Root/
1 | /* |
2 | * Real-Time Scheduling Class (mapped to the SCHED_FIFO and SCHED_RR |
3 | * policies) |
4 | */ |
5 | |
6 | #ifdef CONFIG_RT_GROUP_SCHED |
7 | |
8 | #define rt_entity_is_task(rt_se) (!(rt_se)->my_q) |
9 | |
10 | static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se) |
11 | { |
12 | #ifdef CONFIG_SCHED_DEBUG |
13 | WARN_ON_ONCE(!rt_entity_is_task(rt_se)); |
14 | #endif |
15 | return container_of(rt_se, struct task_struct, rt); |
16 | } |
17 | |
18 | static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq) |
19 | { |
20 | return rt_rq->rq; |
21 | } |
22 | |
23 | static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se) |
24 | { |
25 | return rt_se->rt_rq; |
26 | } |
27 | |
28 | #else /* CONFIG_RT_GROUP_SCHED */ |
29 | |
30 | #define rt_entity_is_task(rt_se) (1) |
31 | |
32 | static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se) |
33 | { |
34 | return container_of(rt_se, struct task_struct, rt); |
35 | } |
36 | |
37 | static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq) |
38 | { |
39 | return container_of(rt_rq, struct rq, rt); |
40 | } |
41 | |
42 | static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se) |
43 | { |
44 | struct task_struct *p = rt_task_of(rt_se); |
45 | struct rq *rq = task_rq(p); |
46 | |
47 | return &rq->rt; |
48 | } |
49 | |
50 | #endif /* CONFIG_RT_GROUP_SCHED */ |
51 | |
52 | #ifdef CONFIG_SMP |
53 | |
54 | static inline int rt_overloaded(struct rq *rq) |
55 | { |
56 | return atomic_read(&rq->rd->rto_count); |
57 | } |
58 | |
59 | static inline void rt_set_overload(struct rq *rq) |
60 | { |
61 | if (!rq->online) |
62 | return; |
63 | |
64 | cpumask_set_cpu(rq->cpu, rq->rd->rto_mask); |
65 | /* |
66 | * Make sure the mask is visible before we set |
67 | * the overload count. That is checked to determine |
68 | * if we should look at the mask. It would be a shame |
69 | * if we looked at the mask, but the mask was not |
70 | * updated yet. |
71 | */ |
72 | wmb(); |
73 | atomic_inc(&rq->rd->rto_count); |
74 | } |
75 | |
76 | static inline void rt_clear_overload(struct rq *rq) |
77 | { |
78 | if (!rq->online) |
79 | return; |
80 | |
81 | /* the order here really doesn't matter */ |
82 | atomic_dec(&rq->rd->rto_count); |
83 | cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask); |
84 | } |
85 | |
86 | static void update_rt_migration(struct rt_rq *rt_rq) |
87 | { |
88 | if (rt_rq->rt_nr_migratory && rt_rq->rt_nr_total > 1) { |
89 | if (!rt_rq->overloaded) { |
90 | rt_set_overload(rq_of_rt_rq(rt_rq)); |
91 | rt_rq->overloaded = 1; |
92 | } |
93 | } else if (rt_rq->overloaded) { |
94 | rt_clear_overload(rq_of_rt_rq(rt_rq)); |
95 | rt_rq->overloaded = 0; |
96 | } |
97 | } |
98 | |
99 | static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
100 | { |
101 | if (!rt_entity_is_task(rt_se)) |
102 | return; |
103 | |
104 | rt_rq = &rq_of_rt_rq(rt_rq)->rt; |
105 | |
106 | rt_rq->rt_nr_total++; |
107 | if (rt_se->nr_cpus_allowed > 1) |
108 | rt_rq->rt_nr_migratory++; |
109 | |
110 | update_rt_migration(rt_rq); |
111 | } |
112 | |
113 | static void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
114 | { |
115 | if (!rt_entity_is_task(rt_se)) |
116 | return; |
117 | |
118 | rt_rq = &rq_of_rt_rq(rt_rq)->rt; |
119 | |
120 | rt_rq->rt_nr_total--; |
121 | if (rt_se->nr_cpus_allowed > 1) |
122 | rt_rq->rt_nr_migratory--; |
123 | |
124 | update_rt_migration(rt_rq); |
125 | } |
126 | |
127 | static void enqueue_pushable_task(struct rq *rq, struct task_struct *p) |
128 | { |
129 | plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks); |
130 | plist_node_init(&p->pushable_tasks, p->prio); |
131 | plist_add(&p->pushable_tasks, &rq->rt.pushable_tasks); |
132 | } |
133 | |
134 | static void dequeue_pushable_task(struct rq *rq, struct task_struct *p) |
135 | { |
136 | plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks); |
137 | } |
138 | |
139 | static inline int has_pushable_tasks(struct rq *rq) |
140 | { |
141 | return !plist_head_empty(&rq->rt.pushable_tasks); |
142 | } |
143 | |
144 | #else |
145 | |
146 | static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p) |
147 | { |
148 | } |
149 | |
150 | static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p) |
151 | { |
152 | } |
153 | |
154 | static inline |
155 | void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
156 | { |
157 | } |
158 | |
159 | static inline |
160 | void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
161 | { |
162 | } |
163 | |
164 | #endif /* CONFIG_SMP */ |
165 | |
166 | static inline int on_rt_rq(struct sched_rt_entity *rt_se) |
167 | { |
168 | return !list_empty(&rt_se->run_list); |
169 | } |
170 | |
171 | #ifdef CONFIG_RT_GROUP_SCHED |
172 | |
173 | static inline u64 sched_rt_runtime(struct rt_rq *rt_rq) |
174 | { |
175 | if (!rt_rq->tg) |
176 | return RUNTIME_INF; |
177 | |
178 | return rt_rq->rt_runtime; |
179 | } |
180 | |
181 | static inline u64 sched_rt_period(struct rt_rq *rt_rq) |
182 | { |
183 | return ktime_to_ns(rt_rq->tg->rt_bandwidth.rt_period); |
184 | } |
185 | |
186 | #define for_each_leaf_rt_rq(rt_rq, rq) \ |
187 | list_for_each_entry_rcu(rt_rq, &rq->leaf_rt_rq_list, leaf_rt_rq_list) |
188 | |
189 | #define for_each_sched_rt_entity(rt_se) \ |
190 | for (; rt_se; rt_se = rt_se->parent) |
191 | |
192 | static inline struct rt_rq *group_rt_rq(struct sched_rt_entity *rt_se) |
193 | { |
194 | return rt_se->my_q; |
195 | } |
196 | |
197 | static void enqueue_rt_entity(struct sched_rt_entity *rt_se); |
198 | static void dequeue_rt_entity(struct sched_rt_entity *rt_se); |
199 | |
200 | static void sched_rt_rq_enqueue(struct rt_rq *rt_rq) |
201 | { |
202 | struct task_struct *curr = rq_of_rt_rq(rt_rq)->curr; |
203 | struct sched_rt_entity *rt_se = rt_rq->rt_se; |
204 | |
205 | if (rt_rq->rt_nr_running) { |
206 | if (rt_se && !on_rt_rq(rt_se)) |
207 | enqueue_rt_entity(rt_se); |
208 | if (rt_rq->highest_prio.curr < curr->prio) |
209 | resched_task(curr); |
210 | } |
211 | } |
212 | |
213 | static void sched_rt_rq_dequeue(struct rt_rq *rt_rq) |
214 | { |
215 | struct sched_rt_entity *rt_se = rt_rq->rt_se; |
216 | |
217 | if (rt_se && on_rt_rq(rt_se)) |
218 | dequeue_rt_entity(rt_se); |
219 | } |
220 | |
221 | static inline int rt_rq_throttled(struct rt_rq *rt_rq) |
222 | { |
223 | return rt_rq->rt_throttled && !rt_rq->rt_nr_boosted; |
224 | } |
225 | |
226 | static int rt_se_boosted(struct sched_rt_entity *rt_se) |
227 | { |
228 | struct rt_rq *rt_rq = group_rt_rq(rt_se); |
229 | struct task_struct *p; |
230 | |
231 | if (rt_rq) |
232 | return !!rt_rq->rt_nr_boosted; |
233 | |
234 | p = rt_task_of(rt_se); |
235 | return p->prio != p->normal_prio; |
236 | } |
237 | |
238 | #ifdef CONFIG_SMP |
239 | static inline const struct cpumask *sched_rt_period_mask(void) |
240 | { |
241 | return cpu_rq(smp_processor_id())->rd->span; |
242 | } |
243 | #else |
244 | static inline const struct cpumask *sched_rt_period_mask(void) |
245 | { |
246 | return cpu_online_mask; |
247 | } |
248 | #endif |
249 | |
250 | static inline |
251 | struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu) |
252 | { |
253 | return container_of(rt_b, struct task_group, rt_bandwidth)->rt_rq[cpu]; |
254 | } |
255 | |
256 | static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq) |
257 | { |
258 | return &rt_rq->tg->rt_bandwidth; |
259 | } |
260 | |
261 | #else /* !CONFIG_RT_GROUP_SCHED */ |
262 | |
263 | static inline u64 sched_rt_runtime(struct rt_rq *rt_rq) |
264 | { |
265 | return rt_rq->rt_runtime; |
266 | } |
267 | |
268 | static inline u64 sched_rt_period(struct rt_rq *rt_rq) |
269 | { |
270 | return ktime_to_ns(def_rt_bandwidth.rt_period); |
271 | } |
272 | |
273 | #define for_each_leaf_rt_rq(rt_rq, rq) \ |
274 | for (rt_rq = &rq->rt; rt_rq; rt_rq = NULL) |
275 | |
276 | #define for_each_sched_rt_entity(rt_se) \ |
277 | for (; rt_se; rt_se = NULL) |
278 | |
279 | static inline struct rt_rq *group_rt_rq(struct sched_rt_entity *rt_se) |
280 | { |
281 | return NULL; |
282 | } |
283 | |
284 | static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq) |
285 | { |
286 | if (rt_rq->rt_nr_running) |
287 | resched_task(rq_of_rt_rq(rt_rq)->curr); |
288 | } |
289 | |
290 | static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq) |
291 | { |
292 | } |
293 | |
294 | static inline int rt_rq_throttled(struct rt_rq *rt_rq) |
295 | { |
296 | return rt_rq->rt_throttled; |
297 | } |
298 | |
299 | static inline const struct cpumask *sched_rt_period_mask(void) |
300 | { |
301 | return cpu_online_mask; |
302 | } |
303 | |
304 | static inline |
305 | struct rt_rq *sched_rt_period_rt_rq(struct rt_bandwidth *rt_b, int cpu) |
306 | { |
307 | return &cpu_rq(cpu)->rt; |
308 | } |
309 | |
310 | static inline struct rt_bandwidth *sched_rt_bandwidth(struct rt_rq *rt_rq) |
311 | { |
312 | return &def_rt_bandwidth; |
313 | } |
314 | |
315 | #endif /* CONFIG_RT_GROUP_SCHED */ |
316 | |
317 | #ifdef CONFIG_SMP |
318 | /* |
319 | * We ran out of runtime, see if we can borrow some from our neighbours. |
320 | */ |
321 | static int do_balance_runtime(struct rt_rq *rt_rq) |
322 | { |
323 | struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq); |
324 | struct root_domain *rd = cpu_rq(smp_processor_id())->rd; |
325 | int i, weight, more = 0; |
326 | u64 rt_period; |
327 | |
328 | weight = cpumask_weight(rd->span); |
329 | |
330 | spin_lock(&rt_b->rt_runtime_lock); |
331 | rt_period = ktime_to_ns(rt_b->rt_period); |
332 | for_each_cpu(i, rd->span) { |
333 | struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i); |
334 | s64 diff; |
335 | |
336 | if (iter == rt_rq) |
337 | continue; |
338 | |
339 | spin_lock(&iter->rt_runtime_lock); |
340 | /* |
341 | * Either all rqs have inf runtime and there's nothing to steal |
342 | * or __disable_runtime() below sets a specific rq to inf to |
343 | * indicate its been disabled and disalow stealing. |
344 | */ |
345 | if (iter->rt_runtime == RUNTIME_INF) |
346 | goto next; |
347 | |
348 | /* |
349 | * From runqueues with spare time, take 1/n part of their |
350 | * spare time, but no more than our period. |
351 | */ |
352 | diff = iter->rt_runtime - iter->rt_time; |
353 | if (diff > 0) { |
354 | diff = div_u64((u64)diff, weight); |
355 | if (rt_rq->rt_runtime + diff > rt_period) |
356 | diff = rt_period - rt_rq->rt_runtime; |
357 | iter->rt_runtime -= diff; |
358 | rt_rq->rt_runtime += diff; |
359 | more = 1; |
360 | if (rt_rq->rt_runtime == rt_period) { |
361 | spin_unlock(&iter->rt_runtime_lock); |
362 | break; |
363 | } |
364 | } |
365 | next: |
366 | spin_unlock(&iter->rt_runtime_lock); |
367 | } |
368 | spin_unlock(&rt_b->rt_runtime_lock); |
369 | |
370 | return more; |
371 | } |
372 | |
373 | /* |
374 | * Ensure this RQ takes back all the runtime it lend to its neighbours. |
375 | */ |
376 | static void __disable_runtime(struct rq *rq) |
377 | { |
378 | struct root_domain *rd = rq->rd; |
379 | struct rt_rq *rt_rq; |
380 | |
381 | if (unlikely(!scheduler_running)) |
382 | return; |
383 | |
384 | for_each_leaf_rt_rq(rt_rq, rq) { |
385 | struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq); |
386 | s64 want; |
387 | int i; |
388 | |
389 | spin_lock(&rt_b->rt_runtime_lock); |
390 | spin_lock(&rt_rq->rt_runtime_lock); |
391 | /* |
392 | * Either we're all inf and nobody needs to borrow, or we're |
393 | * already disabled and thus have nothing to do, or we have |
394 | * exactly the right amount of runtime to take out. |
395 | */ |
396 | if (rt_rq->rt_runtime == RUNTIME_INF || |
397 | rt_rq->rt_runtime == rt_b->rt_runtime) |
398 | goto balanced; |
399 | spin_unlock(&rt_rq->rt_runtime_lock); |
400 | |
401 | /* |
402 | * Calculate the difference between what we started out with |
403 | * and what we current have, that's the amount of runtime |
404 | * we lend and now have to reclaim. |
405 | */ |
406 | want = rt_b->rt_runtime - rt_rq->rt_runtime; |
407 | |
408 | /* |
409 | * Greedy reclaim, take back as much as we can. |
410 | */ |
411 | for_each_cpu(i, rd->span) { |
412 | struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i); |
413 | s64 diff; |
414 | |
415 | /* |
416 | * Can't reclaim from ourselves or disabled runqueues. |
417 | */ |
418 | if (iter == rt_rq || iter->rt_runtime == RUNTIME_INF) |
419 | continue; |
420 | |
421 | spin_lock(&iter->rt_runtime_lock); |
422 | if (want > 0) { |
423 | diff = min_t(s64, iter->rt_runtime, want); |
424 | iter->rt_runtime -= diff; |
425 | want -= diff; |
426 | } else { |
427 | iter->rt_runtime -= want; |
428 | want -= want; |
429 | } |
430 | spin_unlock(&iter->rt_runtime_lock); |
431 | |
432 | if (!want) |
433 | break; |
434 | } |
435 | |
436 | spin_lock(&rt_rq->rt_runtime_lock); |
437 | /* |
438 | * We cannot be left wanting - that would mean some runtime |
439 | * leaked out of the system. |
440 | */ |
441 | BUG_ON(want); |
442 | balanced: |
443 | /* |
444 | * Disable all the borrow logic by pretending we have inf |
445 | * runtime - in which case borrowing doesn't make sense. |
446 | */ |
447 | rt_rq->rt_runtime = RUNTIME_INF; |
448 | spin_unlock(&rt_rq->rt_runtime_lock); |
449 | spin_unlock(&rt_b->rt_runtime_lock); |
450 | } |
451 | } |
452 | |
453 | static void disable_runtime(struct rq *rq) |
454 | { |
455 | unsigned long flags; |
456 | |
457 | spin_lock_irqsave(&rq->lock, flags); |
458 | __disable_runtime(rq); |
459 | spin_unlock_irqrestore(&rq->lock, flags); |
460 | } |
461 | |
462 | static void __enable_runtime(struct rq *rq) |
463 | { |
464 | struct rt_rq *rt_rq; |
465 | |
466 | if (unlikely(!scheduler_running)) |
467 | return; |
468 | |
469 | /* |
470 | * Reset each runqueue's bandwidth settings |
471 | */ |
472 | for_each_leaf_rt_rq(rt_rq, rq) { |
473 | struct rt_bandwidth *rt_b = sched_rt_bandwidth(rt_rq); |
474 | |
475 | spin_lock(&rt_b->rt_runtime_lock); |
476 | spin_lock(&rt_rq->rt_runtime_lock); |
477 | rt_rq->rt_runtime = rt_b->rt_runtime; |
478 | rt_rq->rt_time = 0; |
479 | rt_rq->rt_throttled = 0; |
480 | spin_unlock(&rt_rq->rt_runtime_lock); |
481 | spin_unlock(&rt_b->rt_runtime_lock); |
482 | } |
483 | } |
484 | |
485 | static void enable_runtime(struct rq *rq) |
486 | { |
487 | unsigned long flags; |
488 | |
489 | spin_lock_irqsave(&rq->lock, flags); |
490 | __enable_runtime(rq); |
491 | spin_unlock_irqrestore(&rq->lock, flags); |
492 | } |
493 | |
494 | static int balance_runtime(struct rt_rq *rt_rq) |
495 | { |
496 | int more = 0; |
497 | |
498 | if (rt_rq->rt_time > rt_rq->rt_runtime) { |
499 | spin_unlock(&rt_rq->rt_runtime_lock); |
500 | more = do_balance_runtime(rt_rq); |
501 | spin_lock(&rt_rq->rt_runtime_lock); |
502 | } |
503 | |
504 | return more; |
505 | } |
506 | #else /* !CONFIG_SMP */ |
507 | static inline int balance_runtime(struct rt_rq *rt_rq) |
508 | { |
509 | return 0; |
510 | } |
511 | #endif /* CONFIG_SMP */ |
512 | |
513 | static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun) |
514 | { |
515 | int i, idle = 1; |
516 | const struct cpumask *span; |
517 | |
518 | if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF) |
519 | return 1; |
520 | |
521 | span = sched_rt_period_mask(); |
522 | for_each_cpu(i, span) { |
523 | int enqueue = 0; |
524 | struct rt_rq *rt_rq = sched_rt_period_rt_rq(rt_b, i); |
525 | struct rq *rq = rq_of_rt_rq(rt_rq); |
526 | |
527 | spin_lock(&rq->lock); |
528 | if (rt_rq->rt_time) { |
529 | u64 runtime; |
530 | |
531 | spin_lock(&rt_rq->rt_runtime_lock); |
532 | if (rt_rq->rt_throttled) |
533 | balance_runtime(rt_rq); |
534 | runtime = rt_rq->rt_runtime; |
535 | rt_rq->rt_time -= min(rt_rq->rt_time, overrun*runtime); |
536 | if (rt_rq->rt_throttled && rt_rq->rt_time < runtime) { |
537 | rt_rq->rt_throttled = 0; |
538 | enqueue = 1; |
539 | } |
540 | if (rt_rq->rt_time || rt_rq->rt_nr_running) |
541 | idle = 0; |
542 | spin_unlock(&rt_rq->rt_runtime_lock); |
543 | } else if (rt_rq->rt_nr_running) |
544 | idle = 0; |
545 | |
546 | if (enqueue) |
547 | sched_rt_rq_enqueue(rt_rq); |
548 | spin_unlock(&rq->lock); |
549 | } |
550 | |
551 | return idle; |
552 | } |
553 | |
554 | static inline int rt_se_prio(struct sched_rt_entity *rt_se) |
555 | { |
556 | #ifdef CONFIG_RT_GROUP_SCHED |
557 | struct rt_rq *rt_rq = group_rt_rq(rt_se); |
558 | |
559 | if (rt_rq) |
560 | return rt_rq->highest_prio.curr; |
561 | #endif |
562 | |
563 | return rt_task_of(rt_se)->prio; |
564 | } |
565 | |
566 | static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq) |
567 | { |
568 | u64 runtime = sched_rt_runtime(rt_rq); |
569 | |
570 | if (rt_rq->rt_throttled) |
571 | return rt_rq_throttled(rt_rq); |
572 | |
573 | if (sched_rt_runtime(rt_rq) >= sched_rt_period(rt_rq)) |
574 | return 0; |
575 | |
576 | balance_runtime(rt_rq); |
577 | runtime = sched_rt_runtime(rt_rq); |
578 | if (runtime == RUNTIME_INF) |
579 | return 0; |
580 | |
581 | if (rt_rq->rt_time > runtime) { |
582 | rt_rq->rt_throttled = 1; |
583 | if (rt_rq_throttled(rt_rq)) { |
584 | sched_rt_rq_dequeue(rt_rq); |
585 | return 1; |
586 | } |
587 | } |
588 | |
589 | return 0; |
590 | } |
591 | |
592 | /* |
593 | * Update the current task's runtime statistics. Skip current tasks that |
594 | * are not in our scheduling class. |
595 | */ |
596 | static void update_curr_rt(struct rq *rq) |
597 | { |
598 | struct task_struct *curr = rq->curr; |
599 | struct sched_rt_entity *rt_se = &curr->rt; |
600 | struct rt_rq *rt_rq = rt_rq_of_se(rt_se); |
601 | u64 delta_exec; |
602 | |
603 | if (!task_has_rt_policy(curr)) |
604 | return; |
605 | |
606 | delta_exec = rq->clock - curr->se.exec_start; |
607 | if (unlikely((s64)delta_exec < 0)) |
608 | delta_exec = 0; |
609 | |
610 | schedstat_set(curr->se.exec_max, max(curr->se.exec_max, delta_exec)); |
611 | |
612 | curr->se.sum_exec_runtime += delta_exec; |
613 | account_group_exec_runtime(curr, delta_exec); |
614 | |
615 | curr->se.exec_start = rq->clock; |
616 | cpuacct_charge(curr, delta_exec); |
617 | |
618 | sched_rt_avg_update(rq, delta_exec); |
619 | |
620 | if (!rt_bandwidth_enabled()) |
621 | return; |
622 | |
623 | for_each_sched_rt_entity(rt_se) { |
624 | rt_rq = rt_rq_of_se(rt_se); |
625 | |
626 | if (sched_rt_runtime(rt_rq) != RUNTIME_INF) { |
627 | spin_lock(&rt_rq->rt_runtime_lock); |
628 | rt_rq->rt_time += delta_exec; |
629 | if (sched_rt_runtime_exceeded(rt_rq)) |
630 | resched_task(curr); |
631 | spin_unlock(&rt_rq->rt_runtime_lock); |
632 | } |
633 | } |
634 | } |
635 | |
636 | #if defined CONFIG_SMP |
637 | |
638 | static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu); |
639 | |
640 | static inline int next_prio(struct rq *rq) |
641 | { |
642 | struct task_struct *next = pick_next_highest_task_rt(rq, rq->cpu); |
643 | |
644 | if (next && rt_prio(next->prio)) |
645 | return next->prio; |
646 | else |
647 | return MAX_RT_PRIO; |
648 | } |
649 | |
650 | static void |
651 | inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) |
652 | { |
653 | struct rq *rq = rq_of_rt_rq(rt_rq); |
654 | |
655 | if (prio < prev_prio) { |
656 | |
657 | /* |
658 | * If the new task is higher in priority than anything on the |
659 | * run-queue, we know that the previous high becomes our |
660 | * next-highest. |
661 | */ |
662 | rt_rq->highest_prio.next = prev_prio; |
663 | |
664 | if (rq->online) |
665 | cpupri_set(&rq->rd->cpupri, rq->cpu, prio); |
666 | |
667 | } else if (prio == rt_rq->highest_prio.curr) |
668 | /* |
669 | * If the next task is equal in priority to the highest on |
670 | * the run-queue, then we implicitly know that the next highest |
671 | * task cannot be any lower than current |
672 | */ |
673 | rt_rq->highest_prio.next = prio; |
674 | else if (prio < rt_rq->highest_prio.next) |
675 | /* |
676 | * Otherwise, we need to recompute next-highest |
677 | */ |
678 | rt_rq->highest_prio.next = next_prio(rq); |
679 | } |
680 | |
681 | static void |
682 | dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) |
683 | { |
684 | struct rq *rq = rq_of_rt_rq(rt_rq); |
685 | |
686 | if (rt_rq->rt_nr_running && (prio <= rt_rq->highest_prio.next)) |
687 | rt_rq->highest_prio.next = next_prio(rq); |
688 | |
689 | if (rq->online && rt_rq->highest_prio.curr != prev_prio) |
690 | cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr); |
691 | } |
692 | |
693 | #else /* CONFIG_SMP */ |
694 | |
695 | static inline |
696 | void inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) {} |
697 | static inline |
698 | void dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) {} |
699 | |
700 | #endif /* CONFIG_SMP */ |
701 | |
702 | #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED |
703 | static void |
704 | inc_rt_prio(struct rt_rq *rt_rq, int prio) |
705 | { |
706 | int prev_prio = rt_rq->highest_prio.curr; |
707 | |
708 | if (prio < prev_prio) |
709 | rt_rq->highest_prio.curr = prio; |
710 | |
711 | inc_rt_prio_smp(rt_rq, prio, prev_prio); |
712 | } |
713 | |
714 | static void |
715 | dec_rt_prio(struct rt_rq *rt_rq, int prio) |
716 | { |
717 | int prev_prio = rt_rq->highest_prio.curr; |
718 | |
719 | if (rt_rq->rt_nr_running) { |
720 | |
721 | WARN_ON(prio < prev_prio); |
722 | |
723 | /* |
724 | * This may have been our highest task, and therefore |
725 | * we may have some recomputation to do |
726 | */ |
727 | if (prio == prev_prio) { |
728 | struct rt_prio_array *array = &rt_rq->active; |
729 | |
730 | rt_rq->highest_prio.curr = |
731 | sched_find_first_bit(array->bitmap); |
732 | } |
733 | |
734 | } else |
735 | rt_rq->highest_prio.curr = MAX_RT_PRIO; |
736 | |
737 | dec_rt_prio_smp(rt_rq, prio, prev_prio); |
738 | } |
739 | |
740 | #else |
741 | |
742 | static inline void inc_rt_prio(struct rt_rq *rt_rq, int prio) {} |
743 | static inline void dec_rt_prio(struct rt_rq *rt_rq, int prio) {} |
744 | |
745 | #endif /* CONFIG_SMP || CONFIG_RT_GROUP_SCHED */ |
746 | |
747 | #ifdef CONFIG_RT_GROUP_SCHED |
748 | |
749 | static void |
750 | inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
751 | { |
752 | if (rt_se_boosted(rt_se)) |
753 | rt_rq->rt_nr_boosted++; |
754 | |
755 | if (rt_rq->tg) |
756 | start_rt_bandwidth(&rt_rq->tg->rt_bandwidth); |
757 | } |
758 | |
759 | static void |
760 | dec_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
761 | { |
762 | if (rt_se_boosted(rt_se)) |
763 | rt_rq->rt_nr_boosted--; |
764 | |
765 | WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted); |
766 | } |
767 | |
768 | #else /* CONFIG_RT_GROUP_SCHED */ |
769 | |
770 | static void |
771 | inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
772 | { |
773 | start_rt_bandwidth(&def_rt_bandwidth); |
774 | } |
775 | |
776 | static inline |
777 | void dec_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {} |
778 | |
779 | #endif /* CONFIG_RT_GROUP_SCHED */ |
780 | |
781 | static inline |
782 | void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
783 | { |
784 | int prio = rt_se_prio(rt_se); |
785 | |
786 | WARN_ON(!rt_prio(prio)); |
787 | rt_rq->rt_nr_running++; |
788 | |
789 | inc_rt_prio(rt_rq, prio); |
790 | inc_rt_migration(rt_se, rt_rq); |
791 | inc_rt_group(rt_se, rt_rq); |
792 | } |
793 | |
794 | static inline |
795 | void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) |
796 | { |
797 | WARN_ON(!rt_prio(rt_se_prio(rt_se))); |
798 | WARN_ON(!rt_rq->rt_nr_running); |
799 | rt_rq->rt_nr_running--; |
800 | |
801 | dec_rt_prio(rt_rq, rt_se_prio(rt_se)); |
802 | dec_rt_migration(rt_se, rt_rq); |
803 | dec_rt_group(rt_se, rt_rq); |
804 | } |
805 | |
806 | static void __enqueue_rt_entity(struct sched_rt_entity *rt_se) |
807 | { |
808 | struct rt_rq *rt_rq = rt_rq_of_se(rt_se); |
809 | struct rt_prio_array *array = &rt_rq->active; |
810 | struct rt_rq *group_rq = group_rt_rq(rt_se); |
811 | struct list_head *queue = array->queue + rt_se_prio(rt_se); |
812 | |
813 | /* |
814 | * Don't enqueue the group if its throttled, or when empty. |
815 | * The latter is a consequence of the former when a child group |
816 | * get throttled and the current group doesn't have any other |
817 | * active members. |
818 | */ |
819 | if (group_rq && (rt_rq_throttled(group_rq) || !group_rq->rt_nr_running)) |
820 | return; |
821 | |
822 | list_add_tail(&rt_se->run_list, queue); |
823 | __set_bit(rt_se_prio(rt_se), array->bitmap); |
824 | |
825 | inc_rt_tasks(rt_se, rt_rq); |
826 | } |
827 | |
828 | static void __dequeue_rt_entity(struct sched_rt_entity *rt_se) |
829 | { |
830 | struct rt_rq *rt_rq = rt_rq_of_se(rt_se); |
831 | struct rt_prio_array *array = &rt_rq->active; |
832 | |
833 | list_del_init(&rt_se->run_list); |
834 | if (list_empty(array->queue + rt_se_prio(rt_se))) |
835 | __clear_bit(rt_se_prio(rt_se), array->bitmap); |
836 | |
837 | dec_rt_tasks(rt_se, rt_rq); |
838 | } |
839 | |
840 | /* |
841 | * Because the prio of an upper entry depends on the lower |
842 | * entries, we must remove entries top - down. |
843 | */ |
844 | static void dequeue_rt_stack(struct sched_rt_entity *rt_se) |
845 | { |
846 | struct sched_rt_entity *back = NULL; |
847 | |
848 | for_each_sched_rt_entity(rt_se) { |
849 | rt_se->back = back; |
850 | back = rt_se; |
851 | } |
852 | |
853 | for (rt_se = back; rt_se; rt_se = rt_se->back) { |
854 | if (on_rt_rq(rt_se)) |
855 | __dequeue_rt_entity(rt_se); |
856 | } |
857 | } |
858 | |
859 | static void enqueue_rt_entity(struct sched_rt_entity *rt_se) |
860 | { |
861 | dequeue_rt_stack(rt_se); |
862 | for_each_sched_rt_entity(rt_se) |
863 | __enqueue_rt_entity(rt_se); |
864 | } |
865 | |
866 | static void dequeue_rt_entity(struct sched_rt_entity *rt_se) |
867 | { |
868 | dequeue_rt_stack(rt_se); |
869 | |
870 | for_each_sched_rt_entity(rt_se) { |
871 | struct rt_rq *rt_rq = group_rt_rq(rt_se); |
872 | |
873 | if (rt_rq && rt_rq->rt_nr_running) |
874 | __enqueue_rt_entity(rt_se); |
875 | } |
876 | } |
877 | |
878 | /* |
879 | * Adding/removing a task to/from a priority array: |
880 | */ |
881 | static void enqueue_task_rt(struct rq *rq, struct task_struct *p, int wakeup) |
882 | { |
883 | struct sched_rt_entity *rt_se = &p->rt; |
884 | |
885 | if (wakeup) |
886 | rt_se->timeout = 0; |
887 | |
888 | enqueue_rt_entity(rt_se); |
889 | |
890 | if (!task_current(rq, p) && p->rt.nr_cpus_allowed > 1) |
891 | enqueue_pushable_task(rq, p); |
892 | } |
893 | |
894 | static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep) |
895 | { |
896 | struct sched_rt_entity *rt_se = &p->rt; |
897 | |
898 | update_curr_rt(rq); |
899 | dequeue_rt_entity(rt_se); |
900 | |
901 | dequeue_pushable_task(rq, p); |
902 | } |
903 | |
904 | /* |
905 | * Put task to the end of the run list without the overhead of dequeue |
906 | * followed by enqueue. |
907 | */ |
908 | static void |
909 | requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head) |
910 | { |
911 | if (on_rt_rq(rt_se)) { |
912 | struct rt_prio_array *array = &rt_rq->active; |
913 | struct list_head *queue = array->queue + rt_se_prio(rt_se); |
914 | |
915 | if (head) |
916 | list_move(&rt_se->run_list, queue); |
917 | else |
918 | list_move_tail(&rt_se->run_list, queue); |
919 | } |
920 | } |
921 | |
922 | static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head) |
923 | { |
924 | struct sched_rt_entity *rt_se = &p->rt; |
925 | struct rt_rq *rt_rq; |
926 | |
927 | for_each_sched_rt_entity(rt_se) { |
928 | rt_rq = rt_rq_of_se(rt_se); |
929 | requeue_rt_entity(rt_rq, rt_se, head); |
930 | } |
931 | } |
932 | |
933 | static void yield_task_rt(struct rq *rq) |
934 | { |
935 | requeue_task_rt(rq, rq->curr, 0); |
936 | } |
937 | |
938 | #ifdef CONFIG_SMP |
939 | static int find_lowest_rq(struct task_struct *task); |
940 | |
941 | static int select_task_rq_rt(struct task_struct *p, int sd_flag, int flags) |
942 | { |
943 | struct rq *rq = task_rq(p); |
944 | |
945 | if (sd_flag != SD_BALANCE_WAKE) |
946 | return smp_processor_id(); |
947 | |
948 | /* |
949 | * If the current task is an RT task, then |
950 | * try to see if we can wake this RT task up on another |
951 | * runqueue. Otherwise simply start this RT task |
952 | * on its current runqueue. |
953 | * |
954 | * We want to avoid overloading runqueues. Even if |
955 | * the RT task is of higher priority than the current RT task. |
956 | * RT tasks behave differently than other tasks. If |
957 | * one gets preempted, we try to push it off to another queue. |
958 | * So trying to keep a preempting RT task on the same |
959 | * cache hot CPU will force the running RT task to |
960 | * a cold CPU. So we waste all the cache for the lower |
961 | * RT task in hopes of saving some of a RT task |
962 | * that is just being woken and probably will have |
963 | * cold cache anyway. |
964 | */ |
965 | if (unlikely(rt_task(rq->curr)) && |
966 | (p->rt.nr_cpus_allowed > 1)) { |
967 | int cpu = find_lowest_rq(p); |
968 | |
969 | return (cpu == -1) ? task_cpu(p) : cpu; |
970 | } |
971 | |
972 | /* |
973 | * Otherwise, just let it ride on the affined RQ and the |
974 | * post-schedule router will push the preempted task away |
975 | */ |
976 | return task_cpu(p); |
977 | } |
978 | |
979 | static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p) |
980 | { |
981 | if (rq->curr->rt.nr_cpus_allowed == 1) |
982 | return; |
983 | |
984 | if (p->rt.nr_cpus_allowed != 1 |
985 | && cpupri_find(&rq->rd->cpupri, p, NULL)) |
986 | return; |
987 | |
988 | if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL)) |
989 | return; |
990 | |
991 | /* |
992 | * There appears to be other cpus that can accept |
993 | * current and none to run 'p', so lets reschedule |
994 | * to try and push current away: |
995 | */ |
996 | requeue_task_rt(rq, p, 1); |
997 | resched_task(rq->curr); |
998 | } |
999 | |
1000 | #endif /* CONFIG_SMP */ |
1001 | |
1002 | /* |
1003 | * Preempt the current task with a newly woken task if needed: |
1004 | */ |
1005 | static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flags) |
1006 | { |
1007 | if (p->prio < rq->curr->prio) { |
1008 | resched_task(rq->curr); |
1009 | return; |
1010 | } |
1011 | |
1012 | #ifdef CONFIG_SMP |
1013 | /* |
1014 | * If: |
1015 | * |
1016 | * - the newly woken task is of equal priority to the current task |
1017 | * - the newly woken task is non-migratable while current is migratable |
1018 | * - current will be preempted on the next reschedule |
1019 | * |
1020 | * we should check to see if current can readily move to a different |
1021 | * cpu. If so, we will reschedule to allow the push logic to try |
1022 | * to move current somewhere else, making room for our non-migratable |
1023 | * task. |
1024 | */ |
1025 | if (p->prio == rq->curr->prio && !need_resched()) |
1026 | check_preempt_equal_prio(rq, p); |
1027 | #endif |
1028 | } |
1029 | |
1030 | static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq, |
1031 | struct rt_rq *rt_rq) |
1032 | { |
1033 | struct rt_prio_array *array = &rt_rq->active; |
1034 | struct sched_rt_entity *next = NULL; |
1035 | struct list_head *queue; |
1036 | int idx; |
1037 | |
1038 | idx = sched_find_first_bit(array->bitmap); |
1039 | BUG_ON(idx >= MAX_RT_PRIO); |
1040 | |
1041 | queue = array->queue + idx; |
1042 | next = list_entry(queue->next, struct sched_rt_entity, run_list); |
1043 | |
1044 | return next; |
1045 | } |
1046 | |
1047 | static struct task_struct *_pick_next_task_rt(struct rq *rq) |
1048 | { |
1049 | struct sched_rt_entity *rt_se; |
1050 | struct task_struct *p; |
1051 | struct rt_rq *rt_rq; |
1052 | |
1053 | rt_rq = &rq->rt; |
1054 | |
1055 | if (unlikely(!rt_rq->rt_nr_running)) |
1056 | return NULL; |
1057 | |
1058 | if (rt_rq_throttled(rt_rq)) |
1059 | return NULL; |
1060 | |
1061 | do { |
1062 | rt_se = pick_next_rt_entity(rq, rt_rq); |
1063 | BUG_ON(!rt_se); |
1064 | rt_rq = group_rt_rq(rt_se); |
1065 | } while (rt_rq); |
1066 | |
1067 | p = rt_task_of(rt_se); |
1068 | p->se.exec_start = rq->clock; |
1069 | |
1070 | return p; |
1071 | } |
1072 | |
1073 | static struct task_struct *pick_next_task_rt(struct rq *rq) |
1074 | { |
1075 | struct task_struct *p = _pick_next_task_rt(rq); |
1076 | |
1077 | /* The running task is never eligible for pushing */ |
1078 | if (p) |
1079 | dequeue_pushable_task(rq, p); |
1080 | |
1081 | #ifdef CONFIG_SMP |
1082 | /* |
1083 | * We detect this state here so that we can avoid taking the RQ |
1084 | * lock again later if there is no need to push |
1085 | */ |
1086 | rq->post_schedule = has_pushable_tasks(rq); |
1087 | #endif |
1088 | |
1089 | return p; |
1090 | } |
1091 | |
1092 | static void put_prev_task_rt(struct rq *rq, struct task_struct *p) |
1093 | { |
1094 | update_curr_rt(rq); |
1095 | p->se.exec_start = 0; |
1096 | |
1097 | /* |
1098 | * The previous task needs to be made eligible for pushing |
1099 | * if it is still active |
1100 | */ |
1101 | if (p->se.on_rq && p->rt.nr_cpus_allowed > 1) |
1102 | enqueue_pushable_task(rq, p); |
1103 | } |
1104 | |
1105 | #ifdef CONFIG_SMP |
1106 | |
1107 | /* Only try algorithms three times */ |
1108 | #define RT_MAX_TRIES 3 |
1109 | |
1110 | static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep); |
1111 | |
1112 | static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu) |
1113 | { |
1114 | if (!task_running(rq, p) && |
1115 | (cpu < 0 || cpumask_test_cpu(cpu, &p->cpus_allowed)) && |
1116 | (p->rt.nr_cpus_allowed > 1)) |
1117 | return 1; |
1118 | return 0; |
1119 | } |
1120 | |
1121 | /* Return the second highest RT task, NULL otherwise */ |
1122 | static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu) |
1123 | { |
1124 | struct task_struct *next = NULL; |
1125 | struct sched_rt_entity *rt_se; |
1126 | struct rt_prio_array *array; |
1127 | struct rt_rq *rt_rq; |
1128 | int idx; |
1129 | |
1130 | for_each_leaf_rt_rq(rt_rq, rq) { |
1131 | array = &rt_rq->active; |
1132 | idx = sched_find_first_bit(array->bitmap); |
1133 | next_idx: |
1134 | if (idx >= MAX_RT_PRIO) |
1135 | continue; |
1136 | if (next && next->prio < idx) |
1137 | continue; |
1138 | list_for_each_entry(rt_se, array->queue + idx, run_list) { |
1139 | struct task_struct *p = rt_task_of(rt_se); |
1140 | if (pick_rt_task(rq, p, cpu)) { |
1141 | next = p; |
1142 | break; |
1143 | } |
1144 | } |
1145 | if (!next) { |
1146 | idx = find_next_bit(array->bitmap, MAX_RT_PRIO, idx+1); |
1147 | goto next_idx; |
1148 | } |
1149 | } |
1150 | |
1151 | return next; |
1152 | } |
1153 | |
1154 | static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask); |
1155 | |
1156 | static inline int pick_optimal_cpu(int this_cpu, |
1157 | const struct cpumask *mask) |
1158 | { |
1159 | int first; |
1160 | |
1161 | /* "this_cpu" is cheaper to preempt than a remote processor */ |
1162 | if ((this_cpu != -1) && cpumask_test_cpu(this_cpu, mask)) |
1163 | return this_cpu; |
1164 | |
1165 | first = cpumask_first(mask); |
1166 | if (first < nr_cpu_ids) |
1167 | return first; |
1168 | |
1169 | return -1; |
1170 | } |
1171 | |
1172 | static int find_lowest_rq(struct task_struct *task) |
1173 | { |
1174 | struct sched_domain *sd; |
1175 | struct cpumask *lowest_mask = __get_cpu_var(local_cpu_mask); |
1176 | int this_cpu = smp_processor_id(); |
1177 | int cpu = task_cpu(task); |
1178 | cpumask_var_t domain_mask; |
1179 | |
1180 | if (task->rt.nr_cpus_allowed == 1) |
1181 | return -1; /* No other targets possible */ |
1182 | |
1183 | if (!cpupri_find(&task_rq(task)->rd->cpupri, task, lowest_mask)) |
1184 | return -1; /* No targets found */ |
1185 | |
1186 | /* |
1187 | * At this point we have built a mask of cpus representing the |
1188 | * lowest priority tasks in the system. Now we want to elect |
1189 | * the best one based on our affinity and topology. |
1190 | * |
1191 | * We prioritize the last cpu that the task executed on since |
1192 | * it is most likely cache-hot in that location. |
1193 | */ |
1194 | if (cpumask_test_cpu(cpu, lowest_mask)) |
1195 | return cpu; |
1196 | |
1197 | /* |
1198 | * Otherwise, we consult the sched_domains span maps to figure |
1199 | * out which cpu is logically closest to our hot cache data. |
1200 | */ |
1201 | if (this_cpu == cpu) |
1202 | this_cpu = -1; /* Skip this_cpu opt if the same */ |
1203 | |
1204 | if (alloc_cpumask_var(&domain_mask, GFP_ATOMIC)) { |
1205 | for_each_domain(cpu, sd) { |
1206 | if (sd->flags & SD_WAKE_AFFINE) { |
1207 | int best_cpu; |
1208 | |
1209 | cpumask_and(domain_mask, |
1210 | sched_domain_span(sd), |
1211 | lowest_mask); |
1212 | |
1213 | best_cpu = pick_optimal_cpu(this_cpu, |
1214 | domain_mask); |
1215 | |
1216 | if (best_cpu != -1) { |
1217 | free_cpumask_var(domain_mask); |
1218 | return best_cpu; |
1219 | } |
1220 | } |
1221 | } |
1222 | free_cpumask_var(domain_mask); |
1223 | } |
1224 | |
1225 | /* |
1226 | * And finally, if there were no matches within the domains |
1227 | * just give the caller *something* to work with from the compatible |
1228 | * locations. |
1229 | */ |
1230 | return pick_optimal_cpu(this_cpu, lowest_mask); |
1231 | } |
1232 | |
1233 | /* Will lock the rq it finds */ |
1234 | static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq) |
1235 | { |
1236 | struct rq *lowest_rq = NULL; |
1237 | int tries; |
1238 | int cpu; |
1239 | |
1240 | for (tries = 0; tries < RT_MAX_TRIES; tries++) { |
1241 | cpu = find_lowest_rq(task); |
1242 | |
1243 | if ((cpu == -1) || (cpu == rq->cpu)) |
1244 | break; |
1245 | |
1246 | lowest_rq = cpu_rq(cpu); |
1247 | |
1248 | /* if the prio of this runqueue changed, try again */ |
1249 | if (double_lock_balance(rq, lowest_rq)) { |
1250 | /* |
1251 | * We had to unlock the run queue. In |
1252 | * the mean time, task could have |
1253 | * migrated already or had its affinity changed. |
1254 | * Also make sure that it wasn't scheduled on its rq. |
1255 | */ |
1256 | if (unlikely(task_rq(task) != rq || |
1257 | !cpumask_test_cpu(lowest_rq->cpu, |
1258 | &task->cpus_allowed) || |
1259 | task_running(rq, task) || |
1260 | !task->se.on_rq)) { |
1261 | |
1262 | spin_unlock(&lowest_rq->lock); |
1263 | lowest_rq = NULL; |
1264 | break; |
1265 | } |
1266 | } |
1267 | |
1268 | /* If this rq is still suitable use it. */ |
1269 | if (lowest_rq->rt.highest_prio.curr > task->prio) |
1270 | break; |
1271 | |
1272 | /* try again */ |
1273 | double_unlock_balance(rq, lowest_rq); |
1274 | lowest_rq = NULL; |
1275 | } |
1276 | |
1277 | return lowest_rq; |
1278 | } |
1279 | |
1280 | static struct task_struct *pick_next_pushable_task(struct rq *rq) |
1281 | { |
1282 | struct task_struct *p; |
1283 | |
1284 | if (!has_pushable_tasks(rq)) |
1285 | return NULL; |
1286 | |
1287 | p = plist_first_entry(&rq->rt.pushable_tasks, |
1288 | struct task_struct, pushable_tasks); |
1289 | |
1290 | BUG_ON(rq->cpu != task_cpu(p)); |
1291 | BUG_ON(task_current(rq, p)); |
1292 | BUG_ON(p->rt.nr_cpus_allowed <= 1); |
1293 | |
1294 | BUG_ON(!p->se.on_rq); |
1295 | BUG_ON(!rt_task(p)); |
1296 | |
1297 | return p; |
1298 | } |
1299 | |
1300 | /* |
1301 | * If the current CPU has more than one RT task, see if the non |
1302 | * running task can migrate over to a CPU that is running a task |
1303 | * of lesser priority. |
1304 | */ |
1305 | static int push_rt_task(struct rq *rq) |
1306 | { |
1307 | struct task_struct *next_task; |
1308 | struct rq *lowest_rq; |
1309 | |
1310 | if (!rq->rt.overloaded) |
1311 | return 0; |
1312 | |
1313 | next_task = pick_next_pushable_task(rq); |
1314 | if (!next_task) |
1315 | return 0; |
1316 | |
1317 | retry: |
1318 | if (unlikely(next_task == rq->curr)) { |
1319 | WARN_ON(1); |
1320 | return 0; |
1321 | } |
1322 | |
1323 | /* |
1324 | * It's possible that the next_task slipped in of |
1325 | * higher priority than current. If that's the case |
1326 | * just reschedule current. |
1327 | */ |
1328 | if (unlikely(next_task->prio < rq->curr->prio)) { |
1329 | resched_task(rq->curr); |
1330 | return 0; |
1331 | } |
1332 | |
1333 | /* We might release rq lock */ |
1334 | get_task_struct(next_task); |
1335 | |
1336 | /* find_lock_lowest_rq locks the rq if found */ |
1337 | lowest_rq = find_lock_lowest_rq(next_task, rq); |
1338 | if (!lowest_rq) { |
1339 | struct task_struct *task; |
1340 | /* |
1341 | * find lock_lowest_rq releases rq->lock |
1342 | * so it is possible that next_task has migrated. |
1343 | * |
1344 | * We need to make sure that the task is still on the same |
1345 | * run-queue and is also still the next task eligible for |
1346 | * pushing. |
1347 | */ |
1348 | task = pick_next_pushable_task(rq); |
1349 | if (task_cpu(next_task) == rq->cpu && task == next_task) { |
1350 | /* |
1351 | * If we get here, the task hasnt moved at all, but |
1352 | * it has failed to push. We will not try again, |
1353 | * since the other cpus will pull from us when they |
1354 | * are ready. |
1355 | */ |
1356 | dequeue_pushable_task(rq, next_task); |
1357 | goto out; |
1358 | } |
1359 | |
1360 | if (!task) |
1361 | /* No more tasks, just exit */ |
1362 | goto out; |
1363 | |
1364 | /* |
1365 | * Something has shifted, try again. |
1366 | */ |
1367 | put_task_struct(next_task); |
1368 | next_task = task; |
1369 | goto retry; |
1370 | } |
1371 | |
1372 | deactivate_task(rq, next_task, 0); |
1373 | set_task_cpu(next_task, lowest_rq->cpu); |
1374 | activate_task(lowest_rq, next_task, 0); |
1375 | |
1376 | resched_task(lowest_rq->curr); |
1377 | |
1378 | double_unlock_balance(rq, lowest_rq); |
1379 | |
1380 | out: |
1381 | put_task_struct(next_task); |
1382 | |
1383 | return 1; |
1384 | } |
1385 | |
1386 | static void push_rt_tasks(struct rq *rq) |
1387 | { |
1388 | /* push_rt_task will return true if it moved an RT */ |
1389 | while (push_rt_task(rq)) |
1390 | ; |
1391 | } |
1392 | |
1393 | static int pull_rt_task(struct rq *this_rq) |
1394 | { |
1395 | int this_cpu = this_rq->cpu, ret = 0, cpu; |
1396 | struct task_struct *p; |
1397 | struct rq *src_rq; |
1398 | |
1399 | if (likely(!rt_overloaded(this_rq))) |
1400 | return 0; |
1401 | |
1402 | for_each_cpu(cpu, this_rq->rd->rto_mask) { |
1403 | if (this_cpu == cpu) |
1404 | continue; |
1405 | |
1406 | src_rq = cpu_rq(cpu); |
1407 | |
1408 | /* |
1409 | * Don't bother taking the src_rq->lock if the next highest |
1410 | * task is known to be lower-priority than our current task. |
1411 | * This may look racy, but if this value is about to go |
1412 | * logically higher, the src_rq will push this task away. |
1413 | * And if its going logically lower, we do not care |
1414 | */ |
1415 | if (src_rq->rt.highest_prio.next >= |
1416 | this_rq->rt.highest_prio.curr) |
1417 | continue; |
1418 | |
1419 | /* |
1420 | * We can potentially drop this_rq's lock in |
1421 | * double_lock_balance, and another CPU could |
1422 | * alter this_rq |
1423 | */ |
1424 | double_lock_balance(this_rq, src_rq); |
1425 | |
1426 | /* |
1427 | * Are there still pullable RT tasks? |
1428 | */ |
1429 | if (src_rq->rt.rt_nr_running <= 1) |
1430 | goto skip; |
1431 | |
1432 | p = pick_next_highest_task_rt(src_rq, this_cpu); |
1433 | |
1434 | /* |
1435 | * Do we have an RT task that preempts |
1436 | * the to-be-scheduled task? |
1437 | */ |
1438 | if (p && (p->prio < this_rq->rt.highest_prio.curr)) { |
1439 | WARN_ON(p == src_rq->curr); |
1440 | WARN_ON(!p->se.on_rq); |
1441 | |
1442 | /* |
1443 | * There's a chance that p is higher in priority |
1444 | * than what's currently running on its cpu. |
1445 | * This is just that p is wakeing up and hasn't |
1446 | * had a chance to schedule. We only pull |
1447 | * p if it is lower in priority than the |
1448 | * current task on the run queue |
1449 | */ |
1450 | if (p->prio < src_rq->curr->prio) |
1451 | goto skip; |
1452 | |
1453 | ret = 1; |
1454 | |
1455 | deactivate_task(src_rq, p, 0); |
1456 | set_task_cpu(p, this_cpu); |
1457 | activate_task(this_rq, p, 0); |
1458 | /* |
1459 | * We continue with the search, just in |
1460 | * case there's an even higher prio task |
1461 | * in another runqueue. (low likelyhood |
1462 | * but possible) |
1463 | */ |
1464 | } |
1465 | skip: |
1466 | double_unlock_balance(this_rq, src_rq); |
1467 | } |
1468 | |
1469 | return ret; |
1470 | } |
1471 | |
1472 | static void pre_schedule_rt(struct rq *rq, struct task_struct *prev) |
1473 | { |
1474 | /* Try to pull RT tasks here if we lower this rq's prio */ |
1475 | if (unlikely(rt_task(prev)) && rq->rt.highest_prio.curr > prev->prio) |
1476 | pull_rt_task(rq); |
1477 | } |
1478 | |
1479 | static void post_schedule_rt(struct rq *rq) |
1480 | { |
1481 | push_rt_tasks(rq); |
1482 | } |
1483 | |
1484 | /* |
1485 | * If we are not running and we are not going to reschedule soon, we should |
1486 | * try to push tasks away now |
1487 | */ |
1488 | static void task_wake_up_rt(struct rq *rq, struct task_struct *p) |
1489 | { |
1490 | if (!task_running(rq, p) && |
1491 | !test_tsk_need_resched(rq->curr) && |
1492 | has_pushable_tasks(rq) && |
1493 | p->rt.nr_cpus_allowed > 1) |
1494 | push_rt_tasks(rq); |
1495 | } |
1496 | |
1497 | static unsigned long |
1498 | load_balance_rt(struct rq *this_rq, int this_cpu, struct rq *busiest, |
1499 | unsigned long max_load_move, |
1500 | struct sched_domain *sd, enum cpu_idle_type idle, |
1501 | int *all_pinned, int *this_best_prio) |
1502 | { |
1503 | /* don't touch RT tasks */ |
1504 | return 0; |
1505 | } |
1506 | |
1507 | static int |
1508 | move_one_task_rt(struct rq *this_rq, int this_cpu, struct rq *busiest, |
1509 | struct sched_domain *sd, enum cpu_idle_type idle) |
1510 | { |
1511 | /* don't touch RT tasks */ |
1512 | return 0; |
1513 | } |
1514 | |
1515 | static void set_cpus_allowed_rt(struct task_struct *p, |
1516 | const struct cpumask *new_mask) |
1517 | { |
1518 | int weight = cpumask_weight(new_mask); |
1519 | |
1520 | BUG_ON(!rt_task(p)); |
1521 | |
1522 | /* |
1523 | * Update the migration status of the RQ if we have an RT task |
1524 | * which is running AND changing its weight value. |
1525 | */ |
1526 | if (p->se.on_rq && (weight != p->rt.nr_cpus_allowed)) { |
1527 | struct rq *rq = task_rq(p); |
1528 | |
1529 | if (!task_current(rq, p)) { |
1530 | /* |
1531 | * Make sure we dequeue this task from the pushable list |
1532 | * before going further. It will either remain off of |
1533 | * the list because we are no longer pushable, or it |
1534 | * will be requeued. |
1535 | */ |
1536 | if (p->rt.nr_cpus_allowed > 1) |
1537 | dequeue_pushable_task(rq, p); |
1538 | |
1539 | /* |
1540 | * Requeue if our weight is changing and still > 1 |
1541 | */ |
1542 | if (weight > 1) |
1543 | enqueue_pushable_task(rq, p); |
1544 | |
1545 | } |
1546 | |
1547 | if ((p->rt.nr_cpus_allowed <= 1) && (weight > 1)) { |
1548 | rq->rt.rt_nr_migratory++; |
1549 | } else if ((p->rt.nr_cpus_allowed > 1) && (weight <= 1)) { |
1550 | BUG_ON(!rq->rt.rt_nr_migratory); |
1551 | rq->rt.rt_nr_migratory--; |
1552 | } |
1553 | |
1554 | update_rt_migration(&rq->rt); |
1555 | } |
1556 | |
1557 | cpumask_copy(&p->cpus_allowed, new_mask); |
1558 | p->rt.nr_cpus_allowed = weight; |
1559 | } |
1560 | |
1561 | /* Assumes rq->lock is held */ |
1562 | static void rq_online_rt(struct rq *rq) |
1563 | { |
1564 | if (rq->rt.overloaded) |
1565 | rt_set_overload(rq); |
1566 | |
1567 | __enable_runtime(rq); |
1568 | |
1569 | cpupri_set(&rq->rd->cpupri, rq->cpu, rq->rt.highest_prio.curr); |
1570 | } |
1571 | |
1572 | /* Assumes rq->lock is held */ |
1573 | static void rq_offline_rt(struct rq *rq) |
1574 | { |
1575 | if (rq->rt.overloaded) |
1576 | rt_clear_overload(rq); |
1577 | |
1578 | __disable_runtime(rq); |
1579 | |
1580 | cpupri_set(&rq->rd->cpupri, rq->cpu, CPUPRI_INVALID); |
1581 | } |
1582 | |
1583 | /* |
1584 | * When switch from the rt queue, we bring ourselves to a position |
1585 | * that we might want to pull RT tasks from other runqueues. |
1586 | */ |
1587 | static void switched_from_rt(struct rq *rq, struct task_struct *p, |
1588 | int running) |
1589 | { |
1590 | /* |
1591 | * If there are other RT tasks then we will reschedule |
1592 | * and the scheduling of the other RT tasks will handle |
1593 | * the balancing. But if we are the last RT task |
1594 | * we may need to handle the pulling of RT tasks |
1595 | * now. |
1596 | */ |
1597 | if (!rq->rt.rt_nr_running) |
1598 | pull_rt_task(rq); |
1599 | } |
1600 | |
1601 | static inline void init_sched_rt_class(void) |
1602 | { |
1603 | unsigned int i; |
1604 | |
1605 | for_each_possible_cpu(i) |
1606 | zalloc_cpumask_var_node(&per_cpu(local_cpu_mask, i), |
1607 | GFP_KERNEL, cpu_to_node(i)); |
1608 | } |
1609 | #endif /* CONFIG_SMP */ |
1610 | |
1611 | /* |
1612 | * When switching a task to RT, we may overload the runqueue |
1613 | * with RT tasks. In this case we try to push them off to |
1614 | * other runqueues. |
1615 | */ |
1616 | static void switched_to_rt(struct rq *rq, struct task_struct *p, |
1617 | int running) |
1618 | { |
1619 | int check_resched = 1; |
1620 | |
1621 | /* |
1622 | * If we are already running, then there's nothing |
1623 | * that needs to be done. But if we are not running |
1624 | * we may need to preempt the current running task. |
1625 | * If that current running task is also an RT task |
1626 | * then see if we can move to another run queue. |
1627 | */ |
1628 | if (!running) { |
1629 | #ifdef CONFIG_SMP |
1630 | if (rq->rt.overloaded && push_rt_task(rq) && |
1631 | /* Don't resched if we changed runqueues */ |
1632 | rq != task_rq(p)) |
1633 | check_resched = 0; |
1634 | #endif /* CONFIG_SMP */ |
1635 | if (check_resched && p->prio < rq->curr->prio) |
1636 | resched_task(rq->curr); |
1637 | } |
1638 | } |
1639 | |
1640 | /* |
1641 | * Priority of the task has changed. This may cause |
1642 | * us to initiate a push or pull. |
1643 | */ |
1644 | static void prio_changed_rt(struct rq *rq, struct task_struct *p, |
1645 | int oldprio, int running) |
1646 | { |
1647 | if (running) { |
1648 | #ifdef CONFIG_SMP |
1649 | /* |
1650 | * If our priority decreases while running, we |
1651 | * may need to pull tasks to this runqueue. |
1652 | */ |
1653 | if (oldprio < p->prio) |
1654 | pull_rt_task(rq); |
1655 | /* |
1656 | * If there's a higher priority task waiting to run |
1657 | * then reschedule. Note, the above pull_rt_task |
1658 | * can release the rq lock and p could migrate. |
1659 | * Only reschedule if p is still on the same runqueue. |
1660 | */ |
1661 | if (p->prio > rq->rt.highest_prio.curr && rq->curr == p) |
1662 | resched_task(p); |
1663 | #else |
1664 | /* For UP simply resched on drop of prio */ |
1665 | if (oldprio < p->prio) |
1666 | resched_task(p); |
1667 | #endif /* CONFIG_SMP */ |
1668 | } else { |
1669 | /* |
1670 | * This task is not running, but if it is |
1671 | * greater than the current running task |
1672 | * then reschedule. |
1673 | */ |
1674 | if (p->prio < rq->curr->prio) |
1675 | resched_task(rq->curr); |
1676 | } |
1677 | } |
1678 | |
1679 | static void watchdog(struct rq *rq, struct task_struct *p) |
1680 | { |
1681 | unsigned long soft, hard; |
1682 | |
1683 | if (!p->signal) |
1684 | return; |
1685 | |
1686 | soft = p->signal->rlim[RLIMIT_RTTIME].rlim_cur; |
1687 | hard = p->signal->rlim[RLIMIT_RTTIME].rlim_max; |
1688 | |
1689 | if (soft != RLIM_INFINITY) { |
1690 | unsigned long next; |
1691 | |
1692 | p->rt.timeout++; |
1693 | next = DIV_ROUND_UP(min(soft, hard), USEC_PER_SEC/HZ); |
1694 | if (p->rt.timeout > next) |
1695 | p->cputime_expires.sched_exp = p->se.sum_exec_runtime; |
1696 | } |
1697 | } |
1698 | |
1699 | static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued) |
1700 | { |
1701 | update_curr_rt(rq); |
1702 | |
1703 | watchdog(rq, p); |
1704 | |
1705 | /* |
1706 | * RR tasks need a special form of timeslice management. |
1707 | * FIFO tasks have no timeslices. |
1708 | */ |
1709 | if (p->policy != SCHED_RR) |
1710 | return; |
1711 | |
1712 | if (--p->rt.time_slice) |
1713 | return; |
1714 | |
1715 | p->rt.time_slice = DEF_TIMESLICE; |
1716 | |
1717 | /* |
1718 | * Requeue to the end of queue if we are not the only element |
1719 | * on the queue: |
1720 | */ |
1721 | if (p->rt.run_list.prev != p->rt.run_list.next) { |
1722 | requeue_task_rt(rq, p, 0); |
1723 | set_tsk_need_resched(p); |
1724 | } |
1725 | } |
1726 | |
1727 | static void set_curr_task_rt(struct rq *rq) |
1728 | { |
1729 | struct task_struct *p = rq->curr; |
1730 | |
1731 | p->se.exec_start = rq->clock; |
1732 | |
1733 | /* The running task is never eligible for pushing */ |
1734 | dequeue_pushable_task(rq, p); |
1735 | } |
1736 | |
1737 | unsigned int get_rr_interval_rt(struct task_struct *task) |
1738 | { |
1739 | /* |
1740 | * Time slice is 0 for SCHED_FIFO tasks |
1741 | */ |
1742 | if (task->policy == SCHED_RR) |
1743 | return DEF_TIMESLICE; |
1744 | else |
1745 | return 0; |
1746 | } |
1747 | |
1748 | static const struct sched_class rt_sched_class = { |
1749 | .next = &fair_sched_class, |
1750 | .enqueue_task = enqueue_task_rt, |
1751 | .dequeue_task = dequeue_task_rt, |
1752 | .yield_task = yield_task_rt, |
1753 | |
1754 | .check_preempt_curr = check_preempt_curr_rt, |
1755 | |
1756 | .pick_next_task = pick_next_task_rt, |
1757 | .put_prev_task = put_prev_task_rt, |
1758 | |
1759 | #ifdef CONFIG_SMP |
1760 | .select_task_rq = select_task_rq_rt, |
1761 | |
1762 | .load_balance = load_balance_rt, |
1763 | .move_one_task = move_one_task_rt, |
1764 | .set_cpus_allowed = set_cpus_allowed_rt, |
1765 | .rq_online = rq_online_rt, |
1766 | .rq_offline = rq_offline_rt, |
1767 | .pre_schedule = pre_schedule_rt, |
1768 | .post_schedule = post_schedule_rt, |
1769 | .task_wake_up = task_wake_up_rt, |
1770 | .switched_from = switched_from_rt, |
1771 | #endif |
1772 | |
1773 | .set_curr_task = set_curr_task_rt, |
1774 | .task_tick = task_tick_rt, |
1775 | |
1776 | .get_rr_interval = get_rr_interval_rt, |
1777 | |
1778 | .prio_changed = prio_changed_rt, |
1779 | .switched_to = switched_to_rt, |
1780 | }; |
1781 | |
1782 | #ifdef CONFIG_SCHED_DEBUG |
1783 | extern void print_rt_rq(struct seq_file *m, int cpu, struct rt_rq *rt_rq); |
1784 | |
1785 | static void print_rt_stats(struct seq_file *m, int cpu) |
1786 | { |
1787 | struct rt_rq *rt_rq; |
1788 | |
1789 | rcu_read_lock(); |
1790 | for_each_leaf_rt_rq(rt_rq, cpu_rq(cpu)) |
1791 | print_rt_rq(m, cpu, rt_rq); |
1792 | rcu_read_unlock(); |
1793 | } |
1794 | #endif /* CONFIG_SCHED_DEBUG */ |
1795 | |
1796 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9