sched: Strengthen buddies and mitigate buddy induced latencies
[linux-2.6.git] / kernel / sched_fair.c
1 /*
2  * Completely Fair Scheduling (CFS) Class (SCHED_NORMAL/SCHED_BATCH)
3  *
4  *  Copyright (C) 2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
5  *
6  *  Interactivity improvements by Mike Galbraith
7  *  (C) 2007 Mike Galbraith <efault@gmx.de>
8  *
9  *  Various enhancements by Dmitry Adamushko.
10  *  (C) 2007 Dmitry Adamushko <dmitry.adamushko@gmail.com>
11  *
12  *  Group scheduling enhancements by Srivatsa Vaddagiri
13  *  Copyright IBM Corporation, 2007
14  *  Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
15  *
16  *  Scaled math optimizations by Thomas Gleixner
17  *  Copyright (C) 2007, Thomas Gleixner <tglx@linutronix.de>
18  *
19  *  Adaptive scheduling granularity, math enhancements by Peter Zijlstra
20  *  Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
21  */
22
23 #include <linux/latencytop.h>
24
25 /*
26  * Targeted preemption latency for CPU-bound tasks:
27  * (default: 5ms * (1 + ilog(ncpus)), units: nanoseconds)
28  *
29  * NOTE: this latency value is not the same as the concept of
30  * 'timeslice length' - timeslices in CFS are of variable length
31  * and have no persistent notion like in traditional, time-slice
32  * based scheduling concepts.
33  *
34  * (to see the precise effective timeslice length of your workload,
35  *  run vmstat and monitor the context-switches (cs) field)
36  */
37 unsigned int sysctl_sched_latency = 5000000ULL;
38
39 /*
40  * Minimal preemption granularity for CPU-bound tasks:
41  * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds)
42  */
43 unsigned int sysctl_sched_min_granularity = 1000000ULL;
44
45 /*
46  * is kept at sysctl_sched_latency / sysctl_sched_min_granularity
47  */
48 static unsigned int sched_nr_latency = 5;
49
50 /*
51  * After fork, child runs first. If set to 0 (default) then
52  * parent will (try to) run first.
53  */
54 unsigned int sysctl_sched_child_runs_first __read_mostly;
55
56 /*
57  * sys_sched_yield() compat mode
58  *
59  * This option switches the agressive yield implementation of the
60  * old scheduler back on.
61  */
62 unsigned int __read_mostly sysctl_sched_compat_yield;
63
64 /*
65  * SCHED_OTHER wake-up granularity.
66  * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds)
67  *
68  * This option delays the preemption effects of decoupled workloads
69  * and reduces their over-scheduling. Synchronous workloads will still
70  * have immediate wakeup/sleep latencies.
71  */
72 unsigned int sysctl_sched_wakeup_granularity = 1000000UL;
73
74 const_debug unsigned int sysctl_sched_migration_cost = 500000UL;
75
76 static const struct sched_class fair_sched_class;
77
78 /**************************************************************
79  * CFS operations on generic schedulable entities:
80  */
81
82 #ifdef CONFIG_FAIR_GROUP_SCHED
83
84 /* cpu runqueue to which this cfs_rq is attached */
85 static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
86 {
87         return cfs_rq->rq;
88 }
89
90 /* An entity is a task if it doesn't "own" a runqueue */
91 #define entity_is_task(se)      (!se->my_q)
92
93 static inline struct task_struct *task_of(struct sched_entity *se)
94 {
95 #ifdef CONFIG_SCHED_DEBUG
96         WARN_ON_ONCE(!entity_is_task(se));
97 #endif
98         return container_of(se, struct task_struct, se);
99 }
100
101 /* Walk up scheduling entities hierarchy */
102 #define for_each_sched_entity(se) \
103                 for (; se; se = se->parent)
104
105 static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
106 {
107         return p->se.cfs_rq;
108 }
109
110 /* runqueue on which this entity is (to be) queued */
111 static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
112 {
113         return se->cfs_rq;
114 }
115
116 /* runqueue "owned" by this group */
117 static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
118 {
119         return grp->my_q;
120 }
121
122 /* Given a group's cfs_rq on one cpu, return its corresponding cfs_rq on
123  * another cpu ('this_cpu')
124  */
125 static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
126 {
127         return cfs_rq->tg->cfs_rq[this_cpu];
128 }
129
130 /* Iterate thr' all leaf cfs_rq's on a runqueue */
131 #define for_each_leaf_cfs_rq(rq, cfs_rq) \
132         list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)
133
134 /* Do the two (enqueued) entities belong to the same group ? */
135 static inline int
136 is_same_group(struct sched_entity *se, struct sched_entity *pse)
137 {
138         if (se->cfs_rq == pse->cfs_rq)
139                 return 1;
140
141         return 0;
142 }
143
144 static inline struct sched_entity *parent_entity(struct sched_entity *se)
145 {
146         return se->parent;
147 }
148
149 /* return depth at which a sched entity is present in the hierarchy */
150 static inline int depth_se(struct sched_entity *se)
151 {
152         int depth = 0;
153
154         for_each_sched_entity(se)
155                 depth++;
156
157         return depth;
158 }
159
160 static void
161 find_matching_se(struct sched_entity **se, struct sched_entity **pse)
162 {
163         int se_depth, pse_depth;
164
165         /*
166          * preemption test can be made between sibling entities who are in the
167          * same cfs_rq i.e who have a common parent. Walk up the hierarchy of
168          * both tasks until we find their ancestors who are siblings of common
169          * parent.
170          */
171
172         /* First walk up until both entities are at same depth */
173         se_depth = depth_se(*se);
174         pse_depth = depth_se(*pse);
175
176         while (se_depth > pse_depth) {
177                 se_depth--;
178                 *se = parent_entity(*se);
179         }
180
181         while (pse_depth > se_depth) {
182                 pse_depth--;
183                 *pse = parent_entity(*pse);
184         }
185
186         while (!is_same_group(*se, *pse)) {
187                 *se = parent_entity(*se);
188                 *pse = parent_entity(*pse);
189         }
190 }
191
192 #else   /* !CONFIG_FAIR_GROUP_SCHED */
193
194 static inline struct task_struct *task_of(struct sched_entity *se)
195 {
196         return container_of(se, struct task_struct, se);
197 }
198
199 static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
200 {
201         return container_of(cfs_rq, struct rq, cfs);
202 }
203
204 #define entity_is_task(se)      1
205
206 #define for_each_sched_entity(se) \
207                 for (; se; se = NULL)
208
209 static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
210 {
211         return &task_rq(p)->cfs;
212 }
213
214 static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
215 {
216         struct task_struct *p = task_of(se);
217         struct rq *rq = task_rq(p);
218
219         return &rq->cfs;
220 }
221
222 /* runqueue "owned" by this group */
223 static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
224 {
225         return NULL;
226 }
227
228 static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
229 {
230         return &cpu_rq(this_cpu)->cfs;
231 }
232
233 #define for_each_leaf_cfs_rq(rq, cfs_rq) \
234                 for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)
235
236 static inline int
237 is_same_group(struct sched_entity *se, struct sched_entity *pse)
238 {
239         return 1;
240 }
241
242 static inline struct sched_entity *parent_entity(struct sched_entity *se)
243 {
244         return NULL;
245 }
246
247 static inline void
248 find_matching_se(struct sched_entity **se, struct sched_entity **pse)
249 {
250 }
251
252 #endif  /* CONFIG_FAIR_GROUP_SCHED */
253
254
255 /**************************************************************
256  * Scheduling class tree data structure manipulation methods:
257  */
258
259 static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
260 {
261         s64 delta = (s64)(vruntime - min_vruntime);
262         if (delta > 0)
263                 min_vruntime = vruntime;
264
265         return min_vruntime;
266 }
267
268 static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
269 {
270         s64 delta = (s64)(vruntime - min_vruntime);
271         if (delta < 0)
272                 min_vruntime = vruntime;
273
274         return min_vruntime;
275 }
276
277 static inline int entity_before(struct sched_entity *a,
278                                 struct sched_entity *b)
279 {
280         return (s64)(a->vruntime - b->vruntime) < 0;
281 }
282
283 static inline s64 entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
284 {
285         return se->vruntime - cfs_rq->min_vruntime;
286 }
287
288 static void update_min_vruntime(struct cfs_rq *cfs_rq)
289 {
290         u64 vruntime = cfs_rq->min_vruntime;
291
292         if (cfs_rq->curr)
293                 vruntime = cfs_rq->curr->vruntime;
294
295         if (cfs_rq->rb_leftmost) {
296                 struct sched_entity *se = rb_entry(cfs_rq->rb_leftmost,
297                                                    struct sched_entity,
298                                                    run_node);
299
300                 if (!cfs_rq->curr)
301                         vruntime = se->vruntime;
302                 else
303                         vruntime = min_vruntime(vruntime, se->vruntime);
304         }
305
306         cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime);
307 }
308
309 /*
310  * Enqueue an entity into the rb-tree:
311  */
312 static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
313 {
314         struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
315         struct rb_node *parent = NULL;
316         struct sched_entity *entry;
317         s64 key = entity_key(cfs_rq, se);
318         int leftmost = 1;
319
320         /*
321          * Find the right place in the rbtree:
322          */
323         while (*link) {
324                 parent = *link;
325                 entry = rb_entry(parent, struct sched_entity, run_node);
326                 /*
327                  * We dont care about collisions. Nodes with
328                  * the same key stay together.
329                  */
330                 if (key < entity_key(cfs_rq, entry)) {
331                         link = &parent->rb_left;
332                 } else {
333                         link = &parent->rb_right;
334                         leftmost = 0;
335                 }
336         }
337
338         /*
339          * Maintain a cache of leftmost tree entries (it is frequently
340          * used):
341          */
342         if (leftmost)
343                 cfs_rq->rb_leftmost = &se->run_node;
344
345         rb_link_node(&se->run_node, parent, link);
346         rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline);
347 }
348
349 static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
350 {
351         if (cfs_rq->rb_leftmost == &se->run_node) {
352                 struct rb_node *next_node;
353
354                 next_node = rb_next(&se->run_node);
355                 cfs_rq->rb_leftmost = next_node;
356         }
357
358         rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
359 }
360
361 static struct sched_entity *__pick_next_entity(struct cfs_rq *cfs_rq)
362 {
363         struct rb_node *left = cfs_rq->rb_leftmost;
364
365         if (!left)
366                 return NULL;
367
368         return rb_entry(left, struct sched_entity, run_node);
369 }
370
371 static struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
372 {
373         struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
374
375         if (!last)
376                 return NULL;
377
378         return rb_entry(last, struct sched_entity, run_node);
379 }
380
381 /**************************************************************
382  * Scheduling class statistics methods:
383  */
384
385 #ifdef CONFIG_SCHED_DEBUG
386 int sched_nr_latency_handler(struct ctl_table *table, int write,
387                 void __user *buffer, size_t *lenp,
388                 loff_t *ppos)
389 {
390         int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
391
392         if (ret || !write)
393                 return ret;
394
395         sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
396                                         sysctl_sched_min_granularity);
397
398         return 0;
399 }
400 #endif
401
402 /*
403  * delta /= w
404  */
405 static inline unsigned long
406 calc_delta_fair(unsigned long delta, struct sched_entity *se)
407 {
408         if (unlikely(se->load.weight != NICE_0_LOAD))
409                 delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load);
410
411         return delta;
412 }
413
414 /*
415  * The idea is to set a period in which each task runs once.
416  *
417  * When there are too many tasks (sysctl_sched_nr_latency) we have to stretch
418  * this period because otherwise the slices get too small.
419  *
420  * p = (nr <= nl) ? l : l*nr/nl
421  */
422 static u64 __sched_period(unsigned long nr_running)
423 {
424         u64 period = sysctl_sched_latency;
425         unsigned long nr_latency = sched_nr_latency;
426
427         if (unlikely(nr_running > nr_latency)) {
428                 period = sysctl_sched_min_granularity;
429                 period *= nr_running;
430         }
431
432         return period;
433 }
434
435 /*
436  * We calculate the wall-time slice from the period by taking a part
437  * proportional to the weight.
438  *
439  * s = p*P[w/rw]
440  */
441 static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
442 {
443         u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq);
444
445         for_each_sched_entity(se) {
446                 struct load_weight *load;
447                 struct load_weight lw;
448
449                 cfs_rq = cfs_rq_of(se);
450                 load = &cfs_rq->load;
451
452                 if (unlikely(!se->on_rq)) {
453                         lw = cfs_rq->load;
454
455                         update_load_add(&lw, se->load.weight);
456                         load = &lw;
457                 }
458                 slice = calc_delta_mine(slice, se->load.weight, load);
459         }
460         return slice;
461 }
462
463 /*
464  * We calculate the vruntime slice of a to be inserted task
465  *
466  * vs = s/w
467  */
468 static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
469 {
470         return calc_delta_fair(sched_slice(cfs_rq, se), se);
471 }
472
473 /*
474  * Update the current task's runtime statistics. Skip current tasks that
475  * are not in our scheduling class.
476  */
477 static inline void
478 __update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
479               unsigned long delta_exec)
480 {
481         unsigned long delta_exec_weighted;
482
483         schedstat_set(curr->exec_max, max((u64)delta_exec, curr->exec_max));
484
485         curr->sum_exec_runtime += delta_exec;
486         schedstat_add(cfs_rq, exec_clock, delta_exec);
487         delta_exec_weighted = calc_delta_fair(delta_exec, curr);
488         curr->vruntime += delta_exec_weighted;
489         update_min_vruntime(cfs_rq);
490 }
491
492 static void update_curr(struct cfs_rq *cfs_rq)
493 {
494         struct sched_entity *curr = cfs_rq->curr;
495         u64 now = rq_of(cfs_rq)->clock;
496         unsigned long delta_exec;
497
498         if (unlikely(!curr))
499                 return;
500
501         /*
502          * Get the amount of time the current task was running
503          * since the last time we changed load (this cannot
504          * overflow on 32 bits):
505          */
506         delta_exec = (unsigned long)(now - curr->exec_start);
507         if (!delta_exec)
508                 return;
509
510         __update_curr(cfs_rq, curr, delta_exec);
511         curr->exec_start = now;
512
513         if (entity_is_task(curr)) {
514                 struct task_struct *curtask = task_of(curr);
515
516                 trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
517                 cpuacct_charge(curtask, delta_exec);
518                 account_group_exec_runtime(curtask, delta_exec);
519         }
520 }
521
522 static inline void
523 update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
524 {
525         schedstat_set(se->wait_start, rq_of(cfs_rq)->clock);
526 }
527
528 /*
529  * Task is being enqueued - update stats:
530  */
531 static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
532 {
533         /*
534          * Are we enqueueing a waiting task? (for current tasks
535          * a dequeue/enqueue event is a NOP)
536          */
537         if (se != cfs_rq->curr)
538                 update_stats_wait_start(cfs_rq, se);
539 }
540
541 static void
542 update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
543 {
544         schedstat_set(se->wait_max, max(se->wait_max,
545                         rq_of(cfs_rq)->clock - se->wait_start));
546         schedstat_set(se->wait_count, se->wait_count + 1);
547         schedstat_set(se->wait_sum, se->wait_sum +
548                         rq_of(cfs_rq)->clock - se->wait_start);
549 #ifdef CONFIG_SCHEDSTATS
550         if (entity_is_task(se)) {
551                 trace_sched_stat_wait(task_of(se),
552                         rq_of(cfs_rq)->clock - se->wait_start);
553         }
554 #endif
555         schedstat_set(se->wait_start, 0);
556 }
557
558 static inline void
559 update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
560 {
561         /*
562          * Mark the end of the wait period if dequeueing a
563          * waiting task:
564          */
565         if (se != cfs_rq->curr)
566                 update_stats_wait_end(cfs_rq, se);
567 }
568
569 /*
570  * We are picking a new current task - update its stats:
571  */
572 static inline void
573 update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
574 {
575         /*
576          * We are starting a new run period:
577          */
578         se->exec_start = rq_of(cfs_rq)->clock;
579 }
580
581 /**************************************************
582  * Scheduling class queueing methods:
583  */
584
585 #if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
586 static void
587 add_cfs_task_weight(struct cfs_rq *cfs_rq, unsigned long weight)
588 {
589         cfs_rq->task_weight += weight;
590 }
591 #else
592 static inline void
593 add_cfs_task_weight(struct cfs_rq *cfs_rq, unsigned long weight)
594 {
595 }
596 #endif
597
598 static void
599 account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
600 {
601         update_load_add(&cfs_rq->load, se->load.weight);
602         if (!parent_entity(se))
603                 inc_cpu_load(rq_of(cfs_rq), se->load.weight);
604         if (entity_is_task(se)) {
605                 add_cfs_task_weight(cfs_rq, se->load.weight);
606                 list_add(&se->group_node, &cfs_rq->tasks);
607         }
608         cfs_rq->nr_running++;
609         se->on_rq = 1;
610 }
611
612 static void
613 account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
614 {
615         update_load_sub(&cfs_rq->load, se->load.weight);
616         if (!parent_entity(se))
617                 dec_cpu_load(rq_of(cfs_rq), se->load.weight);
618         if (entity_is_task(se)) {
619                 add_cfs_task_weight(cfs_rq, -se->load.weight);
620                 list_del_init(&se->group_node);
621         }
622         cfs_rq->nr_running--;
623         se->on_rq = 0;
624 }
625
626 static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
627 {
628 #ifdef CONFIG_SCHEDSTATS
629         struct task_struct *tsk = NULL;
630
631         if (entity_is_task(se))
632                 tsk = task_of(se);
633
634         if (se->sleep_start) {
635                 u64 delta = rq_of(cfs_rq)->clock - se->sleep_start;
636
637                 if ((s64)delta < 0)
638                         delta = 0;
639
640                 if (unlikely(delta > se->sleep_max))
641                         se->sleep_max = delta;
642
643                 se->sleep_start = 0;
644                 se->sum_sleep_runtime += delta;
645
646                 if (tsk) {
647                         account_scheduler_latency(tsk, delta >> 10, 1);
648                         trace_sched_stat_sleep(tsk, delta);
649                 }
650         }
651         if (se->block_start) {
652                 u64 delta = rq_of(cfs_rq)->clock - se->block_start;
653
654                 if ((s64)delta < 0)
655                         delta = 0;
656
657                 if (unlikely(delta > se->block_max))
658                         se->block_max = delta;
659
660                 se->block_start = 0;
661                 se->sum_sleep_runtime += delta;
662
663                 if (tsk) {
664                         if (tsk->in_iowait) {
665                                 se->iowait_sum += delta;
666                                 se->iowait_count++;
667                                 trace_sched_stat_iowait(tsk, delta);
668                         }
669
670                         /*
671                          * Blocking time is in units of nanosecs, so shift by
672                          * 20 to get a milliseconds-range estimation of the
673                          * amount of time that the task spent sleeping:
674                          */
675                         if (unlikely(prof_on == SLEEP_PROFILING)) {
676                                 profile_hits(SLEEP_PROFILING,
677                                                 (void *)get_wchan(tsk),
678                                                 delta >> 20);
679                         }
680                         account_scheduler_latency(tsk, delta >> 10, 0);
681                 }
682         }
683 #endif
684 }
685
686 static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se)
687 {
688 #ifdef CONFIG_SCHED_DEBUG
689         s64 d = se->vruntime - cfs_rq->min_vruntime;
690
691         if (d < 0)
692                 d = -d;
693
694         if (d > 3*sysctl_sched_latency)
695                 schedstat_inc(cfs_rq, nr_spread_over);
696 #endif
697 }
698
699 static void
700 place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
701 {
702         u64 vruntime = cfs_rq->min_vruntime;
703
704         /*
705          * The 'current' period is already promised to the current tasks,
706          * however the extra weight of the new task will slow them down a
707          * little, place the new task so that it fits in the slot that
708          * stays open at the end.
709          */
710         if (initial && sched_feat(START_DEBIT))
711                 vruntime += sched_vslice(cfs_rq, se);
712
713         /* sleeps up to a single latency don't count. */
714         if (!initial && sched_feat(FAIR_SLEEPERS)) {
715                 unsigned long thresh = sysctl_sched_latency;
716
717                 /*
718                  * Convert the sleeper threshold into virtual time.
719                  * SCHED_IDLE is a special sub-class.  We care about
720                  * fairness only relative to other SCHED_IDLE tasks,
721                  * all of which have the same weight.
722                  */
723                 if (sched_feat(NORMALIZED_SLEEPER) && (!entity_is_task(se) ||
724                                  task_of(se)->policy != SCHED_IDLE))
725                         thresh = calc_delta_fair(thresh, se);
726
727                 /*
728                  * Halve their sleep time's effect, to allow
729                  * for a gentler effect of sleepers:
730                  */
731                 if (sched_feat(GENTLE_FAIR_SLEEPERS))
732                         thresh >>= 1;
733
734                 vruntime -= thresh;
735         }
736
737         /* ensure we never gain time by being placed backwards. */
738         vruntime = max_vruntime(se->vruntime, vruntime);
739
740         se->vruntime = vruntime;
741 }
742
743 static void
744 enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
745 {
746         /*
747          * Update run-time statistics of the 'current'.
748          */
749         update_curr(cfs_rq);
750         account_entity_enqueue(cfs_rq, se);
751
752         if (wakeup) {
753                 place_entity(cfs_rq, se, 0);
754                 enqueue_sleeper(cfs_rq, se);
755         }
756
757         update_stats_enqueue(cfs_rq, se);
758         check_spread(cfs_rq, se);
759         if (se != cfs_rq->curr)
760                 __enqueue_entity(cfs_rq, se);
761 }
762
763 static void __clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
764 {
765         if (!se || cfs_rq->last == se)
766                 cfs_rq->last = NULL;
767
768         if (!se || cfs_rq->next == se)
769                 cfs_rq->next = NULL;
770 }
771
772 static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
773 {
774         for_each_sched_entity(se)
775                 __clear_buddies(cfs_rq_of(se), se);
776 }
777
778 static void
779 dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
780 {
781         /*
782          * Update run-time statistics of the 'current'.
783          */
784         update_curr(cfs_rq);
785
786         update_stats_dequeue(cfs_rq, se);
787         if (sleep) {
788 #ifdef CONFIG_SCHEDSTATS
789                 if (entity_is_task(se)) {
790                         struct task_struct *tsk = task_of(se);
791
792                         if (tsk->state & TASK_INTERRUPTIBLE)
793                                 se->sleep_start = rq_of(cfs_rq)->clock;
794                         if (tsk->state & TASK_UNINTERRUPTIBLE)
795                                 se->block_start = rq_of(cfs_rq)->clock;
796                 }
797 #endif
798         }
799
800         clear_buddies(cfs_rq, se);
801
802         if (se != cfs_rq->curr)
803                 __dequeue_entity(cfs_rq, se);
804         account_entity_dequeue(cfs_rq, se);
805         update_min_vruntime(cfs_rq);
806 }
807
808 /*
809  * Preempt the current task with a newly woken task if needed:
810  */
811 static void
812 check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
813 {
814         unsigned long ideal_runtime, delta_exec;
815
816         ideal_runtime = sched_slice(cfs_rq, curr);
817         delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
818         if (delta_exec > ideal_runtime) {
819                 resched_task(rq_of(cfs_rq)->curr);
820                 /*
821                  * The current task ran long enough, ensure it doesn't get
822                  * re-elected due to buddy favours.
823                  */
824                 clear_buddies(cfs_rq, curr);
825                 return;
826         }
827
828         /*
829          * Ensure that a task that missed wakeup preemption by a
830          * narrow margin doesn't have to wait for a full slice.
831          * This also mitigates buddy induced latencies under load.
832          */
833         if (!sched_feat(WAKEUP_PREEMPT))
834                 return;
835
836         if (delta_exec < sysctl_sched_min_granularity)
837                 return;
838
839         if (cfs_rq->nr_running > 1) {
840                 struct sched_entity *se = __pick_next_entity(cfs_rq);
841                 s64 delta = curr->vruntime - se->vruntime;
842
843                 if (delta > ideal_runtime)
844                         resched_task(rq_of(cfs_rq)->curr);
845         }
846 }
847
848 static void
849 set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
850 {
851         /* 'current' is not kept within the tree. */
852         if (se->on_rq) {
853                 /*
854                  * Any task has to be enqueued before it get to execute on
855                  * a CPU. So account for the time it spent waiting on the
856                  * runqueue.
857                  */
858                 update_stats_wait_end(cfs_rq, se);
859                 __dequeue_entity(cfs_rq, se);
860         }
861
862         update_stats_curr_start(cfs_rq, se);
863         cfs_rq->curr = se;
864 #ifdef CONFIG_SCHEDSTATS
865         /*
866          * Track our maximum slice length, if the CPU's load is at
867          * least twice that of our own weight (i.e. dont track it
868          * when there are only lesser-weight tasks around):
869          */
870         if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
871                 se->slice_max = max(se->slice_max,
872                         se->sum_exec_runtime - se->prev_sum_exec_runtime);
873         }
874 #endif
875         se->prev_sum_exec_runtime = se->sum_exec_runtime;
876 }
877
878 static int
879 wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);
880
881 static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
882 {
883         struct sched_entity *se = __pick_next_entity(cfs_rq);
884         struct sched_entity *left = se;
885
886         if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1)
887                 se = cfs_rq->next;
888
889         /*
890          * Prefer last buddy, try to return the CPU to a preempted task.
891          */
892         if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1)
893                 se = cfs_rq->last;
894
895         clear_buddies(cfs_rq, se);
896
897         return se;
898 }
899
900 static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
901 {
902         /*
903          * If still on the runqueue then deactivate_task()
904          * was not called and update_curr() has to be done:
905          */
906         if (prev->on_rq)
907                 update_curr(cfs_rq);
908
909         check_spread(cfs_rq, prev);
910         if (prev->on_rq) {
911                 update_stats_wait_start(cfs_rq, prev);
912                 /* Put 'current' back into the tree. */
913                 __enqueue_entity(cfs_rq, prev);
914         }
915         cfs_rq->curr = NULL;
916 }
917
918 static void
919 entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
920 {
921         /*
922          * Update run-time statistics of the 'current'.
923          */
924         update_curr(cfs_rq);
925
926 #ifdef CONFIG_SCHED_HRTICK
927         /*
928          * queued ticks are scheduled to match the slice, so don't bother
929          * validating it and just reschedule.
930          */
931         if (queued) {
932                 resched_task(rq_of(cfs_rq)->curr);
933                 return;
934         }
935         /*
936          * don't let the period tick interfere with the hrtick preemption
937          */
938         if (!sched_feat(DOUBLE_TICK) &&
939                         hrtimer_active(&rq_of(cfs_rq)->hrtick_timer))
940                 return;
941 #endif
942
943         if (cfs_rq->nr_running > 1 || !sched_feat(WAKEUP_PREEMPT))
944                 check_preempt_tick(cfs_rq, curr);
945 }
946
947 /**************************************************
948  * CFS operations on tasks:
949  */
950
951 #ifdef CONFIG_SCHED_HRTICK
952 static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
953 {
954         struct sched_entity *se = &p->se;
955         struct cfs_rq *cfs_rq = cfs_rq_of(se);
956
957         WARN_ON(task_rq(p) != rq);
958
959         if (hrtick_enabled(rq) && cfs_rq->nr_running > 1) {
960                 u64 slice = sched_slice(cfs_rq, se);
961                 u64 ran = se->sum_exec_runtime - se->prev_sum_exec_runtime;
962                 s64 delta = slice - ran;
963
964                 if (delta < 0) {
965                         if (rq->curr == p)
966                                 resched_task(p);
967                         return;
968                 }
969
970                 /*
971                  * Don't schedule slices shorter than 10000ns, that just
972                  * doesn't make sense. Rely on vruntime for fairness.
973                  */
974                 if (rq->curr != p)
975                         delta = max_t(s64, 10000LL, delta);
976
977                 hrtick_start(rq, delta);
978         }
979 }
980
981 /*
982  * called from enqueue/dequeue and updates the hrtick when the
983  * current task is from our class and nr_running is low enough
984  * to matter.
985  */
986 static void hrtick_update(struct rq *rq)
987 {
988         struct task_struct *curr = rq->curr;
989
990         if (curr->sched_class != &fair_sched_class)
991                 return;
992
993         if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency)
994                 hrtick_start_fair(rq, curr);
995 }
996 #else /* !CONFIG_SCHED_HRTICK */
997 static inline void
998 hrtick_start_fair(struct rq *rq, struct task_struct *p)
999 {
1000 }
1001
1002 static inline void hrtick_update(struct rq *rq)
1003 {
1004 }
1005 #endif
1006
1007 /*
1008  * The enqueue_task method is called before nr_running is
1009  * increased. Here we update the fair scheduling stats and
1010  * then put the task into the rbtree:
1011  */
1012 static void enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup)
1013 {
1014         struct cfs_rq *cfs_rq;
1015         struct sched_entity *se = &p->se;
1016
1017         for_each_sched_entity(se) {
1018                 if (se->on_rq)
1019                         break;
1020                 cfs_rq = cfs_rq_of(se);
1021                 enqueue_entity(cfs_rq, se, wakeup);
1022                 wakeup = 1;
1023         }
1024
1025         hrtick_update(rq);
1026 }
1027
1028 /*
1029  * The dequeue_task method is called before nr_running is
1030  * decreased. We remove the task from the rbtree and
1031  * update the fair scheduling stats:
1032  */
1033 static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int sleep)
1034 {
1035         struct cfs_rq *cfs_rq;
1036         struct sched_entity *se = &p->se;
1037
1038         for_each_sched_entity(se) {
1039                 cfs_rq = cfs_rq_of(se);
1040                 dequeue_entity(cfs_rq, se, sleep);
1041                 /* Don't dequeue parent if it has other entities besides us */
1042                 if (cfs_rq->load.weight)
1043                         break;
1044                 sleep = 1;
1045         }
1046
1047         hrtick_update(rq);
1048 }
1049
1050 /*
1051  * sched_yield() support is very simple - we dequeue and enqueue.
1052  *
1053  * If compat_yield is turned on then we requeue to the end of the tree.
1054  */
1055 static void yield_task_fair(struct rq *rq)
1056 {
1057         struct task_struct *curr = rq->curr;
1058         struct cfs_rq *cfs_rq = task_cfs_rq(curr);
1059         struct sched_entity *rightmost, *se = &curr->se;
1060
1061         /*
1062          * Are we the only task in the tree?
1063          */
1064         if (unlikely(cfs_rq->nr_running == 1))
1065                 return;
1066
1067         clear_buddies(cfs_rq, se);
1068
1069         if (likely(!sysctl_sched_compat_yield) && curr->policy != SCHED_BATCH) {
1070                 update_rq_clock(rq);
1071                 /*
1072                  * Update run-time statistics of the 'current'.
1073                  */
1074                 update_curr(cfs_rq);
1075
1076                 return;
1077         }
1078         /*
1079          * Find the rightmost entry in the rbtree:
1080          */
1081         rightmost = __pick_last_entity(cfs_rq);
1082         /*
1083          * Already in the rightmost position?
1084          */
1085         if (unlikely(!rightmost || entity_before(rightmost, se)))
1086                 return;
1087
1088         /*
1089          * Minimally necessary key value to be last in the tree:
1090          * Upon rescheduling, sched_class::put_prev_task() will place
1091          * 'current' within the tree based on its new key value.
1092          */
1093         se->vruntime = rightmost->vruntime + 1;
1094 }
1095
1096 #ifdef CONFIG_SMP
1097
1098 #ifdef CONFIG_FAIR_GROUP_SCHED
1099 /*
1100  * effective_load() calculates the load change as seen from the root_task_group
1101  *
1102  * Adding load to a group doesn't make a group heavier, but can cause movement
1103  * of group shares between cpus. Assuming the shares were perfectly aligned one
1104  * can calculate the shift in shares.
1105  *
1106  * The problem is that perfectly aligning the shares is rather expensive, hence
1107  * we try to avoid doing that too often - see update_shares(), which ratelimits
1108  * this change.
1109  *
1110  * We compensate this by not only taking the current delta into account, but
1111  * also considering the delta between when the shares were last adjusted and
1112  * now.
1113  *
1114  * We still saw a performance dip, some tracing learned us that between
1115  * cgroup:/ and cgroup:/foo balancing the number of affine wakeups increased
1116  * significantly. Therefore try to bias the error in direction of failing
1117  * the affine wakeup.
1118  *
1119  */
1120 static long effective_load(struct task_group *tg, int cpu,
1121                 long wl, long wg)
1122 {
1123         struct sched_entity *se = tg->se[cpu];
1124
1125         if (!tg->parent)
1126                 return wl;
1127
1128         /*
1129          * By not taking the decrease of shares on the other cpu into
1130          * account our error leans towards reducing the affine wakeups.
1131          */
1132         if (!wl && sched_feat(ASYM_EFF_LOAD))
1133                 return wl;
1134
1135         for_each_sched_entity(se) {
1136                 long S, rw, s, a, b;
1137                 long more_w;
1138
1139                 /*
1140                  * Instead of using this increment, also add the difference
1141                  * between when the shares were last updated and now.
1142                  */
1143                 more_w = se->my_q->load.weight - se->my_q->rq_weight;
1144                 wl += more_w;
1145                 wg += more_w;
1146
1147                 S = se->my_q->tg->shares;
1148                 s = se->my_q->shares;
1149                 rw = se->my_q->rq_weight;
1150
1151                 a = S*(rw + wl);
1152                 b = S*rw + s*wg;
1153
1154                 wl = s*(a-b);
1155
1156                 if (likely(b))
1157                         wl /= b;
1158
1159                 /*
1160                  * Assume the group is already running and will
1161                  * thus already be accounted for in the weight.
1162                  *
1163                  * That is, moving shares between CPUs, does not
1164                  * alter the group weight.
1165                  */
1166                 wg = 0;
1167         }
1168
1169         return wl;
1170 }
1171
1172 #else
1173
1174 static inline unsigned long effective_load(struct task_group *tg, int cpu,
1175                 unsigned long wl, unsigned long wg)
1176 {
1177         return wl;
1178 }
1179
1180 #endif
1181
1182 static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
1183 {
1184         struct task_struct *curr = current;
1185         unsigned long this_load, load;
1186         int idx, this_cpu, prev_cpu;
1187         unsigned long tl_per_task;
1188         unsigned int imbalance;
1189         struct task_group *tg;
1190         unsigned long weight;
1191         int balanced;
1192
1193         idx       = sd->wake_idx;
1194         this_cpu  = smp_processor_id();
1195         prev_cpu  = task_cpu(p);
1196         load      = source_load(prev_cpu, idx);
1197         this_load = target_load(this_cpu, idx);
1198
1199         if (sync) {
1200                if (sched_feat(SYNC_LESS) &&
1201                    (curr->se.avg_overlap > sysctl_sched_migration_cost ||
1202                     p->se.avg_overlap > sysctl_sched_migration_cost))
1203                        sync = 0;
1204         } else {
1205                 if (sched_feat(SYNC_MORE) &&
1206                     (curr->se.avg_overlap < sysctl_sched_migration_cost &&
1207                      p->se.avg_overlap < sysctl_sched_migration_cost))
1208                         sync = 1;
1209         }
1210
1211         /*
1212          * If sync wakeup then subtract the (maximum possible)
1213          * effect of the currently running task from the load
1214          * of the current CPU:
1215          */
1216         if (sync) {
1217                 tg = task_group(current);
1218                 weight = current->se.load.weight;
1219
1220                 this_load += effective_load(tg, this_cpu, -weight, -weight);
1221                 load += effective_load(tg, prev_cpu, 0, -weight);
1222         }
1223
1224         tg = task_group(p);
1225         weight = p->se.load.weight;
1226
1227         imbalance = 100 + (sd->imbalance_pct - 100) / 2;
1228
1229         /*
1230          * In low-load situations, where prev_cpu is idle and this_cpu is idle
1231          * due to the sync cause above having dropped this_load to 0, we'll
1232          * always have an imbalance, but there's really nothing you can do
1233          * about that, so that's good too.
1234          *
1235          * Otherwise check if either cpus are near enough in load to allow this
1236          * task to be woken on this_cpu.
1237          */
1238         balanced = !this_load ||
1239                 100*(this_load + effective_load(tg, this_cpu, weight, weight)) <=
1240                 imbalance*(load + effective_load(tg, prev_cpu, 0, weight));
1241
1242         /*
1243          * If the currently running task will sleep within
1244          * a reasonable amount of time then attract this newly
1245          * woken task:
1246          */
1247         if (sync && balanced)
1248                 return 1;
1249
1250         schedstat_inc(p, se.nr_wakeups_affine_attempts);
1251         tl_per_task = cpu_avg_load_per_task(this_cpu);
1252
1253         if (balanced ||
1254             (this_load <= load &&
1255              this_load + target_load(prev_cpu, idx) <= tl_per_task)) {
1256                 /*
1257                  * This domain has SD_WAKE_AFFINE and
1258                  * p is cache cold in this domain, and
1259                  * there is no bad imbalance.
1260                  */
1261                 schedstat_inc(sd, ttwu_move_affine);
1262                 schedstat_inc(p, se.nr_wakeups_affine);
1263
1264                 return 1;
1265         }
1266         return 0;
1267 }
1268
1269 /*
1270  * find_idlest_group finds and returns the least busy CPU group within the
1271  * domain.
1272  */
1273 static struct sched_group *
1274 find_idlest_group(struct sched_domain *sd, struct task_struct *p,
1275                   int this_cpu, int load_idx)
1276 {
1277         struct sched_group *idlest = NULL, *this = NULL, *group = sd->groups;
1278         unsigned long min_load = ULONG_MAX, this_load = 0;
1279         int imbalance = 100 + (sd->imbalance_pct-100)/2;
1280
1281         do {
1282                 unsigned long load, avg_load;
1283                 int local_group;
1284                 int i;
1285
1286                 /* Skip over this group if it has no CPUs allowed */
1287                 if (!cpumask_intersects(sched_group_cpus(group),
1288                                         &p->cpus_allowed))
1289                         continue;
1290
1291                 local_group = cpumask_test_cpu(this_cpu,
1292                                                sched_group_cpus(group));
1293
1294                 /* Tally up the load of all CPUs in the group */
1295                 avg_load = 0;
1296
1297                 for_each_cpu(i, sched_group_cpus(group)) {
1298                         /* Bias balancing toward cpus of our domain */
1299                         if (local_group)
1300                                 load = source_load(i, load_idx);
1301                         else
1302                                 load = target_load(i, load_idx);
1303
1304                         avg_load += load;
1305                 }
1306
1307                 /* Adjust by relative CPU power of the group */
1308                 avg_load = (avg_load * SCHED_LOAD_SCALE) / group->cpu_power;
1309
1310                 if (local_group) {
1311                         this_load = avg_load;
1312                         this = group;
1313                 } else if (avg_load < min_load) {
1314                         min_load = avg_load;
1315                         idlest = group;
1316                 }
1317         } while (group = group->next, group != sd->groups);
1318
1319         if (!idlest || 100*this_load < imbalance*min_load)
1320                 return NULL;
1321         return idlest;
1322 }
1323
1324 /*
1325  * find_idlest_cpu - find the idlest cpu among the cpus in group.
1326  */
1327 static int
1328 find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu)
1329 {
1330         unsigned long load, min_load = ULONG_MAX;
1331         int idlest = -1;
1332         int i;
1333
1334         /* Traverse only the allowed CPUs */
1335         for_each_cpu_and(i, sched_group_cpus(group), &p->cpus_allowed) {
1336                 load = weighted_cpuload(i);
1337
1338                 if (load < min_load || (load == min_load && i == this_cpu)) {
1339                         min_load = load;
1340                         idlest = i;
1341                 }
1342         }
1343
1344         return idlest;
1345 }
1346
1347 /*
1348  * sched_balance_self: balance the current task (running on cpu) in domains
1349  * that have the 'flag' flag set. In practice, this is SD_BALANCE_FORK and
1350  * SD_BALANCE_EXEC.
1351  *
1352  * Balance, ie. select the least loaded group.
1353  *
1354  * Returns the target CPU number, or the same CPU if no balancing is needed.
1355  *
1356  * preempt must be disabled.
1357  */
1358 static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
1359 {
1360         struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
1361         int cpu = smp_processor_id();
1362         int prev_cpu = task_cpu(p);
1363         int new_cpu = cpu;
1364         int want_affine = 0;
1365         int want_sd = 1;
1366         int sync = wake_flags & WF_SYNC;
1367
1368         if (sd_flag & SD_BALANCE_WAKE) {
1369                 if (sched_feat(AFFINE_WAKEUPS) &&
1370                     cpumask_test_cpu(cpu, &p->cpus_allowed))
1371                         want_affine = 1;
1372                 new_cpu = prev_cpu;
1373         }
1374
1375         rcu_read_lock();
1376         for_each_domain(cpu, tmp) {
1377                 /*
1378                  * If power savings logic is enabled for a domain, see if we
1379                  * are not overloaded, if so, don't balance wider.
1380                  */
1381                 if (tmp->flags & (SD_POWERSAVINGS_BALANCE|SD_PREFER_LOCAL)) {
1382                         unsigned long power = 0;
1383                         unsigned long nr_running = 0;
1384                         unsigned long capacity;
1385                         int i;
1386
1387                         for_each_cpu(i, sched_domain_span(tmp)) {
1388                                 power += power_of(i);
1389                                 nr_running += cpu_rq(i)->cfs.nr_running;
1390                         }
1391
1392                         capacity = DIV_ROUND_CLOSEST(power, SCHED_LOAD_SCALE);
1393
1394                         if (tmp->flags & SD_POWERSAVINGS_BALANCE)
1395                                 nr_running /= 2;
1396
1397                         if (nr_running < capacity)
1398                                 want_sd = 0;
1399                 }
1400
1401                 if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
1402                     cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
1403
1404                         affine_sd = tmp;
1405                         want_affine = 0;
1406                 }
1407
1408                 if (!want_sd && !want_affine)
1409                         break;
1410
1411                 if (!(tmp->flags & sd_flag))
1412                         continue;
1413
1414                 if (want_sd)
1415                         sd = tmp;
1416         }
1417
1418         if (sched_feat(LB_SHARES_UPDATE)) {
1419                 /*
1420                  * Pick the largest domain to update shares over
1421                  */
1422                 tmp = sd;
1423                 if (affine_sd && (!tmp ||
1424                                   cpumask_weight(sched_domain_span(affine_sd)) >
1425                                   cpumask_weight(sched_domain_span(sd))))
1426                         tmp = affine_sd;
1427
1428                 if (tmp)
1429                         update_shares(tmp);
1430         }
1431
1432         if (affine_sd && wake_affine(affine_sd, p, sync)) {
1433                 new_cpu = cpu;
1434                 goto out;
1435         }
1436
1437         while (sd) {
1438                 int load_idx = sd->forkexec_idx;
1439                 struct sched_group *group;
1440                 int weight;
1441
1442                 if (!(sd->flags & sd_flag)) {
1443                         sd = sd->child;
1444                         continue;
1445                 }
1446
1447                 if (sd_flag & SD_BALANCE_WAKE)
1448                         load_idx = sd->wake_idx;
1449
1450                 group = find_idlest_group(sd, p, cpu, load_idx);
1451                 if (!group) {
1452                         sd = sd->child;
1453                         continue;
1454                 }
1455
1456                 new_cpu = find_idlest_cpu(group, p, cpu);
1457                 if (new_cpu == -1 || new_cpu == cpu) {
1458                         /* Now try balancing at a lower domain level of cpu */
1459                         sd = sd->child;
1460                         continue;
1461                 }
1462
1463                 /* Now try balancing at a lower domain level of new_cpu */
1464                 cpu = new_cpu;
1465                 weight = cpumask_weight(sched_domain_span(sd));
1466                 sd = NULL;
1467                 for_each_domain(cpu, tmp) {
1468                         if (weight <= cpumask_weight(sched_domain_span(tmp)))
1469                                 break;
1470                         if (tmp->flags & sd_flag)
1471                                 sd = tmp;
1472                 }
1473                 /* while loop will break here if sd == NULL */
1474         }
1475
1476 out:
1477         rcu_read_unlock();
1478         return new_cpu;
1479 }
1480 #endif /* CONFIG_SMP */
1481
1482 /*
1483  * Adaptive granularity
1484  *
1485  * se->avg_wakeup gives the average time a task runs until it does a wakeup,
1486  * with the limit of wakeup_gran -- when it never does a wakeup.
1487  *
1488  * So the smaller avg_wakeup is the faster we want this task to preempt,
1489  * but we don't want to treat the preemptee unfairly and therefore allow it
1490  * to run for at least the amount of time we'd like to run.
1491  *
1492  * NOTE: we use 2*avg_wakeup to increase the probability of actually doing one
1493  *
1494  * NOTE: we use *nr_running to scale with load, this nicely matches the
1495  *       degrading latency on load.
1496  */
1497 static unsigned long
1498 adaptive_gran(struct sched_entity *curr, struct sched_entity *se)
1499 {
1500         u64 this_run = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
1501         u64 expected_wakeup = 2*se->avg_wakeup * cfs_rq_of(se)->nr_running;
1502         u64 gran = 0;
1503
1504         if (this_run < expected_wakeup)
1505                 gran = expected_wakeup - this_run;
1506
1507         return min_t(s64, gran, sysctl_sched_wakeup_granularity);
1508 }
1509
1510 static unsigned long
1511 wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
1512 {
1513         unsigned long gran = sysctl_sched_wakeup_granularity;
1514
1515         if (cfs_rq_of(curr)->curr && sched_feat(ADAPTIVE_GRAN))
1516                 gran = adaptive_gran(curr, se);
1517
1518         /*
1519          * Since its curr running now, convert the gran from real-time
1520          * to virtual-time in his units.
1521          */
1522         if (sched_feat(ASYM_GRAN)) {
1523                 /*
1524                  * By using 'se' instead of 'curr' we penalize light tasks, so
1525                  * they get preempted easier. That is, if 'se' < 'curr' then
1526                  * the resulting gran will be larger, therefore penalizing the
1527                  * lighter, if otoh 'se' > 'curr' then the resulting gran will
1528                  * be smaller, again penalizing the lighter task.
1529                  *
1530                  * This is especially important for buddies when the leftmost
1531                  * task is higher priority than the buddy.
1532                  */
1533                 if (unlikely(se->load.weight != NICE_0_LOAD))
1534                         gran = calc_delta_fair(gran, se);
1535         } else {
1536                 if (unlikely(curr->load.weight != NICE_0_LOAD))
1537                         gran = calc_delta_fair(gran, curr);
1538         }
1539
1540         return gran;
1541 }
1542
1543 /*
1544  * Should 'se' preempt 'curr'.
1545  *
1546  *             |s1
1547  *        |s2
1548  *   |s3
1549  *         g
1550  *      |<--->|c
1551  *
1552  *  w(c, s1) = -1
1553  *  w(c, s2) =  0
1554  *  w(c, s3) =  1
1555  *
1556  */
1557 static int
1558 wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se)
1559 {
1560         s64 gran, vdiff = curr->vruntime - se->vruntime;
1561
1562         if (vdiff <= 0)
1563                 return -1;
1564
1565         gran = wakeup_gran(curr, se);
1566         if (vdiff > gran)
1567                 return 1;
1568
1569         return 0;
1570 }
1571
1572 static void set_last_buddy(struct sched_entity *se)
1573 {
1574         if (likely(task_of(se)->policy != SCHED_IDLE)) {
1575                 for_each_sched_entity(se)
1576                         cfs_rq_of(se)->last = se;
1577         }
1578 }
1579
1580 static void set_next_buddy(struct sched_entity *se)
1581 {
1582         if (likely(task_of(se)->policy != SCHED_IDLE)) {
1583                 for_each_sched_entity(se)
1584                         cfs_rq_of(se)->next = se;
1585         }
1586 }
1587
1588 /*
1589  * Preempt the current task with a newly woken task if needed:
1590  */
1591 static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
1592 {
1593         struct task_struct *curr = rq->curr;
1594         struct sched_entity *se = &curr->se, *pse = &p->se;
1595         struct cfs_rq *cfs_rq = task_cfs_rq(curr);
1596         int sync = wake_flags & WF_SYNC;
1597         int scale = cfs_rq->nr_running >= sched_nr_latency;
1598
1599         update_curr(cfs_rq);
1600
1601         if (unlikely(rt_prio(p->prio))) {
1602                 resched_task(curr);
1603                 return;
1604         }
1605
1606         if (unlikely(p->sched_class != &fair_sched_class))
1607                 return;
1608
1609         if (unlikely(se == pse))
1610                 return;
1611
1612         if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK))
1613                 set_next_buddy(pse);
1614
1615         /*
1616          * We can come here with TIF_NEED_RESCHED already set from new task
1617          * wake up path.
1618          */
1619         if (test_tsk_need_resched(curr))
1620                 return;
1621
1622         /*
1623          * Batch and idle tasks do not preempt (their preemption is driven by
1624          * the tick):
1625          */
1626         if (unlikely(p->policy != SCHED_NORMAL))
1627                 return;
1628
1629         /* Idle tasks are by definition preempted by everybody. */
1630         if (unlikely(curr->policy == SCHED_IDLE)) {
1631                 resched_task(curr);
1632                 return;
1633         }
1634
1635         if ((sched_feat(WAKEUP_SYNC) && sync) ||
1636             (sched_feat(WAKEUP_OVERLAP) &&
1637              (se->avg_overlap < sysctl_sched_migration_cost &&
1638               pse->avg_overlap < sysctl_sched_migration_cost))) {
1639                 resched_task(curr);
1640                 return;
1641         }
1642
1643         if (sched_feat(WAKEUP_RUNNING)) {
1644                 if (pse->avg_running < se->avg_running) {
1645                         set_next_buddy(pse);
1646                         resched_task(curr);
1647                         return;
1648                 }
1649         }
1650
1651         if (!sched_feat(WAKEUP_PREEMPT))
1652                 return;
1653
1654         find_matching_se(&se, &pse);
1655
1656         BUG_ON(!pse);
1657
1658         if (wakeup_preempt_entity(se, pse) == 1) {
1659                 resched_task(curr);
1660                 /*
1661                  * Only set the backward buddy when the current task is still
1662                  * on the rq. This can happen when a wakeup gets interleaved
1663                  * with schedule on the ->pre_schedule() or idle_balance()
1664                  * point, either of which can * drop the rq lock.
1665                  *
1666                  * Also, during early boot the idle thread is in the fair class,
1667                  * for obvious reasons its a bad idea to schedule back to it.
1668                  */
1669                 if (unlikely(!se->on_rq || curr == rq->idle))
1670                         return;
1671                 if (sched_feat(LAST_BUDDY) && scale && entity_is_task(se))
1672                         set_last_buddy(se);
1673         }
1674 }
1675
1676 static struct task_struct *pick_next_task_fair(struct rq *rq)
1677 {
1678         struct task_struct *p;
1679         struct cfs_rq *cfs_rq = &rq->cfs;
1680         struct sched_entity *se;
1681
1682         if (unlikely(!cfs_rq->nr_running))
1683                 return NULL;
1684
1685         do {
1686                 se = pick_next_entity(cfs_rq);
1687                 set_next_entity(cfs_rq, se);
1688                 cfs_rq = group_cfs_rq(se);
1689         } while (cfs_rq);
1690
1691         p = task_of(se);
1692         hrtick_start_fair(rq, p);
1693
1694         return p;
1695 }
1696
1697 /*
1698  * Account for a descheduled task:
1699  */
1700 static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
1701 {
1702         struct sched_entity *se = &prev->se;
1703         struct cfs_rq *cfs_rq;
1704
1705         for_each_sched_entity(se) {
1706                 cfs_rq = cfs_rq_of(se);
1707                 put_prev_entity(cfs_rq, se);
1708         }
1709 }
1710
1711 #ifdef CONFIG_SMP
1712 /**************************************************
1713  * Fair scheduling class load-balancing methods:
1714  */
1715
1716 /*
1717  * Load-balancing iterator. Note: while the runqueue stays locked
1718  * during the whole iteration, the current task might be
1719  * dequeued so the iterator has to be dequeue-safe. Here we
1720  * achieve that by always pre-iterating before returning
1721  * the current task:
1722  */
1723 static struct task_struct *
1724 __load_balance_iterator(struct cfs_rq *cfs_rq, struct list_head *next)
1725 {
1726         struct task_struct *p = NULL;
1727         struct sched_entity *se;
1728
1729         if (next == &cfs_rq->tasks)
1730                 return NULL;
1731
1732         se = list_entry(next, struct sched_entity, group_node);
1733         p = task_of(se);
1734         cfs_rq->balance_iterator = next->next;
1735
1736         return p;
1737 }
1738
1739 static struct task_struct *load_balance_start_fair(void *arg)
1740 {
1741         struct cfs_rq *cfs_rq = arg;
1742
1743         return __load_balance_iterator(cfs_rq, cfs_rq->tasks.next);
1744 }
1745
1746 static struct task_struct *load_balance_next_fair(void *arg)
1747 {
1748         struct cfs_rq *cfs_rq = arg;
1749
1750         return __load_balance_iterator(cfs_rq, cfs_rq->balance_iterator);
1751 }
1752
1753 static unsigned long
1754 __load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
1755                 unsigned long max_load_move, struct sched_domain *sd,
1756                 enum cpu_idle_type idle, int *all_pinned, int *this_best_prio,
1757                 struct cfs_rq *cfs_rq)
1758 {
1759         struct rq_iterator cfs_rq_iterator;
1760
1761         cfs_rq_iterator.start = load_balance_start_fair;
1762         cfs_rq_iterator.next = load_balance_next_fair;
1763         cfs_rq_iterator.arg = cfs_rq;
1764
1765         return balance_tasks(this_rq, this_cpu, busiest,
1766                         max_load_move, sd, idle, all_pinned,
1767                         this_best_prio, &cfs_rq_iterator);
1768 }
1769
1770 #ifdef CONFIG_FAIR_GROUP_SCHED
1771 static unsigned long
1772 load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
1773                   unsigned long max_load_move,
1774                   struct sched_domain *sd, enum cpu_idle_type idle,
1775                   int *all_pinned, int *this_best_prio)
1776 {
1777         long rem_load_move = max_load_move;
1778         int busiest_cpu = cpu_of(busiest);
1779         struct task_group *tg;
1780
1781         rcu_read_lock();
1782         update_h_load(busiest_cpu);
1783
1784         list_for_each_entry_rcu(tg, &task_groups, list) {
1785                 struct cfs_rq *busiest_cfs_rq = tg->cfs_rq[busiest_cpu];
1786                 unsigned long busiest_h_load = busiest_cfs_rq->h_load;
1787                 unsigned long busiest_weight = busiest_cfs_rq->load.weight;
1788                 u64 rem_load, moved_load;
1789
1790                 /*
1791                  * empty group
1792                  */
1793                 if (!busiest_cfs_rq->task_weight)
1794                         continue;
1795
1796                 rem_load = (u64)rem_load_move * busiest_weight;
1797                 rem_load = div_u64(rem_load, busiest_h_load + 1);
1798
1799                 moved_load = __load_balance_fair(this_rq, this_cpu, busiest,
1800                                 rem_load, sd, idle, all_pinned, this_best_prio,
1801                                 tg->cfs_rq[busiest_cpu]);
1802
1803                 if (!moved_load)
1804                         continue;
1805
1806                 moved_load *= busiest_h_load;
1807                 moved_load = div_u64(moved_load, busiest_weight + 1);
1808
1809                 rem_load_move -= moved_load;
1810                 if (rem_load_move < 0)
1811                         break;
1812         }
1813         rcu_read_unlock();
1814
1815         return max_load_move - rem_load_move;
1816 }
1817 #else
1818 static unsigned long
1819 load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
1820                   unsigned long max_load_move,
1821                   struct sched_domain *sd, enum cpu_idle_type idle,
1822                   int *all_pinned, int *this_best_prio)
1823 {
1824         return __load_balance_fair(this_rq, this_cpu, busiest,
1825                         max_load_move, sd, idle, all_pinned,
1826                         this_best_prio, &busiest->cfs);
1827 }
1828 #endif
1829
1830 static int
1831 move_one_task_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
1832                    struct sched_domain *sd, enum cpu_idle_type idle)
1833 {
1834         struct cfs_rq *busy_cfs_rq;
1835         struct rq_iterator cfs_rq_iterator;
1836
1837         cfs_rq_iterator.start = load_balance_start_fair;
1838         cfs_rq_iterator.next = load_balance_next_fair;
1839
1840         for_each_leaf_cfs_rq(busiest, busy_cfs_rq) {
1841                 /*
1842                  * pass busy_cfs_rq argument into
1843                  * load_balance_[start|next]_fair iterators
1844                  */
1845                 cfs_rq_iterator.arg = busy_cfs_rq;
1846                 if (iter_move_one_task(this_rq, this_cpu, busiest, sd, idle,
1847                                        &cfs_rq_iterator))
1848                     return 1;
1849         }
1850
1851         return 0;
1852 }
1853 #endif /* CONFIG_SMP */
1854
1855 /*
1856  * scheduler tick hitting a task of our scheduling class:
1857  */
1858 static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
1859 {
1860         struct cfs_rq *cfs_rq;
1861         struct sched_entity *se = &curr->se;
1862
1863         for_each_sched_entity(se) {
1864                 cfs_rq = cfs_rq_of(se);
1865                 entity_tick(cfs_rq, se, queued);
1866         }
1867 }
1868
1869 /*
1870  * Share the fairness runtime between parent and child, thus the
1871  * total amount of pressure for CPU stays equal - new tasks
1872  * get a chance to run but frequent forkers are not allowed to
1873  * monopolize the CPU. Note: the parent runqueue is locked,
1874  * the child is not running yet.
1875  */
1876 static void task_new_fair(struct rq *rq, struct task_struct *p)
1877 {
1878         struct cfs_rq *cfs_rq = task_cfs_rq(p);
1879         struct sched_entity *se = &p->se, *curr = cfs_rq->curr;
1880         int this_cpu = smp_processor_id();
1881
1882         sched_info_queued(p);
1883
1884         update_curr(cfs_rq);
1885         if (curr)
1886                 se->vruntime = curr->vruntime;
1887         place_entity(cfs_rq, se, 1);
1888
1889         /* 'curr' will be NULL if the child belongs to a different group */
1890         if (sysctl_sched_child_runs_first && this_cpu == task_cpu(p) &&
1891                         curr && entity_before(curr, se)) {
1892                 /*
1893                  * Upon rescheduling, sched_class::put_prev_task() will place
1894                  * 'current' within the tree based on its new key value.
1895                  */
1896                 swap(curr->vruntime, se->vruntime);
1897                 resched_task(rq->curr);
1898         }
1899
1900         enqueue_task_fair(rq, p, 0);
1901 }
1902
1903 /*
1904  * Priority of the task has changed. Check to see if we preempt
1905  * the current task.
1906  */
1907 static void prio_changed_fair(struct rq *rq, struct task_struct *p,
1908                               int oldprio, int running)
1909 {
1910         /*
1911          * Reschedule if we are currently running on this runqueue and
1912          * our priority decreased, or if we are not currently running on
1913          * this runqueue and our priority is higher than the current's
1914          */
1915         if (running) {
1916                 if (p->prio > oldprio)
1917                         resched_task(rq->curr);
1918         } else
1919                 check_preempt_curr(rq, p, 0);
1920 }
1921
1922 /*
1923  * We switched to the sched_fair class.
1924  */
1925 static void switched_to_fair(struct rq *rq, struct task_struct *p,
1926                              int running)
1927 {
1928         /*
1929          * We were most likely switched from sched_rt, so
1930          * kick off the schedule if running, otherwise just see
1931          * if we can still preempt the current task.
1932          */
1933         if (running)
1934                 resched_task(rq->curr);
1935         else
1936                 check_preempt_curr(rq, p, 0);
1937 }
1938
1939 /* Account for a task changing its policy or group.
1940  *
1941  * This routine is mostly called to set cfs_rq->curr field when a task
1942  * migrates between groups/classes.
1943  */
1944 static void set_curr_task_fair(struct rq *rq)
1945 {
1946         struct sched_entity *se = &rq->curr->se;
1947
1948         for_each_sched_entity(se)
1949                 set_next_entity(cfs_rq_of(se), se);
1950 }
1951
1952 #ifdef CONFIG_FAIR_GROUP_SCHED
1953 static void moved_group_fair(struct task_struct *p)
1954 {
1955         struct cfs_rq *cfs_rq = task_cfs_rq(p);
1956
1957         update_curr(cfs_rq);
1958         place_entity(cfs_rq, &p->se, 1);
1959 }
1960 #endif
1961
1962 unsigned int get_rr_interval_fair(struct task_struct *task)
1963 {
1964         struct sched_entity *se = &task->se;
1965         unsigned long flags;
1966         struct rq *rq;
1967         unsigned int rr_interval = 0;
1968
1969         /*
1970          * Time slice is 0 for SCHED_OTHER tasks that are on an otherwise
1971          * idle runqueue:
1972          */
1973         rq = task_rq_lock(task, &flags);
1974         if (rq->cfs.load.weight)
1975                 rr_interval = NS_TO_JIFFIES(sched_slice(&rq->cfs, se));
1976         task_rq_unlock(rq, &flags);
1977
1978         return rr_interval;
1979 }
1980
1981 /*
1982  * All the scheduling class methods:
1983  */
1984 static const struct sched_class fair_sched_class = {
1985         .next                   = &idle_sched_class,
1986         .enqueue_task           = enqueue_task_fair,
1987         .dequeue_task           = dequeue_task_fair,
1988         .yield_task             = yield_task_fair,
1989
1990         .check_preempt_curr     = check_preempt_wakeup,
1991
1992         .pick_next_task         = pick_next_task_fair,
1993         .put_prev_task          = put_prev_task_fair,
1994
1995 #ifdef CONFIG_SMP
1996         .select_task_rq         = select_task_rq_fair,
1997
1998         .load_balance           = load_balance_fair,
1999         .move_one_task          = move_one_task_fair,
2000 #endif
2001
2002         .set_curr_task          = set_curr_task_fair,
2003         .task_tick              = task_tick_fair,
2004         .task_new               = task_new_fair,
2005
2006         .prio_changed           = prio_changed_fair,
2007         .switched_to            = switched_to_fair,
2008
2009         .get_rr_interval        = get_rr_interval_fair,
2010
2011 #ifdef CONFIG_FAIR_GROUP_SCHED
2012         .moved_group            = moved_group_fair,
2013 #endif
2014 };
2015
2016 #ifdef CONFIG_SCHED_DEBUG
2017 static void print_cfs_stats(struct seq_file *m, int cpu)
2018 {
2019         struct cfs_rq *cfs_rq;
2020
2021         rcu_read_lock();
2022         for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
2023                 print_cfs_rq(m, cpu, cfs_rq);
2024         rcu_read_unlock();
2025 }
2026 #endif