sched_rt.c 41.8 KB
Newer Older
I
Ingo Molnar 已提交
1 2 3 4 5
/*
 * Real-Time Scheduling Class (mapped to the SCHED_FIFO and SCHED_RR
 * policies)
 */

6 7 8 9
#ifdef CONFIG_RT_GROUP_SCHED

#define rt_entity_is_task(rt_se) (!(rt_se)->my_q)

10 11
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
12 13 14
#ifdef CONFIG_SCHED_DEBUG
	WARN_ON_ONCE(!rt_entity_is_task(rt_se));
#endif
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
	return container_of(rt_se, struct task_struct, rt);
}

static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
{
	return rt_rq->rq;
}

static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se)
{
	return rt_se->rt_rq;
}

#else /* CONFIG_RT_GROUP_SCHED */

30 31
#define rt_entity_is_task(rt_se) (1)

32 33 34 35 36
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
	return container_of(rt_se, struct task_struct, rt);
}

37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
{
	return container_of(rt_rq, struct rq, rt);
}

static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se)
{
	struct task_struct *p = rt_task_of(rt_se);
	struct rq *rq = task_rq(p);

	return &rq->rt;
}

#endif /* CONFIG_RT_GROUP_SCHED */

S
Steven Rostedt 已提交
52
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
53

54
static inline int rt_overloaded(struct rq *rq)
S
Steven Rostedt 已提交
55
{
56
	return atomic_read(&rq->rd->rto_count);
S
Steven Rostedt 已提交
57
}
I
Ingo Molnar 已提交
58

S
Steven Rostedt 已提交
59 60
static inline void rt_set_overload(struct rq *rq)
{
61 62 63
	if (!rq->online)
		return;

64
	cpumask_set_cpu(rq->cpu, rq->rd->rto_mask);
S
Steven Rostedt 已提交
65 66 67 68 69 70 71 72
	/*
	 * Make sure the mask is visible before we set
	 * the overload count. That is checked to determine
	 * if we should look at the mask. It would be a shame
	 * if we looked at the mask, but the mask was not
	 * updated yet.
	 */
	wmb();
73
	atomic_inc(&rq->rd->rto_count);
S
Steven Rostedt 已提交
74
}
I
Ingo Molnar 已提交
75

S
Steven Rostedt 已提交
76 77
static inline void rt_clear_overload(struct rq *rq)
{
78 79 80
	if (!rq->online)
		return;

S
Steven Rostedt 已提交
81
	/* the order here really doesn't matter */
82
	atomic_dec(&rq->rd->rto_count);
83
	cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask);
S
Steven Rostedt 已提交
84
}
85

86
static void update_rt_migration(struct rt_rq *rt_rq)
87
{
88
	if (rt_rq->rt_nr_migratory && rt_rq->rt_nr_total > 1) {
89 90 91
		if (!rt_rq->overloaded) {
			rt_set_overload(rq_of_rt_rq(rt_rq));
			rt_rq->overloaded = 1;
92
		}
93 94 95
	} else if (rt_rq->overloaded) {
		rt_clear_overload(rq_of_rt_rq(rt_rq));
		rt_rq->overloaded = 0;
96
	}
97
}
S
Steven Rostedt 已提交
98

99 100
static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
101 102 103 104 105 106
	if (!rt_entity_is_task(rt_se))
		return;

	rt_rq = &rq_of_rt_rq(rt_rq)->rt;

	rt_rq->rt_nr_total++;
107 108 109 110 111 112 113 114
	if (rt_se->nr_cpus_allowed > 1)
		rt_rq->rt_nr_migratory++;

	update_rt_migration(rt_rq);
}

static void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
115 116 117 118 119 120
	if (!rt_entity_is_task(rt_se))
		return;

	rt_rq = &rq_of_rt_rq(rt_rq)->rt;

	rt_rq->rt_nr_total--;
121 122 123 124 125 126
	if (rt_se->nr_cpus_allowed > 1)
		rt_rq->rt_nr_migratory--;

	update_rt_migration(rt_rq);
}

127 128 129 130 131
static inline int has_pushable_tasks(struct rq *rq)
{
	return !plist_head_empty(&rq->rt.pushable_tasks);
}

132 133 134 135 136
static void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
{
	plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks);
	plist_node_init(&p->pushable_tasks, p->prio);
	plist_add(&p->pushable_tasks, &rq->rt.pushable_tasks);
137 138 139 140

	/* Update the highest prio pushable task */
	if (p->prio < rq->rt.highest_prio.next)
		rq->rt.highest_prio.next = p->prio;
141 142 143 144 145 146
}

static void dequeue_pushable_task(struct rq *rq, struct task_struct *p)
{
	plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks);

147 148 149 150 151 152 153
	/* Update the new highest prio pushable task */
	if (has_pushable_tasks(rq)) {
		p = plist_first_entry(&rq->rt.pushable_tasks,
				      struct task_struct, pushable_tasks);
		rq->rt.highest_prio.next = p->prio;
	} else
		rq->rt.highest_prio.next = MAX_RT_PRIO;
154 155
}

156 157
#else

158
static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
P
Peter Zijlstra 已提交
159
{
P
Peter Zijlstra 已提交
160 161
}

162 163 164 165
static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p)
{
}

166
static inline
167 168 169 170
void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}

171
static inline
172 173 174
void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}
175

S
Steven Rostedt 已提交
176 177
#endif /* CONFIG_SMP */

P
Peter Zijlstra 已提交
178 179 180 181 182
static inline int on_rt_rq(struct sched_rt_entity *rt_se)
{
	return !list_empty(&rt_se->run_list);
}

183
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
184

P
Peter Zijlstra 已提交
185
static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
186 187
{
	if (!rt_rq->tg)
P
Peter Zijlstra 已提交
188
		return RUNTIME_INF;
P
Peter Zijlstra 已提交
189

P
Peter Zijlstra 已提交
190 191 192 193 194 195
	return rt_rq->rt_runtime;
}

static inline u64 sched_rt_period(struct rt_rq *rt_rq)
{
	return ktime_to_ns(rt_rq->tg->rt_bandwidth.rt_period);
P
Peter Zijlstra 已提交
196 197
}

C
Cheng Xu 已提交
198 199
typedef struct task_group *rt_rq_iter_t;

200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216
static inline struct task_group *next_task_group(struct task_group *tg)
{
	do {
		tg = list_entry_rcu(tg->list.next,
			typeof(struct task_group), list);
	} while (&tg->list != &task_groups && task_group_is_autogroup(tg));

	if (&tg->list == &task_groups)
		tg = NULL;

	return tg;
}

#define for_each_rt_rq(rt_rq, iter, rq)					\
	for (iter = container_of(&task_groups, typeof(*iter), list);	\
		(iter = next_task_group(iter)) &&			\
		(rt_rq = iter->rt_rq[cpu_of(rq)]);)
C
Cheng Xu 已提交
217

218 219 220 221 222 223 224 225 226 227 228
static inline void list_add_leaf_rt_rq(struct rt_rq *rt_rq)
{
	list_add_rcu(&rt_rq->leaf_rt_rq_list,
			&rq_of_rt_rq(rt_rq)->leaf_rt_rq_list);
}

static inline void list_del_leaf_rt_rq(struct rt_rq *rt_rq)
{
	list_del_rcu(&rt_rq->leaf_rt_rq_list);
}

P
Peter Zijlstra 已提交
229
#define for_each_leaf_rt_rq(rt_rq, rq) \
230
	list_for_each_entry_rcu(rt_rq, &rq->leaf_rt_rq_list, leaf_rt_rq_list)
P
Peter Zijlstra 已提交
231 232 233 234 235 236 237 238 239

#define for_each_sched_rt_entity(rt_se) \
	for (; rt_se; rt_se = rt_se->parent)

static inline struct rt_rq *group_rt_rq(struct sched_rt_entity *rt_se)
{
	return rt_se->my_q;
}

240
static void enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head);
P
Peter Zijlstra 已提交
241 242
static void dequeue_rt_entity(struct sched_rt_entity *rt_se);

P
Peter Zijlstra 已提交
243
static void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
244
{
245
	struct task_struct *curr = rq_of_rt_rq(rt_rq)->curr;
246 247
	struct sched_rt_entity *rt_se;

248 249 250
	int cpu = cpu_of(rq_of_rt_rq(rt_rq));

	rt_se = rt_rq->tg->rt_se[cpu];
P
Peter Zijlstra 已提交
251

252 253
	if (rt_rq->rt_nr_running) {
		if (rt_se && !on_rt_rq(rt_se))
254
			enqueue_rt_entity(rt_se, false);
255
		if (rt_rq->highest_prio.curr < curr->prio)
256
			resched_task(curr);
P
Peter Zijlstra 已提交
257 258 259
	}
}

P
Peter Zijlstra 已提交
260
static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
261
{
262
	struct sched_rt_entity *rt_se;
263
	int cpu = cpu_of(rq_of_rt_rq(rt_rq));
264

265
	rt_se = rt_rq->tg->rt_se[cpu];
P
Peter Zijlstra 已提交
266 267 268 269 270

	if (rt_se && on_rt_rq(rt_se))
		dequeue_rt_entity(rt_se);
}

P
Peter Zijlstra 已提交
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
static inline int rt_rq_throttled(struct rt_rq *rt_rq)
{
	return rt_rq->rt_throttled && !rt_rq->rt_nr_boosted;
}

static int rt_se_boosted(struct sched_rt_entity *rt_se)
{
	struct rt_rq *rt_rq = group_rt_rq(rt_se);
	struct task_struct *p;

	if (rt_rq)
		return !!rt_rq->rt_nr_boosted;

	p = rt_task_of(rt_se);
	return p->prio != p->normal_prio;
}

288
#ifdef CONFIG_SMP
289
static inline const struct cpumask *sched_rt_period_mask(void)
290 291 292
{
	return cpu_rq(smp_processor_id())->rd->span;
}
P
Peter Zijlstra 已提交
293
#else
294
static inline const struct cpumask *sched_rt_period_mask(void)
295
{
296
	return cpu_online_mask;
297 298
}
#endif
P
Peter Zijlstra 已提交
299

300 301
static inline
struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu)
P
Peter Zijlstra 已提交
302
{
303 304
	return container_of(rt_b, struct task_group, rt_bandwidth)->rt_rq[cpu];
}
P
Peter Zijlstra 已提交
305

P
Peter Zijlstra 已提交
306 307 308 309 310
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &rt_rq->tg->rt_bandwidth;
}

311
#else /* !CONFIG_RT_GROUP_SCHED */
312 313 314

static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
{
P
Peter Zijlstra 已提交
315 316 317 318 319 320
	return rt_rq->rt_runtime;
}

static inline u64 sched_rt_period(struct rt_rq *rt_rq)
{
	return ktime_to_ns(def_rt_bandwidth.rt_period);
P
Peter Zijlstra 已提交
321 322
}

C
Cheng Xu 已提交
323 324 325 326 327
typedef struct rt_rq *rt_rq_iter_t;

#define for_each_rt_rq(rt_rq, iter, rq) \
	for ((void) iter, rt_rq = &rq->rt; rt_rq; rt_rq = NULL)

328 329 330 331 332 333 334 335
static inline void list_add_leaf_rt_rq(struct rt_rq *rt_rq)
{
}

static inline void list_del_leaf_rt_rq(struct rt_rq *rt_rq)
{
}

P
Peter Zijlstra 已提交
336 337 338 339 340 341 342 343 344 345 346
#define for_each_leaf_rt_rq(rt_rq, rq) \
	for (rt_rq = &rq->rt; rt_rq; rt_rq = NULL)

#define for_each_sched_rt_entity(rt_se) \
	for (; rt_se; rt_se = NULL)

static inline struct rt_rq *group_rt_rq(struct sched_rt_entity *rt_se)
{
	return NULL;
}

P
Peter Zijlstra 已提交
347
static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
348
{
349 350
	if (rt_rq->rt_nr_running)
		resched_task(rq_of_rt_rq(rt_rq)->curr);
P
Peter Zijlstra 已提交
351 352
}

P
Peter Zijlstra 已提交
353
static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
354 355 356
{
}

P
Peter Zijlstra 已提交
357 358 359 360
static inline int rt_rq_throttled(struct rt_rq *rt_rq)
{
	return rt_rq->rt_throttled;
}
361

362
static inline const struct cpumask *sched_rt_period_mask(void)
363
{
364
	return cpu_online_mask;
365 366 367 368 369 370 371 372
}

static inline
struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu)
{
	return &cpu_rq(cpu)->rt;
}

P
Peter Zijlstra 已提交
373 374 375 376 377
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &def_rt_bandwidth;
}

378
#endif /* CONFIG_RT_GROUP_SCHED */
379

P
Peter Zijlstra 已提交
380
#ifdef CONFIG_SMP
381 382 383
/*
 * We ran out of runtime, see if we can borrow some from our neighbours.
 */
384
static int do_balance_runtime(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
385 386 387 388 389 390
{
	struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);
	struct root_domain *rd = cpu_rq(smp_processor_id())->rd;
	int i, weight, more = 0;
	u64 rt_period;

391
	weight = cpumask_weight(rd->span);
P
Peter Zijlstra 已提交
392

393
	raw_spin_lock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
394
	rt_period = ktime_to_ns(rt_b->rt_period);
395
	for_each_cpu(i, rd->span) {
P
Peter Zijlstra 已提交
396 397 398 399 400 401
		struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
		s64 diff;

		if (iter == rt_rq)
			continue;

402
		raw_spin_lock(&iter->rt_runtime_lock);
403 404 405 406 407
		/*
		 * Either all rqs have inf runtime and there's nothing to steal
		 * or __disable_runtime() below sets a specific rq to inf to
		 * indicate its been disabled and disalow stealing.
		 */
P
Peter Zijlstra 已提交
408 409 410
		if (iter->rt_runtime == RUNTIME_INF)
			goto next;

411 412 413 414
		/*
		 * From runqueues with spare time, take 1/n part of their
		 * spare time, but no more than our period.
		 */
P
Peter Zijlstra 已提交
415 416
		diff = iter->rt_runtime - iter->rt_time;
		if (diff > 0) {
417
			diff = div_u64((u64)diff, weight);
P
Peter Zijlstra 已提交
418 419 420 421 422 423
			if (rt_rq->rt_runtime + diff > rt_period)
				diff = rt_period - rt_rq->rt_runtime;
			iter->rt_runtime -= diff;
			rt_rq->rt_runtime += diff;
			more = 1;
			if (rt_rq->rt_runtime == rt_period) {
424
				raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
425 426 427
				break;
			}
		}
P
Peter Zijlstra 已提交
428
next:
429
		raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
430
	}
431
	raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
432 433 434

	return more;
}
P
Peter Zijlstra 已提交
435

436 437 438
/*
 * Ensure this RQ takes back all the runtime it lend to its neighbours.
 */
P
Peter Zijlstra 已提交
439 440 441
static void __disable_runtime(struct rq *rq)
{
	struct root_domain *rd = rq->rd;
C
Cheng Xu 已提交
442
	rt_rq_iter_t iter;
P
Peter Zijlstra 已提交
443 444 445 446 447
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

C
Cheng Xu 已提交
448
	for_each_rt_rq(rt_rq, iter, rq) {
P
Peter Zijlstra 已提交
449 450 451 452
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);
		s64 want;
		int i;

453 454
		raw_spin_lock(&rt_b->rt_runtime_lock);
		raw_spin_lock(&rt_rq->rt_runtime_lock);
455 456 457 458 459
		/*
		 * Either we're all inf and nobody needs to borrow, or we're
		 * already disabled and thus have nothing to do, or we have
		 * exactly the right amount of runtime to take out.
		 */
P
Peter Zijlstra 已提交
460 461 462
		if (rt_rq->rt_runtime == RUNTIME_INF ||
				rt_rq->rt_runtime == rt_b->rt_runtime)
			goto balanced;
463
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
464

465 466 467 468 469
		/*
		 * Calculate the difference between what we started out with
		 * and what we current have, that's the amount of runtime
		 * we lend and now have to reclaim.
		 */
P
Peter Zijlstra 已提交
470 471
		want = rt_b->rt_runtime - rt_rq->rt_runtime;

472 473 474
		/*
		 * Greedy reclaim, take back as much as we can.
		 */
475
		for_each_cpu(i, rd->span) {
P
Peter Zijlstra 已提交
476 477 478
			struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
			s64 diff;

479 480 481
			/*
			 * Can't reclaim from ourselves or disabled runqueues.
			 */
482
			if (iter == rt_rq || iter->rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
483 484
				continue;

485
			raw_spin_lock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
486 487 488 489 490 491 492 493
			if (want > 0) {
				diff = min_t(s64, iter->rt_runtime, want);
				iter->rt_runtime -= diff;
				want -= diff;
			} else {
				iter->rt_runtime -= want;
				want -= want;
			}
494
			raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
495 496 497 498 499

			if (!want)
				break;
		}

500
		raw_spin_lock(&rt_rq->rt_runtime_lock);
501 502 503 504
		/*
		 * We cannot be left wanting - that would mean some runtime
		 * leaked out of the system.
		 */
P
Peter Zijlstra 已提交
505 506
		BUG_ON(want);
balanced:
507 508 509 510
		/*
		 * Disable all the borrow logic by pretending we have inf
		 * runtime - in which case borrowing doesn't make sense.
		 */
P
Peter Zijlstra 已提交
511
		rt_rq->rt_runtime = RUNTIME_INF;
512 513
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
		raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
514 515 516 517 518 519 520
	}
}

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

521
	raw_spin_lock_irqsave(&rq->lock, flags);
P
Peter Zijlstra 已提交
522
	__disable_runtime(rq);
523
	raw_spin_unlock_irqrestore(&rq->lock, flags);
P
Peter Zijlstra 已提交
524 525 526 527
}

static void __enable_runtime(struct rq *rq)
{
C
Cheng Xu 已提交
528
	rt_rq_iter_t iter;
P
Peter Zijlstra 已提交
529 530 531 532 533
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

534 535 536
	/*
	 * Reset each runqueue's bandwidth settings
	 */
C
Cheng Xu 已提交
537
	for_each_rt_rq(rt_rq, iter, rq) {
P
Peter Zijlstra 已提交
538 539
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);

540 541
		raw_spin_lock(&rt_b->rt_runtime_lock);
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
542 543
		rt_rq->rt_runtime = rt_b->rt_runtime;
		rt_rq->rt_time = 0;
544
		rt_rq->rt_throttled = 0;
545 546
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
		raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
547 548 549 550 551 552 553
	}
}

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

554
	raw_spin_lock_irqsave(&rq->lock, flags);
P
Peter Zijlstra 已提交
555
	__enable_runtime(rq);
556
	raw_spin_unlock_irqrestore(&rq->lock, flags);
P
Peter Zijlstra 已提交
557 558
}

559 560 561 562 563
static int balance_runtime(struct rt_rq *rt_rq)
{
	int more = 0;

	if (rt_rq->rt_time > rt_rq->rt_runtime) {
564
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
565
		more = do_balance_runtime(rt_rq);
566
		raw_spin_lock(&rt_rq->rt_runtime_lock);
567 568 569 570
	}

	return more;
}
571
#else /* !CONFIG_SMP */
572 573 574 575
static inline int balance_runtime(struct rt_rq *rt_rq)
{
	return 0;
}
576
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
577

578 579 580
static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
{
	int i, idle = 1;
581
	const struct cpumask *span;
582

583
	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
584 585 586
		return 1;

	span = sched_rt_period_mask();
587
	for_each_cpu(i, span) {
588 589 590 591
		int enqueue = 0;
		struct rt_rq *rt_rq = sched_rt_period_rt_rq(rt_b, i);
		struct rq *rq = rq_of_rt_rq(rt_rq);

592
		raw_spin_lock(&rq->lock);
593 594 595
		if (rt_rq->rt_time) {
			u64 runtime;

596
			raw_spin_lock(&rt_rq->rt_runtime_lock);
597 598 599 600 601 602 603
			if (rt_rq->rt_throttled)
				balance_runtime(rt_rq);
			runtime = rt_rq->rt_runtime;
			rt_rq->rt_time -= min(rt_rq->rt_time, overrun*runtime);
			if (rt_rq->rt_throttled && rt_rq->rt_time < runtime) {
				rt_rq->rt_throttled = 0;
				enqueue = 1;
604 605 606 607 608 609 610

				/*
				 * Force a clock update if the CPU was idle,
				 * lest wakeup -> unthrottle time accumulate.
				 */
				if (rt_rq->rt_nr_running && rq->curr == rq->idle)
					rq->skip_clock_update = -1;
611 612 613
			}
			if (rt_rq->rt_time || rt_rq->rt_nr_running)
				idle = 0;
614
			raw_spin_unlock(&rt_rq->rt_runtime_lock);
615
		} else if (rt_rq->rt_nr_running) {
616
			idle = 0;
617 618 619
			if (!rt_rq_throttled(rt_rq))
				enqueue = 1;
		}
620 621 622

		if (enqueue)
			sched_rt_rq_enqueue(rt_rq);
623
		raw_spin_unlock(&rq->lock);
624 625 626 627
	}

	return idle;
}
P
Peter Zijlstra 已提交
628

P
Peter Zijlstra 已提交
629 630
static inline int rt_se_prio(struct sched_rt_entity *rt_se)
{
631
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
632 633 634
	struct rt_rq *rt_rq = group_rt_rq(rt_se);

	if (rt_rq)
635
		return rt_rq->highest_prio.curr;
P
Peter Zijlstra 已提交
636 637 638 639 640
#endif

	return rt_task_of(rt_se)->prio;
}

P
Peter Zijlstra 已提交
641
static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
642
{
P
Peter Zijlstra 已提交
643
	u64 runtime = sched_rt_runtime(rt_rq);
P
Peter Zijlstra 已提交
644 645

	if (rt_rq->rt_throttled)
P
Peter Zijlstra 已提交
646
		return rt_rq_throttled(rt_rq);
P
Peter Zijlstra 已提交
647

P
Peter Zijlstra 已提交
648 649 650
	if (sched_rt_runtime(rt_rq) >= sched_rt_period(rt_rq))
		return 0;

651 652 653 654
	balance_runtime(rt_rq);
	runtime = sched_rt_runtime(rt_rq);
	if (runtime == RUNTIME_INF)
		return 0;
P
Peter Zijlstra 已提交
655

P
Peter Zijlstra 已提交
656
	if (rt_rq->rt_time > runtime) {
P
Peter Zijlstra 已提交
657
		rt_rq->rt_throttled = 1;
P
Peter Zijlstra 已提交
658
		if (rt_rq_throttled(rt_rq)) {
P
Peter Zijlstra 已提交
659
			sched_rt_rq_dequeue(rt_rq);
P
Peter Zijlstra 已提交
660 661
			return 1;
		}
P
Peter Zijlstra 已提交
662 663 664 665 666
	}

	return 0;
}

I
Ingo Molnar 已提交
667 668 669 670
/*
 * Update the current task's runtime statistics. Skip current tasks that
 * are not in our scheduling class.
 */
A
Alexey Dobriyan 已提交
671
static void update_curr_rt(struct rq *rq)
I
Ingo Molnar 已提交
672 673
{
	struct task_struct *curr = rq->curr;
P
Peter Zijlstra 已提交
674 675
	struct sched_rt_entity *rt_se = &curr->rt;
	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
I
Ingo Molnar 已提交
676 677
	u64 delta_exec;

P
Peter Zijlstra 已提交
678
	if (curr->sched_class != &rt_sched_class)
I
Ingo Molnar 已提交
679 680
		return;

681
	delta_exec = rq->clock_task - curr->se.exec_start;
I
Ingo Molnar 已提交
682 683
	if (unlikely((s64)delta_exec < 0))
		delta_exec = 0;
I
Ingo Molnar 已提交
684

685
	schedstat_set(curr->se.statistics.exec_max, max(curr->se.statistics.exec_max, delta_exec));
I
Ingo Molnar 已提交
686 687

	curr->se.sum_exec_runtime += delta_exec;
688 689
	account_group_exec_runtime(curr, delta_exec);

690
	curr->se.exec_start = rq->clock_task;
691
	cpuacct_charge(curr, delta_exec);
P
Peter Zijlstra 已提交
692

693 694
	sched_rt_avg_update(rq, delta_exec);

695 696 697
	if (!rt_bandwidth_enabled())
		return;

D
Dhaval Giani 已提交
698 699 700
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);

701
		if (sched_rt_runtime(rt_rq) != RUNTIME_INF) {
702
			raw_spin_lock(&rt_rq->rt_runtime_lock);
703 704 705
			rt_rq->rt_time += delta_exec;
			if (sched_rt_runtime_exceeded(rt_rq))
				resched_task(curr);
706
			raw_spin_unlock(&rt_rq->rt_runtime_lock);
707
		}
D
Dhaval Giani 已提交
708
	}
I
Ingo Molnar 已提交
709 710
}

711
#if defined CONFIG_SMP
712

713 714
static void
inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
715
{
G
Gregory Haskins 已提交
716
	struct rq *rq = rq_of_rt_rq(rt_rq);
717

718 719
	if (rq->online && prio < prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, prio);
720
}
721

722 723 724 725
static void
dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
{
	struct rq *rq = rq_of_rt_rq(rt_rq);
726

727 728
	if (rq->online && rt_rq->highest_prio.curr != prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr);
729 730
}

731 732
#else /* CONFIG_SMP */

P
Peter Zijlstra 已提交
733
static inline
734 735 736 737 738
void inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) {}
static inline
void dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) {}

#endif /* CONFIG_SMP */
739

740
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756
static void
inc_rt_prio(struct rt_rq *rt_rq, int prio)
{
	int prev_prio = rt_rq->highest_prio.curr;

	if (prio < prev_prio)
		rt_rq->highest_prio.curr = prio;

	inc_rt_prio_smp(rt_rq, prio, prev_prio);
}

static void
dec_rt_prio(struct rt_rq *rt_rq, int prio)
{
	int prev_prio = rt_rq->highest_prio.curr;

P
Peter Zijlstra 已提交
757
	if (rt_rq->rt_nr_running) {
758

759
		WARN_ON(prio < prev_prio);
760

761
		/*
762 763
		 * This may have been our highest task, and therefore
		 * we may have some recomputation to do
764
		 */
765
		if (prio == prev_prio) {
766 767 768
			struct rt_prio_array *array = &rt_rq->active;

			rt_rq->highest_prio.curr =
769
				sched_find_first_bit(array->bitmap);
770 771
		}

772
	} else
773
		rt_rq->highest_prio.curr = MAX_RT_PRIO;
774

775 776
	dec_rt_prio_smp(rt_rq, prio, prev_prio);
}
777

778 779 780 781 782 783
#else

static inline void inc_rt_prio(struct rt_rq *rt_rq, int prio) {}
static inline void dec_rt_prio(struct rt_rq *rt_rq, int prio) {}

#endif /* CONFIG_SMP || CONFIG_RT_GROUP_SCHED */
784

785
#ifdef CONFIG_RT_GROUP_SCHED
786 787 788 789 790 791 792 793 794 795 796 797 798 799

static void
inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
	if (rt_se_boosted(rt_se))
		rt_rq->rt_nr_boosted++;

	if (rt_rq->tg)
		start_rt_bandwidth(&rt_rq->tg->rt_bandwidth);
}

static void
dec_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
P
Peter Zijlstra 已提交
800 801 802 803
	if (rt_se_boosted(rt_se))
		rt_rq->rt_nr_boosted--;

	WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted);
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
}

#else /* CONFIG_RT_GROUP_SCHED */

static void
inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
	start_rt_bandwidth(&def_rt_bandwidth);
}

static inline
void dec_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {}

#endif /* CONFIG_RT_GROUP_SCHED */

static inline
void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
	int prio = rt_se_prio(rt_se);

	WARN_ON(!rt_prio(prio));
	rt_rq->rt_nr_running++;

	inc_rt_prio(rt_rq, prio);
	inc_rt_migration(rt_se, rt_rq);
	inc_rt_group(rt_se, rt_rq);
}

static inline
void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
	WARN_ON(!rt_prio(rt_se_prio(rt_se)));
	WARN_ON(!rt_rq->rt_nr_running);
	rt_rq->rt_nr_running--;

	dec_rt_prio(rt_rq, rt_se_prio(rt_se));
	dec_rt_migration(rt_se, rt_rq);
	dec_rt_group(rt_se, rt_rq);
842 843
}

844
static void __enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head)
I
Ingo Molnar 已提交
845
{
P
Peter Zijlstra 已提交
846 847 848
	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
	struct rt_prio_array *array = &rt_rq->active;
	struct rt_rq *group_rq = group_rt_rq(rt_se);
849
	struct list_head *queue = array->queue + rt_se_prio(rt_se);
I
Ingo Molnar 已提交
850

851 852 853 854 855 856 857
	/*
	 * Don't enqueue the group if its throttled, or when empty.
	 * The latter is a consequence of the former when a child group
	 * get throttled and the current group doesn't have any other
	 * active members.
	 */
	if (group_rq && (rt_rq_throttled(group_rq) || !group_rq->rt_nr_running))
P
Peter Zijlstra 已提交
858
		return;
859

860 861 862
	if (!rt_rq->rt_nr_running)
		list_add_leaf_rt_rq(rt_rq);

863 864 865 866
	if (head)
		list_add(&rt_se->run_list, queue);
	else
		list_add_tail(&rt_se->run_list, queue);
P
Peter Zijlstra 已提交
867
	__set_bit(rt_se_prio(rt_se), array->bitmap);
868

P
Peter Zijlstra 已提交
869 870 871
	inc_rt_tasks(rt_se, rt_rq);
}

872
static void __dequeue_rt_entity(struct sched_rt_entity *rt_se)
P
Peter Zijlstra 已提交
873 874 875 876 877 878 879 880 881
{
	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
	struct rt_prio_array *array = &rt_rq->active;

	list_del_init(&rt_se->run_list);
	if (list_empty(array->queue + rt_se_prio(rt_se)))
		__clear_bit(rt_se_prio(rt_se), array->bitmap);

	dec_rt_tasks(rt_se, rt_rq);
882 883
	if (!rt_rq->rt_nr_running)
		list_del_leaf_rt_rq(rt_rq);
P
Peter Zijlstra 已提交
884 885 886 887 888 889
}

/*
 * Because the prio of an upper entry depends on the lower
 * entries, we must remove entries top - down.
 */
890
static void dequeue_rt_stack(struct sched_rt_entity *rt_se)
P
Peter Zijlstra 已提交
891
{
892
	struct sched_rt_entity *back = NULL;
P
Peter Zijlstra 已提交
893

894 895 896 897 898 899 900
	for_each_sched_rt_entity(rt_se) {
		rt_se->back = back;
		back = rt_se;
	}

	for (rt_se = back; rt_se; rt_se = rt_se->back) {
		if (on_rt_rq(rt_se))
901 902 903 904
			__dequeue_rt_entity(rt_se);
	}
}

905
static void enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head)
906 907 908
{
	dequeue_rt_stack(rt_se);
	for_each_sched_rt_entity(rt_se)
909
		__enqueue_rt_entity(rt_se, head);
910 911 912 913 914 915 916 917 918 919
}

static void dequeue_rt_entity(struct sched_rt_entity *rt_se)
{
	dequeue_rt_stack(rt_se);

	for_each_sched_rt_entity(rt_se) {
		struct rt_rq *rt_rq = group_rt_rq(rt_se);

		if (rt_rq && rt_rq->rt_nr_running)
920
			__enqueue_rt_entity(rt_se, false);
921
	}
I
Ingo Molnar 已提交
922 923 924 925 926
}

/*
 * Adding/removing a task to/from a priority array:
 */
927
static void
928
enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags)
P
Peter Zijlstra 已提交
929 930 931
{
	struct sched_rt_entity *rt_se = &p->rt;

932
	if (flags & ENQUEUE_WAKEUP)
P
Peter Zijlstra 已提交
933 934
		rt_se->timeout = 0;

935
	enqueue_rt_entity(rt_se, flags & ENQUEUE_HEAD);
936

937 938
	if (!task_current(rq, p) && p->rt.nr_cpus_allowed > 1)
		enqueue_pushable_task(rq, p);
939 940

	inc_nr_running(rq);
P
Peter Zijlstra 已提交
941 942
}

943
static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
944
{
P
Peter Zijlstra 已提交
945
	struct sched_rt_entity *rt_se = &p->rt;
I
Ingo Molnar 已提交
946

947
	update_curr_rt(rq);
948
	dequeue_rt_entity(rt_se);
949

950
	dequeue_pushable_task(rq, p);
951 952

	dec_nr_running(rq);
I
Ingo Molnar 已提交
953 954 955 956 957 958
}

/*
 * Put task to the end of the run list without the overhead of dequeue
 * followed by enqueue.
 */
959 960
static void
requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head)
P
Peter Zijlstra 已提交
961
{
962
	if (on_rt_rq(rt_se)) {
963 964 965 966 967 968 969
		struct rt_prio_array *array = &rt_rq->active;
		struct list_head *queue = array->queue + rt_se_prio(rt_se);

		if (head)
			list_move(&rt_se->run_list, queue);
		else
			list_move_tail(&rt_se->run_list, queue);
970
	}
P
Peter Zijlstra 已提交
971 972
}

973
static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head)
I
Ingo Molnar 已提交
974
{
P
Peter Zijlstra 已提交
975 976
	struct sched_rt_entity *rt_se = &p->rt;
	struct rt_rq *rt_rq;
I
Ingo Molnar 已提交
977

P
Peter Zijlstra 已提交
978 979
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);
980
		requeue_rt_entity(rt_rq, rt_se, head);
P
Peter Zijlstra 已提交
981
	}
I
Ingo Molnar 已提交
982 983
}

P
Peter Zijlstra 已提交
984
static void yield_task_rt(struct rq *rq)
I
Ingo Molnar 已提交
985
{
986
	requeue_task_rt(rq, rq->curr, 0);
I
Ingo Molnar 已提交
987 988
}

989
#ifdef CONFIG_SMP
990 991
static int find_lowest_rq(struct task_struct *task);

992
static int
993
select_task_rq_rt(struct task_struct *p, int sd_flag, int flags)
994
{
995 996 997 998 999
	struct task_struct *curr;
	struct rq *rq;
	int cpu;

	cpu = task_cpu(p);
1000 1001 1002 1003 1004

	/* For anything but wake ups, just return the task_cpu */
	if (sd_flag != SD_BALANCE_WAKE && sd_flag != SD_BALANCE_FORK)
		goto out;

1005 1006 1007 1008 1009
	rq = cpu_rq(cpu);

	rcu_read_lock();
	curr = ACCESS_ONCE(rq->curr); /* unlocked access */

1010
	/*
1011
	 * If the current task on @p's runqueue is an RT task, then
1012 1013 1014 1015
	 * try to see if we can wake this RT task up on another
	 * runqueue. Otherwise simply start this RT task
	 * on its current runqueue.
	 *
1016 1017 1018 1019 1020 1021 1022 1023 1024
	 * We want to avoid overloading runqueues. If the woken
	 * task is a higher priority, then it will stay on this CPU
	 * and the lower prio task should be moved to another CPU.
	 * Even though this will probably make the lower prio task
	 * lose its cache, we do not want to bounce a higher task
	 * around just because it gave up its CPU, perhaps for a
	 * lock?
	 *
	 * For equal prio tasks, we just let the scheduler sort it out.
1025 1026 1027 1028 1029 1030
	 *
	 * Otherwise, just let it ride on the affined RQ and the
	 * post-schedule router will push the preempted task away
	 *
	 * This test is optimistic, if we get it wrong the load-balancer
	 * will have to sort it out.
1031
	 */
1032 1033
	if (curr && unlikely(rt_task(curr)) &&
	    (curr->rt.nr_cpus_allowed < 2 ||
1034
	     curr->prio <= p->prio) &&
P
Peter Zijlstra 已提交
1035
	    (p->rt.nr_cpus_allowed > 1)) {
1036
		int target = find_lowest_rq(p);
1037

1038 1039
		if (target != -1)
			cpu = target;
1040
	}
1041
	rcu_read_unlock();
1042

1043
out:
1044
	return cpu;
1045
}
1046 1047 1048 1049 1050 1051

static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
{
	if (rq->curr->rt.nr_cpus_allowed == 1)
		return;

1052
	if (p->rt.nr_cpus_allowed != 1
1053 1054
	    && cpupri_find(&rq->rd->cpupri, p, NULL))
		return;
1055

1056 1057
	if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL))
		return;
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067

	/*
	 * There appears to be other cpus that can accept
	 * current and none to run 'p', so lets reschedule
	 * to try and push current away:
	 */
	requeue_task_rt(rq, p, 1);
	resched_task(rq->curr);
}

1068 1069
#endif /* CONFIG_SMP */

I
Ingo Molnar 已提交
1070 1071 1072
/*
 * Preempt the current task with a newly woken task if needed:
 */
P
Peter Zijlstra 已提交
1073
static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
1074
{
1075
	if (p->prio < rq->curr->prio) {
I
Ingo Molnar 已提交
1076
		resched_task(rq->curr);
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
		return;
	}

#ifdef CONFIG_SMP
	/*
	 * If:
	 *
	 * - the newly woken task is of equal priority to the current task
	 * - the newly woken task is non-migratable while current is migratable
	 * - current will be preempted on the next reschedule
	 *
	 * we should check to see if current can readily move to a different
	 * cpu.  If so, we will reschedule to allow the push logic to try
	 * to move current somewhere else, making room for our non-migratable
	 * task.
	 */
1093
	if (p->prio == rq->curr->prio && !test_tsk_need_resched(rq->curr))
1094
		check_preempt_equal_prio(rq, p);
1095
#endif
I
Ingo Molnar 已提交
1096 1097
}

P
Peter Zijlstra 已提交
1098 1099
static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq,
						   struct rt_rq *rt_rq)
I
Ingo Molnar 已提交
1100
{
P
Peter Zijlstra 已提交
1101 1102
	struct rt_prio_array *array = &rt_rq->active;
	struct sched_rt_entity *next = NULL;
I
Ingo Molnar 已提交
1103 1104 1105 1106
	struct list_head *queue;
	int idx;

	idx = sched_find_first_bit(array->bitmap);
P
Peter Zijlstra 已提交
1107
	BUG_ON(idx >= MAX_RT_PRIO);
I
Ingo Molnar 已提交
1108 1109

	queue = array->queue + idx;
P
Peter Zijlstra 已提交
1110
	next = list_entry(queue->next, struct sched_rt_entity, run_list);
1111

P
Peter Zijlstra 已提交
1112 1113
	return next;
}
I
Ingo Molnar 已提交
1114

1115
static struct task_struct *_pick_next_task_rt(struct rq *rq)
P
Peter Zijlstra 已提交
1116 1117 1118 1119
{
	struct sched_rt_entity *rt_se;
	struct task_struct *p;
	struct rt_rq *rt_rq;
I
Ingo Molnar 已提交
1120

P
Peter Zijlstra 已提交
1121 1122
	rt_rq = &rq->rt;

1123
	if (!rt_rq->rt_nr_running)
P
Peter Zijlstra 已提交
1124 1125
		return NULL;

P
Peter Zijlstra 已提交
1126
	if (rt_rq_throttled(rt_rq))
P
Peter Zijlstra 已提交
1127 1128 1129 1130
		return NULL;

	do {
		rt_se = pick_next_rt_entity(rq, rt_rq);
1131
		BUG_ON(!rt_se);
P
Peter Zijlstra 已提交
1132 1133 1134 1135
		rt_rq = group_rt_rq(rt_se);
	} while (rt_rq);

	p = rt_task_of(rt_se);
1136
	p->se.exec_start = rq->clock_task;
1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148

	return p;
}

static struct task_struct *pick_next_task_rt(struct rq *rq)
{
	struct task_struct *p = _pick_next_task_rt(rq);

	/* The running task is never eligible for pushing */
	if (p)
		dequeue_pushable_task(rq, p);

1149
#ifdef CONFIG_SMP
1150 1151 1152 1153 1154
	/*
	 * We detect this state here so that we can avoid taking the RQ
	 * lock again later if there is no need to push
	 */
	rq->post_schedule = has_pushable_tasks(rq);
1155
#endif
1156

P
Peter Zijlstra 已提交
1157
	return p;
I
Ingo Molnar 已提交
1158 1159
}

1160
static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
I
Ingo Molnar 已提交
1161
{
1162
	update_curr_rt(rq);
1163 1164 1165 1166 1167

	/*
	 * The previous task needs to be made eligible for pushing
	 * if it is still active
	 */
P
Peter Zijlstra 已提交
1168
	if (on_rt_rq(&p->rt) && p->rt.nr_cpus_allowed > 1)
1169
		enqueue_pushable_task(rq, p);
I
Ingo Molnar 已提交
1170 1171
}

1172
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1173

S
Steven Rostedt 已提交
1174 1175 1176 1177 1178
/* Only try algorithms three times */
#define RT_MAX_TRIES 3

static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep);

1179 1180 1181
static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu)
{
	if (!task_running(rq, p) &&
1182
	    (cpu < 0 || cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) &&
P
Peter Zijlstra 已提交
1183
	    (p->rt.nr_cpus_allowed > 1))
1184 1185 1186 1187
		return 1;
	return 0;
}

S
Steven Rostedt 已提交
1188
/* Return the second highest RT task, NULL otherwise */
1189
static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu)
S
Steven Rostedt 已提交
1190
{
P
Peter Zijlstra 已提交
1191 1192 1193 1194
	struct task_struct *next = NULL;
	struct sched_rt_entity *rt_se;
	struct rt_prio_array *array;
	struct rt_rq *rt_rq;
S
Steven Rostedt 已提交
1195 1196
	int idx;

P
Peter Zijlstra 已提交
1197 1198 1199
	for_each_leaf_rt_rq(rt_rq, rq) {
		array = &rt_rq->active;
		idx = sched_find_first_bit(array->bitmap);
P
Peter Zijlstra 已提交
1200
next_idx:
P
Peter Zijlstra 已提交
1201 1202 1203 1204 1205
		if (idx >= MAX_RT_PRIO)
			continue;
		if (next && next->prio < idx)
			continue;
		list_for_each_entry(rt_se, array->queue + idx, run_list) {
1206 1207 1208 1209 1210 1211
			struct task_struct *p;

			if (!rt_entity_is_task(rt_se))
				continue;

			p = rt_task_of(rt_se);
P
Peter Zijlstra 已提交
1212 1213 1214 1215 1216 1217 1218 1219 1220
			if (pick_rt_task(rq, p, cpu)) {
				next = p;
				break;
			}
		}
		if (!next) {
			idx = find_next_bit(array->bitmap, MAX_RT_PRIO, idx+1);
			goto next_idx;
		}
1221 1222
	}

S
Steven Rostedt 已提交
1223 1224 1225
	return next;
}

1226
static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask);
S
Steven Rostedt 已提交
1227

G
Gregory Haskins 已提交
1228 1229 1230
static int find_lowest_rq(struct task_struct *task)
{
	struct sched_domain *sd;
1231
	struct cpumask *lowest_mask = __get_cpu_var(local_cpu_mask);
G
Gregory Haskins 已提交
1232 1233
	int this_cpu = smp_processor_id();
	int cpu      = task_cpu(task);
G
Gregory Haskins 已提交
1234

1235 1236 1237 1238
	/* Make sure the mask is initialized first */
	if (unlikely(!lowest_mask))
		return -1;

1239 1240
	if (task->rt.nr_cpus_allowed == 1)
		return -1; /* No other targets possible */
G
Gregory Haskins 已提交
1241

1242 1243
	if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask))
		return -1; /* No targets found */
G
Gregory Haskins 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252

	/*
	 * At this point we have built a mask of cpus representing the
	 * lowest priority tasks in the system.  Now we want to elect
	 * the best one based on our affinity and topology.
	 *
	 * We prioritize the last cpu that the task executed on since
	 * it is most likely cache-hot in that location.
	 */
1253
	if (cpumask_test_cpu(cpu, lowest_mask))
G
Gregory Haskins 已提交
1254 1255 1256 1257 1258 1259
		return cpu;

	/*
	 * Otherwise, we consult the sched_domains span maps to figure
	 * out which cpu is logically closest to our hot cache data.
	 */
R
Rusty Russell 已提交
1260 1261
	if (!cpumask_test_cpu(this_cpu, lowest_mask))
		this_cpu = -1; /* Skip this_cpu opt if not among lowest */
G
Gregory Haskins 已提交
1262

1263
	rcu_read_lock();
R
Rusty Russell 已提交
1264 1265 1266
	for_each_domain(cpu, sd) {
		if (sd->flags & SD_WAKE_AFFINE) {
			int best_cpu;
G
Gregory Haskins 已提交
1267

R
Rusty Russell 已提交
1268 1269 1270 1271 1272
			/*
			 * "this_cpu" is cheaper to preempt than a
			 * remote processor.
			 */
			if (this_cpu != -1 &&
1273 1274
			    cpumask_test_cpu(this_cpu, sched_domain_span(sd))) {
				rcu_read_unlock();
R
Rusty Russell 已提交
1275
				return this_cpu;
1276
			}
R
Rusty Russell 已提交
1277 1278 1279

			best_cpu = cpumask_first_and(lowest_mask,
						     sched_domain_span(sd));
1280 1281
			if (best_cpu < nr_cpu_ids) {
				rcu_read_unlock();
R
Rusty Russell 已提交
1282
				return best_cpu;
1283
			}
G
Gregory Haskins 已提交
1284 1285
		}
	}
1286
	rcu_read_unlock();
G
Gregory Haskins 已提交
1287 1288 1289 1290 1291 1292

	/*
	 * And finally, if there were no matches within the domains
	 * just give the caller *something* to work with from the compatible
	 * locations.
	 */
R
Rusty Russell 已提交
1293 1294 1295 1296 1297 1298 1299
	if (this_cpu != -1)
		return this_cpu;

	cpu = cpumask_any(lowest_mask);
	if (cpu < nr_cpu_ids)
		return cpu;
	return -1;
1300 1301 1302
}

/* Will lock the rq it finds */
1303
static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq)
1304 1305 1306
{
	struct rq *lowest_rq = NULL;
	int tries;
1307
	int cpu;
S
Steven Rostedt 已提交
1308

1309 1310 1311
	for (tries = 0; tries < RT_MAX_TRIES; tries++) {
		cpu = find_lowest_rq(task);

1312
		if ((cpu == -1) || (cpu == rq->cpu))
S
Steven Rostedt 已提交
1313 1314
			break;

1315 1316
		lowest_rq = cpu_rq(cpu);

S
Steven Rostedt 已提交
1317
		/* if the prio of this runqueue changed, try again */
1318
		if (double_lock_balance(rq, lowest_rq)) {
S
Steven Rostedt 已提交
1319 1320 1321 1322 1323 1324
			/*
			 * We had to unlock the run queue. In
			 * the mean time, task could have
			 * migrated already or had its affinity changed.
			 * Also make sure that it wasn't scheduled on its rq.
			 */
1325
			if (unlikely(task_rq(task) != rq ||
1326
				     !cpumask_test_cpu(lowest_rq->cpu,
1327
						       tsk_cpus_allowed(task)) ||
1328
				     task_running(rq, task) ||
P
Peter Zijlstra 已提交
1329
				     !task->on_rq)) {
1330

1331
				raw_spin_unlock(&lowest_rq->lock);
S
Steven Rostedt 已提交
1332 1333 1334 1335 1336 1337
				lowest_rq = NULL;
				break;
			}
		}

		/* If this rq is still suitable use it. */
1338
		if (lowest_rq->rt.highest_prio.curr > task->prio)
S
Steven Rostedt 已提交
1339 1340 1341
			break;

		/* try again */
1342
		double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1343 1344 1345 1346 1347 1348
		lowest_rq = NULL;
	}

	return lowest_rq;
}

1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362
static struct task_struct *pick_next_pushable_task(struct rq *rq)
{
	struct task_struct *p;

	if (!has_pushable_tasks(rq))
		return NULL;

	p = plist_first_entry(&rq->rt.pushable_tasks,
			      struct task_struct, pushable_tasks);

	BUG_ON(rq->cpu != task_cpu(p));
	BUG_ON(task_current(rq, p));
	BUG_ON(p->rt.nr_cpus_allowed <= 1);

P
Peter Zijlstra 已提交
1363
	BUG_ON(!p->on_rq);
1364 1365 1366 1367 1368
	BUG_ON(!rt_task(p));

	return p;
}

S
Steven Rostedt 已提交
1369 1370 1371 1372 1373
/*
 * If the current CPU has more than one RT task, see if the non
 * running task can migrate over to a CPU that is running a task
 * of lesser priority.
 */
1374
static int push_rt_task(struct rq *rq)
S
Steven Rostedt 已提交
1375 1376 1377
{
	struct task_struct *next_task;
	struct rq *lowest_rq;
1378
	int ret = 0;
S
Steven Rostedt 已提交
1379

G
Gregory Haskins 已提交
1380 1381 1382
	if (!rq->rt.overloaded)
		return 0;

1383
	next_task = pick_next_pushable_task(rq);
S
Steven Rostedt 已提交
1384 1385 1386
	if (!next_task)
		return 0;

P
Peter Zijlstra 已提交
1387
retry:
1388
	if (unlikely(next_task == rq->curr)) {
1389
		WARN_ON(1);
S
Steven Rostedt 已提交
1390
		return 0;
1391
	}
S
Steven Rostedt 已提交
1392 1393 1394 1395 1396 1397

	/*
	 * It's possible that the next_task slipped in of
	 * higher priority than current. If that's the case
	 * just reschedule current.
	 */
1398 1399
	if (unlikely(next_task->prio < rq->curr->prio)) {
		resched_task(rq->curr);
S
Steven Rostedt 已提交
1400 1401 1402
		return 0;
	}

1403
	/* We might release rq lock */
S
Steven Rostedt 已提交
1404 1405 1406
	get_task_struct(next_task);

	/* find_lock_lowest_rq locks the rq if found */
1407
	lowest_rq = find_lock_lowest_rq(next_task, rq);
S
Steven Rostedt 已提交
1408 1409 1410
	if (!lowest_rq) {
		struct task_struct *task;
		/*
1411
		 * find_lock_lowest_rq releases rq->lock
1412 1413 1414 1415 1416
		 * so it is possible that next_task has migrated.
		 *
		 * We need to make sure that the task is still on the same
		 * run-queue and is also still the next task eligible for
		 * pushing.
S
Steven Rostedt 已提交
1417
		 */
1418
		task = pick_next_pushable_task(rq);
1419 1420
		if (task_cpu(next_task) == rq->cpu && task == next_task) {
			/*
1421 1422 1423 1424
			 * The task hasn't migrated, and is still the next
			 * eligible task, but we failed to find a run-queue
			 * to push it to.  Do not retry in this case, since
			 * other cpus will pull from us when ready.
1425 1426
			 */
			goto out;
S
Steven Rostedt 已提交
1427
		}
1428

1429 1430 1431 1432
		if (!task)
			/* No more tasks, just exit */
			goto out;

1433
		/*
1434
		 * Something has shifted, try again.
1435
		 */
1436 1437 1438
		put_task_struct(next_task);
		next_task = task;
		goto retry;
S
Steven Rostedt 已提交
1439 1440
	}

1441
	deactivate_task(rq, next_task, 0);
S
Steven Rostedt 已提交
1442 1443
	set_task_cpu(next_task, lowest_rq->cpu);
	activate_task(lowest_rq, next_task, 0);
1444
	ret = 1;
S
Steven Rostedt 已提交
1445 1446 1447

	resched_task(lowest_rq->curr);

1448
	double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1449 1450 1451 1452

out:
	put_task_struct(next_task);

1453
	return ret;
S
Steven Rostedt 已提交
1454 1455 1456 1457 1458 1459 1460 1461 1462
}

static void push_rt_tasks(struct rq *rq)
{
	/* push_rt_task will return true if it moved an RT */
	while (push_rt_task(rq))
		;
}

1463 1464
static int pull_rt_task(struct rq *this_rq)
{
I
Ingo Molnar 已提交
1465
	int this_cpu = this_rq->cpu, ret = 0, cpu;
1466
	struct task_struct *p;
1467 1468
	struct rq *src_rq;

1469
	if (likely(!rt_overloaded(this_rq)))
1470 1471
		return 0;

1472
	for_each_cpu(cpu, this_rq->rd->rto_mask) {
1473 1474 1475 1476
		if (this_cpu == cpu)
			continue;

		src_rq = cpu_rq(cpu);
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488

		/*
		 * Don't bother taking the src_rq->lock if the next highest
		 * task is known to be lower-priority than our current task.
		 * This may look racy, but if this value is about to go
		 * logically higher, the src_rq will push this task away.
		 * And if its going logically lower, we do not care
		 */
		if (src_rq->rt.highest_prio.next >=
		    this_rq->rt.highest_prio.curr)
			continue;

1489 1490 1491
		/*
		 * We can potentially drop this_rq's lock in
		 * double_lock_balance, and another CPU could
1492
		 * alter this_rq
1493
		 */
1494
		double_lock_balance(this_rq, src_rq);
1495 1496 1497 1498

		/*
		 * Are there still pullable RT tasks?
		 */
M
Mike Galbraith 已提交
1499 1500
		if (src_rq->rt.rt_nr_running <= 1)
			goto skip;
1501 1502 1503 1504 1505 1506 1507

		p = pick_next_highest_task_rt(src_rq, this_cpu);

		/*
		 * Do we have an RT task that preempts
		 * the to-be-scheduled task?
		 */
1508
		if (p && (p->prio < this_rq->rt.highest_prio.curr)) {
1509
			WARN_ON(p == src_rq->curr);
P
Peter Zijlstra 已提交
1510
			WARN_ON(!p->on_rq);
1511 1512 1513 1514 1515 1516 1517

			/*
			 * There's a chance that p is higher in priority
			 * than what's currently running on its cpu.
			 * This is just that p is wakeing up and hasn't
			 * had a chance to schedule. We only pull
			 * p if it is lower in priority than the
1518
			 * current task on the run queue
1519
			 */
1520
			if (p->prio < src_rq->curr->prio)
M
Mike Galbraith 已提交
1521
				goto skip;
1522 1523 1524 1525 1526 1527 1528 1529 1530

			ret = 1;

			deactivate_task(src_rq, p, 0);
			set_task_cpu(p, this_cpu);
			activate_task(this_rq, p, 0);
			/*
			 * We continue with the search, just in
			 * case there's an even higher prio task
L
Lucas De Marchi 已提交
1531
			 * in another runqueue. (low likelihood
1532 1533 1534
			 * but possible)
			 */
		}
P
Peter Zijlstra 已提交
1535
skip:
1536
		double_unlock_balance(this_rq, src_rq);
1537 1538 1539 1540 1541
	}

	return ret;
}

1542
static void pre_schedule_rt(struct rq *rq, struct task_struct *prev)
1543 1544
{
	/* Try to pull RT tasks here if we lower this rq's prio */
Y
Yong Zhang 已提交
1545
	if (rq->rt.highest_prio.curr > prev->prio)
1546 1547 1548
		pull_rt_task(rq);
}

1549
static void post_schedule_rt(struct rq *rq)
S
Steven Rostedt 已提交
1550
{
1551
	push_rt_tasks(rq);
S
Steven Rostedt 已提交
1552 1553
}

1554 1555 1556 1557
/*
 * If we are not running and we are not going to reschedule soon, we should
 * try to push tasks away now
 */
1558
static void task_woken_rt(struct rq *rq, struct task_struct *p)
1559
{
1560
	if (!task_running(rq, p) &&
1561
	    !test_tsk_need_resched(rq->curr) &&
1562
	    has_pushable_tasks(rq) &&
1563
	    p->rt.nr_cpus_allowed > 1 &&
1564
	    rt_task(rq->curr) &&
1565
	    (rq->curr->rt.nr_cpus_allowed < 2 ||
1566
	     rq->curr->prio <= p->prio))
1567 1568 1569
		push_rt_tasks(rq);
}

1570
static void set_cpus_allowed_rt(struct task_struct *p,
1571
				const struct cpumask *new_mask)
1572
{
1573
	int weight = cpumask_weight(new_mask);
1574 1575 1576 1577 1578 1579 1580

	BUG_ON(!rt_task(p));

	/*
	 * Update the migration status of the RQ if we have an RT task
	 * which is running AND changing its weight value.
	 */
P
Peter Zijlstra 已提交
1581
	if (p->on_rq && (weight != p->rt.nr_cpus_allowed)) {
1582 1583
		struct rq *rq = task_rq(p);

1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
		if (!task_current(rq, p)) {
			/*
			 * Make sure we dequeue this task from the pushable list
			 * before going further.  It will either remain off of
			 * the list because we are no longer pushable, or it
			 * will be requeued.
			 */
			if (p->rt.nr_cpus_allowed > 1)
				dequeue_pushable_task(rq, p);

			/*
			 * Requeue if our weight is changing and still > 1
			 */
			if (weight > 1)
				enqueue_pushable_task(rq, p);

		}

P
Peter Zijlstra 已提交
1602
		if ((p->rt.nr_cpus_allowed <= 1) && (weight > 1)) {
1603
			rq->rt.rt_nr_migratory++;
P
Peter Zijlstra 已提交
1604
		} else if ((p->rt.nr_cpus_allowed > 1) && (weight <= 1)) {
1605 1606 1607 1608
			BUG_ON(!rq->rt.rt_nr_migratory);
			rq->rt.rt_nr_migratory--;
		}

1609
		update_rt_migration(&rq->rt);
1610 1611
	}
}
1612

1613
/* Assumes rq->lock is held */
1614
static void rq_online_rt(struct rq *rq)
1615 1616 1617
{
	if (rq->rt.overloaded)
		rt_set_overload(rq);
1618

P
Peter Zijlstra 已提交
1619 1620
	__enable_runtime(rq);

1621
	cpupri_set(&rq->rd->cpupri, rq->cpu, rq->rt.highest_prio.curr);
1622 1623 1624
}

/* Assumes rq->lock is held */
1625
static void rq_offline_rt(struct rq *rq)
1626 1627 1628
{
	if (rq->rt.overloaded)
		rt_clear_overload(rq);
1629

P
Peter Zijlstra 已提交
1630 1631
	__disable_runtime(rq);

1632
	cpupri_set(&rq->rd->cpupri, rq->cpu, CPUPRI_INVALID);
1633
}
1634 1635 1636 1637 1638

/*
 * When switch from the rt queue, we bring ourselves to a position
 * that we might want to pull RT tasks from other runqueues.
 */
P
Peter Zijlstra 已提交
1639
static void switched_from_rt(struct rq *rq, struct task_struct *p)
1640 1641 1642 1643 1644 1645 1646 1647
{
	/*
	 * If there are other RT tasks then we will reschedule
	 * and the scheduling of the other RT tasks will handle
	 * the balancing. But if we are the last RT task
	 * we may need to handle the pulling of RT tasks
	 * now.
	 */
P
Peter Zijlstra 已提交
1648
	if (p->on_rq && !rq->rt.rt_nr_running)
1649 1650
		pull_rt_task(rq);
}
1651 1652 1653 1654 1655 1656

static inline void init_sched_rt_class(void)
{
	unsigned int i;

	for_each_possible_cpu(i)
1657
		zalloc_cpumask_var_node(&per_cpu(local_cpu_mask, i),
1658
					GFP_KERNEL, cpu_to_node(i));
1659
}
1660 1661 1662 1663 1664 1665 1666
#endif /* CONFIG_SMP */

/*
 * When switching a task to RT, we may overload the runqueue
 * with RT tasks. In this case we try to push them off to
 * other runqueues.
 */
P
Peter Zijlstra 已提交
1667
static void switched_to_rt(struct rq *rq, struct task_struct *p)
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677
{
	int check_resched = 1;

	/*
	 * If we are already running, then there's nothing
	 * that needs to be done. But if we are not running
	 * we may need to preempt the current running task.
	 * If that current running task is also an RT task
	 * then see if we can move to another run queue.
	 */
P
Peter Zijlstra 已提交
1678
	if (p->on_rq && rq->curr != p) {
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
#ifdef CONFIG_SMP
		if (rq->rt.overloaded && push_rt_task(rq) &&
		    /* Don't resched if we changed runqueues */
		    rq != task_rq(p))
			check_resched = 0;
#endif /* CONFIG_SMP */
		if (check_resched && p->prio < rq->curr->prio)
			resched_task(rq->curr);
	}
}

/*
 * Priority of the task has changed. This may cause
 * us to initiate a push or pull.
 */
P
Peter Zijlstra 已提交
1694 1695
static void
prio_changed_rt(struct rq *rq, struct task_struct *p, int oldprio)
1696
{
P
Peter Zijlstra 已提交
1697
	if (!p->on_rq)
P
Peter Zijlstra 已提交
1698 1699 1700
		return;

	if (rq->curr == p) {
1701 1702 1703 1704 1705 1706 1707 1708 1709
#ifdef CONFIG_SMP
		/*
		 * If our priority decreases while running, we
		 * may need to pull tasks to this runqueue.
		 */
		if (oldprio < p->prio)
			pull_rt_task(rq);
		/*
		 * If there's a higher priority task waiting to run
1710 1711 1712
		 * then reschedule. Note, the above pull_rt_task
		 * can release the rq lock and p could migrate.
		 * Only reschedule if p is still on the same runqueue.
1713
		 */
1714
		if (p->prio > rq->rt.highest_prio.curr && rq->curr == p)
1715 1716 1717 1718 1719
			resched_task(p);
#else
		/* For UP simply resched on drop of prio */
		if (oldprio < p->prio)
			resched_task(p);
S
Steven Rostedt 已提交
1720
#endif /* CONFIG_SMP */
1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731
	} else {
		/*
		 * This task is not running, but if it is
		 * greater than the current running task
		 * then reschedule.
		 */
		if (p->prio < rq->curr->prio)
			resched_task(rq->curr);
	}
}

1732 1733 1734 1735
static void watchdog(struct rq *rq, struct task_struct *p)
{
	unsigned long soft, hard;

1736 1737 1738
	/* max may change after cur was read, this will be fixed next tick */
	soft = task_rlimit(p, RLIMIT_RTTIME);
	hard = task_rlimit_max(p, RLIMIT_RTTIME);
1739 1740 1741 1742 1743 1744

	if (soft != RLIM_INFINITY) {
		unsigned long next;

		p->rt.timeout++;
		next = DIV_ROUND_UP(min(soft, hard), USEC_PER_SEC/HZ);
1745
		if (p->rt.timeout > next)
1746
			p->cputime_expires.sched_exp = p->se.sum_exec_runtime;
1747 1748
	}
}
I
Ingo Molnar 已提交
1749

P
Peter Zijlstra 已提交
1750
static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued)
I
Ingo Molnar 已提交
1751
{
1752 1753
	update_curr_rt(rq);

1754 1755
	watchdog(rq, p);

I
Ingo Molnar 已提交
1756 1757 1758 1759 1760 1761 1762
	/*
	 * RR tasks need a special form of timeslice management.
	 * FIFO tasks have no timeslices.
	 */
	if (p->policy != SCHED_RR)
		return;

P
Peter Zijlstra 已提交
1763
	if (--p->rt.time_slice)
I
Ingo Molnar 已提交
1764 1765
		return;

P
Peter Zijlstra 已提交
1766
	p->rt.time_slice = DEF_TIMESLICE;
I
Ingo Molnar 已提交
1767

1768 1769 1770 1771
	/*
	 * Requeue to the end of queue if we are not the only element
	 * on the queue:
	 */
P
Peter Zijlstra 已提交
1772
	if (p->rt.run_list.prev != p->rt.run_list.next) {
1773
		requeue_task_rt(rq, p, 0);
1774 1775
		set_tsk_need_resched(p);
	}
I
Ingo Molnar 已提交
1776 1777
}

1778 1779 1780 1781
static void set_curr_task_rt(struct rq *rq)
{
	struct task_struct *p = rq->curr;

1782
	p->se.exec_start = rq->clock_task;
1783 1784 1785

	/* The running task is never eligible for pushing */
	dequeue_pushable_task(rq, p);
1786 1787
}

1788
static unsigned int get_rr_interval_rt(struct rq *rq, struct task_struct *task)
1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
{
	/*
	 * Time slice is 0 for SCHED_FIFO tasks
	 */
	if (task->policy == SCHED_RR)
		return DEF_TIMESLICE;
	else
		return 0;
}

1799
static const struct sched_class rt_sched_class = {
1800
	.next			= &fair_sched_class,
I
Ingo Molnar 已提交
1801 1802 1803 1804 1805 1806 1807 1808 1809
	.enqueue_task		= enqueue_task_rt,
	.dequeue_task		= dequeue_task_rt,
	.yield_task		= yield_task_rt,

	.check_preempt_curr	= check_preempt_curr_rt,

	.pick_next_task		= pick_next_task_rt,
	.put_prev_task		= put_prev_task_rt,

1810
#ifdef CONFIG_SMP
L
Li Zefan 已提交
1811 1812
	.select_task_rq		= select_task_rq_rt,

1813
	.set_cpus_allowed       = set_cpus_allowed_rt,
1814 1815
	.rq_online              = rq_online_rt,
	.rq_offline             = rq_offline_rt,
1816 1817
	.pre_schedule		= pre_schedule_rt,
	.post_schedule		= post_schedule_rt,
1818
	.task_woken		= task_woken_rt,
1819
	.switched_from		= switched_from_rt,
1820
#endif
I
Ingo Molnar 已提交
1821

1822
	.set_curr_task          = set_curr_task_rt,
I
Ingo Molnar 已提交
1823
	.task_tick		= task_tick_rt,
1824

1825 1826
	.get_rr_interval	= get_rr_interval_rt,

1827 1828
	.prio_changed		= prio_changed_rt,
	.switched_to		= switched_to_rt,
I
Ingo Molnar 已提交
1829
};
1830 1831 1832 1833 1834 1835

#ifdef CONFIG_SCHED_DEBUG
extern void print_rt_rq(struct seq_file *m, int cpu, struct rt_rq *rt_rq);

static void print_rt_stats(struct seq_file *m, int cpu)
{
C
Cheng Xu 已提交
1836
	rt_rq_iter_t iter;
1837 1838 1839
	struct rt_rq *rt_rq;

	rcu_read_lock();
C
Cheng Xu 已提交
1840
	for_each_rt_rq(rt_rq, iter, cpu_rq(cpu))
1841 1842 1843
		print_rt_rq(m, cpu, rt_rq);
	rcu_read_unlock();
}
1844
#endif /* CONFIG_SCHED_DEBUG */