sched: revert load_balance_monitor() changes
[linux-2.6.git] / kernel / sched_rt.c
1 /*
2  * Real-Time Scheduling Class (mapped to the SCHED_FIFO and SCHED_RR
3  * policies)
4  */
5
6 #ifdef CONFIG_SMP
7
8 static inline int rt_overloaded(struct rq *rq)
9 {
10         return atomic_read(&rq->rd->rto_count);
11 }
12
13 static inline void rt_set_overload(struct rq *rq)
14 {
15         cpu_set(rq->cpu, rq->rd->rto_mask);
16         /*
17          * Make sure the mask is visible before we set
18          * the overload count. That is checked to determine
19          * if we should look at the mask. It would be a shame
20          * if we looked at the mask, but the mask was not
21          * updated yet.
22          */
23         wmb();
24         atomic_inc(&rq->rd->rto_count);
25 }
26
27 static inline void rt_clear_overload(struct rq *rq)
28 {
29         /* the order here really doesn't matter */
30         atomic_dec(&rq->rd->rto_count);
31         cpu_clear(rq->cpu, rq->rd->rto_mask);
32 }
33
34 static void update_rt_migration(struct rq *rq)
35 {
36         if (rq->rt.rt_nr_migratory && (rq->rt.rt_nr_running > 1)) {
37                 if (!rq->rt.overloaded) {
38                         rt_set_overload(rq);
39                         rq->rt.overloaded = 1;
40                 }
41         } else if (rq->rt.overloaded) {
42                 rt_clear_overload(rq);
43                 rq->rt.overloaded = 0;
44         }
45 }
46 #endif /* CONFIG_SMP */
47
48 static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
49 {
50         return container_of(rt_se, struct task_struct, rt);
51 }
52
53 static inline int on_rt_rq(struct sched_rt_entity *rt_se)
54 {
55         return !list_empty(&rt_se->run_list);
56 }
57
58 #ifdef CONFIG_RT_GROUP_SCHED
59
60 static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
61 {
62         if (!rt_rq->tg)
63                 return RUNTIME_INF;
64
65         return rt_rq->tg->rt_runtime;
66 }
67
68 #define for_each_leaf_rt_rq(rt_rq, rq) \
69         list_for_each_entry(rt_rq, &rq->leaf_rt_rq_list, leaf_rt_rq_list)
70
71 static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
72 {
73         return rt_rq->rq;
74 }
75
76 static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se)
77 {
78         return rt_se->rt_rq;
79 }
80
81 #define for_each_sched_rt_entity(rt_se) \
82         for (; rt_se; rt_se = rt_se->parent)
83
84 static inline struct rt_rq *group_rt_rq(struct sched_rt_entity *rt_se)
85 {
86         return rt_se->my_q;
87 }
88
89 static void enqueue_rt_entity(struct sched_rt_entity *rt_se);
90 static void dequeue_rt_entity(struct sched_rt_entity *rt_se);
91
92 static void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
93 {
94         struct sched_rt_entity *rt_se = rt_rq->rt_se;
95
96         if (rt_se && !on_rt_rq(rt_se) && rt_rq->rt_nr_running) {
97                 struct task_struct *curr = rq_of_rt_rq(rt_rq)->curr;
98
99                 enqueue_rt_entity(rt_se);
100                 if (rt_rq->highest_prio < curr->prio)
101                         resched_task(curr);
102         }
103 }
104
105 static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
106 {
107         struct sched_rt_entity *rt_se = rt_rq->rt_se;
108
109         if (rt_se && on_rt_rq(rt_se))
110                 dequeue_rt_entity(rt_se);
111 }
112
113 static inline int rt_rq_throttled(struct rt_rq *rt_rq)
114 {
115         return rt_rq->rt_throttled && !rt_rq->rt_nr_boosted;
116 }
117
118 static int rt_se_boosted(struct sched_rt_entity *rt_se)
119 {
120         struct rt_rq *rt_rq = group_rt_rq(rt_se);
121         struct task_struct *p;
122
123         if (rt_rq)
124                 return !!rt_rq->rt_nr_boosted;
125
126         p = rt_task_of(rt_se);
127         return p->prio != p->normal_prio;
128 }
129
130 #else
131
132 static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
133 {
134         if (sysctl_sched_rt_runtime == -1)
135                 return RUNTIME_INF;
136
137         return (u64)sysctl_sched_rt_runtime * NSEC_PER_USEC;
138 }
139
140 #define for_each_leaf_rt_rq(rt_rq, rq) \
141         for (rt_rq = &rq->rt; rt_rq; rt_rq = NULL)
142
143 static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
144 {
145         return container_of(rt_rq, struct rq, rt);
146 }
147
148 static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se)
149 {
150         struct task_struct *p = rt_task_of(rt_se);
151         struct rq *rq = task_rq(p);
152
153         return &rq->rt;
154 }
155
156 #define for_each_sched_rt_entity(rt_se) \
157         for (; rt_se; rt_se = NULL)
158
159 static inline struct rt_rq *group_rt_rq(struct sched_rt_entity *rt_se)
160 {
161         return NULL;
162 }
163
164 static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
165 {
166 }
167
168 static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
169 {
170 }
171
172 static inline int rt_rq_throttled(struct rt_rq *rt_rq)
173 {
174         return rt_rq->rt_throttled;
175 }
176 #endif
177
178 static inline int rt_se_prio(struct sched_rt_entity *rt_se)
179 {
180 #ifdef CONFIG_RT_GROUP_SCHED
181         struct rt_rq *rt_rq = group_rt_rq(rt_se);
182
183         if (rt_rq)
184                 return rt_rq->highest_prio;
185 #endif
186
187         return rt_task_of(rt_se)->prio;
188 }
189
190 static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq)
191 {
192         u64 runtime = sched_rt_runtime(rt_rq);
193
194         if (runtime == RUNTIME_INF)
195                 return 0;
196
197         if (rt_rq->rt_throttled)
198                 return rt_rq_throttled(rt_rq);
199
200         if (rt_rq->rt_time > runtime) {
201                 struct rq *rq = rq_of_rt_rq(rt_rq);
202
203                 rq->rt_throttled = 1;
204                 rt_rq->rt_throttled = 1;
205
206                 if (rt_rq_throttled(rt_rq)) {
207                         sched_rt_rq_dequeue(rt_rq);
208                         return 1;
209                 }
210         }
211
212         return 0;
213 }
214
215 static void update_sched_rt_period(struct rq *rq)
216 {
217         struct rt_rq *rt_rq;
218         u64 period;
219
220         while (rq->clock > rq->rt_period_expire) {
221                 period = (u64)sysctl_sched_rt_period * NSEC_PER_USEC;
222                 rq->rt_period_expire += period;
223
224                 for_each_leaf_rt_rq(rt_rq, rq) {
225                         u64 runtime = sched_rt_runtime(rt_rq);
226
227                         rt_rq->rt_time -= min(rt_rq->rt_time, runtime);
228                         if (rt_rq->rt_throttled && rt_rq->rt_time < runtime) {
229                                 rt_rq->rt_throttled = 0;
230                                 sched_rt_rq_enqueue(rt_rq);
231                         }
232                 }
233
234                 rq->rt_throttled = 0;
235         }
236 }
237
238 /*
239  * Update the current task's runtime statistics. Skip current tasks that
240  * are not in our scheduling class.
241  */
242 static void update_curr_rt(struct rq *rq)
243 {
244         struct task_struct *curr = rq->curr;
245         struct sched_rt_entity *rt_se = &curr->rt;
246         struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
247         u64 delta_exec;
248
249         if (!task_has_rt_policy(curr))
250                 return;
251
252         delta_exec = rq->clock - curr->se.exec_start;
253         if (unlikely((s64)delta_exec < 0))
254                 delta_exec = 0;
255
256         schedstat_set(curr->se.exec_max, max(curr->se.exec_max, delta_exec));
257
258         curr->se.sum_exec_runtime += delta_exec;
259         curr->se.exec_start = rq->clock;
260         cpuacct_charge(curr, delta_exec);
261
262         rt_rq->rt_time += delta_exec;
263         if (sched_rt_runtime_exceeded(rt_rq))
264                 resched_task(curr);
265 }
266
267 static inline
268 void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
269 {
270         WARN_ON(!rt_prio(rt_se_prio(rt_se)));
271         rt_rq->rt_nr_running++;
272 #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
273         if (rt_se_prio(rt_se) < rt_rq->highest_prio)
274                 rt_rq->highest_prio = rt_se_prio(rt_se);
275 #endif
276 #ifdef CONFIG_SMP
277         if (rt_se->nr_cpus_allowed > 1) {
278                 struct rq *rq = rq_of_rt_rq(rt_rq);
279                 rq->rt.rt_nr_migratory++;
280         }
281
282         update_rt_migration(rq_of_rt_rq(rt_rq));
283 #endif
284 #ifdef CONFIG_RT_GROUP_SCHED
285         if (rt_se_boosted(rt_se))
286                 rt_rq->rt_nr_boosted++;
287 #endif
288 }
289
290 static inline
291 void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
292 {
293         WARN_ON(!rt_prio(rt_se_prio(rt_se)));
294         WARN_ON(!rt_rq->rt_nr_running);
295         rt_rq->rt_nr_running--;
296 #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
297         if (rt_rq->rt_nr_running) {
298                 struct rt_prio_array *array;
299
300                 WARN_ON(rt_se_prio(rt_se) < rt_rq->highest_prio);
301                 if (rt_se_prio(rt_se) == rt_rq->highest_prio) {
302                         /* recalculate */
303                         array = &rt_rq->active;
304                         rt_rq->highest_prio =
305                                 sched_find_first_bit(array->bitmap);
306                 } /* otherwise leave rq->highest prio alone */
307         } else
308                 rt_rq->highest_prio = MAX_RT_PRIO;
309 #endif
310 #ifdef CONFIG_SMP
311         if (rt_se->nr_cpus_allowed > 1) {
312                 struct rq *rq = rq_of_rt_rq(rt_rq);
313                 rq->rt.rt_nr_migratory--;
314         }
315
316         update_rt_migration(rq_of_rt_rq(rt_rq));
317 #endif /* CONFIG_SMP */
318 #ifdef CONFIG_RT_GROUP_SCHED
319         if (rt_se_boosted(rt_se))
320                 rt_rq->rt_nr_boosted--;
321
322         WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted);
323 #endif
324 }
325
326 static void enqueue_rt_entity(struct sched_rt_entity *rt_se)
327 {
328         struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
329         struct rt_prio_array *array = &rt_rq->active;
330         struct rt_rq *group_rq = group_rt_rq(rt_se);
331
332         if (group_rq && rt_rq_throttled(group_rq))
333                 return;
334
335         list_add_tail(&rt_se->run_list, array->queue + rt_se_prio(rt_se));
336         __set_bit(rt_se_prio(rt_se), array->bitmap);
337
338         inc_rt_tasks(rt_se, rt_rq);
339 }
340
341 static void dequeue_rt_entity(struct sched_rt_entity *rt_se)
342 {
343         struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
344         struct rt_prio_array *array = &rt_rq->active;
345
346         list_del_init(&rt_se->run_list);
347         if (list_empty(array->queue + rt_se_prio(rt_se)))
348                 __clear_bit(rt_se_prio(rt_se), array->bitmap);
349
350         dec_rt_tasks(rt_se, rt_rq);
351 }
352
353 /*
354  * Because the prio of an upper entry depends on the lower
355  * entries, we must remove entries top - down.
356  *
357  * XXX: O(1/2 h^2) because we can only walk up, not down the chain.
358  *      doesn't matter much for now, as h=2 for GROUP_SCHED.
359  */
360 static void dequeue_rt_stack(struct task_struct *p)
361 {
362         struct sched_rt_entity *rt_se, *top_se;
363
364         /*
365          * dequeue all, top - down.
366          */
367         do {
368                 rt_se = &p->rt;
369                 top_se = NULL;
370                 for_each_sched_rt_entity(rt_se) {
371                         if (on_rt_rq(rt_se))
372                                 top_se = rt_se;
373                 }
374                 if (top_se)
375                         dequeue_rt_entity(top_se);
376         } while (top_se);
377 }
378
379 /*
380  * Adding/removing a task to/from a priority array:
381  */
382 static void enqueue_task_rt(struct rq *rq, struct task_struct *p, int wakeup)
383 {
384         struct sched_rt_entity *rt_se = &p->rt;
385
386         if (wakeup)
387                 rt_se->timeout = 0;
388
389         dequeue_rt_stack(p);
390
391         /*
392          * enqueue everybody, bottom - up.
393          */
394         for_each_sched_rt_entity(rt_se)
395                 enqueue_rt_entity(rt_se);
396 }
397
398 static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep)
399 {
400         struct sched_rt_entity *rt_se = &p->rt;
401         struct rt_rq *rt_rq;
402
403         update_curr_rt(rq);
404
405         dequeue_rt_stack(p);
406
407         /*
408          * re-enqueue all non-empty rt_rq entities.
409          */
410         for_each_sched_rt_entity(rt_se) {
411                 rt_rq = group_rt_rq(rt_se);
412                 if (rt_rq && rt_rq->rt_nr_running)
413                         enqueue_rt_entity(rt_se);
414         }
415 }
416
417 /*
418  * Put task to the end of the run list without the overhead of dequeue
419  * followed by enqueue.
420  */
421 static
422 void requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
423 {
424         struct rt_prio_array *array = &rt_rq->active;
425
426         list_move_tail(&rt_se->run_list, array->queue + rt_se_prio(rt_se));
427 }
428
429 static void requeue_task_rt(struct rq *rq, struct task_struct *p)
430 {
431         struct sched_rt_entity *rt_se = &p->rt;
432         struct rt_rq *rt_rq;
433
434         for_each_sched_rt_entity(rt_se) {
435                 rt_rq = rt_rq_of_se(rt_se);
436                 requeue_rt_entity(rt_rq, rt_se);
437         }
438 }
439
440 static void yield_task_rt(struct rq *rq)
441 {
442         requeue_task_rt(rq, rq->curr);
443 }
444
445 #ifdef CONFIG_SMP
446 static int find_lowest_rq(struct task_struct *task);
447
448 static int select_task_rq_rt(struct task_struct *p, int sync)
449 {
450         struct rq *rq = task_rq(p);
451
452         /*
453          * If the current task is an RT task, then
454          * try to see if we can wake this RT task up on another
455          * runqueue. Otherwise simply start this RT task
456          * on its current runqueue.
457          *
458          * We want to avoid overloading runqueues. Even if
459          * the RT task is of higher priority than the current RT task.
460          * RT tasks behave differently than other tasks. If
461          * one gets preempted, we try to push it off to another queue.
462          * So trying to keep a preempting RT task on the same
463          * cache hot CPU will force the running RT task to
464          * a cold CPU. So we waste all the cache for the lower
465          * RT task in hopes of saving some of a RT task
466          * that is just being woken and probably will have
467          * cold cache anyway.
468          */
469         if (unlikely(rt_task(rq->curr)) &&
470             (p->rt.nr_cpus_allowed > 1)) {
471                 int cpu = find_lowest_rq(p);
472
473                 return (cpu == -1) ? task_cpu(p) : cpu;
474         }
475
476         /*
477          * Otherwise, just let it ride on the affined RQ and the
478          * post-schedule router will push the preempted task away
479          */
480         return task_cpu(p);
481 }
482 #endif /* CONFIG_SMP */
483
484 /*
485  * Preempt the current task with a newly woken task if needed:
486  */
487 static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p)
488 {
489         if (p->prio < rq->curr->prio)
490                 resched_task(rq->curr);
491 }
492
493 static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq,
494                                                    struct rt_rq *rt_rq)
495 {
496         struct rt_prio_array *array = &rt_rq->active;
497         struct sched_rt_entity *next = NULL;
498         struct list_head *queue;
499         int idx;
500
501         idx = sched_find_first_bit(array->bitmap);
502         BUG_ON(idx >= MAX_RT_PRIO);
503
504         queue = array->queue + idx;
505         next = list_entry(queue->next, struct sched_rt_entity, run_list);
506
507         return next;
508 }
509
510 static struct task_struct *pick_next_task_rt(struct rq *rq)
511 {
512         struct sched_rt_entity *rt_se;
513         struct task_struct *p;
514         struct rt_rq *rt_rq;
515
516         rt_rq = &rq->rt;
517
518         if (unlikely(!rt_rq->rt_nr_running))
519                 return NULL;
520
521         if (rt_rq_throttled(rt_rq))
522                 return NULL;
523
524         do {
525                 rt_se = pick_next_rt_entity(rq, rt_rq);
526                 BUG_ON(!rt_se);
527                 rt_rq = group_rt_rq(rt_se);
528         } while (rt_rq);
529
530         p = rt_task_of(rt_se);
531         p->se.exec_start = rq->clock;
532         return p;
533 }
534
535 static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
536 {
537         update_curr_rt(rq);
538         p->se.exec_start = 0;
539 }
540
541 #ifdef CONFIG_SMP
542
543 /* Only try algorithms three times */
544 #define RT_MAX_TRIES 3
545
546 static int double_lock_balance(struct rq *this_rq, struct rq *busiest);
547 static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep);
548
549 static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu)
550 {
551         if (!task_running(rq, p) &&
552             (cpu < 0 || cpu_isset(cpu, p->cpus_allowed)) &&
553             (p->rt.nr_cpus_allowed > 1))
554                 return 1;
555         return 0;
556 }
557
558 /* Return the second highest RT task, NULL otherwise */
559 static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu)
560 {
561         struct task_struct *next = NULL;
562         struct sched_rt_entity *rt_se;
563         struct rt_prio_array *array;
564         struct rt_rq *rt_rq;
565         int idx;
566
567         for_each_leaf_rt_rq(rt_rq, rq) {
568                 array = &rt_rq->active;
569                 idx = sched_find_first_bit(array->bitmap);
570  next_idx:
571                 if (idx >= MAX_RT_PRIO)
572                         continue;
573                 if (next && next->prio < idx)
574                         continue;
575                 list_for_each_entry(rt_se, array->queue + idx, run_list) {
576                         struct task_struct *p = rt_task_of(rt_se);
577                         if (pick_rt_task(rq, p, cpu)) {
578                                 next = p;
579                                 break;
580                         }
581                 }
582                 if (!next) {
583                         idx = find_next_bit(array->bitmap, MAX_RT_PRIO, idx+1);
584                         goto next_idx;
585                 }
586         }
587
588         return next;
589 }
590
591 static DEFINE_PER_CPU(cpumask_t, local_cpu_mask);
592
593 static int find_lowest_cpus(struct task_struct *task, cpumask_t *lowest_mask)
594 {
595         int       lowest_prio = -1;
596         int       lowest_cpu  = -1;
597         int       count       = 0;
598         int       cpu;
599
600         cpus_and(*lowest_mask, task_rq(task)->rd->online, task->cpus_allowed);
601
602         /*
603          * Scan each rq for the lowest prio.
604          */
605         for_each_cpu_mask(cpu, *lowest_mask) {
606                 struct rq *rq = cpu_rq(cpu);
607
608                 /* We look for lowest RT prio or non-rt CPU */
609                 if (rq->rt.highest_prio >= MAX_RT_PRIO) {
610                         /*
611                          * if we already found a low RT queue
612                          * and now we found this non-rt queue
613                          * clear the mask and set our bit.
614                          * Otherwise just return the queue as is
615                          * and the count==1 will cause the algorithm
616                          * to use the first bit found.
617                          */
618                         if (lowest_cpu != -1) {
619                                 cpus_clear(*lowest_mask);
620                                 cpu_set(rq->cpu, *lowest_mask);
621                         }
622                         return 1;
623                 }
624
625                 /* no locking for now */
626                 if ((rq->rt.highest_prio > task->prio)
627                     && (rq->rt.highest_prio >= lowest_prio)) {
628                         if (rq->rt.highest_prio > lowest_prio) {
629                                 /* new low - clear old data */
630                                 lowest_prio = rq->rt.highest_prio;
631                                 lowest_cpu = cpu;
632                                 count = 0;
633                         }
634                         count++;
635                 } else
636                         cpu_clear(cpu, *lowest_mask);
637         }
638
639         /*
640          * Clear out all the set bits that represent
641          * runqueues that were of higher prio than
642          * the lowest_prio.
643          */
644         if (lowest_cpu > 0) {
645                 /*
646                  * Perhaps we could add another cpumask op to
647                  * zero out bits. Like cpu_zero_bits(cpumask, nrbits);
648                  * Then that could be optimized to use memset and such.
649                  */
650                 for_each_cpu_mask(cpu, *lowest_mask) {
651                         if (cpu >= lowest_cpu)
652                                 break;
653                         cpu_clear(cpu, *lowest_mask);
654                 }
655         }
656
657         return count;
658 }
659
660 static inline int pick_optimal_cpu(int this_cpu, cpumask_t *mask)
661 {
662         int first;
663
664         /* "this_cpu" is cheaper to preempt than a remote processor */
665         if ((this_cpu != -1) && cpu_isset(this_cpu, *mask))
666                 return this_cpu;
667
668         first = first_cpu(*mask);
669         if (first != NR_CPUS)
670                 return first;
671
672         return -1;
673 }
674
675 static int find_lowest_rq(struct task_struct *task)
676 {
677         struct sched_domain *sd;
678         cpumask_t *lowest_mask = &__get_cpu_var(local_cpu_mask);
679         int this_cpu = smp_processor_id();
680         int cpu      = task_cpu(task);
681         int count    = find_lowest_cpus(task, lowest_mask);
682
683         if (!count)
684                 return -1; /* No targets found */
685
686         /*
687          * There is no sense in performing an optimal search if only one
688          * target is found.
689          */
690         if (count == 1)
691                 return first_cpu(*lowest_mask);
692
693         /*
694          * At this point we have built a mask of cpus representing the
695          * lowest priority tasks in the system.  Now we want to elect
696          * the best one based on our affinity and topology.
697          *
698          * We prioritize the last cpu that the task executed on since
699          * it is most likely cache-hot in that location.
700          */
701         if (cpu_isset(cpu, *lowest_mask))
702                 return cpu;
703
704         /*
705          * Otherwise, we consult the sched_domains span maps to figure
706          * out which cpu is logically closest to our hot cache data.
707          */
708         if (this_cpu == cpu)
709                 this_cpu = -1; /* Skip this_cpu opt if the same */
710
711         for_each_domain(cpu, sd) {
712                 if (sd->flags & SD_WAKE_AFFINE) {
713                         cpumask_t domain_mask;
714                         int       best_cpu;
715
716                         cpus_and(domain_mask, sd->span, *lowest_mask);
717
718                         best_cpu = pick_optimal_cpu(this_cpu,
719                                                     &domain_mask);
720                         if (best_cpu != -1)
721                                 return best_cpu;
722                 }
723         }
724
725         /*
726          * And finally, if there were no matches within the domains
727          * just give the caller *something* to work with from the compatible
728          * locations.
729          */
730         return pick_optimal_cpu(this_cpu, lowest_mask);
731 }
732
733 /* Will lock the rq it finds */
734 static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq)
735 {
736         struct rq *lowest_rq = NULL;
737         int tries;
738         int cpu;
739
740         for (tries = 0; tries < RT_MAX_TRIES; tries++) {
741                 cpu = find_lowest_rq(task);
742
743                 if ((cpu == -1) || (cpu == rq->cpu))
744                         break;
745
746                 lowest_rq = cpu_rq(cpu);
747
748                 /* if the prio of this runqueue changed, try again */
749                 if (double_lock_balance(rq, lowest_rq)) {
750                         /*
751                          * We had to unlock the run queue. In
752                          * the mean time, task could have
753                          * migrated already or had its affinity changed.
754                          * Also make sure that it wasn't scheduled on its rq.
755                          */
756                         if (unlikely(task_rq(task) != rq ||
757                                      !cpu_isset(lowest_rq->cpu,
758                                                 task->cpus_allowed) ||
759                                      task_running(rq, task) ||
760                                      !task->se.on_rq)) {
761
762                                 spin_unlock(&lowest_rq->lock);
763                                 lowest_rq = NULL;
764                                 break;
765                         }
766                 }
767
768                 /* If this rq is still suitable use it. */
769                 if (lowest_rq->rt.highest_prio > task->prio)
770                         break;
771
772                 /* try again */
773                 spin_unlock(&lowest_rq->lock);
774                 lowest_rq = NULL;
775         }
776
777         return lowest_rq;
778 }
779
780 /*
781  * If the current CPU has more than one RT task, see if the non
782  * running task can migrate over to a CPU that is running a task
783  * of lesser priority.
784  */
785 static int push_rt_task(struct rq *rq)
786 {
787         struct task_struct *next_task;
788         struct rq *lowest_rq;
789         int ret = 0;
790         int paranoid = RT_MAX_TRIES;
791
792         if (!rq->rt.overloaded)
793                 return 0;
794
795         next_task = pick_next_highest_task_rt(rq, -1);
796         if (!next_task)
797                 return 0;
798
799  retry:
800         if (unlikely(next_task == rq->curr)) {
801                 WARN_ON(1);
802                 return 0;
803         }
804
805         /*
806          * It's possible that the next_task slipped in of
807          * higher priority than current. If that's the case
808          * just reschedule current.
809          */
810         if (unlikely(next_task->prio < rq->curr->prio)) {
811                 resched_task(rq->curr);
812                 return 0;
813         }
814
815         /* We might release rq lock */
816         get_task_struct(next_task);
817
818         /* find_lock_lowest_rq locks the rq if found */
819         lowest_rq = find_lock_lowest_rq(next_task, rq);
820         if (!lowest_rq) {
821                 struct task_struct *task;
822                 /*
823                  * find lock_lowest_rq releases rq->lock
824                  * so it is possible that next_task has changed.
825                  * If it has, then try again.
826                  */
827                 task = pick_next_highest_task_rt(rq, -1);
828                 if (unlikely(task != next_task) && task && paranoid--) {
829                         put_task_struct(next_task);
830                         next_task = task;
831                         goto retry;
832                 }
833                 goto out;
834         }
835
836         deactivate_task(rq, next_task, 0);
837         set_task_cpu(next_task, lowest_rq->cpu);
838         activate_task(lowest_rq, next_task, 0);
839
840         resched_task(lowest_rq->curr);
841
842         spin_unlock(&lowest_rq->lock);
843
844         ret = 1;
845 out:
846         put_task_struct(next_task);
847
848         return ret;
849 }
850
851 /*
852  * TODO: Currently we just use the second highest prio task on
853  *       the queue, and stop when it can't migrate (or there's
854  *       no more RT tasks).  There may be a case where a lower
855  *       priority RT task has a different affinity than the
856  *       higher RT task. In this case the lower RT task could
857  *       possibly be able to migrate where as the higher priority
858  *       RT task could not.  We currently ignore this issue.
859  *       Enhancements are welcome!
860  */
861 static void push_rt_tasks(struct rq *rq)
862 {
863         /* push_rt_task will return true if it moved an RT */
864         while (push_rt_task(rq))
865                 ;
866 }
867
868 static int pull_rt_task(struct rq *this_rq)
869 {
870         int this_cpu = this_rq->cpu, ret = 0, cpu;
871         struct task_struct *p, *next;
872         struct rq *src_rq;
873
874         if (likely(!rt_overloaded(this_rq)))
875                 return 0;
876
877         next = pick_next_task_rt(this_rq);
878
879         for_each_cpu_mask(cpu, this_rq->rd->rto_mask) {
880                 if (this_cpu == cpu)
881                         continue;
882
883                 src_rq = cpu_rq(cpu);
884                 /*
885                  * We can potentially drop this_rq's lock in
886                  * double_lock_balance, and another CPU could
887                  * steal our next task - hence we must cause
888                  * the caller to recalculate the next task
889                  * in that case:
890                  */
891                 if (double_lock_balance(this_rq, src_rq)) {
892                         struct task_struct *old_next = next;
893
894                         next = pick_next_task_rt(this_rq);
895                         if (next != old_next)
896                                 ret = 1;
897                 }
898
899                 /*
900                  * Are there still pullable RT tasks?
901                  */
902                 if (src_rq->rt.rt_nr_running <= 1)
903                         goto skip;
904
905                 p = pick_next_highest_task_rt(src_rq, this_cpu);
906
907                 /*
908                  * Do we have an RT task that preempts
909                  * the to-be-scheduled task?
910                  */
911                 if (p && (!next || (p->prio < next->prio))) {
912                         WARN_ON(p == src_rq->curr);
913                         WARN_ON(!p->se.on_rq);
914
915                         /*
916                          * There's a chance that p is higher in priority
917                          * than what's currently running on its cpu.
918                          * This is just that p is wakeing up and hasn't
919                          * had a chance to schedule. We only pull
920                          * p if it is lower in priority than the
921                          * current task on the run queue or
922                          * this_rq next task is lower in prio than
923                          * the current task on that rq.
924                          */
925                         if (p->prio < src_rq->curr->prio ||
926                             (next && next->prio < src_rq->curr->prio))
927                                 goto skip;
928
929                         ret = 1;
930
931                         deactivate_task(src_rq, p, 0);
932                         set_task_cpu(p, this_cpu);
933                         activate_task(this_rq, p, 0);
934                         /*
935                          * We continue with the search, just in
936                          * case there's an even higher prio task
937                          * in another runqueue. (low likelyhood
938                          * but possible)
939                          *
940                          * Update next so that we won't pick a task
941                          * on another cpu with a priority lower (or equal)
942                          * than the one we just picked.
943                          */
944                         next = p;
945
946                 }
947  skip:
948                 spin_unlock(&src_rq->lock);
949         }
950
951         return ret;
952 }
953
954 static void pre_schedule_rt(struct rq *rq, struct task_struct *prev)
955 {
956         /* Try to pull RT tasks here if we lower this rq's prio */
957         if (unlikely(rt_task(prev)) && rq->rt.highest_prio > prev->prio)
958                 pull_rt_task(rq);
959 }
960
961 static void post_schedule_rt(struct rq *rq)
962 {
963         /*
964          * If we have more than one rt_task queued, then
965          * see if we can push the other rt_tasks off to other CPUS.
966          * Note we may release the rq lock, and since
967          * the lock was owned by prev, we need to release it
968          * first via finish_lock_switch and then reaquire it here.
969          */
970         if (unlikely(rq->rt.overloaded)) {
971                 spin_lock_irq(&rq->lock);
972                 push_rt_tasks(rq);
973                 spin_unlock_irq(&rq->lock);
974         }
975 }
976
977
978 static void task_wake_up_rt(struct rq *rq, struct task_struct *p)
979 {
980         if (!task_running(rq, p) &&
981             (p->prio >= rq->rt.highest_prio) &&
982             rq->rt.overloaded)
983                 push_rt_tasks(rq);
984 }
985
986 static unsigned long
987 load_balance_rt(struct rq *this_rq, int this_cpu, struct rq *busiest,
988                 unsigned long max_load_move,
989                 struct sched_domain *sd, enum cpu_idle_type idle,
990                 int *all_pinned, int *this_best_prio)
991 {
992         /* don't touch RT tasks */
993         return 0;
994 }
995
996 static int
997 move_one_task_rt(struct rq *this_rq, int this_cpu, struct rq *busiest,
998                  struct sched_domain *sd, enum cpu_idle_type idle)
999 {
1000         /* don't touch RT tasks */
1001         return 0;
1002 }
1003
1004 static void set_cpus_allowed_rt(struct task_struct *p, cpumask_t *new_mask)
1005 {
1006         int weight = cpus_weight(*new_mask);
1007
1008         BUG_ON(!rt_task(p));
1009
1010         /*
1011          * Update the migration status of the RQ if we have an RT task
1012          * which is running AND changing its weight value.
1013          */
1014         if (p->se.on_rq && (weight != p->rt.nr_cpus_allowed)) {
1015                 struct rq *rq = task_rq(p);
1016
1017                 if ((p->rt.nr_cpus_allowed <= 1) && (weight > 1)) {
1018                         rq->rt.rt_nr_migratory++;
1019                 } else if ((p->rt.nr_cpus_allowed > 1) && (weight <= 1)) {
1020                         BUG_ON(!rq->rt.rt_nr_migratory);
1021                         rq->rt.rt_nr_migratory--;
1022                 }
1023
1024                 update_rt_migration(rq);
1025         }
1026
1027         p->cpus_allowed    = *new_mask;
1028         p->rt.nr_cpus_allowed = weight;
1029 }
1030
1031 /* Assumes rq->lock is held */
1032 static void join_domain_rt(struct rq *rq)
1033 {
1034         if (rq->rt.overloaded)
1035                 rt_set_overload(rq);
1036 }
1037
1038 /* Assumes rq->lock is held */
1039 static void leave_domain_rt(struct rq *rq)
1040 {
1041         if (rq->rt.overloaded)
1042                 rt_clear_overload(rq);
1043 }
1044
1045 /*
1046  * When switch from the rt queue, we bring ourselves to a position
1047  * that we might want to pull RT tasks from other runqueues.
1048  */
1049 static void switched_from_rt(struct rq *rq, struct task_struct *p,
1050                            int running)
1051 {
1052         /*
1053          * If there are other RT tasks then we will reschedule
1054          * and the scheduling of the other RT tasks will handle
1055          * the balancing. But if we are the last RT task
1056          * we may need to handle the pulling of RT tasks
1057          * now.
1058          */
1059         if (!rq->rt.rt_nr_running)
1060                 pull_rt_task(rq);
1061 }
1062 #endif /* CONFIG_SMP */
1063
1064 /*
1065  * When switching a task to RT, we may overload the runqueue
1066  * with RT tasks. In this case we try to push them off to
1067  * other runqueues.
1068  */
1069 static void switched_to_rt(struct rq *rq, struct task_struct *p,
1070                            int running)
1071 {
1072         int check_resched = 1;
1073
1074         /*
1075          * If we are already running, then there's nothing
1076          * that needs to be done. But if we are not running
1077          * we may need to preempt the current running task.
1078          * If that current running task is also an RT task
1079          * then see if we can move to another run queue.
1080          */
1081         if (!running) {
1082 #ifdef CONFIG_SMP
1083                 if (rq->rt.overloaded && push_rt_task(rq) &&
1084                     /* Don't resched if we changed runqueues */
1085                     rq != task_rq(p))
1086                         check_resched = 0;
1087 #endif /* CONFIG_SMP */
1088                 if (check_resched && p->prio < rq->curr->prio)
1089                         resched_task(rq->curr);
1090         }
1091 }
1092
1093 /*
1094  * Priority of the task has changed. This may cause
1095  * us to initiate a push or pull.
1096  */
1097 static void prio_changed_rt(struct rq *rq, struct task_struct *p,
1098                             int oldprio, int running)
1099 {
1100         if (running) {
1101 #ifdef CONFIG_SMP
1102                 /*
1103                  * If our priority decreases while running, we
1104                  * may need to pull tasks to this runqueue.
1105                  */
1106                 if (oldprio < p->prio)
1107                         pull_rt_task(rq);
1108                 /*
1109                  * If there's a higher priority task waiting to run
1110                  * then reschedule.
1111                  */
1112                 if (p->prio > rq->rt.highest_prio)
1113                         resched_task(p);
1114 #else
1115                 /* For UP simply resched on drop of prio */
1116                 if (oldprio < p->prio)
1117                         resched_task(p);
1118 #endif /* CONFIG_SMP */
1119         } else {
1120                 /*
1121                  * This task is not running, but if it is
1122                  * greater than the current running task
1123                  * then reschedule.
1124                  */
1125                 if (p->prio < rq->curr->prio)
1126                         resched_task(rq->curr);
1127         }
1128 }
1129
1130 static void watchdog(struct rq *rq, struct task_struct *p)
1131 {
1132         unsigned long soft, hard;
1133
1134         if (!p->signal)
1135                 return;
1136
1137         soft = p->signal->rlim[RLIMIT_RTTIME].rlim_cur;
1138         hard = p->signal->rlim[RLIMIT_RTTIME].rlim_max;
1139
1140         if (soft != RLIM_INFINITY) {
1141                 unsigned long next;
1142
1143                 p->rt.timeout++;
1144                 next = DIV_ROUND_UP(min(soft, hard), USEC_PER_SEC/HZ);
1145                 if (p->rt.timeout > next)
1146                         p->it_sched_expires = p->se.sum_exec_runtime;
1147         }
1148 }
1149
1150 static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued)
1151 {
1152         update_curr_rt(rq);
1153
1154         watchdog(rq, p);
1155
1156         /*
1157          * RR tasks need a special form of timeslice management.
1158          * FIFO tasks have no timeslices.
1159          */
1160         if (p->policy != SCHED_RR)
1161                 return;
1162
1163         if (--p->rt.time_slice)
1164                 return;
1165
1166         p->rt.time_slice = DEF_TIMESLICE;
1167
1168         /*
1169          * Requeue to the end of queue if we are not the only element
1170          * on the queue:
1171          */
1172         if (p->rt.run_list.prev != p->rt.run_list.next) {
1173                 requeue_task_rt(rq, p);
1174                 set_tsk_need_resched(p);
1175         }
1176 }
1177
1178 static void set_curr_task_rt(struct rq *rq)
1179 {
1180         struct task_struct *p = rq->curr;
1181
1182         p->se.exec_start = rq->clock;
1183 }
1184
1185 const struct sched_class rt_sched_class = {
1186         .next                   = &fair_sched_class,
1187         .enqueue_task           = enqueue_task_rt,
1188         .dequeue_task           = dequeue_task_rt,
1189         .yield_task             = yield_task_rt,
1190 #ifdef CONFIG_SMP
1191         .select_task_rq         = select_task_rq_rt,
1192 #endif /* CONFIG_SMP */
1193
1194         .check_preempt_curr     = check_preempt_curr_rt,
1195
1196         .pick_next_task         = pick_next_task_rt,
1197         .put_prev_task          = put_prev_task_rt,
1198
1199 #ifdef CONFIG_SMP
1200         .load_balance           = load_balance_rt,
1201         .move_one_task          = move_one_task_rt,
1202         .set_cpus_allowed       = set_cpus_allowed_rt,
1203         .join_domain            = join_domain_rt,
1204         .leave_domain           = leave_domain_rt,
1205         .pre_schedule           = pre_schedule_rt,
1206         .post_schedule          = post_schedule_rt,
1207         .task_wake_up           = task_wake_up_rt,
1208         .switched_from          = switched_from_rt,
1209 #endif
1210
1211         .set_curr_task          = set_curr_task_rt,
1212         .task_tick              = task_tick_rt,
1213
1214         .prio_changed           = prio_changed_rt,
1215         .switched_to            = switched_to_rt,
1216 };