rt.c 46.4 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 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
#include "sched.h"

#include <linux/slab.h>

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

struct rt_bandwidth def_rt_bandwidth;

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;
}

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;

	raw_spin_lock_init(&rt_b->rt_runtime_lock);

	hrtimer_init(&rt_b->rt_period_timer,
			CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rt_b->rt_period_timer.function = sched_rt_period_timer;
}

static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
{
	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
		return;

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

	raw_spin_lock(&rt_b->rt_runtime_lock);
	start_bandwidth_timer(&rt_b->rt_period_timer, rt_b->rt_period);
	raw_spin_unlock(&rt_b->rt_runtime_lock);
}

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);

#if defined CONFIG_SMP
	rt_rq->highest_prio.curr = MAX_RT_PRIO;
	rt_rq->highest_prio.next = MAX_RT_PRIO;
	rt_rq->rt_nr_migratory = 0;
	rt_rq->overloaded = 0;
	plist_head_init(&rt_rq->pushable_tasks);
#endif

	rt_rq->rt_time = 0;
	rt_rq->rt_throttled = 0;
	rt_rq->rt_runtime = 0;
	raw_spin_lock_init(&rt_rq->rt_runtime_lock);
}

87
#ifdef CONFIG_RT_GROUP_SCHED
88 89 90 91
static void destroy_rt_bandwidth(struct rt_bandwidth *rt_b)
{
	hrtimer_cancel(&rt_b->rt_period_timer);
}
92 93 94

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

95 96
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
97 98 99
#ifdef CONFIG_SCHED_DEBUG
	WARN_ON_ONCE(!rt_entity_is_task(rt_se));
#endif
100 101 102 103 104 105 106 107 108 109 110 111 112
	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;
}

113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
void free_rt_sched_group(struct task_group *tg)
{
	int i;

	if (tg->rt_se)
		destroy_rt_bandwidth(&tg->rt_bandwidth);

	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);
}

void init_tg_rt_entry(struct task_group *tg, struct rt_rq *rt_rq,
		struct sched_rt_entity *rt_se, int cpu,
		struct sched_rt_entity *parent)
{
	struct rq *rq = cpu_rq(cpu);

	rt_rq->highest_prio.curr = MAX_RT_PRIO;
	rt_rq->rt_nr_boosted = 0;
	rt_rq->rq = rq;
	rt_rq->tg = tg;

	tg->rt_rq[cpu] = rt_rq;
	tg->rt_se[cpu] = rt_se;

	if (!rt_se)
		return;

	if (!parent)
		rt_se->rt_rq = &rq->rt;
	else
		rt_se->rt_rq = parent->my_q;

	rt_se->my_q = rt_rq;
	rt_se->parent = parent;
	INIT_LIST_HEAD(&rt_se->run_list);
}

int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
{
	struct rt_rq *rt_rq;
	struct sched_rt_entity *rt_se;
	int i;

	tg->rt_rq = kzalloc(sizeof(rt_rq) * nr_cpu_ids, GFP_KERNEL);
	if (!tg->rt_rq)
		goto err;
	tg->rt_se = kzalloc(sizeof(rt_se) * nr_cpu_ids, GFP_KERNEL);
	if (!tg->rt_se)
		goto err;

	init_rt_bandwidth(&tg->rt_bandwidth,
			ktime_to_ns(def_rt_bandwidth.rt_period), 0);

	for_each_possible_cpu(i) {
		rt_rq = kzalloc_node(sizeof(struct rt_rq),
				     GFP_KERNEL, cpu_to_node(i));
		if (!rt_rq)
			goto err;

		rt_se = kzalloc_node(sizeof(struct sched_rt_entity),
				     GFP_KERNEL, cpu_to_node(i));
		if (!rt_se)
			goto err_free_rq;

		init_rt_rq(rt_rq, cpu_rq(i));
		rt_rq->rt_runtime = tg->rt_bandwidth.rt_runtime;
		init_tg_rt_entry(tg, rt_rq, rt_se, i, parent->rt_se[i]);
	}

	return 1;

err_free_rq:
	kfree(rt_rq);
err:
	return 0;
}

198 199
#else /* CONFIG_RT_GROUP_SCHED */

200 201
#define rt_entity_is_task(rt_se) (1)

202 203 204 205 206
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
	return container_of(rt_se, struct task_struct, rt);
}

207 208 209 210 211 212 213 214 215 216 217 218 219
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;
}

220 221 222 223 224 225
void free_rt_sched_group(struct task_group *tg) { }

int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
{
	return 1;
}
226 227
#endif /* CONFIG_RT_GROUP_SCHED */

S
Steven Rostedt 已提交
228
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
229

230
static inline int rt_overloaded(struct rq *rq)
S
Steven Rostedt 已提交
231
{
232
	return atomic_read(&rq->rd->rto_count);
S
Steven Rostedt 已提交
233
}
I
Ingo Molnar 已提交
234

S
Steven Rostedt 已提交
235 236
static inline void rt_set_overload(struct rq *rq)
{
237 238 239
	if (!rq->online)
		return;

240
	cpumask_set_cpu(rq->cpu, rq->rd->rto_mask);
S
Steven Rostedt 已提交
241 242 243 244 245 246 247 248
	/*
	 * 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();
249
	atomic_inc(&rq->rd->rto_count);
S
Steven Rostedt 已提交
250
}
I
Ingo Molnar 已提交
251

S
Steven Rostedt 已提交
252 253
static inline void rt_clear_overload(struct rq *rq)
{
254 255 256
	if (!rq->online)
		return;

S
Steven Rostedt 已提交
257
	/* the order here really doesn't matter */
258
	atomic_dec(&rq->rd->rto_count);
259
	cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask);
S
Steven Rostedt 已提交
260
}
261

262
static void update_rt_migration(struct rt_rq *rt_rq)
263
{
264
	if (rt_rq->rt_nr_migratory && rt_rq->rt_nr_total > 1) {
265 266 267
		if (!rt_rq->overloaded) {
			rt_set_overload(rq_of_rt_rq(rt_rq));
			rt_rq->overloaded = 1;
268
		}
269 270 271
	} else if (rt_rq->overloaded) {
		rt_clear_overload(rq_of_rt_rq(rt_rq));
		rt_rq->overloaded = 0;
272
	}
273
}
S
Steven Rostedt 已提交
274

275 276
static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
277 278 279 280 281 282
	if (!rt_entity_is_task(rt_se))
		return;

	rt_rq = &rq_of_rt_rq(rt_rq)->rt;

	rt_rq->rt_nr_total++;
283 284 285 286 287 288 289 290
	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)
{
291 292 293 294 295 296
	if (!rt_entity_is_task(rt_se))
		return;

	rt_rq = &rq_of_rt_rq(rt_rq)->rt;

	rt_rq->rt_nr_total--;
297 298 299 300 301 302
	if (rt_se->nr_cpus_allowed > 1)
		rt_rq->rt_nr_migratory--;

	update_rt_migration(rt_rq);
}

303 304 305 306 307
static inline int has_pushable_tasks(struct rq *rq)
{
	return !plist_head_empty(&rq->rt.pushable_tasks);
}

308 309 310 311 312
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);
313 314 315 316

	/* Update the highest prio pushable task */
	if (p->prio < rq->rt.highest_prio.next)
		rq->rt.highest_prio.next = p->prio;
317 318 319 320 321 322
}

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

323 324 325 326 327 328 329
	/* 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;
330 331
}

332 333
#else

334
static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
P
Peter Zijlstra 已提交
335
{
P
Peter Zijlstra 已提交
336 337
}

338 339 340 341
static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p)
{
}

342
static inline
343 344 345 346
void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}

347
static inline
348 349 350
void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}
351

S
Steven Rostedt 已提交
352 353
#endif /* CONFIG_SMP */

P
Peter Zijlstra 已提交
354 355 356 357 358
static inline int on_rt_rq(struct sched_rt_entity *rt_se)
{
	return !list_empty(&rt_se->run_list);
}

359
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
360

P
Peter Zijlstra 已提交
361
static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
362 363
{
	if (!rt_rq->tg)
P
Peter Zijlstra 已提交
364
		return RUNTIME_INF;
P
Peter Zijlstra 已提交
365

P
Peter Zijlstra 已提交
366 367 368 369 370 371
	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
372 373
}

C
Cheng Xu 已提交
374 375
typedef struct task_group *rt_rq_iter_t;

376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
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
393

394 395 396 397 398 399 400 401 402 403 404
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
405
#define for_each_leaf_rt_rq(rt_rq, rq) \
406
	list_for_each_entry_rcu(rt_rq, &rq->leaf_rt_rq_list, leaf_rt_rq_list)
P
Peter Zijlstra 已提交
407 408 409 410 411 412 413 414 415

#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;
}

416
static void enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head);
P
Peter Zijlstra 已提交
417 418
static void dequeue_rt_entity(struct sched_rt_entity *rt_se);

P
Peter Zijlstra 已提交
419
static void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
420
{
421
	struct task_struct *curr = rq_of_rt_rq(rt_rq)->curr;
422 423
	struct sched_rt_entity *rt_se;

424 425 426
	int cpu = cpu_of(rq_of_rt_rq(rt_rq));

	rt_se = rt_rq->tg->rt_se[cpu];
P
Peter Zijlstra 已提交
427

428 429
	if (rt_rq->rt_nr_running) {
		if (rt_se && !on_rt_rq(rt_se))
430
			enqueue_rt_entity(rt_se, false);
431
		if (rt_rq->highest_prio.curr < curr->prio)
432
			resched_task(curr);
P
Peter Zijlstra 已提交
433 434 435
	}
}

P
Peter Zijlstra 已提交
436
static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
437
{
438
	struct sched_rt_entity *rt_se;
439
	int cpu = cpu_of(rq_of_rt_rq(rt_rq));
440

441
	rt_se = rt_rq->tg->rt_se[cpu];
P
Peter Zijlstra 已提交
442 443 444 445 446

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

P
Peter Zijlstra 已提交
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
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;
}

464
#ifdef CONFIG_SMP
465
static inline const struct cpumask *sched_rt_period_mask(void)
466 467 468
{
	return cpu_rq(smp_processor_id())->rd->span;
}
P
Peter Zijlstra 已提交
469
#else
470
static inline const struct cpumask *sched_rt_period_mask(void)
471
{
472
	return cpu_online_mask;
473 474
}
#endif
P
Peter Zijlstra 已提交
475

476 477
static inline
struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu)
P
Peter Zijlstra 已提交
478
{
479 480
	return container_of(rt_b, struct task_group, rt_bandwidth)->rt_rq[cpu];
}
P
Peter Zijlstra 已提交
481

P
Peter Zijlstra 已提交
482 483 484 485 486
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &rt_rq->tg->rt_bandwidth;
}

487
#else /* !CONFIG_RT_GROUP_SCHED */
488 489 490

static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
{
P
Peter Zijlstra 已提交
491 492 493 494 495 496
	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
497 498
}

C
Cheng Xu 已提交
499 500 501 502 503
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)

504 505 506 507 508 509 510 511
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
512 513 514 515 516 517 518 519 520 521 522
#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
523
static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
524
{
525 526
	if (rt_rq->rt_nr_running)
		resched_task(rq_of_rt_rq(rt_rq)->curr);
P
Peter Zijlstra 已提交
527 528
}

P
Peter Zijlstra 已提交
529
static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
530 531 532
{
}

P
Peter Zijlstra 已提交
533 534 535 536
static inline int rt_rq_throttled(struct rt_rq *rt_rq)
{
	return rt_rq->rt_throttled;
}
537

538
static inline const struct cpumask *sched_rt_period_mask(void)
539
{
540
	return cpu_online_mask;
541 542 543 544 545 546 547 548
}

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
549 550 551 552 553
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &def_rt_bandwidth;
}

554
#endif /* CONFIG_RT_GROUP_SCHED */
555

P
Peter Zijlstra 已提交
556
#ifdef CONFIG_SMP
557 558 559
/*
 * We ran out of runtime, see if we can borrow some from our neighbours.
 */
560
static int do_balance_runtime(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
561 562 563 564 565 566
{
	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;

567
	weight = cpumask_weight(rd->span);
P
Peter Zijlstra 已提交
568

569
	raw_spin_lock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
570
	rt_period = ktime_to_ns(rt_b->rt_period);
571
	for_each_cpu(i, rd->span) {
P
Peter Zijlstra 已提交
572 573 574 575 576 577
		struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
		s64 diff;

		if (iter == rt_rq)
			continue;

578
		raw_spin_lock(&iter->rt_runtime_lock);
579 580 581 582 583
		/*
		 * 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
584 585 586
		if (iter->rt_runtime == RUNTIME_INF)
			goto next;

587 588 589 590
		/*
		 * 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
591 592
		diff = iter->rt_runtime - iter->rt_time;
		if (diff > 0) {
593
			diff = div_u64((u64)diff, weight);
P
Peter Zijlstra 已提交
594 595 596 597 598 599
			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) {
600
				raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
601 602 603
				break;
			}
		}
P
Peter Zijlstra 已提交
604
next:
605
		raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
606
	}
607
	raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
608 609 610

	return more;
}
P
Peter Zijlstra 已提交
611

612 613 614
/*
 * Ensure this RQ takes back all the runtime it lend to its neighbours.
 */
P
Peter Zijlstra 已提交
615 616 617
static void __disable_runtime(struct rq *rq)
{
	struct root_domain *rd = rq->rd;
C
Cheng Xu 已提交
618
	rt_rq_iter_t iter;
P
Peter Zijlstra 已提交
619 620 621 622 623
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

C
Cheng Xu 已提交
624
	for_each_rt_rq(rt_rq, iter, rq) {
P
Peter Zijlstra 已提交
625 626 627 628
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);
		s64 want;
		int i;

629 630
		raw_spin_lock(&rt_b->rt_runtime_lock);
		raw_spin_lock(&rt_rq->rt_runtime_lock);
631 632 633 634 635
		/*
		 * 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
636 637 638
		if (rt_rq->rt_runtime == RUNTIME_INF ||
				rt_rq->rt_runtime == rt_b->rt_runtime)
			goto balanced;
639
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
640

641 642 643 644 645
		/*
		 * 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
646 647
		want = rt_b->rt_runtime - rt_rq->rt_runtime;

648 649 650
		/*
		 * Greedy reclaim, take back as much as we can.
		 */
651
		for_each_cpu(i, rd->span) {
P
Peter Zijlstra 已提交
652 653 654
			struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
			s64 diff;

655 656 657
			/*
			 * Can't reclaim from ourselves or disabled runqueues.
			 */
658
			if (iter == rt_rq || iter->rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
659 660
				continue;

661
			raw_spin_lock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
662 663 664 665 666 667 668 669
			if (want > 0) {
				diff = min_t(s64, iter->rt_runtime, want);
				iter->rt_runtime -= diff;
				want -= diff;
			} else {
				iter->rt_runtime -= want;
				want -= want;
			}
670
			raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
671 672 673 674 675

			if (!want)
				break;
		}

676
		raw_spin_lock(&rt_rq->rt_runtime_lock);
677 678 679 680
		/*
		 * We cannot be left wanting - that would mean some runtime
		 * leaked out of the system.
		 */
P
Peter Zijlstra 已提交
681 682
		BUG_ON(want);
balanced:
683 684 685 686
		/*
		 * 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
687
		rt_rq->rt_runtime = RUNTIME_INF;
688 689
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
		raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
690 691 692 693 694 695 696
	}
}

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

697
	raw_spin_lock_irqsave(&rq->lock, flags);
P
Peter Zijlstra 已提交
698
	__disable_runtime(rq);
699
	raw_spin_unlock_irqrestore(&rq->lock, flags);
P
Peter Zijlstra 已提交
700 701 702 703
}

static void __enable_runtime(struct rq *rq)
{
C
Cheng Xu 已提交
704
	rt_rq_iter_t iter;
P
Peter Zijlstra 已提交
705 706 707 708 709
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

710 711 712
	/*
	 * Reset each runqueue's bandwidth settings
	 */
C
Cheng Xu 已提交
713
	for_each_rt_rq(rt_rq, iter, rq) {
P
Peter Zijlstra 已提交
714 715
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);

716 717
		raw_spin_lock(&rt_b->rt_runtime_lock);
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
718 719
		rt_rq->rt_runtime = rt_b->rt_runtime;
		rt_rq->rt_time = 0;
720
		rt_rq->rt_throttled = 0;
721 722
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
		raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
723 724 725 726 727 728 729
	}
}

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

730
	raw_spin_lock_irqsave(&rq->lock, flags);
P
Peter Zijlstra 已提交
731
	__enable_runtime(rq);
732
	raw_spin_unlock_irqrestore(&rq->lock, flags);
P
Peter Zijlstra 已提交
733 734
}

735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756
int update_runtime(struct notifier_block *nfb, unsigned long action, void *hcpu)
{
	int cpu = (int)(long)hcpu;

	switch (action) {
	case CPU_DOWN_PREPARE:
	case CPU_DOWN_PREPARE_FROZEN:
		disable_runtime(cpu_rq(cpu));
		return NOTIFY_OK;

	case CPU_DOWN_FAILED:
	case CPU_DOWN_FAILED_FROZEN:
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
		enable_runtime(cpu_rq(cpu));
		return NOTIFY_OK;

	default:
		return NOTIFY_DONE;
	}
}

757 758 759 760
static int balance_runtime(struct rt_rq *rt_rq)
{
	int more = 0;

761 762 763
	if (!sched_feat(RT_RUNTIME_SHARE))
		return more;

764
	if (rt_rq->rt_time > rt_rq->rt_runtime) {
765
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
766
		more = do_balance_runtime(rt_rq);
767
		raw_spin_lock(&rt_rq->rt_runtime_lock);
768 769 770 771
	}

	return more;
}
772
#else /* !CONFIG_SMP */
773 774 775 776
static inline int balance_runtime(struct rt_rq *rt_rq)
{
	return 0;
}
777
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
778

779 780
static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
{
781
	int i, idle = 1, throttled = 0;
782
	const struct cpumask *span;
783 784

	span = sched_rt_period_mask();
785
	for_each_cpu(i, span) {
786 787 788 789
		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);

790
		raw_spin_lock(&rq->lock);
791 792 793
		if (rt_rq->rt_time) {
			u64 runtime;

794
			raw_spin_lock(&rt_rq->rt_runtime_lock);
795 796 797 798 799 800 801
			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;
802 803 804 805 806 807 808

				/*
				 * 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;
809 810 811
			}
			if (rt_rq->rt_time || rt_rq->rt_nr_running)
				idle = 0;
812
			raw_spin_unlock(&rt_rq->rt_runtime_lock);
813
		} else if (rt_rq->rt_nr_running) {
814
			idle = 0;
815 816 817
			if (!rt_rq_throttled(rt_rq))
				enqueue = 1;
		}
818 819
		if (rt_rq->rt_throttled)
			throttled = 1;
820 821 822

		if (enqueue)
			sched_rt_rq_enqueue(rt_rq);
823
		raw_spin_unlock(&rq->lock);
824 825
	}

826 827 828
	if (!throttled && (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF))
		return 1;

829 830
	return idle;
}
P
Peter Zijlstra 已提交
831

P
Peter Zijlstra 已提交
832 833
static inline int rt_se_prio(struct sched_rt_entity *rt_se)
{
834
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
835 836 837
	struct rt_rq *rt_rq = group_rt_rq(rt_se);

	if (rt_rq)
838
		return rt_rq->highest_prio.curr;
P
Peter Zijlstra 已提交
839 840 841 842 843
#endif

	return rt_task_of(rt_se)->prio;
}

P
Peter Zijlstra 已提交
844
static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
845
{
P
Peter Zijlstra 已提交
846
	u64 runtime = sched_rt_runtime(rt_rq);
P
Peter Zijlstra 已提交
847 848

	if (rt_rq->rt_throttled)
P
Peter Zijlstra 已提交
849
		return rt_rq_throttled(rt_rq);
P
Peter Zijlstra 已提交
850

851
	if (runtime >= sched_rt_period(rt_rq))
P
Peter Zijlstra 已提交
852 853
		return 0;

854 855 856 857
	balance_runtime(rt_rq);
	runtime = sched_rt_runtime(rt_rq);
	if (runtime == RUNTIME_INF)
		return 0;
P
Peter Zijlstra 已提交
858

P
Peter Zijlstra 已提交
859
	if (rt_rq->rt_time > runtime) {
860 861 862 863 864 865 866
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);

		/*
		 * Don't actually throttle groups that have no runtime assigned
		 * but accrue some time due to boosting.
		 */
		if (likely(rt_b->rt_runtime)) {
867 868
			static bool once = false;

869
			rt_rq->rt_throttled = 1;
870 871 872 873 874

			if (!once) {
				once = true;
				printk_sched("sched: RT throttling activated\n");
			}
875 876 877 878 879 880 881 882 883
		} else {
			/*
			 * In case we did anyway, make it go away,
			 * replenishment is a joke, since it will replenish us
			 * with exactly 0 ns.
			 */
			rt_rq->rt_time = 0;
		}

P
Peter Zijlstra 已提交
884
		if (rt_rq_throttled(rt_rq)) {
P
Peter Zijlstra 已提交
885
			sched_rt_rq_dequeue(rt_rq);
P
Peter Zijlstra 已提交
886 887
			return 1;
		}
P
Peter Zijlstra 已提交
888 889 890 891 892
	}

	return 0;
}

I
Ingo Molnar 已提交
893 894 895 896
/*
 * Update the current task's runtime statistics. Skip current tasks that
 * are not in our scheduling class.
 */
A
Alexey Dobriyan 已提交
897
static void update_curr_rt(struct rq *rq)
I
Ingo Molnar 已提交
898 899
{
	struct task_struct *curr = rq->curr;
P
Peter Zijlstra 已提交
900 901
	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
902 903
	u64 delta_exec;

P
Peter Zijlstra 已提交
904
	if (curr->sched_class != &rt_sched_class)
I
Ingo Molnar 已提交
905 906
		return;

907
	delta_exec = rq->clock_task - curr->se.exec_start;
I
Ingo Molnar 已提交
908 909
	if (unlikely((s64)delta_exec < 0))
		delta_exec = 0;
I
Ingo Molnar 已提交
910

911 912
	schedstat_set(curr->se.statistics.exec_max,
		      max(curr->se.statistics.exec_max, delta_exec));
I
Ingo Molnar 已提交
913 914

	curr->se.sum_exec_runtime += delta_exec;
915 916
	account_group_exec_runtime(curr, delta_exec);

917
	curr->se.exec_start = rq->clock_task;
918
	cpuacct_charge(curr, delta_exec);
P
Peter Zijlstra 已提交
919

920 921
	sched_rt_avg_update(rq, delta_exec);

922 923 924
	if (!rt_bandwidth_enabled())
		return;

D
Dhaval Giani 已提交
925 926 927
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);

928
		if (sched_rt_runtime(rt_rq) != RUNTIME_INF) {
929
			raw_spin_lock(&rt_rq->rt_runtime_lock);
930 931 932
			rt_rq->rt_time += delta_exec;
			if (sched_rt_runtime_exceeded(rt_rq))
				resched_task(curr);
933
			raw_spin_unlock(&rt_rq->rt_runtime_lock);
934
		}
D
Dhaval Giani 已提交
935
	}
I
Ingo Molnar 已提交
936 937
}

938
#if defined CONFIG_SMP
939

940 941
static void
inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
942
{
G
Gregory Haskins 已提交
943
	struct rq *rq = rq_of_rt_rq(rt_rq);
944

945 946
	if (rq->online && prio < prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, prio);
947
}
948

949 950 951 952
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);
953

954 955
	if (rq->online && rt_rq->highest_prio.curr != prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr);
956 957
}

958 959
#else /* CONFIG_SMP */

P
Peter Zijlstra 已提交
960
static inline
961 962 963 964 965
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 */
966

967
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
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
984
	if (rt_rq->rt_nr_running) {
985

986
		WARN_ON(prio < prev_prio);
987

988
		/*
989 990
		 * This may have been our highest task, and therefore
		 * we may have some recomputation to do
991
		 */
992
		if (prio == prev_prio) {
993 994 995
			struct rt_prio_array *array = &rt_rq->active;

			rt_rq->highest_prio.curr =
996
				sched_find_first_bit(array->bitmap);
997 998
		}

999
	} else
1000
		rt_rq->highest_prio.curr = MAX_RT_PRIO;
1001

1002 1003
	dec_rt_prio_smp(rt_rq, prio, prev_prio);
}
1004

1005 1006 1007 1008 1009 1010
#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 */
1011

1012
#ifdef CONFIG_RT_GROUP_SCHED
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026

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
1027 1028 1029 1030
	if (rt_se_boosted(rt_se))
		rt_rq->rt_nr_boosted--;

	WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted);
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
}

#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);
1069 1070
}

1071
static void __enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head)
I
Ingo Molnar 已提交
1072
{
P
Peter Zijlstra 已提交
1073 1074 1075
	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);
1076
	struct list_head *queue = array->queue + rt_se_prio(rt_se);
I
Ingo Molnar 已提交
1077

1078 1079 1080 1081 1082 1083 1084
	/*
	 * 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
1085
		return;
1086

1087 1088 1089
	if (!rt_rq->rt_nr_running)
		list_add_leaf_rt_rq(rt_rq);

1090 1091 1092 1093
	if (head)
		list_add(&rt_se->run_list, queue);
	else
		list_add_tail(&rt_se->run_list, queue);
P
Peter Zijlstra 已提交
1094
	__set_bit(rt_se_prio(rt_se), array->bitmap);
1095

P
Peter Zijlstra 已提交
1096 1097 1098
	inc_rt_tasks(rt_se, rt_rq);
}

1099
static void __dequeue_rt_entity(struct sched_rt_entity *rt_se)
P
Peter Zijlstra 已提交
1100 1101 1102 1103 1104 1105 1106 1107 1108
{
	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);
1109 1110
	if (!rt_rq->rt_nr_running)
		list_del_leaf_rt_rq(rt_rq);
P
Peter Zijlstra 已提交
1111 1112 1113 1114 1115 1116
}

/*
 * Because the prio of an upper entry depends on the lower
 * entries, we must remove entries top - down.
 */
1117
static void dequeue_rt_stack(struct sched_rt_entity *rt_se)
P
Peter Zijlstra 已提交
1118
{
1119
	struct sched_rt_entity *back = NULL;
P
Peter Zijlstra 已提交
1120

1121 1122 1123 1124 1125 1126 1127
	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))
1128 1129 1130 1131
			__dequeue_rt_entity(rt_se);
	}
}

1132
static void enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head)
1133 1134 1135
{
	dequeue_rt_stack(rt_se);
	for_each_sched_rt_entity(rt_se)
1136
		__enqueue_rt_entity(rt_se, head);
1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
}

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)
1147
			__enqueue_rt_entity(rt_se, false);
1148
	}
I
Ingo Molnar 已提交
1149 1150 1151 1152 1153
}

/*
 * Adding/removing a task to/from a priority array:
 */
1154
static void
1155
enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags)
P
Peter Zijlstra 已提交
1156 1157 1158
{
	struct sched_rt_entity *rt_se = &p->rt;

1159
	if (flags & ENQUEUE_WAKEUP)
P
Peter Zijlstra 已提交
1160 1161
		rt_se->timeout = 0;

1162
	enqueue_rt_entity(rt_se, flags & ENQUEUE_HEAD);
1163

1164 1165
	if (!task_current(rq, p) && p->rt.nr_cpus_allowed > 1)
		enqueue_pushable_task(rq, p);
1166 1167

	inc_nr_running(rq);
P
Peter Zijlstra 已提交
1168 1169
}

1170
static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
1171
{
P
Peter Zijlstra 已提交
1172
	struct sched_rt_entity *rt_se = &p->rt;
I
Ingo Molnar 已提交
1173

1174
	update_curr_rt(rq);
1175
	dequeue_rt_entity(rt_se);
1176

1177
	dequeue_pushable_task(rq, p);
1178 1179

	dec_nr_running(rq);
I
Ingo Molnar 已提交
1180 1181 1182
}

/*
1183 1184
 * Put task to the head or the end of the run list without the overhead of
 * dequeue followed by enqueue.
I
Ingo Molnar 已提交
1185
 */
1186 1187
static void
requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head)
P
Peter Zijlstra 已提交
1188
{
1189
	if (on_rt_rq(rt_se)) {
1190 1191 1192 1193 1194 1195 1196
		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);
1197
	}
P
Peter Zijlstra 已提交
1198 1199
}

1200
static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head)
I
Ingo Molnar 已提交
1201
{
P
Peter Zijlstra 已提交
1202 1203
	struct sched_rt_entity *rt_se = &p->rt;
	struct rt_rq *rt_rq;
I
Ingo Molnar 已提交
1204

P
Peter Zijlstra 已提交
1205 1206
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);
1207
		requeue_rt_entity(rt_rq, rt_se, head);
P
Peter Zijlstra 已提交
1208
	}
I
Ingo Molnar 已提交
1209 1210
}

P
Peter Zijlstra 已提交
1211
static void yield_task_rt(struct rq *rq)
I
Ingo Molnar 已提交
1212
{
1213
	requeue_task_rt(rq, rq->curr, 0);
I
Ingo Molnar 已提交
1214 1215
}

1216
#ifdef CONFIG_SMP
1217 1218
static int find_lowest_rq(struct task_struct *task);

1219
static int
1220
select_task_rq_rt(struct task_struct *p, int sd_flag, int flags)
1221
{
1222 1223 1224 1225 1226
	struct task_struct *curr;
	struct rq *rq;
	int cpu;

	cpu = task_cpu(p);
1227

1228 1229 1230
	if (p->rt.nr_cpus_allowed == 1)
		goto out;

1231 1232 1233 1234
	/* For anything but wake ups, just return the task_cpu */
	if (sd_flag != SD_BALANCE_WAKE && sd_flag != SD_BALANCE_FORK)
		goto out;

1235 1236 1237 1238 1239
	rq = cpu_rq(cpu);

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

1240
	/*
1241
	 * If the current task on @p's runqueue is an RT task, then
1242 1243 1244 1245
	 * try to see if we can wake this RT task up on another
	 * runqueue. Otherwise simply start this RT task
	 * on its current runqueue.
	 *
1246 1247 1248 1249 1250 1251 1252 1253 1254
	 * 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.
1255 1256 1257 1258 1259 1260
	 *
	 * 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.
1261
	 */
1262 1263
	if (curr && unlikely(rt_task(curr)) &&
	    (curr->rt.nr_cpus_allowed < 2 ||
1264
	     curr->prio <= p->prio) &&
P
Peter Zijlstra 已提交
1265
	    (p->rt.nr_cpus_allowed > 1)) {
1266
		int target = find_lowest_rq(p);
1267

1268 1269
		if (target != -1)
			cpu = target;
1270
	}
1271
	rcu_read_unlock();
1272

1273
out:
1274
	return cpu;
1275
}
1276 1277 1278 1279 1280 1281

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

1282
	if (p->rt.nr_cpus_allowed != 1
1283 1284
	    && cpupri_find(&rq->rd->cpupri, p, NULL))
		return;
1285

1286 1287
	if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL))
		return;
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297

	/*
	 * 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);
}

1298 1299
#endif /* CONFIG_SMP */

I
Ingo Molnar 已提交
1300 1301 1302
/*
 * Preempt the current task with a newly woken task if needed:
 */
P
Peter Zijlstra 已提交
1303
static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
1304
{
1305
	if (p->prio < rq->curr->prio) {
I
Ingo Molnar 已提交
1306
		resched_task(rq->curr);
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
		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.
	 */
1323
	if (p->prio == rq->curr->prio && !test_tsk_need_resched(rq->curr))
1324
		check_preempt_equal_prio(rq, p);
1325
#endif
I
Ingo Molnar 已提交
1326 1327
}

P
Peter Zijlstra 已提交
1328 1329
static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq,
						   struct rt_rq *rt_rq)
I
Ingo Molnar 已提交
1330
{
P
Peter Zijlstra 已提交
1331 1332
	struct rt_prio_array *array = &rt_rq->active;
	struct sched_rt_entity *next = NULL;
I
Ingo Molnar 已提交
1333 1334 1335 1336
	struct list_head *queue;
	int idx;

	idx = sched_find_first_bit(array->bitmap);
P
Peter Zijlstra 已提交
1337
	BUG_ON(idx >= MAX_RT_PRIO);
I
Ingo Molnar 已提交
1338 1339

	queue = array->queue + idx;
P
Peter Zijlstra 已提交
1340
	next = list_entry(queue->next, struct sched_rt_entity, run_list);
1341

P
Peter Zijlstra 已提交
1342 1343
	return next;
}
I
Ingo Molnar 已提交
1344

1345
static struct task_struct *_pick_next_task_rt(struct rq *rq)
P
Peter Zijlstra 已提交
1346 1347 1348 1349
{
	struct sched_rt_entity *rt_se;
	struct task_struct *p;
	struct rt_rq *rt_rq;
I
Ingo Molnar 已提交
1350

P
Peter Zijlstra 已提交
1351 1352
	rt_rq = &rq->rt;

1353
	if (!rt_rq->rt_nr_running)
P
Peter Zijlstra 已提交
1354 1355
		return NULL;

P
Peter Zijlstra 已提交
1356
	if (rt_rq_throttled(rt_rq))
P
Peter Zijlstra 已提交
1357 1358 1359 1360
		return NULL;

	do {
		rt_se = pick_next_rt_entity(rq, rt_rq);
1361
		BUG_ON(!rt_se);
P
Peter Zijlstra 已提交
1362 1363 1364 1365
		rt_rq = group_rt_rq(rt_se);
	} while (rt_rq);

	p = rt_task_of(rt_se);
1366
	p->se.exec_start = rq->clock_task;
1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378

	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);

1379
#ifdef CONFIG_SMP
1380 1381 1382 1383 1384
	/*
	 * 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);
1385
#endif
1386

P
Peter Zijlstra 已提交
1387
	return p;
I
Ingo Molnar 已提交
1388 1389
}

1390
static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
I
Ingo Molnar 已提交
1391
{
1392
	update_curr_rt(rq);
1393 1394 1395 1396 1397

	/*
	 * The previous task needs to be made eligible for pushing
	 * if it is still active
	 */
P
Peter Zijlstra 已提交
1398
	if (on_rt_rq(&p->rt) && p->rt.nr_cpus_allowed > 1)
1399
		enqueue_pushable_task(rq, p);
I
Ingo Molnar 已提交
1400 1401
}

1402
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1403

S
Steven Rostedt 已提交
1404 1405 1406
/* Only try algorithms three times */
#define RT_MAX_TRIES 3

1407 1408 1409
static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu)
{
	if (!task_running(rq, p) &&
1410
	    (cpu < 0 || cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) &&
P
Peter Zijlstra 已提交
1411
	    (p->rt.nr_cpus_allowed > 1))
1412 1413 1414 1415
		return 1;
	return 0;
}

S
Steven Rostedt 已提交
1416
/* Return the second highest RT task, NULL otherwise */
1417
static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu)
S
Steven Rostedt 已提交
1418
{
P
Peter Zijlstra 已提交
1419 1420 1421 1422
	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
1423 1424
	int idx;

P
Peter Zijlstra 已提交
1425 1426 1427
	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
1428
next_idx:
P
Peter Zijlstra 已提交
1429 1430 1431 1432 1433
		if (idx >= MAX_RT_PRIO)
			continue;
		if (next && next->prio < idx)
			continue;
		list_for_each_entry(rt_se, array->queue + idx, run_list) {
1434 1435 1436 1437 1438 1439
			struct task_struct *p;

			if (!rt_entity_is_task(rt_se))
				continue;

			p = rt_task_of(rt_se);
P
Peter Zijlstra 已提交
1440 1441 1442 1443 1444 1445 1446 1447 1448
			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;
		}
1449 1450
	}

S
Steven Rostedt 已提交
1451 1452 1453
	return next;
}

1454
static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask);
S
Steven Rostedt 已提交
1455

G
Gregory Haskins 已提交
1456 1457 1458
static int find_lowest_rq(struct task_struct *task)
{
	struct sched_domain *sd;
1459
	struct cpumask *lowest_mask = __get_cpu_var(local_cpu_mask);
G
Gregory Haskins 已提交
1460 1461
	int this_cpu = smp_processor_id();
	int cpu      = task_cpu(task);
G
Gregory Haskins 已提交
1462

1463 1464 1465 1466
	/* Make sure the mask is initialized first */
	if (unlikely(!lowest_mask))
		return -1;

1467 1468
	if (task->rt.nr_cpus_allowed == 1)
		return -1; /* No other targets possible */
G
Gregory Haskins 已提交
1469

1470 1471
	if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask))
		return -1; /* No targets found */
G
Gregory Haskins 已提交
1472 1473 1474 1475 1476 1477 1478 1479 1480

	/*
	 * 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.
	 */
1481
	if (cpumask_test_cpu(cpu, lowest_mask))
G
Gregory Haskins 已提交
1482 1483 1484 1485 1486 1487
		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
1488 1489
	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
1490

1491
	rcu_read_lock();
R
Rusty Russell 已提交
1492 1493 1494
	for_each_domain(cpu, sd) {
		if (sd->flags & SD_WAKE_AFFINE) {
			int best_cpu;
G
Gregory Haskins 已提交
1495

R
Rusty Russell 已提交
1496 1497 1498 1499 1500
			/*
			 * "this_cpu" is cheaper to preempt than a
			 * remote processor.
			 */
			if (this_cpu != -1 &&
1501 1502
			    cpumask_test_cpu(this_cpu, sched_domain_span(sd))) {
				rcu_read_unlock();
R
Rusty Russell 已提交
1503
				return this_cpu;
1504
			}
R
Rusty Russell 已提交
1505 1506 1507

			best_cpu = cpumask_first_and(lowest_mask,
						     sched_domain_span(sd));
1508 1509
			if (best_cpu < nr_cpu_ids) {
				rcu_read_unlock();
R
Rusty Russell 已提交
1510
				return best_cpu;
1511
			}
G
Gregory Haskins 已提交
1512 1513
		}
	}
1514
	rcu_read_unlock();
G
Gregory Haskins 已提交
1515 1516 1517 1518 1519 1520

	/*
	 * 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
1521 1522 1523 1524 1525 1526 1527
	if (this_cpu != -1)
		return this_cpu;

	cpu = cpumask_any(lowest_mask);
	if (cpu < nr_cpu_ids)
		return cpu;
	return -1;
1528 1529 1530
}

/* Will lock the rq it finds */
1531
static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq)
1532 1533 1534
{
	struct rq *lowest_rq = NULL;
	int tries;
1535
	int cpu;
S
Steven Rostedt 已提交
1536

1537 1538 1539
	for (tries = 0; tries < RT_MAX_TRIES; tries++) {
		cpu = find_lowest_rq(task);

1540
		if ((cpu == -1) || (cpu == rq->cpu))
S
Steven Rostedt 已提交
1541 1542
			break;

1543 1544
		lowest_rq = cpu_rq(cpu);

S
Steven Rostedt 已提交
1545
		/* if the prio of this runqueue changed, try again */
1546
		if (double_lock_balance(rq, lowest_rq)) {
S
Steven Rostedt 已提交
1547 1548 1549 1550 1551 1552
			/*
			 * 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.
			 */
1553
			if (unlikely(task_rq(task) != rq ||
1554
				     !cpumask_test_cpu(lowest_rq->cpu,
1555
						       tsk_cpus_allowed(task)) ||
1556
				     task_running(rq, task) ||
P
Peter Zijlstra 已提交
1557
				     !task->on_rq)) {
1558

1559
				raw_spin_unlock(&lowest_rq->lock);
S
Steven Rostedt 已提交
1560 1561 1562 1563 1564 1565
				lowest_rq = NULL;
				break;
			}
		}

		/* If this rq is still suitable use it. */
1566
		if (lowest_rq->rt.highest_prio.curr > task->prio)
S
Steven Rostedt 已提交
1567 1568 1569
			break;

		/* try again */
1570
		double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1571 1572 1573 1574 1575 1576
		lowest_rq = NULL;
	}

	return lowest_rq;
}

1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590
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
1591
	BUG_ON(!p->on_rq);
1592 1593 1594 1595 1596
	BUG_ON(!rt_task(p));

	return p;
}

S
Steven Rostedt 已提交
1597 1598 1599 1600 1601
/*
 * 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.
 */
1602
static int push_rt_task(struct rq *rq)
S
Steven Rostedt 已提交
1603 1604 1605
{
	struct task_struct *next_task;
	struct rq *lowest_rq;
1606
	int ret = 0;
S
Steven Rostedt 已提交
1607

G
Gregory Haskins 已提交
1608 1609 1610
	if (!rq->rt.overloaded)
		return 0;

1611
	next_task = pick_next_pushable_task(rq);
S
Steven Rostedt 已提交
1612 1613 1614
	if (!next_task)
		return 0;

1615 1616 1617 1618 1619
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
       if (unlikely(task_running(rq, next_task)))
               return 0;
#endif

P
Peter Zijlstra 已提交
1620
retry:
1621
	if (unlikely(next_task == rq->curr)) {
1622
		WARN_ON(1);
S
Steven Rostedt 已提交
1623
		return 0;
1624
	}
S
Steven Rostedt 已提交
1625 1626 1627 1628 1629 1630

	/*
	 * It's possible that the next_task slipped in of
	 * higher priority than current. If that's the case
	 * just reschedule current.
	 */
1631 1632
	if (unlikely(next_task->prio < rq->curr->prio)) {
		resched_task(rq->curr);
S
Steven Rostedt 已提交
1633 1634 1635
		return 0;
	}

1636
	/* We might release rq lock */
S
Steven Rostedt 已提交
1637 1638 1639
	get_task_struct(next_task);

	/* find_lock_lowest_rq locks the rq if found */
1640
	lowest_rq = find_lock_lowest_rq(next_task, rq);
S
Steven Rostedt 已提交
1641 1642 1643
	if (!lowest_rq) {
		struct task_struct *task;
		/*
1644
		 * find_lock_lowest_rq releases rq->lock
1645 1646 1647 1648 1649
		 * 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
1650
		 */
1651
		task = pick_next_pushable_task(rq);
1652 1653
		if (task_cpu(next_task) == rq->cpu && task == next_task) {
			/*
1654 1655 1656 1657
			 * 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.
1658 1659
			 */
			goto out;
S
Steven Rostedt 已提交
1660
		}
1661

1662 1663 1664 1665
		if (!task)
			/* No more tasks, just exit */
			goto out;

1666
		/*
1667
		 * Something has shifted, try again.
1668
		 */
1669 1670 1671
		put_task_struct(next_task);
		next_task = task;
		goto retry;
S
Steven Rostedt 已提交
1672 1673
	}

1674
	deactivate_task(rq, next_task, 0);
S
Steven Rostedt 已提交
1675 1676
	set_task_cpu(next_task, lowest_rq->cpu);
	activate_task(lowest_rq, next_task, 0);
1677
	ret = 1;
S
Steven Rostedt 已提交
1678 1679 1680

	resched_task(lowest_rq->curr);

1681
	double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1682 1683 1684 1685

out:
	put_task_struct(next_task);

1686
	return ret;
S
Steven Rostedt 已提交
1687 1688 1689 1690 1691 1692 1693 1694 1695
}

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

1696 1697
static int pull_rt_task(struct rq *this_rq)
{
I
Ingo Molnar 已提交
1698
	int this_cpu = this_rq->cpu, ret = 0, cpu;
1699
	struct task_struct *p;
1700 1701
	struct rq *src_rq;

1702
	if (likely(!rt_overloaded(this_rq)))
1703 1704
		return 0;

1705
	for_each_cpu(cpu, this_rq->rd->rto_mask) {
1706 1707 1708 1709
		if (this_cpu == cpu)
			continue;

		src_rq = cpu_rq(cpu);
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721

		/*
		 * 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;

1722 1723 1724
		/*
		 * We can potentially drop this_rq's lock in
		 * double_lock_balance, and another CPU could
1725
		 * alter this_rq
1726
		 */
1727
		double_lock_balance(this_rq, src_rq);
1728 1729 1730 1731

		/*
		 * Are there still pullable RT tasks?
		 */
M
Mike Galbraith 已提交
1732 1733
		if (src_rq->rt.rt_nr_running <= 1)
			goto skip;
1734 1735 1736 1737 1738 1739 1740

		p = pick_next_highest_task_rt(src_rq, this_cpu);

		/*
		 * Do we have an RT task that preempts
		 * the to-be-scheduled task?
		 */
1741
		if (p && (p->prio < this_rq->rt.highest_prio.curr)) {
1742
			WARN_ON(p == src_rq->curr);
P
Peter Zijlstra 已提交
1743
			WARN_ON(!p->on_rq);
1744 1745 1746 1747 1748 1749 1750

			/*
			 * 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
1751
			 * current task on the run queue
1752
			 */
1753
			if (p->prio < src_rq->curr->prio)
M
Mike Galbraith 已提交
1754
				goto skip;
1755 1756 1757 1758 1759 1760 1761 1762 1763

			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
1764
			 * in another runqueue. (low likelihood
1765 1766 1767
			 * but possible)
			 */
		}
P
Peter Zijlstra 已提交
1768
skip:
1769
		double_unlock_balance(this_rq, src_rq);
1770 1771 1772 1773 1774
	}

	return ret;
}

1775
static void pre_schedule_rt(struct rq *rq, struct task_struct *prev)
1776 1777
{
	/* Try to pull RT tasks here if we lower this rq's prio */
Y
Yong Zhang 已提交
1778
	if (rq->rt.highest_prio.curr > prev->prio)
1779 1780 1781
		pull_rt_task(rq);
}

1782
static void post_schedule_rt(struct rq *rq)
S
Steven Rostedt 已提交
1783
{
1784
	push_rt_tasks(rq);
S
Steven Rostedt 已提交
1785 1786
}

1787 1788 1789 1790
/*
 * If we are not running and we are not going to reschedule soon, we should
 * try to push tasks away now
 */
1791
static void task_woken_rt(struct rq *rq, struct task_struct *p)
1792
{
1793
	if (!task_running(rq, p) &&
1794
	    !test_tsk_need_resched(rq->curr) &&
1795
	    has_pushable_tasks(rq) &&
1796
	    p->rt.nr_cpus_allowed > 1 &&
1797
	    rt_task(rq->curr) &&
1798
	    (rq->curr->rt.nr_cpus_allowed < 2 ||
1799
	     rq->curr->prio <= p->prio))
1800 1801 1802
		push_rt_tasks(rq);
}

1803
static void set_cpus_allowed_rt(struct task_struct *p,
1804
				const struct cpumask *new_mask)
1805
{
1806 1807
	struct rq *rq;
	int weight;
1808 1809 1810

	BUG_ON(!rt_task(p));

1811 1812
	if (!p->on_rq)
		return;
1813

1814
	weight = cpumask_weight(new_mask);
1815

1816 1817 1818 1819 1820 1821
	/*
	 * Only update if the process changes its state from whether it
	 * can migrate or not.
	 */
	if ((p->rt.nr_cpus_allowed > 1) == (weight > 1))
		return;
1822

1823
	rq = task_rq(p);
1824

1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836
	/*
	 * The process used to be able to migrate OR it can now migrate
	 */
	if (weight <= 1) {
		if (!task_current(rq, p))
			dequeue_pushable_task(rq, p);
		BUG_ON(!rq->rt.rt_nr_migratory);
		rq->rt.rt_nr_migratory--;
	} else {
		if (!task_current(rq, p))
			enqueue_pushable_task(rq, p);
		rq->rt.rt_nr_migratory++;
1837
	}
1838 1839

	update_rt_migration(&rq->rt);
1840
}
1841

1842
/* Assumes rq->lock is held */
1843
static void rq_online_rt(struct rq *rq)
1844 1845 1846
{
	if (rq->rt.overloaded)
		rt_set_overload(rq);
1847

P
Peter Zijlstra 已提交
1848 1849
	__enable_runtime(rq);

1850
	cpupri_set(&rq->rd->cpupri, rq->cpu, rq->rt.highest_prio.curr);
1851 1852 1853
}

/* Assumes rq->lock is held */
1854
static void rq_offline_rt(struct rq *rq)
1855 1856 1857
{
	if (rq->rt.overloaded)
		rt_clear_overload(rq);
1858

P
Peter Zijlstra 已提交
1859 1860
	__disable_runtime(rq);

1861
	cpupri_set(&rq->rd->cpupri, rq->cpu, CPUPRI_INVALID);
1862
}
1863 1864 1865 1866 1867

/*
 * 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
1868
static void switched_from_rt(struct rq *rq, struct task_struct *p)
1869 1870 1871 1872 1873 1874 1875 1876
{
	/*
	 * 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
1877
	if (p->on_rq && !rq->rt.rt_nr_running)
1878 1879
		pull_rt_task(rq);
}
1880

1881
void init_sched_rt_class(void)
1882 1883 1884
{
	unsigned int i;

1885
	for_each_possible_cpu(i) {
1886
		zalloc_cpumask_var_node(&per_cpu(local_cpu_mask, i),
1887
					GFP_KERNEL, cpu_to_node(i));
1888
	}
1889
}
1890 1891 1892 1893 1894 1895 1896
#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
1897
static void switched_to_rt(struct rq *rq, struct task_struct *p)
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
{
	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
1908
	if (p->on_rq && rq->curr != p) {
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
#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
1924 1925
static void
prio_changed_rt(struct rq *rq, struct task_struct *p, int oldprio)
1926
{
P
Peter Zijlstra 已提交
1927
	if (!p->on_rq)
P
Peter Zijlstra 已提交
1928 1929 1930
		return;

	if (rq->curr == p) {
1931 1932 1933 1934 1935 1936 1937 1938 1939
#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
1940 1941 1942
		 * 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.
1943
		 */
1944
		if (p->prio > rq->rt.highest_prio.curr && rq->curr == p)
1945 1946 1947 1948 1949
			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
1950
#endif /* CONFIG_SMP */
1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
	} 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);
	}
}

1962 1963 1964 1965
static void watchdog(struct rq *rq, struct task_struct *p)
{
	unsigned long soft, hard;

1966 1967 1968
	/* 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);
1969 1970 1971 1972 1973 1974

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

		p->rt.timeout++;
		next = DIV_ROUND_UP(min(soft, hard), USEC_PER_SEC/HZ);
1975
		if (p->rt.timeout > next)
1976
			p->cputime_expires.sched_exp = p->se.sum_exec_runtime;
1977 1978
	}
}
I
Ingo Molnar 已提交
1979

P
Peter Zijlstra 已提交
1980
static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued)
I
Ingo Molnar 已提交
1981
{
1982 1983
	update_curr_rt(rq);

1984 1985
	watchdog(rq, p);

I
Ingo Molnar 已提交
1986 1987 1988 1989 1990 1991 1992
	/*
	 * 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
1993
	if (--p->rt.time_slice)
I
Ingo Molnar 已提交
1994 1995
		return;

1996
	p->rt.time_slice = RR_TIMESLICE;
I
Ingo Molnar 已提交
1997

1998 1999 2000 2001
	/*
	 * Requeue to the end of queue if we are not the only element
	 * on the queue:
	 */
P
Peter Zijlstra 已提交
2002
	if (p->rt.run_list.prev != p->rt.run_list.next) {
2003
		requeue_task_rt(rq, p, 0);
2004 2005
		set_tsk_need_resched(p);
	}
I
Ingo Molnar 已提交
2006 2007
}

2008 2009 2010 2011
static void set_curr_task_rt(struct rq *rq)
{
	struct task_struct *p = rq->curr;

2012
	p->se.exec_start = rq->clock_task;
2013 2014 2015

	/* The running task is never eligible for pushing */
	dequeue_pushable_task(rq, p);
2016 2017
}

2018
static unsigned int get_rr_interval_rt(struct rq *rq, struct task_struct *task)
2019 2020 2021 2022 2023
{
	/*
	 * Time slice is 0 for SCHED_FIFO tasks
	 */
	if (task->policy == SCHED_RR)
2024
		return RR_TIMESLICE;
2025 2026 2027 2028
	else
		return 0;
}

2029
const struct sched_class rt_sched_class = {
2030
	.next			= &fair_sched_class,
I
Ingo Molnar 已提交
2031 2032 2033 2034 2035 2036 2037 2038 2039
	.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,

2040
#ifdef CONFIG_SMP
L
Li Zefan 已提交
2041 2042
	.select_task_rq		= select_task_rq_rt,

2043
	.set_cpus_allowed       = set_cpus_allowed_rt,
2044 2045
	.rq_online              = rq_online_rt,
	.rq_offline             = rq_offline_rt,
2046 2047
	.pre_schedule		= pre_schedule_rt,
	.post_schedule		= post_schedule_rt,
2048
	.task_woken		= task_woken_rt,
2049
	.switched_from		= switched_from_rt,
2050
#endif
I
Ingo Molnar 已提交
2051

2052
	.set_curr_task          = set_curr_task_rt,
I
Ingo Molnar 已提交
2053
	.task_tick		= task_tick_rt,
2054

2055 2056
	.get_rr_interval	= get_rr_interval_rt,

2057 2058
	.prio_changed		= prio_changed_rt,
	.switched_to		= switched_to_rt,
I
Ingo Molnar 已提交
2059
};
2060 2061 2062 2063

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

2064
void print_rt_stats(struct seq_file *m, int cpu)
2065
{
C
Cheng Xu 已提交
2066
	rt_rq_iter_t iter;
2067 2068 2069
	struct rt_rq *rt_rq;

	rcu_read_lock();
C
Cheng Xu 已提交
2070
	for_each_rt_rq(rt_rq, iter, cpu_rq(cpu))
2071 2072 2073
		print_rt_rq(m, cpu, rt_rq);
	rcu_read_unlock();
}
2074
#endif /* CONFIG_SCHED_DEBUG */