Commit | Line | Data |
---|---|---|
bf0f6f24 IM |
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> | |
21805085 PZ |
18 | * |
19 | * Adaptive scheduling granularity, math enhancements by Peter Zijlstra | |
20 | * Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com> | |
bf0f6f24 IM |
21 | */ |
22 | ||
9745512c | 23 | #include <linux/latencytop.h> |
1983a922 | 24 | #include <linux/sched.h> |
3436ae12 | 25 | #include <linux/cpumask.h> |
029632fb PZ |
26 | #include <linux/slab.h> |
27 | #include <linux/profile.h> | |
28 | #include <linux/interrupt.h> | |
cbee9f88 | 29 | #include <linux/mempolicy.h> |
e14808b4 | 30 | #include <linux/migrate.h> |
cbee9f88 | 31 | #include <linux/task_work.h> |
029632fb PZ |
32 | |
33 | #include <trace/events/sched.h> | |
34 | ||
35 | #include "sched.h" | |
9745512c | 36 | |
bf0f6f24 | 37 | /* |
21805085 | 38 | * Targeted preemption latency for CPU-bound tasks: |
864616ee | 39 | * (default: 6ms * (1 + ilog(ncpus)), units: nanoseconds) |
bf0f6f24 | 40 | * |
21805085 | 41 | * NOTE: this latency value is not the same as the concept of |
d274a4ce IM |
42 | * 'timeslice length' - timeslices in CFS are of variable length |
43 | * and have no persistent notion like in traditional, time-slice | |
44 | * based scheduling concepts. | |
bf0f6f24 | 45 | * |
d274a4ce IM |
46 | * (to see the precise effective timeslice length of your workload, |
47 | * run vmstat and monitor the context-switches (cs) field) | |
bf0f6f24 | 48 | */ |
21406928 MG |
49 | unsigned int sysctl_sched_latency = 6000000ULL; |
50 | unsigned int normalized_sysctl_sched_latency = 6000000ULL; | |
2bd8e6d4 | 51 | |
1983a922 CE |
52 | /* |
53 | * The initial- and re-scaling of tunables is configurable | |
54 | * (default SCHED_TUNABLESCALING_LOG = *(1+ilog(ncpus)) | |
55 | * | |
56 | * Options are: | |
57 | * SCHED_TUNABLESCALING_NONE - unscaled, always *1 | |
58 | * SCHED_TUNABLESCALING_LOG - scaled logarithmical, *1+ilog(ncpus) | |
59 | * SCHED_TUNABLESCALING_LINEAR - scaled linear, *ncpus | |
60 | */ | |
61 | enum sched_tunable_scaling sysctl_sched_tunable_scaling | |
62 | = SCHED_TUNABLESCALING_LOG; | |
63 | ||
2bd8e6d4 | 64 | /* |
b2be5e96 | 65 | * Minimal preemption granularity for CPU-bound tasks: |
864616ee | 66 | * (default: 0.75 msec * (1 + ilog(ncpus)), units: nanoseconds) |
2bd8e6d4 | 67 | */ |
0bf377bb IM |
68 | unsigned int sysctl_sched_min_granularity = 750000ULL; |
69 | unsigned int normalized_sysctl_sched_min_granularity = 750000ULL; | |
21805085 PZ |
70 | |
71 | /* | |
b2be5e96 PZ |
72 | * is kept at sysctl_sched_latency / sysctl_sched_min_granularity |
73 | */ | |
0bf377bb | 74 | static unsigned int sched_nr_latency = 8; |
b2be5e96 PZ |
75 | |
76 | /* | |
2bba22c5 | 77 | * After fork, child runs first. If set to 0 (default) then |
b2be5e96 | 78 | * parent will (try to) run first. |
21805085 | 79 | */ |
2bba22c5 | 80 | unsigned int sysctl_sched_child_runs_first __read_mostly; |
bf0f6f24 | 81 | |
bf0f6f24 IM |
82 | /* |
83 | * SCHED_OTHER wake-up granularity. | |
172e082a | 84 | * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds) |
bf0f6f24 IM |
85 | * |
86 | * This option delays the preemption effects of decoupled workloads | |
87 | * and reduces their over-scheduling. Synchronous workloads will still | |
88 | * have immediate wakeup/sleep latencies. | |
89 | */ | |
172e082a | 90 | unsigned int sysctl_sched_wakeup_granularity = 1000000UL; |
0bcdcf28 | 91 | unsigned int normalized_sysctl_sched_wakeup_granularity = 1000000UL; |
bf0f6f24 | 92 | |
da84d961 IM |
93 | const_debug unsigned int sysctl_sched_migration_cost = 500000UL; |
94 | ||
a7a4f8a7 PT |
95 | /* |
96 | * The exponential sliding window over which load is averaged for shares | |
97 | * distribution. | |
98 | * (default: 10msec) | |
99 | */ | |
100 | unsigned int __read_mostly sysctl_sched_shares_window = 10000000UL; | |
101 | ||
ec12cb7f PT |
102 | #ifdef CONFIG_CFS_BANDWIDTH |
103 | /* | |
104 | * Amount of runtime to allocate from global (tg) to local (per-cfs_rq) pool | |
105 | * each time a cfs_rq requests quota. | |
106 | * | |
107 | * Note: in the case that the slice exceeds the runtime remaining (either due | |
108 | * to consumption or the quota being specified to be smaller than the slice) | |
109 | * we will always only issue the remaining available time. | |
110 | * | |
111 | * default: 5 msec, units: microseconds | |
112 | */ | |
113 | unsigned int sysctl_sched_cfs_bandwidth_slice = 5000UL; | |
114 | #endif | |
115 | ||
029632fb PZ |
116 | /* |
117 | * Increase the granularity value when there are more CPUs, | |
118 | * because with more CPUs the 'effective latency' as visible | |
119 | * to users decreases. But the relationship is not linear, | |
120 | * so pick a second-best guess by going with the log2 of the | |
121 | * number of CPUs. | |
122 | * | |
123 | * This idea comes from the SD scheduler of Con Kolivas: | |
124 | */ | |
125 | static int get_update_sysctl_factor(void) | |
126 | { | |
127 | unsigned int cpus = min_t(int, num_online_cpus(), 8); | |
128 | unsigned int factor; | |
129 | ||
130 | switch (sysctl_sched_tunable_scaling) { | |
131 | case SCHED_TUNABLESCALING_NONE: | |
132 | factor = 1; | |
133 | break; | |
134 | case SCHED_TUNABLESCALING_LINEAR: | |
135 | factor = cpus; | |
136 | break; | |
137 | case SCHED_TUNABLESCALING_LOG: | |
138 | default: | |
139 | factor = 1 + ilog2(cpus); | |
140 | break; | |
141 | } | |
142 | ||
143 | return factor; | |
144 | } | |
145 | ||
146 | static void update_sysctl(void) | |
147 | { | |
148 | unsigned int factor = get_update_sysctl_factor(); | |
149 | ||
150 | #define SET_SYSCTL(name) \ | |
151 | (sysctl_##name = (factor) * normalized_sysctl_##name) | |
152 | SET_SYSCTL(sched_min_granularity); | |
153 | SET_SYSCTL(sched_latency); | |
154 | SET_SYSCTL(sched_wakeup_granularity); | |
155 | #undef SET_SYSCTL | |
156 | } | |
157 | ||
158 | void sched_init_granularity(void) | |
159 | { | |
160 | update_sysctl(); | |
161 | } | |
162 | ||
163 | #if BITS_PER_LONG == 32 | |
164 | # define WMULT_CONST (~0UL) | |
165 | #else | |
166 | # define WMULT_CONST (1UL << 32) | |
167 | #endif | |
168 | ||
169 | #define WMULT_SHIFT 32 | |
170 | ||
171 | /* | |
172 | * Shift right and round: | |
173 | */ | |
174 | #define SRR(x, y) (((x) + (1UL << ((y) - 1))) >> (y)) | |
175 | ||
176 | /* | |
177 | * delta *= weight / lw | |
178 | */ | |
179 | static unsigned long | |
180 | calc_delta_mine(unsigned long delta_exec, unsigned long weight, | |
181 | struct load_weight *lw) | |
182 | { | |
183 | u64 tmp; | |
184 | ||
185 | /* | |
186 | * weight can be less than 2^SCHED_LOAD_RESOLUTION for task group sched | |
187 | * entities since MIN_SHARES = 2. Treat weight as 1 if less than | |
188 | * 2^SCHED_LOAD_RESOLUTION. | |
189 | */ | |
190 | if (likely(weight > (1UL << SCHED_LOAD_RESOLUTION))) | |
191 | tmp = (u64)delta_exec * scale_load_down(weight); | |
192 | else | |
193 | tmp = (u64)delta_exec; | |
194 | ||
195 | if (!lw->inv_weight) { | |
196 | unsigned long w = scale_load_down(lw->weight); | |
197 | ||
198 | if (BITS_PER_LONG > 32 && unlikely(w >= WMULT_CONST)) | |
199 | lw->inv_weight = 1; | |
200 | else if (unlikely(!w)) | |
201 | lw->inv_weight = WMULT_CONST; | |
202 | else | |
203 | lw->inv_weight = WMULT_CONST / w; | |
204 | } | |
205 | ||
206 | /* | |
207 | * Check whether we'd overflow the 64-bit multiplication: | |
208 | */ | |
209 | if (unlikely(tmp > WMULT_CONST)) | |
210 | tmp = SRR(SRR(tmp, WMULT_SHIFT/2) * lw->inv_weight, | |
211 | WMULT_SHIFT/2); | |
212 | else | |
213 | tmp = SRR(tmp * lw->inv_weight, WMULT_SHIFT); | |
214 | ||
215 | return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX); | |
216 | } | |
217 | ||
218 | ||
219 | const struct sched_class fair_sched_class; | |
a4c2f00f | 220 | |
bf0f6f24 IM |
221 | /************************************************************** |
222 | * CFS operations on generic schedulable entities: | |
223 | */ | |
224 | ||
62160e3f | 225 | #ifdef CONFIG_FAIR_GROUP_SCHED |
bf0f6f24 | 226 | |
62160e3f | 227 | /* cpu runqueue to which this cfs_rq is attached */ |
bf0f6f24 IM |
228 | static inline struct rq *rq_of(struct cfs_rq *cfs_rq) |
229 | { | |
62160e3f | 230 | return cfs_rq->rq; |
bf0f6f24 IM |
231 | } |
232 | ||
62160e3f IM |
233 | /* An entity is a task if it doesn't "own" a runqueue */ |
234 | #define entity_is_task(se) (!se->my_q) | |
bf0f6f24 | 235 | |
8f48894f PZ |
236 | static inline struct task_struct *task_of(struct sched_entity *se) |
237 | { | |
238 | #ifdef CONFIG_SCHED_DEBUG | |
239 | WARN_ON_ONCE(!entity_is_task(se)); | |
240 | #endif | |
241 | return container_of(se, struct task_struct, se); | |
242 | } | |
243 | ||
b758149c PZ |
244 | /* Walk up scheduling entities hierarchy */ |
245 | #define for_each_sched_entity(se) \ | |
246 | for (; se; se = se->parent) | |
247 | ||
248 | static inline struct cfs_rq *task_cfs_rq(struct task_struct *p) | |
249 | { | |
250 | return p->se.cfs_rq; | |
251 | } | |
252 | ||
253 | /* runqueue on which this entity is (to be) queued */ | |
254 | static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se) | |
255 | { | |
256 | return se->cfs_rq; | |
257 | } | |
258 | ||
259 | /* runqueue "owned" by this group */ | |
260 | static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp) | |
261 | { | |
262 | return grp->my_q; | |
263 | } | |
264 | ||
aff3e498 PT |
265 | static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, |
266 | int force_update); | |
9ee474f5 | 267 | |
3d4b47b4 PZ |
268 | static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq) |
269 | { | |
270 | if (!cfs_rq->on_list) { | |
67e86250 PT |
271 | /* |
272 | * Ensure we either appear before our parent (if already | |
273 | * enqueued) or force our parent to appear after us when it is | |
274 | * enqueued. The fact that we always enqueue bottom-up | |
275 | * reduces this to two cases. | |
276 | */ | |
277 | if (cfs_rq->tg->parent && | |
278 | cfs_rq->tg->parent->cfs_rq[cpu_of(rq_of(cfs_rq))]->on_list) { | |
279 | list_add_rcu(&cfs_rq->leaf_cfs_rq_list, | |
280 | &rq_of(cfs_rq)->leaf_cfs_rq_list); | |
281 | } else { | |
282 | list_add_tail_rcu(&cfs_rq->leaf_cfs_rq_list, | |
3d4b47b4 | 283 | &rq_of(cfs_rq)->leaf_cfs_rq_list); |
67e86250 | 284 | } |
3d4b47b4 PZ |
285 | |
286 | cfs_rq->on_list = 1; | |
9ee474f5 | 287 | /* We should have no load, but we need to update last_decay. */ |
aff3e498 | 288 | update_cfs_rq_blocked_load(cfs_rq, 0); |
3d4b47b4 PZ |
289 | } |
290 | } | |
291 | ||
292 | static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq) | |
293 | { | |
294 | if (cfs_rq->on_list) { | |
295 | list_del_rcu(&cfs_rq->leaf_cfs_rq_list); | |
296 | cfs_rq->on_list = 0; | |
297 | } | |
298 | } | |
299 | ||
b758149c PZ |
300 | /* Iterate thr' all leaf cfs_rq's on a runqueue */ |
301 | #define for_each_leaf_cfs_rq(rq, cfs_rq) \ | |
302 | list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list) | |
303 | ||
304 | /* Do the two (enqueued) entities belong to the same group ? */ | |
305 | static inline int | |
306 | is_same_group(struct sched_entity *se, struct sched_entity *pse) | |
307 | { | |
308 | if (se->cfs_rq == pse->cfs_rq) | |
309 | return 1; | |
310 | ||
311 | return 0; | |
312 | } | |
313 | ||
314 | static inline struct sched_entity *parent_entity(struct sched_entity *se) | |
315 | { | |
316 | return se->parent; | |
317 | } | |
318 | ||
464b7527 PZ |
319 | /* return depth at which a sched entity is present in the hierarchy */ |
320 | static inline int depth_se(struct sched_entity *se) | |
321 | { | |
322 | int depth = 0; | |
323 | ||
324 | for_each_sched_entity(se) | |
325 | depth++; | |
326 | ||
327 | return depth; | |
328 | } | |
329 | ||
330 | static void | |
331 | find_matching_se(struct sched_entity **se, struct sched_entity **pse) | |
332 | { | |
333 | int se_depth, pse_depth; | |
334 | ||
335 | /* | |
336 | * preemption test can be made between sibling entities who are in the | |
337 | * same cfs_rq i.e who have a common parent. Walk up the hierarchy of | |
338 | * both tasks until we find their ancestors who are siblings of common | |
339 | * parent. | |
340 | */ | |
341 | ||
342 | /* First walk up until both entities are at same depth */ | |
343 | se_depth = depth_se(*se); | |
344 | pse_depth = depth_se(*pse); | |
345 | ||
346 | while (se_depth > pse_depth) { | |
347 | se_depth--; | |
348 | *se = parent_entity(*se); | |
349 | } | |
350 | ||
351 | while (pse_depth > se_depth) { | |
352 | pse_depth--; | |
353 | *pse = parent_entity(*pse); | |
354 | } | |
355 | ||
356 | while (!is_same_group(*se, *pse)) { | |
357 | *se = parent_entity(*se); | |
358 | *pse = parent_entity(*pse); | |
359 | } | |
360 | } | |
361 | ||
8f48894f PZ |
362 | #else /* !CONFIG_FAIR_GROUP_SCHED */ |
363 | ||
364 | static inline struct task_struct *task_of(struct sched_entity *se) | |
365 | { | |
366 | return container_of(se, struct task_struct, se); | |
367 | } | |
bf0f6f24 | 368 | |
62160e3f IM |
369 | static inline struct rq *rq_of(struct cfs_rq *cfs_rq) |
370 | { | |
371 | return container_of(cfs_rq, struct rq, cfs); | |
bf0f6f24 IM |
372 | } |
373 | ||
374 | #define entity_is_task(se) 1 | |
375 | ||
b758149c PZ |
376 | #define for_each_sched_entity(se) \ |
377 | for (; se; se = NULL) | |
bf0f6f24 | 378 | |
b758149c | 379 | static inline struct cfs_rq *task_cfs_rq(struct task_struct *p) |
bf0f6f24 | 380 | { |
b758149c | 381 | return &task_rq(p)->cfs; |
bf0f6f24 IM |
382 | } |
383 | ||
b758149c PZ |
384 | static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se) |
385 | { | |
386 | struct task_struct *p = task_of(se); | |
387 | struct rq *rq = task_rq(p); | |
388 | ||
389 | return &rq->cfs; | |
390 | } | |
391 | ||
392 | /* runqueue "owned" by this group */ | |
393 | static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp) | |
394 | { | |
395 | return NULL; | |
396 | } | |
397 | ||
3d4b47b4 PZ |
398 | static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq) |
399 | { | |
400 | } | |
401 | ||
402 | static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq) | |
403 | { | |
404 | } | |
405 | ||
b758149c PZ |
406 | #define for_each_leaf_cfs_rq(rq, cfs_rq) \ |
407 | for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL) | |
408 | ||
409 | static inline int | |
410 | is_same_group(struct sched_entity *se, struct sched_entity *pse) | |
411 | { | |
412 | return 1; | |
413 | } | |
414 | ||
415 | static inline struct sched_entity *parent_entity(struct sched_entity *se) | |
416 | { | |
417 | return NULL; | |
418 | } | |
419 | ||
464b7527 PZ |
420 | static inline void |
421 | find_matching_se(struct sched_entity **se, struct sched_entity **pse) | |
422 | { | |
423 | } | |
424 | ||
b758149c PZ |
425 | #endif /* CONFIG_FAIR_GROUP_SCHED */ |
426 | ||
6c16a6dc PZ |
427 | static __always_inline |
428 | void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec); | |
bf0f6f24 IM |
429 | |
430 | /************************************************************** | |
431 | * Scheduling class tree data structure manipulation methods: | |
432 | */ | |
433 | ||
0702e3eb | 434 | static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime) |
02e0431a | 435 | { |
368059a9 PZ |
436 | s64 delta = (s64)(vruntime - min_vruntime); |
437 | if (delta > 0) | |
02e0431a PZ |
438 | min_vruntime = vruntime; |
439 | ||
440 | return min_vruntime; | |
441 | } | |
442 | ||
0702e3eb | 443 | static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime) |
b0ffd246 PZ |
444 | { |
445 | s64 delta = (s64)(vruntime - min_vruntime); | |
446 | if (delta < 0) | |
447 | min_vruntime = vruntime; | |
448 | ||
449 | return min_vruntime; | |
450 | } | |
451 | ||
54fdc581 FC |
452 | static inline int entity_before(struct sched_entity *a, |
453 | struct sched_entity *b) | |
454 | { | |
455 | return (s64)(a->vruntime - b->vruntime) < 0; | |
456 | } | |
457 | ||
1af5f730 PZ |
458 | static void update_min_vruntime(struct cfs_rq *cfs_rq) |
459 | { | |
460 | u64 vruntime = cfs_rq->min_vruntime; | |
461 | ||
462 | if (cfs_rq->curr) | |
463 | vruntime = cfs_rq->curr->vruntime; | |
464 | ||
465 | if (cfs_rq->rb_leftmost) { | |
466 | struct sched_entity *se = rb_entry(cfs_rq->rb_leftmost, | |
467 | struct sched_entity, | |
468 | run_node); | |
469 | ||
e17036da | 470 | if (!cfs_rq->curr) |
1af5f730 PZ |
471 | vruntime = se->vruntime; |
472 | else | |
473 | vruntime = min_vruntime(vruntime, se->vruntime); | |
474 | } | |
475 | ||
476 | cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime); | |
3fe1698b PZ |
477 | #ifndef CONFIG_64BIT |
478 | smp_wmb(); | |
479 | cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime; | |
480 | #endif | |
1af5f730 PZ |
481 | } |
482 | ||
bf0f6f24 IM |
483 | /* |
484 | * Enqueue an entity into the rb-tree: | |
485 | */ | |
0702e3eb | 486 | static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 IM |
487 | { |
488 | struct rb_node **link = &cfs_rq->tasks_timeline.rb_node; | |
489 | struct rb_node *parent = NULL; | |
490 | struct sched_entity *entry; | |
bf0f6f24 IM |
491 | int leftmost = 1; |
492 | ||
493 | /* | |
494 | * Find the right place in the rbtree: | |
495 | */ | |
496 | while (*link) { | |
497 | parent = *link; | |
498 | entry = rb_entry(parent, struct sched_entity, run_node); | |
499 | /* | |
500 | * We dont care about collisions. Nodes with | |
501 | * the same key stay together. | |
502 | */ | |
2bd2d6f2 | 503 | if (entity_before(se, entry)) { |
bf0f6f24 IM |
504 | link = &parent->rb_left; |
505 | } else { | |
506 | link = &parent->rb_right; | |
507 | leftmost = 0; | |
508 | } | |
509 | } | |
510 | ||
511 | /* | |
512 | * Maintain a cache of leftmost tree entries (it is frequently | |
513 | * used): | |
514 | */ | |
1af5f730 | 515 | if (leftmost) |
57cb499d | 516 | cfs_rq->rb_leftmost = &se->run_node; |
bf0f6f24 IM |
517 | |
518 | rb_link_node(&se->run_node, parent, link); | |
519 | rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline); | |
bf0f6f24 IM |
520 | } |
521 | ||
0702e3eb | 522 | static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 523 | { |
3fe69747 PZ |
524 | if (cfs_rq->rb_leftmost == &se->run_node) { |
525 | struct rb_node *next_node; | |
3fe69747 PZ |
526 | |
527 | next_node = rb_next(&se->run_node); | |
528 | cfs_rq->rb_leftmost = next_node; | |
3fe69747 | 529 | } |
e9acbff6 | 530 | |
bf0f6f24 | 531 | rb_erase(&se->run_node, &cfs_rq->tasks_timeline); |
bf0f6f24 IM |
532 | } |
533 | ||
029632fb | 534 | struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq) |
bf0f6f24 | 535 | { |
f4b6755f PZ |
536 | struct rb_node *left = cfs_rq->rb_leftmost; |
537 | ||
538 | if (!left) | |
539 | return NULL; | |
540 | ||
541 | return rb_entry(left, struct sched_entity, run_node); | |
bf0f6f24 IM |
542 | } |
543 | ||
ac53db59 RR |
544 | static struct sched_entity *__pick_next_entity(struct sched_entity *se) |
545 | { | |
546 | struct rb_node *next = rb_next(&se->run_node); | |
547 | ||
548 | if (!next) | |
549 | return NULL; | |
550 | ||
551 | return rb_entry(next, struct sched_entity, run_node); | |
552 | } | |
553 | ||
554 | #ifdef CONFIG_SCHED_DEBUG | |
029632fb | 555 | struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq) |
aeb73b04 | 556 | { |
7eee3e67 | 557 | struct rb_node *last = rb_last(&cfs_rq->tasks_timeline); |
aeb73b04 | 558 | |
70eee74b BS |
559 | if (!last) |
560 | return NULL; | |
7eee3e67 IM |
561 | |
562 | return rb_entry(last, struct sched_entity, run_node); | |
aeb73b04 PZ |
563 | } |
564 | ||
bf0f6f24 IM |
565 | /************************************************************** |
566 | * Scheduling class statistics methods: | |
567 | */ | |
568 | ||
acb4a848 | 569 | int sched_proc_update_handler(struct ctl_table *table, int write, |
8d65af78 | 570 | void __user *buffer, size_t *lenp, |
b2be5e96 PZ |
571 | loff_t *ppos) |
572 | { | |
8d65af78 | 573 | int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); |
acb4a848 | 574 | int factor = get_update_sysctl_factor(); |
b2be5e96 PZ |
575 | |
576 | if (ret || !write) | |
577 | return ret; | |
578 | ||
579 | sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency, | |
580 | sysctl_sched_min_granularity); | |
581 | ||
acb4a848 CE |
582 | #define WRT_SYSCTL(name) \ |
583 | (normalized_sysctl_##name = sysctl_##name / (factor)) | |
584 | WRT_SYSCTL(sched_min_granularity); | |
585 | WRT_SYSCTL(sched_latency); | |
586 | WRT_SYSCTL(sched_wakeup_granularity); | |
acb4a848 CE |
587 | #undef WRT_SYSCTL |
588 | ||
b2be5e96 PZ |
589 | return 0; |
590 | } | |
591 | #endif | |
647e7cac | 592 | |
a7be37ac | 593 | /* |
f9c0b095 | 594 | * delta /= w |
a7be37ac PZ |
595 | */ |
596 | static inline unsigned long | |
597 | calc_delta_fair(unsigned long delta, struct sched_entity *se) | |
598 | { | |
f9c0b095 PZ |
599 | if (unlikely(se->load.weight != NICE_0_LOAD)) |
600 | delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load); | |
a7be37ac PZ |
601 | |
602 | return delta; | |
603 | } | |
604 | ||
647e7cac IM |
605 | /* |
606 | * The idea is to set a period in which each task runs once. | |
607 | * | |
532b1858 | 608 | * When there are too many tasks (sched_nr_latency) we have to stretch |
647e7cac IM |
609 | * this period because otherwise the slices get too small. |
610 | * | |
611 | * p = (nr <= nl) ? l : l*nr/nl | |
612 | */ | |
4d78e7b6 PZ |
613 | static u64 __sched_period(unsigned long nr_running) |
614 | { | |
615 | u64 period = sysctl_sched_latency; | |
b2be5e96 | 616 | unsigned long nr_latency = sched_nr_latency; |
4d78e7b6 PZ |
617 | |
618 | if (unlikely(nr_running > nr_latency)) { | |
4bf0b771 | 619 | period = sysctl_sched_min_granularity; |
4d78e7b6 | 620 | period *= nr_running; |
4d78e7b6 PZ |
621 | } |
622 | ||
623 | return period; | |
624 | } | |
625 | ||
647e7cac IM |
626 | /* |
627 | * We calculate the wall-time slice from the period by taking a part | |
628 | * proportional to the weight. | |
629 | * | |
f9c0b095 | 630 | * s = p*P[w/rw] |
647e7cac | 631 | */ |
6d0f0ebd | 632 | static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se) |
21805085 | 633 | { |
0a582440 | 634 | u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq); |
f9c0b095 | 635 | |
0a582440 | 636 | for_each_sched_entity(se) { |
6272d68c | 637 | struct load_weight *load; |
3104bf03 | 638 | struct load_weight lw; |
6272d68c LM |
639 | |
640 | cfs_rq = cfs_rq_of(se); | |
641 | load = &cfs_rq->load; | |
f9c0b095 | 642 | |
0a582440 | 643 | if (unlikely(!se->on_rq)) { |
3104bf03 | 644 | lw = cfs_rq->load; |
0a582440 MG |
645 | |
646 | update_load_add(&lw, se->load.weight); | |
647 | load = &lw; | |
648 | } | |
649 | slice = calc_delta_mine(slice, se->load.weight, load); | |
650 | } | |
651 | return slice; | |
bf0f6f24 IM |
652 | } |
653 | ||
647e7cac | 654 | /* |
ac884dec | 655 | * We calculate the vruntime slice of a to be inserted task |
647e7cac | 656 | * |
f9c0b095 | 657 | * vs = s/w |
647e7cac | 658 | */ |
f9c0b095 | 659 | static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se) |
67e9fb2a | 660 | { |
f9c0b095 | 661 | return calc_delta_fair(sched_slice(cfs_rq, se), se); |
a7be37ac PZ |
662 | } |
663 | ||
bf0f6f24 IM |
664 | /* |
665 | * Update the current task's runtime statistics. Skip current tasks that | |
666 | * are not in our scheduling class. | |
667 | */ | |
668 | static inline void | |
8ebc91d9 IM |
669 | __update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr, |
670 | unsigned long delta_exec) | |
bf0f6f24 | 671 | { |
bbdba7c0 | 672 | unsigned long delta_exec_weighted; |
bf0f6f24 | 673 | |
41acab88 LDM |
674 | schedstat_set(curr->statistics.exec_max, |
675 | max((u64)delta_exec, curr->statistics.exec_max)); | |
bf0f6f24 IM |
676 | |
677 | curr->sum_exec_runtime += delta_exec; | |
7a62eabc | 678 | schedstat_add(cfs_rq, exec_clock, delta_exec); |
a7be37ac | 679 | delta_exec_weighted = calc_delta_fair(delta_exec, curr); |
88ec22d3 | 680 | |
e9acbff6 | 681 | curr->vruntime += delta_exec_weighted; |
1af5f730 | 682 | update_min_vruntime(cfs_rq); |
bf0f6f24 IM |
683 | } |
684 | ||
b7cc0896 | 685 | static void update_curr(struct cfs_rq *cfs_rq) |
bf0f6f24 | 686 | { |
429d43bc | 687 | struct sched_entity *curr = cfs_rq->curr; |
305e6835 | 688 | u64 now = rq_of(cfs_rq)->clock_task; |
bf0f6f24 IM |
689 | unsigned long delta_exec; |
690 | ||
691 | if (unlikely(!curr)) | |
692 | return; | |
693 | ||
694 | /* | |
695 | * Get the amount of time the current task was running | |
696 | * since the last time we changed load (this cannot | |
697 | * overflow on 32 bits): | |
698 | */ | |
8ebc91d9 | 699 | delta_exec = (unsigned long)(now - curr->exec_start); |
34f28ecd PZ |
700 | if (!delta_exec) |
701 | return; | |
bf0f6f24 | 702 | |
8ebc91d9 IM |
703 | __update_curr(cfs_rq, curr, delta_exec); |
704 | curr->exec_start = now; | |
d842de87 SV |
705 | |
706 | if (entity_is_task(curr)) { | |
707 | struct task_struct *curtask = task_of(curr); | |
708 | ||
f977bb49 | 709 | trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime); |
d842de87 | 710 | cpuacct_charge(curtask, delta_exec); |
f06febc9 | 711 | account_group_exec_runtime(curtask, delta_exec); |
d842de87 | 712 | } |
ec12cb7f PT |
713 | |
714 | account_cfs_rq_runtime(cfs_rq, delta_exec); | |
bf0f6f24 IM |
715 | } |
716 | ||
717 | static inline void | |
5870db5b | 718 | update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 719 | { |
41acab88 | 720 | schedstat_set(se->statistics.wait_start, rq_of(cfs_rq)->clock); |
bf0f6f24 IM |
721 | } |
722 | ||
bf0f6f24 IM |
723 | /* |
724 | * Task is being enqueued - update stats: | |
725 | */ | |
d2417e5a | 726 | static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 727 | { |
bf0f6f24 IM |
728 | /* |
729 | * Are we enqueueing a waiting task? (for current tasks | |
730 | * a dequeue/enqueue event is a NOP) | |
731 | */ | |
429d43bc | 732 | if (se != cfs_rq->curr) |
5870db5b | 733 | update_stats_wait_start(cfs_rq, se); |
bf0f6f24 IM |
734 | } |
735 | ||
bf0f6f24 | 736 | static void |
9ef0a961 | 737 | update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 738 | { |
41acab88 LDM |
739 | schedstat_set(se->statistics.wait_max, max(se->statistics.wait_max, |
740 | rq_of(cfs_rq)->clock - se->statistics.wait_start)); | |
741 | schedstat_set(se->statistics.wait_count, se->statistics.wait_count + 1); | |
742 | schedstat_set(se->statistics.wait_sum, se->statistics.wait_sum + | |
743 | rq_of(cfs_rq)->clock - se->statistics.wait_start); | |
768d0c27 PZ |
744 | #ifdef CONFIG_SCHEDSTATS |
745 | if (entity_is_task(se)) { | |
746 | trace_sched_stat_wait(task_of(se), | |
41acab88 | 747 | rq_of(cfs_rq)->clock - se->statistics.wait_start); |
768d0c27 PZ |
748 | } |
749 | #endif | |
41acab88 | 750 | schedstat_set(se->statistics.wait_start, 0); |
bf0f6f24 IM |
751 | } |
752 | ||
753 | static inline void | |
19b6a2e3 | 754 | update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 755 | { |
bf0f6f24 IM |
756 | /* |
757 | * Mark the end of the wait period if dequeueing a | |
758 | * waiting task: | |
759 | */ | |
429d43bc | 760 | if (se != cfs_rq->curr) |
9ef0a961 | 761 | update_stats_wait_end(cfs_rq, se); |
bf0f6f24 IM |
762 | } |
763 | ||
764 | /* | |
765 | * We are picking a new current task - update its stats: | |
766 | */ | |
767 | static inline void | |
79303e9e | 768 | update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 IM |
769 | { |
770 | /* | |
771 | * We are starting a new run period: | |
772 | */ | |
305e6835 | 773 | se->exec_start = rq_of(cfs_rq)->clock_task; |
bf0f6f24 IM |
774 | } |
775 | ||
bf0f6f24 IM |
776 | /************************************************** |
777 | * Scheduling class queueing methods: | |
778 | */ | |
779 | ||
cbee9f88 PZ |
780 | #ifdef CONFIG_NUMA_BALANCING |
781 | /* | |
6e5fb223 | 782 | * numa task sample period in ms |
cbee9f88 | 783 | */ |
6e5fb223 | 784 | unsigned int sysctl_numa_balancing_scan_period_min = 100; |
b8593bfd MG |
785 | unsigned int sysctl_numa_balancing_scan_period_max = 100*50; |
786 | unsigned int sysctl_numa_balancing_scan_period_reset = 100*600; | |
6e5fb223 PZ |
787 | |
788 | /* Portion of address space to scan in MB */ | |
789 | unsigned int sysctl_numa_balancing_scan_size = 256; | |
cbee9f88 | 790 | |
4b96a29b PZ |
791 | /* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */ |
792 | unsigned int sysctl_numa_balancing_scan_delay = 1000; | |
793 | ||
cbee9f88 PZ |
794 | static void task_numa_placement(struct task_struct *p) |
795 | { | |
796 | int seq = ACCESS_ONCE(p->mm->numa_scan_seq); | |
797 | ||
798 | if (p->numa_scan_seq == seq) | |
799 | return; | |
800 | p->numa_scan_seq = seq; | |
801 | ||
802 | /* FIXME: Scheduling placement policy hints go here */ | |
803 | } | |
804 | ||
805 | /* | |
806 | * Got a PROT_NONE fault for a page on @node. | |
807 | */ | |
b8593bfd | 808 | void task_numa_fault(int node, int pages, bool migrated) |
cbee9f88 PZ |
809 | { |
810 | struct task_struct *p = current; | |
811 | ||
1a687c2e MG |
812 | if (!sched_feat_numa(NUMA)) |
813 | return; | |
814 | ||
cbee9f88 PZ |
815 | /* FIXME: Allocate task-specific structure for placement policy here */ |
816 | ||
fb003b80 | 817 | /* |
b8593bfd MG |
818 | * If pages are properly placed (did not migrate) then scan slower. |
819 | * This is reset periodically in case of phase changes | |
fb003b80 | 820 | */ |
b8593bfd MG |
821 | if (!migrated) |
822 | p->numa_scan_period = min(sysctl_numa_balancing_scan_period_max, | |
823 | p->numa_scan_period + jiffies_to_msecs(10)); | |
fb003b80 | 824 | |
cbee9f88 PZ |
825 | task_numa_placement(p); |
826 | } | |
827 | ||
6e5fb223 PZ |
828 | static void reset_ptenuma_scan(struct task_struct *p) |
829 | { | |
830 | ACCESS_ONCE(p->mm->numa_scan_seq)++; | |
831 | p->mm->numa_scan_offset = 0; | |
832 | } | |
833 | ||
cbee9f88 PZ |
834 | /* |
835 | * The expensive part of numa migration is done from task_work context. | |
836 | * Triggered from task_tick_numa(). | |
837 | */ | |
838 | void task_numa_work(struct callback_head *work) | |
839 | { | |
840 | unsigned long migrate, next_scan, now = jiffies; | |
841 | struct task_struct *p = current; | |
842 | struct mm_struct *mm = p->mm; | |
6e5fb223 | 843 | struct vm_area_struct *vma; |
9f40604c MG |
844 | unsigned long start, end; |
845 | long pages; | |
cbee9f88 PZ |
846 | |
847 | WARN_ON_ONCE(p != container_of(work, struct task_struct, numa_work)); | |
848 | ||
849 | work->next = work; /* protect against double add */ | |
850 | /* | |
851 | * Who cares about NUMA placement when they're dying. | |
852 | * | |
853 | * NOTE: make sure not to dereference p->mm before this check, | |
854 | * exit_task_work() happens _after_ exit_mm() so we could be called | |
855 | * without p->mm even though we still had it when we enqueued this | |
856 | * work. | |
857 | */ | |
858 | if (p->flags & PF_EXITING) | |
859 | return; | |
860 | ||
5bca2303 MG |
861 | /* |
862 | * We do not care about task placement until a task runs on a node | |
863 | * other than the first one used by the address space. This is | |
864 | * largely because migrations are driven by what CPU the task | |
865 | * is running on. If it's never scheduled on another node, it'll | |
866 | * not migrate so why bother trapping the fault. | |
867 | */ | |
868 | if (mm->first_nid == NUMA_PTE_SCAN_INIT) | |
869 | mm->first_nid = numa_node_id(); | |
870 | if (mm->first_nid != NUMA_PTE_SCAN_ACTIVE) { | |
871 | /* Are we running on a new node yet? */ | |
872 | if (numa_node_id() == mm->first_nid && | |
873 | !sched_feat_numa(NUMA_FORCE)) | |
874 | return; | |
875 | ||
876 | mm->first_nid = NUMA_PTE_SCAN_ACTIVE; | |
877 | } | |
878 | ||
b8593bfd MG |
879 | /* |
880 | * Reset the scan period if enough time has gone by. Objective is that | |
881 | * scanning will be reduced if pages are properly placed. As tasks | |
882 | * can enter different phases this needs to be re-examined. Lacking | |
883 | * proper tracking of reference behaviour, this blunt hammer is used. | |
884 | */ | |
885 | migrate = mm->numa_next_reset; | |
886 | if (time_after(now, migrate)) { | |
887 | p->numa_scan_period = sysctl_numa_balancing_scan_period_min; | |
888 | next_scan = now + msecs_to_jiffies(sysctl_numa_balancing_scan_period_reset); | |
889 | xchg(&mm->numa_next_reset, next_scan); | |
890 | } | |
891 | ||
cbee9f88 PZ |
892 | /* |
893 | * Enforce maximal scan/migration frequency.. | |
894 | */ | |
895 | migrate = mm->numa_next_scan; | |
896 | if (time_before(now, migrate)) | |
897 | return; | |
898 | ||
899 | if (p->numa_scan_period == 0) | |
900 | p->numa_scan_period = sysctl_numa_balancing_scan_period_min; | |
901 | ||
fb003b80 | 902 | next_scan = now + msecs_to_jiffies(p->numa_scan_period); |
cbee9f88 PZ |
903 | if (cmpxchg(&mm->numa_next_scan, migrate, next_scan) != migrate) |
904 | return; | |
905 | ||
e14808b4 MG |
906 | /* |
907 | * Do not set pte_numa if the current running node is rate-limited. | |
908 | * This loses statistics on the fault but if we are unwilling to | |
909 | * migrate to this node, it is less likely we can do useful work | |
910 | */ | |
911 | if (migrate_ratelimited(numa_node_id())) | |
912 | return; | |
913 | ||
9f40604c MG |
914 | start = mm->numa_scan_offset; |
915 | pages = sysctl_numa_balancing_scan_size; | |
916 | pages <<= 20 - PAGE_SHIFT; /* MB in pages */ | |
917 | if (!pages) | |
918 | return; | |
cbee9f88 | 919 | |
6e5fb223 | 920 | down_read(&mm->mmap_sem); |
9f40604c | 921 | vma = find_vma(mm, start); |
6e5fb223 PZ |
922 | if (!vma) { |
923 | reset_ptenuma_scan(p); | |
9f40604c | 924 | start = 0; |
6e5fb223 PZ |
925 | vma = mm->mmap; |
926 | } | |
9f40604c | 927 | for (; vma; vma = vma->vm_next) { |
6e5fb223 PZ |
928 | if (!vma_migratable(vma)) |
929 | continue; | |
930 | ||
931 | /* Skip small VMAs. They are not likely to be of relevance */ | |
221392c3 | 932 | if (vma->vm_end - vma->vm_start < HPAGE_SIZE) |
6e5fb223 PZ |
933 | continue; |
934 | ||
9f40604c MG |
935 | do { |
936 | start = max(start, vma->vm_start); | |
937 | end = ALIGN(start + (pages << PAGE_SHIFT), HPAGE_SIZE); | |
938 | end = min(end, vma->vm_end); | |
939 | pages -= change_prot_numa(vma, start, end); | |
6e5fb223 | 940 | |
9f40604c MG |
941 | start = end; |
942 | if (pages <= 0) | |
943 | goto out; | |
944 | } while (end != vma->vm_end); | |
cbee9f88 | 945 | } |
6e5fb223 | 946 | |
9f40604c | 947 | out: |
6e5fb223 PZ |
948 | /* |
949 | * It is possible to reach the end of the VMA list but the last few VMAs are | |
950 | * not guaranteed to the vma_migratable. If they are not, we would find the | |
951 | * !migratable VMA on the next scan but not reset the scanner to the start | |
952 | * so check it now. | |
953 | */ | |
954 | if (vma) | |
9f40604c | 955 | mm->numa_scan_offset = start; |
6e5fb223 PZ |
956 | else |
957 | reset_ptenuma_scan(p); | |
958 | up_read(&mm->mmap_sem); | |
cbee9f88 PZ |
959 | } |
960 | ||
961 | /* | |
962 | * Drive the periodic memory faults.. | |
963 | */ | |
964 | void task_tick_numa(struct rq *rq, struct task_struct *curr) | |
965 | { | |
966 | struct callback_head *work = &curr->numa_work; | |
967 | u64 period, now; | |
968 | ||
969 | /* | |
970 | * We don't care about NUMA placement if we don't have memory. | |
971 | */ | |
972 | if (!curr->mm || (curr->flags & PF_EXITING) || work->next != work) | |
973 | return; | |
974 | ||
975 | /* | |
976 | * Using runtime rather than walltime has the dual advantage that | |
977 | * we (mostly) drive the selection from busy threads and that the | |
978 | * task needs to have done some actual work before we bother with | |
979 | * NUMA placement. | |
980 | */ | |
981 | now = curr->se.sum_exec_runtime; | |
982 | period = (u64)curr->numa_scan_period * NSEC_PER_MSEC; | |
983 | ||
984 | if (now - curr->node_stamp > period) { | |
4b96a29b PZ |
985 | if (!curr->node_stamp) |
986 | curr->numa_scan_period = sysctl_numa_balancing_scan_period_min; | |
cbee9f88 PZ |
987 | curr->node_stamp = now; |
988 | ||
989 | if (!time_before(jiffies, curr->mm->numa_next_scan)) { | |
990 | init_task_work(work, task_numa_work); /* TODO: move this into sched_fork() */ | |
991 | task_work_add(curr, work, true); | |
992 | } | |
993 | } | |
994 | } | |
995 | #else | |
996 | static void task_tick_numa(struct rq *rq, struct task_struct *curr) | |
997 | { | |
998 | } | |
999 | #endif /* CONFIG_NUMA_BALANCING */ | |
1000 | ||
30cfdcfc DA |
1001 | static void |
1002 | account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) | |
1003 | { | |
1004 | update_load_add(&cfs_rq->load, se->load.weight); | |
c09595f6 | 1005 | if (!parent_entity(se)) |
029632fb | 1006 | update_load_add(&rq_of(cfs_rq)->load, se->load.weight); |
367456c7 PZ |
1007 | #ifdef CONFIG_SMP |
1008 | if (entity_is_task(se)) | |
eb95308e | 1009 | list_add(&se->group_node, &rq_of(cfs_rq)->cfs_tasks); |
367456c7 | 1010 | #endif |
30cfdcfc | 1011 | cfs_rq->nr_running++; |
30cfdcfc DA |
1012 | } |
1013 | ||
1014 | static void | |
1015 | account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) | |
1016 | { | |
1017 | update_load_sub(&cfs_rq->load, se->load.weight); | |
c09595f6 | 1018 | if (!parent_entity(se)) |
029632fb | 1019 | update_load_sub(&rq_of(cfs_rq)->load, se->load.weight); |
367456c7 | 1020 | if (entity_is_task(se)) |
b87f1724 | 1021 | list_del_init(&se->group_node); |
30cfdcfc | 1022 | cfs_rq->nr_running--; |
30cfdcfc DA |
1023 | } |
1024 | ||
3ff6dcac YZ |
1025 | #ifdef CONFIG_FAIR_GROUP_SCHED |
1026 | # ifdef CONFIG_SMP | |
cf5f0acf PZ |
1027 | static inline long calc_tg_weight(struct task_group *tg, struct cfs_rq *cfs_rq) |
1028 | { | |
1029 | long tg_weight; | |
1030 | ||
1031 | /* | |
1032 | * Use this CPU's actual weight instead of the last load_contribution | |
1033 | * to gain a more accurate current total weight. See | |
1034 | * update_cfs_rq_load_contribution(). | |
1035 | */ | |
82958366 PT |
1036 | tg_weight = atomic64_read(&tg->load_avg); |
1037 | tg_weight -= cfs_rq->tg_load_contrib; | |
cf5f0acf PZ |
1038 | tg_weight += cfs_rq->load.weight; |
1039 | ||
1040 | return tg_weight; | |
1041 | } | |
1042 | ||
6d5ab293 | 1043 | static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg) |
3ff6dcac | 1044 | { |
cf5f0acf | 1045 | long tg_weight, load, shares; |
3ff6dcac | 1046 | |
cf5f0acf | 1047 | tg_weight = calc_tg_weight(tg, cfs_rq); |
6d5ab293 | 1048 | load = cfs_rq->load.weight; |
3ff6dcac | 1049 | |
3ff6dcac | 1050 | shares = (tg->shares * load); |
cf5f0acf PZ |
1051 | if (tg_weight) |
1052 | shares /= tg_weight; | |
3ff6dcac YZ |
1053 | |
1054 | if (shares < MIN_SHARES) | |
1055 | shares = MIN_SHARES; | |
1056 | if (shares > tg->shares) | |
1057 | shares = tg->shares; | |
1058 | ||
1059 | return shares; | |
1060 | } | |
3ff6dcac | 1061 | # else /* CONFIG_SMP */ |
6d5ab293 | 1062 | static inline long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg) |
3ff6dcac YZ |
1063 | { |
1064 | return tg->shares; | |
1065 | } | |
3ff6dcac | 1066 | # endif /* CONFIG_SMP */ |
2069dd75 PZ |
1067 | static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, |
1068 | unsigned long weight) | |
1069 | { | |
19e5eebb PT |
1070 | if (se->on_rq) { |
1071 | /* commit outstanding execution time */ | |
1072 | if (cfs_rq->curr == se) | |
1073 | update_curr(cfs_rq); | |
2069dd75 | 1074 | account_entity_dequeue(cfs_rq, se); |
19e5eebb | 1075 | } |
2069dd75 PZ |
1076 | |
1077 | update_load_set(&se->load, weight); | |
1078 | ||
1079 | if (se->on_rq) | |
1080 | account_entity_enqueue(cfs_rq, se); | |
1081 | } | |
1082 | ||
82958366 PT |
1083 | static inline int throttled_hierarchy(struct cfs_rq *cfs_rq); |
1084 | ||
6d5ab293 | 1085 | static void update_cfs_shares(struct cfs_rq *cfs_rq) |
2069dd75 PZ |
1086 | { |
1087 | struct task_group *tg; | |
1088 | struct sched_entity *se; | |
3ff6dcac | 1089 | long shares; |
2069dd75 | 1090 | |
2069dd75 PZ |
1091 | tg = cfs_rq->tg; |
1092 | se = tg->se[cpu_of(rq_of(cfs_rq))]; | |
64660c86 | 1093 | if (!se || throttled_hierarchy(cfs_rq)) |
2069dd75 | 1094 | return; |
3ff6dcac YZ |
1095 | #ifndef CONFIG_SMP |
1096 | if (likely(se->load.weight == tg->shares)) | |
1097 | return; | |
1098 | #endif | |
6d5ab293 | 1099 | shares = calc_cfs_shares(cfs_rq, tg); |
2069dd75 PZ |
1100 | |
1101 | reweight_entity(cfs_rq_of(se), se, shares); | |
1102 | } | |
1103 | #else /* CONFIG_FAIR_GROUP_SCHED */ | |
6d5ab293 | 1104 | static inline void update_cfs_shares(struct cfs_rq *cfs_rq) |
2069dd75 PZ |
1105 | { |
1106 | } | |
1107 | #endif /* CONFIG_FAIR_GROUP_SCHED */ | |
1108 | ||
f4e26b12 PT |
1109 | /* Only depends on SMP, FAIR_GROUP_SCHED may be removed when useful in lb */ |
1110 | #if defined(CONFIG_SMP) && defined(CONFIG_FAIR_GROUP_SCHED) | |
5b51f2f8 PT |
1111 | /* |
1112 | * We choose a half-life close to 1 scheduling period. | |
1113 | * Note: The tables below are dependent on this value. | |
1114 | */ | |
1115 | #define LOAD_AVG_PERIOD 32 | |
1116 | #define LOAD_AVG_MAX 47742 /* maximum possible load avg */ | |
1117 | #define LOAD_AVG_MAX_N 345 /* number of full periods to produce LOAD_MAX_AVG */ | |
1118 | ||
1119 | /* Precomputed fixed inverse multiplies for multiplication by y^n */ | |
1120 | static const u32 runnable_avg_yN_inv[] = { | |
1121 | 0xffffffff, 0xfa83b2da, 0xf5257d14, 0xefe4b99a, 0xeac0c6e6, 0xe5b906e6, | |
1122 | 0xe0ccdeeb, 0xdbfbb796, 0xd744fcc9, 0xd2a81d91, 0xce248c14, 0xc9b9bd85, | |
1123 | 0xc5672a10, 0xc12c4cc9, 0xbd08a39e, 0xb8fbaf46, 0xb504f333, 0xb123f581, | |
1124 | 0xad583ee9, 0xa9a15ab4, 0xa5fed6a9, 0xa2704302, 0x9ef5325f, 0x9b8d39b9, | |
1125 | 0x9837f050, 0x94f4efa8, 0x91c3d373, 0x8ea4398a, 0x8b95c1e3, 0x88980e80, | |
1126 | 0x85aac367, 0x82cd8698, | |
1127 | }; | |
1128 | ||
1129 | /* | |
1130 | * Precomputed \Sum y^k { 1<=k<=n }. These are floor(true_value) to prevent | |
1131 | * over-estimates when re-combining. | |
1132 | */ | |
1133 | static const u32 runnable_avg_yN_sum[] = { | |
1134 | 0, 1002, 1982, 2941, 3880, 4798, 5697, 6576, 7437, 8279, 9103, | |
1135 | 9909,10698,11470,12226,12966,13690,14398,15091,15769,16433,17082, | |
1136 | 17718,18340,18949,19545,20128,20698,21256,21802,22336,22859,23371, | |
1137 | }; | |
1138 | ||
9d85f21c PT |
1139 | /* |
1140 | * Approximate: | |
1141 | * val * y^n, where y^32 ~= 0.5 (~1 scheduling period) | |
1142 | */ | |
1143 | static __always_inline u64 decay_load(u64 val, u64 n) | |
1144 | { | |
5b51f2f8 PT |
1145 | unsigned int local_n; |
1146 | ||
1147 | if (!n) | |
1148 | return val; | |
1149 | else if (unlikely(n > LOAD_AVG_PERIOD * 63)) | |
1150 | return 0; | |
1151 | ||
1152 | /* after bounds checking we can collapse to 32-bit */ | |
1153 | local_n = n; | |
1154 | ||
1155 | /* | |
1156 | * As y^PERIOD = 1/2, we can combine | |
1157 | * y^n = 1/2^(n/PERIOD) * k^(n%PERIOD) | |
1158 | * With a look-up table which covers k^n (n<PERIOD) | |
1159 | * | |
1160 | * To achieve constant time decay_load. | |
1161 | */ | |
1162 | if (unlikely(local_n >= LOAD_AVG_PERIOD)) { | |
1163 | val >>= local_n / LOAD_AVG_PERIOD; | |
1164 | local_n %= LOAD_AVG_PERIOD; | |
9d85f21c PT |
1165 | } |
1166 | ||
5b51f2f8 PT |
1167 | val *= runnable_avg_yN_inv[local_n]; |
1168 | /* We don't use SRR here since we always want to round down. */ | |
1169 | return val >> 32; | |
1170 | } | |
1171 | ||
1172 | /* | |
1173 | * For updates fully spanning n periods, the contribution to runnable | |
1174 | * average will be: \Sum 1024*y^n | |
1175 | * | |
1176 | * We can compute this reasonably efficiently by combining: | |
1177 | * y^PERIOD = 1/2 with precomputed \Sum 1024*y^n {for n <PERIOD} | |
1178 | */ | |
1179 | static u32 __compute_runnable_contrib(u64 n) | |
1180 | { | |
1181 | u32 contrib = 0; | |
1182 | ||
1183 | if (likely(n <= LOAD_AVG_PERIOD)) | |
1184 | return runnable_avg_yN_sum[n]; | |
1185 | else if (unlikely(n >= LOAD_AVG_MAX_N)) | |
1186 | return LOAD_AVG_MAX; | |
1187 | ||
1188 | /* Compute \Sum k^n combining precomputed values for k^i, \Sum k^j */ | |
1189 | do { | |
1190 | contrib /= 2; /* y^LOAD_AVG_PERIOD = 1/2 */ | |
1191 | contrib += runnable_avg_yN_sum[LOAD_AVG_PERIOD]; | |
1192 | ||
1193 | n -= LOAD_AVG_PERIOD; | |
1194 | } while (n > LOAD_AVG_PERIOD); | |
1195 | ||
1196 | contrib = decay_load(contrib, n); | |
1197 | return contrib + runnable_avg_yN_sum[n]; | |
9d85f21c PT |
1198 | } |
1199 | ||
1200 | /* | |
1201 | * We can represent the historical contribution to runnable average as the | |
1202 | * coefficients of a geometric series. To do this we sub-divide our runnable | |
1203 | * history into segments of approximately 1ms (1024us); label the segment that | |
1204 | * occurred N-ms ago p_N, with p_0 corresponding to the current period, e.g. | |
1205 | * | |
1206 | * [<- 1024us ->|<- 1024us ->|<- 1024us ->| ... | |
1207 | * p0 p1 p2 | |
1208 | * (now) (~1ms ago) (~2ms ago) | |
1209 | * | |
1210 | * Let u_i denote the fraction of p_i that the entity was runnable. | |
1211 | * | |
1212 | * We then designate the fractions u_i as our co-efficients, yielding the | |
1213 | * following representation of historical load: | |
1214 | * u_0 + u_1*y + u_2*y^2 + u_3*y^3 + ... | |
1215 | * | |
1216 | * We choose y based on the with of a reasonably scheduling period, fixing: | |
1217 | * y^32 = 0.5 | |
1218 | * | |
1219 | * This means that the contribution to load ~32ms ago (u_32) will be weighted | |
1220 | * approximately half as much as the contribution to load within the last ms | |
1221 | * (u_0). | |
1222 | * | |
1223 | * When a period "rolls over" and we have new u_0`, multiplying the previous | |
1224 | * sum again by y is sufficient to update: | |
1225 | * load_avg = u_0` + y*(u_0 + u_1*y + u_2*y^2 + ... ) | |
1226 | * = u_0 + u_1*y + u_2*y^2 + ... [re-labeling u_i --> u_{i+1}] | |
1227 | */ | |
1228 | static __always_inline int __update_entity_runnable_avg(u64 now, | |
1229 | struct sched_avg *sa, | |
1230 | int runnable) | |
1231 | { | |
5b51f2f8 PT |
1232 | u64 delta, periods; |
1233 | u32 runnable_contrib; | |
9d85f21c PT |
1234 | int delta_w, decayed = 0; |
1235 | ||
1236 | delta = now - sa->last_runnable_update; | |
1237 | /* | |
1238 | * This should only happen when time goes backwards, which it | |
1239 | * unfortunately does during sched clock init when we swap over to TSC. | |
1240 | */ | |
1241 | if ((s64)delta < 0) { | |
1242 | sa->last_runnable_update = now; | |
1243 | return 0; | |
1244 | } | |
1245 | ||
1246 | /* | |
1247 | * Use 1024ns as the unit of measurement since it's a reasonable | |
1248 | * approximation of 1us and fast to compute. | |
1249 | */ | |
1250 | delta >>= 10; | |
1251 | if (!delta) | |
1252 | return 0; | |
1253 | sa->last_runnable_update = now; | |
1254 | ||
1255 | /* delta_w is the amount already accumulated against our next period */ | |
1256 | delta_w = sa->runnable_avg_period % 1024; | |
1257 | if (delta + delta_w >= 1024) { | |
1258 | /* period roll-over */ | |
1259 | decayed = 1; | |
1260 | ||
1261 | /* | |
1262 | * Now that we know we're crossing a period boundary, figure | |
1263 | * out how much from delta we need to complete the current | |
1264 | * period and accrue it. | |
1265 | */ | |
1266 | delta_w = 1024 - delta_w; | |
5b51f2f8 PT |
1267 | if (runnable) |
1268 | sa->runnable_avg_sum += delta_w; | |
1269 | sa->runnable_avg_period += delta_w; | |
1270 | ||
1271 | delta -= delta_w; | |
1272 | ||
1273 | /* Figure out how many additional periods this update spans */ | |
1274 | periods = delta / 1024; | |
1275 | delta %= 1024; | |
1276 | ||
1277 | sa->runnable_avg_sum = decay_load(sa->runnable_avg_sum, | |
1278 | periods + 1); | |
1279 | sa->runnable_avg_period = decay_load(sa->runnable_avg_period, | |
1280 | periods + 1); | |
1281 | ||
1282 | /* Efficiently calculate \sum (1..n_period) 1024*y^i */ | |
1283 | runnable_contrib = __compute_runnable_contrib(periods); | |
1284 | if (runnable) | |
1285 | sa->runnable_avg_sum += runnable_contrib; | |
1286 | sa->runnable_avg_period += runnable_contrib; | |
9d85f21c PT |
1287 | } |
1288 | ||
1289 | /* Remainder of delta accrued against u_0` */ | |
1290 | if (runnable) | |
1291 | sa->runnable_avg_sum += delta; | |
1292 | sa->runnable_avg_period += delta; | |
1293 | ||
1294 | return decayed; | |
1295 | } | |
1296 | ||
9ee474f5 | 1297 | /* Synchronize an entity's decay with its parenting cfs_rq.*/ |
aff3e498 | 1298 | static inline u64 __synchronize_entity_decay(struct sched_entity *se) |
9ee474f5 PT |
1299 | { |
1300 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
1301 | u64 decays = atomic64_read(&cfs_rq->decay_counter); | |
1302 | ||
1303 | decays -= se->avg.decay_count; | |
1304 | if (!decays) | |
aff3e498 | 1305 | return 0; |
9ee474f5 PT |
1306 | |
1307 | se->avg.load_avg_contrib = decay_load(se->avg.load_avg_contrib, decays); | |
1308 | se->avg.decay_count = 0; | |
aff3e498 PT |
1309 | |
1310 | return decays; | |
9ee474f5 PT |
1311 | } |
1312 | ||
c566e8e9 PT |
1313 | #ifdef CONFIG_FAIR_GROUP_SCHED |
1314 | static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq, | |
1315 | int force_update) | |
1316 | { | |
1317 | struct task_group *tg = cfs_rq->tg; | |
1318 | s64 tg_contrib; | |
1319 | ||
1320 | tg_contrib = cfs_rq->runnable_load_avg + cfs_rq->blocked_load_avg; | |
1321 | tg_contrib -= cfs_rq->tg_load_contrib; | |
1322 | ||
1323 | if (force_update || abs64(tg_contrib) > cfs_rq->tg_load_contrib / 8) { | |
1324 | atomic64_add(tg_contrib, &tg->load_avg); | |
1325 | cfs_rq->tg_load_contrib += tg_contrib; | |
1326 | } | |
1327 | } | |
8165e145 | 1328 | |
bb17f655 PT |
1329 | /* |
1330 | * Aggregate cfs_rq runnable averages into an equivalent task_group | |
1331 | * representation for computing load contributions. | |
1332 | */ | |
1333 | static inline void __update_tg_runnable_avg(struct sched_avg *sa, | |
1334 | struct cfs_rq *cfs_rq) | |
1335 | { | |
1336 | struct task_group *tg = cfs_rq->tg; | |
1337 | long contrib; | |
1338 | ||
1339 | /* The fraction of a cpu used by this cfs_rq */ | |
1340 | contrib = div_u64(sa->runnable_avg_sum << NICE_0_SHIFT, | |
1341 | sa->runnable_avg_period + 1); | |
1342 | contrib -= cfs_rq->tg_runnable_contrib; | |
1343 | ||
1344 | if (abs(contrib) > cfs_rq->tg_runnable_contrib / 64) { | |
1345 | atomic_add(contrib, &tg->runnable_avg); | |
1346 | cfs_rq->tg_runnable_contrib += contrib; | |
1347 | } | |
1348 | } | |
1349 | ||
8165e145 PT |
1350 | static inline void __update_group_entity_contrib(struct sched_entity *se) |
1351 | { | |
1352 | struct cfs_rq *cfs_rq = group_cfs_rq(se); | |
1353 | struct task_group *tg = cfs_rq->tg; | |
bb17f655 PT |
1354 | int runnable_avg; |
1355 | ||
8165e145 PT |
1356 | u64 contrib; |
1357 | ||
1358 | contrib = cfs_rq->tg_load_contrib * tg->shares; | |
1359 | se->avg.load_avg_contrib = div64_u64(contrib, | |
1360 | atomic64_read(&tg->load_avg) + 1); | |
bb17f655 PT |
1361 | |
1362 | /* | |
1363 | * For group entities we need to compute a correction term in the case | |
1364 | * that they are consuming <1 cpu so that we would contribute the same | |
1365 | * load as a task of equal weight. | |
1366 | * | |
1367 | * Explicitly co-ordinating this measurement would be expensive, but | |
1368 | * fortunately the sum of each cpus contribution forms a usable | |
1369 | * lower-bound on the true value. | |
1370 | * | |
1371 | * Consider the aggregate of 2 contributions. Either they are disjoint | |
1372 | * (and the sum represents true value) or they are disjoint and we are | |
1373 | * understating by the aggregate of their overlap. | |
1374 | * | |
1375 | * Extending this to N cpus, for a given overlap, the maximum amount we | |
1376 | * understand is then n_i(n_i+1)/2 * w_i where n_i is the number of | |
1377 | * cpus that overlap for this interval and w_i is the interval width. | |
1378 | * | |
1379 | * On a small machine; the first term is well-bounded which bounds the | |
1380 | * total error since w_i is a subset of the period. Whereas on a | |
1381 | * larger machine, while this first term can be larger, if w_i is the | |
1382 | * of consequential size guaranteed to see n_i*w_i quickly converge to | |
1383 | * our upper bound of 1-cpu. | |
1384 | */ | |
1385 | runnable_avg = atomic_read(&tg->runnable_avg); | |
1386 | if (runnable_avg < NICE_0_LOAD) { | |
1387 | se->avg.load_avg_contrib *= runnable_avg; | |
1388 | se->avg.load_avg_contrib >>= NICE_0_SHIFT; | |
1389 | } | |
8165e145 | 1390 | } |
c566e8e9 PT |
1391 | #else |
1392 | static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq, | |
1393 | int force_update) {} | |
bb17f655 PT |
1394 | static inline void __update_tg_runnable_avg(struct sched_avg *sa, |
1395 | struct cfs_rq *cfs_rq) {} | |
8165e145 | 1396 | static inline void __update_group_entity_contrib(struct sched_entity *se) {} |
c566e8e9 PT |
1397 | #endif |
1398 | ||
8165e145 PT |
1399 | static inline void __update_task_entity_contrib(struct sched_entity *se) |
1400 | { | |
1401 | u32 contrib; | |
1402 | ||
1403 | /* avoid overflowing a 32-bit type w/ SCHED_LOAD_SCALE */ | |
1404 | contrib = se->avg.runnable_avg_sum * scale_load_down(se->load.weight); | |
1405 | contrib /= (se->avg.runnable_avg_period + 1); | |
1406 | se->avg.load_avg_contrib = scale_load(contrib); | |
1407 | } | |
1408 | ||
2dac754e PT |
1409 | /* Compute the current contribution to load_avg by se, return any delta */ |
1410 | static long __update_entity_load_avg_contrib(struct sched_entity *se) | |
1411 | { | |
1412 | long old_contrib = se->avg.load_avg_contrib; | |
1413 | ||
8165e145 PT |
1414 | if (entity_is_task(se)) { |
1415 | __update_task_entity_contrib(se); | |
1416 | } else { | |
bb17f655 | 1417 | __update_tg_runnable_avg(&se->avg, group_cfs_rq(se)); |
8165e145 PT |
1418 | __update_group_entity_contrib(se); |
1419 | } | |
2dac754e PT |
1420 | |
1421 | return se->avg.load_avg_contrib - old_contrib; | |
1422 | } | |
1423 | ||
9ee474f5 PT |
1424 | static inline void subtract_blocked_load_contrib(struct cfs_rq *cfs_rq, |
1425 | long load_contrib) | |
1426 | { | |
1427 | if (likely(load_contrib < cfs_rq->blocked_load_avg)) | |
1428 | cfs_rq->blocked_load_avg -= load_contrib; | |
1429 | else | |
1430 | cfs_rq->blocked_load_avg = 0; | |
1431 | } | |
1432 | ||
f1b17280 PT |
1433 | static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq); |
1434 | ||
9d85f21c | 1435 | /* Update a sched_entity's runnable average */ |
9ee474f5 PT |
1436 | static inline void update_entity_load_avg(struct sched_entity *se, |
1437 | int update_cfs_rq) | |
9d85f21c | 1438 | { |
2dac754e PT |
1439 | struct cfs_rq *cfs_rq = cfs_rq_of(se); |
1440 | long contrib_delta; | |
f1b17280 | 1441 | u64 now; |
2dac754e | 1442 | |
f1b17280 PT |
1443 | /* |
1444 | * For a group entity we need to use their owned cfs_rq_clock_task() in | |
1445 | * case they are the parent of a throttled hierarchy. | |
1446 | */ | |
1447 | if (entity_is_task(se)) | |
1448 | now = cfs_rq_clock_task(cfs_rq); | |
1449 | else | |
1450 | now = cfs_rq_clock_task(group_cfs_rq(se)); | |
1451 | ||
1452 | if (!__update_entity_runnable_avg(now, &se->avg, se->on_rq)) | |
2dac754e PT |
1453 | return; |
1454 | ||
1455 | contrib_delta = __update_entity_load_avg_contrib(se); | |
9ee474f5 PT |
1456 | |
1457 | if (!update_cfs_rq) | |
1458 | return; | |
1459 | ||
2dac754e PT |
1460 | if (se->on_rq) |
1461 | cfs_rq->runnable_load_avg += contrib_delta; | |
9ee474f5 PT |
1462 | else |
1463 | subtract_blocked_load_contrib(cfs_rq, -contrib_delta); | |
1464 | } | |
1465 | ||
1466 | /* | |
1467 | * Decay the load contributed by all blocked children and account this so that | |
1468 | * their contribution may appropriately discounted when they wake up. | |
1469 | */ | |
aff3e498 | 1470 | static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update) |
9ee474f5 | 1471 | { |
f1b17280 | 1472 | u64 now = cfs_rq_clock_task(cfs_rq) >> 20; |
9ee474f5 PT |
1473 | u64 decays; |
1474 | ||
1475 | decays = now - cfs_rq->last_decay; | |
aff3e498 | 1476 | if (!decays && !force_update) |
9ee474f5 PT |
1477 | return; |
1478 | ||
aff3e498 PT |
1479 | if (atomic64_read(&cfs_rq->removed_load)) { |
1480 | u64 removed_load = atomic64_xchg(&cfs_rq->removed_load, 0); | |
1481 | subtract_blocked_load_contrib(cfs_rq, removed_load); | |
1482 | } | |
9ee474f5 | 1483 | |
aff3e498 PT |
1484 | if (decays) { |
1485 | cfs_rq->blocked_load_avg = decay_load(cfs_rq->blocked_load_avg, | |
1486 | decays); | |
1487 | atomic64_add(decays, &cfs_rq->decay_counter); | |
1488 | cfs_rq->last_decay = now; | |
1489 | } | |
c566e8e9 PT |
1490 | |
1491 | __update_cfs_rq_tg_load_contrib(cfs_rq, force_update); | |
9d85f21c | 1492 | } |
18bf2805 BS |
1493 | |
1494 | static inline void update_rq_runnable_avg(struct rq *rq, int runnable) | |
1495 | { | |
1496 | __update_entity_runnable_avg(rq->clock_task, &rq->avg, runnable); | |
bb17f655 | 1497 | __update_tg_runnable_avg(&rq->avg, &rq->cfs); |
18bf2805 | 1498 | } |
2dac754e PT |
1499 | |
1500 | /* Add the load generated by se into cfs_rq's child load-average */ | |
1501 | static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq, | |
9ee474f5 PT |
1502 | struct sched_entity *se, |
1503 | int wakeup) | |
2dac754e | 1504 | { |
aff3e498 PT |
1505 | /* |
1506 | * We track migrations using entity decay_count <= 0, on a wake-up | |
1507 | * migration we use a negative decay count to track the remote decays | |
1508 | * accumulated while sleeping. | |
1509 | */ | |
1510 | if (unlikely(se->avg.decay_count <= 0)) { | |
9ee474f5 | 1511 | se->avg.last_runnable_update = rq_of(cfs_rq)->clock_task; |
aff3e498 PT |
1512 | if (se->avg.decay_count) { |
1513 | /* | |
1514 | * In a wake-up migration we have to approximate the | |
1515 | * time sleeping. This is because we can't synchronize | |
1516 | * clock_task between the two cpus, and it is not | |
1517 | * guaranteed to be read-safe. Instead, we can | |
1518 | * approximate this using our carried decays, which are | |
1519 | * explicitly atomically readable. | |
1520 | */ | |
1521 | se->avg.last_runnable_update -= (-se->avg.decay_count) | |
1522 | << 20; | |
1523 | update_entity_load_avg(se, 0); | |
1524 | /* Indicate that we're now synchronized and on-rq */ | |
1525 | se->avg.decay_count = 0; | |
1526 | } | |
9ee474f5 PT |
1527 | wakeup = 0; |
1528 | } else { | |
1529 | __synchronize_entity_decay(se); | |
1530 | } | |
1531 | ||
aff3e498 PT |
1532 | /* migrated tasks did not contribute to our blocked load */ |
1533 | if (wakeup) { | |
9ee474f5 | 1534 | subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib); |
aff3e498 PT |
1535 | update_entity_load_avg(se, 0); |
1536 | } | |
9ee474f5 | 1537 | |
2dac754e | 1538 | cfs_rq->runnable_load_avg += se->avg.load_avg_contrib; |
aff3e498 PT |
1539 | /* we force update consideration on load-balancer moves */ |
1540 | update_cfs_rq_blocked_load(cfs_rq, !wakeup); | |
2dac754e PT |
1541 | } |
1542 | ||
9ee474f5 PT |
1543 | /* |
1544 | * Remove se's load from this cfs_rq child load-average, if the entity is | |
1545 | * transitioning to a blocked state we track its projected decay using | |
1546 | * blocked_load_avg. | |
1547 | */ | |
2dac754e | 1548 | static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f5 PT |
1549 | struct sched_entity *se, |
1550 | int sleep) | |
2dac754e | 1551 | { |
9ee474f5 | 1552 | update_entity_load_avg(se, 1); |
aff3e498 PT |
1553 | /* we force update consideration on load-balancer moves */ |
1554 | update_cfs_rq_blocked_load(cfs_rq, !sleep); | |
9ee474f5 | 1555 | |
2dac754e | 1556 | cfs_rq->runnable_load_avg -= se->avg.load_avg_contrib; |
9ee474f5 PT |
1557 | if (sleep) { |
1558 | cfs_rq->blocked_load_avg += se->avg.load_avg_contrib; | |
1559 | se->avg.decay_count = atomic64_read(&cfs_rq->decay_counter); | |
1560 | } /* migrations, e.g. sleep=0 leave decay_count == 0 */ | |
2dac754e | 1561 | } |
9d85f21c | 1562 | #else |
9ee474f5 PT |
1563 | static inline void update_entity_load_avg(struct sched_entity *se, |
1564 | int update_cfs_rq) {} | |
18bf2805 | 1565 | static inline void update_rq_runnable_avg(struct rq *rq, int runnable) {} |
2dac754e | 1566 | static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f5 PT |
1567 | struct sched_entity *se, |
1568 | int wakeup) {} | |
2dac754e | 1569 | static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f5 PT |
1570 | struct sched_entity *se, |
1571 | int sleep) {} | |
aff3e498 PT |
1572 | static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, |
1573 | int force_update) {} | |
9d85f21c PT |
1574 | #endif |
1575 | ||
2396af69 | 1576 | static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 1577 | { |
bf0f6f24 | 1578 | #ifdef CONFIG_SCHEDSTATS |
e414314c PZ |
1579 | struct task_struct *tsk = NULL; |
1580 | ||
1581 | if (entity_is_task(se)) | |
1582 | tsk = task_of(se); | |
1583 | ||
41acab88 LDM |
1584 | if (se->statistics.sleep_start) { |
1585 | u64 delta = rq_of(cfs_rq)->clock - se->statistics.sleep_start; | |
bf0f6f24 IM |
1586 | |
1587 | if ((s64)delta < 0) | |
1588 | delta = 0; | |
1589 | ||
41acab88 LDM |
1590 | if (unlikely(delta > se->statistics.sleep_max)) |
1591 | se->statistics.sleep_max = delta; | |
bf0f6f24 | 1592 | |
8c79a045 | 1593 | se->statistics.sleep_start = 0; |
41acab88 | 1594 | se->statistics.sum_sleep_runtime += delta; |
9745512c | 1595 | |
768d0c27 | 1596 | if (tsk) { |
e414314c | 1597 | account_scheduler_latency(tsk, delta >> 10, 1); |
768d0c27 PZ |
1598 | trace_sched_stat_sleep(tsk, delta); |
1599 | } | |
bf0f6f24 | 1600 | } |
41acab88 LDM |
1601 | if (se->statistics.block_start) { |
1602 | u64 delta = rq_of(cfs_rq)->clock - se->statistics.block_start; | |
bf0f6f24 IM |
1603 | |
1604 | if ((s64)delta < 0) | |
1605 | delta = 0; | |
1606 | ||
41acab88 LDM |
1607 | if (unlikely(delta > se->statistics.block_max)) |
1608 | se->statistics.block_max = delta; | |
bf0f6f24 | 1609 | |
8c79a045 | 1610 | se->statistics.block_start = 0; |
41acab88 | 1611 | se->statistics.sum_sleep_runtime += delta; |
30084fbd | 1612 | |
e414314c | 1613 | if (tsk) { |
8f0dfc34 | 1614 | if (tsk->in_iowait) { |
41acab88 LDM |
1615 | se->statistics.iowait_sum += delta; |
1616 | se->statistics.iowait_count++; | |
768d0c27 | 1617 | trace_sched_stat_iowait(tsk, delta); |
8f0dfc34 AV |
1618 | } |
1619 | ||
b781a602 AV |
1620 | trace_sched_stat_blocked(tsk, delta); |
1621 | ||
e414314c PZ |
1622 | /* |
1623 | * Blocking time is in units of nanosecs, so shift by | |
1624 | * 20 to get a milliseconds-range estimation of the | |
1625 | * amount of time that the task spent sleeping: | |
1626 | */ | |
1627 | if (unlikely(prof_on == SLEEP_PROFILING)) { | |
1628 | profile_hits(SLEEP_PROFILING, | |
1629 | (void *)get_wchan(tsk), | |
1630 | delta >> 20); | |
1631 | } | |
1632 | account_scheduler_latency(tsk, delta >> 10, 0); | |
30084fbd | 1633 | } |
bf0f6f24 IM |
1634 | } |
1635 | #endif | |
1636 | } | |
1637 | ||
ddc97297 PZ |
1638 | static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se) |
1639 | { | |
1640 | #ifdef CONFIG_SCHED_DEBUG | |
1641 | s64 d = se->vruntime - cfs_rq->min_vruntime; | |
1642 | ||
1643 | if (d < 0) | |
1644 | d = -d; | |
1645 | ||
1646 | if (d > 3*sysctl_sched_latency) | |
1647 | schedstat_inc(cfs_rq, nr_spread_over); | |
1648 | #endif | |
1649 | } | |
1650 | ||
aeb73b04 PZ |
1651 | static void |
1652 | place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) | |
1653 | { | |
1af5f730 | 1654 | u64 vruntime = cfs_rq->min_vruntime; |
94dfb5e7 | 1655 | |
2cb8600e PZ |
1656 | /* |
1657 | * The 'current' period is already promised to the current tasks, | |
1658 | * however the extra weight of the new task will slow them down a | |
1659 | * little, place the new task so that it fits in the slot that | |
1660 | * stays open at the end. | |
1661 | */ | |
94dfb5e7 | 1662 | if (initial && sched_feat(START_DEBIT)) |
f9c0b095 | 1663 | vruntime += sched_vslice(cfs_rq, se); |
aeb73b04 | 1664 | |
a2e7a7eb | 1665 | /* sleeps up to a single latency don't count. */ |
5ca9880c | 1666 | if (!initial) { |
a2e7a7eb | 1667 | unsigned long thresh = sysctl_sched_latency; |
a7be37ac | 1668 | |
a2e7a7eb MG |
1669 | /* |
1670 | * Halve their sleep time's effect, to allow | |
1671 | * for a gentler effect of sleepers: | |
1672 | */ | |
1673 | if (sched_feat(GENTLE_FAIR_SLEEPERS)) | |
1674 | thresh >>= 1; | |
51e0304c | 1675 | |
a2e7a7eb | 1676 | vruntime -= thresh; |
aeb73b04 PZ |
1677 | } |
1678 | ||
b5d9d734 MG |
1679 | /* ensure we never gain time by being placed backwards. */ |
1680 | vruntime = max_vruntime(se->vruntime, vruntime); | |
1681 | ||
67e9fb2a | 1682 | se->vruntime = vruntime; |
aeb73b04 PZ |
1683 | } |
1684 | ||
d3d9dc33 PT |
1685 | static void check_enqueue_throttle(struct cfs_rq *cfs_rq); |
1686 | ||
bf0f6f24 | 1687 | static void |
88ec22d3 | 1688 | enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) |
bf0f6f24 | 1689 | { |
88ec22d3 PZ |
1690 | /* |
1691 | * Update the normalized vruntime before updating min_vruntime | |
1692 | * through callig update_curr(). | |
1693 | */ | |
371fd7e7 | 1694 | if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING)) |
88ec22d3 PZ |
1695 | se->vruntime += cfs_rq->min_vruntime; |
1696 | ||
bf0f6f24 | 1697 | /* |
a2a2d680 | 1698 | * Update run-time statistics of the 'current'. |
bf0f6f24 | 1699 | */ |
b7cc0896 | 1700 | update_curr(cfs_rq); |
f269ae04 | 1701 | enqueue_entity_load_avg(cfs_rq, se, flags & ENQUEUE_WAKEUP); |
17bc14b7 LT |
1702 | account_entity_enqueue(cfs_rq, se); |
1703 | update_cfs_shares(cfs_rq); | |
bf0f6f24 | 1704 | |
88ec22d3 | 1705 | if (flags & ENQUEUE_WAKEUP) { |
aeb73b04 | 1706 | place_entity(cfs_rq, se, 0); |
2396af69 | 1707 | enqueue_sleeper(cfs_rq, se); |
e9acbff6 | 1708 | } |
bf0f6f24 | 1709 | |
d2417e5a | 1710 | update_stats_enqueue(cfs_rq, se); |
ddc97297 | 1711 | check_spread(cfs_rq, se); |
83b699ed SV |
1712 | if (se != cfs_rq->curr) |
1713 | __enqueue_entity(cfs_rq, se); | |
2069dd75 | 1714 | se->on_rq = 1; |
3d4b47b4 | 1715 | |
d3d9dc33 | 1716 | if (cfs_rq->nr_running == 1) { |
3d4b47b4 | 1717 | list_add_leaf_cfs_rq(cfs_rq); |
d3d9dc33 PT |
1718 | check_enqueue_throttle(cfs_rq); |
1719 | } | |
bf0f6f24 IM |
1720 | } |
1721 | ||
2c13c919 | 1722 | static void __clear_buddies_last(struct sched_entity *se) |
2002c695 | 1723 | { |
2c13c919 RR |
1724 | for_each_sched_entity(se) { |
1725 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
1726 | if (cfs_rq->last == se) | |
1727 | cfs_rq->last = NULL; | |
1728 | else | |
1729 | break; | |
1730 | } | |
1731 | } | |
2002c695 | 1732 | |
2c13c919 RR |
1733 | static void __clear_buddies_next(struct sched_entity *se) |
1734 | { | |
1735 | for_each_sched_entity(se) { | |
1736 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
1737 | if (cfs_rq->next == se) | |
1738 | cfs_rq->next = NULL; | |
1739 | else | |
1740 | break; | |
1741 | } | |
2002c695 PZ |
1742 | } |
1743 | ||
ac53db59 RR |
1744 | static void __clear_buddies_skip(struct sched_entity *se) |
1745 | { | |
1746 | for_each_sched_entity(se) { | |
1747 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
1748 | if (cfs_rq->skip == se) | |
1749 | cfs_rq->skip = NULL; | |
1750 | else | |
1751 | break; | |
1752 | } | |
1753 | } | |
1754 | ||
a571bbea PZ |
1755 | static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se) |
1756 | { | |
2c13c919 RR |
1757 | if (cfs_rq->last == se) |
1758 | __clear_buddies_last(se); | |
1759 | ||
1760 | if (cfs_rq->next == se) | |
1761 | __clear_buddies_next(se); | |
ac53db59 RR |
1762 | |
1763 | if (cfs_rq->skip == se) | |
1764 | __clear_buddies_skip(se); | |
a571bbea PZ |
1765 | } |
1766 | ||
6c16a6dc | 1767 | static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq); |
d8b4986d | 1768 | |
bf0f6f24 | 1769 | static void |
371fd7e7 | 1770 | dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) |
bf0f6f24 | 1771 | { |
a2a2d680 DA |
1772 | /* |
1773 | * Update run-time statistics of the 'current'. | |
1774 | */ | |
1775 | update_curr(cfs_rq); | |
17bc14b7 | 1776 | dequeue_entity_load_avg(cfs_rq, se, flags & DEQUEUE_SLEEP); |
a2a2d680 | 1777 | |
19b6a2e3 | 1778 | update_stats_dequeue(cfs_rq, se); |
371fd7e7 | 1779 | if (flags & DEQUEUE_SLEEP) { |
67e9fb2a | 1780 | #ifdef CONFIG_SCHEDSTATS |
bf0f6f24 IM |
1781 | if (entity_is_task(se)) { |
1782 | struct task_struct *tsk = task_of(se); | |
1783 | ||
1784 | if (tsk->state & TASK_INTERRUPTIBLE) | |
41acab88 | 1785 | se->statistics.sleep_start = rq_of(cfs_rq)->clock; |
bf0f6f24 | 1786 | if (tsk->state & TASK_UNINTERRUPTIBLE) |
41acab88 | 1787 | se->statistics.block_start = rq_of(cfs_rq)->clock; |
bf0f6f24 | 1788 | } |
db36cc7d | 1789 | #endif |
67e9fb2a PZ |
1790 | } |
1791 | ||
2002c695 | 1792 | clear_buddies(cfs_rq, se); |
4793241b | 1793 | |
83b699ed | 1794 | if (se != cfs_rq->curr) |
30cfdcfc | 1795 | __dequeue_entity(cfs_rq, se); |
17bc14b7 | 1796 | se->on_rq = 0; |
30cfdcfc | 1797 | account_entity_dequeue(cfs_rq, se); |
88ec22d3 PZ |
1798 | |
1799 | /* | |
1800 | * Normalize the entity after updating the min_vruntime because the | |
1801 | * update can refer to the ->curr item and we need to reflect this | |
1802 | * movement in our normalized position. | |
1803 | */ | |
371fd7e7 | 1804 | if (!(flags & DEQUEUE_SLEEP)) |
88ec22d3 | 1805 | se->vruntime -= cfs_rq->min_vruntime; |
1e876231 | 1806 | |
d8b4986d PT |
1807 | /* return excess runtime on last dequeue */ |
1808 | return_cfs_rq_runtime(cfs_rq); | |
1809 | ||
1e876231 | 1810 | update_min_vruntime(cfs_rq); |
17bc14b7 | 1811 | update_cfs_shares(cfs_rq); |
bf0f6f24 IM |
1812 | } |
1813 | ||
1814 | /* | |
1815 | * Preempt the current task with a newly woken task if needed: | |
1816 | */ | |
7c92e54f | 1817 | static void |
2e09bf55 | 1818 | check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr) |
bf0f6f24 | 1819 | { |
11697830 | 1820 | unsigned long ideal_runtime, delta_exec; |
f4cfb33e WX |
1821 | struct sched_entity *se; |
1822 | s64 delta; | |
11697830 | 1823 | |
6d0f0ebd | 1824 | ideal_runtime = sched_slice(cfs_rq, curr); |
11697830 | 1825 | delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime; |
a9f3e2b5 | 1826 | if (delta_exec > ideal_runtime) { |
bf0f6f24 | 1827 | resched_task(rq_of(cfs_rq)->curr); |
a9f3e2b5 MG |
1828 | /* |
1829 | * The current task ran long enough, ensure it doesn't get | |
1830 | * re-elected due to buddy favours. | |
1831 | */ | |
1832 | clear_buddies(cfs_rq, curr); | |
f685ceac MG |
1833 | return; |
1834 | } | |
1835 | ||
1836 | /* | |
1837 | * Ensure that a task that missed wakeup preemption by a | |
1838 | * narrow margin doesn't have to wait for a full slice. | |
1839 | * This also mitigates buddy induced latencies under load. | |
1840 | */ | |
f685ceac MG |
1841 | if (delta_exec < sysctl_sched_min_granularity) |
1842 | return; | |
1843 | ||
f4cfb33e WX |
1844 | se = __pick_first_entity(cfs_rq); |
1845 | delta = curr->vruntime - se->vruntime; | |
f685ceac | 1846 | |
f4cfb33e WX |
1847 | if (delta < 0) |
1848 | return; | |
d7d82944 | 1849 | |
f4cfb33e WX |
1850 | if (delta > ideal_runtime) |
1851 | resched_task(rq_of(cfs_rq)->curr); | |
bf0f6f24 IM |
1852 | } |
1853 | ||
83b699ed | 1854 | static void |
8494f412 | 1855 | set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24 | 1856 | { |
83b699ed SV |
1857 | /* 'current' is not kept within the tree. */ |
1858 | if (se->on_rq) { | |
1859 | /* | |
1860 | * Any task has to be enqueued before it get to execute on | |
1861 | * a CPU. So account for the time it spent waiting on the | |
1862 | * runqueue. | |
1863 | */ | |
1864 | update_stats_wait_end(cfs_rq, se); | |
1865 | __dequeue_entity(cfs_rq, se); | |
1866 | } | |
1867 | ||
79303e9e | 1868 | update_stats_curr_start(cfs_rq, se); |
429d43bc | 1869 | cfs_rq->curr = se; |
eba1ed4b IM |
1870 | #ifdef CONFIG_SCHEDSTATS |
1871 | /* | |
1872 | * Track our maximum slice length, if the CPU's load is at | |
1873 | * least twice that of our own weight (i.e. dont track it | |
1874 | * when there are only lesser-weight tasks around): | |
1875 | */ | |
495eca49 | 1876 | if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) { |
41acab88 | 1877 | se->statistics.slice_max = max(se->statistics.slice_max, |
eba1ed4b IM |
1878 | se->sum_exec_runtime - se->prev_sum_exec_runtime); |
1879 | } | |
1880 | #endif | |
4a55b450 | 1881 | se->prev_sum_exec_runtime = se->sum_exec_runtime; |
bf0f6f24 IM |
1882 | } |
1883 | ||
3f3a4904 PZ |
1884 | static int |
1885 | wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se); | |
1886 | ||
ac53db59 RR |
1887 | /* |
1888 | * Pick the next process, keeping these things in mind, in this order: | |
1889 | * 1) keep things fair between processes/task groups | |
1890 | * 2) pick the "next" process, since someone really wants that to run | |
1891 | * 3) pick the "last" process, for cache locality | |
1892 | * 4) do not run the "skip" process, if something else is available | |
1893 | */ | |
f4b6755f | 1894 | static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq) |
aa2ac252 | 1895 | { |
ac53db59 | 1896 | struct sched_entity *se = __pick_first_entity(cfs_rq); |
f685ceac | 1897 | struct sched_entity *left = se; |
f4b6755f | 1898 | |
ac53db59 RR |
1899 | /* |
1900 | * Avoid running the skip buddy, if running something else can | |
1901 | * be done without getting too unfair. | |
1902 | */ | |
1903 | if (cfs_rq->skip == se) { | |
1904 | struct sched_entity *second = __pick_next_entity(se); | |
1905 | if (second && wakeup_preempt_entity(second, left) < 1) | |
1906 | se = second; | |
1907 | } | |
aa2ac252 | 1908 | |
f685ceac MG |
1909 | /* |
1910 | * Prefer last buddy, try to return the CPU to a preempted task. | |
1911 | */ | |
1912 | if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1) | |
1913 | se = cfs_rq->last; | |
1914 | ||
ac53db59 RR |
1915 | /* |
1916 | * Someone really wants this to run. If it's not unfair, run it. | |
1917 | */ | |
1918 | if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1) | |
1919 | se = cfs_rq->next; | |
1920 | ||
f685ceac | 1921 | clear_buddies(cfs_rq, se); |
4793241b PZ |
1922 | |
1923 | return se; | |
aa2ac252 PZ |
1924 | } |
1925 | ||
d3d9dc33 PT |
1926 | static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq); |
1927 | ||
ab6cde26 | 1928 | static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev) |
bf0f6f24 IM |
1929 | { |
1930 | /* | |
1931 | * If still on the runqueue then deactivate_task() | |
1932 | * was not called and update_curr() has to be done: | |
1933 | */ | |
1934 | if (prev->on_rq) | |
b7cc0896 | 1935 | update_curr(cfs_rq); |
bf0f6f24 | 1936 | |
d3d9dc33 PT |
1937 | /* throttle cfs_rqs exceeding runtime */ |
1938 | check_cfs_rq_runtime(cfs_rq); | |
1939 | ||
ddc97297 | 1940 | check_spread(cfs_rq, prev); |
30cfdcfc | 1941 | if (prev->on_rq) { |
5870db5b | 1942 | update_stats_wait_start(cfs_rq, prev); |
30cfdcfc DA |
1943 | /* Put 'current' back into the tree. */ |
1944 | __enqueue_entity(cfs_rq, prev); | |
9d85f21c | 1945 | /* in !on_rq case, update occurred at dequeue */ |
9ee474f5 | 1946 | update_entity_load_avg(prev, 1); |
30cfdcfc | 1947 | } |
429d43bc | 1948 | cfs_rq->curr = NULL; |
bf0f6f24 IM |
1949 | } |
1950 | ||
8f4d37ec PZ |
1951 | static void |
1952 | entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued) | |
bf0f6f24 | 1953 | { |
bf0f6f24 | 1954 | /* |
30cfdcfc | 1955 | * Update run-time statistics of the 'current'. |
bf0f6f24 | 1956 | */ |
30cfdcfc | 1957 | update_curr(cfs_rq); |
bf0f6f24 | 1958 | |
9d85f21c PT |
1959 | /* |
1960 | * Ensure that runnable average is periodically updated. | |
1961 | */ | |
9ee474f5 | 1962 | update_entity_load_avg(curr, 1); |
aff3e498 | 1963 | update_cfs_rq_blocked_load(cfs_rq, 1); |
9d85f21c | 1964 | |
8f4d37ec PZ |
1965 | #ifdef CONFIG_SCHED_HRTICK |
1966 | /* | |
1967 | * queued ticks are scheduled to match the slice, so don't bother | |
1968 | * validating it and just reschedule. | |
1969 | */ | |
983ed7a6 HH |
1970 | if (queued) { |
1971 | resched_task(rq_of(cfs_rq)->curr); | |
1972 | return; | |
1973 | } | |
8f4d37ec PZ |
1974 | /* |
1975 | * don't let the period tick interfere with the hrtick preemption | |
1976 | */ | |
1977 | if (!sched_feat(DOUBLE_TICK) && | |
1978 | hrtimer_active(&rq_of(cfs_rq)->hrtick_timer)) | |
1979 | return; | |
1980 | #endif | |
1981 | ||
2c2efaed | 1982 | if (cfs_rq->nr_running > 1) |
2e09bf55 | 1983 | check_preempt_tick(cfs_rq, curr); |
bf0f6f24 IM |
1984 | } |
1985 | ||
ab84d31e PT |
1986 | |
1987 | /************************************************** | |
1988 | * CFS bandwidth control machinery | |
1989 | */ | |
1990 | ||
1991 | #ifdef CONFIG_CFS_BANDWIDTH | |
029632fb PZ |
1992 | |
1993 | #ifdef HAVE_JUMP_LABEL | |
c5905afb | 1994 | static struct static_key __cfs_bandwidth_used; |
029632fb PZ |
1995 | |
1996 | static inline bool cfs_bandwidth_used(void) | |
1997 | { | |
c5905afb | 1998 | return static_key_false(&__cfs_bandwidth_used); |
029632fb PZ |
1999 | } |
2000 | ||
2001 | void account_cfs_bandwidth_used(int enabled, int was_enabled) | |
2002 | { | |
2003 | /* only need to count groups transitioning between enabled/!enabled */ | |
2004 | if (enabled && !was_enabled) | |
c5905afb | 2005 | static_key_slow_inc(&__cfs_bandwidth_used); |
029632fb | 2006 | else if (!enabled && was_enabled) |
c5905afb | 2007 | static_key_slow_dec(&__cfs_bandwidth_used); |
029632fb PZ |
2008 | } |
2009 | #else /* HAVE_JUMP_LABEL */ | |
2010 | static bool cfs_bandwidth_used(void) | |
2011 | { | |
2012 | return true; | |
2013 | } | |
2014 | ||
2015 | void account_cfs_bandwidth_used(int enabled, int was_enabled) {} | |
2016 | #endif /* HAVE_JUMP_LABEL */ | |
2017 | ||
ab84d31e PT |
2018 | /* |
2019 | * default period for cfs group bandwidth. | |
2020 | * default: 0.1s, units: nanoseconds | |
2021 | */ | |
2022 | static inline u64 default_cfs_period(void) | |
2023 | { | |
2024 | return 100000000ULL; | |
2025 | } | |
ec12cb7f PT |
2026 | |
2027 | static inline u64 sched_cfs_bandwidth_slice(void) | |
2028 | { | |
2029 | return (u64)sysctl_sched_cfs_bandwidth_slice * NSEC_PER_USEC; | |
2030 | } | |
2031 | ||
a9cf55b2 PT |
2032 | /* |
2033 | * Replenish runtime according to assigned quota and update expiration time. | |
2034 | * We use sched_clock_cpu directly instead of rq->clock to avoid adding | |
2035 | * additional synchronization around rq->lock. | |
2036 | * | |
2037 | * requires cfs_b->lock | |
2038 | */ | |
029632fb | 2039 | void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b) |
a9cf55b2 PT |
2040 | { |
2041 | u64 now; | |
2042 | ||
2043 | if (cfs_b->quota == RUNTIME_INF) | |
2044 | return; | |
2045 | ||
2046 | now = sched_clock_cpu(smp_processor_id()); | |
2047 | cfs_b->runtime = cfs_b->quota; | |
2048 | cfs_b->runtime_expires = now + ktime_to_ns(cfs_b->period); | |
2049 | } | |
2050 | ||
029632fb PZ |
2051 | static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg) |
2052 | { | |
2053 | return &tg->cfs_bandwidth; | |
2054 | } | |
2055 | ||
f1b17280 PT |
2056 | /* rq->task_clock normalized against any time this cfs_rq has spent throttled */ |
2057 | static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq) | |
2058 | { | |
2059 | if (unlikely(cfs_rq->throttle_count)) | |
2060 | return cfs_rq->throttled_clock_task; | |
2061 | ||
2062 | return rq_of(cfs_rq)->clock_task - cfs_rq->throttled_clock_task_time; | |
2063 | } | |
2064 | ||
85dac906 PT |
2065 | /* returns 0 on failure to allocate runtime */ |
2066 | static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq) | |
ec12cb7f PT |
2067 | { |
2068 | struct task_group *tg = cfs_rq->tg; | |
2069 | struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg); | |
a9cf55b2 | 2070 | u64 amount = 0, min_amount, expires; |
ec12cb7f PT |
2071 | |
2072 | /* note: this is a positive sum as runtime_remaining <= 0 */ | |
2073 | min_amount = sched_cfs_bandwidth_slice() - cfs_rq->runtime_remaining; | |
2074 | ||
2075 | raw_spin_lock(&cfs_b->lock); | |
2076 | if (cfs_b->quota == RUNTIME_INF) | |
2077 | amount = min_amount; | |
58088ad0 | 2078 | else { |
a9cf55b2 PT |
2079 | /* |
2080 | * If the bandwidth pool has become inactive, then at least one | |
2081 | * period must have elapsed since the last consumption. | |
2082 | * Refresh the global state and ensure bandwidth timer becomes | |
2083 | * active. | |
2084 | */ | |
2085 | if (!cfs_b->timer_active) { | |
2086 | __refill_cfs_bandwidth_runtime(cfs_b); | |
58088ad0 | 2087 | __start_cfs_bandwidth(cfs_b); |
a9cf55b2 | 2088 | } |
58088ad0 PT |
2089 | |
2090 | if (cfs_b->runtime > 0) { | |
2091 | amount = min(cfs_b->runtime, min_amount); | |
2092 | cfs_b->runtime -= amount; | |
2093 | cfs_b->idle = 0; | |
2094 | } | |
ec12cb7f | 2095 | } |
a9cf55b2 | 2096 | expires = cfs_b->runtime_expires; |
ec12cb7f PT |
2097 | raw_spin_unlock(&cfs_b->lock); |
2098 | ||
2099 | cfs_rq->runtime_remaining += amount; | |
a9cf55b2 PT |
2100 | /* |
2101 | * we may have advanced our local expiration to account for allowed | |
2102 | * spread between our sched_clock and the one on which runtime was | |
2103 | * issued. | |
2104 | */ | |
2105 | if ((s64)(expires - cfs_rq->runtime_expires) > 0) | |
2106 | cfs_rq->runtime_expires = expires; | |
85dac906 PT |
2107 | |
2108 | return cfs_rq->runtime_remaining > 0; | |
ec12cb7f PT |
2109 | } |
2110 | ||
a9cf55b2 PT |
2111 | /* |
2112 | * Note: This depends on the synchronization provided by sched_clock and the | |
2113 | * fact that rq->clock snapshots this value. | |
2114 | */ | |
2115 | static void expire_cfs_rq_runtime(struct cfs_rq *cfs_rq) | |
ec12cb7f | 2116 | { |
a9cf55b2 PT |
2117 | struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); |
2118 | struct rq *rq = rq_of(cfs_rq); | |
2119 | ||
2120 | /* if the deadline is ahead of our clock, nothing to do */ | |
2121 | if (likely((s64)(rq->clock - cfs_rq->runtime_expires) < 0)) | |
ec12cb7f PT |
2122 | return; |
2123 | ||
a9cf55b2 PT |
2124 | if (cfs_rq->runtime_remaining < 0) |
2125 | return; | |
2126 | ||
2127 | /* | |
2128 | * If the local deadline has passed we have to consider the | |
2129 | * possibility that our sched_clock is 'fast' and the global deadline | |
2130 | * has not truly expired. | |
2131 | * | |
2132 | * Fortunately we can check determine whether this the case by checking | |
2133 | * whether the global deadline has advanced. | |
2134 | */ | |
2135 | ||
2136 | if ((s64)(cfs_rq->runtime_expires - cfs_b->runtime_expires) >= 0) { | |
2137 | /* extend local deadline, drift is bounded above by 2 ticks */ | |
2138 | cfs_rq->runtime_expires += TICK_NSEC; | |
2139 | } else { | |
2140 | /* global deadline is ahead, expiration has passed */ | |
2141 | cfs_rq->runtime_remaining = 0; | |
2142 | } | |
2143 | } | |
2144 | ||
2145 | static void __account_cfs_rq_runtime(struct cfs_rq *cfs_rq, | |
2146 | unsigned long delta_exec) | |
2147 | { | |
2148 | /* dock delta_exec before expiring quota (as it could span periods) */ | |
ec12cb7f | 2149 | cfs_rq->runtime_remaining -= delta_exec; |
a9cf55b2 PT |
2150 | expire_cfs_rq_runtime(cfs_rq); |
2151 | ||
2152 | if (likely(cfs_rq->runtime_remaining > 0)) | |
ec12cb7f PT |
2153 | return; |
2154 | ||
85dac906 PT |
2155 | /* |
2156 | * if we're unable to extend our runtime we resched so that the active | |
2157 | * hierarchy can be throttled | |
2158 | */ | |
2159 | if (!assign_cfs_rq_runtime(cfs_rq) && likely(cfs_rq->curr)) | |
2160 | resched_task(rq_of(cfs_rq)->curr); | |
ec12cb7f PT |
2161 | } |
2162 | ||
6c16a6dc PZ |
2163 | static __always_inline |
2164 | void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec) | |
ec12cb7f | 2165 | { |
56f570e5 | 2166 | if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled) |
ec12cb7f PT |
2167 | return; |
2168 | ||
2169 | __account_cfs_rq_runtime(cfs_rq, delta_exec); | |
2170 | } | |
2171 | ||
85dac906 PT |
2172 | static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq) |
2173 | { | |
56f570e5 | 2174 | return cfs_bandwidth_used() && cfs_rq->throttled; |
85dac906 PT |
2175 | } |
2176 | ||
64660c86 PT |
2177 | /* check whether cfs_rq, or any parent, is throttled */ |
2178 | static inline int throttled_hierarchy(struct cfs_rq *cfs_rq) | |
2179 | { | |
56f570e5 | 2180 | return cfs_bandwidth_used() && cfs_rq->throttle_count; |
64660c86 PT |
2181 | } |
2182 | ||
2183 | /* | |
2184 | * Ensure that neither of the group entities corresponding to src_cpu or | |
2185 | * dest_cpu are members of a throttled hierarchy when performing group | |
2186 | * load-balance operations. | |
2187 | */ | |
2188 | static inline int throttled_lb_pair(struct task_group *tg, | |
2189 | int src_cpu, int dest_cpu) | |
2190 | { | |
2191 | struct cfs_rq *src_cfs_rq, *dest_cfs_rq; | |
2192 | ||
2193 | src_cfs_rq = tg->cfs_rq[src_cpu]; | |
2194 | dest_cfs_rq = tg->cfs_rq[dest_cpu]; | |
2195 | ||
2196 | return throttled_hierarchy(src_cfs_rq) || | |
2197 | throttled_hierarchy(dest_cfs_rq); | |
2198 | } | |
2199 | ||
2200 | /* updated child weight may affect parent so we have to do this bottom up */ | |
2201 | static int tg_unthrottle_up(struct task_group *tg, void *data) | |
2202 | { | |
2203 | struct rq *rq = data; | |
2204 | struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)]; | |
2205 | ||
2206 | cfs_rq->throttle_count--; | |
2207 | #ifdef CONFIG_SMP | |
2208 | if (!cfs_rq->throttle_count) { | |
f1b17280 PT |
2209 | /* adjust cfs_rq_clock_task() */ |
2210 | cfs_rq->throttled_clock_task_time += rq->clock_task - | |
2211 | cfs_rq->throttled_clock_task; | |
64660c86 PT |
2212 | } |
2213 | #endif | |
2214 | ||
2215 | return 0; | |
2216 | } | |
2217 | ||
2218 | static int tg_throttle_down(struct task_group *tg, void *data) | |
2219 | { | |
2220 | struct rq *rq = data; | |
2221 | struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)]; | |
2222 | ||
82958366 PT |
2223 | /* group is entering throttled state, stop time */ |
2224 | if (!cfs_rq->throttle_count) | |
f1b17280 | 2225 | cfs_rq->throttled_clock_task = rq->clock_task; |
64660c86 PT |
2226 | cfs_rq->throttle_count++; |
2227 | ||
2228 | return 0; | |
2229 | } | |
2230 | ||
d3d9dc33 | 2231 | static void throttle_cfs_rq(struct cfs_rq *cfs_rq) |
85dac906 PT |
2232 | { |
2233 | struct rq *rq = rq_of(cfs_rq); | |
2234 | struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); | |
2235 | struct sched_entity *se; | |
2236 | long task_delta, dequeue = 1; | |
2237 | ||
2238 | se = cfs_rq->tg->se[cpu_of(rq_of(cfs_rq))]; | |
2239 | ||
f1b17280 | 2240 | /* freeze hierarchy runnable averages while throttled */ |
64660c86 PT |
2241 | rcu_read_lock(); |
2242 | walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq); | |
2243 | rcu_read_unlock(); | |
85dac906 PT |
2244 | |
2245 | task_delta = cfs_rq->h_nr_running; | |
2246 | for_each_sched_entity(se) { | |
2247 | struct cfs_rq *qcfs_rq = cfs_rq_of(se); | |
2248 | /* throttled entity or throttle-on-deactivate */ | |
2249 | if (!se->on_rq) | |
2250 | break; | |
2251 | ||
2252 | if (dequeue) | |
2253 | dequeue_entity(qcfs_rq, se, DEQUEUE_SLEEP); | |
2254 | qcfs_rq->h_nr_running -= task_delta; | |
2255 | ||
2256 | if (qcfs_rq->load.weight) | |
2257 | dequeue = 0; | |
2258 | } | |
2259 | ||
2260 | if (!se) | |
2261 | rq->nr_running -= task_delta; | |
2262 | ||
2263 | cfs_rq->throttled = 1; | |
f1b17280 | 2264 | cfs_rq->throttled_clock = rq->clock; |
85dac906 PT |
2265 | raw_spin_lock(&cfs_b->lock); |
2266 | list_add_tail_rcu(&cfs_rq->throttled_list, &cfs_b->throttled_cfs_rq); | |
2267 | raw_spin_unlock(&cfs_b->lock); | |
2268 | } | |
2269 | ||
029632fb | 2270 | void unthrottle_cfs_rq(struct cfs_rq *cfs_rq) |
671fd9da PT |
2271 | { |
2272 | struct rq *rq = rq_of(cfs_rq); | |
2273 | struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); | |
2274 | struct sched_entity *se; | |
2275 | int enqueue = 1; | |
2276 | long task_delta; | |
2277 | ||
2278 | se = cfs_rq->tg->se[cpu_of(rq_of(cfs_rq))]; | |
2279 | ||
2280 | cfs_rq->throttled = 0; | |
2281 | raw_spin_lock(&cfs_b->lock); | |
f1b17280 | 2282 | cfs_b->throttled_time += rq->clock - cfs_rq->throttled_clock; |
671fd9da PT |
2283 | list_del_rcu(&cfs_rq->throttled_list); |
2284 | raw_spin_unlock(&cfs_b->lock); | |
2285 | ||
64660c86 PT |
2286 | update_rq_clock(rq); |
2287 | /* update hierarchical throttle state */ | |
2288 | walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq); | |
2289 | ||
671fd9da PT |
2290 | if (!cfs_rq->load.weight) |
2291 | return; | |
2292 | ||
2293 | task_delta = cfs_rq->h_nr_running; | |
2294 | for_each_sched_entity(se) { | |
2295 | if (se->on_rq) | |
2296 | enqueue = 0; | |
2297 | ||
2298 | cfs_rq = cfs_rq_of(se); | |
2299 | if (enqueue) | |
2300 | enqueue_entity(cfs_rq, se, ENQUEUE_WAKEUP); | |
2301 | cfs_rq->h_nr_running += task_delta; | |
2302 | ||
2303 | if (cfs_rq_throttled(cfs_rq)) | |
2304 | break; | |
2305 | } | |
2306 | ||
2307 | if (!se) | |
2308 | rq->nr_running += task_delta; | |
2309 | ||
2310 | /* determine whether we need to wake up potentially idle cpu */ | |
2311 | if (rq->curr == rq->idle && rq->cfs.nr_running) | |
2312 | resched_task(rq->curr); | |
2313 | } | |
2314 | ||
2315 | static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, | |
2316 | u64 remaining, u64 expires) | |
2317 | { | |
2318 | struct cfs_rq *cfs_rq; | |
2319 | u64 runtime = remaining; | |
2320 | ||
2321 | rcu_read_lock(); | |
2322 | list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq, | |
2323 | throttled_list) { | |
2324 | struct rq *rq = rq_of(cfs_rq); | |
2325 | ||
2326 | raw_spin_lock(&rq->lock); | |
2327 | if (!cfs_rq_throttled(cfs_rq)) | |
2328 | goto next; | |
2329 | ||
2330 | runtime = -cfs_rq->runtime_remaining + 1; | |
2331 | if (runtime > remaining) | |
2332 | runtime = remaining; | |
2333 | remaining -= runtime; | |
2334 | ||
2335 | cfs_rq->runtime_remaining += runtime; | |
2336 | cfs_rq->runtime_expires = expires; | |
2337 | ||
2338 | /* we check whether we're throttled above */ | |
2339 | if (cfs_rq->runtime_remaining > 0) | |
2340 | unthrottle_cfs_rq(cfs_rq); | |
2341 | ||
2342 | next: | |
2343 | raw_spin_unlock(&rq->lock); | |
2344 | ||
2345 | if (!remaining) | |
2346 | break; | |
2347 | } | |
2348 | rcu_read_unlock(); | |
2349 | ||
2350 | return remaining; | |
2351 | } | |
2352 | ||
58088ad0 PT |
2353 | /* |
2354 | * Responsible for refilling a task_group's bandwidth and unthrottling its | |
2355 | * cfs_rqs as appropriate. If there has been no activity within the last | |
2356 | * period the timer is deactivated until scheduling resumes; cfs_b->idle is | |
2357 | * used to track this state. | |
2358 | */ | |
2359 | static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun) | |
2360 | { | |
671fd9da PT |
2361 | u64 runtime, runtime_expires; |
2362 | int idle = 1, throttled; | |
58088ad0 PT |
2363 | |
2364 | raw_spin_lock(&cfs_b->lock); | |
2365 | /* no need to continue the timer with no bandwidth constraint */ | |
2366 | if (cfs_b->quota == RUNTIME_INF) | |
2367 | goto out_unlock; | |
2368 | ||
671fd9da PT |
2369 | throttled = !list_empty(&cfs_b->throttled_cfs_rq); |
2370 | /* idle depends on !throttled (for the case of a large deficit) */ | |
2371 | idle = cfs_b->idle && !throttled; | |
e8da1b18 | 2372 | cfs_b->nr_periods += overrun; |
671fd9da | 2373 | |
a9cf55b2 PT |
2374 | /* if we're going inactive then everything else can be deferred */ |
2375 | if (idle) | |
2376 | goto out_unlock; | |
2377 | ||
2378 | __refill_cfs_bandwidth_runtime(cfs_b); | |
2379 | ||
671fd9da PT |
2380 | if (!throttled) { |
2381 | /* mark as potentially idle for the upcoming period */ | |
2382 | cfs_b->idle = 1; | |
2383 | goto out_unlock; | |
2384 | } | |
2385 | ||
e8da1b18 NR |
2386 | /* account preceding periods in which throttling occurred */ |
2387 | cfs_b->nr_throttled += overrun; | |
2388 | ||
671fd9da PT |
2389 | /* |
2390 | * There are throttled entities so we must first use the new bandwidth | |
2391 | * to unthrottle them before making it generally available. This | |
2392 | * ensures that all existing debts will be paid before a new cfs_rq is | |
2393 | * allowed to run. | |
2394 | */ | |
2395 | runtime = cfs_b->runtime; | |
2396 | runtime_expires = cfs_b->runtime_expires; | |
2397 | cfs_b->runtime = 0; | |
2398 | ||
2399 | /* | |
2400 | * This check is repeated as we are holding onto the new bandwidth | |
2401 | * while we unthrottle. This can potentially race with an unthrottled | |
2402 | * group trying to acquire new bandwidth from the global pool. | |
2403 | */ | |
2404 | while (throttled && runtime > 0) { | |
2405 | raw_spin_unlock(&cfs_b->lock); | |
2406 | /* we can't nest cfs_b->lock while distributing bandwidth */ | |
2407 | runtime = distribute_cfs_runtime(cfs_b, runtime, | |
2408 | runtime_expires); | |
2409 | raw_spin_lock(&cfs_b->lock); | |
2410 | ||
2411 | throttled = !list_empty(&cfs_b->throttled_cfs_rq); | |
2412 | } | |
58088ad0 | 2413 | |
671fd9da PT |
2414 | /* return (any) remaining runtime */ |
2415 | cfs_b->runtime = runtime; | |
2416 | /* | |
2417 | * While we are ensured activity in the period following an | |
2418 | * unthrottle, this also covers the case in which the new bandwidth is | |
2419 | * insufficient to cover the existing bandwidth deficit. (Forcing the | |
2420 | * timer to remain active while there are any throttled entities.) | |
2421 | */ | |
2422 | cfs_b->idle = 0; | |
58088ad0 PT |
2423 | out_unlock: |
2424 | if (idle) | |
2425 | cfs_b->timer_active = 0; | |
2426 | raw_spin_unlock(&cfs_b->lock); | |
2427 | ||
2428 | return idle; | |
2429 | } | |
d3d9dc33 | 2430 | |
d8b4986d PT |
2431 | /* a cfs_rq won't donate quota below this amount */ |
2432 | static const u64 min_cfs_rq_runtime = 1 * NSEC_PER_MSEC; | |
2433 | /* minimum remaining period time to redistribute slack quota */ | |
2434 | static const u64 min_bandwidth_expiration = 2 * NSEC_PER_MSEC; | |
2435 | /* how long we wait to gather additional slack before distributing */ | |
2436 | static const u64 cfs_bandwidth_slack_period = 5 * NSEC_PER_MSEC; | |
2437 | ||
2438 | /* are we near the end of the current quota period? */ | |
2439 | static int runtime_refresh_within(struct cfs_bandwidth *cfs_b, u64 min_expire) | |
2440 | { | |
2441 | struct hrtimer *refresh_timer = &cfs_b->period_timer; | |
2442 | u64 remaining; | |
2443 | ||
2444 | /* if the call-back is running a quota refresh is already occurring */ | |
2445 | if (hrtimer_callback_running(refresh_timer)) | |
2446 | return 1; | |
2447 | ||
2448 | /* is a quota refresh about to occur? */ | |
2449 | remaining = ktime_to_ns(hrtimer_expires_remaining(refresh_timer)); | |
2450 | if (remaining < min_expire) | |
2451 | return 1; | |
2452 | ||
2453 | return 0; | |
2454 | } | |
2455 | ||
2456 | static void start_cfs_slack_bandwidth(struct cfs_bandwidth *cfs_b) | |
2457 | { | |
2458 | u64 min_left = cfs_bandwidth_slack_period + min_bandwidth_expiration; | |
2459 | ||
2460 | /* if there's a quota refresh soon don't bother with slack */ | |
2461 | if (runtime_refresh_within(cfs_b, min_left)) | |
2462 | return; | |
2463 | ||
2464 | start_bandwidth_timer(&cfs_b->slack_timer, | |
2465 | ns_to_ktime(cfs_bandwidth_slack_period)); | |
2466 | } | |
2467 | ||
2468 | /* we know any runtime found here is valid as update_curr() precedes return */ | |
2469 | static void __return_cfs_rq_runtime(struct cfs_rq *cfs_rq) | |
2470 | { | |
2471 | struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); | |
2472 | s64 slack_runtime = cfs_rq->runtime_remaining - min_cfs_rq_runtime; | |
2473 | ||
2474 | if (slack_runtime <= 0) | |
2475 | return; | |
2476 | ||
2477 | raw_spin_lock(&cfs_b->lock); | |
2478 | if (cfs_b->quota != RUNTIME_INF && | |
2479 | cfs_rq->runtime_expires == cfs_b->runtime_expires) { | |
2480 | cfs_b->runtime += slack_runtime; | |
2481 | ||
2482 | /* we are under rq->lock, defer unthrottling using a timer */ | |
2483 | if (cfs_b->runtime > sched_cfs_bandwidth_slice() && | |
2484 | !list_empty(&cfs_b->throttled_cfs_rq)) | |
2485 | start_cfs_slack_bandwidth(cfs_b); | |
2486 | } | |
2487 | raw_spin_unlock(&cfs_b->lock); | |
2488 | ||
2489 | /* even if it's not valid for return we don't want to try again */ | |
2490 | cfs_rq->runtime_remaining -= slack_runtime; | |
2491 | } | |
2492 | ||
2493 | static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) | |
2494 | { | |
56f570e5 PT |
2495 | if (!cfs_bandwidth_used()) |
2496 | return; | |
2497 | ||
fccfdc6f | 2498 | if (!cfs_rq->runtime_enabled || cfs_rq->nr_running) |
d8b4986d PT |
2499 | return; |
2500 | ||
2501 | __return_cfs_rq_runtime(cfs_rq); | |
2502 | } | |
2503 | ||
2504 | /* | |
2505 | * This is done with a timer (instead of inline with bandwidth return) since | |
2506 | * it's necessary to juggle rq->locks to unthrottle their respective cfs_rqs. | |
2507 | */ | |
2508 | static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b) | |
2509 | { | |
2510 | u64 runtime = 0, slice = sched_cfs_bandwidth_slice(); | |
2511 | u64 expires; | |
2512 | ||
2513 | /* confirm we're still not at a refresh boundary */ | |
2514 | if (runtime_refresh_within(cfs_b, min_bandwidth_expiration)) | |
2515 | return; | |
2516 | ||
2517 | raw_spin_lock(&cfs_b->lock); | |
2518 | if (cfs_b->quota != RUNTIME_INF && cfs_b->runtime > slice) { | |
2519 | runtime = cfs_b->runtime; | |
2520 | cfs_b->runtime = 0; | |
2521 | } | |
2522 | expires = cfs_b->runtime_expires; | |
2523 | raw_spin_unlock(&cfs_b->lock); | |
2524 | ||
2525 | if (!runtime) | |
2526 | return; | |
2527 | ||
2528 | runtime = distribute_cfs_runtime(cfs_b, runtime, expires); | |
2529 | ||
2530 | raw_spin_lock(&cfs_b->lock); | |
2531 | if (expires == cfs_b->runtime_expires) | |
2532 | cfs_b->runtime = runtime; | |
2533 | raw_spin_unlock(&cfs_b->lock); | |
2534 | } | |
2535 | ||
d3d9dc33 PT |
2536 | /* |
2537 | * When a group wakes up we want to make sure that its quota is not already | |
2538 | * expired/exceeded, otherwise it may be allowed to steal additional ticks of | |
2539 | * runtime as update_curr() throttling can not not trigger until it's on-rq. | |
2540 | */ | |
2541 | static void check_enqueue_throttle(struct cfs_rq *cfs_rq) | |
2542 | { | |
56f570e5 PT |
2543 | if (!cfs_bandwidth_used()) |
2544 | return; | |
2545 | ||
d3d9dc33 PT |
2546 | /* an active group must be handled by the update_curr()->put() path */ |
2547 | if (!cfs_rq->runtime_enabled || cfs_rq->curr) | |
2548 | return; | |
2549 | ||
2550 | /* ensure the group is not already throttled */ | |
2551 | if (cfs_rq_throttled(cfs_rq)) | |
2552 | return; | |
2553 | ||
2554 | /* update runtime allocation */ | |
2555 | account_cfs_rq_runtime(cfs_rq, 0); | |
2556 | if (cfs_rq->runtime_remaining <= 0) | |
2557 | throttle_cfs_rq(cfs_rq); | |
2558 | } | |
2559 | ||
2560 | /* conditionally throttle active cfs_rq's from put_prev_entity() */ | |
2561 | static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) | |
2562 | { | |
56f570e5 PT |
2563 | if (!cfs_bandwidth_used()) |
2564 | return; | |
2565 | ||
d3d9dc33 PT |
2566 | if (likely(!cfs_rq->runtime_enabled || cfs_rq->runtime_remaining > 0)) |
2567 | return; | |
2568 | ||
2569 | /* | |
2570 | * it's possible for a throttled entity to be forced into a running | |
2571 | * state (e.g. set_curr_task), in this case we're finished. | |
2572 | */ | |
2573 | if (cfs_rq_throttled(cfs_rq)) | |
2574 | return; | |
2575 | ||
2576 | throttle_cfs_rq(cfs_rq); | |
2577 | } | |
029632fb PZ |
2578 | |
2579 | static inline u64 default_cfs_period(void); | |
2580 | static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun); | |
2581 | static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b); | |
2582 | ||
2583 | static enum hrtimer_restart sched_cfs_slack_timer(struct hrtimer *timer) | |
2584 | { | |
2585 | struct cfs_bandwidth *cfs_b = | |
2586 | container_of(timer, struct cfs_bandwidth, slack_timer); | |
2587 | do_sched_cfs_slack_timer(cfs_b); | |
2588 | ||
2589 | return HRTIMER_NORESTART; | |
2590 | } | |
2591 | ||
2592 | static enum hrtimer_restart sched_cfs_period_timer(struct hrtimer *timer) | |
2593 | { | |
2594 | struct cfs_bandwidth *cfs_b = | |
2595 | container_of(timer, struct cfs_bandwidth, period_timer); | |
2596 | ktime_t now; | |
2597 | int overrun; | |
2598 | int idle = 0; | |
2599 | ||
2600 | for (;;) { | |
2601 | now = hrtimer_cb_get_time(timer); | |
2602 | overrun = hrtimer_forward(timer, now, cfs_b->period); | |
2603 | ||
2604 | if (!overrun) | |
2605 | break; | |
2606 | ||
2607 | idle = do_sched_cfs_period_timer(cfs_b, overrun); | |
2608 | } | |
2609 | ||
2610 | return idle ? HRTIMER_NORESTART : HRTIMER_RESTART; | |
2611 | } | |
2612 | ||
2613 | void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b) | |
2614 | { | |
2615 | raw_spin_lock_init(&cfs_b->lock); | |
2616 | cfs_b->runtime = 0; | |
2617 | cfs_b->quota = RUNTIME_INF; | |
2618 | cfs_b->period = ns_to_ktime(default_cfs_period()); | |
2619 | ||
2620 | INIT_LIST_HEAD(&cfs_b->throttled_cfs_rq); | |
2621 | hrtimer_init(&cfs_b->period_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); | |
2622 | cfs_b->period_timer.function = sched_cfs_period_timer; | |
2623 | hrtimer_init(&cfs_b->slack_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); | |
2624 | cfs_b->slack_timer.function = sched_cfs_slack_timer; | |
2625 | } | |
2626 | ||
2627 | static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq) | |
2628 | { | |
2629 | cfs_rq->runtime_enabled = 0; | |
2630 | INIT_LIST_HEAD(&cfs_rq->throttled_list); | |
2631 | } | |
2632 | ||
2633 | /* requires cfs_b->lock, may release to reprogram timer */ | |
2634 | void __start_cfs_bandwidth(struct cfs_bandwidth *cfs_b) | |
2635 | { | |
2636 | /* | |
2637 | * The timer may be active because we're trying to set a new bandwidth | |
2638 | * period or because we're racing with the tear-down path | |
2639 | * (timer_active==0 becomes visible before the hrtimer call-back | |
2640 | * terminates). In either case we ensure that it's re-programmed | |
2641 | */ | |
2642 | while (unlikely(hrtimer_active(&cfs_b->period_timer))) { | |
2643 | raw_spin_unlock(&cfs_b->lock); | |
2644 | /* ensure cfs_b->lock is available while we wait */ | |
2645 | hrtimer_cancel(&cfs_b->period_timer); | |
2646 | ||
2647 | raw_spin_lock(&cfs_b->lock); | |
2648 | /* if someone else restarted the timer then we're done */ | |
2649 | if (cfs_b->timer_active) | |
2650 | return; | |
2651 | } | |
2652 | ||
2653 | cfs_b->timer_active = 1; | |
2654 | start_bandwidth_timer(&cfs_b->period_timer, cfs_b->period); | |
2655 | } | |
2656 | ||
2657 | static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) | |
2658 | { | |
2659 | hrtimer_cancel(&cfs_b->period_timer); | |
2660 | hrtimer_cancel(&cfs_b->slack_timer); | |
2661 | } | |
2662 | ||
a4c96ae3 | 2663 | static void unthrottle_offline_cfs_rqs(struct rq *rq) |
029632fb PZ |
2664 | { |
2665 | struct cfs_rq *cfs_rq; | |
2666 | ||
2667 | for_each_leaf_cfs_rq(rq, cfs_rq) { | |
2668 | struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); | |
2669 | ||
2670 | if (!cfs_rq->runtime_enabled) | |
2671 | continue; | |
2672 | ||
2673 | /* | |
2674 | * clock_task is not advancing so we just need to make sure | |
2675 | * there's some valid quota amount | |
2676 | */ | |
2677 | cfs_rq->runtime_remaining = cfs_b->quota; | |
2678 | if (cfs_rq_throttled(cfs_rq)) | |
2679 | unthrottle_cfs_rq(cfs_rq); | |
2680 | } | |
2681 | } | |
2682 | ||
2683 | #else /* CONFIG_CFS_BANDWIDTH */ | |
f1b17280 PT |
2684 | static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq) |
2685 | { | |
2686 | return rq_of(cfs_rq)->clock_task; | |
2687 | } | |
2688 | ||
2689 | static void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, | |
2690 | unsigned long delta_exec) {} | |
d3d9dc33 PT |
2691 | static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) {} |
2692 | static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {} | |
6c16a6dc | 2693 | static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {} |
85dac906 PT |
2694 | |
2695 | static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq) | |
2696 | { | |
2697 | return 0; | |
2698 | } | |
64660c86 PT |
2699 | |
2700 | static inline int throttled_hierarchy(struct cfs_rq *cfs_rq) | |
2701 | { | |
2702 | return 0; | |
2703 | } | |
2704 | ||
2705 | static inline int throttled_lb_pair(struct task_group *tg, | |
2706 | int src_cpu, int dest_cpu) | |
2707 | { | |
2708 | return 0; | |
2709 | } | |
029632fb PZ |
2710 | |
2711 | void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {} | |
2712 | ||
2713 | #ifdef CONFIG_FAIR_GROUP_SCHED | |
2714 | static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq) {} | |
ab84d31e PT |
2715 | #endif |
2716 | ||
029632fb PZ |
2717 | static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg) |
2718 | { | |
2719 | return NULL; | |
2720 | } | |
2721 | static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {} | |
a4c96ae3 | 2722 | static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {} |
029632fb PZ |
2723 | |
2724 | #endif /* CONFIG_CFS_BANDWIDTH */ | |
2725 | ||
bf0f6f24 IM |
2726 | /************************************************** |
2727 | * CFS operations on tasks: | |
2728 | */ | |
2729 | ||
8f4d37ec PZ |
2730 | #ifdef CONFIG_SCHED_HRTICK |
2731 | static void hrtick_start_fair(struct rq *rq, struct task_struct *p) | |
2732 | { | |
8f4d37ec PZ |
2733 | struct sched_entity *se = &p->se; |
2734 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
2735 | ||
2736 | WARN_ON(task_rq(p) != rq); | |
2737 | ||
b39e66ea | 2738 | if (cfs_rq->nr_running > 1) { |
8f4d37ec PZ |
2739 | u64 slice = sched_slice(cfs_rq, se); |
2740 | u64 ran = se->sum_exec_runtime - se->prev_sum_exec_runtime; | |
2741 | s64 delta = slice - ran; | |
2742 | ||
2743 | if (delta < 0) { | |
2744 | if (rq->curr == p) | |
2745 | resched_task(p); | |
2746 | return; | |
2747 | } | |
2748 | ||
2749 | /* | |
2750 | * Don't schedule slices shorter than 10000ns, that just | |
2751 | * doesn't make sense. Rely on vruntime for fairness. | |
2752 | */ | |
31656519 | 2753 | if (rq->curr != p) |
157124c1 | 2754 | delta = max_t(s64, 10000LL, delta); |
8f4d37ec | 2755 | |
31656519 | 2756 | hrtick_start(rq, delta); |
8f4d37ec PZ |
2757 | } |
2758 | } | |
a4c2f00f PZ |
2759 | |
2760 | /* | |
2761 | * called from enqueue/dequeue and updates the hrtick when the | |
2762 | * current task is from our class and nr_running is low enough | |
2763 | * to matter. | |
2764 | */ | |
2765 | static void hrtick_update(struct rq *rq) | |
2766 | { | |
2767 | struct task_struct *curr = rq->curr; | |
2768 | ||
b39e66ea | 2769 | if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class) |
a4c2f00f PZ |
2770 | return; |
2771 | ||
2772 | if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency) | |
2773 | hrtick_start_fair(rq, curr); | |
2774 | } | |
55e12e5e | 2775 | #else /* !CONFIG_SCHED_HRTICK */ |
8f4d37ec PZ |
2776 | static inline void |
2777 | hrtick_start_fair(struct rq *rq, struct task_struct *p) | |
2778 | { | |
2779 | } | |
a4c2f00f PZ |
2780 | |
2781 | static inline void hrtick_update(struct rq *rq) | |
2782 | { | |
2783 | } | |
8f4d37ec PZ |
2784 | #endif |
2785 | ||
bf0f6f24 IM |
2786 | /* |
2787 | * The enqueue_task method is called before nr_running is | |
2788 | * increased. Here we update the fair scheduling stats and | |
2789 | * then put the task into the rbtree: | |
2790 | */ | |
ea87bb78 | 2791 | static void |
371fd7e7 | 2792 | enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) |
bf0f6f24 IM |
2793 | { |
2794 | struct cfs_rq *cfs_rq; | |
62fb1851 | 2795 | struct sched_entity *se = &p->se; |
bf0f6f24 IM |
2796 | |
2797 | for_each_sched_entity(se) { | |
62fb1851 | 2798 | if (se->on_rq) |
bf0f6f24 IM |
2799 | break; |
2800 | cfs_rq = cfs_rq_of(se); | |
88ec22d3 | 2801 | enqueue_entity(cfs_rq, se, flags); |
85dac906 PT |
2802 | |
2803 | /* | |
2804 | * end evaluation on encountering a throttled cfs_rq | |
2805 | * | |
2806 | * note: in the case of encountering a throttled cfs_rq we will | |
2807 | * post the final h_nr_running increment below. | |
2808 | */ | |
2809 | if (cfs_rq_throttled(cfs_rq)) | |
2810 | break; | |
953bfcd1 | 2811 | cfs_rq->h_nr_running++; |
85dac906 | 2812 | |
88ec22d3 | 2813 | flags = ENQUEUE_WAKEUP; |
bf0f6f24 | 2814 | } |
8f4d37ec | 2815 | |
2069dd75 | 2816 | for_each_sched_entity(se) { |
0f317143 | 2817 | cfs_rq = cfs_rq_of(se); |
953bfcd1 | 2818 | cfs_rq->h_nr_running++; |
2069dd75 | 2819 | |
85dac906 PT |
2820 | if (cfs_rq_throttled(cfs_rq)) |
2821 | break; | |
2822 | ||
17bc14b7 | 2823 | update_cfs_shares(cfs_rq); |
9ee474f5 | 2824 | update_entity_load_avg(se, 1); |
2069dd75 PZ |
2825 | } |
2826 | ||
18bf2805 BS |
2827 | if (!se) { |
2828 | update_rq_runnable_avg(rq, rq->nr_running); | |
85dac906 | 2829 | inc_nr_running(rq); |
18bf2805 | 2830 | } |
a4c2f00f | 2831 | hrtick_update(rq); |
bf0f6f24 IM |
2832 | } |
2833 | ||
2f36825b VP |
2834 | static void set_next_buddy(struct sched_entity *se); |
2835 | ||
bf0f6f24 IM |
2836 | /* |
2837 | * The dequeue_task method is called before nr_running is | |
2838 | * decreased. We remove the task from the rbtree and | |
2839 | * update the fair scheduling stats: | |
2840 | */ | |
371fd7e7 | 2841 | static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) |
bf0f6f24 IM |
2842 | { |
2843 | struct cfs_rq *cfs_rq; | |
62fb1851 | 2844 | struct sched_entity *se = &p->se; |
2f36825b | 2845 | int task_sleep = flags & DEQUEUE_SLEEP; |
bf0f6f24 IM |
2846 | |
2847 | for_each_sched_entity(se) { | |
2848 | cfs_rq = cfs_rq_of(se); | |
371fd7e7 | 2849 | dequeue_entity(cfs_rq, se, flags); |
85dac906 PT |
2850 | |
2851 | /* | |
2852 | * end evaluation on encountering a throttled cfs_rq | |
2853 | * | |
2854 | * note: in the case of encountering a throttled cfs_rq we will | |
2855 | * post the final h_nr_running decrement below. | |
2856 | */ | |
2857 | if (cfs_rq_throttled(cfs_rq)) | |
2858 | break; | |
953bfcd1 | 2859 | cfs_rq->h_nr_running--; |
2069dd75 | 2860 | |
bf0f6f24 | 2861 | /* Don't dequeue parent if it has other entities besides us */ |
2f36825b VP |
2862 | if (cfs_rq->load.weight) { |
2863 | /* | |
2864 | * Bias pick_next to pick a task from this cfs_rq, as | |
2865 | * p is sleeping when it is within its sched_slice. | |
2866 | */ | |
2867 | if (task_sleep && parent_entity(se)) | |
2868 | set_next_buddy(parent_entity(se)); | |
9598c82d PT |
2869 | |
2870 | /* avoid re-evaluating load for this entity */ | |
2871 | se = parent_entity(se); | |
bf0f6f24 | 2872 | break; |
2f36825b | 2873 | } |
371fd7e7 | 2874 | flags |= DEQUEUE_SLEEP; |
bf0f6f24 | 2875 | } |
8f4d37ec | 2876 | |
2069dd75 | 2877 | for_each_sched_entity(se) { |
0f317143 | 2878 | cfs_rq = cfs_rq_of(se); |
953bfcd1 | 2879 | cfs_rq->h_nr_running--; |
2069dd75 | 2880 | |
85dac906 PT |
2881 | if (cfs_rq_throttled(cfs_rq)) |
2882 | break; | |
2883 | ||
17bc14b7 | 2884 | update_cfs_shares(cfs_rq); |
9ee474f5 | 2885 | update_entity_load_avg(se, 1); |
2069dd75 PZ |
2886 | } |
2887 | ||
18bf2805 | 2888 | if (!se) { |
85dac906 | 2889 | dec_nr_running(rq); |
18bf2805 BS |
2890 | update_rq_runnable_avg(rq, 1); |
2891 | } | |
a4c2f00f | 2892 | hrtick_update(rq); |
bf0f6f24 IM |
2893 | } |
2894 | ||
e7693a36 | 2895 | #ifdef CONFIG_SMP |
029632fb PZ |
2896 | /* Used instead of source_load when we know the type == 0 */ |
2897 | static unsigned long weighted_cpuload(const int cpu) | |
2898 | { | |
2899 | return cpu_rq(cpu)->load.weight; | |
2900 | } | |
2901 | ||
2902 | /* | |
2903 | * Return a low guess at the load of a migration-source cpu weighted | |
2904 | * according to the scheduling class and "nice" value. | |
2905 | * | |
2906 | * We want to under-estimate the load of migration sources, to | |
2907 | * balance conservatively. | |
2908 | */ | |
2909 | static unsigned long source_load(int cpu, int type) | |
2910 | { | |
2911 | struct rq *rq = cpu_rq(cpu); | |
2912 | unsigned long total = weighted_cpuload(cpu); | |
2913 | ||
2914 | if (type == 0 || !sched_feat(LB_BIAS)) | |
2915 | return total; | |
2916 | ||
2917 | return min(rq->cpu_load[type-1], total); | |
2918 | } | |
2919 | ||
2920 | /* | |
2921 | * Return a high guess at the load of a migration-target cpu weighted | |
2922 | * according to the scheduling class and "nice" value. | |
2923 | */ | |
2924 | static unsigned long target_load(int cpu, int type) | |
2925 | { | |
2926 | struct rq *rq = cpu_rq(cpu); | |
2927 | unsigned long total = weighted_cpuload(cpu); | |
2928 | ||
2929 | if (type == 0 || !sched_feat(LB_BIAS)) | |
2930 | return total; | |
2931 | ||
2932 | return max(rq->cpu_load[type-1], total); | |
2933 | } | |
2934 | ||
2935 | static unsigned long power_of(int cpu) | |
2936 | { | |
2937 | return cpu_rq(cpu)->cpu_power; | |
2938 | } | |
2939 | ||
2940 | static unsigned long cpu_avg_load_per_task(int cpu) | |
2941 | { | |
2942 | struct rq *rq = cpu_rq(cpu); | |
2943 | unsigned long nr_running = ACCESS_ONCE(rq->nr_running); | |
2944 | ||
2945 | if (nr_running) | |
2946 | return rq->load.weight / nr_running; | |
2947 | ||
2948 | return 0; | |
2949 | } | |
2950 | ||
098fb9db | 2951 | |
74f8e4b2 | 2952 | static void task_waking_fair(struct task_struct *p) |
88ec22d3 PZ |
2953 | { |
2954 | struct sched_entity *se = &p->se; | |
2955 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
3fe1698b PZ |
2956 | u64 min_vruntime; |
2957 | ||
2958 | #ifndef CONFIG_64BIT | |
2959 | u64 min_vruntime_copy; | |
88ec22d3 | 2960 | |
3fe1698b PZ |
2961 | do { |
2962 | min_vruntime_copy = cfs_rq->min_vruntime_copy; | |
2963 | smp_rmb(); | |
2964 | min_vruntime = cfs_rq->min_vruntime; | |
2965 | } while (min_vruntime != min_vruntime_copy); | |
2966 | #else | |
2967 | min_vruntime = cfs_rq->min_vruntime; | |
2968 | #endif | |
88ec22d3 | 2969 | |
3fe1698b | 2970 | se->vruntime -= min_vruntime; |
88ec22d3 PZ |
2971 | } |
2972 | ||
bb3469ac | 2973 | #ifdef CONFIG_FAIR_GROUP_SCHED |
f5bfb7d9 PZ |
2974 | /* |
2975 | * effective_load() calculates the load change as seen from the root_task_group | |
2976 | * | |
2977 | * Adding load to a group doesn't make a group heavier, but can cause movement | |
2978 | * of group shares between cpus. Assuming the shares were perfectly aligned one | |
2979 | * can calculate the shift in shares. | |
cf5f0acf PZ |
2980 | * |
2981 | * Calculate the effective load difference if @wl is added (subtracted) to @tg | |
2982 | * on this @cpu and results in a total addition (subtraction) of @wg to the | |
2983 | * total group weight. | |
2984 | * | |
2985 | * Given a runqueue weight distribution (rw_i) we can compute a shares | |
2986 | * distribution (s_i) using: | |
2987 | * | |
2988 | * s_i = rw_i / \Sum rw_j (1) | |
2989 | * | |
2990 | * Suppose we have 4 CPUs and our @tg is a direct child of the root group and | |
2991 | * has 7 equal weight tasks, distributed as below (rw_i), with the resulting | |
2992 | * shares distribution (s_i): | |
2993 | * | |
2994 | * rw_i = { 2, 4, 1, 0 } | |
2995 | * s_i = { 2/7, 4/7, 1/7, 0 } | |
2996 | * | |
2997 | * As per wake_affine() we're interested in the load of two CPUs (the CPU the | |
2998 | * task used to run on and the CPU the waker is running on), we need to | |
2999 | * compute the effect of waking a task on either CPU and, in case of a sync | |
3000 | * wakeup, compute the effect of the current task going to sleep. | |
3001 | * | |
3002 | * So for a change of @wl to the local @cpu with an overall group weight change | |
3003 | * of @wl we can compute the new shares distribution (s'_i) using: | |
3004 | * | |
3005 | * s'_i = (rw_i + @wl) / (@wg + \Sum rw_j) (2) | |
3006 | * | |
3007 | * Suppose we're interested in CPUs 0 and 1, and want to compute the load | |
3008 | * differences in waking a task to CPU 0. The additional task changes the | |
3009 | * weight and shares distributions like: | |
3010 | * | |
3011 | * rw'_i = { 3, 4, 1, 0 } | |
3012 | * s'_i = { 3/8, 4/8, 1/8, 0 } | |
3013 | * | |
3014 | * We can then compute the difference in effective weight by using: | |
3015 | * | |
3016 | * dw_i = S * (s'_i - s_i) (3) | |
3017 | * | |
3018 | * Where 'S' is the group weight as seen by its parent. | |
3019 | * | |
3020 | * Therefore the effective change in loads on CPU 0 would be 5/56 (3/8 - 2/7) | |
3021 | * times the weight of the group. The effect on CPU 1 would be -4/56 (4/8 - | |
3022 | * 4/7) times the weight of the group. | |
f5bfb7d9 | 3023 | */ |
2069dd75 | 3024 | static long effective_load(struct task_group *tg, int cpu, long wl, long wg) |
bb3469ac | 3025 | { |
4be9daaa | 3026 | struct sched_entity *se = tg->se[cpu]; |
f1d239f7 | 3027 | |
cf5f0acf | 3028 | if (!tg->parent) /* the trivial, non-cgroup case */ |
f1d239f7 PZ |
3029 | return wl; |
3030 | ||
4be9daaa | 3031 | for_each_sched_entity(se) { |
cf5f0acf | 3032 | long w, W; |
4be9daaa | 3033 | |
977dda7c | 3034 | tg = se->my_q->tg; |
bb3469ac | 3035 | |
cf5f0acf PZ |
3036 | /* |
3037 | * W = @wg + \Sum rw_j | |
3038 | */ | |
3039 | W = wg + calc_tg_weight(tg, se->my_q); | |
4be9daaa | 3040 | |
cf5f0acf PZ |
3041 | /* |
3042 | * w = rw_i + @wl | |
3043 | */ | |
3044 | w = se->my_q->load.weight + wl; | |
940959e9 | 3045 | |
cf5f0acf PZ |
3046 | /* |
3047 | * wl = S * s'_i; see (2) | |
3048 | */ | |
3049 | if (W > 0 && w < W) | |
3050 | wl = (w * tg->shares) / W; | |
977dda7c PT |
3051 | else |
3052 | wl = tg->shares; | |
940959e9 | 3053 | |
cf5f0acf PZ |
3054 | /* |
3055 | * Per the above, wl is the new se->load.weight value; since | |
3056 | * those are clipped to [MIN_SHARES, ...) do so now. See | |
3057 | * calc_cfs_shares(). | |
3058 | */ | |
977dda7c PT |
3059 | if (wl < MIN_SHARES) |
3060 | wl = MIN_SHARES; | |
cf5f0acf PZ |
3061 | |
3062 | /* | |
3063 | * wl = dw_i = S * (s'_i - s_i); see (3) | |
3064 | */ | |
977dda7c | 3065 | wl -= se->load.weight; |
cf5f0acf PZ |
3066 | |
3067 | /* | |
3068 | * Recursively apply this logic to all parent groups to compute | |
3069 | * the final effective load change on the root group. Since | |
3070 | * only the @tg group gets extra weight, all parent groups can | |
3071 | * only redistribute existing shares. @wl is the shift in shares | |
3072 | * resulting from this level per the above. | |
3073 | */ | |
4be9daaa | 3074 | wg = 0; |
4be9daaa | 3075 | } |
bb3469ac | 3076 | |
4be9daaa | 3077 | return wl; |
bb3469ac PZ |
3078 | } |
3079 | #else | |
4be9daaa | 3080 | |
83378269 PZ |
3081 | static inline unsigned long effective_load(struct task_group *tg, int cpu, |
3082 | unsigned long wl, unsigned long wg) | |
4be9daaa | 3083 | { |
83378269 | 3084 | return wl; |
bb3469ac | 3085 | } |
4be9daaa | 3086 | |
bb3469ac PZ |
3087 | #endif |
3088 | ||
c88d5910 | 3089 | static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) |
098fb9db | 3090 | { |
e37b6a7b | 3091 | s64 this_load, load; |
c88d5910 | 3092 | int idx, this_cpu, prev_cpu; |
098fb9db | 3093 | unsigned long tl_per_task; |
c88d5910 | 3094 | struct task_group *tg; |
83378269 | 3095 | unsigned long weight; |
b3137bc8 | 3096 | int balanced; |
098fb9db | 3097 | |
c88d5910 PZ |
3098 | idx = sd->wake_idx; |
3099 | this_cpu = smp_processor_id(); | |
3100 | prev_cpu = task_cpu(p); | |
3101 | load = source_load(prev_cpu, idx); | |
3102 | this_load = target_load(this_cpu, idx); | |
098fb9db | 3103 | |
b3137bc8 MG |
3104 | /* |
3105 | * If sync wakeup then subtract the (maximum possible) | |
3106 | * effect of the currently running task from the load | |
3107 | * of the current CPU: | |
3108 | */ | |
83378269 PZ |
3109 | if (sync) { |
3110 | tg = task_group(current); | |
3111 | weight = current->se.load.weight; | |
3112 | ||
c88d5910 | 3113 | this_load += effective_load(tg, this_cpu, -weight, -weight); |
83378269 PZ |
3114 | load += effective_load(tg, prev_cpu, 0, -weight); |
3115 | } | |
b3137bc8 | 3116 | |
83378269 PZ |
3117 | tg = task_group(p); |
3118 | weight = p->se.load.weight; | |
b3137bc8 | 3119 | |
71a29aa7 PZ |
3120 | /* |
3121 | * In low-load situations, where prev_cpu is idle and this_cpu is idle | |
c88d5910 PZ |
3122 | * due to the sync cause above having dropped this_load to 0, we'll |
3123 | * always have an imbalance, but there's really nothing you can do | |
3124 | * about that, so that's good too. | |
71a29aa7 PZ |
3125 | * |
3126 | * Otherwise check if either cpus are near enough in load to allow this | |
3127 | * task to be woken on this_cpu. | |
3128 | */ | |
e37b6a7b PT |
3129 | if (this_load > 0) { |
3130 | s64 this_eff_load, prev_eff_load; | |
e51fd5e2 PZ |
3131 | |
3132 | this_eff_load = 100; | |
3133 | this_eff_load *= power_of(prev_cpu); | |
3134 | this_eff_load *= this_load + | |
3135 | effective_load(tg, this_cpu, weight, weight); | |
3136 | ||
3137 | prev_eff_load = 100 + (sd->imbalance_pct - 100) / 2; | |
3138 | prev_eff_load *= power_of(this_cpu); | |
3139 | prev_eff_load *= load + effective_load(tg, prev_cpu, 0, weight); | |
3140 | ||
3141 | balanced = this_eff_load <= prev_eff_load; | |
3142 | } else | |
3143 | balanced = true; | |
b3137bc8 | 3144 | |
098fb9db | 3145 | /* |
4ae7d5ce IM |
3146 | * If the currently running task will sleep within |
3147 | * a reasonable amount of time then attract this newly | |
3148 | * woken task: | |
098fb9db | 3149 | */ |
2fb7635c PZ |
3150 | if (sync && balanced) |
3151 | return 1; | |
098fb9db | 3152 | |
41acab88 | 3153 | schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts); |
098fb9db IM |
3154 | tl_per_task = cpu_avg_load_per_task(this_cpu); |
3155 | ||
c88d5910 PZ |
3156 | if (balanced || |
3157 | (this_load <= load && | |
3158 | this_load + target_load(prev_cpu, idx) <= tl_per_task)) { | |
098fb9db IM |
3159 | /* |
3160 | * This domain has SD_WAKE_AFFINE and | |
3161 | * p is cache cold in this domain, and | |
3162 | * there is no bad imbalance. | |
3163 | */ | |
c88d5910 | 3164 | schedstat_inc(sd, ttwu_move_affine); |
41acab88 | 3165 | schedstat_inc(p, se.statistics.nr_wakeups_affine); |
098fb9db IM |
3166 | |
3167 | return 1; | |
3168 | } | |
3169 | return 0; | |
3170 | } | |
3171 | ||
aaee1203 PZ |
3172 | /* |
3173 | * find_idlest_group finds and returns the least busy CPU group within the | |
3174 | * domain. | |
3175 | */ | |
3176 | static struct sched_group * | |
78e7ed53 | 3177 | find_idlest_group(struct sched_domain *sd, struct task_struct *p, |
5158f4e4 | 3178 | int this_cpu, int load_idx) |
e7693a36 | 3179 | { |
b3bd3de6 | 3180 | struct sched_group *idlest = NULL, *group = sd->groups; |
aaee1203 | 3181 | unsigned long min_load = ULONG_MAX, this_load = 0; |
aaee1203 | 3182 | int imbalance = 100 + (sd->imbalance_pct-100)/2; |
e7693a36 | 3183 | |
aaee1203 PZ |
3184 | do { |
3185 | unsigned long load, avg_load; | |
3186 | int local_group; | |
3187 | int i; | |
e7693a36 | 3188 | |
aaee1203 PZ |
3189 | /* Skip over this group if it has no CPUs allowed */ |
3190 | if (!cpumask_intersects(sched_group_cpus(group), | |
fa17b507 | 3191 | tsk_cpus_allowed(p))) |
aaee1203 PZ |
3192 | continue; |
3193 | ||
3194 | local_group = cpumask_test_cpu(this_cpu, | |
3195 | sched_group_cpus(group)); | |
3196 | ||
3197 | /* Tally up the load of all CPUs in the group */ | |
3198 | avg_load = 0; | |
3199 | ||
3200 | for_each_cpu(i, sched_group_cpus(group)) { | |
3201 | /* Bias balancing toward cpus of our domain */ | |
3202 | if (local_group) | |
3203 | load = source_load(i, load_idx); | |
3204 | else | |
3205 | load = target_load(i, load_idx); | |
3206 | ||
3207 | avg_load += load; | |
3208 | } | |
3209 | ||
3210 | /* Adjust by relative CPU power of the group */ | |
9c3f75cb | 3211 | avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power; |
aaee1203 PZ |
3212 | |
3213 | if (local_group) { | |
3214 | this_load = avg_load; | |
aaee1203 PZ |
3215 | } else if (avg_load < min_load) { |
3216 | min_load = avg_load; | |
3217 | idlest = group; | |
3218 | } | |
3219 | } while (group = group->next, group != sd->groups); | |
3220 | ||
3221 | if (!idlest || 100*this_load < imbalance*min_load) | |
3222 | return NULL; | |
3223 | return idlest; | |
3224 | } | |
3225 | ||
3226 | /* | |
3227 | * find_idlest_cpu - find the idlest cpu among the cpus in group. | |
3228 | */ | |
3229 | static int | |
3230 | find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu) | |
3231 | { | |
3232 | unsigned long load, min_load = ULONG_MAX; | |
3233 | int idlest = -1; | |
3234 | int i; | |
3235 | ||
3236 | /* Traverse only the allowed CPUs */ | |
fa17b507 | 3237 | for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) { |
aaee1203 PZ |
3238 | load = weighted_cpuload(i); |
3239 | ||
3240 | if (load < min_load || (load == min_load && i == this_cpu)) { | |
3241 | min_load = load; | |
3242 | idlest = i; | |
e7693a36 GH |
3243 | } |
3244 | } | |
3245 | ||
aaee1203 PZ |
3246 | return idlest; |
3247 | } | |
e7693a36 | 3248 | |
a50bde51 PZ |
3249 | /* |
3250 | * Try and locate an idle CPU in the sched_domain. | |
3251 | */ | |
99bd5e2f | 3252 | static int select_idle_sibling(struct task_struct *p, int target) |
a50bde51 PZ |
3253 | { |
3254 | int cpu = smp_processor_id(); | |
3255 | int prev_cpu = task_cpu(p); | |
99bd5e2f | 3256 | struct sched_domain *sd; |
37407ea7 LT |
3257 | struct sched_group *sg; |
3258 | int i; | |
a50bde51 PZ |
3259 | |
3260 | /* | |
99bd5e2f SS |
3261 | * If the task is going to be woken-up on this cpu and if it is |
3262 | * already idle, then it is the right target. | |
a50bde51 | 3263 | */ |
99bd5e2f SS |
3264 | if (target == cpu && idle_cpu(cpu)) |
3265 | return cpu; | |
3266 | ||
3267 | /* | |
3268 | * If the task is going to be woken-up on the cpu where it previously | |
3269 | * ran and if it is currently idle, then it the right target. | |
3270 | */ | |
3271 | if (target == prev_cpu && idle_cpu(prev_cpu)) | |
fe3bcfe1 | 3272 | return prev_cpu; |
a50bde51 PZ |
3273 | |
3274 | /* | |
37407ea7 | 3275 | * Otherwise, iterate the domains and find an elegible idle cpu. |
a50bde51 | 3276 | */ |
518cd623 | 3277 | sd = rcu_dereference(per_cpu(sd_llc, target)); |
970e1789 | 3278 | for_each_lower_domain(sd) { |
37407ea7 LT |
3279 | sg = sd->groups; |
3280 | do { | |
3281 | if (!cpumask_intersects(sched_group_cpus(sg), | |
3282 | tsk_cpus_allowed(p))) | |
3283 | goto next; | |
3284 | ||
3285 | for_each_cpu(i, sched_group_cpus(sg)) { | |
3286 | if (!idle_cpu(i)) | |
3287 | goto next; | |
3288 | } | |
970e1789 | 3289 | |
37407ea7 LT |
3290 | target = cpumask_first_and(sched_group_cpus(sg), |
3291 | tsk_cpus_allowed(p)); | |
3292 | goto done; | |
3293 | next: | |
3294 | sg = sg->next; | |
3295 | } while (sg != sd->groups); | |
3296 | } | |
3297 | done: | |
a50bde51 PZ |
3298 | return target; |
3299 | } | |
3300 | ||
aaee1203 PZ |
3301 | /* |
3302 | * sched_balance_self: balance the current task (running on cpu) in domains | |
3303 | * that have the 'flag' flag set. In practice, this is SD_BALANCE_FORK and | |
3304 | * SD_BALANCE_EXEC. | |
3305 | * | |
3306 | * Balance, ie. select the least loaded group. | |
3307 | * | |
3308 | * Returns the target CPU number, or the same CPU if no balancing is needed. | |
3309 | * | |
3310 | * preempt must be disabled. | |
3311 | */ | |
0017d735 | 3312 | static int |
7608dec2 | 3313 | select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags) |
aaee1203 | 3314 | { |
29cd8bae | 3315 | struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL; |
c88d5910 PZ |
3316 | int cpu = smp_processor_id(); |
3317 | int prev_cpu = task_cpu(p); | |
3318 | int new_cpu = cpu; | |
99bd5e2f | 3319 | int want_affine = 0; |
5158f4e4 | 3320 | int sync = wake_flags & WF_SYNC; |
c88d5910 | 3321 | |
29baa747 | 3322 | if (p->nr_cpus_allowed == 1) |
76854c7e MG |
3323 | return prev_cpu; |
3324 | ||
0763a660 | 3325 | if (sd_flag & SD_BALANCE_WAKE) { |
fa17b507 | 3326 | if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) |
c88d5910 PZ |
3327 | want_affine = 1; |
3328 | new_cpu = prev_cpu; | |
3329 | } | |
aaee1203 | 3330 | |
dce840a0 | 3331 | rcu_read_lock(); |
aaee1203 | 3332 | for_each_domain(cpu, tmp) { |
e4f42888 PZ |
3333 | if (!(tmp->flags & SD_LOAD_BALANCE)) |
3334 | continue; | |
3335 | ||
fe3bcfe1 | 3336 | /* |
99bd5e2f SS |
3337 | * If both cpu and prev_cpu are part of this domain, |
3338 | * cpu is a valid SD_WAKE_AFFINE target. | |
fe3bcfe1 | 3339 | */ |
99bd5e2f SS |
3340 | if (want_affine && (tmp->flags & SD_WAKE_AFFINE) && |
3341 | cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) { | |
3342 | affine_sd = tmp; | |
29cd8bae | 3343 | break; |
f03542a7 | 3344 | } |
29cd8bae | 3345 | |
f03542a7 | 3346 | if (tmp->flags & sd_flag) |
29cd8bae PZ |
3347 | sd = tmp; |
3348 | } | |
3349 | ||
8b911acd | 3350 | if (affine_sd) { |
f03542a7 | 3351 | if (cpu != prev_cpu && wake_affine(affine_sd, p, sync)) |
dce840a0 PZ |
3352 | prev_cpu = cpu; |
3353 | ||
3354 | new_cpu = select_idle_sibling(p, prev_cpu); | |
3355 | goto unlock; | |
8b911acd | 3356 | } |
e7693a36 | 3357 | |
aaee1203 | 3358 | while (sd) { |
5158f4e4 | 3359 | int load_idx = sd->forkexec_idx; |
aaee1203 | 3360 | struct sched_group *group; |
c88d5910 | 3361 | int weight; |
098fb9db | 3362 | |
0763a660 | 3363 | if (!(sd->flags & sd_flag)) { |
aaee1203 PZ |
3364 | sd = sd->child; |
3365 | continue; | |
3366 | } | |
098fb9db | 3367 | |
5158f4e4 PZ |
3368 | if (sd_flag & SD_BALANCE_WAKE) |
3369 | load_idx = sd->wake_idx; | |
098fb9db | 3370 | |
5158f4e4 | 3371 | group = find_idlest_group(sd, p, cpu, load_idx); |
aaee1203 PZ |
3372 | if (!group) { |
3373 | sd = sd->child; | |
3374 | continue; | |
3375 | } | |
4ae7d5ce | 3376 | |
d7c33c49 | 3377 | new_cpu = find_idlest_cpu(group, p, cpu); |
aaee1203 PZ |
3378 | if (new_cpu == -1 || new_cpu == cpu) { |
3379 | /* Now try balancing at a lower domain level of cpu */ | |
3380 | sd = sd->child; | |
3381 | continue; | |
e7693a36 | 3382 | } |
aaee1203 PZ |
3383 | |
3384 | /* Now try balancing at a lower domain level of new_cpu */ | |
3385 | cpu = new_cpu; | |
669c55e9 | 3386 | weight = sd->span_weight; |
aaee1203 PZ |
3387 | sd = NULL; |
3388 | for_each_domain(cpu, tmp) { | |
669c55e9 | 3389 | if (weight <= tmp->span_weight) |
aaee1203 | 3390 | break; |
0763a660 | 3391 | if (tmp->flags & sd_flag) |
aaee1203 PZ |
3392 | sd = tmp; |
3393 | } | |
3394 | /* while loop will break here if sd == NULL */ | |
e7693a36 | 3395 | } |
dce840a0 PZ |
3396 | unlock: |
3397 | rcu_read_unlock(); | |
e7693a36 | 3398 | |
c88d5910 | 3399 | return new_cpu; |
e7693a36 | 3400 | } |
0a74bef8 | 3401 | |
f4e26b12 PT |
3402 | /* |
3403 | * Load-tracking only depends on SMP, FAIR_GROUP_SCHED dependency below may be | |
3404 | * removed when useful for applications beyond shares distribution (e.g. | |
3405 | * load-balance). | |
3406 | */ | |
3407 | #ifdef CONFIG_FAIR_GROUP_SCHED | |
0a74bef8 PT |
3408 | /* |
3409 | * Called immediately before a task is migrated to a new cpu; task_cpu(p) and | |
3410 | * cfs_rq_of(p) references at time of call are still valid and identify the | |
3411 | * previous cpu. However, the caller only guarantees p->pi_lock is held; no | |
3412 | * other assumptions, including the state of rq->lock, should be made. | |
3413 | */ | |
3414 | static void | |
3415 | migrate_task_rq_fair(struct task_struct *p, int next_cpu) | |
3416 | { | |
aff3e498 PT |
3417 | struct sched_entity *se = &p->se; |
3418 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
3419 | ||
3420 | /* | |
3421 | * Load tracking: accumulate removed load so that it can be processed | |
3422 | * when we next update owning cfs_rq under rq->lock. Tasks contribute | |
3423 | * to blocked load iff they have a positive decay-count. It can never | |
3424 | * be negative here since on-rq tasks have decay-count == 0. | |
3425 | */ | |
3426 | if (se->avg.decay_count) { | |
3427 | se->avg.decay_count = -__synchronize_entity_decay(se); | |
3428 | atomic64_add(se->avg.load_avg_contrib, &cfs_rq->removed_load); | |
3429 | } | |
0a74bef8 | 3430 | } |
f4e26b12 | 3431 | #endif |
e7693a36 GH |
3432 | #endif /* CONFIG_SMP */ |
3433 | ||
e52fb7c0 PZ |
3434 | static unsigned long |
3435 | wakeup_gran(struct sched_entity *curr, struct sched_entity *se) | |
0bbd3336 PZ |
3436 | { |
3437 | unsigned long gran = sysctl_sched_wakeup_granularity; | |
3438 | ||
3439 | /* | |
e52fb7c0 PZ |
3440 | * Since its curr running now, convert the gran from real-time |
3441 | * to virtual-time in his units. | |
13814d42 MG |
3442 | * |
3443 | * By using 'se' instead of 'curr' we penalize light tasks, so | |
3444 | * they get preempted easier. That is, if 'se' < 'curr' then | |
3445 | * the resulting gran will be larger, therefore penalizing the | |
3446 | * lighter, if otoh 'se' > 'curr' then the resulting gran will | |
3447 | * be smaller, again penalizing the lighter task. | |
3448 | * | |
3449 | * This is especially important for buddies when the leftmost | |
3450 | * task is higher priority than the buddy. | |
0bbd3336 | 3451 | */ |
f4ad9bd2 | 3452 | return calc_delta_fair(gran, se); |
0bbd3336 PZ |
3453 | } |
3454 | ||
464b7527 PZ |
3455 | /* |
3456 | * Should 'se' preempt 'curr'. | |
3457 | * | |
3458 | * |s1 | |
3459 | * |s2 | |
3460 | * |s3 | |
3461 | * g | |
3462 | * |<--->|c | |
3463 | * | |
3464 | * w(c, s1) = -1 | |
3465 | * w(c, s2) = 0 | |
3466 | * w(c, s3) = 1 | |
3467 | * | |
3468 | */ | |
3469 | static int | |
3470 | wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se) | |
3471 | { | |
3472 | s64 gran, vdiff = curr->vruntime - se->vruntime; | |
3473 | ||
3474 | if (vdiff <= 0) | |
3475 | return -1; | |
3476 | ||
e52fb7c0 | 3477 | gran = wakeup_gran(curr, se); |
464b7527 PZ |
3478 | if (vdiff > gran) |
3479 | return 1; | |
3480 | ||
3481 | return 0; | |
3482 | } | |
3483 | ||
02479099 PZ |
3484 | static void set_last_buddy(struct sched_entity *se) |
3485 | { | |
69c80f3e VP |
3486 | if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE)) |
3487 | return; | |
3488 | ||
3489 | for_each_sched_entity(se) | |
3490 | cfs_rq_of(se)->last = se; | |
02479099 PZ |
3491 | } |
3492 | ||
3493 | static void set_next_buddy(struct sched_entity *se) | |
3494 | { | |
69c80f3e VP |
3495 | if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE)) |
3496 | return; | |
3497 | ||
3498 | for_each_sched_entity(se) | |
3499 | cfs_rq_of(se)->next = se; | |
02479099 PZ |
3500 | } |
3501 | ||
ac53db59 RR |
3502 | static void set_skip_buddy(struct sched_entity *se) |
3503 | { | |
69c80f3e VP |
3504 | for_each_sched_entity(se) |
3505 | cfs_rq_of(se)->skip = se; | |
ac53db59 RR |
3506 | } |
3507 | ||
bf0f6f24 IM |
3508 | /* |
3509 | * Preempt the current task with a newly woken task if needed: | |
3510 | */ | |
5a9b86f6 | 3511 | static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) |
bf0f6f24 IM |
3512 | { |
3513 | struct task_struct *curr = rq->curr; | |
8651a86c | 3514 | struct sched_entity *se = &curr->se, *pse = &p->se; |
03e89e45 | 3515 | struct cfs_rq *cfs_rq = task_cfs_rq(curr); |
f685ceac | 3516 | int scale = cfs_rq->nr_running >= sched_nr_latency; |
2f36825b | 3517 | int next_buddy_marked = 0; |
bf0f6f24 | 3518 | |
4ae7d5ce IM |
3519 | if (unlikely(se == pse)) |
3520 | return; | |
3521 | ||
5238cdd3 | 3522 | /* |
ddcdf6e7 | 3523 | * This is possible from callers such as move_task(), in which we |
5238cdd3 PT |
3524 | * unconditionally check_prempt_curr() after an enqueue (which may have |
3525 | * lead to a throttle). This both saves work and prevents false | |
3526 | * next-buddy nomination below. | |
3527 | */ | |
3528 | if (unlikely(throttled_hierarchy(cfs_rq_of(pse)))) | |
3529 | return; | |
3530 | ||
2f36825b | 3531 | if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) { |
3cb63d52 | 3532 | set_next_buddy(pse); |
2f36825b VP |
3533 | next_buddy_marked = 1; |
3534 | } | |
57fdc26d | 3535 | |
aec0a514 BR |
3536 | /* |
3537 | * We can come here with TIF_NEED_RESCHED already set from new task | |
3538 | * wake up path. | |
5238cdd3 PT |
3539 | * |
3540 | * Note: this also catches the edge-case of curr being in a throttled | |
3541 | * group (e.g. via set_curr_task), since update_curr() (in the | |
3542 | * enqueue of curr) will have resulted in resched being set. This | |
3543 | * prevents us from potentially nominating it as a false LAST_BUDDY | |
3544 | * below. | |
aec0a514 BR |
3545 | */ |
3546 | if (test_tsk_need_resched(curr)) | |
3547 | return; | |
3548 | ||
a2f5c9ab DH |
3549 | /* Idle tasks are by definition preempted by non-idle tasks. */ |
3550 | if (unlikely(curr->policy == SCHED_IDLE) && | |
3551 | likely(p->policy != SCHED_IDLE)) | |
3552 | goto preempt; | |
3553 | ||
91c234b4 | 3554 | /* |
a2f5c9ab DH |
3555 | * Batch and idle tasks do not preempt non-idle tasks (their preemption |
3556 | * is driven by the tick): | |
91c234b4 | 3557 | */ |
8ed92e51 | 3558 | if (unlikely(p->policy != SCHED_NORMAL) || !sched_feat(WAKEUP_PREEMPTION)) |
91c234b4 | 3559 | return; |
bf0f6f24 | 3560 | |
464b7527 | 3561 | find_matching_se(&se, &pse); |
9bbd7374 | 3562 | update_curr(cfs_rq_of(se)); |
002f128b | 3563 | BUG_ON(!pse); |
2f36825b VP |
3564 | if (wakeup_preempt_entity(se, pse) == 1) { |
3565 | /* | |
3566 | * Bias pick_next to pick the sched entity that is | |
3567 | * triggering this preemption. | |
3568 | */ | |
3569 | if (!next_buddy_marked) | |
3570 | set_next_buddy(pse); | |
3a7e73a2 | 3571 | goto preempt; |
2f36825b | 3572 | } |
464b7527 | 3573 | |
3a7e73a2 | 3574 | return; |
a65ac745 | 3575 | |
3a7e73a2 PZ |
3576 | preempt: |
3577 | resched_task(curr); | |
3578 | /* | |
3579 | * Only set the backward buddy when the current task is still | |
3580 | * on the rq. This can happen when a wakeup gets interleaved | |
3581 | * with schedule on the ->pre_schedule() or idle_balance() | |
3582 | * point, either of which can * drop the rq lock. | |
3583 | * | |
3584 | * Also, during early boot the idle thread is in the fair class, | |
3585 | * for obvious reasons its a bad idea to schedule back to it. | |
3586 | */ | |
3587 | if (unlikely(!se->on_rq || curr == rq->idle)) | |
3588 | return; | |
3589 | ||
3590 | if (sched_feat(LAST_BUDDY) && scale && entity_is_task(se)) | |
3591 | set_last_buddy(se); | |
bf0f6f24 IM |
3592 | } |
3593 | ||
fb8d4724 | 3594 | static struct task_struct *pick_next_task_fair(struct rq *rq) |
bf0f6f24 | 3595 | { |
8f4d37ec | 3596 | struct task_struct *p; |
bf0f6f24 IM |
3597 | struct cfs_rq *cfs_rq = &rq->cfs; |
3598 | struct sched_entity *se; | |
3599 | ||
36ace27e | 3600 | if (!cfs_rq->nr_running) |
bf0f6f24 IM |
3601 | return NULL; |
3602 | ||
3603 | do { | |
9948f4b2 | 3604 | se = pick_next_entity(cfs_rq); |
f4b6755f | 3605 | set_next_entity(cfs_rq, se); |
bf0f6f24 IM |
3606 | cfs_rq = group_cfs_rq(se); |
3607 | } while (cfs_rq); | |
3608 | ||
8f4d37ec | 3609 | p = task_of(se); |
b39e66ea MG |
3610 | if (hrtick_enabled(rq)) |
3611 | hrtick_start_fair(rq, p); | |
8f4d37ec PZ |
3612 | |
3613 | return p; | |
bf0f6f24 IM |
3614 | } |
3615 | ||
3616 | /* | |
3617 | * Account for a descheduled task: | |
3618 | */ | |
31ee529c | 3619 | static void put_prev_task_fair(struct rq *rq, struct task_struct *prev) |
bf0f6f24 IM |
3620 | { |
3621 | struct sched_entity *se = &prev->se; | |
3622 | struct cfs_rq *cfs_rq; | |
3623 | ||
3624 | for_each_sched_entity(se) { | |
3625 | cfs_rq = cfs_rq_of(se); | |
ab6cde26 | 3626 | put_prev_entity(cfs_rq, se); |
bf0f6f24 IM |
3627 | } |
3628 | } | |
3629 | ||
ac53db59 RR |
3630 | /* |
3631 | * sched_yield() is very simple | |
3632 | * | |
3633 | * The magic of dealing with the ->skip buddy is in pick_next_entity. | |
3634 | */ | |
3635 | static void yield_task_fair(struct rq *rq) | |
3636 | { | |
3637 | struct task_struct *curr = rq->curr; | |
3638 | struct cfs_rq *cfs_rq = task_cfs_rq(curr); | |
3639 | struct sched_entity *se = &curr->se; | |
3640 | ||
3641 | /* | |
3642 | * Are we the only task in the tree? | |
3643 | */ | |
3644 | if (unlikely(rq->nr_running == 1)) | |
3645 | return; | |
3646 | ||
3647 | clear_buddies(cfs_rq, se); | |
3648 | ||
3649 | if (curr->policy != SCHED_BATCH) { | |
3650 | update_rq_clock(rq); | |
3651 | /* | |
3652 | * Update run-time statistics of the 'current'. | |
3653 | */ | |
3654 | update_curr(cfs_rq); | |
916671c0 MG |
3655 | /* |
3656 | * Tell update_rq_clock() that we've just updated, | |
3657 | * so we don't do microscopic update in schedule() | |
3658 | * and double the fastpath cost. | |
3659 | */ | |
3660 | rq->skip_clock_update = 1; | |
ac53db59 RR |
3661 | } |
3662 | ||
3663 | set_skip_buddy(se); | |
3664 | } | |
3665 | ||
d95f4122 MG |
3666 | static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt) |
3667 | { | |
3668 | struct sched_entity *se = &p->se; | |
3669 | ||
5238cdd3 PT |
3670 | /* throttled hierarchies are not runnable */ |
3671 | if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se))) | |
d95f4122 MG |
3672 | return false; |
3673 | ||
3674 | /* Tell the scheduler that we'd really like pse to run next. */ | |
3675 | set_next_buddy(se); | |
3676 | ||
d95f4122 MG |
3677 | yield_task_fair(rq); |
3678 | ||
3679 | return true; | |
3680 | } | |
3681 | ||
681f3e68 | 3682 | #ifdef CONFIG_SMP |
bf0f6f24 | 3683 | /************************************************** |
e9c84cb8 PZ |
3684 | * Fair scheduling class load-balancing methods. |
3685 | * | |
3686 | * BASICS | |
3687 | * | |
3688 | * The purpose of load-balancing is to achieve the same basic fairness the | |
3689 | * per-cpu scheduler provides, namely provide a proportional amount of compute | |
3690 | * time to each task. This is expressed in the following equation: | |
3691 | * | |
3692 | * W_i,n/P_i == W_j,n/P_j for all i,j (1) | |
3693 | * | |
3694 | * Where W_i,n is the n-th weight average for cpu i. The instantaneous weight | |
3695 | * W_i,0 is defined as: | |
3696 | * | |
3697 | * W_i,0 = \Sum_j w_i,j (2) | |
3698 | * | |
3699 | * Where w_i,j is the weight of the j-th runnable task on cpu i. This weight | |
3700 | * is derived from the nice value as per prio_to_weight[]. | |
3701 | * | |
3702 | * The weight average is an exponential decay average of the instantaneous | |
3703 | * weight: | |
3704 | * | |
3705 | * W'_i,n = (2^n - 1) / 2^n * W_i,n + 1 / 2^n * W_i,0 (3) | |
3706 | * | |
3707 | * P_i is the cpu power (or compute capacity) of cpu i, typically it is the | |
3708 | * fraction of 'recent' time available for SCHED_OTHER task execution. But it | |
3709 | * can also include other factors [XXX]. | |
3710 | * | |
3711 | * To achieve this balance we define a measure of imbalance which follows | |
3712 | * directly from (1): | |
3713 | * | |
3714 | * imb_i,j = max{ avg(W/P), W_i/P_i } - min{ avg(W/P), W_j/P_j } (4) | |
3715 | * | |
3716 | * We them move tasks around to minimize the imbalance. In the continuous | |
3717 | * function space it is obvious this converges, in the discrete case we get | |
3718 | * a few fun cases generally called infeasible weight scenarios. | |
3719 | * | |
3720 | * [XXX expand on: | |
3721 | * - infeasible weights; | |
3722 | * - local vs global optima in the discrete case. ] | |
3723 | * | |
3724 | * | |
3725 | * SCHED DOMAINS | |
3726 | * | |
3727 | * In order to solve the imbalance equation (4), and avoid the obvious O(n^2) | |
3728 | * for all i,j solution, we create a tree of cpus that follows the hardware | |
3729 | * topology where each level pairs two lower groups (or better). This results | |
3730 | * in O(log n) layers. Furthermore we reduce the number of cpus going up the | |
3731 | * tree to only the first of the previous level and we decrease the frequency | |
3732 | * of load-balance at each level inv. proportional to the number of cpus in | |
3733 | * the groups. | |
3734 | * | |
3735 | * This yields: | |
3736 | * | |
3737 | * log_2 n 1 n | |
3738 | * \Sum { --- * --- * 2^i } = O(n) (5) | |
3739 | * i = 0 2^i 2^i | |
3740 | * `- size of each group | |
3741 | * | | `- number of cpus doing load-balance | |
3742 | * | `- freq | |
3743 | * `- sum over all levels | |
3744 | * | |
3745 | * Coupled with a limit on how many tasks we can migrate every balance pass, | |
3746 | * this makes (5) the runtime complexity of the balancer. | |
3747 | * | |
3748 | * An important property here is that each CPU is still (indirectly) connected | |
3749 | * to every other cpu in at most O(log n) steps: | |
3750 | * | |
3751 | * The adjacency matrix of the resulting graph is given by: | |
3752 | * | |
3753 | * log_2 n | |
3754 | * A_i,j = \Union (i % 2^k == 0) && i / 2^(k+1) == j / 2^(k+1) (6) | |
3755 | * k = 0 | |
3756 | * | |
3757 | * And you'll find that: | |
3758 | * | |
3759 | * A^(log_2 n)_i,j != 0 for all i,j (7) | |
3760 | * | |
3761 | * Showing there's indeed a path between every cpu in at most O(log n) steps. | |
3762 | * The task movement gives a factor of O(m), giving a convergence complexity | |
3763 | * of: | |
3764 | * | |
3765 | * O(nm log n), n := nr_cpus, m := nr_tasks (8) | |
3766 | * | |
3767 | * | |
3768 | * WORK CONSERVING | |
3769 | * | |
3770 | * In order to avoid CPUs going idle while there's still work to do, new idle | |
3771 | * balancing is more aggressive and has the newly idle cpu iterate up the domain | |
3772 | * tree itself instead of relying on other CPUs to bring it work. | |
3773 | * | |
3774 | * This adds some complexity to both (5) and (8) but it reduces the total idle | |
3775 | * time. | |
3776 | * | |
3777 | * [XXX more?] | |
3778 | * | |
3779 | * | |
3780 | * CGROUPS | |
3781 | * | |
3782 | * Cgroups make a horror show out of (2), instead of a simple sum we get: | |
3783 | * | |
3784 | * s_k,i | |
3785 | * W_i,0 = \Sum_j \Prod_k w_k * ----- (9) | |
3786 | * S_k | |
3787 | * | |
3788 | * Where | |
3789 | * | |
3790 | * s_k,i = \Sum_j w_i,j,k and S_k = \Sum_i s_k,i (10) | |
3791 | * | |
3792 | * w_i,j,k is the weight of the j-th runnable task in the k-th cgroup on cpu i. | |
3793 | * | |
3794 | * The big problem is S_k, its a global sum needed to compute a local (W_i) | |
3795 | * property. | |
3796 | * | |
3797 | * [XXX write more on how we solve this.. _after_ merging pjt's patches that | |
3798 | * rewrite all of this once again.] | |
3799 | */ | |
bf0f6f24 | 3800 | |
ed387b78 HS |
3801 | static unsigned long __read_mostly max_load_balance_interval = HZ/10; |
3802 | ||
ddcdf6e7 | 3803 | #define LBF_ALL_PINNED 0x01 |
367456c7 | 3804 | #define LBF_NEED_BREAK 0x02 |
88b8dac0 | 3805 | #define LBF_SOME_PINNED 0x04 |
ddcdf6e7 PZ |
3806 | |
3807 | struct lb_env { | |
3808 | struct sched_domain *sd; | |
3809 | ||
ddcdf6e7 | 3810 | struct rq *src_rq; |
85c1e7da | 3811 | int src_cpu; |
ddcdf6e7 PZ |
3812 | |
3813 | int dst_cpu; | |
3814 | struct rq *dst_rq; | |
3815 | ||
88b8dac0 SV |
3816 | struct cpumask *dst_grpmask; |
3817 | int new_dst_cpu; | |
ddcdf6e7 | 3818 | enum cpu_idle_type idle; |
bd939f45 | 3819 | long imbalance; |
b9403130 MW |
3820 | /* The set of CPUs under consideration for load-balancing */ |
3821 | struct cpumask *cpus; | |
3822 | ||
ddcdf6e7 | 3823 | unsigned int flags; |
367456c7 PZ |
3824 | |
3825 | unsigned int loop; | |
3826 | unsigned int loop_break; | |
3827 | unsigned int loop_max; | |
ddcdf6e7 PZ |
3828 | }; |
3829 | ||
1e3c88bd | 3830 | /* |
ddcdf6e7 | 3831 | * move_task - move a task from one runqueue to another runqueue. |
1e3c88bd PZ |
3832 | * Both runqueues must be locked. |
3833 | */ | |
ddcdf6e7 | 3834 | static void move_task(struct task_struct *p, struct lb_env *env) |
1e3c88bd | 3835 | { |
ddcdf6e7 PZ |
3836 | deactivate_task(env->src_rq, p, 0); |
3837 | set_task_cpu(p, env->dst_cpu); | |
3838 | activate_task(env->dst_rq, p, 0); | |
3839 | check_preempt_curr(env->dst_rq, p, 0); | |
1e3c88bd PZ |
3840 | } |
3841 | ||
029632fb PZ |
3842 | /* |
3843 | * Is this task likely cache-hot: | |
3844 | */ | |
3845 | static int | |
3846 | task_hot(struct task_struct *p, u64 now, struct sched_domain *sd) | |
3847 | { | |
3848 | s64 delta; | |
3849 | ||
3850 | if (p->sched_class != &fair_sched_class) | |
3851 | return 0; | |
3852 | ||
3853 | if (unlikely(p->policy == SCHED_IDLE)) | |
3854 | return 0; | |
3855 | ||
3856 | /* | |
3857 | * Buddy candidates are cache hot: | |
3858 | */ | |
3859 | if (sched_feat(CACHE_HOT_BUDDY) && this_rq()->nr_running && | |
3860 | (&p->se == cfs_rq_of(&p->se)->next || | |
3861 | &p->se == cfs_rq_of(&p->se)->last)) | |
3862 | return 1; | |
3863 | ||
3864 | if (sysctl_sched_migration_cost == -1) | |
3865 | return 1; | |
3866 | if (sysctl_sched_migration_cost == 0) | |
3867 | return 0; | |
3868 | ||
3869 | delta = now - p->se.exec_start; | |
3870 | ||
3871 | return delta < (s64)sysctl_sched_migration_cost; | |
3872 | } | |
3873 | ||
1e3c88bd PZ |
3874 | /* |
3875 | * can_migrate_task - may task p from runqueue rq be migrated to this_cpu? | |
3876 | */ | |
3877 | static | |
8e45cb54 | 3878 | int can_migrate_task(struct task_struct *p, struct lb_env *env) |
1e3c88bd PZ |
3879 | { |
3880 | int tsk_cache_hot = 0; | |
3881 | /* | |
3882 | * We do not migrate tasks that are: | |
3883 | * 1) running (obviously), or | |
3884 | * 2) cannot be migrated to this CPU due to cpus_allowed, or | |
3885 | * 3) are cache-hot on their current CPU. | |
3886 | */ | |
ddcdf6e7 | 3887 | if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) { |
88b8dac0 SV |
3888 | int new_dst_cpu; |
3889 | ||
41acab88 | 3890 | schedstat_inc(p, se.statistics.nr_failed_migrations_affine); |
88b8dac0 SV |
3891 | |
3892 | /* | |
3893 | * Remember if this task can be migrated to any other cpu in | |
3894 | * our sched_group. We may want to revisit it if we couldn't | |
3895 | * meet load balance goals by pulling other tasks on src_cpu. | |
3896 | * | |
3897 | * Also avoid computing new_dst_cpu if we have already computed | |
3898 | * one in current iteration. | |
3899 | */ | |
3900 | if (!env->dst_grpmask || (env->flags & LBF_SOME_PINNED)) | |
3901 | return 0; | |
3902 | ||
3903 | new_dst_cpu = cpumask_first_and(env->dst_grpmask, | |
3904 | tsk_cpus_allowed(p)); | |
3905 | if (new_dst_cpu < nr_cpu_ids) { | |
3906 | env->flags |= LBF_SOME_PINNED; | |
3907 | env->new_dst_cpu = new_dst_cpu; | |
3908 | } | |
1e3c88bd PZ |
3909 | return 0; |
3910 | } | |
88b8dac0 SV |
3911 | |
3912 | /* Record that we found atleast one task that could run on dst_cpu */ | |
8e45cb54 | 3913 | env->flags &= ~LBF_ALL_PINNED; |
1e3c88bd | 3914 | |
ddcdf6e7 | 3915 | if (task_running(env->src_rq, p)) { |
41acab88 | 3916 | schedstat_inc(p, se.statistics.nr_failed_migrations_running); |
1e3c88bd PZ |
3917 | return 0; |
3918 | } | |
3919 | ||
3920 | /* | |
3921 | * Aggressive migration if: | |
3922 | * 1) task is cache cold, or | |
3923 | * 2) too many balance attempts have failed. | |
3924 | */ | |
3925 | ||
ddcdf6e7 | 3926 | tsk_cache_hot = task_hot(p, env->src_rq->clock_task, env->sd); |
1e3c88bd | 3927 | if (!tsk_cache_hot || |
8e45cb54 | 3928 | env->sd->nr_balance_failed > env->sd->cache_nice_tries) { |
1e3c88bd PZ |
3929 | #ifdef CONFIG_SCHEDSTATS |
3930 | if (tsk_cache_hot) { | |
8e45cb54 | 3931 | schedstat_inc(env->sd, lb_hot_gained[env->idle]); |
41acab88 | 3932 | schedstat_inc(p, se.statistics.nr_forced_migrations); |
1e3c88bd PZ |
3933 | } |
3934 | #endif | |
3935 | return 1; | |
3936 | } | |
3937 | ||
3938 | if (tsk_cache_hot) { | |
41acab88 | 3939 | schedstat_inc(p, se.statistics.nr_failed_migrations_hot); |
1e3c88bd PZ |
3940 | return 0; |
3941 | } | |
3942 | return 1; | |
3943 | } | |
3944 | ||
897c395f PZ |
3945 | /* |
3946 | * move_one_task tries to move exactly one task from busiest to this_rq, as | |
3947 | * part of active balancing operations within "domain". | |
3948 | * Returns 1 if successful and 0 otherwise. | |
3949 | * | |
3950 | * Called with both runqueues locked. | |
3951 | */ | |
8e45cb54 | 3952 | static int move_one_task(struct lb_env *env) |
897c395f PZ |
3953 | { |
3954 | struct task_struct *p, *n; | |
897c395f | 3955 | |
367456c7 PZ |
3956 | list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) { |
3957 | if (throttled_lb_pair(task_group(p), env->src_rq->cpu, env->dst_cpu)) | |
3958 | continue; | |
897c395f | 3959 | |
367456c7 PZ |
3960 | if (!can_migrate_task(p, env)) |
3961 | continue; | |
897c395f | 3962 | |
367456c7 PZ |
3963 | move_task(p, env); |
3964 | /* | |
3965 | * Right now, this is only the second place move_task() | |
3966 | * is called, so we can safely collect move_task() | |
3967 | * stats here rather than inside move_task(). | |
3968 | */ | |
3969 | schedstat_inc(env->sd, lb_gained[env->idle]); | |
3970 | return 1; | |
897c395f | 3971 | } |
897c395f PZ |
3972 | return 0; |
3973 | } | |
3974 | ||
367456c7 PZ |
3975 | static unsigned long task_h_load(struct task_struct *p); |
3976 | ||
eb95308e PZ |
3977 | static const unsigned int sched_nr_migrate_break = 32; |
3978 | ||
5d6523eb | 3979 | /* |
bd939f45 | 3980 | * move_tasks tries to move up to imbalance weighted load from busiest to |
5d6523eb PZ |
3981 | * this_rq, as part of a balancing operation within domain "sd". |
3982 | * Returns 1 if successful and 0 otherwise. | |
3983 | * | |
3984 | * Called with both runqueues locked. | |
3985 | */ | |
3986 | static int move_tasks(struct lb_env *env) | |
1e3c88bd | 3987 | { |
5d6523eb PZ |
3988 | struct list_head *tasks = &env->src_rq->cfs_tasks; |
3989 | struct task_struct *p; | |
367456c7 PZ |
3990 | unsigned long load; |
3991 | int pulled = 0; | |
1e3c88bd | 3992 | |
bd939f45 | 3993 | if (env->imbalance <= 0) |
5d6523eb | 3994 | return 0; |
1e3c88bd | 3995 | |
5d6523eb PZ |
3996 | while (!list_empty(tasks)) { |
3997 | p = list_first_entry(tasks, struct task_struct, se.group_node); | |
1e3c88bd | 3998 | |
367456c7 PZ |
3999 | env->loop++; |
4000 | /* We've more or less seen every task there is, call it quits */ | |
5d6523eb | 4001 | if (env->loop > env->loop_max) |
367456c7 | 4002 | break; |
5d6523eb PZ |
4003 | |
4004 | /* take a breather every nr_migrate tasks */ | |
367456c7 | 4005 | if (env->loop > env->loop_break) { |
eb95308e | 4006 | env->loop_break += sched_nr_migrate_break; |
8e45cb54 | 4007 | env->flags |= LBF_NEED_BREAK; |
ee00e66f | 4008 | break; |
a195f004 | 4009 | } |
1e3c88bd | 4010 | |
5d6523eb | 4011 | if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu)) |
367456c7 PZ |
4012 | goto next; |
4013 | ||
4014 | load = task_h_load(p); | |
5d6523eb | 4015 | |
eb95308e | 4016 | if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed) |
367456c7 PZ |
4017 | goto next; |
4018 | ||
bd939f45 | 4019 | if ((load / 2) > env->imbalance) |
367456c7 | 4020 | goto next; |
1e3c88bd | 4021 | |
367456c7 PZ |
4022 | if (!can_migrate_task(p, env)) |
4023 | goto next; | |
1e3c88bd | 4024 | |
ddcdf6e7 | 4025 | move_task(p, env); |
ee00e66f | 4026 | pulled++; |
bd939f45 | 4027 | env->imbalance -= load; |
1e3c88bd PZ |
4028 | |
4029 | #ifdef CONFIG_PREEMPT | |
ee00e66f PZ |
4030 | /* |
4031 | * NEWIDLE balancing is a source of latency, so preemptible | |
4032 | * kernels will stop after the first task is pulled to minimize | |
4033 | * the critical section. | |
4034 | */ | |
5d6523eb | 4035 | if (env->idle == CPU_NEWLY_IDLE) |
ee00e66f | 4036 | break; |
1e3c88bd PZ |
4037 | #endif |
4038 | ||
ee00e66f PZ |
4039 | /* |
4040 | * We only want to steal up to the prescribed amount of | |
4041 | * weighted load. | |
4042 | */ | |
bd939f45 | 4043 | if (env->imbalance <= 0) |
ee00e66f | 4044 | break; |
367456c7 PZ |
4045 | |
4046 | continue; | |
4047 | next: | |
5d6523eb | 4048 | list_move_tail(&p->se.group_node, tasks); |
1e3c88bd | 4049 | } |
5d6523eb | 4050 | |
1e3c88bd | 4051 | /* |
ddcdf6e7 PZ |
4052 | * Right now, this is one of only two places move_task() is called, |
4053 | * so we can safely collect move_task() stats here rather than | |
4054 | * inside move_task(). | |
1e3c88bd | 4055 | */ |
8e45cb54 | 4056 | schedstat_add(env->sd, lb_gained[env->idle], pulled); |
1e3c88bd | 4057 | |
5d6523eb | 4058 | return pulled; |
1e3c88bd PZ |
4059 | } |
4060 | ||
230059de | 4061 | #ifdef CONFIG_FAIR_GROUP_SCHED |
9e3081ca PZ |
4062 | /* |
4063 | * update tg->load_weight by folding this cpu's load_avg | |
4064 | */ | |
48a16753 | 4065 | static void __update_blocked_averages_cpu(struct task_group *tg, int cpu) |
9e3081ca | 4066 | { |
48a16753 PT |
4067 | struct sched_entity *se = tg->se[cpu]; |
4068 | struct cfs_rq *cfs_rq = tg->cfs_rq[cpu]; | |
9e3081ca | 4069 | |
48a16753 PT |
4070 | /* throttled entities do not contribute to load */ |
4071 | if (throttled_hierarchy(cfs_rq)) | |
4072 | return; | |
9e3081ca | 4073 | |
aff3e498 | 4074 | update_cfs_rq_blocked_load(cfs_rq, 1); |
9e3081ca | 4075 | |
82958366 PT |
4076 | if (se) { |
4077 | update_entity_load_avg(se, 1); | |
4078 | /* | |
4079 | * We pivot on our runnable average having decayed to zero for | |
4080 | * list removal. This generally implies that all our children | |
4081 | * have also been removed (modulo rounding error or bandwidth | |
4082 | * control); however, such cases are rare and we can fix these | |
4083 | * at enqueue. | |
4084 | * | |
4085 | * TODO: fix up out-of-order children on enqueue. | |
4086 | */ | |
4087 | if (!se->avg.runnable_avg_sum && !cfs_rq->nr_running) | |
4088 | list_del_leaf_cfs_rq(cfs_rq); | |
4089 | } else { | |
48a16753 | 4090 | struct rq *rq = rq_of(cfs_rq); |
82958366 PT |
4091 | update_rq_runnable_avg(rq, rq->nr_running); |
4092 | } | |
9e3081ca PZ |
4093 | } |
4094 | ||
48a16753 | 4095 | static void update_blocked_averages(int cpu) |
9e3081ca | 4096 | { |
9e3081ca | 4097 | struct rq *rq = cpu_rq(cpu); |
48a16753 PT |
4098 | struct cfs_rq *cfs_rq; |
4099 | unsigned long flags; | |
9e3081ca | 4100 | |
48a16753 PT |
4101 | raw_spin_lock_irqsave(&rq->lock, flags); |
4102 | update_rq_clock(rq); | |
9763b67f PZ |
4103 | /* |
4104 | * Iterates the task_group tree in a bottom up fashion, see | |
4105 | * list_add_leaf_cfs_rq() for details. | |
4106 | */ | |
64660c86 | 4107 | for_each_leaf_cfs_rq(rq, cfs_rq) { |
48a16753 PT |
4108 | /* |
4109 | * Note: We may want to consider periodically releasing | |
4110 | * rq->lock about these updates so that creating many task | |
4111 | * groups does not result in continually extending hold time. | |
4112 | */ | |
4113 | __update_blocked_averages_cpu(cfs_rq->tg, rq->cpu); | |
64660c86 | 4114 | } |
48a16753 PT |
4115 | |
4116 | raw_spin_unlock_irqrestore(&rq->lock, flags); | |
9e3081ca PZ |
4117 | } |
4118 | ||
9763b67f PZ |
4119 | /* |
4120 | * Compute the cpu's hierarchical load factor for each task group. | |
4121 | * This needs to be done in a top-down fashion because the load of a child | |
4122 | * group is a fraction of its parents load. | |
4123 | */ | |
4124 | static int tg_load_down(struct task_group *tg, void *data) | |
4125 | { | |
4126 | unsigned long load; | |
4127 | long cpu = (long)data; | |
4128 | ||
4129 | if (!tg->parent) { | |
4130 | load = cpu_rq(cpu)->load.weight; | |
4131 | } else { | |
4132 | load = tg->parent->cfs_rq[cpu]->h_load; | |
4133 | load *= tg->se[cpu]->load.weight; | |
4134 | load /= tg->parent->cfs_rq[cpu]->load.weight + 1; | |
4135 | } | |
4136 | ||
4137 | tg->cfs_rq[cpu]->h_load = load; | |
4138 | ||
4139 | return 0; | |
4140 | } | |
4141 | ||
4142 | static void update_h_load(long cpu) | |
4143 | { | |
a35b6466 PZ |
4144 | struct rq *rq = cpu_rq(cpu); |
4145 | unsigned long now = jiffies; | |
4146 | ||
4147 | if (rq->h_load_throttle == now) | |
4148 | return; | |
4149 | ||
4150 | rq->h_load_throttle = now; | |
4151 | ||
367456c7 | 4152 | rcu_read_lock(); |
9763b67f | 4153 | walk_tg_tree(tg_load_down, tg_nop, (void *)cpu); |
367456c7 | 4154 | rcu_read_unlock(); |
9763b67f PZ |
4155 | } |
4156 | ||
367456c7 | 4157 | static unsigned long task_h_load(struct task_struct *p) |
230059de | 4158 | { |
367456c7 PZ |
4159 | struct cfs_rq *cfs_rq = task_cfs_rq(p); |
4160 | unsigned long load; | |
230059de | 4161 | |
367456c7 PZ |
4162 | load = p->se.load.weight; |
4163 | load = div_u64(load * cfs_rq->h_load, cfs_rq->load.weight + 1); | |
230059de | 4164 | |
367456c7 | 4165 | return load; |
230059de PZ |
4166 | } |
4167 | #else | |
48a16753 | 4168 | static inline void update_blocked_averages(int cpu) |
9e3081ca PZ |
4169 | { |
4170 | } | |
4171 | ||
367456c7 | 4172 | static inline void update_h_load(long cpu) |
230059de | 4173 | { |
230059de | 4174 | } |
230059de | 4175 | |
367456c7 | 4176 | static unsigned long task_h_load(struct task_struct *p) |
1e3c88bd | 4177 | { |
367456c7 | 4178 | return p->se.load.weight; |
1e3c88bd | 4179 | } |
230059de | 4180 | #endif |
1e3c88bd | 4181 | |
1e3c88bd PZ |
4182 | /********** Helpers for find_busiest_group ************************/ |
4183 | /* | |
4184 | * sd_lb_stats - Structure to store the statistics of a sched_domain | |
4185 | * during load balancing. | |
4186 | */ | |
4187 | struct sd_lb_stats { | |
4188 | struct sched_group *busiest; /* Busiest group in this sd */ | |
4189 | struct sched_group *this; /* Local group in this sd */ | |
4190 | unsigned long total_load; /* Total load of all groups in sd */ | |
4191 | unsigned long total_pwr; /* Total power of all groups in sd */ | |
4192 | unsigned long avg_load; /* Average load across all groups in sd */ | |
4193 | ||
4194 | /** Statistics of this group */ | |
4195 | unsigned long this_load; | |
4196 | unsigned long this_load_per_task; | |
4197 | unsigned long this_nr_running; | |
fab47622 | 4198 | unsigned long this_has_capacity; |
aae6d3dd | 4199 | unsigned int this_idle_cpus; |
1e3c88bd PZ |
4200 | |
4201 | /* Statistics of the busiest group */ | |
aae6d3dd | 4202 | unsigned int busiest_idle_cpus; |
1e3c88bd PZ |
4203 | unsigned long max_load; |
4204 | unsigned long busiest_load_per_task; | |
4205 | unsigned long busiest_nr_running; | |
dd5feea1 | 4206 | unsigned long busiest_group_capacity; |
fab47622 | 4207 | unsigned long busiest_has_capacity; |
aae6d3dd | 4208 | unsigned int busiest_group_weight; |
1e3c88bd PZ |
4209 | |
4210 | int group_imb; /* Is there imbalance in this sd */ | |
1e3c88bd PZ |
4211 | }; |
4212 | ||
4213 | /* | |
4214 | * sg_lb_stats - stats of a sched_group required for load_balancing | |
4215 | */ | |
4216 | struct sg_lb_stats { | |
4217 | unsigned long avg_load; /*Avg load across the CPUs of the group */ | |
4218 | unsigned long group_load; /* Total load over the CPUs of the group */ | |
4219 | unsigned long sum_nr_running; /* Nr tasks running in the group */ | |
4220 | unsigned long sum_weighted_load; /* Weighted load of group's tasks */ | |
4221 | unsigned long group_capacity; | |
aae6d3dd SS |
4222 | unsigned long idle_cpus; |
4223 | unsigned long group_weight; | |
1e3c88bd | 4224 | int group_imb; /* Is there an imbalance in the group ? */ |
fab47622 | 4225 | int group_has_capacity; /* Is there extra capacity in the group? */ |
1e3c88bd PZ |
4226 | }; |
4227 | ||
1e3c88bd PZ |
4228 | /** |
4229 | * get_sd_load_idx - Obtain the load index for a given sched domain. | |
4230 | * @sd: The sched_domain whose load_idx is to be obtained. | |
4231 | * @idle: The Idle status of the CPU for whose sd load_icx is obtained. | |
4232 | */ | |
4233 | static inline int get_sd_load_idx(struct sched_domain *sd, | |
4234 | enum cpu_idle_type idle) | |
4235 | { | |
4236 | int load_idx; | |
4237 | ||
4238 | switch (idle) { | |
4239 | case CPU_NOT_IDLE: | |
4240 | load_idx = sd->busy_idx; | |
4241 | break; | |
4242 | ||
4243 | case CPU_NEWLY_IDLE: | |
4244 | load_idx = sd->newidle_idx; | |
4245 | break; | |
4246 | default: | |
4247 | load_idx = sd->idle_idx; | |
4248 | break; | |
4249 | } | |
4250 | ||
4251 | return load_idx; | |
4252 | } | |
4253 | ||
1e3c88bd PZ |
4254 | unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu) |
4255 | { | |
1399fa78 | 4256 | return SCHED_POWER_SCALE; |
1e3c88bd PZ |
4257 | } |
4258 | ||
4259 | unsigned long __weak arch_scale_freq_power(struct sched_domain *sd, int cpu) | |
4260 | { | |
4261 | return default_scale_freq_power(sd, cpu); | |
4262 | } | |
4263 | ||
4264 | unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu) | |
4265 | { | |
669c55e9 | 4266 | unsigned long weight = sd->span_weight; |
1e3c88bd PZ |
4267 | unsigned long smt_gain = sd->smt_gain; |
4268 | ||
4269 | smt_gain /= weight; | |
4270 | ||
4271 | return smt_gain; | |
4272 | } | |
4273 | ||
4274 | unsigned long __weak arch_scale_smt_power(struct sched_domain *sd, int cpu) | |
4275 | { | |
4276 | return default_scale_smt_power(sd, cpu); | |
4277 | } | |
4278 | ||
4279 | unsigned long scale_rt_power(int cpu) | |
4280 | { | |
4281 | struct rq *rq = cpu_rq(cpu); | |
b654f7de | 4282 | u64 total, available, age_stamp, avg; |
1e3c88bd | 4283 | |
b654f7de PZ |
4284 | /* |
4285 | * Since we're reading these variables without serialization make sure | |
4286 | * we read them once before doing sanity checks on them. | |
4287 | */ | |
4288 | age_stamp = ACCESS_ONCE(rq->age_stamp); | |
4289 | avg = ACCESS_ONCE(rq->rt_avg); | |
4290 | ||
4291 | total = sched_avg_period() + (rq->clock - age_stamp); | |
aa483808 | 4292 | |
b654f7de | 4293 | if (unlikely(total < avg)) { |
aa483808 VP |
4294 | /* Ensures that power won't end up being negative */ |
4295 | available = 0; | |
4296 | } else { | |
b654f7de | 4297 | available = total - avg; |
aa483808 | 4298 | } |
1e3c88bd | 4299 | |
1399fa78 NR |
4300 | if (unlikely((s64)total < SCHED_POWER_SCALE)) |
4301 | total = SCHED_POWER_SCALE; | |
1e3c88bd | 4302 | |
1399fa78 | 4303 | total >>= SCHED_POWER_SHIFT; |
1e3c88bd PZ |
4304 | |
4305 | return div_u64(available, total); | |
4306 | } | |
4307 | ||
4308 | static void update_cpu_power(struct sched_domain *sd, int cpu) | |
4309 | { | |
669c55e9 | 4310 | unsigned long weight = sd->span_weight; |
1399fa78 | 4311 | unsigned long power = SCHED_POWER_SCALE; |
1e3c88bd PZ |
4312 | struct sched_group *sdg = sd->groups; |
4313 | ||
1e3c88bd PZ |
4314 | if ((sd->flags & SD_SHARE_CPUPOWER) && weight > 1) { |
4315 | if (sched_feat(ARCH_POWER)) | |
4316 | power *= arch_scale_smt_power(sd, cpu); | |
4317 | else | |
4318 | power *= default_scale_smt_power(sd, cpu); | |
4319 | ||
1399fa78 | 4320 | power >>= SCHED_POWER_SHIFT; |
1e3c88bd PZ |
4321 | } |
4322 | ||
9c3f75cb | 4323 | sdg->sgp->power_orig = power; |
9d5efe05 SV |
4324 | |
4325 | if (sched_feat(ARCH_POWER)) | |
4326 | power *= arch_scale_freq_power(sd, cpu); | |
4327 | else | |
4328 | power *= default_scale_freq_power(sd, cpu); | |
4329 | ||
1399fa78 | 4330 | power >>= SCHED_POWER_SHIFT; |
9d5efe05 | 4331 | |
1e3c88bd | 4332 | power *= scale_rt_power(cpu); |
1399fa78 | 4333 | power >>= SCHED_POWER_SHIFT; |
1e3c88bd PZ |
4334 | |
4335 | if (!power) | |
4336 | power = 1; | |
4337 | ||
e51fd5e2 | 4338 | cpu_rq(cpu)->cpu_power = power; |
9c3f75cb | 4339 | sdg->sgp->power = power; |
1e3c88bd PZ |
4340 | } |
4341 | ||
029632fb | 4342 | void update_group_power(struct sched_domain *sd, int cpu) |
1e3c88bd PZ |
4343 | { |
4344 | struct sched_domain *child = sd->child; | |
4345 | struct sched_group *group, *sdg = sd->groups; | |
4346 | unsigned long power; | |
4ec4412e VG |
4347 | unsigned long interval; |
4348 | ||
4349 | interval = msecs_to_jiffies(sd->balance_interval); | |
4350 | interval = clamp(interval, 1UL, max_load_balance_interval); | |
4351 | sdg->sgp->next_update = jiffies + interval; | |
1e3c88bd PZ |
4352 | |
4353 | if (!child) { | |
4354 | update_cpu_power(sd, cpu); | |
4355 | return; | |
4356 | } | |
4357 | ||
4358 | power = 0; | |
4359 | ||
74a5ce20 PZ |
4360 | if (child->flags & SD_OVERLAP) { |
4361 | /* | |
4362 | * SD_OVERLAP domains cannot assume that child groups | |
4363 | * span the current group. | |
4364 | */ | |
4365 | ||
4366 | for_each_cpu(cpu, sched_group_cpus(sdg)) | |
4367 | power += power_of(cpu); | |
4368 | } else { | |
4369 | /* | |
4370 | * !SD_OVERLAP domains can assume that child groups | |
4371 | * span the current group. | |
4372 | */ | |
4373 | ||
4374 | group = child->groups; | |
4375 | do { | |
4376 | power += group->sgp->power; | |
4377 | group = group->next; | |
4378 | } while (group != child->groups); | |
4379 | } | |
1e3c88bd | 4380 | |
c3decf0d | 4381 | sdg->sgp->power_orig = sdg->sgp->power = power; |
1e3c88bd PZ |
4382 | } |
4383 | ||
9d5efe05 SV |
4384 | /* |
4385 | * Try and fix up capacity for tiny siblings, this is needed when | |
4386 | * things like SD_ASYM_PACKING need f_b_g to select another sibling | |
4387 | * which on its own isn't powerful enough. | |
4388 | * | |
4389 | * See update_sd_pick_busiest() and check_asym_packing(). | |
4390 | */ | |
4391 | static inline int | |
4392 | fix_small_capacity(struct sched_domain *sd, struct sched_group *group) | |
4393 | { | |
4394 | /* | |
1399fa78 | 4395 | * Only siblings can have significantly less than SCHED_POWER_SCALE |
9d5efe05 | 4396 | */ |
a6c75f2f | 4397 | if (!(sd->flags & SD_SHARE_CPUPOWER)) |
9d5efe05 SV |
4398 | return 0; |
4399 | ||
4400 | /* | |
4401 | * If ~90% of the cpu_power is still there, we're good. | |
4402 | */ | |
9c3f75cb | 4403 | if (group->sgp->power * 32 > group->sgp->power_orig * 29) |
9d5efe05 SV |
4404 | return 1; |
4405 | ||
4406 | return 0; | |
4407 | } | |
4408 | ||
1e3c88bd PZ |
4409 | /** |
4410 | * update_sg_lb_stats - Update sched_group's statistics for load balancing. | |
cd96891d | 4411 | * @env: The load balancing environment. |
1e3c88bd | 4412 | * @group: sched_group whose statistics are to be updated. |
1e3c88bd | 4413 | * @load_idx: Load index of sched_domain of this_cpu for load calc. |
1e3c88bd | 4414 | * @local_group: Does group contain this_cpu. |
1e3c88bd PZ |
4415 | * @balance: Should we balance. |
4416 | * @sgs: variable to hold the statistics for this group. | |
4417 | */ | |
bd939f45 PZ |
4418 | static inline void update_sg_lb_stats(struct lb_env *env, |
4419 | struct sched_group *group, int load_idx, | |
b9403130 | 4420 | int local_group, int *balance, struct sg_lb_stats *sgs) |
1e3c88bd | 4421 | { |
e44bc5c5 PZ |
4422 | unsigned long nr_running, max_nr_running, min_nr_running; |
4423 | unsigned long load, max_cpu_load, min_cpu_load; | |
04f733b4 | 4424 | unsigned int balance_cpu = -1, first_idle_cpu = 0; |
dd5feea1 | 4425 | unsigned long avg_load_per_task = 0; |
bd939f45 | 4426 | int i; |
1e3c88bd | 4427 | |
871e35bc | 4428 | if (local_group) |
c1174876 | 4429 | balance_cpu = group_balance_cpu(group); |
1e3c88bd PZ |
4430 | |
4431 | /* Tally up the load of all CPUs in the group */ | |
1e3c88bd PZ |
4432 | max_cpu_load = 0; |
4433 | min_cpu_load = ~0UL; | |
2582f0eb | 4434 | max_nr_running = 0; |
e44bc5c5 | 4435 | min_nr_running = ~0UL; |
1e3c88bd | 4436 | |
b9403130 | 4437 | for_each_cpu_and(i, sched_group_cpus(group), env->cpus) { |
1e3c88bd PZ |
4438 | struct rq *rq = cpu_rq(i); |
4439 | ||
e44bc5c5 PZ |
4440 | nr_running = rq->nr_running; |
4441 | ||
1e3c88bd PZ |
4442 | /* Bias balancing toward cpus of our domain */ |
4443 | if (local_group) { | |
c1174876 PZ |
4444 | if (idle_cpu(i) && !first_idle_cpu && |
4445 | cpumask_test_cpu(i, sched_group_mask(group))) { | |
04f733b4 | 4446 | first_idle_cpu = 1; |
1e3c88bd PZ |
4447 | balance_cpu = i; |
4448 | } | |
04f733b4 PZ |
4449 | |
4450 | load = target_load(i, load_idx); | |
1e3c88bd PZ |
4451 | } else { |
4452 | load = source_load(i, load_idx); | |
e44bc5c5 | 4453 | if (load > max_cpu_load) |
1e3c88bd PZ |
4454 | max_cpu_load = load; |
4455 | if (min_cpu_load > load) | |
4456 | min_cpu_load = load; | |
e44bc5c5 PZ |
4457 | |
4458 | if (nr_running > max_nr_running) | |
4459 | max_nr_running = nr_running; | |
4460 | if (min_nr_running > nr_running) | |
4461 | min_nr_running = nr_running; | |
1e3c88bd PZ |
4462 | } |
4463 | ||
4464 | sgs->group_load += load; | |
e44bc5c5 | 4465 | sgs->sum_nr_running += nr_running; |
1e3c88bd | 4466 | sgs->sum_weighted_load += weighted_cpuload(i); |
aae6d3dd SS |
4467 | if (idle_cpu(i)) |
4468 | sgs->idle_cpus++; | |
1e3c88bd PZ |
4469 | } |
4470 | ||
4471 | /* | |
4472 | * First idle cpu or the first cpu(busiest) in this sched group | |
4473 | * is eligible for doing load balancing at this and above | |
4474 | * domains. In the newly idle case, we will allow all the cpu's | |
4475 | * to do the newly idle load balance. | |
4476 | */ | |
4ec4412e | 4477 | if (local_group) { |
bd939f45 | 4478 | if (env->idle != CPU_NEWLY_IDLE) { |
04f733b4 | 4479 | if (balance_cpu != env->dst_cpu) { |
4ec4412e VG |
4480 | *balance = 0; |
4481 | return; | |
4482 | } | |
bd939f45 | 4483 | update_group_power(env->sd, env->dst_cpu); |
4ec4412e | 4484 | } else if (time_after_eq(jiffies, group->sgp->next_update)) |
bd939f45 | 4485 | update_group_power(env->sd, env->dst_cpu); |
1e3c88bd PZ |
4486 | } |
4487 | ||
4488 | /* Adjust by relative CPU power of the group */ | |
9c3f75cb | 4489 | sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power; |
1e3c88bd | 4490 | |
1e3c88bd PZ |
4491 | /* |
4492 | * Consider the group unbalanced when the imbalance is larger | |
866ab43e | 4493 | * than the average weight of a task. |
1e3c88bd PZ |
4494 | * |
4495 | * APZ: with cgroup the avg task weight can vary wildly and | |
4496 | * might not be a suitable number - should we keep a | |
4497 | * normalized nr_running number somewhere that negates | |
4498 | * the hierarchy? | |
4499 | */ | |
dd5feea1 SS |
4500 | if (sgs->sum_nr_running) |
4501 | avg_load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running; | |
1e3c88bd | 4502 | |
e44bc5c5 PZ |
4503 | if ((max_cpu_load - min_cpu_load) >= avg_load_per_task && |
4504 | (max_nr_running - min_nr_running) > 1) | |
1e3c88bd PZ |
4505 | sgs->group_imb = 1; |
4506 | ||
9c3f75cb | 4507 | sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power, |
1399fa78 | 4508 | SCHED_POWER_SCALE); |
9d5efe05 | 4509 | if (!sgs->group_capacity) |
bd939f45 | 4510 | sgs->group_capacity = fix_small_capacity(env->sd, group); |
aae6d3dd | 4511 | sgs->group_weight = group->group_weight; |
fab47622 NR |
4512 | |
4513 | if (sgs->group_capacity > sgs->sum_nr_running) | |
4514 | sgs->group_has_capacity = 1; | |
1e3c88bd PZ |
4515 | } |
4516 | ||
532cb4c4 MN |
4517 | /** |
4518 | * update_sd_pick_busiest - return 1 on busiest group | |
cd96891d | 4519 | * @env: The load balancing environment. |
532cb4c4 MN |
4520 | * @sds: sched_domain statistics |
4521 | * @sg: sched_group candidate to be checked for being the busiest | |
b6b12294 | 4522 | * @sgs: sched_group statistics |
532cb4c4 MN |
4523 | * |
4524 | * Determine if @sg is a busier group than the previously selected | |
4525 | * busiest group. | |
4526 | */ | |
bd939f45 | 4527 | static bool update_sd_pick_busiest(struct lb_env *env, |
532cb4c4 MN |
4528 | struct sd_lb_stats *sds, |
4529 | struct sched_group *sg, | |
bd939f45 | 4530 | struct sg_lb_stats *sgs) |
532cb4c4 MN |
4531 | { |
4532 | if (sgs->avg_load <= sds->max_load) | |
4533 | return false; | |
4534 | ||
4535 | if (sgs->sum_nr_running > sgs->group_capacity) | |
4536 | return true; | |
4537 | ||
4538 | if (sgs->group_imb) | |
4539 | return true; | |
4540 | ||
4541 | /* | |
4542 | * ASYM_PACKING needs to move all the work to the lowest | |
4543 | * numbered CPUs in the group, therefore mark all groups | |
4544 | * higher than ourself as busy. | |
4545 | */ | |
bd939f45 PZ |
4546 | if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running && |
4547 | env->dst_cpu < group_first_cpu(sg)) { | |
532cb4c4 MN |
4548 | if (!sds->busiest) |
4549 | return true; | |
4550 | ||
4551 | if (group_first_cpu(sds->busiest) > group_first_cpu(sg)) | |
4552 | return true; | |
4553 | } | |
4554 | ||
4555 | return false; | |
4556 | } | |
4557 | ||
1e3c88bd | 4558 | /** |
461819ac | 4559 | * update_sd_lb_stats - Update sched_domain's statistics for load balancing. |
cd96891d | 4560 | * @env: The load balancing environment. |
1e3c88bd PZ |
4561 | * @balance: Should we balance. |
4562 | * @sds: variable to hold the statistics for this sched_domain. | |
4563 | */ | |
bd939f45 | 4564 | static inline void update_sd_lb_stats(struct lb_env *env, |
b9403130 | 4565 | int *balance, struct sd_lb_stats *sds) |
1e3c88bd | 4566 | { |
bd939f45 PZ |
4567 | struct sched_domain *child = env->sd->child; |
4568 | struct sched_group *sg = env->sd->groups; | |
1e3c88bd PZ |
4569 | struct sg_lb_stats sgs; |
4570 | int load_idx, prefer_sibling = 0; | |
4571 | ||
4572 | if (child && child->flags & SD_PREFER_SIBLING) | |
4573 | prefer_sibling = 1; | |
4574 | ||
bd939f45 | 4575 | load_idx = get_sd_load_idx(env->sd, env->idle); |
1e3c88bd PZ |
4576 | |
4577 | do { | |
4578 | int local_group; | |
4579 | ||
bd939f45 | 4580 | local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg)); |
1e3c88bd | 4581 | memset(&sgs, 0, sizeof(sgs)); |
b9403130 | 4582 | update_sg_lb_stats(env, sg, load_idx, local_group, balance, &sgs); |
1e3c88bd | 4583 | |
8f190fb3 | 4584 | if (local_group && !(*balance)) |
1e3c88bd PZ |
4585 | return; |
4586 | ||
4587 | sds->total_load += sgs.group_load; | |
9c3f75cb | 4588 | sds->total_pwr += sg->sgp->power; |
1e3c88bd PZ |
4589 | |
4590 | /* | |
4591 | * In case the child domain prefers tasks go to siblings | |
532cb4c4 | 4592 | * first, lower the sg capacity to one so that we'll try |
75dd321d NR |
4593 | * and move all the excess tasks away. We lower the capacity |
4594 | * of a group only if the local group has the capacity to fit | |
4595 | * these excess tasks, i.e. nr_running < group_capacity. The | |
4596 | * extra check prevents the case where you always pull from the | |
4597 | * heaviest group when it is already under-utilized (possible | |
4598 | * with a large weight task outweighs the tasks on the system). | |
1e3c88bd | 4599 | */ |
75dd321d | 4600 | if (prefer_sibling && !local_group && sds->this_has_capacity) |
1e3c88bd PZ |
4601 | sgs.group_capacity = min(sgs.group_capacity, 1UL); |
4602 | ||
4603 | if (local_group) { | |
4604 | sds->this_load = sgs.avg_load; | |
532cb4c4 | 4605 | sds->this = sg; |
1e3c88bd PZ |
4606 | sds->this_nr_running = sgs.sum_nr_running; |
4607 | sds->this_load_per_task = sgs.sum_weighted_load; | |
fab47622 | 4608 | sds->this_has_capacity = sgs.group_has_capacity; |
aae6d3dd | 4609 | sds->this_idle_cpus = sgs.idle_cpus; |
bd939f45 | 4610 | } else if (update_sd_pick_busiest(env, sds, sg, &sgs)) { |
1e3c88bd | 4611 | sds->max_load = sgs.avg_load; |
532cb4c4 | 4612 | sds->busiest = sg; |
1e3c88bd | 4613 | sds->busiest_nr_running = sgs.sum_nr_running; |
aae6d3dd | 4614 | sds->busiest_idle_cpus = sgs.idle_cpus; |
dd5feea1 | 4615 | sds->busiest_group_capacity = sgs.group_capacity; |
1e3c88bd | 4616 | sds->busiest_load_per_task = sgs.sum_weighted_load; |
fab47622 | 4617 | sds->busiest_has_capacity = sgs.group_has_capacity; |
aae6d3dd | 4618 | sds->busiest_group_weight = sgs.group_weight; |
1e3c88bd PZ |
4619 | sds->group_imb = sgs.group_imb; |
4620 | } | |
4621 | ||
532cb4c4 | 4622 | sg = sg->next; |
bd939f45 | 4623 | } while (sg != env->sd->groups); |
532cb4c4 MN |
4624 | } |
4625 | ||
532cb4c4 MN |
4626 | /** |
4627 | * check_asym_packing - Check to see if the group is packed into the | |
4628 | * sched doman. | |
4629 | * | |
4630 | * This is primarily intended to used at the sibling level. Some | |
4631 | * cores like POWER7 prefer to use lower numbered SMT threads. In the | |
4632 | * case of POWER7, it can move to lower SMT modes only when higher | |
4633 | * threads are idle. When in lower SMT modes, the threads will | |
4634 | * perform better since they share less core resources. Hence when we | |
4635 | * have idle threads, we want them to be the higher ones. | |
4636 | * | |
4637 | * This packing function is run on idle threads. It checks to see if | |
4638 | * the busiest CPU in this domain (core in the P7 case) has a higher | |
4639 | * CPU number than the packing function is being run on. Here we are | |
4640 | * assuming lower CPU number will be equivalent to lower a SMT thread | |
4641 | * number. | |
4642 | * | |
b6b12294 MN |
4643 | * Returns 1 when packing is required and a task should be moved to |
4644 | * this CPU. The amount of the imbalance is returned in *imbalance. | |
4645 | * | |
cd96891d | 4646 | * @env: The load balancing environment. |
532cb4c4 | 4647 | * @sds: Statistics of the sched_domain which is to be packed |
532cb4c4 | 4648 | */ |
bd939f45 | 4649 | static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds) |
532cb4c4 MN |
4650 | { |
4651 | int busiest_cpu; | |
4652 | ||
bd939f45 | 4653 | if (!(env->sd->flags & SD_ASYM_PACKING)) |
532cb4c4 MN |
4654 | return 0; |
4655 | ||
4656 | if (!sds->busiest) | |
4657 | return 0; | |
4658 | ||
4659 | busiest_cpu = group_first_cpu(sds->busiest); | |
bd939f45 | 4660 | if (env->dst_cpu > busiest_cpu) |
532cb4c4 MN |
4661 | return 0; |
4662 | ||
bd939f45 PZ |
4663 | env->imbalance = DIV_ROUND_CLOSEST( |
4664 | sds->max_load * sds->busiest->sgp->power, SCHED_POWER_SCALE); | |
4665 | ||
532cb4c4 | 4666 | return 1; |
1e3c88bd PZ |
4667 | } |
4668 | ||
4669 | /** | |
4670 | * fix_small_imbalance - Calculate the minor imbalance that exists | |
4671 | * amongst the groups of a sched_domain, during | |
4672 | * load balancing. | |
cd96891d | 4673 | * @env: The load balancing environment. |
1e3c88bd | 4674 | * @sds: Statistics of the sched_domain whose imbalance is to be calculated. |
1e3c88bd | 4675 | */ |
bd939f45 PZ |
4676 | static inline |
4677 | void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds) | |
1e3c88bd PZ |
4678 | { |
4679 | unsigned long tmp, pwr_now = 0, pwr_move = 0; | |
4680 | unsigned int imbn = 2; | |
dd5feea1 | 4681 | unsigned long scaled_busy_load_per_task; |
1e3c88bd PZ |
4682 | |
4683 | if (sds->this_nr_running) { | |
4684 | sds->this_load_per_task /= sds->this_nr_running; | |
4685 | if (sds->busiest_load_per_task > | |
4686 | sds->this_load_per_task) | |
4687 | imbn = 1; | |
bd939f45 | 4688 | } else { |
1e3c88bd | 4689 | sds->this_load_per_task = |
bd939f45 PZ |
4690 | cpu_avg_load_per_task(env->dst_cpu); |
4691 | } | |
1e3c88bd | 4692 | |
dd5feea1 | 4693 | scaled_busy_load_per_task = sds->busiest_load_per_task |
1399fa78 | 4694 | * SCHED_POWER_SCALE; |
9c3f75cb | 4695 | scaled_busy_load_per_task /= sds->busiest->sgp->power; |
dd5feea1 SS |
4696 | |
4697 | if (sds->max_load - sds->this_load + scaled_busy_load_per_task >= | |
4698 | (scaled_busy_load_per_task * imbn)) { | |
bd939f45 | 4699 | env->imbalance = sds->busiest_load_per_task; |
1e3c88bd PZ |
4700 | return; |
4701 | } | |
4702 | ||
4703 | /* | |
4704 | * OK, we don't have enough imbalance to justify moving tasks, | |
4705 | * however we may be able to increase total CPU power used by | |
4706 | * moving them. | |
4707 | */ | |
4708 | ||
9c3f75cb | 4709 | pwr_now += sds->busiest->sgp->power * |
1e3c88bd | 4710 | min(sds->busiest_load_per_task, sds->max_load); |
9c3f75cb | 4711 | pwr_now += sds->this->sgp->power * |
1e3c88bd | 4712 | min(sds->this_load_per_task, sds->this_load); |
1399fa78 | 4713 | pwr_now /= SCHED_POWER_SCALE; |
1e3c88bd PZ |
4714 | |
4715 | /* Amount of load we'd subtract */ | |
1399fa78 | 4716 | tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) / |
9c3f75cb | 4717 | sds->busiest->sgp->power; |
1e3c88bd | 4718 | if (sds->max_load > tmp) |
9c3f75cb | 4719 | pwr_move += sds->busiest->sgp->power * |
1e3c88bd PZ |
4720 | min(sds->busiest_load_per_task, sds->max_load - tmp); |
4721 | ||
4722 | /* Amount of load we'd add */ | |
9c3f75cb | 4723 | if (sds->max_load * sds->busiest->sgp->power < |
1399fa78 | 4724 | sds->busiest_load_per_task * SCHED_POWER_SCALE) |
9c3f75cb PZ |
4725 | tmp = (sds->max_load * sds->busiest->sgp->power) / |
4726 | sds->this->sgp->power; | |
1e3c88bd | 4727 | else |
1399fa78 | 4728 | tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) / |
9c3f75cb PZ |
4729 | sds->this->sgp->power; |
4730 | pwr_move += sds->this->sgp->power * | |
1e3c88bd | 4731 | min(sds->this_load_per_task, sds->this_load + tmp); |
1399fa78 | 4732 | pwr_move /= SCHED_POWER_SCALE; |
1e3c88bd PZ |
4733 | |
4734 | /* Move if we gain throughput */ | |
4735 | if (pwr_move > pwr_now) | |
bd939f45 | 4736 | env->imbalance = sds->busiest_load_per_task; |
1e3c88bd PZ |
4737 | } |
4738 | ||
4739 | /** | |
4740 | * calculate_imbalance - Calculate the amount of imbalance present within the | |
4741 | * groups of a given sched_domain during load balance. | |
bd939f45 | 4742 | * @env: load balance environment |
1e3c88bd | 4743 | * @sds: statistics of the sched_domain whose imbalance is to be calculated. |
1e3c88bd | 4744 | */ |
bd939f45 | 4745 | static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds) |
1e3c88bd | 4746 | { |
dd5feea1 SS |
4747 | unsigned long max_pull, load_above_capacity = ~0UL; |
4748 | ||
4749 | sds->busiest_load_per_task /= sds->busiest_nr_running; | |
4750 | if (sds->group_imb) { | |
4751 | sds->busiest_load_per_task = | |
4752 | min(sds->busiest_load_per_task, sds->avg_load); | |
4753 | } | |
4754 | ||
1e3c88bd PZ |
4755 | /* |
4756 | * In the presence of smp nice balancing, certain scenarios can have | |
4757 | * max load less than avg load(as we skip the groups at or below | |
4758 | * its cpu_power, while calculating max_load..) | |
4759 | */ | |
4760 | if (sds->max_load < sds->avg_load) { | |
bd939f45 PZ |
4761 | env->imbalance = 0; |
4762 | return fix_small_imbalance(env, sds); | |
1e3c88bd PZ |
4763 | } |
4764 | ||
dd5feea1 SS |
4765 | if (!sds->group_imb) { |
4766 | /* | |
4767 | * Don't want to pull so many tasks that a group would go idle. | |
4768 | */ | |
4769 | load_above_capacity = (sds->busiest_nr_running - | |
4770 | sds->busiest_group_capacity); | |
4771 | ||
1399fa78 | 4772 | load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE); |
dd5feea1 | 4773 | |
9c3f75cb | 4774 | load_above_capacity /= sds->busiest->sgp->power; |
dd5feea1 SS |
4775 | } |
4776 | ||
4777 | /* | |
4778 | * We're trying to get all the cpus to the average_load, so we don't | |
4779 | * want to push ourselves above the average load, nor do we wish to | |
4780 | * reduce the max loaded cpu below the average load. At the same time, | |
4781 | * we also don't want to reduce the group load below the group capacity | |
4782 | * (so that we can implement power-savings policies etc). Thus we look | |
4783 | * for the minimum possible imbalance. | |
4784 | * Be careful of negative numbers as they'll appear as very large values | |
4785 | * with unsigned longs. | |
4786 | */ | |
4787 | max_pull = min(sds->max_load - sds->avg_load, load_above_capacity); | |
1e3c88bd PZ |
4788 | |
4789 | /* How much load to actually move to equalise the imbalance */ | |
bd939f45 | 4790 | env->imbalance = min(max_pull * sds->busiest->sgp->power, |
9c3f75cb | 4791 | (sds->avg_load - sds->this_load) * sds->this->sgp->power) |
1399fa78 | 4792 | / SCHED_POWER_SCALE; |
1e3c88bd PZ |
4793 | |
4794 | /* | |
4795 | * if *imbalance is less than the average load per runnable task | |
25985edc | 4796 | * there is no guarantee that any tasks will be moved so we'll have |
1e3c88bd PZ |
4797 | * a think about bumping its value to force at least one task to be |
4798 | * moved | |
4799 | */ | |
bd939f45 PZ |
4800 | if (env->imbalance < sds->busiest_load_per_task) |
4801 | return fix_small_imbalance(env, sds); | |
1e3c88bd PZ |
4802 | |
4803 | } | |
fab47622 | 4804 | |
1e3c88bd PZ |
4805 | /******* find_busiest_group() helpers end here *********************/ |
4806 | ||
4807 | /** | |
4808 | * find_busiest_group - Returns the busiest group within the sched_domain | |
4809 | * if there is an imbalance. If there isn't an imbalance, and | |
4810 | * the user has opted for power-savings, it returns a group whose | |
4811 | * CPUs can be put to idle by rebalancing those tasks elsewhere, if | |
4812 | * such a group exists. | |
4813 | * | |
4814 | * Also calculates the amount of weighted load which should be moved | |
4815 | * to restore balance. | |
4816 | * | |
cd96891d | 4817 | * @env: The load balancing environment. |
1e3c88bd PZ |
4818 | * @balance: Pointer to a variable indicating if this_cpu |
4819 | * is the appropriate cpu to perform load balancing at this_level. | |
4820 | * | |
4821 | * Returns: - the busiest group if imbalance exists. | |
4822 | * - If no imbalance and user has opted for power-savings balance, | |
4823 | * return the least loaded group whose CPUs can be | |
4824 | * put to idle by rebalancing its tasks onto our group. | |
4825 | */ | |
4826 | static struct sched_group * | |
b9403130 | 4827 | find_busiest_group(struct lb_env *env, int *balance) |
1e3c88bd PZ |
4828 | { |
4829 | struct sd_lb_stats sds; | |
4830 | ||
4831 | memset(&sds, 0, sizeof(sds)); | |
4832 | ||
4833 | /* | |
4834 | * Compute the various statistics relavent for load balancing at | |
4835 | * this level. | |
4836 | */ | |
b9403130 | 4837 | update_sd_lb_stats(env, balance, &sds); |
1e3c88bd | 4838 | |
cc57aa8f PZ |
4839 | /* |
4840 | * this_cpu is not the appropriate cpu to perform load balancing at | |
4841 | * this level. | |
1e3c88bd | 4842 | */ |
8f190fb3 | 4843 | if (!(*balance)) |
1e3c88bd PZ |
4844 | goto ret; |
4845 | ||
bd939f45 PZ |
4846 | if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) && |
4847 | check_asym_packing(env, &sds)) | |
532cb4c4 MN |
4848 | return sds.busiest; |
4849 | ||
cc57aa8f | 4850 | /* There is no busy sibling group to pull tasks from */ |
1e3c88bd PZ |
4851 | if (!sds.busiest || sds.busiest_nr_running == 0) |
4852 | goto out_balanced; | |
4853 | ||
1399fa78 | 4854 | sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr; |
b0432d8f | 4855 | |
866ab43e PZ |
4856 | /* |
4857 | * If the busiest group is imbalanced the below checks don't | |
4858 | * work because they assumes all things are equal, which typically | |
4859 | * isn't true due to cpus_allowed constraints and the like. | |
4860 | */ | |
4861 | if (sds.group_imb) | |
4862 | goto force_balance; | |
4863 | ||
cc57aa8f | 4864 | /* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */ |
bd939f45 | 4865 | if (env->idle == CPU_NEWLY_IDLE && sds.this_has_capacity && |
fab47622 NR |
4866 | !sds.busiest_has_capacity) |
4867 | goto force_balance; | |
4868 | ||
cc57aa8f PZ |
4869 | /* |
4870 | * If the local group is more busy than the selected busiest group | |
4871 | * don't try and pull any tasks. | |
4872 | */ | |
1e3c88bd PZ |
4873 | if (sds.this_load >= sds.max_load) |
4874 | goto out_balanced; | |
4875 | ||
cc57aa8f PZ |
4876 | /* |
4877 | * Don't pull any tasks if this group is already above the domain | |
4878 | * average load. | |
4879 | */ | |
1e3c88bd PZ |
4880 | if (sds.this_load >= sds.avg_load) |
4881 | goto out_balanced; | |
4882 | ||
bd939f45 | 4883 | if (env->idle == CPU_IDLE) { |
aae6d3dd SS |
4884 | /* |
4885 | * This cpu is idle. If the busiest group load doesn't | |
4886 | * have more tasks than the number of available cpu's and | |
4887 | * there is no imbalance between this and busiest group | |
4888 | * wrt to idle cpu's, it is balanced. | |
4889 | */ | |
c186fafe | 4890 | if ((sds.this_idle_cpus <= sds.busiest_idle_cpus + 1) && |
aae6d3dd SS |
4891 | sds.busiest_nr_running <= sds.busiest_group_weight) |
4892 | goto out_balanced; | |
c186fafe PZ |
4893 | } else { |
4894 | /* | |
4895 | * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use | |
4896 | * imbalance_pct to be conservative. | |
4897 | */ | |
bd939f45 | 4898 | if (100 * sds.max_load <= env->sd->imbalance_pct * sds.this_load) |
c186fafe | 4899 | goto out_balanced; |
aae6d3dd | 4900 | } |
1e3c88bd | 4901 | |
fab47622 | 4902 | force_balance: |
1e3c88bd | 4903 | /* Looks like there is an imbalance. Compute it */ |
bd939f45 | 4904 | calculate_imbalance(env, &sds); |
1e3c88bd PZ |
4905 | return sds.busiest; |
4906 | ||
4907 | out_balanced: | |
1e3c88bd | 4908 | ret: |
bd939f45 | 4909 | env->imbalance = 0; |
1e3c88bd PZ |
4910 | return NULL; |
4911 | } | |
4912 | ||
4913 | /* | |
4914 | * find_busiest_queue - find the busiest runqueue among the cpus in group. | |
4915 | */ | |
bd939f45 | 4916 | static struct rq *find_busiest_queue(struct lb_env *env, |
b9403130 | 4917 | struct sched_group *group) |
1e3c88bd PZ |
4918 | { |
4919 | struct rq *busiest = NULL, *rq; | |
4920 | unsigned long max_load = 0; | |
4921 | int i; | |
4922 | ||
4923 | for_each_cpu(i, sched_group_cpus(group)) { | |
4924 | unsigned long power = power_of(i); | |
1399fa78 NR |
4925 | unsigned long capacity = DIV_ROUND_CLOSEST(power, |
4926 | SCHED_POWER_SCALE); | |
1e3c88bd PZ |
4927 | unsigned long wl; |
4928 | ||
9d5efe05 | 4929 | if (!capacity) |
bd939f45 | 4930 | capacity = fix_small_capacity(env->sd, group); |
9d5efe05 | 4931 | |
b9403130 | 4932 | if (!cpumask_test_cpu(i, env->cpus)) |
1e3c88bd PZ |
4933 | continue; |
4934 | ||
4935 | rq = cpu_rq(i); | |
6e40f5bb | 4936 | wl = weighted_cpuload(i); |
1e3c88bd | 4937 | |
6e40f5bb TG |
4938 | /* |
4939 | * When comparing with imbalance, use weighted_cpuload() | |
4940 | * which is not scaled with the cpu power. | |
4941 | */ | |
bd939f45 | 4942 | if (capacity && rq->nr_running == 1 && wl > env->imbalance) |
1e3c88bd PZ |
4943 | continue; |
4944 | ||
6e40f5bb TG |
4945 | /* |
4946 | * For the load comparisons with the other cpu's, consider | |
4947 | * the weighted_cpuload() scaled with the cpu power, so that | |
4948 | * the load can be moved away from the cpu that is potentially | |
4949 | * running at a lower capacity. | |
4950 | */ | |
1399fa78 | 4951 | wl = (wl * SCHED_POWER_SCALE) / power; |
6e40f5bb | 4952 | |
1e3c88bd PZ |
4953 | if (wl > max_load) { |
4954 | max_load = wl; | |
4955 | busiest = rq; | |
4956 | } | |
4957 | } | |
4958 | ||
4959 | return busiest; | |
4960 | } | |
4961 | ||
4962 | /* | |
4963 | * Max backoff if we encounter pinned tasks. Pretty arbitrary value, but | |
4964 | * so long as it is large enough. | |
4965 | */ | |
4966 | #define MAX_PINNED_INTERVAL 512 | |
4967 | ||
4968 | /* Working cpumask for load_balance and load_balance_newidle. */ | |
029632fb | 4969 | DEFINE_PER_CPU(cpumask_var_t, load_balance_tmpmask); |
1e3c88bd | 4970 | |
bd939f45 | 4971 | static int need_active_balance(struct lb_env *env) |
1af3ed3d | 4972 | { |
bd939f45 PZ |
4973 | struct sched_domain *sd = env->sd; |
4974 | ||
4975 | if (env->idle == CPU_NEWLY_IDLE) { | |
532cb4c4 MN |
4976 | |
4977 | /* | |
4978 | * ASYM_PACKING needs to force migrate tasks from busy but | |
4979 | * higher numbered CPUs in order to pack all tasks in the | |
4980 | * lowest numbered CPUs. | |
4981 | */ | |
bd939f45 | 4982 | if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu) |
532cb4c4 | 4983 | return 1; |
1af3ed3d PZ |
4984 | } |
4985 | ||
4986 | return unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2); | |
4987 | } | |
4988 | ||
969c7921 TH |
4989 | static int active_load_balance_cpu_stop(void *data); |
4990 | ||
1e3c88bd PZ |
4991 | /* |
4992 | * Check this_cpu to ensure it is balanced within domain. Attempt to move | |
4993 | * tasks if there is an imbalance. | |
4994 | */ | |
4995 | static int load_balance(int this_cpu, struct rq *this_rq, | |
4996 | struct sched_domain *sd, enum cpu_idle_type idle, | |
4997 | int *balance) | |
4998 | { | |
88b8dac0 SV |
4999 | int ld_moved, cur_ld_moved, active_balance = 0; |
5000 | int lb_iterations, max_lb_iterations; | |
1e3c88bd | 5001 | struct sched_group *group; |
1e3c88bd PZ |
5002 | struct rq *busiest; |
5003 | unsigned long flags; | |
5004 | struct cpumask *cpus = __get_cpu_var(load_balance_tmpmask); | |
5005 | ||
8e45cb54 PZ |
5006 | struct lb_env env = { |
5007 | .sd = sd, | |
ddcdf6e7 PZ |
5008 | .dst_cpu = this_cpu, |
5009 | .dst_rq = this_rq, | |
88b8dac0 | 5010 | .dst_grpmask = sched_group_cpus(sd->groups), |
8e45cb54 | 5011 | .idle = idle, |
eb95308e | 5012 | .loop_break = sched_nr_migrate_break, |
b9403130 | 5013 | .cpus = cpus, |
8e45cb54 PZ |
5014 | }; |
5015 | ||
1e3c88bd | 5016 | cpumask_copy(cpus, cpu_active_mask); |
88b8dac0 | 5017 | max_lb_iterations = cpumask_weight(env.dst_grpmask); |
1e3c88bd | 5018 | |
1e3c88bd PZ |
5019 | schedstat_inc(sd, lb_count[idle]); |
5020 | ||
5021 | redo: | |
b9403130 | 5022 | group = find_busiest_group(&env, balance); |
1e3c88bd PZ |
5023 | |
5024 | if (*balance == 0) | |
5025 | goto out_balanced; | |
5026 | ||
5027 | if (!group) { | |
5028 | schedstat_inc(sd, lb_nobusyg[idle]); | |
5029 | goto out_balanced; | |
5030 | } | |
5031 | ||
b9403130 | 5032 | busiest = find_busiest_queue(&env, group); |
1e3c88bd PZ |
5033 | if (!busiest) { |
5034 | schedstat_inc(sd, lb_nobusyq[idle]); | |
5035 | goto out_balanced; | |
5036 | } | |
5037 | ||
78feefc5 | 5038 | BUG_ON(busiest == env.dst_rq); |
1e3c88bd | 5039 | |
bd939f45 | 5040 | schedstat_add(sd, lb_imbalance[idle], env.imbalance); |
1e3c88bd PZ |
5041 | |
5042 | ld_moved = 0; | |
88b8dac0 | 5043 | lb_iterations = 1; |
1e3c88bd PZ |
5044 | if (busiest->nr_running > 1) { |
5045 | /* | |
5046 | * Attempt to move tasks. If find_busiest_group has found | |
5047 | * an imbalance but busiest->nr_running <= 1, the group is | |
5048 | * still unbalanced. ld_moved simply stays zero, so it is | |
5049 | * correctly treated as an imbalance. | |
5050 | */ | |
8e45cb54 | 5051 | env.flags |= LBF_ALL_PINNED; |
c82513e5 PZ |
5052 | env.src_cpu = busiest->cpu; |
5053 | env.src_rq = busiest; | |
5054 | env.loop_max = min(sysctl_sched_nr_migrate, busiest->nr_running); | |
8e45cb54 | 5055 | |
a35b6466 | 5056 | update_h_load(env.src_cpu); |
5d6523eb | 5057 | more_balance: |
1e3c88bd | 5058 | local_irq_save(flags); |
78feefc5 | 5059 | double_rq_lock(env.dst_rq, busiest); |
88b8dac0 SV |
5060 | |
5061 | /* | |
5062 | * cur_ld_moved - load moved in current iteration | |
5063 | * ld_moved - cumulative load moved across iterations | |
5064 | */ | |
5065 | cur_ld_moved = move_tasks(&env); | |
5066 | ld_moved += cur_ld_moved; | |
78feefc5 | 5067 | double_rq_unlock(env.dst_rq, busiest); |
1e3c88bd PZ |
5068 | local_irq_restore(flags); |
5069 | ||
5d6523eb PZ |
5070 | if (env.flags & LBF_NEED_BREAK) { |
5071 | env.flags &= ~LBF_NEED_BREAK; | |
5072 | goto more_balance; | |
5073 | } | |
5074 | ||
1e3c88bd PZ |
5075 | /* |
5076 | * some other cpu did the load balance for us. | |
5077 | */ | |
88b8dac0 SV |
5078 | if (cur_ld_moved && env.dst_cpu != smp_processor_id()) |
5079 | resched_cpu(env.dst_cpu); | |
5080 | ||
5081 | /* | |
5082 | * Revisit (affine) tasks on src_cpu that couldn't be moved to | |
5083 | * us and move them to an alternate dst_cpu in our sched_group | |
5084 | * where they can run. The upper limit on how many times we | |
5085 | * iterate on same src_cpu is dependent on number of cpus in our | |
5086 | * sched_group. | |
5087 | * | |
5088 | * This changes load balance semantics a bit on who can move | |
5089 | * load to a given_cpu. In addition to the given_cpu itself | |
5090 | * (or a ilb_cpu acting on its behalf where given_cpu is | |
5091 | * nohz-idle), we now have balance_cpu in a position to move | |
5092 | * load to given_cpu. In rare situations, this may cause | |
5093 | * conflicts (balance_cpu and given_cpu/ilb_cpu deciding | |
5094 | * _independently_ and at _same_ time to move some load to | |
5095 | * given_cpu) causing exceess load to be moved to given_cpu. | |
5096 | * This however should not happen so much in practice and | |
5097 | * moreover subsequent load balance cycles should correct the | |
5098 | * excess load moved. | |
5099 | */ | |
5100 | if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0 && | |
5101 | lb_iterations++ < max_lb_iterations) { | |
5102 | ||
78feefc5 | 5103 | env.dst_rq = cpu_rq(env.new_dst_cpu); |
88b8dac0 SV |
5104 | env.dst_cpu = env.new_dst_cpu; |
5105 | env.flags &= ~LBF_SOME_PINNED; | |
5106 | env.loop = 0; | |
5107 | env.loop_break = sched_nr_migrate_break; | |
5108 | /* | |
5109 | * Go back to "more_balance" rather than "redo" since we | |
5110 | * need to continue with same src_cpu. | |
5111 | */ | |
5112 | goto more_balance; | |
5113 | } | |
1e3c88bd PZ |
5114 | |
5115 | /* All tasks on this runqueue were pinned by CPU affinity */ | |
8e45cb54 | 5116 | if (unlikely(env.flags & LBF_ALL_PINNED)) { |
1e3c88bd | 5117 | cpumask_clear_cpu(cpu_of(busiest), cpus); |
bbf18b19 PN |
5118 | if (!cpumask_empty(cpus)) { |
5119 | env.loop = 0; | |
5120 | env.loop_break = sched_nr_migrate_break; | |
1e3c88bd | 5121 | goto redo; |
bbf18b19 | 5122 | } |
1e3c88bd PZ |
5123 | goto out_balanced; |
5124 | } | |
5125 | } | |
5126 | ||
5127 | if (!ld_moved) { | |
5128 | schedstat_inc(sd, lb_failed[idle]); | |
58b26c4c VP |
5129 | /* |
5130 | * Increment the failure counter only on periodic balance. | |
5131 | * We do not want newidle balance, which can be very | |
5132 | * frequent, pollute the failure counter causing | |
5133 | * excessive cache_hot migrations and active balances. | |
5134 | */ | |
5135 | if (idle != CPU_NEWLY_IDLE) | |
5136 | sd->nr_balance_failed++; | |
1e3c88bd | 5137 | |
bd939f45 | 5138 | if (need_active_balance(&env)) { |
1e3c88bd PZ |
5139 | raw_spin_lock_irqsave(&busiest->lock, flags); |
5140 | ||
969c7921 TH |
5141 | /* don't kick the active_load_balance_cpu_stop, |
5142 | * if the curr task on busiest cpu can't be | |
5143 | * moved to this_cpu | |
1e3c88bd PZ |
5144 | */ |
5145 | if (!cpumask_test_cpu(this_cpu, | |
fa17b507 | 5146 | tsk_cpus_allowed(busiest->curr))) { |
1e3c88bd PZ |
5147 | raw_spin_unlock_irqrestore(&busiest->lock, |
5148 | flags); | |
8e45cb54 | 5149 | env.flags |= LBF_ALL_PINNED; |
1e3c88bd PZ |
5150 | goto out_one_pinned; |
5151 | } | |
5152 | ||
969c7921 TH |
5153 | /* |
5154 | * ->active_balance synchronizes accesses to | |
5155 | * ->active_balance_work. Once set, it's cleared | |
5156 | * only after active load balance is finished. | |
5157 | */ | |
1e3c88bd PZ |
5158 | if (!busiest->active_balance) { |
5159 | busiest->active_balance = 1; | |
5160 | busiest->push_cpu = this_cpu; | |
5161 | active_balance = 1; | |
5162 | } | |
5163 | raw_spin_unlock_irqrestore(&busiest->lock, flags); | |
969c7921 | 5164 | |
bd939f45 | 5165 | if (active_balance) { |
969c7921 TH |
5166 | stop_one_cpu_nowait(cpu_of(busiest), |
5167 | active_load_balance_cpu_stop, busiest, | |
5168 | &busiest->active_balance_work); | |
bd939f45 | 5169 | } |
1e3c88bd PZ |
5170 | |
5171 | /* | |
5172 | * We've kicked active balancing, reset the failure | |
5173 | * counter. | |
5174 | */ | |
5175 | sd->nr_balance_failed = sd->cache_nice_tries+1; | |
5176 | } | |
5177 | } else | |
5178 | sd->nr_balance_failed = 0; | |
5179 | ||
5180 | if (likely(!active_balance)) { | |
5181 | /* We were unbalanced, so reset the balancing interval */ | |
5182 | sd->balance_interval = sd->min_interval; | |
5183 | } else { | |
5184 | /* | |
5185 | * If we've begun active balancing, start to back off. This | |
5186 | * case may not be covered by the all_pinned logic if there | |
5187 | * is only 1 task on the busy runqueue (because we don't call | |
5188 | * move_tasks). | |
5189 | */ | |
5190 | if (sd->balance_interval < sd->max_interval) | |
5191 | sd->balance_interval *= 2; | |
5192 | } | |
5193 | ||
1e3c88bd PZ |
5194 | goto out; |
5195 | ||
5196 | out_balanced: | |
5197 | schedstat_inc(sd, lb_balanced[idle]); | |
5198 | ||
5199 | sd->nr_balance_failed = 0; | |
5200 | ||
5201 | out_one_pinned: | |
5202 | /* tune up the balancing interval */ | |
8e45cb54 | 5203 | if (((env.flags & LBF_ALL_PINNED) && |
5b54b56b | 5204 | sd->balance_interval < MAX_PINNED_INTERVAL) || |
1e3c88bd PZ |
5205 | (sd->balance_interval < sd->max_interval)) |
5206 | sd->balance_interval *= 2; | |
5207 | ||
46e49b38 | 5208 | ld_moved = 0; |
1e3c88bd | 5209 | out: |
1e3c88bd PZ |
5210 | return ld_moved; |
5211 | } | |
5212 | ||
1e3c88bd PZ |
5213 | /* |
5214 | * idle_balance is called by schedule() if this_cpu is about to become | |
5215 | * idle. Attempts to pull tasks from other CPUs. | |
5216 | */ | |
029632fb | 5217 | void idle_balance(int this_cpu, struct rq *this_rq) |
1e3c88bd PZ |
5218 | { |
5219 | struct sched_domain *sd; | |
5220 | int pulled_task = 0; | |
5221 | unsigned long next_balance = jiffies + HZ; | |
5222 | ||
5223 | this_rq->idle_stamp = this_rq->clock; | |
5224 | ||
5225 | if (this_rq->avg_idle < sysctl_sched_migration_cost) | |
5226 | return; | |
5227 | ||
18bf2805 BS |
5228 | update_rq_runnable_avg(this_rq, 1); |
5229 | ||
f492e12e PZ |
5230 | /* |
5231 | * Drop the rq->lock, but keep IRQ/preempt disabled. | |
5232 | */ | |
5233 | raw_spin_unlock(&this_rq->lock); | |
5234 | ||
48a16753 | 5235 | update_blocked_averages(this_cpu); |
dce840a0 | 5236 | rcu_read_lock(); |
1e3c88bd PZ |
5237 | for_each_domain(this_cpu, sd) { |
5238 | unsigned long interval; | |
f492e12e | 5239 | int balance = 1; |
1e3c88bd PZ |
5240 | |
5241 | if (!(sd->flags & SD_LOAD_BALANCE)) | |
5242 | continue; | |
5243 | ||
f492e12e | 5244 | if (sd->flags & SD_BALANCE_NEWIDLE) { |
1e3c88bd | 5245 | /* If we've pulled tasks over stop searching: */ |
f492e12e PZ |
5246 | pulled_task = load_balance(this_cpu, this_rq, |
5247 | sd, CPU_NEWLY_IDLE, &balance); | |
5248 | } | |
1e3c88bd PZ |
5249 | |
5250 | interval = msecs_to_jiffies(sd->balance_interval); | |
5251 | if (time_after(next_balance, sd->last_balance + interval)) | |
5252 | next_balance = sd->last_balance + interval; | |
d5ad140b NR |
5253 | if (pulled_task) { |
5254 | this_rq->idle_stamp = 0; | |
1e3c88bd | 5255 | break; |
d5ad140b | 5256 | } |
1e3c88bd | 5257 | } |
dce840a0 | 5258 | rcu_read_unlock(); |
f492e12e PZ |
5259 | |
5260 | raw_spin_lock(&this_rq->lock); | |
5261 | ||
1e3c88bd PZ |
5262 | if (pulled_task || time_after(jiffies, this_rq->next_balance)) { |
5263 | /* | |
5264 | * We are going idle. next_balance may be set based on | |
5265 | * a busy processor. So reset next_balance. | |
5266 | */ | |
5267 | this_rq->next_balance = next_balance; | |
5268 | } | |
5269 | } | |
5270 | ||
5271 | /* | |
969c7921 TH |
5272 | * active_load_balance_cpu_stop is run by cpu stopper. It pushes |
5273 | * running tasks off the busiest CPU onto idle CPUs. It requires at | |
5274 | * least 1 task to be running on each physical CPU where possible, and | |
5275 | * avoids physical / logical imbalances. | |
1e3c88bd | 5276 | */ |
969c7921 | 5277 | static int active_load_balance_cpu_stop(void *data) |
1e3c88bd | 5278 | { |
969c7921 TH |
5279 | struct rq *busiest_rq = data; |
5280 | int busiest_cpu = cpu_of(busiest_rq); | |
1e3c88bd | 5281 | int target_cpu = busiest_rq->push_cpu; |
969c7921 | 5282 | struct rq *target_rq = cpu_rq(target_cpu); |
1e3c88bd | 5283 | struct sched_domain *sd; |
969c7921 TH |
5284 | |
5285 | raw_spin_lock_irq(&busiest_rq->lock); | |
5286 | ||
5287 | /* make sure the requested cpu hasn't gone down in the meantime */ | |
5288 | if (unlikely(busiest_cpu != smp_processor_id() || | |
5289 | !busiest_rq->active_balance)) | |
5290 | goto out_unlock; | |
1e3c88bd PZ |
5291 | |
5292 | /* Is there any task to move? */ | |
5293 | if (busiest_rq->nr_running <= 1) | |
969c7921 | 5294 | goto out_unlock; |
1e3c88bd PZ |
5295 | |
5296 | /* | |
5297 | * This condition is "impossible", if it occurs | |
5298 | * we need to fix it. Originally reported by | |
5299 | * Bjorn Helgaas on a 128-cpu setup. | |
5300 | */ | |
5301 | BUG_ON(busiest_rq == target_rq); | |
5302 | ||
5303 | /* move a task from busiest_rq to target_rq */ | |
5304 | double_lock_balance(busiest_rq, target_rq); | |
1e3c88bd PZ |
5305 | |
5306 | /* Search for an sd spanning us and the target CPU. */ | |
dce840a0 | 5307 | rcu_read_lock(); |
1e3c88bd PZ |
5308 | for_each_domain(target_cpu, sd) { |
5309 | if ((sd->flags & SD_LOAD_BALANCE) && | |
5310 | cpumask_test_cpu(busiest_cpu, sched_domain_span(sd))) | |
5311 | break; | |
5312 | } | |
5313 | ||
5314 | if (likely(sd)) { | |
8e45cb54 PZ |
5315 | struct lb_env env = { |
5316 | .sd = sd, | |
ddcdf6e7 PZ |
5317 | .dst_cpu = target_cpu, |
5318 | .dst_rq = target_rq, | |
5319 | .src_cpu = busiest_rq->cpu, | |
5320 | .src_rq = busiest_rq, | |
8e45cb54 PZ |
5321 | .idle = CPU_IDLE, |
5322 | }; | |
5323 | ||
1e3c88bd PZ |
5324 | schedstat_inc(sd, alb_count); |
5325 | ||
8e45cb54 | 5326 | if (move_one_task(&env)) |
1e3c88bd PZ |
5327 | schedstat_inc(sd, alb_pushed); |
5328 | else | |
5329 | schedstat_inc(sd, alb_failed); | |
5330 | } | |
dce840a0 | 5331 | rcu_read_unlock(); |
1e3c88bd | 5332 | double_unlock_balance(busiest_rq, target_rq); |
969c7921 TH |
5333 | out_unlock: |
5334 | busiest_rq->active_balance = 0; | |
5335 | raw_spin_unlock_irq(&busiest_rq->lock); | |
5336 | return 0; | |
1e3c88bd PZ |
5337 | } |
5338 | ||
5339 | #ifdef CONFIG_NO_HZ | |
83cd4fe2 VP |
5340 | /* |
5341 | * idle load balancing details | |
83cd4fe2 VP |
5342 | * - When one of the busy CPUs notice that there may be an idle rebalancing |
5343 | * needed, they will kick the idle load balancer, which then does idle | |
5344 | * load balancing for all the idle CPUs. | |
5345 | */ | |
1e3c88bd | 5346 | static struct { |
83cd4fe2 | 5347 | cpumask_var_t idle_cpus_mask; |
0b005cf5 | 5348 | atomic_t nr_cpus; |
83cd4fe2 VP |
5349 | unsigned long next_balance; /* in jiffy units */ |
5350 | } nohz ____cacheline_aligned; | |
1e3c88bd | 5351 | |
8e7fbcbc | 5352 | static inline int find_new_ilb(int call_cpu) |
1e3c88bd | 5353 | { |
0b005cf5 | 5354 | int ilb = cpumask_first(nohz.idle_cpus_mask); |
1e3c88bd | 5355 | |
786d6dc7 SS |
5356 | if (ilb < nr_cpu_ids && idle_cpu(ilb)) |
5357 | return ilb; | |
5358 | ||
5359 | return nr_cpu_ids; | |
1e3c88bd | 5360 | } |
1e3c88bd | 5361 | |
83cd4fe2 VP |
5362 | /* |
5363 | * Kick a CPU to do the nohz balancing, if it is time for it. We pick the | |
5364 | * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle | |
5365 | * CPU (if there is one). | |
5366 | */ | |
5367 | static void nohz_balancer_kick(int cpu) | |
5368 | { | |
5369 | int ilb_cpu; | |
5370 | ||
5371 | nohz.next_balance++; | |
5372 | ||
0b005cf5 | 5373 | ilb_cpu = find_new_ilb(cpu); |
83cd4fe2 | 5374 | |
0b005cf5 SS |
5375 | if (ilb_cpu >= nr_cpu_ids) |
5376 | return; | |
83cd4fe2 | 5377 | |
cd490c5b | 5378 | if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu))) |
1c792db7 SS |
5379 | return; |
5380 | /* | |
5381 | * Use smp_send_reschedule() instead of resched_cpu(). | |
5382 | * This way we generate a sched IPI on the target cpu which | |
5383 | * is idle. And the softirq performing nohz idle load balance | |
5384 | * will be run before returning from the IPI. | |
5385 | */ | |
5386 | smp_send_reschedule(ilb_cpu); | |
83cd4fe2 VP |
5387 | return; |
5388 | } | |
5389 | ||
c1cc017c | 5390 | static inline void nohz_balance_exit_idle(int cpu) |
71325960 SS |
5391 | { |
5392 | if (unlikely(test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))) { | |
5393 | cpumask_clear_cpu(cpu, nohz.idle_cpus_mask); | |
5394 | atomic_dec(&nohz.nr_cpus); | |
5395 | clear_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)); | |
5396 | } | |
5397 | } | |
5398 | ||
69e1e811 SS |
5399 | static inline void set_cpu_sd_state_busy(void) |
5400 | { | |
5401 | struct sched_domain *sd; | |
5402 | int cpu = smp_processor_id(); | |
5403 | ||
5404 | if (!test_bit(NOHZ_IDLE, nohz_flags(cpu))) | |
5405 | return; | |
5406 | clear_bit(NOHZ_IDLE, nohz_flags(cpu)); | |
5407 | ||
5408 | rcu_read_lock(); | |
5409 | for_each_domain(cpu, sd) | |
5410 | atomic_inc(&sd->groups->sgp->nr_busy_cpus); | |
5411 | rcu_read_unlock(); | |
5412 | } | |
5413 | ||
5414 | void set_cpu_sd_state_idle(void) | |
5415 | { | |
5416 | struct sched_domain *sd; | |
5417 | int cpu = smp_processor_id(); | |
5418 | ||
5419 | if (test_bit(NOHZ_IDLE, nohz_flags(cpu))) | |
5420 | return; | |
5421 | set_bit(NOHZ_IDLE, nohz_flags(cpu)); | |
5422 | ||
5423 | rcu_read_lock(); | |
5424 | for_each_domain(cpu, sd) | |
5425 | atomic_dec(&sd->groups->sgp->nr_busy_cpus); | |
5426 | rcu_read_unlock(); | |
5427 | } | |
5428 | ||
1e3c88bd | 5429 | /* |
c1cc017c | 5430 | * This routine will record that the cpu is going idle with tick stopped. |
0b005cf5 | 5431 | * This info will be used in performing idle load balancing in the future. |
1e3c88bd | 5432 | */ |
c1cc017c | 5433 | void nohz_balance_enter_idle(int cpu) |
1e3c88bd | 5434 | { |
71325960 SS |
5435 | /* |
5436 | * If this cpu is going down, then nothing needs to be done. | |
5437 | */ | |
5438 | if (!cpu_active(cpu)) | |
5439 | return; | |
5440 | ||
c1cc017c AS |
5441 | if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu))) |
5442 | return; | |
1e3c88bd | 5443 | |
c1cc017c AS |
5444 | cpumask_set_cpu(cpu, nohz.idle_cpus_mask); |
5445 | atomic_inc(&nohz.nr_cpus); | |
5446 | set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)); | |
1e3c88bd | 5447 | } |
71325960 SS |
5448 | |
5449 | static int __cpuinit sched_ilb_notifier(struct notifier_block *nfb, | |
5450 | unsigned long action, void *hcpu) | |
5451 | { | |
5452 | switch (action & ~CPU_TASKS_FROZEN) { | |
5453 | case CPU_DYING: | |
c1cc017c | 5454 | nohz_balance_exit_idle(smp_processor_id()); |
71325960 SS |
5455 | return NOTIFY_OK; |
5456 | default: | |
5457 | return NOTIFY_DONE; | |
5458 | } | |
5459 | } | |
1e3c88bd PZ |
5460 | #endif |
5461 | ||
5462 | static DEFINE_SPINLOCK(balancing); | |
5463 | ||
49c022e6 PZ |
5464 | /* |
5465 | * Scale the max load_balance interval with the number of CPUs in the system. | |
5466 | * This trades load-balance latency on larger machines for less cross talk. | |
5467 | */ | |
029632fb | 5468 | void update_max_interval(void) |
49c022e6 PZ |
5469 | { |
5470 | max_load_balance_interval = HZ*num_online_cpus()/10; | |
5471 | } | |
5472 | ||
1e3c88bd PZ |
5473 | /* |
5474 | * It checks each scheduling domain to see if it is due to be balanced, | |
5475 | * and initiates a balancing operation if so. | |
5476 | * | |
5477 | * Balancing parameters are set up in arch_init_sched_domains. | |
5478 | */ | |
5479 | static void rebalance_domains(int cpu, enum cpu_idle_type idle) | |
5480 | { | |
5481 | int balance = 1; | |
5482 | struct rq *rq = cpu_rq(cpu); | |
5483 | unsigned long interval; | |
04f733b4 | 5484 | struct sched_domain *sd; |
1e3c88bd PZ |
5485 | /* Earliest time when we have to do rebalance again */ |
5486 | unsigned long next_balance = jiffies + 60*HZ; | |
5487 | int update_next_balance = 0; | |
5488 | int need_serialize; | |
5489 | ||
48a16753 | 5490 | update_blocked_averages(cpu); |
2069dd75 | 5491 | |
dce840a0 | 5492 | rcu_read_lock(); |
1e3c88bd PZ |
5493 | for_each_domain(cpu, sd) { |
5494 | if (!(sd->flags & SD_LOAD_BALANCE)) | |
5495 | continue; | |
5496 | ||
5497 | interval = sd->balance_interval; | |
5498 | if (idle != CPU_IDLE) | |
5499 | interval *= sd->busy_factor; | |
5500 | ||
5501 | /* scale ms to jiffies */ | |
5502 | interval = msecs_to_jiffies(interval); | |
49c022e6 | 5503 | interval = clamp(interval, 1UL, max_load_balance_interval); |
1e3c88bd PZ |
5504 | |
5505 | need_serialize = sd->flags & SD_SERIALIZE; | |
5506 | ||
5507 | if (need_serialize) { | |
5508 | if (!spin_trylock(&balancing)) | |
5509 | goto out; | |
5510 | } | |
5511 | ||
5512 | if (time_after_eq(jiffies, sd->last_balance + interval)) { | |
5513 | if (load_balance(cpu, rq, sd, idle, &balance)) { | |
5514 | /* | |
5515 | * We've pulled tasks over so either we're no | |
c186fafe | 5516 | * longer idle. |
1e3c88bd PZ |
5517 | */ |
5518 | idle = CPU_NOT_IDLE; | |
5519 | } | |
5520 | sd->last_balance = jiffies; | |
5521 | } | |
5522 | if (need_serialize) | |
5523 | spin_unlock(&balancing); | |
5524 | out: | |
5525 | if (time_after(next_balance, sd->last_balance + interval)) { | |
5526 | next_balance = sd->last_balance + interval; | |
5527 | update_next_balance = 1; | |
5528 | } | |
5529 | ||
5530 | /* | |
5531 | * Stop the load balance at this level. There is another | |
5532 | * CPU in our sched group which is doing load balancing more | |
5533 | * actively. | |
5534 | */ | |
5535 | if (!balance) | |
5536 | break; | |
5537 | } | |
dce840a0 | 5538 | rcu_read_unlock(); |
1e3c88bd PZ |
5539 | |
5540 | /* | |
5541 | * next_balance will be updated only when there is a need. | |
5542 | * When the cpu is attached to null domain for ex, it will not be | |
5543 | * updated. | |
5544 | */ | |
5545 | if (likely(update_next_balance)) | |
5546 | rq->next_balance = next_balance; | |
5547 | } | |
5548 | ||
83cd4fe2 | 5549 | #ifdef CONFIG_NO_HZ |
1e3c88bd | 5550 | /* |
83cd4fe2 | 5551 | * In CONFIG_NO_HZ case, the idle balance kickee will do the |
1e3c88bd PZ |
5552 | * rebalancing for all the cpus for whom scheduler ticks are stopped. |
5553 | */ | |
83cd4fe2 VP |
5554 | static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) |
5555 | { | |
5556 | struct rq *this_rq = cpu_rq(this_cpu); | |
5557 | struct rq *rq; | |
5558 | int balance_cpu; | |
5559 | ||
1c792db7 SS |
5560 | if (idle != CPU_IDLE || |
5561 | !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu))) | |
5562 | goto end; | |
83cd4fe2 VP |
5563 | |
5564 | for_each_cpu(balance_cpu, nohz.idle_cpus_mask) { | |
8a6d42d1 | 5565 | if (balance_cpu == this_cpu || !idle_cpu(balance_cpu)) |
83cd4fe2 VP |
5566 | continue; |
5567 | ||
5568 | /* | |
5569 | * If this cpu gets work to do, stop the load balancing | |
5570 | * work being done for other cpus. Next load | |
5571 | * balancing owner will pick it up. | |
5572 | */ | |
1c792db7 | 5573 | if (need_resched()) |
83cd4fe2 | 5574 | break; |
83cd4fe2 | 5575 | |
5ed4f1d9 VG |
5576 | rq = cpu_rq(balance_cpu); |
5577 | ||
5578 | raw_spin_lock_irq(&rq->lock); | |
5579 | update_rq_clock(rq); | |
5580 | update_idle_cpu_load(rq); | |
5581 | raw_spin_unlock_irq(&rq->lock); | |
83cd4fe2 VP |
5582 | |
5583 | rebalance_domains(balance_cpu, CPU_IDLE); | |
5584 | ||
83cd4fe2 VP |
5585 | if (time_after(this_rq->next_balance, rq->next_balance)) |
5586 | this_rq->next_balance = rq->next_balance; | |
5587 | } | |
5588 | nohz.next_balance = this_rq->next_balance; | |
1c792db7 SS |
5589 | end: |
5590 | clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)); | |
83cd4fe2 VP |
5591 | } |
5592 | ||
5593 | /* | |
0b005cf5 SS |
5594 | * Current heuristic for kicking the idle load balancer in the presence |
5595 | * of an idle cpu is the system. | |
5596 | * - This rq has more than one task. | |
5597 | * - At any scheduler domain level, this cpu's scheduler group has multiple | |
5598 | * busy cpu's exceeding the group's power. | |
5599 | * - For SD_ASYM_PACKING, if the lower numbered cpu's in the scheduler | |
5600 | * domain span are idle. | |
83cd4fe2 VP |
5601 | */ |
5602 | static inline int nohz_kick_needed(struct rq *rq, int cpu) | |
5603 | { | |
5604 | unsigned long now = jiffies; | |
0b005cf5 | 5605 | struct sched_domain *sd; |
83cd4fe2 | 5606 | |
1c792db7 | 5607 | if (unlikely(idle_cpu(cpu))) |
83cd4fe2 VP |
5608 | return 0; |
5609 | ||
1c792db7 SS |
5610 | /* |
5611 | * We may be recently in ticked or tickless idle mode. At the first | |
5612 | * busy tick after returning from idle, we will update the busy stats. | |
5613 | */ | |
69e1e811 | 5614 | set_cpu_sd_state_busy(); |
c1cc017c | 5615 | nohz_balance_exit_idle(cpu); |
0b005cf5 SS |
5616 | |
5617 | /* | |
5618 | * None are in tickless mode and hence no need for NOHZ idle load | |
5619 | * balancing. | |
5620 | */ | |
5621 | if (likely(!atomic_read(&nohz.nr_cpus))) | |
5622 | return 0; | |
1c792db7 SS |
5623 | |
5624 | if (time_before(now, nohz.next_balance)) | |
83cd4fe2 VP |
5625 | return 0; |
5626 | ||
0b005cf5 SS |
5627 | if (rq->nr_running >= 2) |
5628 | goto need_kick; | |
83cd4fe2 | 5629 | |
067491b7 | 5630 | rcu_read_lock(); |
0b005cf5 SS |
5631 | for_each_domain(cpu, sd) { |
5632 | struct sched_group *sg = sd->groups; | |
5633 | struct sched_group_power *sgp = sg->sgp; | |
5634 | int nr_busy = atomic_read(&sgp->nr_busy_cpus); | |
83cd4fe2 | 5635 | |
0b005cf5 | 5636 | if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1) |
067491b7 | 5637 | goto need_kick_unlock; |
0b005cf5 SS |
5638 | |
5639 | if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight | |
5640 | && (cpumask_first_and(nohz.idle_cpus_mask, | |
5641 | sched_domain_span(sd)) < cpu)) | |
067491b7 | 5642 | goto need_kick_unlock; |
0b005cf5 SS |
5643 | |
5644 | if (!(sd->flags & (SD_SHARE_PKG_RESOURCES | SD_ASYM_PACKING))) | |
5645 | break; | |
83cd4fe2 | 5646 | } |
067491b7 | 5647 | rcu_read_unlock(); |
83cd4fe2 | 5648 | return 0; |
067491b7 PZ |
5649 | |
5650 | need_kick_unlock: | |
5651 | rcu_read_unlock(); | |
0b005cf5 SS |
5652 | need_kick: |
5653 | return 1; | |
83cd4fe2 VP |
5654 | } |
5655 | #else | |
5656 | static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) { } | |
5657 | #endif | |
5658 | ||
5659 | /* | |
5660 | * run_rebalance_domains is triggered when needed from the scheduler tick. | |
5661 | * Also triggered for nohz idle balancing (with nohz_balancing_kick set). | |
5662 | */ | |
1e3c88bd PZ |
5663 | static void run_rebalance_domains(struct softirq_action *h) |
5664 | { | |
5665 | int this_cpu = smp_processor_id(); | |
5666 | struct rq *this_rq = cpu_rq(this_cpu); | |
6eb57e0d | 5667 | enum cpu_idle_type idle = this_rq->idle_balance ? |
1e3c88bd PZ |
5668 | CPU_IDLE : CPU_NOT_IDLE; |
5669 | ||
5670 | rebalance_domains(this_cpu, idle); | |
5671 | ||
1e3c88bd | 5672 | /* |
83cd4fe2 | 5673 | * If this cpu has a pending nohz_balance_kick, then do the |
1e3c88bd PZ |
5674 | * balancing on behalf of the other idle cpus whose ticks are |
5675 | * stopped. | |
5676 | */ | |
83cd4fe2 | 5677 | nohz_idle_balance(this_cpu, idle); |
1e3c88bd PZ |
5678 | } |
5679 | ||
5680 | static inline int on_null_domain(int cpu) | |
5681 | { | |
90a6501f | 5682 | return !rcu_dereference_sched(cpu_rq(cpu)->sd); |
1e3c88bd PZ |
5683 | } |
5684 | ||
5685 | /* | |
5686 | * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. | |
1e3c88bd | 5687 | */ |
029632fb | 5688 | void trigger_load_balance(struct rq *rq, int cpu) |
1e3c88bd | 5689 | { |
1e3c88bd PZ |
5690 | /* Don't need to rebalance while attached to NULL domain */ |
5691 | if (time_after_eq(jiffies, rq->next_balance) && | |
5692 | likely(!on_null_domain(cpu))) | |
5693 | raise_softirq(SCHED_SOFTIRQ); | |
83cd4fe2 | 5694 | #ifdef CONFIG_NO_HZ |
1c792db7 | 5695 | if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu))) |
83cd4fe2 VP |
5696 | nohz_balancer_kick(cpu); |
5697 | #endif | |
1e3c88bd PZ |
5698 | } |
5699 | ||
0bcdcf28 CE |
5700 | static void rq_online_fair(struct rq *rq) |
5701 | { | |
5702 | update_sysctl(); | |
5703 | } | |
5704 | ||
5705 | static void rq_offline_fair(struct rq *rq) | |
5706 | { | |
5707 | update_sysctl(); | |
a4c96ae3 PB |
5708 | |
5709 | /* Ensure any throttled groups are reachable by pick_next_task */ | |
5710 | unthrottle_offline_cfs_rqs(rq); | |
0bcdcf28 CE |
5711 | } |
5712 | ||
55e12e5e | 5713 | #endif /* CONFIG_SMP */ |
e1d1484f | 5714 | |
bf0f6f24 IM |
5715 | /* |
5716 | * scheduler tick hitting a task of our scheduling class: | |
5717 | */ | |
8f4d37ec | 5718 | static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued) |
bf0f6f24 IM |
5719 | { |
5720 | struct cfs_rq *cfs_rq; | |
5721 | struct sched_entity *se = &curr->se; | |
5722 | ||
5723 | for_each_sched_entity(se) { | |
5724 | cfs_rq = cfs_rq_of(se); | |
8f4d37ec | 5725 | entity_tick(cfs_rq, se, queued); |
bf0f6f24 | 5726 | } |
18bf2805 | 5727 | |
cbee9f88 PZ |
5728 | if (sched_feat_numa(NUMA)) |
5729 | task_tick_numa(rq, curr); | |
3d59eebc | 5730 | |
18bf2805 | 5731 | update_rq_runnable_avg(rq, 1); |
bf0f6f24 IM |
5732 | } |
5733 | ||
5734 | /* | |
cd29fe6f PZ |
5735 | * called on fork with the child task as argument from the parent's context |
5736 | * - child not yet on the tasklist | |
5737 | * - preemption disabled | |
bf0f6f24 | 5738 | */ |
cd29fe6f | 5739 | static void task_fork_fair(struct task_struct *p) |
bf0f6f24 | 5740 | { |
4fc420c9 DN |
5741 | struct cfs_rq *cfs_rq; |
5742 | struct sched_entity *se = &p->se, *curr; | |
00bf7bfc | 5743 | int this_cpu = smp_processor_id(); |
cd29fe6f PZ |
5744 | struct rq *rq = this_rq(); |
5745 | unsigned long flags; | |
5746 | ||
05fa785c | 5747 | raw_spin_lock_irqsave(&rq->lock, flags); |
bf0f6f24 | 5748 | |
861d034e PZ |
5749 | update_rq_clock(rq); |
5750 | ||
4fc420c9 DN |
5751 | cfs_rq = task_cfs_rq(current); |
5752 | curr = cfs_rq->curr; | |
5753 | ||
b0a0f667 PM |
5754 | if (unlikely(task_cpu(p) != this_cpu)) { |
5755 | rcu_read_lock(); | |
cd29fe6f | 5756 | __set_task_cpu(p, this_cpu); |
b0a0f667 PM |
5757 | rcu_read_unlock(); |
5758 | } | |
bf0f6f24 | 5759 | |
7109c442 | 5760 | update_curr(cfs_rq); |
cd29fe6f | 5761 | |
b5d9d734 MG |
5762 | if (curr) |
5763 | se->vruntime = curr->vruntime; | |
aeb73b04 | 5764 | place_entity(cfs_rq, se, 1); |
4d78e7b6 | 5765 | |
cd29fe6f | 5766 | if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) { |
87fefa38 | 5767 | /* |
edcb60a3 IM |
5768 | * Upon rescheduling, sched_class::put_prev_task() will place |
5769 | * 'current' within the tree based on its new key value. | |
5770 | */ | |
4d78e7b6 | 5771 | swap(curr->vruntime, se->vruntime); |
aec0a514 | 5772 | resched_task(rq->curr); |
4d78e7b6 | 5773 | } |
bf0f6f24 | 5774 | |
88ec22d3 PZ |
5775 | se->vruntime -= cfs_rq->min_vruntime; |
5776 | ||
05fa785c | 5777 | raw_spin_unlock_irqrestore(&rq->lock, flags); |
bf0f6f24 IM |
5778 | } |
5779 | ||
cb469845 SR |
5780 | /* |
5781 | * Priority of the task has changed. Check to see if we preempt | |
5782 | * the current task. | |
5783 | */ | |
da7a735e PZ |
5784 | static void |
5785 | prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio) | |
cb469845 | 5786 | { |
da7a735e PZ |
5787 | if (!p->se.on_rq) |
5788 | return; | |
5789 | ||
cb469845 SR |
5790 | /* |
5791 | * Reschedule if we are currently running on this runqueue and | |
5792 | * our priority decreased, or if we are not currently running on | |
5793 | * this runqueue and our priority is higher than the current's | |
5794 | */ | |
da7a735e | 5795 | if (rq->curr == p) { |
cb469845 SR |
5796 | if (p->prio > oldprio) |
5797 | resched_task(rq->curr); | |
5798 | } else | |
15afe09b | 5799 | check_preempt_curr(rq, p, 0); |
cb469845 SR |
5800 | } |
5801 | ||
da7a735e PZ |
5802 | static void switched_from_fair(struct rq *rq, struct task_struct *p) |
5803 | { | |
5804 | struct sched_entity *se = &p->se; | |
5805 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
5806 | ||
5807 | /* | |
5808 | * Ensure the task's vruntime is normalized, so that when its | |
5809 | * switched back to the fair class the enqueue_entity(.flags=0) will | |
5810 | * do the right thing. | |
5811 | * | |
5812 | * If it was on_rq, then the dequeue_entity(.flags=0) will already | |
5813 | * have normalized the vruntime, if it was !on_rq, then only when | |
5814 | * the task is sleeping will it still have non-normalized vruntime. | |
5815 | */ | |
5816 | if (!se->on_rq && p->state != TASK_RUNNING) { | |
5817 | /* | |
5818 | * Fix up our vruntime so that the current sleep doesn't | |
5819 | * cause 'unlimited' sleep bonus. | |
5820 | */ | |
5821 | place_entity(cfs_rq, se, 0); | |
5822 | se->vruntime -= cfs_rq->min_vruntime; | |
5823 | } | |
9ee474f5 PT |
5824 | |
5825 | #if defined(CONFIG_FAIR_GROUP_SCHED) && defined(CONFIG_SMP) | |
5826 | /* | |
5827 | * Remove our load from contribution when we leave sched_fair | |
5828 | * and ensure we don't carry in an old decay_count if we | |
5829 | * switch back. | |
5830 | */ | |
5831 | if (p->se.avg.decay_count) { | |
5832 | struct cfs_rq *cfs_rq = cfs_rq_of(&p->se); | |
5833 | __synchronize_entity_decay(&p->se); | |
5834 | subtract_blocked_load_contrib(cfs_rq, | |
5835 | p->se.avg.load_avg_contrib); | |
5836 | } | |
5837 | #endif | |
da7a735e PZ |
5838 | } |
5839 | ||
cb469845 SR |
5840 | /* |
5841 | * We switched to the sched_fair class. | |
5842 | */ | |
da7a735e | 5843 | static void switched_to_fair(struct rq *rq, struct task_struct *p) |
cb469845 | 5844 | { |
da7a735e PZ |
5845 | if (!p->se.on_rq) |
5846 | return; | |
5847 | ||
cb469845 SR |
5848 | /* |
5849 | * We were most likely switched from sched_rt, so | |
5850 | * kick off the schedule if running, otherwise just see | |
5851 | * if we can still preempt the current task. | |
5852 | */ | |
da7a735e | 5853 | if (rq->curr == p) |
cb469845 SR |
5854 | resched_task(rq->curr); |
5855 | else | |
15afe09b | 5856 | check_preempt_curr(rq, p, 0); |
cb469845 SR |
5857 | } |
5858 | ||
83b699ed SV |
5859 | /* Account for a task changing its policy or group. |
5860 | * | |
5861 | * This routine is mostly called to set cfs_rq->curr field when a task | |
5862 | * migrates between groups/classes. | |
5863 | */ | |
5864 | static void set_curr_task_fair(struct rq *rq) | |
5865 | { | |
5866 | struct sched_entity *se = &rq->curr->se; | |
5867 | ||
ec12cb7f PT |
5868 | for_each_sched_entity(se) { |
5869 | struct cfs_rq *cfs_rq = cfs_rq_of(se); | |
5870 | ||
5871 | set_next_entity(cfs_rq, se); | |
5872 | /* ensure bandwidth has been allocated on our new cfs_rq */ | |
5873 | account_cfs_rq_runtime(cfs_rq, 0); | |
5874 | } | |
83b699ed SV |
5875 | } |
5876 | ||
029632fb PZ |
5877 | void init_cfs_rq(struct cfs_rq *cfs_rq) |
5878 | { | |
5879 | cfs_rq->tasks_timeline = RB_ROOT; | |
029632fb PZ |
5880 | cfs_rq->min_vruntime = (u64)(-(1LL << 20)); |
5881 | #ifndef CONFIG_64BIT | |
5882 | cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime; | |
5883 | #endif | |
9ee474f5 PT |
5884 | #if defined(CONFIG_FAIR_GROUP_SCHED) && defined(CONFIG_SMP) |
5885 | atomic64_set(&cfs_rq->decay_counter, 1); | |
aff3e498 | 5886 | atomic64_set(&cfs_rq->removed_load, 0); |
9ee474f5 | 5887 | #endif |
029632fb PZ |
5888 | } |
5889 | ||
810b3817 | 5890 | #ifdef CONFIG_FAIR_GROUP_SCHED |
b2b5ce02 | 5891 | static void task_move_group_fair(struct task_struct *p, int on_rq) |
810b3817 | 5892 | { |
aff3e498 | 5893 | struct cfs_rq *cfs_rq; |
b2b5ce02 PZ |
5894 | /* |
5895 | * If the task was not on the rq at the time of this cgroup movement | |
5896 | * it must have been asleep, sleeping tasks keep their ->vruntime | |
5897 | * absolute on their old rq until wakeup (needed for the fair sleeper | |
5898 | * bonus in place_entity()). | |
5899 | * | |
5900 | * If it was on the rq, we've just 'preempted' it, which does convert | |
5901 | * ->vruntime to a relative base. | |
5902 | * | |
5903 | * Make sure both cases convert their relative position when migrating | |
5904 | * to another cgroup's rq. This does somewhat interfere with the | |
5905 | * fair sleeper stuff for the first placement, but who cares. | |
5906 | */ | |
7ceff013 DN |
5907 | /* |
5908 | * When !on_rq, vruntime of the task has usually NOT been normalized. | |
5909 | * But there are some cases where it has already been normalized: | |
5910 | * | |
5911 | * - Moving a forked child which is waiting for being woken up by | |
5912 | * wake_up_new_task(). | |
62af3783 DN |
5913 | * - Moving a task which has been woken up by try_to_wake_up() and |
5914 | * waiting for actually being woken up by sched_ttwu_pending(). | |
7ceff013 DN |
5915 | * |
5916 | * To prevent boost or penalty in the new cfs_rq caused by delta | |
5917 | * min_vruntime between the two cfs_rqs, we skip vruntime adjustment. | |
5918 | */ | |
62af3783 | 5919 | if (!on_rq && (!p->se.sum_exec_runtime || p->state == TASK_WAKING)) |
7ceff013 DN |
5920 | on_rq = 1; |
5921 | ||
b2b5ce02 PZ |
5922 | if (!on_rq) |
5923 | p->se.vruntime -= cfs_rq_of(&p->se)->min_vruntime; | |
5924 | set_task_rq(p, task_cpu(p)); | |
aff3e498 PT |
5925 | if (!on_rq) { |
5926 | cfs_rq = cfs_rq_of(&p->se); | |
5927 | p->se.vruntime += cfs_rq->min_vruntime; | |
5928 | #ifdef CONFIG_SMP | |
5929 | /* | |
5930 | * migrate_task_rq_fair() will have removed our previous | |
5931 | * contribution, but we must synchronize for ongoing future | |
5932 | * decay. | |
5933 | */ | |
5934 | p->se.avg.decay_count = atomic64_read(&cfs_rq->decay_counter); | |
5935 | cfs_rq->blocked_load_avg += p->se.avg.load_avg_contrib; | |
5936 | #endif | |
5937 | } | |
810b3817 | 5938 | } |
029632fb PZ |
5939 | |
5940 | void free_fair_sched_group(struct task_group *tg) | |
5941 | { | |
5942 | int i; | |
5943 | ||
5944 | destroy_cfs_bandwidth(tg_cfs_bandwidth(tg)); | |
5945 | ||
5946 | for_each_possible_cpu(i) { | |
5947 | if (tg->cfs_rq) | |
5948 | kfree(tg->cfs_rq[i]); | |
5949 | if (tg->se) | |
5950 | kfree(tg->se[i]); | |
5951 | } | |
5952 | ||
5953 | kfree(tg->cfs_rq); | |
5954 | kfree(tg->se); | |
5955 | } | |
5956 | ||
5957 | int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) | |
5958 | { | |
5959 | struct cfs_rq *cfs_rq; | |
5960 | struct sched_entity *se; | |
5961 | int i; | |
5962 | ||
5963 | tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL); | |
5964 | if (!tg->cfs_rq) | |
5965 | goto err; | |
5966 | tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL); | |
5967 | if (!tg->se) | |
5968 | goto err; | |
5969 | ||
5970 | tg->shares = NICE_0_LOAD; | |
5971 | ||
5972 | init_cfs_bandwidth(tg_cfs_bandwidth(tg)); | |
5973 | ||
5974 | for_each_possible_cpu(i) { | |
5975 | cfs_rq = kzalloc_node(sizeof(struct cfs_rq), | |
5976 | GFP_KERNEL, cpu_to_node(i)); | |
5977 | if (!cfs_rq) | |
5978 | goto err; | |
5979 | ||
5980 | se = kzalloc_node(sizeof(struct sched_entity), | |
5981 | GFP_KERNEL, cpu_to_node(i)); | |
5982 | if (!se) | |
5983 | goto err_free_rq; | |
5984 | ||
5985 | init_cfs_rq(cfs_rq); | |
5986 | init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]); | |
5987 | } | |
5988 | ||
5989 | return 1; | |
5990 | ||
5991 | err_free_rq: | |
5992 | kfree(cfs_rq); | |
5993 | err: | |
5994 | return 0; | |
5995 | } | |
5996 | ||
5997 | void unregister_fair_sched_group(struct task_group *tg, int cpu) | |
5998 | { | |
5999 | struct rq *rq = cpu_rq(cpu); | |
6000 | unsigned long flags; | |
6001 | ||
6002 | /* | |
6003 | * Only empty task groups can be destroyed; so we can speculatively | |
6004 | * check on_list without danger of it being re-added. | |
6005 | */ | |
6006 | if (!tg->cfs_rq[cpu]->on_list) | |
6007 | return; | |
6008 | ||
6009 | raw_spin_lock_irqsave(&rq->lock, flags); | |
6010 | list_del_leaf_cfs_rq(tg->cfs_rq[cpu]); | |
6011 | raw_spin_unlock_irqrestore(&rq->lock, flags); | |
6012 | } | |
6013 | ||
6014 | void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq, | |
6015 | struct sched_entity *se, int cpu, | |
6016 | struct sched_entity *parent) | |
6017 | { | |
6018 | struct rq *rq = cpu_rq(cpu); | |
6019 | ||
6020 | cfs_rq->tg = tg; | |
6021 | cfs_rq->rq = rq; | |
029632fb PZ |
6022 | init_cfs_rq_runtime(cfs_rq); |
6023 | ||
6024 | tg->cfs_rq[cpu] = cfs_rq; | |
6025 | tg->se[cpu] = se; | |
6026 | ||
6027 | /* se could be NULL for root_task_group */ | |
6028 | if (!se) | |
6029 | return; | |
6030 | ||
6031 | if (!parent) | |
6032 | se->cfs_rq = &rq->cfs; | |
6033 | else | |
6034 | se->cfs_rq = parent->my_q; | |
6035 | ||
6036 | se->my_q = cfs_rq; | |
6037 | update_load_set(&se->load, 0); | |
6038 | se->parent = parent; | |
6039 | } | |
6040 | ||
6041 | static DEFINE_MUTEX(shares_mutex); | |
6042 | ||
6043 | int sched_group_set_shares(struct task_group *tg, unsigned long shares) | |
6044 | { | |
6045 | int i; | |
6046 | unsigned long flags; | |
6047 | ||
6048 | /* | |
6049 | * We can't change the weight of the root cgroup. | |
6050 | */ | |
6051 | if (!tg->se[0]) | |
6052 | return -EINVAL; | |
6053 | ||
6054 | shares = clamp(shares, scale_load(MIN_SHARES), scale_load(MAX_SHARES)); | |
6055 | ||
6056 | mutex_lock(&shares_mutex); | |
6057 | if (tg->shares == shares) | |
6058 | goto done; | |
6059 | ||
6060 | tg->shares = shares; | |
6061 | for_each_possible_cpu(i) { | |
6062 | struct rq *rq = cpu_rq(i); | |
6063 | struct sched_entity *se; | |
6064 | ||
6065 | se = tg->se[i]; | |
6066 | /* Propagate contribution to hierarchy */ | |
6067 | raw_spin_lock_irqsave(&rq->lock, flags); | |
17bc14b7 | 6068 | for_each_sched_entity(se) |
029632fb PZ |
6069 | update_cfs_shares(group_cfs_rq(se)); |
6070 | raw_spin_unlock_irqrestore(&rq->lock, flags); | |
6071 | } | |
6072 | ||
6073 | done: | |
6074 | mutex_unlock(&shares_mutex); | |
6075 | return 0; | |
6076 | } | |
6077 | #else /* CONFIG_FAIR_GROUP_SCHED */ | |
6078 | ||
6079 | void free_fair_sched_group(struct task_group *tg) { } | |
6080 | ||
6081 | int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) | |
6082 | { | |
6083 | return 1; | |
6084 | } | |
6085 | ||
6086 | void unregister_fair_sched_group(struct task_group *tg, int cpu) { } | |
6087 | ||
6088 | #endif /* CONFIG_FAIR_GROUP_SCHED */ | |
6089 | ||
810b3817 | 6090 | |
6d686f45 | 6091 | static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task) |
0d721cea PW |
6092 | { |
6093 | struct sched_entity *se = &task->se; | |
0d721cea PW |
6094 | unsigned int rr_interval = 0; |
6095 | ||
6096 | /* | |
6097 | * Time slice is 0 for SCHED_OTHER tasks that are on an otherwise | |
6098 | * idle runqueue: | |
6099 | */ | |
0d721cea PW |
6100 | if (rq->cfs.load.weight) |
6101 | rr_interval = NS_TO_JIFFIES(sched_slice(&rq->cfs, se)); | |
0d721cea PW |
6102 | |
6103 | return rr_interval; | |
6104 | } | |
6105 | ||
bf0f6f24 IM |
6106 | /* |
6107 | * All the scheduling class methods: | |
6108 | */ | |
029632fb | 6109 | const struct sched_class fair_sched_class = { |
5522d5d5 | 6110 | .next = &idle_sched_class, |
bf0f6f24 IM |
6111 | .enqueue_task = enqueue_task_fair, |
6112 | .dequeue_task = dequeue_task_fair, | |
6113 | .yield_task = yield_task_fair, | |
d95f4122 | 6114 | .yield_to_task = yield_to_task_fair, |
bf0f6f24 | 6115 | |
2e09bf55 | 6116 | .check_preempt_curr = check_preempt_wakeup, |
bf0f6f24 IM |
6117 | |
6118 | .pick_next_task = pick_next_task_fair, | |
6119 | .put_prev_task = put_prev_task_fair, | |
6120 | ||
681f3e68 | 6121 | #ifdef CONFIG_SMP |
4ce72a2c | 6122 | .select_task_rq = select_task_rq_fair, |
f4e26b12 | 6123 | #ifdef CONFIG_FAIR_GROUP_SCHED |
0a74bef8 | 6124 | .migrate_task_rq = migrate_task_rq_fair, |
f4e26b12 | 6125 | #endif |
0bcdcf28 CE |
6126 | .rq_online = rq_online_fair, |
6127 | .rq_offline = rq_offline_fair, | |
88ec22d3 PZ |
6128 | |
6129 | .task_waking = task_waking_fair, | |
681f3e68 | 6130 | #endif |
bf0f6f24 | 6131 | |
83b699ed | 6132 | .set_curr_task = set_curr_task_fair, |
bf0f6f24 | 6133 | .task_tick = task_tick_fair, |
cd29fe6f | 6134 | .task_fork = task_fork_fair, |
cb469845 SR |
6135 | |
6136 | .prio_changed = prio_changed_fair, | |
da7a735e | 6137 | .switched_from = switched_from_fair, |
cb469845 | 6138 | .switched_to = switched_to_fair, |
810b3817 | 6139 | |
0d721cea PW |
6140 | .get_rr_interval = get_rr_interval_fair, |
6141 | ||
810b3817 | 6142 | #ifdef CONFIG_FAIR_GROUP_SCHED |
b2b5ce02 | 6143 | .task_move_group = task_move_group_fair, |
810b3817 | 6144 | #endif |
bf0f6f24 IM |
6145 | }; |
6146 | ||
6147 | #ifdef CONFIG_SCHED_DEBUG | |
029632fb | 6148 | void print_cfs_stats(struct seq_file *m, int cpu) |
bf0f6f24 | 6149 | { |
bf0f6f24 IM |
6150 | struct cfs_rq *cfs_rq; |
6151 | ||
5973e5b9 | 6152 | rcu_read_lock(); |
c3b64f1e | 6153 | for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq) |
5cef9eca | 6154 | print_cfs_rq(m, cpu, cfs_rq); |
5973e5b9 | 6155 | rcu_read_unlock(); |
bf0f6f24 IM |
6156 | } |
6157 | #endif | |
029632fb PZ |
6158 | |
6159 | __init void init_sched_fair_class(void) | |
6160 | { | |
6161 | #ifdef CONFIG_SMP | |
6162 | open_softirq(SCHED_SOFTIRQ, run_rebalance_domains); | |
6163 | ||
6164 | #ifdef CONFIG_NO_HZ | |
554cecaf | 6165 | nohz.next_balance = jiffies; |
029632fb | 6166 | zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT); |
71325960 | 6167 | cpu_notifier(sched_ilb_notifier, 0); |
029632fb PZ |
6168 | #endif |
6169 | #endif /* SMP */ | |
6170 | ||
6171 | } |