sched_rt.c 40.3 KB
Newer Older
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
static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
{
	return container_of(rt_se, struct task_struct, rt);
}

#ifdef CONFIG_RT_GROUP_SCHED

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

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

#else /* CONFIG_RT_GROUP_SCHED */

static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
{
	return container_of(rt_rq, struct rq, rt);
}

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

	return &rq->rt;
}

#endif /* CONFIG_RT_GROUP_SCHED */

40
#ifdef CONFIG_SMP
41

42
static inline int rt_overloaded(struct rq *rq)
43
{
44
	return atomic_read(&rq->rd->rto_count);
45
}
46

47
48
static inline void rt_set_overload(struct rq *rq)
{
49
50
51
	if (!rq->online)
		return;

52
	cpumask_set_cpu(rq->cpu, rq->rd->rto_mask);
53
54
55
56
57
58
59
60
	/*
	 * 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();
61
	atomic_inc(&rq->rd->rto_count);
62
}
63

64
65
static inline void rt_clear_overload(struct rq *rq)
{
66
67
68
	if (!rq->online)
		return;

69
	/* the order here really doesn't matter */
70
	atomic_dec(&rq->rd->rto_count);
71
	cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask);
72
}
73

74
static void update_rt_migration(struct rt_rq *rt_rq)
75
{
76
77
78
79
	if (rt_rq->rt_nr_migratory && (rt_rq->rt_nr_running > 1)) {
		if (!rt_rq->overloaded) {
			rt_set_overload(rq_of_rt_rq(rt_rq));
			rt_rq->overloaded = 1;
80
		}
81
82
83
	} else if (rt_rq->overloaded) {
		rt_clear_overload(rq_of_rt_rq(rt_rq));
		rt_rq->overloaded = 0;
84
	}
85
}
86

87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
	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)
{
	if (rt_se->nr_cpus_allowed > 1)
		rt_rq->rt_nr_migratory--;

	update_rt_migration(rt_rq);
}

103
104
105
106
107
108
109
110
111
112
113
114
115
116
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);
}

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

#else

117
static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
Peter Zijlstra's avatar
Peter Zijlstra committed
118
{
Peter Zijlstra's avatar
Peter Zijlstra committed
119
120
}

121
122
123
124
static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p)
{
}

125
static inline
126
127
128
129
void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}

130
static inline
131
132
133
void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
{
}
134

135
136
#endif /* CONFIG_SMP */

Peter Zijlstra's avatar
Peter Zijlstra committed
137
138
139
140
141
static inline int on_rt_rq(struct sched_rt_entity *rt_se)
{
	return !list_empty(&rt_se->run_list);
}

142
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
143

Peter Zijlstra's avatar
Peter Zijlstra committed
144
static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
145
146
{
	if (!rt_rq->tg)
Peter Zijlstra's avatar
Peter Zijlstra committed
147
		return RUNTIME_INF;
Peter Zijlstra's avatar
Peter Zijlstra committed
148

149
150
151
152
153
154
	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);
Peter Zijlstra's avatar
Peter Zijlstra committed
155
156
157
}

#define for_each_leaf_rt_rq(rt_rq, rq) \
158
	list_for_each_entry_rcu(rt_rq, &rq->leaf_rt_rq_list, leaf_rt_rq_list)
Peter Zijlstra's avatar
Peter Zijlstra committed
159
160
161
162
163
164
165
166
167
168
169
170

#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;
}

static void enqueue_rt_entity(struct sched_rt_entity *rt_se);
static void dequeue_rt_entity(struct sched_rt_entity *rt_se);

Peter Zijlstra's avatar
Peter Zijlstra committed
171
static void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
172
{
173
	struct task_struct *curr = rq_of_rt_rq(rt_rq)->curr;
Peter Zijlstra's avatar
Peter Zijlstra committed
174
175
	struct sched_rt_entity *rt_se = rt_rq->rt_se;

176
177
178
	if (rt_rq->rt_nr_running) {
		if (rt_se && !on_rt_rq(rt_se))
			enqueue_rt_entity(rt_se);
179
		if (rt_rq->highest_prio.curr < curr->prio)
180
			resched_task(curr);
Peter Zijlstra's avatar
Peter Zijlstra committed
181
182
183
	}
}

Peter Zijlstra's avatar
Peter Zijlstra committed
184
static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
185
186
187
188
189
190
191
{
	struct sched_rt_entity *rt_se = rt_rq->rt_se;

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

Peter Zijlstra's avatar
Peter Zijlstra committed
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
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;
}

209
#ifdef CONFIG_SMP
210
static inline const struct cpumask *sched_rt_period_mask(void)
211
212
213
{
	return cpu_rq(smp_processor_id())->rd->span;
}
Peter Zijlstra's avatar
Peter Zijlstra committed
214
#else
215
static inline const struct cpumask *sched_rt_period_mask(void)
216
{
217
	return cpu_online_mask;
218
219
}
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
220

221
222
static inline
struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
223
{
224
225
	return container_of(rt_b, struct task_group, rt_bandwidth)->rt_rq[cpu];
}
Peter Zijlstra's avatar
Peter Zijlstra committed
226

227
228
229
230
231
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &rt_rq->tg->rt_bandwidth;
}

232
#else /* !CONFIG_RT_GROUP_SCHED */
233
234
235

static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
{
236
237
238
239
240
241
	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);
Peter Zijlstra's avatar
Peter Zijlstra committed
242
243
244
245
246
247
248
249
250
251
252
253
254
}

#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;
}

Peter Zijlstra's avatar
Peter Zijlstra committed
255
static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
256
{
257
258
	if (rt_rq->rt_nr_running)
		resched_task(rq_of_rt_rq(rt_rq)->curr);
Peter Zijlstra's avatar
Peter Zijlstra committed
259
260
}

Peter Zijlstra's avatar
Peter Zijlstra committed
261
static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
262
263
264
{
}

Peter Zijlstra's avatar
Peter Zijlstra committed
265
266
267
268
static inline int rt_rq_throttled(struct rt_rq *rt_rq)
{
	return rt_rq->rt_throttled;
}
269

270
static inline const struct cpumask *sched_rt_period_mask(void)
271
{
272
	return cpu_online_mask;
273
274
275
276
277
278
279
280
}

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

281
282
283
284
285
static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq)
{
	return &def_rt_bandwidth;
}

286
#endif /* CONFIG_RT_GROUP_SCHED */
287

288
#ifdef CONFIG_SMP
289
290
291
/*
 * We ran out of runtime, see if we can borrow some from our neighbours.
 */
292
static int do_balance_runtime(struct rt_rq *rt_rq)
293
294
295
296
297
298
{
	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;

299
	weight = cpumask_weight(rd->span);
300
301
302

	spin_lock(&rt_b->rt_runtime_lock);
	rt_period = ktime_to_ns(rt_b->rt_period);
303
	for_each_cpu(i, rd->span) {
304
305
306
307
308
309
310
		struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
		s64 diff;

		if (iter == rt_rq)
			continue;

		spin_lock(&iter->rt_runtime_lock);
311
312
313
314
315
		/*
		 * 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.
		 */
316
317
318
		if (iter->rt_runtime == RUNTIME_INF)
			goto next;

319
320
321
322
		/*
		 * From runqueues with spare time, take 1/n part of their
		 * spare time, but no more than our period.
		 */
323
324
		diff = iter->rt_runtime - iter->rt_time;
		if (diff > 0) {
325
			diff = div_u64((u64)diff, weight);
326
327
328
329
330
331
332
333
334
335
			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) {
				spin_unlock(&iter->rt_runtime_lock);
				break;
			}
		}
336
next:
337
338
339
340
341
342
		spin_unlock(&iter->rt_runtime_lock);
	}
	spin_unlock(&rt_b->rt_runtime_lock);

	return more;
}
343

344
345
346
/*
 * Ensure this RQ takes back all the runtime it lend to its neighbours.
 */
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
static void __disable_runtime(struct rq *rq)
{
	struct root_domain *rd = rq->rd;
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

	for_each_leaf_rt_rq(rt_rq, rq) {
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);
		s64 want;
		int i;

		spin_lock(&rt_b->rt_runtime_lock);
		spin_lock(&rt_rq->rt_runtime_lock);
362
363
364
365
366
		/*
		 * 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.
		 */
367
368
369
370
371
		if (rt_rq->rt_runtime == RUNTIME_INF ||
				rt_rq->rt_runtime == rt_b->rt_runtime)
			goto balanced;
		spin_unlock(&rt_rq->rt_runtime_lock);

372
373
374
375
376
		/*
		 * 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.
		 */
377
378
		want = rt_b->rt_runtime - rt_rq->rt_runtime;

379
380
381
		/*
		 * Greedy reclaim, take back as much as we can.
		 */
382
		for_each_cpu(i, rd->span) {
383
384
385
			struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
			s64 diff;

386
387
388
			/*
			 * Can't reclaim from ourselves or disabled runqueues.
			 */
389
			if (iter == rt_rq || iter->rt_runtime == RUNTIME_INF)
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
				continue;

			spin_lock(&iter->rt_runtime_lock);
			if (want > 0) {
				diff = min_t(s64, iter->rt_runtime, want);
				iter->rt_runtime -= diff;
				want -= diff;
			} else {
				iter->rt_runtime -= want;
				want -= want;
			}
			spin_unlock(&iter->rt_runtime_lock);

			if (!want)
				break;
		}

		spin_lock(&rt_rq->rt_runtime_lock);
408
409
410
411
		/*
		 * We cannot be left wanting - that would mean some runtime
		 * leaked out of the system.
		 */
412
413
		BUG_ON(want);
balanced:
414
415
416
417
		/*
		 * Disable all the borrow logic by pretending we have inf
		 * runtime - in which case borrowing doesn't make sense.
		 */
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
		rt_rq->rt_runtime = RUNTIME_INF;
		spin_unlock(&rt_rq->rt_runtime_lock);
		spin_unlock(&rt_b->rt_runtime_lock);
	}
}

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

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

static void __enable_runtime(struct rq *rq)
{
	struct rt_rq *rt_rq;

	if (unlikely(!scheduler_running))
		return;

440
441
442
	/*
	 * Reset each runqueue's bandwidth settings
	 */
443
444
445
446
447
448
449
	for_each_leaf_rt_rq(rt_rq, rq) {
		struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq);

		spin_lock(&rt_b->rt_runtime_lock);
		spin_lock(&rt_rq->rt_runtime_lock);
		rt_rq->rt_runtime = rt_b->rt_runtime;
		rt_rq->rt_time = 0;
450
		rt_rq->rt_throttled = 0;
451
452
453
454
455
456
457
458
459
460
461
462
463
464
		spin_unlock(&rt_rq->rt_runtime_lock);
		spin_unlock(&rt_b->rt_runtime_lock);
	}
}

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

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

465
466
467
468
469
470
471
472
473
474
475
476
static int balance_runtime(struct rt_rq *rt_rq)
{
	int more = 0;

	if (rt_rq->rt_time > rt_rq->rt_runtime) {
		spin_unlock(&rt_rq->rt_runtime_lock);
		more = do_balance_runtime(rt_rq);
		spin_lock(&rt_rq->rt_runtime_lock);
	}

	return more;
}
477
#else /* !CONFIG_SMP */
478
479
480
481
static inline int balance_runtime(struct rt_rq *rt_rq)
{
	return 0;
}
482
#endif /* CONFIG_SMP */
483

484
485
486
static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
{
	int i, idle = 1;
487
	const struct cpumask *span;
488

489
	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
490
491
492
		return 1;

	span = sched_rt_period_mask();
493
	for_each_cpu(i, span) {
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
		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);

		spin_lock(&rq->lock);
		if (rt_rq->rt_time) {
			u64 runtime;

			spin_lock(&rt_rq->rt_runtime_lock);
			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;
			}
			if (rt_rq->rt_time || rt_rq->rt_nr_running)
				idle = 0;
			spin_unlock(&rt_rq->rt_runtime_lock);
514
515
		} else if (rt_rq->rt_nr_running)
			idle = 0;
516
517
518
519
520
521
522
523

		if (enqueue)
			sched_rt_rq_enqueue(rt_rq);
		spin_unlock(&rq->lock);
	}

	return idle;
}
524

Peter Zijlstra's avatar
Peter Zijlstra committed
525
526
static inline int rt_se_prio(struct sched_rt_entity *rt_se)
{
527
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
528
529
530
	struct rt_rq *rt_rq = group_rt_rq(rt_se);

	if (rt_rq)
531
		return rt_rq->highest_prio.curr;
Peter Zijlstra's avatar
Peter Zijlstra committed
532
533
534
535
536
#endif

	return rt_task_of(rt_se)->prio;
}

Peter Zijlstra's avatar
Peter Zijlstra committed
537
static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
538
{
Peter Zijlstra's avatar
Peter Zijlstra committed
539
	u64 runtime = sched_rt_runtime(rt_rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
540
541

	if (rt_rq->rt_throttled)
Peter Zijlstra's avatar
Peter Zijlstra committed
542
		return rt_rq_throttled(rt_rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
543

544
545
546
	if (sched_rt_runtime(rt_rq) >= sched_rt_period(rt_rq))
		return 0;

547
548
549
550
	balance_runtime(rt_rq);
	runtime = sched_rt_runtime(rt_rq);
	if (runtime == RUNTIME_INF)
		return 0;
551

Peter Zijlstra's avatar
Peter Zijlstra committed
552
	if (rt_rq->rt_time > runtime) {
Peter Zijlstra's avatar
Peter Zijlstra committed
553
		rt_rq->rt_throttled = 1;
Peter Zijlstra's avatar
Peter Zijlstra committed
554
		if (rt_rq_throttled(rt_rq)) {
Peter Zijlstra's avatar
Peter Zijlstra committed
555
			sched_rt_rq_dequeue(rt_rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
556
557
			return 1;
		}
Peter Zijlstra's avatar
Peter Zijlstra committed
558
559
560
561
562
	}

	return 0;
}

563
564
565
566
/*
 * Update the current task's runtime statistics. Skip current tasks that
 * are not in our scheduling class.
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
567
static void update_curr_rt(struct rq *rq)
568
569
{
	struct task_struct *curr = rq->curr;
Peter Zijlstra's avatar
Peter Zijlstra committed
570
571
	struct sched_rt_entity *rt_se = &curr->rt;
	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
572
573
574
575
576
	u64 delta_exec;

	if (!task_has_rt_policy(curr))
		return;

577
	delta_exec = rq->clock - curr->se.exec_start;
578
579
	if (unlikely((s64)delta_exec < 0))
		delta_exec = 0;
Ingo Molnar's avatar
Ingo Molnar committed
580
581

	schedstat_set(curr->se.exec_max, max(curr->se.exec_max, delta_exec));
582
583

	curr->se.sum_exec_runtime += delta_exec;
584
585
	account_group_exec_runtime(curr, delta_exec);

586
	curr->se.exec_start = rq->clock;
587
	cpuacct_charge(curr, delta_exec);
Peter Zijlstra's avatar
Peter Zijlstra committed
588

589
590
591
	if (!rt_bandwidth_enabled())
		return;

Dhaval Giani's avatar
Dhaval Giani committed
592
593
594
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);

595
		if (sched_rt_runtime(rt_rq) != RUNTIME_INF) {
596
			spin_lock(&rt_rq->rt_runtime_lock);
597
598
599
			rt_rq->rt_time += delta_exec;
			if (sched_rt_runtime_exceeded(rt_rq))
				resched_task(curr);
600
			spin_unlock(&rt_rq->rt_runtime_lock);
601
		}
Dhaval Giani's avatar
Dhaval Giani committed
602
	}
603
604
}

605
#if defined CONFIG_SMP
606
607
608
609

static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu);

static inline int next_prio(struct rq *rq)
610
{
611
612
613
614
615
616
617
618
	struct task_struct *next = pick_next_highest_task_rt(rq, rq->cpu);

	if (next && rt_prio(next->prio))
		return next->prio;
	else
		return MAX_RT_PRIO;
}

619
620
static void
inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
621
{
622
	struct rq *rq = rq_of_rt_rq(rt_rq);
623

624
	if (prio < prev_prio) {
625

626
627
		/*
		 * If the new task is higher in priority than anything on the
628
629
		 * run-queue, we know that the previous high becomes our
		 * next-highest.
630
		 */
631
		rt_rq->highest_prio.next = prev_prio;
632
633

		if (rq->online)
634
			cpupri_set(&rq->rd->cpupri, rq->cpu, prio);
Ingo Molnar's avatar
Ingo Molnar committed
635

636
637
638
639
640
641
642
643
644
645
646
647
	} else if (prio == rt_rq->highest_prio.curr)
		/*
		 * If the next task is equal in priority to the highest on
		 * the run-queue, then we implicitly know that the next highest
		 * task cannot be any lower than current
		 */
		rt_rq->highest_prio.next = prio;
	else if (prio < rt_rq->highest_prio.next)
		/*
		 * Otherwise, we need to recompute next-highest
		 */
		rt_rq->highest_prio.next = next_prio(rq);
648
}
649

650
651
652
653
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);
654

655
656
657
658
659
	if (rt_rq->rt_nr_running && (prio <= rt_rq->highest_prio.next))
		rt_rq->highest_prio.next = next_prio(rq);

	if (rq->online && rt_rq->highest_prio.curr != prev_prio)
		cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr);
660
661
}

662
663
#else /* CONFIG_SMP */

Peter Zijlstra's avatar
Peter Zijlstra committed
664
static inline
665
666
667
668
669
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 */
670

671
#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
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;

Peter Zijlstra's avatar
Peter Zijlstra committed
688
	if (rt_rq->rt_nr_running) {
689

690
		WARN_ON(prio < prev_prio);
691

692
		/*
693
694
		 * This may have been our highest task, and therefore
		 * we may have some recomputation to do
695
		 */
696
		if (prio == prev_prio) {
697
698
699
			struct rt_prio_array *array = &rt_rq->active;

			rt_rq->highest_prio.curr =
700
				sched_find_first_bit(array->bitmap);
701
702
		}

703
	} else
704
		rt_rq->highest_prio.curr = MAX_RT_PRIO;
705

706
707
	dec_rt_prio_smp(rt_rq, prio, prev_prio);
}
708

709
710
711
712
713
714
#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 */
715

716
#ifdef CONFIG_RT_GROUP_SCHED
717
718
719
720
721
722
723
724
725
726
727
728
729
730

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)
{
Peter Zijlstra's avatar
Peter Zijlstra committed
731
732
733
734
	if (rt_se_boosted(rt_se))
		rt_rq->rt_nr_boosted--;

	WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted);
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
}

#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);
773
774
}

775
static void __enqueue_rt_entity(struct sched_rt_entity *rt_se)
776
{
Peter Zijlstra's avatar
Peter Zijlstra committed
777
778
779
	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);
780
	struct list_head *queue = array->queue + rt_se_prio(rt_se);
781

782
783
784
785
786
787
788
	/*
	 * 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))
Peter Zijlstra's avatar
Peter Zijlstra committed
789
		return;
790

791
	list_add_tail(&rt_se->run_list, queue);
Peter Zijlstra's avatar
Peter Zijlstra committed
792
	__set_bit(rt_se_prio(rt_se), array->bitmap);
793

Peter Zijlstra's avatar
Peter Zijlstra committed
794
795
796
	inc_rt_tasks(rt_se, rt_rq);
}

797
static void __dequeue_rt_entity(struct sched_rt_entity *rt_se)
Peter Zijlstra's avatar
Peter Zijlstra committed
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
{
	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.
 */
813
static void dequeue_rt_stack(struct sched_rt_entity *rt_se)
Peter Zijlstra's avatar
Peter Zijlstra committed
814
{
815
	struct sched_rt_entity *back = NULL;
Peter Zijlstra's avatar
Peter Zijlstra committed
816

817
818
819
820
821
822
823
	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))
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
			__dequeue_rt_entity(rt_se);
	}
}

static void enqueue_rt_entity(struct sched_rt_entity *rt_se)
{
	dequeue_rt_stack(rt_se);
	for_each_sched_rt_entity(rt_se)
		__enqueue_rt_entity(rt_se);
}

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)
			__enqueue_rt_entity(rt_se);
844
	}
845
846
847
848
849
}

/*
 * Adding/removing a task to/from a priority array:
 */
Peter Zijlstra's avatar
Peter Zijlstra committed
850
851
852
853
854
855
856
static void enqueue_task_rt(struct rq *rq, struct task_struct *p, int wakeup)
{
	struct sched_rt_entity *rt_se = &p->rt;

	if (wakeup)
		rt_se->timeout = 0;

857
	enqueue_rt_entity(rt_se);
858

859
860
861
	if (!task_current(rq, p) && p->rt.nr_cpus_allowed > 1)
		enqueue_pushable_task(rq, p);

862
	inc_cpu_load(rq, p->se.load.weight);
Peter Zijlstra's avatar
Peter Zijlstra committed
863
864
}

865
static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep)
866
{
Peter Zijlstra's avatar
Peter Zijlstra committed
867
	struct sched_rt_entity *rt_se = &p->rt;
868

869
	update_curr_rt(rq);
870
	dequeue_rt_entity(rt_se);
871

872
873
	dequeue_pushable_task(rq, p);

874
	dec_cpu_load(rq, p->se.load.weight);
875
876
877
878
879
880
}

/*
 * Put task to the end of the run list without the overhead of dequeue
 * followed by enqueue.
 */
881
882
static void
requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head)
Peter Zijlstra's avatar
Peter Zijlstra committed
883
{
884
	if (on_rt_rq(rt_se)) {
885
886
887
888
889
890
891
		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);
892
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
893
894
}

895
static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head)
896
{
Peter Zijlstra's avatar
Peter Zijlstra committed
897
898
	struct sched_rt_entity *rt_se = &p->rt;
	struct rt_rq *rt_rq;
899

Peter Zijlstra's avatar
Peter Zijlstra committed
900
901
	for_each_sched_rt_entity(rt_se) {
		rt_rq = rt_rq_of_se(rt_se);
902
		requeue_rt_entity(rt_rq, rt_se, head);
Peter Zijlstra's avatar
Peter Zijlstra committed
903
	}
904
905
}

Peter Zijlstra's avatar
Peter Zijlstra committed
906
static void yield_task_rt(struct rq *rq)
907
{
908
	requeue_task_rt(rq, rq->curr, 0);
909
910
}

911
#ifdef CONFIG_SMP
912
913
static int find_lowest_rq(struct task_struct *task);

914
915
static int select_task_rq_rt(struct task_struct *p, int sync)
{
916
917
918
	struct rq *rq = task_rq(p);

	/*
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
	 * If the current task is an RT task, then
	 * try to see if we can wake this RT task up on another
	 * runqueue. Otherwise simply start this RT task
	 * on its current runqueue.
	 *
	 * We want to avoid overloading runqueues. Even if
	 * the RT task is of higher priority than the current RT task.
	 * RT tasks behave differently than other tasks. If
	 * one gets preempted, we try to push it off to another queue.
	 * So trying to keep a preempting RT task on the same
	 * cache hot CPU will force the running RT task to
	 * a cold CPU. So we waste all the cache for the lower
	 * RT task in hopes of saving some of a RT task
	 * that is just being woken and probably will have
	 * cold cache anyway.
934
	 */
935
	if (unlikely(rt_task(rq->curr)) &&
Peter Zijlstra's avatar
Peter Zijlstra committed
936
	    (p->rt.nr_cpus_allowed > 1)) {
937
938
939
940
941
942
943
944
945
		int cpu = find_lowest_rq(p);

		return (cpu == -1) ? task_cpu(p) : cpu;
	}

	/*
	 * Otherwise, just let it ride on the affined RQ and the
	 * post-schedule router will push the preempted task away
	 */
946
947
	return task_cpu(p);
}
948
949
950
951
952
953

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

954
	if (p->rt.nr_cpus_allowed != 1
955
956
	    && cpupri_find(&rq->rd->cpupri, p, NULL))
		return;
957

958
959
	if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL))
		return;
960
961
962
963
964
965
966
967
968
969

	/*
	 * 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);
}

970
971
#endif /* CONFIG_SMP */

972
973
974
/*
 * Preempt the current task with a newly woken task if needed:
 */
975
static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int sync)
976
{
977
	if (p->prio < rq->curr->prio) {
978
		resched_task(rq->curr);
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
		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.
	 */
995
996
	if (p->prio == rq->curr->prio && !need_resched())
		check_preempt_equal_prio(rq, p);
997
#endif
998
999
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1000
1001
static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq,
						   struct rt_rq *rt_rq)
1002
{
Peter Zijlstra's avatar
Peter Zijlstra committed
1003
1004
	struct rt_prio_array *array = &rt_rq->active;
	struct sched_rt_entity *next = NULL;
1005
1006
1007
1008
	struct list_head *queue;
	int idx;

	idx = sched_find_first_bit(array->bitmap);
Peter Zijlstra's avatar
Peter Zijlstra committed
1009
	BUG_ON(idx >= MAX_RT_PRIO);
1010
1011

	queue = array->queue + idx;
Peter Zijlstra's avatar
Peter Zijlstra committed
1012
	next = list_entry(queue->next, struct sched_rt_entity, run_list);
1013

Peter Zijlstra's avatar
Peter Zijlstra committed
1014
1015
	return next;
}
1016

1017
static struct task_struct *_pick_next_task_rt(struct rq *rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
1018
1019
1020
1021
{
	struct sched_rt_entity *rt_se;
	struct task_struct *p;
	struct rt_rq *rt_rq;
1022

Peter Zijlstra's avatar
Peter Zijlstra committed
1023
1024
1025
1026
1027
	rt_rq = &rq->rt;

	if (unlikely(!rt_rq->rt_nr_running))
		return NULL;

Peter Zijlstra's avatar
Peter Zijlstra committed
1028
	if (rt_rq_throttled(rt_rq))
Peter Zijlstra's avatar
Peter Zijlstra committed
1029
1030
1031
1032
		return NULL;

	do {
		rt_se = pick_next_rt_entity(rq, rt_rq);
1033
		BUG_ON(!rt_se);
Peter Zijlstra's avatar
Peter Zijlstra committed
1034
1035
1036
1037
1038
		rt_rq = group_rt_rq(rt_se);
	} while (rt_rq);

	p = rt_task_of(rt_se);
	p->se.exec_start = rq->clock;
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050

	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);

Peter Zijlstra's avatar
Peter Zijlstra committed
1051
	return p;
1052
1053
}

1054
static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
1055
{
1056
	update_curr_rt(rq);
1057
	p->se.exec_start = 0;
1058
1059
1060
1061
1062
1063
1064

	/*
	 * The previous task needs to be made eligible for pushing
	 * if it is still active
	 */
	if (p->se.on_rq && p->rt.nr_cpus_allowed > 1)
		enqueue_pushable_task(rq, p);
1065
1066
}

1067
#ifdef CONFIG_SMP
Peter Zijlstra's avatar
Peter Zijlstra committed
1068

Steven Rostedt's avatar
Steven Rostedt committed
1069
1070
1071
1072
1073
/* Only try algorithms three times */
#define RT_MAX_TRIES 3

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

1074
1075
1076
static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu)
{
	if (!task_running(rq, p) &&
1077
	    (cpu < 0 || cpumask_test_cpu(cpu, &p->cpus_allowed)) &&
Peter Zijlstra's avatar
Peter Zijlstra committed
1078
	    (p->rt.nr_cpus_allowed > 1))
1079
1080
1081
1082
		return 1;
	return 0;
}

Steven Rostedt's avatar
Steven Rostedt committed
1083
/* Return the second highest RT task, NULL otherwise */
1084
static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu)
Steven Rostedt's avatar
Steven Rostedt committed
1085
{
Peter Zijlstra's avatar
Peter Zijlstra committed
1086
1087
1088
1089
	struct task_struct *next = NULL;
	struct sched_rt_entity *rt_se;
	struct rt_prio_array *array;
	struct rt_rq *rt_rq;
Steven Rostedt's avatar
Steven Rostedt committed
1090
1091
	int idx;

Peter Zijlstra's avatar
Peter Zijlstra committed
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
	for_each_leaf_rt_rq(rt_rq, rq) {
		array = &rt_rq->active;
		idx = sched_find_first_bit(array->bitmap);
 next_idx:
		if (idx >= MAX_RT_PRIO)
			continue;
		if (next && next->prio < idx)
			continue;
		list_for_each_entry(rt_se, array->queue + idx, run_list) {
			struct task_struct *p = rt_task_of(rt_se);
			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;
		}
1111
1112
	}

Steven Rostedt's avatar
Steven Rostedt committed
1113
1114
1115
	return next;
}

1116
static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask);
Steven Rostedt's avatar
Steven Rostedt committed
1117

1118
1119
static inline int pick_optimal_cpu(int this_cpu,
				   const struct cpumask *mask)
Gregory Haskins's avatar
Gregory Haskins committed
1120
1121
1122
1123
{
	int first;

	/* "this_cpu" is cheaper to preempt than a remote processor */
1124
	if ((this_cpu != -1) && cpumask_test_cpu(this_cpu, mask))
Gregory Haskins's avatar
Gregory Haskins committed
1125
1126
		return this_cpu;

1127
1128
	first = cpumask_first(mask);
	if (first < nr_cpu_ids)
Gregory Haskins's avatar
Gregory Haskins committed
1129
1130
1131
1132
1133
1134
1135
1136
		return first;

	return -1;
}

static int find_lowest_rq(struct task_struct *task)
{
	struct sched_domain *sd;
1137
	struct cpumask *lowest_mask = __get_cpu_var(local_cpu_mask);
Gregory Haskins's avatar
Gregory Haskins committed
1138
1139
	int this_cpu = smp_processor_id();
	int cpu      = task_cpu(task);
1140
	cpumask_var_t domain_mask;
Gregory Haskins's avatar
Gregory Haskins committed
1141

1142
1143
	if (task->rt.nr_cpus_allowed == 1)
		return -1; /* No other targets possible */
Gregory Haskins's avatar
Gregory Haskins committed
1144

1145
1146
	if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask))
		return -1; /* No targets found */
Gregory Haskins's avatar
Gregory Haskins committed
1147

1148
1149
1150
1151
1152
	/*
	 * Only consider CPUs that are usable for migration.
	 * I guess we might want to change cpupri_find() to ignore those
	 * in the first place.
	 */
1153
	cpumask_and(lowest_mask, lowest_mask, cpu_active_mask);
1154

Gregory Haskins's avatar
Gregory Haskins committed
1155
1156
1157
1158
1159
1160
1161
1162
	/*
	 * 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.
	 */
1163
	if (cpumask_test_cpu(cpu, lowest_mask))
Gregory Haskins's avatar
Gregory Haskins committed
1164
1165
1166
1167
1168
1169
1170
1171
1172
		return cpu;

	/*
	 * Otherwise, we consult the sched_domains span maps to figure
	 * out which cpu is logically closest to our hot cache data.
	 */
	if (this_cpu == cpu)
		this_cpu = -1; /* Skip this_cpu opt if the same */

1173
1174
1175
1176
	if (alloc_cpumask_var(&domain_mask, GFP_ATOMIC)) {
		for_each_domain(cpu, sd) {
			if (sd->flags & SD_WAKE_AFFINE) {
				int best_cpu;
Gregory Haskins's avatar
Gregory Haskins committed
1177

1178
1179
1180
				cpumask_and(domain_mask,
					    sched_domain_span(sd),
					    lowest_mask);
Gregory Haskins's avatar
Gregory Haskins committed
1181

1182
1183
				best_cpu = pick_optimal_cpu(this_cpu,
							    domain_mask);
Gregory Haskins's avatar
Gregory Haskins committed
1184

1185
1186
1187
1188
1189
				if (best_cpu != -1) {
					free_cpumask_var(domain_mask);
					return best_cpu;
				}
			}
Gregory Haskins's avatar
Gregory Haskins committed
1190
		}
1191
		free_cpumask_var(domain_mask);
Gregory Haskins's avatar
Gregory Haskins committed
1192
1193
1194
1195
1196
1197
1198
1199
	}

	/*
	 * And finally, if there were no matches within the domains
	 * just give the caller *something* to work with from the compatible
	 * locations.
	 */
	return pick_optimal_cpu(this_cpu, lowest_mask);
1200
1201
1202
}

/* Will lock the rq it finds */
1203
static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq)
1204
1205
1206
{
	struct rq *lowest_rq = NULL;
	int tries;
1207
	int cpu;
Steven Rostedt's avatar
Steven Rostedt committed
1208

1209
1210
1211
	for (tries = 0; tries < RT_MAX_TRIES; tries++) {
		cpu = find_lowest_rq(task);

Gregory Haskins's avatar