmasks);
6985 6986
		int j;

6987 6988 6989 6990 6991
		*nodemask = node_to_cpumask(i);
		cpus_clear(*covered);

		cpus_and(*nodemask, *nodemask, *cpu_map);
		if (cpus_empty(*nodemask)) {
6992
			sched_group_nodes[i] = NULL;
6993
			continue;
6994
		}
6995

6996
		sched_domain_node_span(i, domainspan);
6997
		cpus_and(*domainspan, *domainspan, *cpu_map);
6998

6999
		sg = kmalloc_node(sizeof(struct sched_group), GFP_KERNEL, i);
7000 7001 7002 7003 7004
		if (!sg) {
			printk(KERN_WARNING "Can not alloc domain group for "
				"node %d\n", i);
			goto error;
		}
7005
		sched_group_nodes[i] = sg;
7006
		for_each_cpu_mask(j, *nodemask) {
7007
			struct sched_domain *sd;
I
Ingo Molnar 已提交
7008

7009 7010 7011
			sd = &per_cpu(node_domains, j);
			sd->groups = sg;
		}
7012
		sg->__cpu_power = 0;
7013
		sg->cpumask = *nodemask;
7014
		sg->next = sg;
7015
		cpus_or(*covered, *covered, *nodemask);
7016 7017 7018
		prev = sg;

		for (j = 0; j < MAX_NUMNODES; j++) {
7019
			SCHED_CPUMASK_VAR(notcovered, allmasks);
7020
			int n = (i + j) % MAX_NUMNODES;
7021
			node_to_cpumask_ptr(pnodemask, n);
7022

7023 7024 7025 7026
			cpus_complement(*notcovered, *covered);
			cpus_and(*tmpmask, *notcovered, *cpu_map);
			cpus_and(*tmpmask, *tmpmask, *domainspan);
			if (cpus_empty(*tmpmask))
7027 7028
				break;

7029 7030
			cpus_and(*tmpmask, *tmpmask, *pnodemask);
			if (cpus_empty(*tmpmask))
7031 7032
				continue;

7033 7034
			sg = kmalloc_node(sizeof(struct sched_group),
					  GFP_KERNEL, i);
7035 7036 7037
			if (!sg) {
				printk(KERN_WARNING
				"Can not alloc domain group for node %d\n", j);
7038
				goto error;
7039
			}
7040
			sg->__cpu_power = 0;
7041
			sg->cpumask = *tmpmask;
7042
			sg->next = prev->next;
7043
			cpus_or(*covered, *covered, *tmpmask);
7044 7045 7046 7047
			prev->next = sg;
			prev = sg;
		}
	}
L
Linus Torvalds 已提交
7048 7049 7050
#endif

	/* Calculate CPU power for physical packages and nodes */
7051
#ifdef CONFIG_SCHED_SMT
7052
	for_each_cpu_mask(i, *cpu_map) {
I
Ingo Molnar 已提交
7053 7054
		struct sched_domain *sd = &per_cpu(cpu_domains, i);

7055
		init_sched_groups_power(i, sd);
7056
	}
L
Linus Torvalds 已提交
7057
#endif
7058
#ifdef CONFIG_SCHED_MC
7059
	for_each_cpu_mask(i, *cpu_map) {
I
Ingo Molnar 已提交
7060 7061
		struct sched_domain *sd = &per_cpu(core_domains, i);

7062
		init_sched_groups_power(i, sd);
7063 7064
	}
#endif
7065

7066
	for_each_cpu_mask(i, *cpu_map) {
I
Ingo Molnar 已提交
7067 7068
		struct sched_domain *sd = &per_cpu(phys_domains, i);

7069
		init_sched_groups_power(i, sd);
L
Linus Torvalds 已提交
7070 7071
	}

7072
#ifdef CONFIG_NUMA
7073 7074
	for (i = 0; i < MAX_NUMNODES; i++)
		init_numa_sched_groups_power(sched_group_nodes[i]);
7075

7076 7077
	if (sd_allnodes) {
		struct sched_group *sg;
7078

7079 7080
		cpu_to_allnodes_group(first_cpu(*cpu_map), cpu_map, &sg,
								tmpmask);
7081 7082
		init_numa_sched_groups_power(sg);
	}
7083 7084
#endif

L
Linus Torvalds 已提交
7085
	/* Attach the domains */
7086
	for_each_cpu_mask(i, *cpu_map) {
L
Linus Torvalds 已提交
7087 7088 7089
		struct sched_domain *sd;
#ifdef CONFIG_SCHED_SMT
		sd = &per_cpu(cpu_domains, i);
7090 7091
#elif defined(CONFIG_SCHED_MC)
		sd = &per_cpu(core_domains, i);
L
Linus Torvalds 已提交
7092 7093 7094
#else
		sd = &per_cpu(phys_domains, i);
#endif
G
Gregory Haskins 已提交
7095
		cpu_attach_domain(sd, rd, i);
L
Linus Torvalds 已提交
7096
	}
7097

7098
	SCHED_CPUMASK_FREE((void *)allmasks);
7099 7100
	return 0;

7101
#ifdef CONFIG_NUMA
7102
error:
7103 7104
	free_sched_groups(cpu_map, tmpmask);
	SCHED_CPUMASK_FREE((void *)allmasks);
7105
	return -ENOMEM;
7106
#endif
L
Linus Torvalds 已提交
7107
}
P
Paul Jackson 已提交
7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118

static cpumask_t *doms_cur;	/* current sched domains */
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
 * cpumask_t) fails, then fallback to a single sched domain,
 * as determined by the single cpumask_t fallback_doms.
 */
static cpumask_t fallback_doms;

7119 7120 7121 7122
void __attribute__((weak)) arch_update_cpu_topology(void)
{
}

7123
/*
I
Ingo Molnar 已提交
7124
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
7125 7126
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
7127
 */
7128
static int arch_init_sched_domains(const cpumask_t *cpu_map)
7129
{
7130 7131
	int err;

7132
	arch_update_cpu_topology();
P
Paul Jackson 已提交
7133 7134 7135 7136 7137
	ndoms_cur = 1;
	doms_cur = kmalloc(sizeof(cpumask_t), GFP_KERNEL);
	if (!doms_cur)
		doms_cur = &fallback_doms;
	cpus_andnot(*doms_cur, *cpu_map, cpu_isolated_map);
7138
	err = build_sched_domains(doms_cur);
7139
	register_sched_domain_sysctl();
7140 7141

	return err;
7142 7143
}

7144 7145
static void arch_destroy_sched_domains(const cpumask_t *cpu_map,
				       cpumask_t *tmpmask)
L
Linus Torvalds 已提交
7146
{
7147
	free_sched_groups(cpu_map, tmpmask);
7148
}
L
Linus Torvalds 已提交
7149

7150 7151 7152 7153
/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
7154
static void detach_destroy_domains(const cpumask_t *cpu_map)
7155
{
7156
	cpumask_t tmpmask;
7157 7158
	int i;

7159 7160
	unregister_sched_domain_sysctl();

7161
	for_each_cpu_mask(i, *cpu_map)
G
Gregory Haskins 已提交
7162
		cpu_attach_domain(NULL, &def_root_domain, i);
7163
	synchronize_sched();
7164
	arch_destroy_sched_domains(cpu_map, &tmpmask);
7165 7166
}

P
Paul Jackson 已提交
7167 7168
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
7169
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
7170 7171 7172 7173
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
 * 'doms_new' is an array of cpumask_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
7174 7175 7176
 * 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
7177 7178 7179
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
I
Ingo Molnar 已提交
7180 7181
 * The passed in 'doms_new' should be kmalloc'd. This routine takes
 * ownership of it and will kfree it when done with it. If the caller
P
Paul Jackson 已提交
7182 7183 7184 7185 7186 7187 7188 7189 7190 7191
 * failed the kmalloc call, then it can pass in doms_new == NULL,
 * and partition_sched_domains() will fallback to the single partition
 * 'fallback_doms'.
 *
 * Call with hotplug lock held
 */
void partition_sched_domains(int ndoms_new, cpumask_t *doms_new)
{
	int i, j;

7192 7193
	lock_doms_cur();

7194 7195 7196
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

P
Paul Jackson 已提交
7197 7198 7199 7200 7201 7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220 7221 7222 7223 7224 7225 7226 7227 7228 7229 7230 7231
	if (doms_new == NULL) {
		ndoms_new = 1;
		doms_new = &fallback_doms;
		cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map);
	}

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
		for (j = 0; j < ndoms_new; j++) {
			if (cpus_equal(doms_cur[i], doms_new[j]))
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
		detach_destroy_domains(doms_cur + i);
match1:
		;
	}

	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
		for (j = 0; j < ndoms_cur; j++) {
			if (cpus_equal(doms_new[i], doms_cur[j]))
				goto match2;
		}
		/* no match - add a new doms_new */
		build_sched_domains(doms_new + i);
match2:
		;
	}

	/* Remember the new sched domains */
	if (doms_cur != &fallback_doms)
		kfree(doms_cur);
	doms_cur = doms_new;
	ndoms_cur = ndoms_new;
7232 7233

	register_sched_domain_sysctl();
7234 7235

	unlock_doms_cur();
P
Paul Jackson 已提交
7236 7237
}

7238
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
7239
int arch_reinit_sched_domains(void)
7240 7241 7242
{
	int err;

7243
	get_online_cpus();
7244 7245
	detach_destroy_domains(&cpu_online_map);
	err = arch_init_sched_domains(&cpu_online_map);
7246
	put_online_cpus();
7247 7248 7249 7250 7251 7252 7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270 7271 7272

	return err;
}

static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
{
	int ret;

	if (buf[0] != '0' && buf[0] != '1')
		return -EINVAL;

	if (smt)
		sched_smt_power_savings = (buf[0] == '1');
	else
		sched_mc_power_savings = (buf[0] == '1');

	ret = arch_reinit_sched_domains();

	return ret ? ret : count;
}

#ifdef CONFIG_SCHED_MC
static ssize_t sched_mc_power_savings_show(struct sys_device *dev, char *page)
{
	return sprintf(page, "%u\n", sched_mc_power_savings);
}
7273 7274
static ssize_t sched_mc_power_savings_store(struct sys_device *dev,
					    const char *buf, size_t count)
7275 7276 7277
{
	return sched_power_savings_store(buf, count, 0);
}
A
Adrian Bunk 已提交
7278 7279
static SYSDEV_ATTR(sched_mc_power_savings, 0644, sched_mc_power_savings_show,
		   sched_mc_power_savings_store);
7280 7281 7282 7283 7284 7285 7286
#endif

#ifdef CONFIG_SCHED_SMT
static ssize_t sched_smt_power_savings_show(struct sys_device *dev, char *page)
{
	return sprintf(page, "%u\n", sched_smt_power_savings);
}
7287 7288
static ssize_t sched_smt_power_savings_store(struct sys_device *dev,
					     const char *buf, size_t count)
7289 7290 7291
{
	return sched_power_savings_store(buf, count, 1);
}
A
Adrian Bunk 已提交
7292 7293 7294 7295 7296 7297 7298 7299 7300 7301 7302 7303 7304 7305 7306 7307 7308 7309 7310 7311
static SYSDEV_ATTR(sched_smt_power_savings, 0644, sched_smt_power_savings_show,
		   sched_smt_power_savings_store);
#endif

int sched_create_sysfs_power_savings_entries(struct sysdev_class *cls)
{
	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;
}
7312 7313
#endif

L
Linus Torvalds 已提交
7314
/*
I
Ingo Molnar 已提交
7315
 * Force a reinitialization of the sched domains hierarchy. The domains
L
Linus Torvalds 已提交
7316
 * and groups cannot be updated in place without racing with the balancing
N
Nick Piggin 已提交
7317
 * code, so we temporarily attach all running cpus to the NULL domain
L
Linus Torvalds 已提交
7318 7319 7320 7321 7322 7323 7324
 * which will prevent rebalancing while the sched domains are recalculated.
 */
static int update_sched_domains(struct notifier_block *nfb,
				unsigned long action, void *hcpu)
{
	switch (action) {
	case CPU_UP_PREPARE:
7325
	case CPU_UP_PREPARE_FROZEN:
L
Linus Torvalds 已提交
7326
	case CPU_DOWN_PREPARE:
7327
	case CPU_DOWN_PREPARE_FROZEN:
7328
		detach_destroy_domains(&cpu_online_map);
L
Linus Torvalds 已提交
7329 7330 7331
		return NOTIFY_OK;

	case CPU_UP_CANCELED:
7332
	case CPU_UP_CANCELED_FROZEN:
L
Linus Torvalds 已提交
7333
	case CPU_DOWN_FAILED:
7334
	case CPU_DOWN_FAILED_FROZEN:
L
Linus Torvalds 已提交
7335
	case CPU_ONLINE:
7336
	case CPU_ONLINE_FROZEN:
L
Linus Torvalds 已提交
7337
	case CPU_DEAD:
7338
	case CPU_DEAD_FROZEN:
L
Linus Torvalds 已提交
7339 7340 7341 7342 7343 7344 7345 7346 7347
		/*
		 * Fall through and re-initialise the domains.
		 */
		break;
	default:
		return NOTIFY_DONE;
	}

	/* The hotplug lock is already held by cpu_up/cpu_down */
7348
	arch_init_sched_domains(&cpu_online_map);
L
Linus Torvalds 已提交
7349 7350 7351 7352 7353 7354

	return NOTIFY_OK;
}

void __init sched_init_smp(void)
{
7355 7356
	cpumask_t non_isolated_cpus;

7357 7358 7359 7360 7361
#if defined(CONFIG_NUMA)
	sched_group_nodes_bycpu = kzalloc(nr_cpu_ids * sizeof(void **),
								GFP_KERNEL);
	BUG_ON(sched_group_nodes_bycpu == NULL);
#endif
7362
	get_online_cpus();
7363
	arch_init_sched_domains(&cpu_online_map);
7364
	cpus_andnot(non_isolated_cpus, cpu_possible_map, cpu_isolated_map);
7365 7366
	if (cpus_empty(non_isolated_cpus))
		cpu_set(smp_processor_id(), non_isolated_cpus);
7367
	put_online_cpus();
L
Linus Torvalds 已提交
7368 7369
	/* XXX: Theoretical race here - CPU may be hotplugged now */
	hotcpu_notifier(update_sched_domains, 0);
7370 7371

	/* Move init over to a non-isolated CPU */
7372
	if (set_cpus_allowed_ptr(current, &non_isolated_cpus) < 0)
7373
		BUG();
I
Ingo Molnar 已提交
7374
	sched_init_granularity();
L
Linus Torvalds 已提交
7375 7376 7377 7378
}
#else
void __init sched_init_smp(void)
{
7379 7380 7381 7382 7383
#if defined(CONFIG_NUMA)
	sched_group_nodes_bycpu = kzalloc(nr_cpu_ids * sizeof(void **),
								GFP_KERNEL);
	BUG_ON(sched_group_nodes_bycpu == NULL);
#endif
I
Ingo Molnar 已提交
7384
	sched_init_granularity();
L
Linus Torvalds 已提交
7385 7386 7387 7388 7389 7390 7391 7392 7393 7394
}
#endif /* CONFIG_SMP */

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
7395
static void init_cfs_rq(struct cfs_rq *cfs_rq, struct rq *rq)
I
Ingo Molnar 已提交
7396 7397 7398 7399 7400
{
	cfs_rq->tasks_timeline = RB_ROOT;
#ifdef CONFIG_FAIR_GROUP_SCHED
	cfs_rq->rq = rq;
#endif
P
Peter Zijlstra 已提交
7401
	cfs_rq->min_vruntime = (u64)(-(1LL << 20));
I
Ingo Molnar 已提交
7402 7403
}

P
Peter Zijlstra 已提交
7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416
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);

7417
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7418 7419
	rt_rq->highest_prio = MAX_RT_PRIO;
#endif
P
Peter Zijlstra 已提交
7420 7421 7422 7423 7424 7425 7426
#ifdef CONFIG_SMP
	rt_rq->rt_nr_migratory = 0;
	rt_rq->overloaded = 0;
#endif

	rt_rq->rt_time = 0;
	rt_rq->rt_throttled = 0;
P
Peter Zijlstra 已提交
7427 7428
	rt_rq->rt_runtime = 0;
	spin_lock_init(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7429

7430
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7431
	rt_rq->rt_nr_boosted = 0;
P
Peter Zijlstra 已提交
7432 7433
	rt_rq->rq = rq;
#endif
P
Peter Zijlstra 已提交
7434 7435
}

P
Peter Zijlstra 已提交
7436 7437 7438 7439 7440 7441 7442 7443 7444 7445 7446 7447 7448 7449 7450 7451 7452 7453
#ifdef CONFIG_FAIR_GROUP_SCHED
static void init_tg_cfs_entry(struct rq *rq, struct task_group *tg,
		struct cfs_rq *cfs_rq, struct sched_entity *se,
		int cpu, int add)
{
	tg->cfs_rq[cpu] = cfs_rq;
	init_cfs_rq(cfs_rq, rq);
	cfs_rq->tg = tg;
	if (add)
		list_add(&cfs_rq->leaf_cfs_rq_list, &rq->leaf_cfs_rq_list);

	tg->se[cpu] = se;
	se->cfs_rq = &rq->cfs;
	se->my_q = cfs_rq;
	se->load.weight = tg->shares;
	se->load.inv_weight = div64_64(1ULL<<32, se->load.weight);
	se->parent = NULL;
}
7454
#endif
P
Peter Zijlstra 已提交
7455

7456
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7457 7458 7459 7460 7461 7462 7463 7464
static void init_tg_rt_entry(struct rq *rq, struct task_group *tg,
		struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
		int cpu, int add)
{
	tg->rt_rq[cpu] = rt_rq;
	init_rt_rq(rt_rq, rq);
	rt_rq->tg = tg;
	rt_rq->rt_se = rt_se;
P
Peter Zijlstra 已提交
7465
	rt_rq->rt_runtime = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7466 7467 7468 7469 7470 7471 7472 7473 7474 7475 7476
	if (add)
		list_add(&rt_rq->leaf_rt_rq_list, &rq->leaf_rt_rq_list);

	tg->rt_se[cpu] = rt_se;
	rt_se->rt_rq = &rq->rt;
	rt_se->my_q = rt_rq;
	rt_se->parent = NULL;
	INIT_LIST_HEAD(&rt_se->run_list);
}
#endif

L
Linus Torvalds 已提交
7477 7478
void __init sched_init(void)
{
I
Ingo Molnar 已提交
7479
	int i, j;
7480 7481 7482 7483 7484 7485 7486 7487 7488 7489 7490 7491 7492 7493 7494 7495 7496 7497 7498 7499 7500 7501 7502 7503 7504 7505 7506 7507 7508
	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 **);
#endif
	/*
	 * As sched_init() is called before page_alloc is setup,
	 * we use alloc_bootmem().
	 */
	if (alloc_size) {
		ptr = (unsigned long)alloc_bootmem_low(alloc_size);

#ifdef CONFIG_FAIR_GROUP_SCHED
		init_task_group.se = (struct sched_entity **)ptr;
		ptr += nr_cpu_ids * sizeof(void **);

		init_task_group.cfs_rq = (struct cfs_rq **)ptr;
		ptr += nr_cpu_ids * sizeof(void **);
#endif
#ifdef CONFIG_RT_GROUP_SCHED
		init_task_group.rt_se = (struct sched_rt_entity **)ptr;
		ptr += nr_cpu_ids * sizeof(void **);

		init_task_group.rt_rq = (struct rt_rq **)ptr;
#endif
	}
I
Ingo Molnar 已提交
7509

G
Gregory Haskins 已提交
7510 7511 7512 7513
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

7514 7515 7516 7517 7518 7519 7520 7521
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
	init_rt_bandwidth(&init_task_group.rt_bandwidth,
			global_rt_period(), global_rt_runtime());
#endif

7522
#ifdef CONFIG_GROUP_SCHED
P
Peter Zijlstra 已提交
7523 7524 7525
	list_add(&init_task_group.list, &task_groups);
#endif

7526
	for_each_possible_cpu(i) {
7527
		struct rq *rq;
L
Linus Torvalds 已提交
7528 7529 7530

		rq = cpu_rq(i);
		spin_lock_init(&rq->lock);
7531
		lockdep_set_class(&rq->lock, &rq->rq_lock_key);
N
Nick Piggin 已提交
7532
		rq->nr_running = 0;
I
Ingo Molnar 已提交
7533
		rq->clock = 1;
7534
		update_last_tick_seen(rq);
I
Ingo Molnar 已提交
7535
		init_cfs_rq(&rq->cfs, rq);
P
Peter Zijlstra 已提交
7536
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
7537
#ifdef CONFIG_FAIR_GROUP_SCHED
7538
		init_task_group.shares = init_task_group_load;
P
Peter Zijlstra 已提交
7539 7540 7541 7542 7543
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
		init_tg_cfs_entry(rq, &init_task_group,
				&per_cpu(init_cfs_rq, i),
				&per_cpu(init_sched_entity, i), i, 1);

7544 7545
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7546 7547 7548 7549
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
		init_tg_rt_entry(rq, &init_task_group,
				&per_cpu(init_rt_rq, i),
				&per_cpu(init_sched_rt_entity, i), i, 1);
P
Peter Zijlstra 已提交
7550 7551
#else
		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
I
Ingo Molnar 已提交
7552
#endif
L
Linus Torvalds 已提交
7553

I
Ingo Molnar 已提交
7554 7555
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
L
Linus Torvalds 已提交
7556
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
7557
		rq->sd = NULL;
G
Gregory Haskins 已提交
7558
		rq->rd = NULL;
L
Linus Torvalds 已提交
7559
		rq->active_balance = 0;
I
Ingo Molnar 已提交
7560
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
7561
		rq->push_cpu = 0;
7562
		rq->cpu = i;
L
Linus Torvalds 已提交
7563 7564
		rq->migration_thread = NULL;
		INIT_LIST_HEAD(&rq->migration_queue);
7565
		rq_attach_root(rq, &def_root_domain);
L
Linus Torvalds 已提交
7566
#endif
P
Peter Zijlstra 已提交
7567
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
7568 7569 7570
		atomic_set(&rq->nr_iowait, 0);
	}

7571
	set_load_weight(&init_task);
7572

7573 7574 7575 7576
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

7577 7578 7579 7580
#ifdef CONFIG_SMP
	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains, NULL);
#endif

7581 7582 7583 7584
#ifdef CONFIG_RT_MUTEXES
	plist_head_init(&init_task.pi_waiters, &init_task.pi_lock);
#endif

L
Linus Torvalds 已提交
7585 7586 7587 7588 7589 7590 7591 7592 7593 7594 7595 7596 7597
	/*
	 * 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());
I
Ingo Molnar 已提交
7598 7599 7600 7601
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
7602 7603

	scheduler_running = 1;
L
Linus Torvalds 已提交
7604 7605 7606 7607 7608
}

#ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
void __might_sleep(char *file, int line)
{
7609
#ifdef in_atomic
L
Linus Torvalds 已提交
7610 7611 7612 7613 7614 7615 7616
	static unsigned long prev_jiffy;	/* ratelimiting */

	if ((in_atomic() || irqs_disabled()) &&
	    system_state == SYSTEM_RUNNING && !oops_in_progress) {
		if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
			return;
		prev_jiffy = jiffies;
7617
		printk(KERN_ERR "BUG: sleeping function called from invalid"
L
Linus Torvalds 已提交
7618 7619 7620
				" context at %s:%d\n", file, line);
		printk("in_atomic():%d, irqs_disabled():%d\n",
			in_atomic(), irqs_disabled());
7621
		debug_show_held_locks(current);
7622 7623
		if (irqs_disabled())
			print_irqtrace_events(current);
L
Linus Torvalds 已提交
7624 7625 7626 7627 7628 7629 7630 7631
		dump_stack();
	}
#endif
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642 7643 7644 7645
static void normalize_task(struct rq *rq, struct task_struct *p)
{
	int on_rq;
	update_rq_clock(rq);
	on_rq = p->se.on_rq;
	if (on_rq)
		deactivate_task(rq, p, 0);
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
		activate_task(rq, p, 0);
		resched_task(rq->curr);
	}
}

L
Linus Torvalds 已提交
7646 7647
void normalize_rt_tasks(void)
{
7648
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
7649
	unsigned long flags;
7650
	struct rq *rq;
L
Linus Torvalds 已提交
7651

7652
	read_lock_irqsave(&tasklist_lock, flags);
7653
	do_each_thread(g, p) {
7654 7655 7656 7657 7658 7659
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
7660 7661
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
I
Ingo Molnar 已提交
7662 7663 7664
		p->se.wait_start		= 0;
		p->se.sleep_start		= 0;
		p->se.block_start		= 0;
I
Ingo Molnar 已提交
7665
#endif
I
Ingo Molnar 已提交
7666 7667 7668 7669 7670 7671 7672 7673 7674
		task_rq(p)->clock		= 0;

		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
7675
			continue;
I
Ingo Molnar 已提交
7676
		}
L
Linus Torvalds 已提交
7677

7678
		spin_lock(&p->pi_lock);
7679
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
7680

7681
		normalize_task(rq, p);
7682

7683
		__task_rq_unlock(rq);
7684
		spin_unlock(&p->pi_lock);
7685 7686
	} while_each_thread(g, p);

7687
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
7688 7689 7690
}

#endif /* CONFIG_MAGIC_SYSRQ */
7691 7692 7693 7694 7695 7696 7697 7698 7699 7700 7701 7702 7703 7704 7705 7706 7707 7708

#ifdef CONFIG_IA64
/*
 * These functions are only useful for the IA64 MCA handling.
 *
 * 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!
 */
7709
struct task_struct *curr_task(int cpu)
7710 7711 7712 7713 7714 7715 7716 7717 7718 7719
{
	return cpu_curr(cpu);
}

/**
 * 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
7720 7721
 * 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
7722 7723 7724 7725 7726 7727 7728
 * 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!
 */
7729
void set_curr_task(int cpu, struct task_struct *p)
7730 7731 7732 7733 7734
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
7735

7736 7737
#ifdef CONFIG_FAIR_GROUP_SCHED
static void free_fair_sched_group(struct task_group *tg)
P
Peter Zijlstra 已提交
7738 7739 7740 7741 7742 7743 7744 7745 7746 7747 7748 7749 7750 7751
{
	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);
}

7752
static int alloc_fair_sched_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7753 7754 7755
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se;
7756
	struct rq *rq;
S
Srivatsa Vaddagiri 已提交
7757 7758
	int i;

7759
	tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL);
S
Srivatsa Vaddagiri 已提交
7760 7761
	if (!tg->cfs_rq)
		goto err;
7762
	tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL);
S
Srivatsa Vaddagiri 已提交
7763 7764
	if (!tg->se)
		goto err;
7765 7766

	tg->shares = NICE_0_LOAD;
S
Srivatsa Vaddagiri 已提交
7767 7768

	for_each_possible_cpu(i) {
7769
		rq = cpu_rq(i);
S
Srivatsa Vaddagiri 已提交
7770

P
Peter Zijlstra 已提交
7771 7772
		cfs_rq = kmalloc_node(sizeof(struct cfs_rq),
				GFP_KERNEL|__GFP_ZERO, cpu_to_node(i));
S
Srivatsa Vaddagiri 已提交
7773 7774 7775
		if (!cfs_rq)
			goto err;

P
Peter Zijlstra 已提交
7776 7777
		se = kmalloc_node(sizeof(struct sched_entity),
				GFP_KERNEL|__GFP_ZERO, cpu_to_node(i));
S
Srivatsa Vaddagiri 已提交
7778 7779 7780
		if (!se)
			goto err;

7781
		init_tg_cfs_entry(rq, tg, cfs_rq, se, i, 0);
7782 7783 7784 7785 7786 7787 7788 7789 7790 7791 7792 7793 7794 7795 7796 7797 7798 7799 7800 7801 7802 7803 7804 7805 7806 7807 7808 7809 7810 7811 7812 7813 7814 7815 7816
	}

	return 1;

 err:
	return 0;
}

static inline void register_fair_sched_group(struct task_group *tg, int cpu)
{
	list_add_rcu(&tg->cfs_rq[cpu]->leaf_cfs_rq_list,
			&cpu_rq(cpu)->leaf_cfs_rq_list);
}

static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
	list_del_rcu(&tg->cfs_rq[cpu]->leaf_cfs_rq_list);
}
#else
static inline void free_fair_sched_group(struct task_group *tg)
{
}

static inline int alloc_fair_sched_group(struct task_group *tg)
{
	return 1;
}

static inline void register_fair_sched_group(struct task_group *tg, int cpu)
{
}

static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
}
7817 7818 7819
#endif

#ifdef CONFIG_RT_GROUP_SCHED
7820 7821 7822 7823
static void free_rt_sched_group(struct task_group *tg)
{
	int i;

7824 7825
	destroy_rt_bandwidth(&tg->rt_bandwidth);

7826 7827 7828 7829 7830 7831 7832 7833 7834 7835 7836 7837 7838 7839 7840 7841 7842 7843
	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);
}

static int alloc_rt_sched_group(struct task_group *tg)
{
	struct rt_rq *rt_rq;
	struct sched_rt_entity *rt_se;
	struct rq *rq;
	int i;

7844
	tg->rt_rq = kzalloc(sizeof(rt_rq) * nr_cpu_ids, GFP_KERNEL);
7845 7846
	if (!tg->rt_rq)
		goto err;
7847
	tg->rt_se = kzalloc(sizeof(rt_se) * nr_cpu_ids, GFP_KERNEL);
7848 7849 7850
	if (!tg->rt_se)
		goto err;

7851 7852
	init_rt_bandwidth(&tg->rt_bandwidth,
			ktime_to_ns(def_rt_bandwidth.rt_period), 0);
7853 7854 7855 7856

	for_each_possible_cpu(i) {
		rq = cpu_rq(i);

P
Peter Zijlstra 已提交
7857 7858 7859 7860
		rt_rq = kmalloc_node(sizeof(struct rt_rq),
				GFP_KERNEL|__GFP_ZERO, cpu_to_node(i));
		if (!rt_rq)
			goto err;
S
Srivatsa Vaddagiri 已提交
7861

P
Peter Zijlstra 已提交
7862 7863 7864 7865
		rt_se = kmalloc_node(sizeof(struct sched_rt_entity),
				GFP_KERNEL|__GFP_ZERO, cpu_to_node(i));
		if (!rt_se)
			goto err;
S
Srivatsa Vaddagiri 已提交
7866

P
Peter Zijlstra 已提交
7867
		init_tg_rt_entry(rq, tg, rt_rq, rt_se, i, 0);
S
Srivatsa Vaddagiri 已提交
7868 7869
	}

7870 7871 7872 7873 7874 7875 7876 7877 7878 7879 7880 7881 7882 7883 7884 7885 7886 7887 7888 7889 7890 7891 7892 7893 7894 7895 7896 7897 7898 7899 7900 7901 7902 7903 7904
	return 1;

 err:
	return 0;
}

static inline void register_rt_sched_group(struct task_group *tg, int cpu)
{
	list_add_rcu(&tg->rt_rq[cpu]->leaf_rt_rq_list,
			&cpu_rq(cpu)->leaf_rt_rq_list);
}

static inline void unregister_rt_sched_group(struct task_group *tg, int cpu)
{
	list_del_rcu(&tg->rt_rq[cpu]->leaf_rt_rq_list);
}
#else
static inline void free_rt_sched_group(struct task_group *tg)
{
}

static inline int alloc_rt_sched_group(struct task_group *tg)
{
	return 1;
}

static inline void register_rt_sched_group(struct task_group *tg, int cpu)
{
}

static inline void unregister_rt_sched_group(struct task_group *tg, int cpu)
{
}
#endif

7905
#ifdef CONFIG_GROUP_SCHED
7906 7907 7908 7909 7910 7911 7912 7913 7914 7915 7916 7917 7918 7919 7920 7921 7922 7923 7924 7925 7926 7927 7928 7929
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
struct task_group *sched_create_group(void)
{
	struct task_group *tg;
	unsigned long flags;
	int i;

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

	if (!alloc_fair_sched_group(tg))
		goto err;

	if (!alloc_rt_sched_group(tg))
		goto err;

7930
	spin_lock_irqsave(&task_group_lock, flags);
7931
	for_each_possible_cpu(i) {
7932 7933
		register_fair_sched_group(tg, i);
		register_rt_sched_group(tg, i);
7934
	}
P
Peter Zijlstra 已提交
7935
	list_add_rcu(&tg->list, &task_groups);
7936
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7937

7938
	return tg;
S
Srivatsa Vaddagiri 已提交
7939 7940

err:
P
Peter Zijlstra 已提交
7941
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
7942 7943 7944
	return ERR_PTR(-ENOMEM);
}

7945
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
7946
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
7947 7948
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
7949
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
7950 7951
}

7952
/* Destroy runqueue etc associated with a task group */
7953
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7954
{
7955
	unsigned long flags;
7956
	int i;
S
Srivatsa Vaddagiri 已提交
7957

7958
	spin_lock_irqsave(&task_group_lock, flags);
7959
	for_each_possible_cpu(i) {
7960 7961
		unregister_fair_sched_group(tg, i);
		unregister_rt_sched_group(tg, i);
7962
	}
P
Peter Zijlstra 已提交
7963
	list_del_rcu(&tg->list);
7964
	spin_unlock_irqrestore(&task_group_lock, flags);
7965 7966

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
7967
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
7968 7969
}

7970
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
7971 7972 7973
 *	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.
7974 7975
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
7976 7977 7978 7979 7980 7981 7982 7983 7984
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

	update_rq_clock(rq);

7985
	running = task_current(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7986 7987
	on_rq = tsk->se.on_rq;

7988
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
7989
		dequeue_task(rq, tsk, 0);
7990 7991
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7992

P
Peter Zijlstra 已提交
7993
	set_task_rq(tsk, task_cpu(tsk));
S
Srivatsa Vaddagiri 已提交
7994

P
Peter Zijlstra 已提交
7995 7996 7997 7998 7999
#ifdef CONFIG_FAIR_GROUP_SCHED
	if (tsk->sched_class->moved_group)
		tsk->sched_class->moved_group(tsk);
#endif

8000 8001 8002
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
8003
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
8004 8005 8006

	task_rq_unlock(rq, &flags);
}
8007
#endif
S
Srivatsa Vaddagiri 已提交
8008

8009
#ifdef CONFIG_FAIR_GROUP_SCHED
S
Srivatsa Vaddagiri 已提交
8010 8011 8012 8013 8014 8015
static void set_se_shares(struct sched_entity *se, unsigned long shares)
{
	struct cfs_rq *cfs_rq = se->cfs_rq;
	struct rq *rq = cfs_rq->rq;
	int on_rq;

8016
	spin_lock_irq(&rq->lock);
S
Srivatsa Vaddagiri 已提交
8017 8018

	on_rq = se->on_rq;
8019
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
8020 8021 8022 8023 8024
		dequeue_entity(cfs_rq, se, 0);

	se->load.weight = shares;
	se->load.inv_weight = div64_64((1ULL<<32), shares);

8025
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
8026
		enqueue_entity(cfs_rq, se, 0);
8027 8028

	spin_unlock_irq(&rq->lock);
S
Srivatsa Vaddagiri 已提交
8029 8030
}

8031 8032
static DEFINE_MUTEX(shares_mutex);

8033
int sched_group_set_shares(struct task_group *tg, unsigned long shares)
S
Srivatsa Vaddagiri 已提交
8034 8035
{
	int i;
8036
	unsigned long flags;
8037

8038 8039 8040 8041 8042 8043 8044 8045
	/*
	 * A weight of 0 or 1 can cause arithmetics problems.
	 * (The default weight is 1024 - so there's no practical
	 *  limitation from this.)
	 */
	if (shares < 2)
		shares = 2;

8046
	mutex_lock(&shares_mutex);
8047
	if (tg->shares == shares)
8048
		goto done;
S
Srivatsa Vaddagiri 已提交
8049

8050
	spin_lock_irqsave(&task_group_lock, flags);
8051 8052
	for_each_possible_cpu(i)
		unregister_fair_sched_group(tg, i);
8053
	spin_unlock_irqrestore(&task_group_lock, flags);
8054 8055 8056 8057 8058 8059 8060 8061

	/* wait for any ongoing reference to this group to finish */
	synchronize_sched();

	/*
	 * Now we are free to modify the group's share on each cpu
	 * w/o tripping rebalance_share or load_balance_fair.
	 */
8062
	tg->shares = shares;
8063
	for_each_possible_cpu(i)
8064
		set_se_shares(tg->se[i], shares);
S
Srivatsa Vaddagiri 已提交
8065

8066 8067 8068 8069
	/*
	 * Enable load balance activity on this group, by inserting it back on
	 * each cpu's rq->leaf_cfs_rq_list.
	 */
8070
	spin_lock_irqsave(&task_group_lock, flags);
8071 8072
	for_each_possible_cpu(i)
		register_fair_sched_group(tg, i);
8073
	spin_unlock_irqrestore(&task_group_lock, flags);
8074
done:
8075
	mutex_unlock(&shares_mutex);
8076
	return 0;
S
Srivatsa Vaddagiri 已提交
8077 8078
}

8079 8080 8081 8082
unsigned long sched_group_shares(struct task_group *tg)
{
	return tg->shares;
}
8083
#endif
8084

8085
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8086
/*
P
Peter Zijlstra 已提交
8087
 * Ensure that the real time constraints are schedulable.
P
Peter Zijlstra 已提交
8088
 */
P
Peter Zijlstra 已提交
8089 8090 8091 8092 8093 8094 8095
static DEFINE_MUTEX(rt_constraints_mutex);

static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
		return 1ULL << 16;

8096
	return div64_64(runtime << 16, period);
P
Peter Zijlstra 已提交
8097 8098 8099
}

static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
P
Peter Zijlstra 已提交
8100 8101 8102
{
	struct task_group *tgi;
	unsigned long total = 0;
P
Peter Zijlstra 已提交
8103
	unsigned long global_ratio =
8104
		to_ratio(global_rt_period(), global_rt_runtime());
P
Peter Zijlstra 已提交
8105 8106

	rcu_read_lock();
P
Peter Zijlstra 已提交
8107 8108 8109
	list_for_each_entry_rcu(tgi, &task_groups, list) {
		if (tgi == tg)
			continue;
P
Peter Zijlstra 已提交
8110

8111 8112
		total += to_ratio(ktime_to_ns(tgi->rt_bandwidth.rt_period),
				tgi->rt_bandwidth.rt_runtime);
P
Peter Zijlstra 已提交
8113 8114
	}
	rcu_read_unlock();
P
Peter Zijlstra 已提交
8115

P
Peter Zijlstra 已提交
8116
	return total + to_ratio(period, runtime) < global_ratio;
P
Peter Zijlstra 已提交
8117 8118
}

8119 8120 8121 8122 8123 8124 8125 8126 8127 8128 8129
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
{
	struct task_struct *g, *p;
	do_each_thread(g, p) {
		if (rt_task(p) && rt_rq_of_se(&p->rt)->tg == tg)
			return 1;
	} while_each_thread(g, p);
	return 0;
}

8130 8131
static int tg_set_bandwidth(struct task_group *tg,
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
8132
{
P
Peter Zijlstra 已提交
8133
	int i, err = 0;
P
Peter Zijlstra 已提交
8134 8135

	mutex_lock(&rt_constraints_mutex);
8136
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
8137
	if (rt_runtime == 0 && tg_has_rt_tasks(tg)) {
8138 8139 8140
		err = -EBUSY;
		goto unlock;
	}
P
Peter Zijlstra 已提交
8141 8142 8143 8144
	if (!__rt_schedulable(tg, rt_period, rt_runtime)) {
		err = -EINVAL;
		goto unlock;
	}
P
Peter Zijlstra 已提交
8145 8146

	spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
8147 8148
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
8149 8150 8151 8152 8153 8154 8155 8156 8157

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

		spin_lock(&rt_rq->rt_runtime_lock);
		rt_rq->rt_runtime = rt_runtime;
		spin_unlock(&rt_rq->rt_runtime_lock);
	}
	spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
8158
 unlock:
8159
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
8160 8161 8162
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
8163 8164
}

8165 8166 8167 8168 8169 8170 8171 8172 8173 8174 8175 8176
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
8177 8178 8179 8180
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

8181
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
8182 8183
		return -1;

8184
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
8185 8186 8187
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
8188 8189 8190 8191 8192 8193 8194 8195 8196 8197 8198 8199 8200 8201 8202 8203 8204 8205 8206 8207 8208 8209 8210 8211 8212 8213 8214 8215 8216 8217 8218 8219 8220 8221

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;

	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)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
	if (!__rt_schedulable(NULL, 1, 0))
		ret = -EINVAL;
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
#else
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
8222 8223 8224 8225 8226 8227 8228 8229 8230 8231 8232 8233 8234
	unsigned long flags;
	int i;

	spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

		spin_lock(&rt_rq->rt_runtime_lock);
		rt_rq->rt_runtime = global_rt_runtime();
		spin_unlock(&rt_rq->rt_runtime_lock);
	}
	spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);

8235 8236
	return 0;
}
8237
#endif
8238 8239 8240 8241 8242 8243 8244 8245 8246 8247 8248 8249 8250 8251 8252 8253 8254 8255 8256 8257 8258 8259 8260 8261 8262 8263 8264 8265 8266 8267

int sched_rt_handler(struct ctl_table *table, int write,
		struct file *filp, void __user *buffer, size_t *lenp,
		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;

	ret = proc_dointvec(table, write, filp, buffer, lenp, ppos);

	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;
}
8268

8269
#ifdef CONFIG_CGROUP_SCHED
8270 8271

/* return corresponding task_group object of a cgroup */
8272
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
8273
{
8274 8275
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
8276 8277 8278
}

static struct cgroup_subsys_state *
8279
cpu_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
8280 8281 8282
{
	struct task_group *tg;

8283
	if (!cgrp->parent) {
8284
		/* This is early initialization for the top cgroup */
8285
		init_task_group.css.cgroup = cgrp;
8286 8287 8288 8289
		return &init_task_group.css;
	}

	/* we support only 1-level deep hierarchical scheduler atm */
8290
	if (cgrp->parent->parent)
8291 8292 8293 8294 8295 8296 8297
		return ERR_PTR(-EINVAL);

	tg = sched_create_group();
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	/* Bind the cgroup to task_group object we just created */
8298
	tg->css.cgroup = cgrp;
8299 8300 8301 8302

	return &tg->css;
}

I
Ingo Molnar 已提交
8303 8304
static void
cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
8305
{
8306
	struct task_group *tg = cgroup_tg(cgrp);
8307 8308 8309 8310

	sched_destroy_group(tg);
}

I
Ingo Molnar 已提交
8311 8312 8313
static int
cpu_cgroup_can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
		      struct task_struct *tsk)
8314
{
8315 8316
#ifdef CONFIG_RT_GROUP_SCHED
	/* Don't accept realtime tasks when there is no way for them to run */
8317
	if (rt_task(tsk) && cgroup_tg(cgrp)->rt_bandwidth.rt_runtime == 0)
8318 8319
		return -EINVAL;
#else
8320 8321 8322
	/* We don't support RT-tasks being in separate groups */
	if (tsk->sched_class != &fair_sched_class)
		return -EINVAL;
8323
#endif
8324 8325 8326 8327 8328

	return 0;
}

static void
8329
cpu_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
8330 8331 8332 8333 8334
			struct cgroup *old_cont, struct task_struct *tsk)
{
	sched_move_task(tsk);
}

8335
#ifdef CONFIG_FAIR_GROUP_SCHED
8336 8337
static int cpu_shares_write_uint(struct cgroup *cgrp, struct cftype *cftype,
				u64 shareval)
8338
{
8339
	return sched_group_set_shares(cgroup_tg(cgrp), shareval);
8340 8341
}

8342
static u64 cpu_shares_read_uint(struct cgroup *cgrp, struct cftype *cft)
8343
{
8344
	struct task_group *tg = cgroup_tg(cgrp);
8345 8346 8347

	return (u64) tg->shares;
}
8348
#endif
8349

8350
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8351
static ssize_t cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
P
Peter Zijlstra 已提交
8352 8353 8354
				struct file *file,
				const char __user *userbuf,
				size_t nbytes, loff_t *unused_ppos)
P
Peter Zijlstra 已提交
8355
{
P
Peter Zijlstra 已提交
8356 8357 8358 8359 8360 8361 8362 8363 8364 8365 8366 8367 8368 8369 8370 8371 8372 8373 8374 8375 8376 8377 8378 8379 8380 8381
	char buffer[64];
	int retval = 0;
	s64 val;
	char *end;

	if (!nbytes)
		return -EINVAL;
	if (nbytes >= sizeof(buffer))
		return -E2BIG;
	if (copy_from_user(buffer, userbuf, nbytes))
		return -EFAULT;

	buffer[nbytes] = 0;     /* nul-terminate */

	/* strip newline if necessary */
	if (nbytes && (buffer[nbytes-1] == '\n'))
		buffer[nbytes-1] = 0;
	val = simple_strtoll(buffer, &end, 0);
	if (*end)
		return -EINVAL;

	/* Pass to subsystem */
	retval = sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
	if (!retval)
		retval = nbytes;
	return retval;
P
Peter Zijlstra 已提交
8382 8383
}

P
Peter Zijlstra 已提交
8384 8385 8386 8387
static ssize_t cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft,
				   struct file *file,
				   char __user *buf, size_t nbytes,
				   loff_t *ppos)
P
Peter Zijlstra 已提交
8388
{
P
Peter Zijlstra 已提交
8389 8390 8391
	char tmp[64];
	long val = sched_group_rt_runtime(cgroup_tg(cgrp));
	int len = sprintf(tmp, "%ld\n", val);
P
Peter Zijlstra 已提交
8392

P
Peter Zijlstra 已提交
8393
	return simple_read_from_buffer(buf, nbytes, ppos, tmp, len);
P
Peter Zijlstra 已提交
8394
}
8395 8396 8397 8398 8399 8400 8401 8402 8403 8404 8405

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));
}
8406
#endif
P
Peter Zijlstra 已提交
8407

8408
static struct cftype cpu_files[] = {
8409
#ifdef CONFIG_FAIR_GROUP_SCHED
8410 8411 8412 8413 8414
	{
		.name = "shares",
		.read_uint = cpu_shares_read_uint,
		.write_uint = cpu_shares_write_uint,
	},
8415 8416
#endif
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8417
	{
P
Peter Zijlstra 已提交
8418 8419 8420
		.name = "rt_runtime_us",
		.read = cpu_rt_runtime_read,
		.write = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
8421
	},
8422 8423 8424 8425 8426
	{
		.name = "rt_period_us",
		.read_uint = cpu_rt_period_read_uint,
		.write_uint = cpu_rt_period_write_uint,
	},
8427
#endif
8428 8429 8430 8431
};

static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
{
8432
	return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
8433 8434 8435
}

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
8436 8437 8438 8439 8440 8441 8442
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
	.populate	= cpu_cgroup_populate,
	.subsys_id	= cpu_cgroup_subsys_id,
8443 8444 8445
	.early_init	= 1,
};

8446
#endif	/* CONFIG_CGROUP_SCHED */
8447 8448 8449 8450 8451 8452 8453 8454 8455 8456 8457 8458 8459 8460 8461 8462 8463 8464 8465 8466

#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).
 */

/* track cpu usage of a group of tasks */
struct cpuacct {
	struct cgroup_subsys_state css;
	/* cpuusage holds pointer to a u64-type object on every cpu */
	u64 *cpuusage;
};

struct cgroup_subsys cpuacct_subsys;

/* return cpu accounting group corresponding to this container */
8467
static inline struct cpuacct *cgroup_ca(struct cgroup *cgrp)
8468
{
8469
	return container_of(cgroup_subsys_state(cgrp, cpuacct_subsys_id),
8470 8471 8472 8473 8474 8475 8476 8477 8478 8479 8480 8481
			    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(
8482
	struct cgroup_subsys *ss, struct cgroup *cgrp)
8483 8484 8485 8486 8487 8488 8489 8490 8491 8492 8493 8494 8495 8496 8497 8498
{
	struct cpuacct *ca = kzalloc(sizeof(*ca), GFP_KERNEL);

	if (!ca)
		return ERR_PTR(-ENOMEM);

	ca->cpuusage = alloc_percpu(u64);
	if (!ca->cpuusage) {
		kfree(ca);
		return ERR_PTR(-ENOMEM);
	}

	return &ca->css;
}

/* destroy an existing cpu accounting group */
I
Ingo Molnar 已提交
8499
static void
8500
cpuacct_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
8501
{
8502
	struct cpuacct *ca = cgroup_ca(cgrp);
8503 8504 8505 8506 8507 8508

	free_percpu(ca->cpuusage);
	kfree(ca);
}

/* return total cpu usage (in nanoseconds) of a group */
8509
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
8510
{
8511
	struct cpuacct *ca = cgroup_ca(cgrp);
8512 8513 8514 8515 8516 8517 8518 8519 8520 8521 8522 8523 8524 8525 8526 8527 8528 8529
	u64 totalcpuusage = 0;
	int i;

	for_each_possible_cpu(i) {
		u64 *cpuusage = percpu_ptr(ca->cpuusage, i);

		/*
		 * Take rq->lock to make 64-bit addition safe on 32-bit
		 * platforms.
		 */
		spin_lock_irq(&cpu_rq(i)->lock);
		totalcpuusage += *cpuusage;
		spin_unlock_irq(&cpu_rq(i)->lock);
	}

	return totalcpuusage;
}

8530 8531 8532 8533 8534 8535 8536 8537 8538 8539 8540 8541 8542 8543 8544 8545 8546 8547 8548 8549 8550 8551 8552
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;
	}

	for_each_possible_cpu(i) {
		u64 *cpuusage = percpu_ptr(ca->cpuusage, i);

		spin_lock_irq(&cpu_rq(i)->lock);
		*cpuusage = 0;
		spin_unlock_irq(&cpu_rq(i)->lock);
	}
out:
	return err;
}

8553 8554 8555 8556
static struct cftype files[] = {
	{
		.name = "usage",
		.read_uint = cpuusage_read,
8557
		.write_uint = cpuusage_write,
8558 8559 8560
	},
};

8561
static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
8562
{
8563
	return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
8564 8565 8566 8567 8568 8569 8570 8571 8572 8573 8574 8575 8576 8577 8578 8579 8580 8581 8582 8583 8584 8585 8586 8587 8588 8589 8590 8591 8592 8593
}

/*
 * 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;

	if (!cpuacct_subsys.active)
		return;

	ca = task_ca(tsk);
	if (ca) {
		u64 *cpuusage = percpu_ptr(ca->cpuusage, task_cpu(tsk));

		*cpuusage += cputime;
	}
}

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 */