blob: 08ae848b71d4643429f67ef312a7bfd78a32aa0a [file] [log] [blame]
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001/*
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>
Peter Zijlstra21805082007-08-25 18:41:53 +020018 *
19 * Adaptive scheduling granularity, math enhancements by Peter Zijlstra
20 * Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020021 */
22
Arjan van de Ven97455122008-01-25 21:08:34 +010023#include <linux/latencytop.h>
24
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020025/*
Peter Zijlstra21805082007-08-25 18:41:53 +020026 * Targeted preemption latency for CPU-bound tasks:
Zou Nan hai722aab02007-11-26 21:21:49 +010027 * (default: 20ms * (1 + ilog(ncpus)), units: nanoseconds)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020028 *
Peter Zijlstra21805082007-08-25 18:41:53 +020029 * NOTE: this latency value is not the same as the concept of
Ingo Molnard274a4c2007-10-15 17:00:14 +020030 * 'timeslice length' - timeslices in CFS are of variable length
31 * and have no persistent notion like in traditional, time-slice
32 * based scheduling concepts.
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020033 *
Ingo Molnard274a4c2007-10-15 17:00:14 +020034 * (to see the precise effective timeslice length of your workload,
35 * run vmstat and monitor the context-switches (cs) field)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020036 */
Ingo Molnar19978ca2007-11-09 22:39:38 +010037unsigned int sysctl_sched_latency = 20000000ULL;
Ingo Molnar2bd8e6d2007-10-15 17:00:02 +020038
39/*
Peter Zijlstrab2be5e92007-11-09 22:39:37 +010040 * Minimal preemption granularity for CPU-bound tasks:
Zou Nan hai722aab02007-11-26 21:21:49 +010041 * (default: 4 msec * (1 + ilog(ncpus)), units: nanoseconds)
Peter Zijlstrab2be5e92007-11-09 22:39:37 +010042 */
Zou Nan hai722aab02007-11-26 21:21:49 +010043unsigned int sysctl_sched_min_granularity = 4000000ULL;
Peter Zijlstrab2be5e92007-11-09 22:39:37 +010044
45/*
46 * is kept at sysctl_sched_latency / sysctl_sched_min_granularity
47 */
Zou Nan hai722aab02007-11-26 21:21:49 +010048static unsigned int sched_nr_latency = 5;
Peter Zijlstrab2be5e92007-11-09 22:39:37 +010049
50/*
Ingo Molnar2bd8e6d2007-10-15 17:00:02 +020051 * After fork, child runs first. (default) If set to 0 then
52 * parent will (try to) run first.
53 */
54const_debug unsigned int sysctl_sched_child_runs_first = 1;
Peter Zijlstra21805082007-08-25 18:41:53 +020055
56/*
Ingo Molnar1799e352007-09-19 23:34:46 +020057 * sys_sched_yield() compat mode
58 *
59 * This option switches the agressive yield implementation of the
60 * old scheduler back on.
61 */
62unsigned int __read_mostly sysctl_sched_compat_yield;
63
64/*
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020065 * SCHED_OTHER wake-up granularity.
Peter Zijlstra0bbd3332008-04-19 19:44:57 +020066 * (default: 10 msec * (1 + ilog(ncpus)), units: nanoseconds)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020067 *
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 */
Peter Zijlstra0bbd3332008-04-19 19:44:57 +020072unsigned int sysctl_sched_wakeup_granularity = 10000000UL;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020073
Ingo Molnarda84d962007-10-15 17:00:18 +020074const_debug unsigned int sysctl_sched_migration_cost = 500000UL;
75
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020076/**************************************************************
77 * CFS operations on generic schedulable entities:
78 */
79
Peter Zijlstrab7581492008-04-19 19:45:00 +020080static inline struct task_struct *task_of(struct sched_entity *se)
81{
82 return container_of(se, struct task_struct, se);
83}
84
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020085#ifdef CONFIG_FAIR_GROUP_SCHED
86
87/* cpu runqueue to which this cfs_rq is attached */
88static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
89{
90 return cfs_rq->rq;
91}
92
Ingo Molnarbf0f6f22007-07-09 18:51:58 +020093/* An entity is a task if it doesn't "own" a runqueue */
94#define entity_is_task(se) (!se->my_q)
95
Peter Zijlstrab7581492008-04-19 19:45:00 +020096/* Walk up scheduling entities hierarchy */
97#define for_each_sched_entity(se) \
98 for (; se; se = se->parent)
99
100static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
101{
102 return p->se.cfs_rq;
103}
104
105/* runqueue on which this entity is (to be) queued */
106static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
107{
108 return se->cfs_rq;
109}
110
111/* runqueue "owned" by this group */
112static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
113{
114 return grp->my_q;
115}
116
117/* Given a group's cfs_rq on one cpu, return its corresponding cfs_rq on
118 * another cpu ('this_cpu')
119 */
120static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
121{
122 return cfs_rq->tg->cfs_rq[this_cpu];
123}
124
125/* Iterate thr' all leaf cfs_rq's on a runqueue */
126#define for_each_leaf_cfs_rq(rq, cfs_rq) \
127 list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)
128
129/* Do the two (enqueued) entities belong to the same group ? */
130static inline int
131is_same_group(struct sched_entity *se, struct sched_entity *pse)
132{
133 if (se->cfs_rq == pse->cfs_rq)
134 return 1;
135
136 return 0;
137}
138
139static inline struct sched_entity *parent_entity(struct sched_entity *se)
140{
141 return se->parent;
142}
143
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200144#else /* CONFIG_FAIR_GROUP_SCHED */
145
146static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
147{
148 return container_of(cfs_rq, struct rq, cfs);
149}
150
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200151#define entity_is_task(se) 1
152
Peter Zijlstrab7581492008-04-19 19:45:00 +0200153#define for_each_sched_entity(se) \
154 for (; se; se = NULL)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200155
Peter Zijlstrab7581492008-04-19 19:45:00 +0200156static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200157{
Peter Zijlstrab7581492008-04-19 19:45:00 +0200158 return &task_rq(p)->cfs;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200159}
160
Peter Zijlstrab7581492008-04-19 19:45:00 +0200161static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
162{
163 struct task_struct *p = task_of(se);
164 struct rq *rq = task_rq(p);
165
166 return &rq->cfs;
167}
168
169/* runqueue "owned" by this group */
170static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
171{
172 return NULL;
173}
174
175static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
176{
177 return &cpu_rq(this_cpu)->cfs;
178}
179
180#define for_each_leaf_cfs_rq(rq, cfs_rq) \
181 for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)
182
183static inline int
184is_same_group(struct sched_entity *se, struct sched_entity *pse)
185{
186 return 1;
187}
188
189static inline struct sched_entity *parent_entity(struct sched_entity *se)
190{
191 return NULL;
192}
193
194#endif /* CONFIG_FAIR_GROUP_SCHED */
195
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200196
197/**************************************************************
198 * Scheduling class tree data structure manipulation methods:
199 */
200
Ingo Molnar0702e3e2007-10-15 17:00:14 +0200201static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
Peter Zijlstra02e04312007-10-15 17:00:07 +0200202{
Peter Zijlstra368059a2007-10-15 17:00:11 +0200203 s64 delta = (s64)(vruntime - min_vruntime);
204 if (delta > 0)
Peter Zijlstra02e04312007-10-15 17:00:07 +0200205 min_vruntime = vruntime;
206
207 return min_vruntime;
208}
209
Ingo Molnar0702e3e2007-10-15 17:00:14 +0200210static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
Peter Zijlstrab0ffd242007-10-15 17:00:12 +0200211{
212 s64 delta = (s64)(vruntime - min_vruntime);
213 if (delta < 0)
214 min_vruntime = vruntime;
215
216 return min_vruntime;
217}
218
Ingo Molnar0702e3e2007-10-15 17:00:14 +0200219static inline s64 entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
Peter Zijlstra9014623c2007-10-15 17:00:05 +0200220{
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200221 return se->vruntime - cfs_rq->min_vruntime;
Peter Zijlstra9014623c2007-10-15 17:00:05 +0200222}
223
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200224/*
225 * Enqueue an entity into the rb-tree:
226 */
Ingo Molnar0702e3e2007-10-15 17:00:14 +0200227static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200228{
229 struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
230 struct rb_node *parent = NULL;
231 struct sched_entity *entry;
Peter Zijlstra9014623c2007-10-15 17:00:05 +0200232 s64 key = entity_key(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200233 int leftmost = 1;
234
235 /*
236 * Find the right place in the rbtree:
237 */
238 while (*link) {
239 parent = *link;
240 entry = rb_entry(parent, struct sched_entity, run_node);
241 /*
242 * We dont care about collisions. Nodes with
243 * the same key stay together.
244 */
Peter Zijlstra9014623c2007-10-15 17:00:05 +0200245 if (key < entity_key(cfs_rq, entry)) {
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200246 link = &parent->rb_left;
247 } else {
248 link = &parent->rb_right;
249 leftmost = 0;
250 }
251 }
252
253 /*
254 * Maintain a cache of leftmost tree entries (it is frequently
255 * used):
256 */
Peter Zijlstra3fe69742008-03-14 20:55:51 +0100257 if (leftmost) {
Ingo Molnar57cb4992007-10-15 17:00:11 +0200258 cfs_rq->rb_leftmost = &se->run_node;
Peter Zijlstra3fe69742008-03-14 20:55:51 +0100259 /*
260 * maintain cfs_rq->min_vruntime to be a monotonic increasing
261 * value tracking the leftmost vruntime in the tree.
262 */
263 cfs_rq->min_vruntime =
264 max_vruntime(cfs_rq->min_vruntime, se->vruntime);
265 }
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200266
267 rb_link_node(&se->run_node, parent, link);
268 rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200269}
270
Ingo Molnar0702e3e2007-10-15 17:00:14 +0200271static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200272{
Peter Zijlstra3fe69742008-03-14 20:55:51 +0100273 if (cfs_rq->rb_leftmost == &se->run_node) {
274 struct rb_node *next_node;
275 struct sched_entity *next;
276
277 next_node = rb_next(&se->run_node);
278 cfs_rq->rb_leftmost = next_node;
279
280 if (next_node) {
281 next = rb_entry(next_node,
282 struct sched_entity, run_node);
283 cfs_rq->min_vruntime =
284 max_vruntime(cfs_rq->min_vruntime,
285 next->vruntime);
286 }
287 }
Ingo Molnare9acbff2007-10-15 17:00:04 +0200288
Peter Zijlstraaa2ac252008-03-14 21:12:12 +0100289 if (cfs_rq->next == se)
290 cfs_rq->next = NULL;
291
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200292 rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200293}
294
295static inline struct rb_node *first_fair(struct cfs_rq *cfs_rq)
296{
297 return cfs_rq->rb_leftmost;
298}
299
300static struct sched_entity *__pick_next_entity(struct cfs_rq *cfs_rq)
301{
302 return rb_entry(first_fair(cfs_rq), struct sched_entity, run_node);
303}
304
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200305static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
306{
Ingo Molnar7eee3e62008-02-22 10:32:21 +0100307 struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200308
Balbir Singh70eee742008-02-22 13:25:53 +0530309 if (!last)
310 return NULL;
Ingo Molnar7eee3e62008-02-22 10:32:21 +0100311
312 return rb_entry(last, struct sched_entity, run_node);
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200313}
314
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200315/**************************************************************
316 * Scheduling class statistics methods:
317 */
318
Peter Zijlstrab2be5e92007-11-09 22:39:37 +0100319#ifdef CONFIG_SCHED_DEBUG
320int sched_nr_latency_handler(struct ctl_table *table, int write,
321 struct file *filp, void __user *buffer, size_t *lenp,
322 loff_t *ppos)
323{
324 int ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos);
325
326 if (ret || !write)
327 return ret;
328
329 sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
330 sysctl_sched_min_granularity);
331
332 return 0;
333}
334#endif
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200335
336/*
337 * The idea is to set a period in which each task runs once.
338 *
339 * When there are too many tasks (sysctl_sched_nr_latency) we have to stretch
340 * this period because otherwise the slices get too small.
341 *
342 * p = (nr <= nl) ? l : l*nr/nl
343 */
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +0200344static u64 __sched_period(unsigned long nr_running)
345{
346 u64 period = sysctl_sched_latency;
Peter Zijlstrab2be5e92007-11-09 22:39:37 +0100347 unsigned long nr_latency = sched_nr_latency;
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +0200348
349 if (unlikely(nr_running > nr_latency)) {
Peter Zijlstra4bf0b772008-01-25 21:08:21 +0100350 period = sysctl_sched_min_granularity;
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +0200351 period *= nr_running;
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +0200352 }
353
354 return period;
355}
356
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200357/*
358 * We calculate the wall-time slice from the period by taking a part
359 * proportional to the weight.
360 *
361 * s = p*w/rw
362 */
Peter Zijlstra6d0f0eb2007-10-15 17:00:05 +0200363static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
Peter Zijlstra21805082007-08-25 18:41:53 +0200364{
Ingo Molnarf9305d42008-05-29 11:23:17 +0200365 u64 slice = __sched_period(cfs_rq->nr_running);
366
367 for_each_sched_entity(se) {
368 cfs_rq = cfs_rq_of(se);
369
370 slice *= se->load.weight;
371 do_div(slice, cfs_rq->load.weight);
372 }
373
374
375 return slice;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200376}
377
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200378/*
Peter Zijlstraac884de2008-04-19 19:45:00 +0200379 * We calculate the vruntime slice of a to be inserted task
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200380 *
Ingo Molnarf9305d42008-05-29 11:23:17 +0200381 * vs = s/w = p/rw
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200382 */
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200383static u64 sched_vslice_add(struct cfs_rq *cfs_rq, struct sched_entity *se)
384{
Peter Zijlstraac884de2008-04-19 19:45:00 +0200385 unsigned long nr_running = cfs_rq->nr_running;
Ingo Molnarf9305d42008-05-29 11:23:17 +0200386 unsigned long weight;
387 u64 vslice;
Peter Zijlstraac884de2008-04-19 19:45:00 +0200388
389 if (!se->on_rq)
390 nr_running++;
391
Ingo Molnarf9305d42008-05-29 11:23:17 +0200392 vslice = __sched_period(nr_running);
Peter Zijlstraac884de2008-04-19 19:45:00 +0200393
394 for_each_sched_entity(se) {
Ingo Molnarf9305d42008-05-29 11:23:17 +0200395 cfs_rq = cfs_rq_of(se);
Peter Zijlstraac884de2008-04-19 19:45:00 +0200396
Ingo Molnarf9305d42008-05-29 11:23:17 +0200397 weight = cfs_rq->load.weight;
398 if (!se->on_rq)
399 weight += se->load.weight;
Peter Zijlstraac884de2008-04-19 19:45:00 +0200400
Ingo Molnarf9305d42008-05-29 11:23:17 +0200401 vslice *= NICE_0_LOAD;
402 do_div(vslice, weight);
Peter Zijlstraac884de2008-04-19 19:45:00 +0200403 }
404
Ingo Molnarf9305d42008-05-29 11:23:17 +0200405 return vslice;
Peter Zijlstra67e9fb22007-10-15 17:00:10 +0200406}
407
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200408/*
409 * Update the current task's runtime statistics. Skip current tasks that
410 * are not in our scheduling class.
411 */
412static inline void
Ingo Molnar8ebc91d2007-10-15 17:00:03 +0200413__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
414 unsigned long delta_exec)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200415{
Ingo Molnarbbdba7c2007-10-15 17:00:06 +0200416 unsigned long delta_exec_weighted;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200417
Ingo Molnar8179ca232007-08-02 17:41:40 +0200418 schedstat_set(curr->exec_max, max((u64)delta_exec, curr->exec_max));
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200419
420 curr->sum_exec_runtime += delta_exec;
Ingo Molnar7a62eab2007-10-15 17:00:06 +0200421 schedstat_add(cfs_rq, exec_clock, delta_exec);
Ingo Molnarf9305d42008-05-29 11:23:17 +0200422 delta_exec_weighted = delta_exec;
423 if (unlikely(curr->load.weight != NICE_0_LOAD)) {
424 delta_exec_weighted = calc_delta_fair(delta_exec_weighted,
425 &curr->load);
426 }
Ingo Molnare9acbff2007-10-15 17:00:04 +0200427 curr->vruntime += delta_exec_weighted;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200428}
429
Ingo Molnarb7cc0892007-08-09 11:16:47 +0200430static void update_curr(struct cfs_rq *cfs_rq)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200431{
Ingo Molnar429d43b2007-10-15 17:00:03 +0200432 struct sched_entity *curr = cfs_rq->curr;
Ingo Molnar8ebc91d2007-10-15 17:00:03 +0200433 u64 now = rq_of(cfs_rq)->clock;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200434 unsigned long delta_exec;
435
436 if (unlikely(!curr))
437 return;
438
439 /*
440 * Get the amount of time the current task was running
441 * since the last time we changed load (this cannot
442 * overflow on 32 bits):
443 */
Ingo Molnar8ebc91d2007-10-15 17:00:03 +0200444 delta_exec = (unsigned long)(now - curr->exec_start);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200445
Ingo Molnar8ebc91d2007-10-15 17:00:03 +0200446 __update_curr(cfs_rq, curr, delta_exec);
447 curr->exec_start = now;
Srivatsa Vaddagirid842de82007-12-02 20:04:49 +0100448
449 if (entity_is_task(curr)) {
450 struct task_struct *curtask = task_of(curr);
451
452 cpuacct_charge(curtask, delta_exec);
453 }
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200454}
455
456static inline void
Ingo Molnar5870db52007-08-09 11:16:47 +0200457update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200458{
Ingo Molnard2819182007-08-09 11:16:47 +0200459 schedstat_set(se->wait_start, rq_of(cfs_rq)->clock);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200460}
461
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200462/*
463 * Task is being enqueued - update stats:
464 */
Ingo Molnard2417e52007-08-09 11:16:47 +0200465static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200466{
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200467 /*
468 * Are we enqueueing a waiting task? (for current tasks
469 * a dequeue/enqueue event is a NOP)
470 */
Ingo Molnar429d43b2007-10-15 17:00:03 +0200471 if (se != cfs_rq->curr)
Ingo Molnar5870db52007-08-09 11:16:47 +0200472 update_stats_wait_start(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200473}
474
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200475static void
Ingo Molnar9ef0a962007-08-09 11:16:47 +0200476update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200477{
Ingo Molnarbbdba7c2007-10-15 17:00:06 +0200478 schedstat_set(se->wait_max, max(se->wait_max,
479 rq_of(cfs_rq)->clock - se->wait_start));
Arjan van de Ven6d082592008-01-25 21:08:35 +0100480 schedstat_set(se->wait_count, se->wait_count + 1);
481 schedstat_set(se->wait_sum, se->wait_sum +
482 rq_of(cfs_rq)->clock - se->wait_start);
Ingo Molnar6cfb0d52007-08-02 17:41:40 +0200483 schedstat_set(se->wait_start, 0);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200484}
485
486static inline void
Ingo Molnar19b6a2e2007-08-09 11:16:48 +0200487update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200488{
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200489 /*
490 * Mark the end of the wait period if dequeueing a
491 * waiting task:
492 */
Ingo Molnar429d43b2007-10-15 17:00:03 +0200493 if (se != cfs_rq->curr)
Ingo Molnar9ef0a962007-08-09 11:16:47 +0200494 update_stats_wait_end(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200495}
496
497/*
498 * We are picking a new current task - update its stats:
499 */
500static inline void
Ingo Molnar79303e92007-08-09 11:16:47 +0200501update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200502{
503 /*
504 * We are starting a new run period:
505 */
Ingo Molnard2819182007-08-09 11:16:47 +0200506 se->exec_start = rq_of(cfs_rq)->clock;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200507}
508
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200509/**************************************************
510 * Scheduling class queueing methods:
511 */
512
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200513static void
514account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
515{
516 update_load_add(&cfs_rq->load, se->load.weight);
517 cfs_rq->nr_running++;
518 se->on_rq = 1;
Peter Zijlstra4a55bd52008-04-19 19:45:00 +0200519 list_add(&se->group_node, &cfs_rq->tasks);
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200520}
521
522static void
523account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
524{
525 update_load_sub(&cfs_rq->load, se->load.weight);
526 cfs_rq->nr_running--;
527 se->on_rq = 0;
Peter Zijlstra4a55bd52008-04-19 19:45:00 +0200528 list_del_init(&se->group_node);
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200529}
530
Ingo Molnar2396af62007-08-09 11:16:48 +0200531static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200532{
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200533#ifdef CONFIG_SCHEDSTATS
534 if (se->sleep_start) {
Ingo Molnard2819182007-08-09 11:16:47 +0200535 u64 delta = rq_of(cfs_rq)->clock - se->sleep_start;
Arjan van de Ven97455122008-01-25 21:08:34 +0100536 struct task_struct *tsk = task_of(se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200537
538 if ((s64)delta < 0)
539 delta = 0;
540
541 if (unlikely(delta > se->sleep_max))
542 se->sleep_max = delta;
543
544 se->sleep_start = 0;
545 se->sum_sleep_runtime += delta;
Arjan van de Ven97455122008-01-25 21:08:34 +0100546
547 account_scheduler_latency(tsk, delta >> 10, 1);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200548 }
549 if (se->block_start) {
Ingo Molnard2819182007-08-09 11:16:47 +0200550 u64 delta = rq_of(cfs_rq)->clock - se->block_start;
Arjan van de Ven97455122008-01-25 21:08:34 +0100551 struct task_struct *tsk = task_of(se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200552
553 if ((s64)delta < 0)
554 delta = 0;
555
556 if (unlikely(delta > se->block_max))
557 se->block_max = delta;
558
559 se->block_start = 0;
560 se->sum_sleep_runtime += delta;
Ingo Molnar30084fb2007-10-02 14:13:08 +0200561
562 /*
563 * Blocking time is in units of nanosecs, so shift by 20 to
564 * get a milliseconds-range estimation of the amount of
565 * time that the task spent sleeping:
566 */
567 if (unlikely(prof_on == SLEEP_PROFILING)) {
Ingo Molnare22f5bb2007-10-15 17:00:06 +0200568
Ingo Molnar30084fb2007-10-02 14:13:08 +0200569 profile_hits(SLEEP_PROFILING, (void *)get_wchan(tsk),
570 delta >> 20);
571 }
Arjan van de Ven97455122008-01-25 21:08:34 +0100572 account_scheduler_latency(tsk, delta >> 10, 0);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200573 }
574#endif
575}
576
Peter Zijlstraddc97292007-10-15 17:00:10 +0200577static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se)
578{
579#ifdef CONFIG_SCHED_DEBUG
580 s64 d = se->vruntime - cfs_rq->min_vruntime;
581
582 if (d < 0)
583 d = -d;
584
585 if (d > 3*sysctl_sched_latency)
586 schedstat_inc(cfs_rq, nr_spread_over);
587#endif
588}
589
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200590static void
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200591place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
592{
Peter Zijlstra67e9fb22007-10-15 17:00:10 +0200593 u64 vruntime;
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200594
Peter Zijlstra3fe69742008-03-14 20:55:51 +0100595 if (first_fair(cfs_rq)) {
596 vruntime = min_vruntime(cfs_rq->min_vruntime,
597 __pick_next_entity(cfs_rq)->vruntime);
598 } else
599 vruntime = cfs_rq->min_vruntime;
Peter Zijlstra94dfb5e2007-10-15 17:00:05 +0200600
Peter Zijlstra2cb86002007-11-09 22:39:37 +0100601 /*
602 * The 'current' period is already promised to the current tasks,
603 * however the extra weight of the new task will slow them down a
604 * little, place the new task so that it fits in the slot that
605 * stays open at the end.
606 */
Peter Zijlstra94dfb5e2007-10-15 17:00:05 +0200607 if (initial && sched_feat(START_DEBIT))
Ingo Molnar647e7ca2007-10-15 17:00:13 +0200608 vruntime += sched_vslice_add(cfs_rq, se);
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200609
Ingo Molnar8465e792007-10-15 17:00:11 +0200610 if (!initial) {
Peter Zijlstra2cb86002007-11-09 22:39:37 +0100611 /* sleeps upto a single latency don't count. */
Ingo Molnarf9305d42008-05-29 11:23:17 +0200612 if (sched_feat(NEW_FAIR_SLEEPERS))
613 vruntime -= sysctl_sched_latency;
Ingo Molnar94359f02007-10-15 17:00:11 +0200614
Peter Zijlstra2cb86002007-11-09 22:39:37 +0100615 /* ensure we never gain time by being placed backwards. */
616 vruntime = max_vruntime(se->vruntime, vruntime);
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200617 }
618
Peter Zijlstra67e9fb22007-10-15 17:00:10 +0200619 se->vruntime = vruntime;
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200620}
621
622static void
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +0200623enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200624{
625 /*
Dmitry Adamushkoa2a2d682007-10-15 17:00:13 +0200626 * Update run-time statistics of the 'current'.
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200627 */
Ingo Molnarb7cc0892007-08-09 11:16:47 +0200628 update_curr(cfs_rq);
Peter Zijlstraa9922412008-05-05 23:56:17 +0200629 account_entity_enqueue(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200630
Ingo Molnare9acbff2007-10-15 17:00:04 +0200631 if (wakeup) {
Peter Zijlstraaeb73b02007-10-15 17:00:05 +0200632 place_entity(cfs_rq, se, 0);
Ingo Molnar2396af62007-08-09 11:16:48 +0200633 enqueue_sleeper(cfs_rq, se);
Ingo Molnare9acbff2007-10-15 17:00:04 +0200634 }
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200635
Ingo Molnard2417e52007-08-09 11:16:47 +0200636 update_stats_enqueue(cfs_rq, se);
Peter Zijlstraddc97292007-10-15 17:00:10 +0200637 check_spread(cfs_rq, se);
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +0200638 if (se != cfs_rq->curr)
639 __enqueue_entity(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200640}
641
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +0100642static void update_avg(u64 *avg, u64 sample)
643{
644 s64 diff = sample - *avg;
645 *avg += diff >> 3;
646}
647
648static void update_avg_stats(struct cfs_rq *cfs_rq, struct sched_entity *se)
649{
650 if (!se->last_wakeup)
651 return;
652
653 update_avg(&se->avg_overlap, se->sum_exec_runtime - se->last_wakeup);
654 se->last_wakeup = 0;
655}
656
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200657static void
Ingo Molnar525c2712007-08-09 11:16:48 +0200658dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200659{
Dmitry Adamushkoa2a2d682007-10-15 17:00:13 +0200660 /*
661 * Update run-time statistics of the 'current'.
662 */
663 update_curr(cfs_rq);
664
Ingo Molnar19b6a2e2007-08-09 11:16:48 +0200665 update_stats_dequeue(cfs_rq, se);
Dmitry Adamushkodb36cc72007-10-15 17:00:06 +0200666 if (sleep) {
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +0100667 update_avg_stats(cfs_rq, se);
Peter Zijlstra67e9fb22007-10-15 17:00:10 +0200668#ifdef CONFIG_SCHEDSTATS
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200669 if (entity_is_task(se)) {
670 struct task_struct *tsk = task_of(se);
671
672 if (tsk->state & TASK_INTERRUPTIBLE)
Ingo Molnard2819182007-08-09 11:16:47 +0200673 se->sleep_start = rq_of(cfs_rq)->clock;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200674 if (tsk->state & TASK_UNINTERRUPTIBLE)
Ingo Molnard2819182007-08-09 11:16:47 +0200675 se->block_start = rq_of(cfs_rq)->clock;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200676 }
Dmitry Adamushkodb36cc72007-10-15 17:00:06 +0200677#endif
Peter Zijlstra67e9fb22007-10-15 17:00:10 +0200678 }
679
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +0200680 if (se != cfs_rq->curr)
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200681 __dequeue_entity(cfs_rq, se);
682 account_entity_dequeue(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200683}
684
685/*
686 * Preempt the current task with a newly woken task if needed:
687 */
Peter Zijlstra7c92e542007-09-05 14:32:49 +0200688static void
Ingo Molnar2e09bf52007-10-15 17:00:05 +0200689check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200690{
Peter Zijlstra11697832007-09-05 14:32:49 +0200691 unsigned long ideal_runtime, delta_exec;
692
Peter Zijlstra6d0f0eb2007-10-15 17:00:05 +0200693 ideal_runtime = sched_slice(cfs_rq, curr);
Peter Zijlstra11697832007-09-05 14:32:49 +0200694 delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
Ingo Molnar3e3e13f2007-11-09 22:39:39 +0100695 if (delta_exec > ideal_runtime)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200696 resched_task(rq_of(cfs_rq)->curr);
697}
698
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +0200699static void
Ingo Molnar8494f412007-08-09 11:16:48 +0200700set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200701{
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +0200702 /* 'current' is not kept within the tree. */
703 if (se->on_rq) {
704 /*
705 * Any task has to be enqueued before it get to execute on
706 * a CPU. So account for the time it spent waiting on the
707 * runqueue.
708 */
709 update_stats_wait_end(cfs_rq, se);
710 __dequeue_entity(cfs_rq, se);
711 }
712
Ingo Molnar79303e92007-08-09 11:16:47 +0200713 update_stats_curr_start(cfs_rq, se);
Ingo Molnar429d43b2007-10-15 17:00:03 +0200714 cfs_rq->curr = se;
Ingo Molnareba1ed42007-10-15 17:00:02 +0200715#ifdef CONFIG_SCHEDSTATS
716 /*
717 * Track our maximum slice length, if the CPU's load is at
718 * least twice that of our own weight (i.e. dont track it
719 * when there are only lesser-weight tasks around):
720 */
Dmitry Adamushko495eca42007-10-15 17:00:06 +0200721 if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
Ingo Molnareba1ed42007-10-15 17:00:02 +0200722 se->slice_max = max(se->slice_max,
723 se->sum_exec_runtime - se->prev_sum_exec_runtime);
724 }
725#endif
Peter Zijlstra4a55b452007-09-05 14:32:49 +0200726 se->prev_sum_exec_runtime = se->sum_exec_runtime;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200727}
728
Peter Zijlstra0bbd3332008-04-19 19:44:57 +0200729static int
730wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);
731
Peter Zijlstraaa2ac252008-03-14 21:12:12 +0100732static struct sched_entity *
733pick_next(struct cfs_rq *cfs_rq, struct sched_entity *se)
734{
Peter Zijlstraaa2ac252008-03-14 21:12:12 +0100735 if (!cfs_rq->next)
736 return se;
737
Peter Zijlstra0bbd3332008-04-19 19:44:57 +0200738 if (wakeup_preempt_entity(cfs_rq->next, se) != 0)
Peter Zijlstraaa2ac252008-03-14 21:12:12 +0100739 return se;
740
741 return cfs_rq->next;
742}
743
Ingo Molnar9948f4b2007-08-09 11:16:48 +0200744static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200745{
Dmitry Adamushko08ec3df2007-10-15 17:00:13 +0200746 struct sched_entity *se = NULL;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200747
Dmitry Adamushko08ec3df2007-10-15 17:00:13 +0200748 if (first_fair(cfs_rq)) {
749 se = __pick_next_entity(cfs_rq);
Peter Zijlstraaa2ac252008-03-14 21:12:12 +0100750 se = pick_next(cfs_rq, se);
Dmitry Adamushko08ec3df2007-10-15 17:00:13 +0200751 set_next_entity(cfs_rq, se);
752 }
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200753
754 return se;
755}
756
Ingo Molnarab6cde22007-08-09 11:16:48 +0200757static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200758{
759 /*
760 * If still on the runqueue then deactivate_task()
761 * was not called and update_curr() has to be done:
762 */
763 if (prev->on_rq)
Ingo Molnarb7cc0892007-08-09 11:16:47 +0200764 update_curr(cfs_rq);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200765
Peter Zijlstraddc97292007-10-15 17:00:10 +0200766 check_spread(cfs_rq, prev);
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200767 if (prev->on_rq) {
Ingo Molnar5870db52007-08-09 11:16:47 +0200768 update_stats_wait_start(cfs_rq, prev);
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200769 /* Put 'current' back into the tree. */
770 __enqueue_entity(cfs_rq, prev);
771 }
Ingo Molnar429d43b2007-10-15 17:00:03 +0200772 cfs_rq->curr = NULL;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200773}
774
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +0100775static void
776entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200777{
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200778 /*
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200779 * Update run-time statistics of the 'current'.
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200780 */
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200781 update_curr(cfs_rq);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200782
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +0100783#ifdef CONFIG_SCHED_HRTICK
784 /*
785 * queued ticks are scheduled to match the slice, so don't bother
786 * validating it and just reschedule.
787 */
Harvey Harrison983ed7a2008-04-24 18:17:55 -0700788 if (queued) {
789 resched_task(rq_of(cfs_rq)->curr);
790 return;
791 }
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +0100792 /*
793 * don't let the period tick interfere with the hrtick preemption
794 */
795 if (!sched_feat(DOUBLE_TICK) &&
796 hrtimer_active(&rq_of(cfs_rq)->hrtick_timer))
797 return;
798#endif
799
Peter Zijlstrace6c1312007-10-15 17:00:14 +0200800 if (cfs_rq->nr_running > 1 || !sched_feat(WAKEUP_PREEMPT))
Ingo Molnar2e09bf52007-10-15 17:00:05 +0200801 check_preempt_tick(cfs_rq, curr);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200802}
803
804/**************************************************
805 * CFS operations on tasks:
806 */
807
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +0100808#ifdef CONFIG_SCHED_HRTICK
809static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
810{
811 int requeue = rq->curr == p;
812 struct sched_entity *se = &p->se;
813 struct cfs_rq *cfs_rq = cfs_rq_of(se);
814
815 WARN_ON(task_rq(p) != rq);
816
817 if (hrtick_enabled(rq) && cfs_rq->nr_running > 1) {
818 u64 slice = sched_slice(cfs_rq, se);
819 u64 ran = se->sum_exec_runtime - se->prev_sum_exec_runtime;
820 s64 delta = slice - ran;
821
822 if (delta < 0) {
823 if (rq->curr == p)
824 resched_task(p);
825 return;
826 }
827
828 /*
829 * Don't schedule slices shorter than 10000ns, that just
830 * doesn't make sense. Rely on vruntime for fairness.
831 */
832 if (!requeue)
833 delta = max(10000LL, delta);
834
835 hrtick_start(rq, delta, requeue);
836 }
837}
838#else
839static inline void
840hrtick_start_fair(struct rq *rq, struct task_struct *p)
841{
842}
843#endif
844
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200845/*
846 * The enqueue_task method is called before nr_running is
847 * increased. Here we update the fair scheduling stats and
848 * then put the task into the rbtree:
849 */
Ingo Molnarfd390f62007-08-09 11:16:48 +0200850static void enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200851{
852 struct cfs_rq *cfs_rq;
Peter Zijlstra62fb1852008-02-25 17:34:02 +0100853 struct sched_entity *se = &p->se;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200854
855 for_each_sched_entity(se) {
Peter Zijlstra62fb1852008-02-25 17:34:02 +0100856 if (se->on_rq)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200857 break;
858 cfs_rq = cfs_rq_of(se);
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +0200859 enqueue_entity(cfs_rq, se, wakeup);
Srivatsa Vaddagirib9fa3df2007-10-15 17:00:12 +0200860 wakeup = 1;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200861 }
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +0100862
863 hrtick_start_fair(rq, rq->curr);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200864}
865
866/*
867 * The dequeue_task method is called before nr_running is
868 * decreased. We remove the task from the rbtree and
869 * update the fair scheduling stats:
870 */
Ingo Molnarf02231e2007-08-09 11:16:48 +0200871static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int sleep)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200872{
873 struct cfs_rq *cfs_rq;
Peter Zijlstra62fb1852008-02-25 17:34:02 +0100874 struct sched_entity *se = &p->se;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200875
876 for_each_sched_entity(se) {
877 cfs_rq = cfs_rq_of(se);
Ingo Molnar525c2712007-08-09 11:16:48 +0200878 dequeue_entity(cfs_rq, se, sleep);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200879 /* Don't dequeue parent if it has other entities besides us */
Peter Zijlstra62fb1852008-02-25 17:34:02 +0100880 if (cfs_rq->load.weight)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200881 break;
Srivatsa Vaddagirib9fa3df2007-10-15 17:00:12 +0200882 sleep = 1;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200883 }
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +0100884
885 hrtick_start_fair(rq, rq->curr);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200886}
887
888/*
Ingo Molnar1799e352007-09-19 23:34:46 +0200889 * sched_yield() support is very simple - we dequeue and enqueue.
890 *
891 * If compat_yield is turned on then we requeue to the end of the tree.
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200892 */
Dmitry Adamushko4530d7a2007-10-15 17:00:08 +0200893static void yield_task_fair(struct rq *rq)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200894{
Ingo Molnardb292ca2007-12-04 17:04:39 +0100895 struct task_struct *curr = rq->curr;
896 struct cfs_rq *cfs_rq = task_cfs_rq(curr);
897 struct sched_entity *rightmost, *se = &curr->se;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200898
899 /*
Ingo Molnar1799e352007-09-19 23:34:46 +0200900 * Are we the only task in the tree?
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200901 */
Ingo Molnar1799e352007-09-19 23:34:46 +0200902 if (unlikely(cfs_rq->nr_running == 1))
903 return;
904
Ingo Molnardb292ca2007-12-04 17:04:39 +0100905 if (likely(!sysctl_sched_compat_yield) && curr->policy != SCHED_BATCH) {
Peter Zijlstra3e51f332008-05-03 18:29:28 +0200906 update_rq_clock(rq);
Ingo Molnar1799e352007-09-19 23:34:46 +0200907 /*
Dmitry Adamushkoa2a2d682007-10-15 17:00:13 +0200908 * Update run-time statistics of the 'current'.
Ingo Molnar1799e352007-09-19 23:34:46 +0200909 */
Dmitry Adamushko2b1e3152007-10-15 17:00:12 +0200910 update_curr(cfs_rq);
Ingo Molnar1799e352007-09-19 23:34:46 +0200911
912 return;
913 }
914 /*
915 * Find the rightmost entry in the rbtree:
916 */
Dmitry Adamushko2b1e3152007-10-15 17:00:12 +0200917 rightmost = __pick_last_entity(cfs_rq);
Ingo Molnar1799e352007-09-19 23:34:46 +0200918 /*
919 * Already in the rightmost position?
920 */
Peter Zijlstra79b3fef2008-02-18 13:39:37 +0100921 if (unlikely(!rightmost || rightmost->vruntime < se->vruntime))
Ingo Molnar1799e352007-09-19 23:34:46 +0200922 return;
923
924 /*
925 * Minimally necessary key value to be last in the tree:
Dmitry Adamushko2b1e3152007-10-15 17:00:12 +0200926 * Upon rescheduling, sched_class::put_prev_task() will place
927 * 'current' within the tree based on its new key value.
Ingo Molnar1799e352007-09-19 23:34:46 +0200928 */
Dmitry Adamushko30cfdcf2007-10-15 17:00:07 +0200929 se->vruntime = rightmost->vruntime + 1;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +0200930}
931
932/*
Gregory Haskinse7693a32008-01-25 21:08:09 +0100933 * wake_idle() will wake a task on an idle cpu if task->cpu is
934 * not idle and an idle cpu is available. The span of cpus to
935 * search starts with cpus closest then further out as needed,
936 * so we always favor a closer, idle cpu.
937 *
938 * Returns the CPU we should wake onto.
939 */
940#if defined(ARCH_HAS_SCHED_WAKE_IDLE)
941static int wake_idle(int cpu, struct task_struct *p)
942{
943 cpumask_t tmp;
944 struct sched_domain *sd;
945 int i;
946
947 /*
948 * If it is idle, then it is the best cpu to run this task.
949 *
950 * This cpu is also the best, if it has more than one task already.
951 * Siblings must be also busy(in most cases) as they didn't already
952 * pickup the extra load from this cpu and hence we need not check
953 * sibling runqueue info. This will avoid the checks and cache miss
954 * penalities associated with that.
955 */
Gregory Haskins104f6452008-04-28 12:40:01 -0400956 if (idle_cpu(cpu) || cpu_rq(cpu)->cfs.nr_running > 1)
Gregory Haskinse7693a32008-01-25 21:08:09 +0100957 return cpu;
958
959 for_each_domain(cpu, sd) {
Hidetoshi Seto1d3504f2008-04-15 14:04:23 +0900960 if ((sd->flags & SD_WAKE_IDLE)
961 || ((sd->flags & SD_WAKE_IDLE_FAR)
962 && !task_hot(p, task_rq(p)->clock, sd))) {
Gregory Haskinse7693a32008-01-25 21:08:09 +0100963 cpus_and(tmp, sd->span, p->cpus_allowed);
964 for_each_cpu_mask(i, tmp) {
965 if (idle_cpu(i)) {
966 if (i != task_cpu(p)) {
967 schedstat_inc(p,
968 se.nr_wakeups_idle);
969 }
970 return i;
971 }
972 }
973 } else {
974 break;
975 }
976 }
977 return cpu;
978}
979#else
980static inline int wake_idle(int cpu, struct task_struct *p)
981{
982 return cpu;
983}
984#endif
985
986#ifdef CONFIG_SMP
Ingo Molnar098fb9d2008-03-16 20:36:10 +0100987
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +0100988static const struct sched_class fair_sched_class;
989
Ingo Molnar098fb9d2008-03-16 20:36:10 +0100990static int
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +0100991wake_affine(struct rq *rq, struct sched_domain *this_sd, struct rq *this_rq,
992 struct task_struct *p, int prev_cpu, int this_cpu, int sync,
993 int idx, unsigned long load, unsigned long this_load,
Ingo Molnar098fb9d2008-03-16 20:36:10 +0100994 unsigned int imbalance)
995{
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +0100996 struct task_struct *curr = this_rq->curr;
Ingo Molnar098fb9d2008-03-16 20:36:10 +0100997 unsigned long tl = this_load;
998 unsigned long tl_per_task;
Mike Galbraithb3137bc2008-05-29 11:11:41 +0200999 int balanced;
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001000
Mike Galbraithb3137bc2008-05-29 11:11:41 +02001001 if (!(this_sd->flags & SD_WAKE_AFFINE) || !sched_feat(AFFINE_WAKEUPS))
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001002 return 0;
1003
1004 /*
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001005 * If sync wakeup then subtract the (maximum possible)
1006 * effect of the currently running task from the load
1007 * of the current CPU:
1008 */
1009 if (sync)
1010 tl -= current->se.load.weight;
1011
Mike Galbraithb3137bc2008-05-29 11:11:41 +02001012 balanced = 100*(tl + p->se.load.weight) <= imbalance*load;
1013
1014 /*
1015 * If the currently running task will sleep within
1016 * a reasonable amount of time then attract this newly
1017 * woken task:
1018 */
1019 if (sync && balanced && curr->sched_class == &fair_sched_class) {
1020 if (curr->se.avg_overlap < sysctl_sched_migration_cost &&
1021 p->se.avg_overlap < sysctl_sched_migration_cost)
1022 return 1;
1023 }
1024
1025 schedstat_inc(p, se.nr_wakeups_affine_attempts);
1026 tl_per_task = cpu_avg_load_per_task(this_cpu);
1027
Ingo Molnarac192d32008-03-16 20:56:26 +01001028 if ((tl <= load && tl + target_load(prev_cpu, idx) <= tl_per_task) ||
Mike Galbraithb3137bc2008-05-29 11:11:41 +02001029 balanced) {
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001030 /*
1031 * This domain has SD_WAKE_AFFINE and
1032 * p is cache cold in this domain, and
1033 * there is no bad imbalance.
1034 */
1035 schedstat_inc(this_sd, ttwu_move_affine);
1036 schedstat_inc(p, se.nr_wakeups_affine);
1037
1038 return 1;
1039 }
1040 return 0;
1041}
1042
Gregory Haskinse7693a32008-01-25 21:08:09 +01001043static int select_task_rq_fair(struct task_struct *p, int sync)
1044{
Gregory Haskinse7693a32008-01-25 21:08:09 +01001045 struct sched_domain *sd, *this_sd = NULL;
Ingo Molnarac192d32008-03-16 20:56:26 +01001046 int prev_cpu, this_cpu, new_cpu;
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001047 unsigned long load, this_load;
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +01001048 struct rq *rq, *this_rq;
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001049 unsigned int imbalance;
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001050 int idx;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001051
Ingo Molnarac192d32008-03-16 20:56:26 +01001052 prev_cpu = task_cpu(p);
1053 rq = task_rq(p);
1054 this_cpu = smp_processor_id();
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +01001055 this_rq = cpu_rq(this_cpu);
Ingo Molnarac192d32008-03-16 20:56:26 +01001056 new_cpu = prev_cpu;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001057
Ingo Molnarac192d32008-03-16 20:56:26 +01001058 /*
1059 * 'this_sd' is the first domain that both
1060 * this_cpu and prev_cpu are present in:
1061 */
Gregory Haskinse7693a32008-01-25 21:08:09 +01001062 for_each_domain(this_cpu, sd) {
Ingo Molnarac192d32008-03-16 20:56:26 +01001063 if (cpu_isset(prev_cpu, sd->span)) {
Gregory Haskinse7693a32008-01-25 21:08:09 +01001064 this_sd = sd;
1065 break;
1066 }
1067 }
1068
1069 if (unlikely(!cpu_isset(this_cpu, p->cpus_allowed)))
Ingo Molnarf4827382008-03-16 21:21:47 +01001070 goto out;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001071
1072 /*
1073 * Check for affine wakeup and passive balancing possibilities.
1074 */
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001075 if (!this_sd)
Ingo Molnarf4827382008-03-16 21:21:47 +01001076 goto out;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001077
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001078 idx = this_sd->wake_idx;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001079
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001080 imbalance = 100 + (this_sd->imbalance_pct - 100) / 2;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001081
Ingo Molnarac192d32008-03-16 20:56:26 +01001082 load = source_load(prev_cpu, idx);
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001083 this_load = target_load(this_cpu, idx);
Gregory Haskinse7693a32008-01-25 21:08:09 +01001084
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +01001085 if (wake_affine(rq, this_sd, this_rq, p, prev_cpu, this_cpu, sync, idx,
1086 load, this_load, imbalance))
1087 return this_cpu;
1088
1089 if (prev_cpu == this_cpu)
Ingo Molnarf4827382008-03-16 21:21:47 +01001090 goto out;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001091
Ingo Molnar098fb9d2008-03-16 20:36:10 +01001092 /*
1093 * Start passive balancing when half the imbalance_pct
1094 * limit is reached.
1095 */
1096 if (this_sd->flags & SD_WAKE_BALANCE) {
1097 if (imbalance*this_load <= 100*load) {
1098 schedstat_inc(this_sd, ttwu_move_balance);
1099 schedstat_inc(p, se.nr_wakeups_passive);
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +01001100 return this_cpu;
Gregory Haskinse7693a32008-01-25 21:08:09 +01001101 }
1102 }
1103
Ingo Molnarf4827382008-03-16 21:21:47 +01001104out:
Gregory Haskinse7693a32008-01-25 21:08:09 +01001105 return wake_idle(new_cpu, p);
1106}
1107#endif /* CONFIG_SMP */
1108
Peter Zijlstra0bbd3332008-04-19 19:44:57 +02001109static unsigned long wakeup_gran(struct sched_entity *se)
1110{
1111 unsigned long gran = sysctl_sched_wakeup_granularity;
1112
1113 /*
Ingo Molnarf9305d42008-05-29 11:23:17 +02001114 * More easily preempt - nice tasks, while not making
1115 * it harder for + nice tasks.
Peter Zijlstra0bbd3332008-04-19 19:44:57 +02001116 */
Ingo Molnarf9305d42008-05-29 11:23:17 +02001117 if (unlikely(se->load.weight > NICE_0_LOAD))
1118 gran = calc_delta_fair(gran, &se->load);
Peter Zijlstra0bbd3332008-04-19 19:44:57 +02001119
1120 return gran;
1121}
1122
1123/*
1124 * Should 'se' preempt 'curr'.
1125 *
1126 * |s1
1127 * |s2
1128 * |s3
1129 * g
1130 * |<--->|c
1131 *
1132 * w(c, s1) = -1
1133 * w(c, s2) = 0
1134 * w(c, s3) = 1
1135 *
1136 */
1137static int
1138wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se)
1139{
1140 s64 gran, vdiff = curr->vruntime - se->vruntime;
1141
1142 if (vdiff < 0)
1143 return -1;
1144
1145 gran = wakeup_gran(curr);
1146 if (vdiff > gran)
1147 return 1;
1148
1149 return 0;
1150}
Gregory Haskinse7693a32008-01-25 21:08:09 +01001151
Dhaval Giani354d60c2008-04-19 19:44:59 +02001152/* return depth at which a sched entity is present in the hierarchy */
1153static inline int depth_se(struct sched_entity *se)
1154{
1155 int depth = 0;
1156
1157 for_each_sched_entity(se)
1158 depth++;
1159
1160 return depth;
1161}
1162
Gregory Haskinse7693a32008-01-25 21:08:09 +01001163/*
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001164 * Preempt the current task with a newly woken task if needed:
1165 */
Ingo Molnar2e09bf52007-10-15 17:00:05 +02001166static void check_preempt_wakeup(struct rq *rq, struct task_struct *p)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001167{
1168 struct task_struct *curr = rq->curr;
Srivatsa Vaddagirifad095a2007-10-15 17:00:12 +02001169 struct cfs_rq *cfs_rq = task_cfs_rq(curr);
Srivatsa Vaddagiri8651a862007-10-15 17:00:12 +02001170 struct sched_entity *se = &curr->se, *pse = &p->se;
Dhaval Giani354d60c2008-04-19 19:44:59 +02001171 int se_depth, pse_depth;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001172
1173 if (unlikely(rt_prio(p->prio))) {
Ingo Molnara8e504d2007-08-09 11:16:47 +02001174 update_rq_clock(rq);
Ingo Molnarb7cc0892007-08-09 11:16:47 +02001175 update_curr(cfs_rq);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001176 resched_task(curr);
1177 return;
1178 }
Peter Zijlstraaa2ac252008-03-14 21:12:12 +01001179
Ingo Molnar4ae7d5c2008-03-19 01:42:00 +01001180 se->last_wakeup = se->sum_exec_runtime;
1181 if (unlikely(se == pse))
1182 return;
1183
Peter Zijlstraaa2ac252008-03-14 21:12:12 +01001184 cfs_rq_of(pse)->next = pse;
1185
Ingo Molnar91c234b2007-10-15 17:00:18 +02001186 /*
1187 * Batch tasks do not preempt (their preemption is driven by
1188 * the tick):
1189 */
1190 if (unlikely(p->policy == SCHED_BATCH))
1191 return;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001192
Ingo Molnar77d9cc42007-11-09 22:39:39 +01001193 if (!sched_feat(WAKEUP_PREEMPT))
1194 return;
Peter Zijlstrace6c1312007-10-15 17:00:14 +02001195
Dhaval Giani354d60c2008-04-19 19:44:59 +02001196 /*
1197 * preemption test can be made between sibling entities who are in the
1198 * same cfs_rq i.e who have a common parent. Walk up the hierarchy of
1199 * both tasks until we find their ancestors who are siblings of common
1200 * parent.
1201 */
1202
1203 /* First walk up until both entities are at same depth */
1204 se_depth = depth_se(se);
1205 pse_depth = depth_se(pse);
1206
1207 while (se_depth > pse_depth) {
1208 se_depth--;
1209 se = parent_entity(se);
1210 }
1211
1212 while (pse_depth > se_depth) {
1213 pse_depth--;
1214 pse = parent_entity(pse);
1215 }
1216
Ingo Molnar77d9cc42007-11-09 22:39:39 +01001217 while (!is_same_group(se, pse)) {
1218 se = parent_entity(se);
1219 pse = parent_entity(pse);
Ingo Molnar2e09bf52007-10-15 17:00:05 +02001220 }
Ingo Molnar77d9cc42007-11-09 22:39:39 +01001221
Peter Zijlstra0bbd3332008-04-19 19:44:57 +02001222 if (wakeup_preempt_entity(se, pse) == 1)
Ingo Molnar77d9cc42007-11-09 22:39:39 +01001223 resched_task(curr);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001224}
1225
Ingo Molnarfb8d4722007-08-09 11:16:48 +02001226static struct task_struct *pick_next_task_fair(struct rq *rq)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001227{
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +01001228 struct task_struct *p;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001229 struct cfs_rq *cfs_rq = &rq->cfs;
1230 struct sched_entity *se;
1231
1232 if (unlikely(!cfs_rq->nr_running))
1233 return NULL;
1234
1235 do {
Ingo Molnar9948f4b2007-08-09 11:16:48 +02001236 se = pick_next_entity(cfs_rq);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001237 cfs_rq = group_cfs_rq(se);
1238 } while (cfs_rq);
1239
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +01001240 p = task_of(se);
1241 hrtick_start_fair(rq, p);
1242
1243 return p;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001244}
1245
1246/*
1247 * Account for a descheduled task:
1248 */
Ingo Molnar31ee5292007-08-09 11:16:49 +02001249static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001250{
1251 struct sched_entity *se = &prev->se;
1252 struct cfs_rq *cfs_rq;
1253
1254 for_each_sched_entity(se) {
1255 cfs_rq = cfs_rq_of(se);
Ingo Molnarab6cde22007-08-09 11:16:48 +02001256 put_prev_entity(cfs_rq, se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001257 }
1258}
1259
Peter Williams681f3e62007-10-24 18:23:51 +02001260#ifdef CONFIG_SMP
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001261/**************************************************
1262 * Fair scheduling class load-balancing methods:
1263 */
1264
1265/*
1266 * Load-balancing iterator. Note: while the runqueue stays locked
1267 * during the whole iteration, the current task might be
1268 * dequeued so the iterator has to be dequeue-safe. Here we
1269 * achieve that by always pre-iterating before returning
1270 * the current task:
1271 */
Alexey Dobriyana9957442007-10-15 17:00:13 +02001272static struct task_struct *
Peter Zijlstra4a55bd52008-04-19 19:45:00 +02001273__load_balance_iterator(struct cfs_rq *cfs_rq, struct list_head *next)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001274{
Dhaval Giani354d60c2008-04-19 19:44:59 +02001275 struct task_struct *p = NULL;
1276 struct sched_entity *se;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001277
Peter Zijlstra4a55bd52008-04-19 19:45:00 +02001278 if (next == &cfs_rq->tasks)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001279 return NULL;
1280
Dhaval Giani354d60c2008-04-19 19:44:59 +02001281 /* Skip over entities that are not tasks */
1282 do {
Peter Zijlstra4a55bd52008-04-19 19:45:00 +02001283 se = list_entry(next, struct sched_entity, group_node);
1284 next = next->next;
1285 } while (next != &cfs_rq->tasks && !entity_is_task(se));
Dhaval Giani354d60c2008-04-19 19:44:59 +02001286
Peter Zijlstra4a55bd52008-04-19 19:45:00 +02001287 if (next == &cfs_rq->tasks)
1288 return NULL;
1289
1290 cfs_rq->balance_iterator = next;
Dhaval Giani354d60c2008-04-19 19:44:59 +02001291
1292 if (entity_is_task(se))
1293 p = task_of(se);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001294
1295 return p;
1296}
1297
1298static struct task_struct *load_balance_start_fair(void *arg)
1299{
1300 struct cfs_rq *cfs_rq = arg;
1301
Peter Zijlstra4a55bd52008-04-19 19:45:00 +02001302 return __load_balance_iterator(cfs_rq, cfs_rq->tasks.next);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001303}
1304
1305static struct task_struct *load_balance_next_fair(void *arg)
1306{
1307 struct cfs_rq *cfs_rq = arg;
1308
Peter Zijlstra4a55bd52008-04-19 19:45:00 +02001309 return __load_balance_iterator(cfs_rq, cfs_rq->balance_iterator);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001310}
1311
Ingo Molnar6363ca52008-05-29 11:28:57 +02001312#ifdef CONFIG_FAIR_GROUP_SCHED
1313static int cfs_rq_best_prio(struct cfs_rq *cfs_rq)
Peter Zijlstra62fb1852008-02-25 17:34:02 +01001314{
Ingo Molnar6363ca52008-05-29 11:28:57 +02001315 struct sched_entity *curr;
1316 struct task_struct *p;
1317
1318 if (!cfs_rq->nr_running || !first_fair(cfs_rq))
1319 return MAX_PRIO;
1320
1321 curr = cfs_rq->curr;
1322 if (!curr)
1323 curr = __pick_next_entity(cfs_rq);
1324
1325 p = task_of(curr);
1326
1327 return p->prio;
1328}
1329#endif
1330
1331static unsigned long
1332load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
1333 unsigned long max_load_move,
1334 struct sched_domain *sd, enum cpu_idle_type idle,
1335 int *all_pinned, int *this_best_prio)
1336{
1337 struct cfs_rq *busy_cfs_rq;
1338 long rem_load_move = max_load_move;
Peter Zijlstra18d95a22008-04-19 19:45:00 +02001339 struct rq_iterator cfs_rq_iterator;
Peter Zijlstra62fb1852008-02-25 17:34:02 +01001340
Peter Zijlstra18d95a22008-04-19 19:45:00 +02001341 cfs_rq_iterator.start = load_balance_start_fair;
1342 cfs_rq_iterator.next = load_balance_next_fair;
Peter Zijlstra62fb1852008-02-25 17:34:02 +01001343
Ingo Molnar6363ca52008-05-29 11:28:57 +02001344 for_each_leaf_cfs_rq(busiest, busy_cfs_rq) {
Peter Zijlstra18d95a22008-04-19 19:45:00 +02001345#ifdef CONFIG_FAIR_GROUP_SCHED
Ingo Molnar6363ca52008-05-29 11:28:57 +02001346 struct cfs_rq *this_cfs_rq;
Peter Zijlstra62fb1852008-02-25 17:34:02 +01001347 long imbalance;
Ingo Molnar6363ca52008-05-29 11:28:57 +02001348 unsigned long maxload;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001349
Ingo Molnar6363ca52008-05-29 11:28:57 +02001350 this_cfs_rq = cpu_cfs_rq(busy_cfs_rq, this_cpu);
1351
1352 imbalance = busy_cfs_rq->load.weight - this_cfs_rq->load.weight;
1353 /* Don't pull if this_cfs_rq has more load than busy_cfs_rq */
1354 if (imbalance <= 0)
1355 continue;
1356
1357 /* Don't pull more than imbalance/2 */
1358 imbalance /= 2;
1359 maxload = min(rem_load_move, imbalance);
1360
1361 *this_best_prio = cfs_rq_best_prio(this_cfs_rq);
1362#else
1363# define maxload rem_load_move
1364#endif
Peter Zijlstra18d95a22008-04-19 19:45:00 +02001365 /*
Ingo Molnar6363ca52008-05-29 11:28:57 +02001366 * pass busy_cfs_rq argument into
1367 * load_balance_[start|next]_fair iterators
Peter Zijlstra18d95a22008-04-19 19:45:00 +02001368 */
Ingo Molnar6363ca52008-05-29 11:28:57 +02001369 cfs_rq_iterator.arg = busy_cfs_rq;
1370 rem_load_move -= balance_tasks(this_rq, this_cpu, busiest,
1371 maxload, sd, idle, all_pinned,
1372 this_best_prio,
1373 &cfs_rq_iterator);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001374
Ingo Molnar6363ca52008-05-29 11:28:57 +02001375 if (rem_load_move <= 0)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001376 break;
1377 }
1378
Peter Williams43010652007-08-09 11:16:46 +02001379 return max_load_move - rem_load_move;
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001380}
1381
Peter Williamse1d14842007-10-24 18:23:51 +02001382static int
1383move_one_task_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
1384 struct sched_domain *sd, enum cpu_idle_type idle)
1385{
1386 struct cfs_rq *busy_cfs_rq;
1387 struct rq_iterator cfs_rq_iterator;
1388
1389 cfs_rq_iterator.start = load_balance_start_fair;
1390 cfs_rq_iterator.next = load_balance_next_fair;
1391
1392 for_each_leaf_cfs_rq(busiest, busy_cfs_rq) {
1393 /*
1394 * pass busy_cfs_rq argument into
1395 * load_balance_[start|next]_fair iterators
1396 */
1397 cfs_rq_iterator.arg = busy_cfs_rq;
1398 if (iter_move_one_task(this_rq, this_cpu, busiest, sd, idle,
1399 &cfs_rq_iterator))
1400 return 1;
1401 }
1402
1403 return 0;
1404}
Peter Williams681f3e62007-10-24 18:23:51 +02001405#endif
Peter Williamse1d14842007-10-24 18:23:51 +02001406
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001407/*
1408 * scheduler tick hitting a task of our scheduling class:
1409 */
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +01001410static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001411{
1412 struct cfs_rq *cfs_rq;
1413 struct sched_entity *se = &curr->se;
1414
1415 for_each_sched_entity(se) {
1416 cfs_rq = cfs_rq_of(se);
Peter Zijlstra8f4d37e2008-01-25 21:08:29 +01001417 entity_tick(cfs_rq, se, queued);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001418 }
1419}
1420
Ingo Molnar8eb172d2007-10-29 21:18:11 +01001421#define swap(a, b) do { typeof(a) tmp = (a); (a) = (b); (b) = tmp; } while (0)
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +02001422
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001423/*
1424 * Share the fairness runtime between parent and child, thus the
1425 * total amount of pressure for CPU stays equal - new tasks
1426 * get a chance to run but frequent forkers are not allowed to
1427 * monopolize the CPU. Note: the parent runqueue is locked,
1428 * the child is not running yet.
1429 */
Ingo Molnaree0827d2007-08-09 11:16:49 +02001430static void task_new_fair(struct rq *rq, struct task_struct *p)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001431{
1432 struct cfs_rq *cfs_rq = task_cfs_rq(p);
Ingo Molnar429d43b2007-10-15 17:00:03 +02001433 struct sched_entity *se = &p->se, *curr = cfs_rq->curr;
Ingo Molnar00bf7bf2007-10-15 17:00:14 +02001434 int this_cpu = smp_processor_id();
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001435
1436 sched_info_queued(p);
1437
Ting Yang7109c4422007-08-28 12:53:24 +02001438 update_curr(cfs_rq);
Peter Zijlstraaeb73b02007-10-15 17:00:05 +02001439 place_entity(cfs_rq, se, 1);
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +02001440
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001441 /* 'curr' will be NULL if the child belongs to a different group */
Ingo Molnar00bf7bf2007-10-15 17:00:14 +02001442 if (sysctl_sched_child_runs_first && this_cpu == task_cpu(p) &&
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001443 curr && curr->vruntime < se->vruntime) {
Dmitry Adamushko87fefa32007-10-15 17:00:08 +02001444 /*
Ingo Molnaredcb60a2007-10-15 17:00:08 +02001445 * Upon rescheduling, sched_class::put_prev_task() will place
1446 * 'current' within the tree based on its new key value.
1447 */
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +02001448 swap(curr->vruntime, se->vruntime);
Peter Zijlstra4d78e7b2007-10-15 17:00:04 +02001449 }
1450
Srivatsa Vaddagirib9dca1e2007-10-17 16:55:11 +02001451 enqueue_task_fair(rq, p, 0);
Ingo Molnarbb61c212007-10-15 17:00:02 +02001452 resched_task(rq->curr);
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001453}
1454
Steven Rostedtcb469842008-01-25 21:08:22 +01001455/*
1456 * Priority of the task has changed. Check to see if we preempt
1457 * the current task.
1458 */
1459static void prio_changed_fair(struct rq *rq, struct task_struct *p,
1460 int oldprio, int running)
1461{
1462 /*
1463 * Reschedule if we are currently running on this runqueue and
1464 * our priority decreased, or if we are not currently running on
1465 * this runqueue and our priority is higher than the current's
1466 */
1467 if (running) {
1468 if (p->prio > oldprio)
1469 resched_task(rq->curr);
1470 } else
1471 check_preempt_curr(rq, p);
1472}
1473
1474/*
1475 * We switched to the sched_fair class.
1476 */
1477static void switched_to_fair(struct rq *rq, struct task_struct *p,
1478 int running)
1479{
1480 /*
1481 * We were most likely switched from sched_rt, so
1482 * kick off the schedule if running, otherwise just see
1483 * if we can still preempt the current task.
1484 */
1485 if (running)
1486 resched_task(rq->curr);
1487 else
1488 check_preempt_curr(rq, p);
1489}
1490
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +02001491/* Account for a task changing its policy or group.
1492 *
1493 * This routine is mostly called to set cfs_rq->curr field when a task
1494 * migrates between groups/classes.
1495 */
1496static void set_curr_task_fair(struct rq *rq)
1497{
1498 struct sched_entity *se = &rq->curr->se;
1499
1500 for_each_sched_entity(se)
1501 set_next_entity(cfs_rq_of(se), se);
1502}
1503
Peter Zijlstra810b3812008-02-29 15:21:01 -05001504#ifdef CONFIG_FAIR_GROUP_SCHED
1505static void moved_group_fair(struct task_struct *p)
1506{
1507 struct cfs_rq *cfs_rq = task_cfs_rq(p);
1508
1509 update_curr(cfs_rq);
1510 place_entity(cfs_rq, &p->se, 1);
1511}
1512#endif
1513
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001514/*
1515 * All the scheduling class methods:
1516 */
Ingo Molnar5522d5d2007-10-15 17:00:12 +02001517static const struct sched_class fair_sched_class = {
1518 .next = &idle_sched_class,
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001519 .enqueue_task = enqueue_task_fair,
1520 .dequeue_task = dequeue_task_fair,
1521 .yield_task = yield_task_fair,
Gregory Haskinse7693a32008-01-25 21:08:09 +01001522#ifdef CONFIG_SMP
1523 .select_task_rq = select_task_rq_fair,
1524#endif /* CONFIG_SMP */
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001525
Ingo Molnar2e09bf52007-10-15 17:00:05 +02001526 .check_preempt_curr = check_preempt_wakeup,
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001527
1528 .pick_next_task = pick_next_task_fair,
1529 .put_prev_task = put_prev_task_fair,
1530
Peter Williams681f3e62007-10-24 18:23:51 +02001531#ifdef CONFIG_SMP
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001532 .load_balance = load_balance_fair,
Peter Williamse1d14842007-10-24 18:23:51 +02001533 .move_one_task = move_one_task_fair,
Peter Williams681f3e62007-10-24 18:23:51 +02001534#endif
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001535
Srivatsa Vaddagiri83b699e2007-10-15 17:00:08 +02001536 .set_curr_task = set_curr_task_fair,
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001537 .task_tick = task_tick_fair,
1538 .task_new = task_new_fair,
Steven Rostedtcb469842008-01-25 21:08:22 +01001539
1540 .prio_changed = prio_changed_fair,
1541 .switched_to = switched_to_fair,
Peter Zijlstra810b3812008-02-29 15:21:01 -05001542
1543#ifdef CONFIG_FAIR_GROUP_SCHED
1544 .moved_group = moved_group_fair,
1545#endif
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001546};
1547
1548#ifdef CONFIG_SCHED_DEBUG
Ingo Molnar5cef9ec2007-08-09 11:16:47 +02001549static void print_cfs_stats(struct seq_file *m, int cpu)
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001550{
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001551 struct cfs_rq *cfs_rq;
1552
Peter Zijlstra5973e5b2008-01-25 21:08:34 +01001553 rcu_read_lock();
Ingo Molnarc3b64f12007-08-09 11:16:51 +02001554 for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
Ingo Molnar5cef9ec2007-08-09 11:16:47 +02001555 print_cfs_rq(m, cpu, cfs_rq);
Peter Zijlstra5973e5b2008-01-25 21:08:34 +01001556 rcu_read_unlock();
Ingo Molnarbf0f6f22007-07-09 18:51:58 +02001557}
1558#endif