rt.c 48.1 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
#include "sched.h"

#include <linux/slab.h>

10 11
int sched_rr_timeslice = RR_TIMESLICE;

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
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
82 83
	/* We start is dequeued state, because no RT tasks are queued */
	rt_rq->rt_queued = 0;
84 85 86 87 88 89 90

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

91
#ifdef CONFIG_RT_GROUP_SCHED
92 93 94 95
static void destroy_rt_bandwidth(struct rt_bandwidth *rt_b)
{
	hrtimer_cancel(&rt_b->rt_period_timer);
}
96 97 98

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

99 100
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
101 102 103
#ifdef CONFIG_SCHED_DEBUG
	WARN_ON_ONCE(!rt_entity_is_task(rt_se));
#endif
104 105 106 107 108 109 110 111 112 113 114 115 116
	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;
}

117 118 119 120 121 122 123
static inline struct rq *rq_of_rt_se(struct sched_rt_entity *rt_se)
{
	struct rt_rq *rt_rq = rt_se->rt_rq;

	return rt_rq->rq;
}

124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208
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;
}

209 210
#else /* CONFIG_RT_GROUP_SCHED */

211 212
#define rt_entity_is_task(rt_se) (1)

213 214 215 216 217
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
	return container_of(rt_se, struct task_struct, rt);
}

218 219 220 221 222
static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
{
	return container_of(rt_rq, struct rq, rt);
}

223
static inline struct rq *rq_of_rt_se(struct sched_rt_entity *rt_se)
224 225
{
	struct task_struct *p = rt_task_of(rt_se);
226 227 228 229 230 231 232

	return task_rq(p);
}

static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se)
{
	struct rq *rq = rq_of_rt_se(rt_se);
233 234 235 236

	return &rq->rt;
}

237 238 239 240 241 242
void free_rt_sched_group(struct task_group *tg) { }

int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
{
	return 1;
}
243 244
#endif /* CONFIG_RT_GROUP_SCHED */

S
Steven Rostedt 已提交
245
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
246

247 248
static int pull_rt_task(struct rq *this_rq);

P
Peter Zijlstra 已提交
249 250 251 252 253 254
static inline bool need_pull_rt_task(struct rq *rq, struct task_struct *prev)
{
	/* Try to pull RT tasks here if we lower this rq's prio */
	return rq->rt.highest_prio.curr > prev->prio;
}

255
static inline int rt_overloaded(struct rq *rq)
S
Steven Rostedt 已提交
256
{
257
	return atomic_read(&rq->rd->rto_count);
S
Steven Rostedt 已提交
258
}
I
Ingo Molnar 已提交
259

S
Steven Rostedt 已提交
260 261
static inline void rt_set_overload(struct rq *rq)
{
262 263 264
	if (!rq->online)
		return;

265
	cpumask_set_cpu(rq->cpu, rq->rd->rto_mask);
S
Steven Rostedt 已提交
266 267 268 269 270 271
	/*
	 * 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.
P
Peter Zijlstra 已提交
272 273
	 *
	 * Matched by the barrier in pull_rt_task().
S
Steven Rostedt 已提交
274
	 */
P
Peter Zijlstra 已提交
275
	smp_wmb();
276
	atomic_inc(&rq->rd->rto_count);
S
Steven Rostedt 已提交
277
}
I
Ingo Molnar 已提交
278

S
Steven Rostedt 已提交
279 280
static inline void rt_clear_overload(struct rq *rq)
{
281 282 283
	if (!rq->online)
		return;

S
Steven Rostedt 已提交
284
	/* the order here really doesn't matter */
285
	atomic_dec(&rq->rd->rto_count);
286
	cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask);
S
Steven Rostedt 已提交
287
}
288

289
static void update_rt_migration(struct rt_rq *rt_rq)
290
{
291
	if (rt_rq->rt_nr_migratory && rt_rq->rt_nr_total > 1) {
292 293 294
		if (!rt_rq->overloaded) {
			rt_set_overload(rq_of_rt_rq(rt_rq));
			rt_rq->overloaded = 1;
295
		}
296 297 298
	} else if (rt_rq->overloaded) {
		rt_clear_overload(rq_of_rt_rq(rt_rq));
		rt_rq->overloaded = 0;
299
	}
300
}
S
Steven Rostedt 已提交
301

302 303
static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
304 305
	struct task_struct *p;

306 307 308
	if (!rt_entity_is_task(rt_se))
		return;

309
	p = rt_task_of(rt_se);
310 311 312
	rt_rq = &rq_of_rt_rq(rt_rq)->rt;

	rt_rq->rt_nr_total++;
313
	if (p->nr_cpus_allowed > 1)
314 315 316 317 318 319 320
		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)
{
321 322
	struct task_struct *p;

323 324 325
	if (!rt_entity_is_task(rt_se))
		return;

326
	p = rt_task_of(rt_se);
327 328 329
	rt_rq = &rq_of_rt_rq(rt_rq)->rt;

	rt_rq->rt_nr_total--;
330
	if (p->nr_cpus_allowed > 1)
331 332 333 334 335
		rt_rq->rt_nr_migratory--;

	update_rt_migration(rt_rq);
}

336 337 338 339 340
static inline int has_pushable_tasks(struct rq *rq)
{
	return !plist_head_empty(&rq->rt.pushable_tasks);
}

P
Peter Zijlstra 已提交
341 342 343 344 345 346 347 348 349
static inline void set_post_schedule(struct rq *rq)
{
	/*
	 * 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);
}

350 351 352 353 354
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);
355 356 357 358

	/* Update the highest prio pushable task */
	if (p->prio < rq->rt.highest_prio.next)
		rq->rt.highest_prio.next = p->prio;
359 360 361 362 363 364
}

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

365 366 367 368 369 370 371
	/* 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;
372 373
}

374 375
#else

376
static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
P
Peter Zijlstra 已提交
377
{
P
Peter Zijlstra 已提交
378 379
}

380 381 382 383
static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p)
{
}

384
static inline
385 386 387 388
void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}

389
static inline
390 391 392
void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}
393

P
Peter Zijlstra 已提交
394 395 396 397 398 399 400 401 402 403 404 405 406
static inline bool need_pull_rt_task(struct rq *rq, struct task_struct *prev)
{
	return false;
}

static inline int pull_rt_task(struct rq *this_rq)
{
	return 0;
}

static inline void set_post_schedule(struct rq *rq)
{
}
S
Steven Rostedt 已提交
407 408
#endif /* CONFIG_SMP */

409 410 411
static void enqueue_top_rt_rq(struct rt_rq *rt_rq);
static void dequeue_top_rt_rq(struct rt_rq *rt_rq);

P
Peter Zijlstra 已提交
412 413 414 415 416
static inline int on_rt_rq(struct sched_rt_entity *rt_se)
{
	return !list_empty(&rt_se->run_list);
}

417
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
418

P
Peter Zijlstra 已提交
419
static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
420 421
{
	if (!rt_rq->tg)
P
Peter Zijlstra 已提交
422
		return RUNTIME_INF;
P
Peter Zijlstra 已提交
423

P
Peter Zijlstra 已提交
424 425 426 427 428 429
	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
430 431
}

C
Cheng Xu 已提交
432 433
typedef struct task_group *rt_rq_iter_t;

434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
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
451

P
Peter Zijlstra 已提交
452 453 454 455 456 457 458 459
#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;
}

460
static void enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head);
P
Peter Zijlstra 已提交
461 462
static void dequeue_rt_entity(struct sched_rt_entity *rt_se);

P
Peter Zijlstra 已提交
463
static void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
464
{
465
	struct task_struct *curr = rq_of_rt_rq(rt_rq)->curr;
466 467
	struct sched_rt_entity *rt_se;

468 469 470
	int cpu = cpu_of(rq_of_rt_rq(rt_rq));

	rt_se = rt_rq->tg->rt_se[cpu];
P
Peter Zijlstra 已提交
471

472
	if (rt_rq->rt_nr_running) {
473 474 475
		if (!rt_se)
			enqueue_top_rt_rq(rt_rq);
		else if (!on_rt_rq(rt_se))
476
			enqueue_rt_entity(rt_se, false);
477

478
		if (rt_rq->highest_prio.curr < curr->prio)
479
			resched_task(curr);
P
Peter Zijlstra 已提交
480 481 482
	}
}

P
Peter Zijlstra 已提交
483
static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
484
{
485
	struct sched_rt_entity *rt_se;
486
	int cpu = cpu_of(rq_of_rt_rq(rt_rq));
487

488
	rt_se = rt_rq->tg->rt_se[cpu];
P
Peter Zijlstra 已提交
489

490 491 492
	if (!rt_se)
		dequeue_top_rt_rq(rt_rq);
	else if (on_rt_rq(rt_se))
P
Peter Zijlstra 已提交
493 494 495
		dequeue_rt_entity(rt_se);
}

496 497 498 499 500
static inline int rt_rq_throttled(struct rt_rq *rt_rq)
{
	return rt_rq->rt_throttled && !rt_rq->rt_nr_boosted;
}

P
Peter Zijlstra 已提交
501 502 503 504 505 506 507 508 509 510 511 512
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;
}

513
#ifdef CONFIG_SMP
514
static inline const struct cpumask *sched_rt_period_mask(void)
515
{
N
Nathan Zimmer 已提交
516
	return this_rq()->rd->span;
517
}
P
Peter Zijlstra 已提交
518
#else
519
static inline const struct cpumask *sched_rt_period_mask(void)
520
{
521
	return cpu_online_mask;
522 523
}
#endif
P
Peter Zijlstra 已提交
524

525 526
static inline
struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu)
P
Peter Zijlstra 已提交
527
{
528 529
	return container_of(rt_b, struct task_group, rt_bandwidth)->rt_rq[cpu];
}
P
Peter Zijlstra 已提交
530

P
Peter Zijlstra 已提交
531 532 533 534 535
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &rt_rq->tg->rt_bandwidth;
}

536
#else /* !CONFIG_RT_GROUP_SCHED */
537 538 539

static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
{
P
Peter Zijlstra 已提交
540 541 542 543 544 545
	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
546 547
}

C
Cheng Xu 已提交
548 549 550 551 552
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)

P
Peter Zijlstra 已提交
553 554 555 556 557 558 559 560
#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
561
static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
562
{
563 564 565 566 567 568 569
	struct rq *rq = rq_of_rt_rq(rt_rq);

	if (!rt_rq->rt_nr_running)
		return;

	enqueue_top_rt_rq(rt_rq);
	resched_task(rq->curr);
P
Peter Zijlstra 已提交
570 571
}

P
Peter Zijlstra 已提交
572
static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
573
{
574
	dequeue_top_rt_rq(rt_rq);
P
Peter Zijlstra 已提交
575 576
}

577 578 579 580 581
static inline int rt_rq_throttled(struct rt_rq *rt_rq)
{
	return rt_rq->rt_throttled;
}

582
static inline const struct cpumask *sched_rt_period_mask(void)
583
{
584
	return cpu_online_mask;
585 586 587 588 589 590 591 592
}

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
593 594 595 596 597
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &def_rt_bandwidth;
}

598
#endif /* CONFIG_RT_GROUP_SCHED */
599

600 601 602 603 604 605 606 607
bool sched_rt_bandwidth_account(struct rt_rq *rt_rq)
{
	struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);

	return (hrtimer_active(&rt_b->rt_period_timer) ||
		rt_rq->rt_time < rt_b->rt_runtime);
}

P
Peter Zijlstra 已提交
608
#ifdef CONFIG_SMP
609 610 611
/*
 * We ran out of runtime, see if we can borrow some from our neighbours.
 */
612
static int do_balance_runtime(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
613 614
{
	struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);
615
	struct root_domain *rd = rq_of_rt_rq(rt_rq)->rd;
P
Peter Zijlstra 已提交
616 617 618
	int i, weight, more = 0;
	u64 rt_period;

619
	weight = cpumask_weight(rd->span);
P
Peter Zijlstra 已提交
620

621
	raw_spin_lock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
622
	rt_period = ktime_to_ns(rt_b->rt_period);
623
	for_each_cpu(i, rd->span) {
P
Peter Zijlstra 已提交
624 625 626 627 628 629
		struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
		s64 diff;

		if (iter == rt_rq)
			continue;

630
		raw_spin_lock(&iter->rt_runtime_lock);
631 632 633 634 635
		/*
		 * 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
636 637 638
		if (iter->rt_runtime == RUNTIME_INF)
			goto next;

639 640 641 642
		/*
		 * 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
643 644
		diff = iter->rt_runtime - iter->rt_time;
		if (diff > 0) {
645
			diff = div_u64((u64)diff, weight);
P
Peter Zijlstra 已提交
646 647 648 649 650 651
			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) {
652
				raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
653 654 655
				break;
			}
		}
P
Peter Zijlstra 已提交
656
next:
657
		raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
658
	}
659
	raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
660 661 662

	return more;
}
P
Peter Zijlstra 已提交
663

664 665 666
/*
 * Ensure this RQ takes back all the runtime it lend to its neighbours.
 */
P
Peter Zijlstra 已提交
667 668 669
static void __disable_runtime(struct rq *rq)
{
	struct root_domain *rd = rq->rd;
C
Cheng Xu 已提交
670
	rt_rq_iter_t iter;
P
Peter Zijlstra 已提交
671 672 673 674 675
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

C
Cheng Xu 已提交
676
	for_each_rt_rq(rt_rq, iter, rq) {
P
Peter Zijlstra 已提交
677 678 679 680
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);
		s64 want;
		int i;

681 682
		raw_spin_lock(&rt_b->rt_runtime_lock);
		raw_spin_lock(&rt_rq->rt_runtime_lock);
683 684 685 686 687
		/*
		 * 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
688 689 690
		if (rt_rq->rt_runtime == RUNTIME_INF ||
				rt_rq->rt_runtime == rt_b->rt_runtime)
			goto balanced;
691
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
692

693 694 695 696 697
		/*
		 * 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
698 699
		want = rt_b->rt_runtime - rt_rq->rt_runtime;

700 701 702
		/*
		 * Greedy reclaim, take back as much as we can.
		 */
703
		for_each_cpu(i, rd->span) {
P
Peter Zijlstra 已提交
704 705 706
			struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
			s64 diff;

707 708 709
			/*
			 * Can't reclaim from ourselves or disabled runqueues.
			 */
710
			if (iter == rt_rq || iter->rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
711 712
				continue;

713
			raw_spin_lock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
714 715 716 717 718 719 720 721
			if (want > 0) {
				diff = min_t(s64, iter->rt_runtime, want);
				iter->rt_runtime -= diff;
				want -= diff;
			} else {
				iter->rt_runtime -= want;
				want -= want;
			}
722
			raw_spin_unlock(&iter->rt_runtime_lock);
P
Peter Zijlstra 已提交
723 724 725 726 727

			if (!want)
				break;
		}

728
		raw_spin_lock(&rt_rq->rt_runtime_lock);
729 730 731 732
		/*
		 * We cannot be left wanting - that would mean some runtime
		 * leaked out of the system.
		 */
P
Peter Zijlstra 已提交
733 734
		BUG_ON(want);
balanced:
735 736 737 738
		/*
		 * 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
739
		rt_rq->rt_runtime = RUNTIME_INF;
740
		rt_rq->rt_throttled = 0;
741 742
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
		raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
743 744 745 746 747
	}
}

static void __enable_runtime(struct rq *rq)
{
C
Cheng Xu 已提交
748
	rt_rq_iter_t iter;
P
Peter Zijlstra 已提交
749 750 751 752 753
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

754 755 756
	/*
	 * Reset each runqueue's bandwidth settings
	 */
C
Cheng Xu 已提交
757
	for_each_rt_rq(rt_rq, iter, rq) {
P
Peter Zijlstra 已提交
758 759
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);

760 761
		raw_spin_lock(&rt_b->rt_runtime_lock);
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
762 763
		rt_rq->rt_runtime = rt_b->rt_runtime;
		rt_rq->rt_time = 0;
764
		rt_rq->rt_throttled = 0;
765 766
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
		raw_spin_unlock(&rt_b->rt_runtime_lock);
P
Peter Zijlstra 已提交
767 768 769
	}
}

770 771 772 773
static int balance_runtime(struct rt_rq *rt_rq)
{
	int more = 0;

774 775 776
	if (!sched_feat(RT_RUNTIME_SHARE))
		return more;

777
	if (rt_rq->rt_time > rt_rq->rt_runtime) {
778
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
779
		more = do_balance_runtime(rt_rq);
780
		raw_spin_lock(&rt_rq->rt_runtime_lock);
781 782 783 784
	}

	return more;
}
785
#else /* !CONFIG_SMP */
786 787 788 789
static inline int balance_runtime(struct rt_rq *rt_rq)
{
	return 0;
}
790
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
791

792 793
static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
{
794
	int i, idle = 1, throttled = 0;
795
	const struct cpumask *span;
796 797

	span = sched_rt_period_mask();
798 799 800 801 802 803 804 805 806 807 808 809 810
#ifdef CONFIG_RT_GROUP_SCHED
	/*
	 * FIXME: isolated CPUs should really leave the root task group,
	 * whether they are isolcpus or were isolated via cpusets, lest
	 * the timer run on a CPU which does not service all runqueues,
	 * potentially leaving other CPUs indefinitely throttled.  If
	 * isolation is really required, the user will turn the throttle
	 * off to kill the perturbations it causes anyway.  Meanwhile,
	 * this maintains functionality for boot and/or troubleshooting.
	 */
	if (rt_b == &root_task_group.rt_bandwidth)
		span = cpu_online_mask;
#endif
811
	for_each_cpu(i, span) {
812 813 814 815
		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);

816
		raw_spin_lock(&rq->lock);
817 818 819
		if (rt_rq->rt_time) {
			u64 runtime;

820
			raw_spin_lock(&rt_rq->rt_runtime_lock);
821 822 823 824 825 826 827
			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;
828 829 830 831 832 833 834

				/*
				 * 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;
835 836 837
			}
			if (rt_rq->rt_time || rt_rq->rt_nr_running)
				idle = 0;
838
			raw_spin_unlock(&rt_rq->rt_runtime_lock);
839
		} else if (rt_rq->rt_nr_running) {
840
			idle = 0;
841 842 843
			if (!rt_rq_throttled(rt_rq))
				enqueue = 1;
		}
844 845
		if (rt_rq->rt_throttled)
			throttled = 1;
846 847 848

		if (enqueue)
			sched_rt_rq_enqueue(rt_rq);
849
		raw_spin_unlock(&rq->lock);
850 851
	}

852 853 854
	if (!throttled && (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF))
		return 1;

855 856
	return idle;
}
P
Peter Zijlstra 已提交
857

P
Peter Zijlstra 已提交
858 859
static inline int rt_se_prio(struct sched_rt_entity *rt_se)
{
860
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
861 862 863
	struct rt_rq *rt_rq = group_rt_rq(rt_se);

	if (rt_rq)
864
		return rt_rq->highest_prio.curr;
P
Peter Zijlstra 已提交
865 866 867 868 869
#endif

	return rt_task_of(rt_se)->prio;
}

P
Peter Zijlstra 已提交
870
static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq)
P
Peter Zijlstra 已提交
871
{
P
Peter Zijlstra 已提交
872
	u64 runtime = sched_rt_runtime(rt_rq);
P
Peter Zijlstra 已提交
873 874

	if (rt_rq->rt_throttled)
P
Peter Zijlstra 已提交
875
		return rt_rq_throttled(rt_rq);
P
Peter Zijlstra 已提交
876

877
	if (runtime >= sched_rt_period(rt_rq))
P
Peter Zijlstra 已提交
878 879
		return 0;

880 881 882 883
	balance_runtime(rt_rq);
	runtime = sched_rt_runtime(rt_rq);
	if (runtime == RUNTIME_INF)
		return 0;
P
Peter Zijlstra 已提交
884

P
Peter Zijlstra 已提交
885
	if (rt_rq->rt_time > runtime) {
886 887 888 889 890 891 892 893
		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)) {
			rt_rq->rt_throttled = 1;
J
John Stultz 已提交
894
			printk_deferred_once("sched: RT throttling activated\n");
895 896 897 898 899 900 901 902 903
		} 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
904
		if (rt_rq_throttled(rt_rq)) {
P
Peter Zijlstra 已提交
905
			sched_rt_rq_dequeue(rt_rq);
P
Peter Zijlstra 已提交
906 907
			return 1;
		}
P
Peter Zijlstra 已提交
908 909 910 911 912
	}

	return 0;
}

I
Ingo Molnar 已提交
913 914 915 916
/*
 * Update the current task's runtime statistics. Skip current tasks that
 * are not in our scheduling class.
 */
A
Alexey Dobriyan 已提交
917
static void update_curr_rt(struct rq *rq)
I
Ingo Molnar 已提交
918 919
{
	struct task_struct *curr = rq->curr;
P
Peter Zijlstra 已提交
920
	struct sched_rt_entity *rt_se = &curr->rt;
I
Ingo Molnar 已提交
921 922
	u64 delta_exec;

P
Peter Zijlstra 已提交
923
	if (curr->sched_class != &rt_sched_class)
I
Ingo Molnar 已提交
924 925
		return;

926
	delta_exec = rq_clock_task(rq) - curr->se.exec_start;
927 928
	if (unlikely((s64)delta_exec <= 0))
		return;
I
Ingo Molnar 已提交
929

930 931
	schedstat_set(curr->se.statistics.exec_max,
		      max(curr->se.statistics.exec_max, delta_exec));
I
Ingo Molnar 已提交
932 933

	curr->se.sum_exec_runtime += delta_exec;
934 935
	account_group_exec_runtime(curr, delta_exec);

936
	curr->se.exec_start = rq_clock_task(rq);
937
	cpuacct_charge(curr, delta_exec);
P
Peter Zijlstra 已提交
938

939 940
	sched_rt_avg_update(rq, delta_exec);

941 942 943
	if (!rt_bandwidth_enabled())
		return;

D
Dhaval Giani 已提交
944
	for_each_sched_rt_entity(rt_se) {
945
		struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
D
Dhaval Giani 已提交
946

947
		if (sched_rt_runtime(rt_rq) != RUNTIME_INF) {
948
			raw_spin_lock(&rt_rq->rt_runtime_lock);
949 950 951
			rt_rq->rt_time += delta_exec;
			if (sched_rt_runtime_exceeded(rt_rq))
				resched_task(curr);
952
			raw_spin_unlock(&rt_rq->rt_runtime_lock);
953
		}
D
Dhaval Giani 已提交
954
	}
I
Ingo Molnar 已提交
955 956
}

957 958 959 960 961 962 963 964 965 966 967 968
static void
dequeue_top_rt_rq(struct rt_rq *rt_rq)
{
	struct rq *rq = rq_of_rt_rq(rt_rq);

	BUG_ON(&rq->rt != rt_rq);

	if (!rt_rq->rt_queued)
		return;

	BUG_ON(!rq->nr_running);

969
	sub_nr_running(rq, rt_rq->rt_nr_running);
970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
	rt_rq->rt_queued = 0;
}

static void
enqueue_top_rt_rq(struct rt_rq *rt_rq)
{
	struct rq *rq = rq_of_rt_rq(rt_rq);

	BUG_ON(&rq->rt != rt_rq);

	if (rt_rq->rt_queued)
		return;
	if (rt_rq_throttled(rt_rq) || !rt_rq->rt_nr_running)
		return;

985
	add_nr_running(rq, rt_rq->rt_nr_running);
986 987 988
	rt_rq->rt_queued = 1;
}

989
#if defined CONFIG_SMP
990

991 992
static void
inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
993
{
G
Gregory Haskins 已提交
994
	struct rq *rq = rq_of_rt_rq(rt_rq);
995

996 997 998 999 1000 1001 1002
#ifdef CONFIG_RT_GROUP_SCHED
	/*
	 * Change rq's cpupri only if rt_rq is the top queue.
	 */
	if (&rq->rt != rt_rq)
		return;
#endif
1003 1004
	if (rq->online && prio < prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, prio);
1005
}
1006

1007 1008 1009 1010
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);
1011

1012 1013 1014 1015 1016 1017 1018
#ifdef CONFIG_RT_GROUP_SCHED
	/*
	 * Change rq's cpupri only if rt_rq is the top queue.
	 */
	if (&rq->rt != rt_rq)
		return;
#endif
1019 1020
	if (rq->online && rt_rq->highest_prio.curr != prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr);
1021 1022
}

1023 1024
#else /* CONFIG_SMP */

P
Peter Zijlstra 已提交
1025
static inline
1026 1027 1028 1029 1030
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 */
1031

1032
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
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
1049
	if (rt_rq->rt_nr_running) {
1050

1051
		WARN_ON(prio < prev_prio);
1052

1053
		/*
1054 1055
		 * This may have been our highest task, and therefore
		 * we may have some recomputation to do
1056
		 */
1057
		if (prio == prev_prio) {
1058 1059 1060
			struct rt_prio_array *array = &rt_rq->active;

			rt_rq->highest_prio.curr =
1061
				sched_find_first_bit(array->bitmap);
1062 1063
		}

1064
	} else
1065
		rt_rq->highest_prio.curr = MAX_RT_PRIO;
1066

1067 1068
	dec_rt_prio_smp(rt_rq, prio, prev_prio);
}
1069

1070 1071 1072 1073 1074 1075
#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 */
1076

1077
#ifdef CONFIG_RT_GROUP_SCHED
1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091

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
1092 1093 1094 1095
	if (rt_se_boosted(rt_se))
		rt_rq->rt_nr_boosted--;

	WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted);
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
}

#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 */

1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
static inline
unsigned int rt_se_nr_running(struct sched_rt_entity *rt_se)
{
	struct rt_rq *group_rq = group_rt_rq(rt_se);

	if (group_rq)
		return group_rq->rt_nr_running;
	else
		return 1;
}

1122 1123 1124 1125 1126 1127
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));
1128
	rt_rq->rt_nr_running += rt_se_nr_running(rt_se);
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139

	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);
1140
	rt_rq->rt_nr_running -= rt_se_nr_running(rt_se);
1141 1142 1143 1144

	dec_rt_prio(rt_rq, rt_se_prio(rt_se));
	dec_rt_migration(rt_se, rt_rq);
	dec_rt_group(rt_se, rt_rq);
1145 1146
}

1147
static void __enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head)
I
Ingo Molnar 已提交
1148
{
P
Peter Zijlstra 已提交
1149 1150 1151
	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);
1152
	struct list_head *queue = array->queue + rt_se_prio(rt_se);
I
Ingo Molnar 已提交
1153

1154 1155 1156 1157 1158 1159 1160
	/*
	 * 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
1161
		return;
1162

1163 1164 1165 1166
	if (head)
		list_add(&rt_se->run_list, queue);
	else
		list_add_tail(&rt_se->run_list, queue);
P
Peter Zijlstra 已提交
1167
	__set_bit(rt_se_prio(rt_se), array->bitmap);
1168

P
Peter Zijlstra 已提交
1169 1170 1171
	inc_rt_tasks(rt_se, rt_rq);
}

1172
static void __dequeue_rt_entity(struct sched_rt_entity *rt_se)
P
Peter Zijlstra 已提交
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187
{
	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);
}

/*
 * Because the prio of an upper entry depends on the lower
 * entries, we must remove entries top - down.
 */
1188
static void dequeue_rt_stack(struct sched_rt_entity *rt_se)
P
Peter Zijlstra 已提交
1189
{
1190
	struct sched_rt_entity *back = NULL;
P
Peter Zijlstra 已提交
1191

1192 1193 1194 1195 1196
	for_each_sched_rt_entity(rt_se) {
		rt_se->back = back;
		back = rt_se;
	}

1197 1198
	dequeue_top_rt_rq(rt_rq_of_se(back));

1199 1200
	for (rt_se = back; rt_se; rt_se = rt_se->back) {
		if (on_rt_rq(rt_se))
1201 1202 1203 1204
			__dequeue_rt_entity(rt_se);
	}
}

1205
static void enqueue_rt_entity(struct sched_rt_entity *rt_se, bool head)
1206
{
1207 1208
	struct rq *rq = rq_of_rt_se(rt_se);

1209 1210
	dequeue_rt_stack(rt_se);
	for_each_sched_rt_entity(rt_se)
1211
		__enqueue_rt_entity(rt_se, head);
1212
	enqueue_top_rt_rq(&rq->rt);
1213 1214 1215 1216
}

static void dequeue_rt_entity(struct sched_rt_entity *rt_se)
{
1217 1218
	struct rq *rq = rq_of_rt_se(rt_se);

1219 1220 1221 1222 1223 1224
	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)
1225
			__enqueue_rt_entity(rt_se, false);
1226
	}
1227
	enqueue_top_rt_rq(&rq->rt);
I
Ingo Molnar 已提交
1228 1229 1230 1231 1232
}

/*
 * Adding/removing a task to/from a priority array:
 */
1233
static void
1234
enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags)
P
Peter Zijlstra 已提交
1235 1236 1237
{
	struct sched_rt_entity *rt_se = &p->rt;

1238
	if (flags & ENQUEUE_WAKEUP)
P
Peter Zijlstra 已提交
1239 1240
		rt_se->timeout = 0;

1241
	enqueue_rt_entity(rt_se, flags & ENQUEUE_HEAD);
1242

1243
	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
1244
		enqueue_pushable_task(rq, p);
P
Peter Zijlstra 已提交
1245 1246
}

1247
static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
1248
{
P
Peter Zijlstra 已提交
1249
	struct sched_rt_entity *rt_se = &p->rt;
I
Ingo Molnar 已提交
1250

1251
	update_curr_rt(rq);
1252
	dequeue_rt_entity(rt_se);
1253

1254
	dequeue_pushable_task(rq, p);
I
Ingo Molnar 已提交
1255 1256 1257
}

/*
1258 1259
 * 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
1260
 */
1261 1262
static void
requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head)
P
Peter Zijlstra 已提交
1263
{
1264
	if (on_rt_rq(rt_se)) {
1265 1266 1267 1268 1269 1270 1271
		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);
1272
	}
P
Peter Zijlstra 已提交
1273 1274
}

1275
static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head)
I
Ingo Molnar 已提交
1276
{
P
Peter Zijlstra 已提交
1277 1278
	struct sched_rt_entity *rt_se = &p->rt;
	struct rt_rq *rt_rq;
I
Ingo Molnar 已提交
1279

P
Peter Zijlstra 已提交
1280 1281
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);
1282
		requeue_rt_entity(rt_rq, rt_se, head);
P
Peter Zijlstra 已提交
1283
	}
I
Ingo Molnar 已提交
1284 1285
}

P
Peter Zijlstra 已提交
1286
static void yield_task_rt(struct rq *rq)
I
Ingo Molnar 已提交
1287
{
1288
	requeue_task_rt(rq, rq->curr, 0);
I
Ingo Molnar 已提交
1289 1290
}

1291
#ifdef CONFIG_SMP
1292 1293
static int find_lowest_rq(struct task_struct *task);

1294
static int
1295
select_task_rq_rt(struct task_struct *p, int cpu, int sd_flag, int flags)
1296
{
1297 1298
	struct task_struct *curr;
	struct rq *rq;
1299

1300
	if (p->nr_cpus_allowed == 1)
1301 1302
		goto out;

1303 1304 1305 1306
	/* For anything but wake ups, just return the task_cpu */
	if (sd_flag != SD_BALANCE_WAKE && sd_flag != SD_BALANCE_FORK)
		goto out;

1307 1308 1309 1310 1311
	rq = cpu_rq(cpu);

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

1312
	/*
1313
	 * If the current task on @p's runqueue is an RT task, then
1314 1315 1316 1317
	 * try to see if we can wake this RT task up on another
	 * runqueue. Otherwise simply start this RT task
	 * on its current runqueue.
	 *
1318 1319 1320 1321 1322 1323 1324 1325 1326
	 * 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.
1327 1328 1329 1330 1331 1332
	 *
	 * 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.
1333
	 */
1334
	if (curr && unlikely(rt_task(curr)) &&
1335
	    (curr->nr_cpus_allowed < 2 ||
1336
	     curr->prio <= p->prio)) {
1337
		int target = find_lowest_rq(p);
1338

1339 1340
		if (target != -1)
			cpu = target;
1341
	}
1342
	rcu_read_unlock();
1343

1344
out:
1345
	return cpu;
1346
}
1347 1348 1349

static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
{
1350
	if (rq->curr->nr_cpus_allowed == 1)
1351 1352
		return;

1353
	if (p->nr_cpus_allowed != 1
1354 1355
	    && cpupri_find(&rq->rd->cpupri, p, NULL))
		return;
1356

1357 1358
	if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL))
		return;
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368

	/*
	 * 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);
}

1369 1370
#endif /* CONFIG_SMP */

I
Ingo Molnar 已提交
1371 1372 1373
/*
 * Preempt the current task with a newly woken task if needed:
 */
P
Peter Zijlstra 已提交
1374
static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flags)
I
Ingo Molnar 已提交
1375
{
1376
	if (p->prio < rq->curr->prio) {
I
Ingo Molnar 已提交
1377
		resched_task(rq->curr);
1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
		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.
	 */
1394
	if (p->prio == rq->curr->prio && !test_tsk_need_resched(rq->curr))
1395
		check_preempt_equal_prio(rq, p);
1396
#endif
I
Ingo Molnar 已提交
1397 1398
}

P
Peter Zijlstra 已提交
1399 1400
static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq,
						   struct rt_rq *rt_rq)
I
Ingo Molnar 已提交
1401
{
P
Peter Zijlstra 已提交
1402 1403
	struct rt_prio_array *array = &rt_rq->active;
	struct sched_rt_entity *next = NULL;
I
Ingo Molnar 已提交
1404 1405 1406 1407
	struct list_head *queue;
	int idx;

	idx = sched_find_first_bit(array->bitmap);
P
Peter Zijlstra 已提交
1408
	BUG_ON(idx >= MAX_RT_PRIO);
I
Ingo Molnar 已提交
1409 1410

	queue = array->queue + idx;
P
Peter Zijlstra 已提交
1411
	next = list_entry(queue->next, struct sched_rt_entity, run_list);
1412

P
Peter Zijlstra 已提交
1413 1414
	return next;
}
I
Ingo Molnar 已提交
1415

1416
static struct task_struct *_pick_next_task_rt(struct rq *rq)
P
Peter Zijlstra 已提交
1417 1418 1419
{
	struct sched_rt_entity *rt_se;
	struct task_struct *p;
1420
	struct rt_rq *rt_rq  = &rq->rt;
P
Peter Zijlstra 已提交
1421 1422 1423

	do {
		rt_se = pick_next_rt_entity(rq, rt_rq);
1424
		BUG_ON(!rt_se);
P
Peter Zijlstra 已提交
1425 1426 1427 1428
		rt_rq = group_rt_rq(rt_se);
	} while (rt_rq);

	p = rt_task_of(rt_se);
1429
	p->se.exec_start = rq_clock_task(rq);
1430 1431 1432 1433

	return p;
}

1434 1435
static struct task_struct *
pick_next_task_rt(struct rq *rq, struct task_struct *prev)
1436
{
1437 1438 1439
	struct task_struct *p;
	struct rt_rq *rt_rq = &rq->rt;

1440
	if (need_pull_rt_task(rq, prev)) {
1441
		pull_rt_task(rq);
1442 1443
		/*
		 * pull_rt_task() can drop (and re-acquire) rq->lock; this
1444 1445
		 * means a dl or stop task can slip in, in which case we need
		 * to re-start task selection.
1446
		 */
1447 1448
		if (unlikely((rq->stop && rq->stop->on_rq) ||
			     rq->dl.dl_nr_running))
1449 1450
			return RETRY_TASK;
	}
1451

1452 1453 1454 1455 1456 1457 1458
	/*
	 * We may dequeue prev's rt_rq in put_prev_task().
	 * So, we update time before rt_nr_running check.
	 */
	if (prev->sched_class == &rt_sched_class)
		update_curr_rt(rq);

1459
	if (!rt_rq->rt_queued)
1460 1461
		return NULL;

1462
	put_prev_task(rq, prev);
1463 1464

	p = _pick_next_task_rt(rq);
1465 1466 1467 1468 1469

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

P
Peter Zijlstra 已提交
1470
	set_post_schedule(rq);
1471

P
Peter Zijlstra 已提交
1472
	return p;
I
Ingo Molnar 已提交
1473 1474
}

1475
static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
I
Ingo Molnar 已提交
1476
{
1477
	update_curr_rt(rq);
1478 1479 1480 1481 1482

	/*
	 * The previous task needs to be made eligible for pushing
	 * if it is still active
	 */
1483
	if (on_rt_rq(&p->rt) && p->nr_cpus_allowed > 1)
1484
		enqueue_pushable_task(rq, p);
I
Ingo Molnar 已提交
1485 1486
}

1487
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1488

S
Steven Rostedt 已提交
1489 1490 1491
/* Only try algorithms three times */
#define RT_MAX_TRIES 3

1492 1493 1494
static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu)
{
	if (!task_running(rq, p) &&
1495
	    cpumask_test_cpu(cpu, tsk_cpus_allowed(p)))
1496 1497 1498 1499
		return 1;
	return 0;
}

1500 1501 1502 1503 1504
/*
 * Return the highest pushable rq's task, which is suitable to be executed
 * on the cpu, NULL otherwise
 */
static struct task_struct *pick_highest_pushable_task(struct rq *rq, int cpu)
S
Steven Rostedt 已提交
1505
{
1506 1507
	struct plist_head *head = &rq->rt.pushable_tasks;
	struct task_struct *p;
1508

1509 1510
	if (!has_pushable_tasks(rq))
		return NULL;
1511

1512 1513 1514
	plist_for_each_entry(p, head, pushable_tasks) {
		if (pick_rt_task(rq, p, cpu))
			return p;
1515 1516
	}

1517
	return NULL;
S
Steven Rostedt 已提交
1518 1519
}

1520
static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask);
S
Steven Rostedt 已提交
1521

G
Gregory Haskins 已提交
1522 1523 1524
static int find_lowest_rq(struct task_struct *task)
{
	struct sched_domain *sd;
1525
	struct cpumask *lowest_mask = __get_cpu_var(local_cpu_mask);
G
Gregory Haskins 已提交
1526 1527
	int this_cpu = smp_processor_id();
	int cpu      = task_cpu(task);
G
Gregory Haskins 已提交
1528

1529 1530 1531 1532
	/* Make sure the mask is initialized first */
	if (unlikely(!lowest_mask))
		return -1;

1533
	if (task->nr_cpus_allowed == 1)
1534
		return -1; /* No other targets possible */
G
Gregory Haskins 已提交
1535

1536 1537
	if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask))
		return -1; /* No targets found */
G
Gregory Haskins 已提交
1538 1539 1540 1541 1542 1543 1544 1545 1546

	/*
	 * 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.
	 */
1547
	if (cpumask_test_cpu(cpu, lowest_mask))
G
Gregory Haskins 已提交
1548 1549 1550 1551 1552 1553
		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
1554 1555
	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
1556

1557
	rcu_read_lock();
R
Rusty Russell 已提交
1558 1559 1560
	for_each_domain(cpu, sd) {
		if (sd->flags & SD_WAKE_AFFINE) {
			int best_cpu;
G
Gregory Haskins 已提交
1561

R
Rusty Russell 已提交
1562 1563 1564 1565 1566
			/*
			 * "this_cpu" is cheaper to preempt than a
			 * remote processor.
			 */
			if (this_cpu != -1 &&
1567 1568
			    cpumask_test_cpu(this_cpu, sched_domain_span(sd))) {
				rcu_read_unlock();
R
Rusty Russell 已提交
1569
				return this_cpu;
1570
			}
R
Rusty Russell 已提交
1571 1572 1573

			best_cpu = cpumask_first_and(lowest_mask,
						     sched_domain_span(sd));
1574 1575
			if (best_cpu < nr_cpu_ids) {
				rcu_read_unlock();
R
Rusty Russell 已提交
1576
				return best_cpu;
1577
			}
G
Gregory Haskins 已提交
1578 1579
		}
	}
1580
	rcu_read_unlock();
G
Gregory Haskins 已提交
1581 1582 1583 1584 1585 1586

	/*
	 * 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
1587 1588 1589 1590 1591 1592 1593
	if (this_cpu != -1)
		return this_cpu;

	cpu = cpumask_any(lowest_mask);
	if (cpu < nr_cpu_ids)
		return cpu;
	return -1;
1594 1595 1596
}

/* Will lock the rq it finds */
1597
static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq)
1598 1599 1600
{
	struct rq *lowest_rq = NULL;
	int tries;
1601
	int cpu;
S
Steven Rostedt 已提交
1602

1603 1604 1605
	for (tries = 0; tries < RT_MAX_TRIES; tries++) {
		cpu = find_lowest_rq(task);

1606
		if ((cpu == -1) || (cpu == rq->cpu))
S
Steven Rostedt 已提交
1607 1608
			break;

1609 1610
		lowest_rq = cpu_rq(cpu);

S
Steven Rostedt 已提交
1611
		/* if the prio of this runqueue changed, try again */
1612
		if (double_lock_balance(rq, lowest_rq)) {
S
Steven Rostedt 已提交
1613 1614 1615 1616 1617 1618
			/*
			 * 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.
			 */
1619
			if (unlikely(task_rq(task) != rq ||
1620
				     !cpumask_test_cpu(lowest_rq->cpu,
1621
						       tsk_cpus_allowed(task)) ||
1622
				     task_running(rq, task) ||
P
Peter Zijlstra 已提交
1623
				     !task->on_rq)) {
1624

1625
				double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1626 1627 1628 1629 1630 1631
				lowest_rq = NULL;
				break;
			}
		}

		/* If this rq is still suitable use it. */
1632
		if (lowest_rq->rt.highest_prio.curr > task->prio)
S
Steven Rostedt 已提交
1633 1634 1635
			break;

		/* try again */
1636
		double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1637 1638 1639 1640 1641 1642
		lowest_rq = NULL;
	}

	return lowest_rq;
}

1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654
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));
1655
	BUG_ON(p->nr_cpus_allowed <= 1);
1656

P
Peter Zijlstra 已提交
1657
	BUG_ON(!p->on_rq);
1658 1659 1660 1661 1662
	BUG_ON(!rt_task(p));

	return p;
}

S
Steven Rostedt 已提交
1663 1664 1665 1666 1667
/*
 * 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.
 */
1668
static int push_rt_task(struct rq *rq)
S
Steven Rostedt 已提交
1669 1670 1671
{
	struct task_struct *next_task;
	struct rq *lowest_rq;
1672
	int ret = 0;
S
Steven Rostedt 已提交
1673

G
Gregory Haskins 已提交
1674 1675 1676
	if (!rq->rt.overloaded)
		return 0;

1677
	next_task = pick_next_pushable_task(rq);
S
Steven Rostedt 已提交
1678 1679 1680
	if (!next_task)
		return 0;

P
Peter Zijlstra 已提交
1681
retry:
1682
	if (unlikely(next_task == rq->curr)) {
1683
		WARN_ON(1);
S
Steven Rostedt 已提交
1684
		return 0;
1685
	}
S
Steven Rostedt 已提交
1686 1687 1688 1689 1690 1691

	/*
	 * It's possible that the next_task slipped in of
	 * higher priority than current. If that's the case
	 * just reschedule current.
	 */
1692 1693
	if (unlikely(next_task->prio < rq->curr->prio)) {
		resched_task(rq->curr);
S
Steven Rostedt 已提交
1694 1695 1696
		return 0;
	}

1697
	/* We might release rq lock */
S
Steven Rostedt 已提交
1698 1699 1700
	get_task_struct(next_task);

	/* find_lock_lowest_rq locks the rq if found */
1701
	lowest_rq = find_lock_lowest_rq(next_task, rq);
S
Steven Rostedt 已提交
1702 1703 1704
	if (!lowest_rq) {
		struct task_struct *task;
		/*
1705
		 * find_lock_lowest_rq releases rq->lock
1706 1707 1708 1709 1710
		 * 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
1711
		 */
1712
		task = pick_next_pushable_task(rq);
1713 1714
		if (task_cpu(next_task) == rq->cpu && task == next_task) {
			/*
1715 1716 1717 1718
			 * 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.
1719 1720
			 */
			goto out;
S
Steven Rostedt 已提交
1721
		}
1722

1723 1724 1725 1726
		if (!task)
			/* No more tasks, just exit */
			goto out;

1727
		/*
1728
		 * Something has shifted, try again.
1729
		 */
1730 1731 1732
		put_task_struct(next_task);
		next_task = task;
		goto retry;
S
Steven Rostedt 已提交
1733 1734
	}

1735
	deactivate_task(rq, next_task, 0);
S
Steven Rostedt 已提交
1736 1737
	set_task_cpu(next_task, lowest_rq->cpu);
	activate_task(lowest_rq, next_task, 0);
1738
	ret = 1;
S
Steven Rostedt 已提交
1739 1740 1741

	resched_task(lowest_rq->curr);

1742
	double_unlock_balance(rq, lowest_rq);
S
Steven Rostedt 已提交
1743 1744 1745 1746

out:
	put_task_struct(next_task);

1747
	return ret;
S
Steven Rostedt 已提交
1748 1749 1750 1751 1752 1753 1754 1755 1756
}

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

1757 1758
static int pull_rt_task(struct rq *this_rq)
{
I
Ingo Molnar 已提交
1759
	int this_cpu = this_rq->cpu, ret = 0, cpu;
1760
	struct task_struct *p;
1761 1762
	struct rq *src_rq;

1763
	if (likely(!rt_overloaded(this_rq)))
1764 1765
		return 0;

P
Peter Zijlstra 已提交
1766 1767 1768 1769 1770 1771
	/*
	 * Match the barrier from rt_set_overloaded; this guarantees that if we
	 * see overloaded we must also see the rto_mask bit.
	 */
	smp_rmb();

1772
	for_each_cpu(cpu, this_rq->rd->rto_mask) {
1773 1774 1775 1776
		if (this_cpu == cpu)
			continue;

		src_rq = cpu_rq(cpu);
1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788

		/*
		 * 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;

1789 1790 1791
		/*
		 * We can potentially drop this_rq's lock in
		 * double_lock_balance, and another CPU could
1792
		 * alter this_rq
1793
		 */
1794
		double_lock_balance(this_rq, src_rq);
1795 1796

		/*
1797 1798
		 * We can pull only a task, which is pushable
		 * on its rq, and no others.
1799
		 */
1800
		p = pick_highest_pushable_task(src_rq, this_cpu);
1801 1802 1803 1804 1805

		/*
		 * Do we have an RT task that preempts
		 * the to-be-scheduled task?
		 */
1806
		if (p && (p->prio < this_rq->rt.highest_prio.curr)) {
1807
			WARN_ON(p == src_rq->curr);
P
Peter Zijlstra 已提交
1808
			WARN_ON(!p->on_rq);
1809 1810 1811 1812 1813 1814 1815

			/*
			 * 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
1816
			 * current task on the run queue
1817
			 */
1818
			if (p->prio < src_rq->curr->prio)
M
Mike Galbraith 已提交
1819
				goto skip;
1820 1821 1822 1823 1824 1825 1826 1827 1828

			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
1829
			 * in another runqueue. (low likelihood
1830 1831 1832
			 * but possible)
			 */
		}
P
Peter Zijlstra 已提交
1833
skip:
1834
		double_unlock_balance(this_rq, src_rq);
1835 1836 1837 1838 1839
	}

	return ret;
}

1840
static void post_schedule_rt(struct rq *rq)
S
Steven Rostedt 已提交
1841
{
1842
	push_rt_tasks(rq);
S
Steven Rostedt 已提交
1843 1844
}

1845 1846 1847 1848
/*
 * If we are not running and we are not going to reschedule soon, we should
 * try to push tasks away now
 */
1849
static void task_woken_rt(struct rq *rq, struct task_struct *p)
1850
{
1851
	if (!task_running(rq, p) &&
1852
	    !test_tsk_need_resched(rq->curr) &&
1853
	    has_pushable_tasks(rq) &&
1854
	    p->nr_cpus_allowed > 1 &&
1855
	    (dl_task(rq->curr) || rt_task(rq->curr)) &&
1856
	    (rq->curr->nr_cpus_allowed < 2 ||
1857
	     rq->curr->prio <= p->prio))
1858 1859 1860
		push_rt_tasks(rq);
}

1861
static void set_cpus_allowed_rt(struct task_struct *p,
1862
				const struct cpumask *new_mask)
1863
{
1864 1865
	struct rq *rq;
	int weight;
1866 1867 1868

	BUG_ON(!rt_task(p));

1869 1870
	if (!p->on_rq)
		return;
1871

1872
	weight = cpumask_weight(new_mask);
1873

1874 1875 1876 1877
	/*
	 * Only update if the process changes its state from whether it
	 * can migrate or not.
	 */
1878
	if ((p->nr_cpus_allowed > 1) == (weight > 1))
1879
		return;
1880

1881
	rq = task_rq(p);
1882

1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
	/*
	 * 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++;
1895
	}
1896 1897

	update_rt_migration(&rq->rt);
1898
}
1899

1900
/* Assumes rq->lock is held */
1901
static void rq_online_rt(struct rq *rq)
1902 1903 1904
{
	if (rq->rt.overloaded)
		rt_set_overload(rq);
1905

P
Peter Zijlstra 已提交
1906 1907
	__enable_runtime(rq);

1908
	cpupri_set(&rq->rd->cpupri, rq->cpu, rq->rt.highest_prio.curr);
1909 1910 1911
}

/* Assumes rq->lock is held */
1912
static void rq_offline_rt(struct rq *rq)
1913 1914 1915
{
	if (rq->rt.overloaded)
		rt_clear_overload(rq);
1916

P
Peter Zijlstra 已提交
1917 1918
	__disable_runtime(rq);

1919
	cpupri_set(&rq->rd->cpupri, rq->cpu, CPUPRI_INVALID);
1920
}
1921 1922 1923 1924 1925

/*
 * 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
1926
static void switched_from_rt(struct rq *rq, struct task_struct *p)
1927 1928 1929 1930 1931 1932 1933 1934
{
	/*
	 * 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.
	 */
1935 1936 1937 1938 1939
	if (!p->on_rq || rq->rt.rt_nr_running)
		return;

	if (pull_rt_task(rq))
		resched_task(rq->curr);
1940
}
1941

1942
void __init init_sched_rt_class(void)
1943 1944 1945
{
	unsigned int i;

1946
	for_each_possible_cpu(i) {
1947
		zalloc_cpumask_var_node(&per_cpu(local_cpu_mask, i),
1948
					GFP_KERNEL, cpu_to_node(i));
1949
	}
1950
}
1951 1952 1953 1954 1955 1956 1957
#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
1958
static void switched_to_rt(struct rq *rq, struct task_struct *p)
1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
{
	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
1969
	if (p->on_rq && rq->curr != p) {
1970
#ifdef CONFIG_SMP
1971
		if (p->nr_cpus_allowed > 1 && rq->rt.overloaded &&
1972
		    /* Don't resched if we changed runqueues */
1973
		    push_rt_task(rq) && rq != task_rq(p))
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
			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
1985 1986
static void
prio_changed_rt(struct rq *rq, struct task_struct *p, int oldprio)
1987
{
P
Peter Zijlstra 已提交
1988
	if (!p->on_rq)
P
Peter Zijlstra 已提交
1989 1990 1991
		return;

	if (rq->curr == p) {
1992 1993 1994 1995 1996 1997 1998 1999 2000
#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
2001 2002 2003
		 * 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.
2004
		 */
2005
		if (p->prio > rq->rt.highest_prio.curr && rq->curr == p)
2006 2007 2008 2009 2010
			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
2011
#endif /* CONFIG_SMP */
2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
	} 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);
	}
}

2023 2024 2025 2026
static void watchdog(struct rq *rq, struct task_struct *p)
{
	unsigned long soft, hard;

2027 2028 2029
	/* 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);
2030 2031 2032 2033

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

2034 2035 2036 2037 2038
		if (p->rt.watchdog_stamp != jiffies) {
			p->rt.timeout++;
			p->rt.watchdog_stamp = jiffies;
		}

2039
		next = DIV_ROUND_UP(min(soft, hard), USEC_PER_SEC/HZ);
2040
		if (p->rt.timeout > next)
2041
			p->cputime_expires.sched_exp = p->se.sum_exec_runtime;
2042 2043
	}
}
I
Ingo Molnar 已提交
2044

P
Peter Zijlstra 已提交
2045
static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued)
I
Ingo Molnar 已提交
2046
{
2047 2048
	struct sched_rt_entity *rt_se = &p->rt;

2049 2050
	update_curr_rt(rq);

2051 2052
	watchdog(rq, p);

I
Ingo Molnar 已提交
2053 2054 2055 2056 2057 2058 2059
	/*
	 * 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
2060
	if (--p->rt.time_slice)
I
Ingo Molnar 已提交
2061 2062
		return;

2063
	p->rt.time_slice = sched_rr_timeslice;
I
Ingo Molnar 已提交
2064

2065
	/*
L
Li Bin 已提交
2066 2067
	 * Requeue to the end of queue if we (and all of our ancestors) are not
	 * the only element on the queue
2068
	 */
2069 2070 2071 2072 2073 2074
	for_each_sched_rt_entity(rt_se) {
		if (rt_se->run_list.prev != rt_se->run_list.next) {
			requeue_task_rt(rq, p, 0);
			set_tsk_need_resched(p);
			return;
		}
2075
	}
I
Ingo Molnar 已提交
2076 2077
}

2078 2079 2080 2081
static void set_curr_task_rt(struct rq *rq)
{
	struct task_struct *p = rq->curr;

2082
	p->se.exec_start = rq_clock_task(rq);
2083 2084 2085

	/* The running task is never eligible for pushing */
	dequeue_pushable_task(rq, p);
2086 2087
}

2088
static unsigned int get_rr_interval_rt(struct rq *rq, struct task_struct *task)
2089 2090 2091 2092 2093
{
	/*
	 * Time slice is 0 for SCHED_FIFO tasks
	 */
	if (task->policy == SCHED_RR)
2094
		return sched_rr_timeslice;
2095 2096 2097 2098
	else
		return 0;
}

2099
const struct sched_class rt_sched_class = {
2100
	.next			= &fair_sched_class,
I
Ingo Molnar 已提交
2101 2102 2103 2104 2105 2106 2107 2108 2109
	.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,

2110
#ifdef CONFIG_SMP
L
Li Zefan 已提交
2111 2112
	.select_task_rq		= select_task_rq_rt,

2113
	.set_cpus_allowed       = set_cpus_allowed_rt,
2114 2115
	.rq_online              = rq_online_rt,
	.rq_offline             = rq_offline_rt,
2116
	.post_schedule		= post_schedule_rt,
2117
	.task_woken		= task_woken_rt,
2118
	.switched_from		= switched_from_rt,
2119
#endif
I
Ingo Molnar 已提交
2120

2121
	.set_curr_task          = set_curr_task_rt,
I
Ingo Molnar 已提交
2122
	.task_tick		= task_tick_rt,
2123

2124 2125
	.get_rr_interval	= get_rr_interval_rt,

2126 2127
	.prio_changed		= prio_changed_rt,
	.switched_to		= switched_to_rt,
I
Ingo Molnar 已提交
2128
};
2129 2130 2131 2132

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

2133
void print_rt_stats(struct seq_file *m, int cpu)
2134
{
C
Cheng Xu 已提交
2135
	rt_rq_iter_t iter;
2136 2137 2138
	struct rt_rq *rt_rq;

	rcu_read_lock();
C
Cheng Xu 已提交
2139
	for_each_rt_rq(rt_rq, iter, cpu_rq(cpu))
2140 2141 2142
		print_rt_rq(m, cpu, rt_rq);
	rcu_read_unlock();
}
2143
#endif /* CONFIG_SCHED_DEBUG */