Blame view
kernel/sched/fair.c
198 KB
bf0f6f24a sched: cfs core, ... |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
/* * Completely Fair Scheduling (CFS) Class (SCHED_NORMAL/SCHED_BATCH) * * Copyright (C) 2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com> * * Interactivity improvements by Mike Galbraith * (C) 2007 Mike Galbraith <efault@gmx.de> * * Various enhancements by Dmitry Adamushko. * (C) 2007 Dmitry Adamushko <dmitry.adamushko@gmail.com> * * Group scheduling enhancements by Srivatsa Vaddagiri * Copyright IBM Corporation, 2007 * Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com> * * Scaled math optimizations by Thomas Gleixner * Copyright (C) 2007, Thomas Gleixner <tglx@linutronix.de> |
218050855 sched: adaptive s... |
18 19 20 |
* * Adaptive scheduling granularity, math enhancements by Peter Zijlstra * Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com> |
bf0f6f24a sched: cfs core, ... |
21 |
*/ |
9745512ce sched: latencytop... |
22 |
#include <linux/latencytop.h> |
1983a922a sched: Make tunab... |
23 |
#include <linux/sched.h> |
3436ae129 sched: Fix rebala... |
24 |
#include <linux/cpumask.h> |
029632fbb sched: Make separ... |
25 26 27 |
#include <linux/slab.h> #include <linux/profile.h> #include <linux/interrupt.h> |
cbee9f88e mm: numa: Add fau... |
28 |
#include <linux/mempolicy.h> |
e14808b49 mm: numa: Rate li... |
29 |
#include <linux/migrate.h> |
cbee9f88e mm: numa: Add fau... |
30 |
#include <linux/task_work.h> |
029632fbb sched: Make separ... |
31 32 33 34 |
#include <trace/events/sched.h> #include "sched.h" |
9745512ce sched: latencytop... |
35 |
|
bf0f6f24a sched: cfs core, ... |
36 |
/* |
218050855 sched: adaptive s... |
37 |
* Targeted preemption latency for CPU-bound tasks: |
864616ee6 sched: Comment up... |
38 |
* (default: 6ms * (1 + ilog(ncpus)), units: nanoseconds) |
bf0f6f24a sched: cfs core, ... |
39 |
* |
218050855 sched: adaptive s... |
40 |
* NOTE: this latency value is not the same as the concept of |
d274a4cee sched: update com... |
41 42 43 |
* 'timeslice length' - timeslices in CFS are of variable length * and have no persistent notion like in traditional, time-slice * based scheduling concepts. |
bf0f6f24a sched: cfs core, ... |
44 |
* |
d274a4cee sched: update com... |
45 46 |
* (to see the precise effective timeslice length of your workload, * run vmstat and monitor the context-switches (cs) field) |
bf0f6f24a sched: cfs core, ... |
47 |
*/ |
21406928a sched: Tweak sche... |
48 49 |
unsigned int sysctl_sched_latency = 6000000ULL; unsigned int normalized_sysctl_sched_latency = 6000000ULL; |
2bd8e6d42 sched: use consta... |
50 51 |
/* |
1983a922a sched: Make tunab... |
52 53 54 55 56 57 58 59 60 61 62 63 |
* The initial- and re-scaling of tunables is configurable * (default SCHED_TUNABLESCALING_LOG = *(1+ilog(ncpus)) * * Options are: * SCHED_TUNABLESCALING_NONE - unscaled, always *1 * SCHED_TUNABLESCALING_LOG - scaled logarithmical, *1+ilog(ncpus) * SCHED_TUNABLESCALING_LINEAR - scaled linear, *ncpus */ enum sched_tunable_scaling sysctl_sched_tunable_scaling = SCHED_TUNABLESCALING_LOG; /* |
b2be5e96d sched: reintroduc... |
64 |
* Minimal preemption granularity for CPU-bound tasks: |
864616ee6 sched: Comment up... |
65 |
* (default: 0.75 msec * (1 + ilog(ncpus)), units: nanoseconds) |
2bd8e6d42 sched: use consta... |
66 |
*/ |
0bf377bbb sched: Improve la... |
67 68 |
unsigned int sysctl_sched_min_granularity = 750000ULL; unsigned int normalized_sysctl_sched_min_granularity = 750000ULL; |
218050855 sched: adaptive s... |
69 70 |
/* |
b2be5e96d sched: reintroduc... |
71 72 |
* is kept at sysctl_sched_latency / sysctl_sched_min_granularity */ |
0bf377bbb sched: Improve la... |
73 |
static unsigned int sched_nr_latency = 8; |
b2be5e96d sched: reintroduc... |
74 75 |
/* |
2bba22c50 sched: Turn off c... |
76 |
* After fork, child runs first. If set to 0 (default) then |
b2be5e96d sched: reintroduc... |
77 |
* parent will (try to) run first. |
218050855 sched: adaptive s... |
78 |
*/ |
2bba22c50 sched: Turn off c... |
79 |
unsigned int sysctl_sched_child_runs_first __read_mostly; |
bf0f6f24a sched: cfs core, ... |
80 81 |
/* |
bf0f6f24a sched: cfs core, ... |
82 |
* SCHED_OTHER wake-up granularity. |
172e082a9 sched: Re-tune th... |
83 |
* (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds) |
bf0f6f24a sched: cfs core, ... |
84 85 86 87 88 |
* * This option delays the preemption effects of decoupled workloads * and reduces their over-scheduling. Synchronous workloads will still * have immediate wakeup/sleep latencies. */ |
172e082a9 sched: Re-tune th... |
89 |
unsigned int sysctl_sched_wakeup_granularity = 1000000UL; |
0bcdcf28c sched: Fix missin... |
90 |
unsigned int normalized_sysctl_sched_wakeup_granularity = 1000000UL; |
bf0f6f24a sched: cfs core, ... |
91 |
|
da84d9617 sched: reintroduc... |
92 |
const_debug unsigned int sysctl_sched_migration_cost = 500000UL; |
a7a4f8a75 sched: Add sysctl... |
93 94 95 96 97 98 |
/* * The exponential sliding window over which load is averaged for shares * distribution. * (default: 10msec) */ unsigned int __read_mostly sysctl_sched_shares_window = 10000000UL; |
ec12cb7f3 sched: Accumulate... |
99 100 101 102 103 104 105 106 107 108 109 110 111 |
#ifdef CONFIG_CFS_BANDWIDTH /* * Amount of runtime to allocate from global (tg) to local (per-cfs_rq) pool * each time a cfs_rq requests quota. * * Note: in the case that the slice exceeds the runtime remaining (either due * to consumption or the quota being specified to be smaller than the slice) * we will always only issue the remaining available time. * * default: 5 msec, units: microseconds */ unsigned int sysctl_sched_cfs_bandwidth_slice = 5000UL; #endif |
8527632dc sched: Move updat... |
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 |
static inline void update_load_add(struct load_weight *lw, unsigned long inc) { lw->weight += inc; lw->inv_weight = 0; } static inline void update_load_sub(struct load_weight *lw, unsigned long dec) { lw->weight -= dec; lw->inv_weight = 0; } static inline void update_load_set(struct load_weight *lw, unsigned long w) { lw->weight = w; lw->inv_weight = 0; } |
029632fbb sched: Make separ... |
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 |
/* * Increase the granularity value when there are more CPUs, * because with more CPUs the 'effective latency' as visible * to users decreases. But the relationship is not linear, * so pick a second-best guess by going with the log2 of the * number of CPUs. * * This idea comes from the SD scheduler of Con Kolivas: */ static int get_update_sysctl_factor(void) { unsigned int cpus = min_t(int, num_online_cpus(), 8); unsigned int factor; switch (sysctl_sched_tunable_scaling) { case SCHED_TUNABLESCALING_NONE: factor = 1; break; case SCHED_TUNABLESCALING_LINEAR: factor = cpus; break; case SCHED_TUNABLESCALING_LOG: default: factor = 1 + ilog2(cpus); break; } return factor; } static void update_sysctl(void) { unsigned int factor = get_update_sysctl_factor(); #define SET_SYSCTL(name) \ (sysctl_##name = (factor) * normalized_sysctl_##name) SET_SYSCTL(sched_min_granularity); SET_SYSCTL(sched_latency); SET_SYSCTL(sched_wakeup_granularity); #undef SET_SYSCTL } void sched_init_granularity(void) { update_sysctl(); } |
9dbdb1555 sched/fair: Rewor... |
175 |
#define WMULT_CONST (~0U) |
029632fbb sched: Make separ... |
176 |
#define WMULT_SHIFT 32 |
9dbdb1555 sched/fair: Rewor... |
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 |
static void __update_inv_weight(struct load_weight *lw) { unsigned long w; if (likely(lw->inv_weight)) return; w = scale_load_down(lw->weight); if (BITS_PER_LONG > 32 && unlikely(w >= WMULT_CONST)) lw->inv_weight = 1; else if (unlikely(!w)) lw->inv_weight = WMULT_CONST; else lw->inv_weight = WMULT_CONST / w; } |
029632fbb sched: Make separ... |
193 194 |
/* |
9dbdb1555 sched/fair: Rewor... |
195 196 197 198 199 200 201 202 203 204 |
* delta_exec * weight / lw.weight * OR * (delta_exec * (weight * lw->inv_weight)) >> WMULT_SHIFT * * Either weight := NICE_0_LOAD and lw \e prio_to_wmult[], in which case * we're guaranteed shift stays positive because inv_weight is guaranteed to * fit 32 bits, and NICE_0_LOAD gives another 10 bits; therefore shift >= 22. * * Or, weight =< lw.weight (because lw.weight is the runqueue weight), thus * weight/lw.weight <= 1, and therefore our shift will also be positive. |
029632fbb sched: Make separ... |
205 |
*/ |
9dbdb1555 sched/fair: Rewor... |
206 |
static u64 __calc_delta(u64 delta_exec, unsigned long weight, struct load_weight *lw) |
029632fbb sched: Make separ... |
207 |
{ |
9dbdb1555 sched/fair: Rewor... |
208 209 |
u64 fact = scale_load_down(weight); int shift = WMULT_SHIFT; |
029632fbb sched: Make separ... |
210 |
|
9dbdb1555 sched/fair: Rewor... |
211 |
__update_inv_weight(lw); |
029632fbb sched: Make separ... |
212 |
|
9dbdb1555 sched/fair: Rewor... |
213 214 215 216 217 |
if (unlikely(fact >> 32)) { while (fact >> 32) { fact >>= 1; shift--; } |
029632fbb sched: Make separ... |
218 |
} |
9dbdb1555 sched/fair: Rewor... |
219 220 |
/* hint to use a 32x32->64 mul */ fact = (u64)(u32)fact * lw->inv_weight; |
029632fbb sched: Make separ... |
221 |
|
9dbdb1555 sched/fair: Rewor... |
222 223 224 225 |
while (fact >> 32) { fact >>= 1; shift--; } |
029632fbb sched: Make separ... |
226 |
|
9dbdb1555 sched/fair: Rewor... |
227 |
return mul_u64_u32_shr(delta_exec, fact, shift); |
029632fbb sched: Make separ... |
228 229 230 231 |
} const struct sched_class fair_sched_class; |
a4c2f00f5 sched: fair sched... |
232 |
|
bf0f6f24a sched: cfs core, ... |
233 234 235 |
/************************************************************** * CFS operations on generic schedulable entities: */ |
62160e3f4 sched: track cfs_... |
236 |
#ifdef CONFIG_FAIR_GROUP_SCHED |
bf0f6f24a sched: cfs core, ... |
237 |
|
62160e3f4 sched: track cfs_... |
238 |
/* cpu runqueue to which this cfs_rq is attached */ |
bf0f6f24a sched: cfs core, ... |
239 240 |
static inline struct rq *rq_of(struct cfs_rq *cfs_rq) { |
62160e3f4 sched: track cfs_... |
241 |
return cfs_rq->rq; |
bf0f6f24a sched: cfs core, ... |
242 |
} |
62160e3f4 sched: track cfs_... |
243 244 |
/* An entity is a task if it doesn't "own" a runqueue */ #define entity_is_task(se) (!se->my_q) |
bf0f6f24a sched: cfs core, ... |
245 |
|
8f48894fc sched: Add debug ... |
246 247 248 249 250 251 252 |
static inline struct task_struct *task_of(struct sched_entity *se) { #ifdef CONFIG_SCHED_DEBUG WARN_ON_ONCE(!entity_is_task(se)); #endif return container_of(se, struct task_struct, se); } |
b758149c0 sched: prepatory ... |
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 |
/* Walk up scheduling entities hierarchy */ #define for_each_sched_entity(se) \ for (; se; se = se->parent) static inline struct cfs_rq *task_cfs_rq(struct task_struct *p) { return p->se.cfs_rq; } /* runqueue on which this entity is (to be) queued */ static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se) { return se->cfs_rq; } /* runqueue "owned" by this group */ static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp) { return grp->my_q; } |
aff3e4988 sched: Account fo... |
273 274 |
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update); |
9ee474f55 sched: Maintain t... |
275 |
|
3d4b47b4b sched: Implement ... |
276 277 278 |
static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq) { if (!cfs_rq->on_list) { |
67e86250f sched: Introduce ... |
279 280 281 282 283 284 285 286 287 288 289 290 |
/* * Ensure we either appear before our parent (if already * enqueued) or force our parent to appear after us when it is * enqueued. The fact that we always enqueue bottom-up * reduces this to two cases. */ if (cfs_rq->tg->parent && cfs_rq->tg->parent->cfs_rq[cpu_of(rq_of(cfs_rq))]->on_list) { list_add_rcu(&cfs_rq->leaf_cfs_rq_list, &rq_of(cfs_rq)->leaf_cfs_rq_list); } else { list_add_tail_rcu(&cfs_rq->leaf_cfs_rq_list, |
3d4b47b4b sched: Implement ... |
291 |
&rq_of(cfs_rq)->leaf_cfs_rq_list); |
67e86250f sched: Introduce ... |
292 |
} |
3d4b47b4b sched: Implement ... |
293 294 |
cfs_rq->on_list = 1; |
9ee474f55 sched: Maintain t... |
295 |
/* We should have no load, but we need to update last_decay. */ |
aff3e4988 sched: Account fo... |
296 |
update_cfs_rq_blocked_load(cfs_rq, 0); |
3d4b47b4b sched: Implement ... |
297 298 299 300 301 302 303 304 305 306 |
} } static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq) { if (cfs_rq->on_list) { list_del_rcu(&cfs_rq->leaf_cfs_rq_list); cfs_rq->on_list = 0; } } |
b758149c0 sched: prepatory ... |
307 308 309 310 311 |
/* Iterate thr' all leaf cfs_rq's on a runqueue */ #define for_each_leaf_cfs_rq(rq, cfs_rq) \ list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list) /* Do the two (enqueued) entities belong to the same group ? */ |
fed14d45f sched/fair: Track... |
312 |
static inline struct cfs_rq * |
b758149c0 sched: prepatory ... |
313 314 315 |
is_same_group(struct sched_entity *se, struct sched_entity *pse) { if (se->cfs_rq == pse->cfs_rq) |
fed14d45f sched/fair: Track... |
316 |
return se->cfs_rq; |
b758149c0 sched: prepatory ... |
317 |
|
fed14d45f sched/fair: Track... |
318 |
return NULL; |
b758149c0 sched: prepatory ... |
319 320 321 322 323 324 |
} static inline struct sched_entity *parent_entity(struct sched_entity *se) { return se->parent; } |
464b75273 sched: re-instate... |
325 326 327 328 329 330 331 332 333 334 335 336 337 |
static void find_matching_se(struct sched_entity **se, struct sched_entity **pse) { int se_depth, pse_depth; /* * preemption test can be made between sibling entities who are in the * same cfs_rq i.e who have a common parent. Walk up the hierarchy of * both tasks until we find their ancestors who are siblings of common * parent. */ /* First walk up until both entities are at same depth */ |
fed14d45f sched/fair: Track... |
338 339 |
se_depth = (*se)->depth; pse_depth = (*pse)->depth; |
464b75273 sched: re-instate... |
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 |
while (se_depth > pse_depth) { se_depth--; *se = parent_entity(*se); } while (pse_depth > se_depth) { pse_depth--; *pse = parent_entity(*pse); } while (!is_same_group(*se, *pse)) { *se = parent_entity(*se); *pse = parent_entity(*pse); } } |
8f48894fc sched: Add debug ... |
356 357 358 359 360 361 |
#else /* !CONFIG_FAIR_GROUP_SCHED */ static inline struct task_struct *task_of(struct sched_entity *se) { return container_of(se, struct task_struct, se); } |
bf0f6f24a sched: cfs core, ... |
362 |
|
62160e3f4 sched: track cfs_... |
363 364 365 |
static inline struct rq *rq_of(struct cfs_rq *cfs_rq) { return container_of(cfs_rq, struct rq, cfs); |
bf0f6f24a sched: cfs core, ... |
366 367 368 |
} #define entity_is_task(se) 1 |
b758149c0 sched: prepatory ... |
369 370 |
#define for_each_sched_entity(se) \ for (; se; se = NULL) |
bf0f6f24a sched: cfs core, ... |
371 |
|
b758149c0 sched: prepatory ... |
372 |
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p) |
bf0f6f24a sched: cfs core, ... |
373 |
{ |
b758149c0 sched: prepatory ... |
374 |
return &task_rq(p)->cfs; |
bf0f6f24a sched: cfs core, ... |
375 |
} |
b758149c0 sched: prepatory ... |
376 377 378 379 380 381 382 383 384 385 386 387 388 |
static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se) { struct task_struct *p = task_of(se); struct rq *rq = task_rq(p); return &rq->cfs; } /* runqueue "owned" by this group */ static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp) { return NULL; } |
3d4b47b4b sched: Implement ... |
389 390 391 392 393 394 395 |
static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq) { } static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq) { } |
b758149c0 sched: prepatory ... |
396 397 |
#define for_each_leaf_cfs_rq(rq, cfs_rq) \ for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL) |
b758149c0 sched: prepatory ... |
398 399 400 401 |
static inline struct sched_entity *parent_entity(struct sched_entity *se) { return NULL; } |
464b75273 sched: re-instate... |
402 403 404 405 |
static inline void find_matching_se(struct sched_entity **se, struct sched_entity **pse) { } |
b758149c0 sched: prepatory ... |
406 |
#endif /* CONFIG_FAIR_GROUP_SCHED */ |
6c16a6dcb sched: Fix compil... |
407 |
static __always_inline |
9dbdb1555 sched/fair: Rewor... |
408 |
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec); |
bf0f6f24a sched: cfs core, ... |
409 410 411 412 |
/************************************************************** * Scheduling class tree data structure manipulation methods: */ |
1bf08230f sched: Fix variab... |
413 |
static inline u64 max_vruntime(u64 max_vruntime, u64 vruntime) |
02e0431a3 sched: better min... |
414 |
{ |
1bf08230f sched: Fix variab... |
415 |
s64 delta = (s64)(vruntime - max_vruntime); |
368059a97 sched: max_vrunti... |
416 |
if (delta > 0) |
1bf08230f sched: Fix variab... |
417 |
max_vruntime = vruntime; |
02e0431a3 sched: better min... |
418 |
|
1bf08230f sched: Fix variab... |
419 |
return max_vruntime; |
02e0431a3 sched: better min... |
420 |
} |
0702e3ebc sched: cleanup: f... |
421 |
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime) |
b0ffd246e sched: clean up m... |
422 423 424 425 426 427 428 |
{ s64 delta = (s64)(vruntime - min_vruntime); if (delta < 0) min_vruntime = vruntime; return min_vruntime; } |
54fdc5816 sched: Account fo... |
429 430 431 432 433 |
static inline int entity_before(struct sched_entity *a, struct sched_entity *b) { return (s64)(a->vruntime - b->vruntime) < 0; } |
1af5f730f sched: more accur... |
434 435 436 437 438 439 440 441 442 443 444 |
static void update_min_vruntime(struct cfs_rq *cfs_rq) { u64 vruntime = cfs_rq->min_vruntime; if (cfs_rq->curr) vruntime = cfs_rq->curr->vruntime; if (cfs_rq->rb_leftmost) { struct sched_entity *se = rb_entry(cfs_rq->rb_leftmost, struct sched_entity, run_node); |
e17036dac sched: fix update... |
445 |
if (!cfs_rq->curr) |
1af5f730f sched: more accur... |
446 447 448 449 |
vruntime = se->vruntime; else vruntime = min_vruntime(vruntime, se->vruntime); } |
1bf08230f sched: Fix variab... |
450 |
/* ensure we never gain time by being placed backwards. */ |
1af5f730f sched: more accur... |
451 |
cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime); |
3fe1698b7 sched: Deal with ... |
452 453 454 455 |
#ifndef CONFIG_64BIT smp_wmb(); cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime; #endif |
1af5f730f sched: more accur... |
456 |
} |
bf0f6f24a sched: cfs core, ... |
457 458 459 |
/* * Enqueue an entity into the rb-tree: */ |
0702e3ebc sched: cleanup: f... |
460 |
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
461 462 463 464 |
{ struct rb_node **link = &cfs_rq->tasks_timeline.rb_node; struct rb_node *parent = NULL; struct sched_entity *entry; |
bf0f6f24a sched: cfs core, ... |
465 466 467 468 469 470 471 472 473 474 475 476 |
int leftmost = 1; /* * Find the right place in the rbtree: */ while (*link) { parent = *link; entry = rb_entry(parent, struct sched_entity, run_node); /* * We dont care about collisions. Nodes with * the same key stay together. */ |
2bd2d6f2d sched: Replace us... |
477 |
if (entity_before(se, entry)) { |
bf0f6f24a sched: cfs core, ... |
478 479 480 481 482 483 484 485 486 487 488 |
link = &parent->rb_left; } else { link = &parent->rb_right; leftmost = 0; } } /* * Maintain a cache of leftmost tree entries (it is frequently * used): */ |
1af5f730f sched: more accur... |
489 |
if (leftmost) |
57cb499df sched: remove set... |
490 |
cfs_rq->rb_leftmost = &se->run_node; |
bf0f6f24a sched: cfs core, ... |
491 492 493 |
rb_link_node(&se->run_node, parent, link); rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline); |
bf0f6f24a sched: cfs core, ... |
494 |
} |
0702e3ebc sched: cleanup: f... |
495 |
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
496 |
{ |
3fe69747d sched: min_vrunti... |
497 498 |
if (cfs_rq->rb_leftmost == &se->run_node) { struct rb_node *next_node; |
3fe69747d sched: min_vrunti... |
499 500 501 |
next_node = rb_next(&se->run_node); cfs_rq->rb_leftmost = next_node; |
3fe69747d sched: min_vrunti... |
502 |
} |
e9acbff64 sched: introduce ... |
503 |
|
bf0f6f24a sched: cfs core, ... |
504 |
rb_erase(&se->run_node, &cfs_rq->tasks_timeline); |
bf0f6f24a sched: cfs core, ... |
505 |
} |
029632fbb sched: Make separ... |
506 |
struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq) |
bf0f6f24a sched: cfs core, ... |
507 |
{ |
f4b6755fb sched: cleanup fa... |
508 509 510 511 512 513 |
struct rb_node *left = cfs_rq->rb_leftmost; if (!left) return NULL; return rb_entry(left, struct sched_entity, run_node); |
bf0f6f24a sched: cfs core, ... |
514 |
} |
ac53db596 sched: Use a budd... |
515 516 517 518 519 520 521 522 523 524 525 |
static struct sched_entity *__pick_next_entity(struct sched_entity *se) { struct rb_node *next = rb_next(&se->run_node); if (!next) return NULL; return rb_entry(next, struct sched_entity, run_node); } #ifdef CONFIG_SCHED_DEBUG |
029632fbb sched: Make separ... |
526 |
struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq) |
aeb73b040 sched: clean up n... |
527 |
{ |
7eee3e677 sched: clean up _... |
528 |
struct rb_node *last = rb_last(&cfs_rq->tasks_timeline); |
aeb73b040 sched: clean up n... |
529 |
|
70eee74b7 sched: remove dup... |
530 531 |
if (!last) return NULL; |
7eee3e677 sched: clean up _... |
532 533 |
return rb_entry(last, struct sched_entity, run_node); |
aeb73b040 sched: clean up n... |
534 |
} |
bf0f6f24a sched: cfs core, ... |
535 536 537 |
/************************************************************** * Scheduling class statistics methods: */ |
acb4a848d sched: Update nor... |
538 |
int sched_proc_update_handler(struct ctl_table *table, int write, |
8d65af789 sysctl: remove "s... |
539 |
void __user *buffer, size_t *lenp, |
b2be5e96d sched: reintroduc... |
540 541 |
loff_t *ppos) { |
8d65af789 sysctl: remove "s... |
542 |
int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); |
acb4a848d sched: Update nor... |
543 |
int factor = get_update_sysctl_factor(); |
b2be5e96d sched: reintroduc... |
544 545 546 547 548 549 |
if (ret || !write) return ret; sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency, sysctl_sched_min_granularity); |
acb4a848d sched: Update nor... |
550 551 552 553 554 |
#define WRT_SYSCTL(name) \ (normalized_sysctl_##name = sysctl_##name / (factor)) WRT_SYSCTL(sched_min_granularity); WRT_SYSCTL(sched_latency); WRT_SYSCTL(sched_wakeup_granularity); |
acb4a848d sched: Update nor... |
555 |
#undef WRT_SYSCTL |
b2be5e96d sched: reintroduc... |
556 557 558 |
return 0; } #endif |
647e7cac2 sched: vslice fix... |
559 560 |
/* |
f9c0b0950 sched: revert bac... |
561 |
* delta /= w |
a7be37ac8 sched: revert the... |
562 |
*/ |
9dbdb1555 sched/fair: Rewor... |
563 |
static inline u64 calc_delta_fair(u64 delta, struct sched_entity *se) |
a7be37ac8 sched: revert the... |
564 |
{ |
f9c0b0950 sched: revert bac... |
565 |
if (unlikely(se->load.weight != NICE_0_LOAD)) |
9dbdb1555 sched/fair: Rewor... |
566 |
delta = __calc_delta(delta, NICE_0_LOAD, &se->load); |
a7be37ac8 sched: revert the... |
567 568 569 570 571 |
return delta; } /* |
647e7cac2 sched: vslice fix... |
572 573 |
* The idea is to set a period in which each task runs once. * |
532b1858c sched: Fix __sche... |
574 |
* When there are too many tasks (sched_nr_latency) we have to stretch |
647e7cac2 sched: vslice fix... |
575 576 577 578 |
* this period because otherwise the slices get too small. * * p = (nr <= nl) ? l : l*nr/nl */ |
4d78e7b65 sched: new task p... |
579 580 581 |
static u64 __sched_period(unsigned long nr_running) { u64 period = sysctl_sched_latency; |
b2be5e96d sched: reintroduc... |
582 |
unsigned long nr_latency = sched_nr_latency; |
4d78e7b65 sched: new task p... |
583 584 |
if (unlikely(nr_running > nr_latency)) { |
4bf0b7715 sched: remove do_... |
585 |
period = sysctl_sched_min_granularity; |
4d78e7b65 sched: new task p... |
586 |
period *= nr_running; |
4d78e7b65 sched: new task p... |
587 588 589 590 |
} return period; } |
647e7cac2 sched: vslice fix... |
591 592 593 594 |
/* * We calculate the wall-time slice from the period by taking a part * proportional to the weight. * |
f9c0b0950 sched: revert bac... |
595 |
* s = p*P[w/rw] |
647e7cac2 sched: vslice fix... |
596 |
*/ |
6d0f0ebd0 sched: simplify a... |
597 |
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se) |
218050855 sched: adaptive s... |
598 |
{ |
0a582440f sched: fix sched_... |
599 |
u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq); |
f9c0b0950 sched: revert bac... |
600 |
|
0a582440f sched: fix sched_... |
601 |
for_each_sched_entity(se) { |
6272d68cc sched: sched_slic... |
602 |
struct load_weight *load; |
3104bf03a sched: Fix out of... |
603 |
struct load_weight lw; |
6272d68cc sched: sched_slic... |
604 605 606 |
cfs_rq = cfs_rq_of(se); load = &cfs_rq->load; |
f9c0b0950 sched: revert bac... |
607 |
|
0a582440f sched: fix sched_... |
608 |
if (unlikely(!se->on_rq)) { |
3104bf03a sched: Fix out of... |
609 |
lw = cfs_rq->load; |
0a582440f sched: fix sched_... |
610 611 612 613 |
update_load_add(&lw, se->load.weight); load = &lw; } |
9dbdb1555 sched/fair: Rewor... |
614 |
slice = __calc_delta(slice, se->load.weight, load); |
0a582440f sched: fix sched_... |
615 616 |
} return slice; |
bf0f6f24a sched: cfs core, ... |
617 |
} |
647e7cac2 sched: vslice fix... |
618 |
/* |
660cc00f8 sched: Spelling fix |
619 |
* We calculate the vruntime slice of a to-be-inserted task. |
647e7cac2 sched: vslice fix... |
620 |
* |
f9c0b0950 sched: revert bac... |
621 |
* vs = s/w |
647e7cac2 sched: vslice fix... |
622 |
*/ |
f9c0b0950 sched: revert bac... |
623 |
static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se) |
67e9fb2a3 sched: add vslice |
624 |
{ |
f9c0b0950 sched: revert bac... |
625 |
return calc_delta_fair(sched_slice(cfs_rq, se), se); |
a7be37ac8 sched: revert the... |
626 |
} |
a75cdaa91 sched: Set an ini... |
627 |
#ifdef CONFIG_SMP |
fb13c7ee0 sched/numa: Use a... |
628 |
static unsigned long task_h_load(struct task_struct *p); |
a75cdaa91 sched: Set an ini... |
629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 |
static inline void __update_task_entity_contrib(struct sched_entity *se); /* Give new task start runnable values to heavy its load in infant time */ void init_task_runnable_average(struct task_struct *p) { u32 slice; p->se.avg.decay_count = 0; slice = sched_slice(task_cfs_rq(p), &p->se) >> 10; p->se.avg.runnable_avg_sum = slice; p->se.avg.runnable_avg_period = slice; __update_task_entity_contrib(&p->se); } #else void init_task_runnable_average(struct task_struct *p) { } #endif |
a7be37ac8 sched: revert the... |
647 |
/* |
9dbdb1555 sched/fair: Rewor... |
648 |
* Update the current task's runtime statistics. |
bf0f6f24a sched: cfs core, ... |
649 |
*/ |
b7cc08965 sched: remove the... |
650 |
static void update_curr(struct cfs_rq *cfs_rq) |
bf0f6f24a sched: cfs core, ... |
651 |
{ |
429d43bcc sched: cleanup: s... |
652 |
struct sched_entity *curr = cfs_rq->curr; |
78becc270 sched: Use an acc... |
653 |
u64 now = rq_clock_task(rq_of(cfs_rq)); |
9dbdb1555 sched/fair: Rewor... |
654 |
u64 delta_exec; |
bf0f6f24a sched: cfs core, ... |
655 656 657 |
if (unlikely(!curr)) return; |
9dbdb1555 sched/fair: Rewor... |
658 659 |
delta_exec = now - curr->exec_start; if (unlikely((s64)delta_exec <= 0)) |
34f28ecd0 sched: optimize u... |
660 |
return; |
bf0f6f24a sched: cfs core, ... |
661 |
|
8ebc91d93 sched: remove sta... |
662 |
curr->exec_start = now; |
d842de871 sched: cpu accoun... |
663 |
|
9dbdb1555 sched/fair: Rewor... |
664 665 666 667 668 669 670 671 |
schedstat_set(curr->statistics.exec_max, max(delta_exec, curr->statistics.exec_max)); curr->sum_exec_runtime += delta_exec; schedstat_add(cfs_rq, exec_clock, delta_exec); curr->vruntime += calc_delta_fair(delta_exec, curr); update_min_vruntime(cfs_rq); |
d842de871 sched: cpu accoun... |
672 673 |
if (entity_is_task(curr)) { struct task_struct *curtask = task_of(curr); |
f977bb493 perf_counter, sch... |
674 |
trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime); |
d842de871 sched: cpu accoun... |
675 |
cpuacct_charge(curtask, delta_exec); |
f06febc96 timers: fix itime... |
676 |
account_group_exec_runtime(curtask, delta_exec); |
d842de871 sched: cpu accoun... |
677 |
} |
ec12cb7f3 sched: Accumulate... |
678 679 |
account_cfs_rq_runtime(cfs_rq, delta_exec); |
bf0f6f24a sched: cfs core, ... |
680 681 682 |
} static inline void |
5870db5b8 sched: remove the... |
683 |
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
684 |
{ |
78becc270 sched: Use an acc... |
685 |
schedstat_set(se->statistics.wait_start, rq_clock(rq_of(cfs_rq))); |
bf0f6f24a sched: cfs core, ... |
686 |
} |
bf0f6f24a sched: cfs core, ... |
687 688 689 |
/* * Task is being enqueued - update stats: */ |
d2417e5a3 sched: remove the... |
690 |
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
691 |
{ |
bf0f6f24a sched: cfs core, ... |
692 693 694 695 |
/* * Are we enqueueing a waiting task? (for current tasks * a dequeue/enqueue event is a NOP) */ |
429d43bcc sched: cleanup: s... |
696 |
if (se != cfs_rq->curr) |
5870db5b8 sched: remove the... |
697 |
update_stats_wait_start(cfs_rq, se); |
bf0f6f24a sched: cfs core, ... |
698 |
} |
bf0f6f24a sched: cfs core, ... |
699 |
static void |
9ef0a9615 sched: remove the... |
700 |
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
701 |
{ |
41acab885 sched: Implement ... |
702 |
schedstat_set(se->statistics.wait_max, max(se->statistics.wait_max, |
78becc270 sched: Use an acc... |
703 |
rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start)); |
41acab885 sched: Implement ... |
704 705 |
schedstat_set(se->statistics.wait_count, se->statistics.wait_count + 1); schedstat_set(se->statistics.wait_sum, se->statistics.wait_sum + |
78becc270 sched: Use an acc... |
706 |
rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start); |
768d0c272 sched: Add wait, ... |
707 708 709 |
#ifdef CONFIG_SCHEDSTATS if (entity_is_task(se)) { trace_sched_stat_wait(task_of(se), |
78becc270 sched: Use an acc... |
710 |
rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start); |
768d0c272 sched: Add wait, ... |
711 712 |
} #endif |
41acab885 sched: Implement ... |
713 |
schedstat_set(se->statistics.wait_start, 0); |
bf0f6f24a sched: cfs core, ... |
714 715 716 |
} static inline void |
19b6a2e37 sched: remove the... |
717 |
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
718 |
{ |
bf0f6f24a sched: cfs core, ... |
719 720 721 722 |
/* * Mark the end of the wait period if dequeueing a * waiting task: */ |
429d43bcc sched: cleanup: s... |
723 |
if (se != cfs_rq->curr) |
9ef0a9615 sched: remove the... |
724 |
update_stats_wait_end(cfs_rq, se); |
bf0f6f24a sched: cfs core, ... |
725 726 727 728 729 730 |
} /* * We are picking a new current task - update its stats: */ static inline void |
79303e9e0 sched: remove the... |
731 |
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
732 733 734 735 |
{ /* * We are starting a new run period: */ |
78becc270 sched: Use an acc... |
736 |
se->exec_start = rq_clock_task(rq_of(cfs_rq)); |
bf0f6f24a sched: cfs core, ... |
737 |
} |
bf0f6f24a sched: cfs core, ... |
738 739 740 |
/************************************************** * Scheduling class queueing methods: */ |
cbee9f88e mm: numa: Add fau... |
741 742 |
#ifdef CONFIG_NUMA_BALANCING /* |
598f0ec0b sched/numa: Set t... |
743 744 745 |
* Approximate time to scan a full NUMA task in ms. The task scan period is * calculated based on the tasks virtual memory size and * numa_balancing_scan_size. |
cbee9f88e mm: numa: Add fau... |
746 |
*/ |
598f0ec0b sched/numa: Set t... |
747 748 |
unsigned int sysctl_numa_balancing_scan_period_min = 1000; unsigned int sysctl_numa_balancing_scan_period_max = 60000; |
6e5fb223e mm: sched: numa: ... |
749 750 751 |
/* Portion of address space to scan in MB */ unsigned int sysctl_numa_balancing_scan_size = 256; |
cbee9f88e mm: numa: Add fau... |
752 |
|
4b96a29ba mm: sched: numa: ... |
753 754 |
/* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */ unsigned int sysctl_numa_balancing_scan_delay = 1000; |
598f0ec0b sched/numa: Set t... |
755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 |
static unsigned int task_nr_scan_windows(struct task_struct *p) { unsigned long rss = 0; unsigned long nr_scan_pages; /* * Calculations based on RSS as non-present and empty pages are skipped * by the PTE scanner and NUMA hinting faults should be trapped based * on resident pages */ nr_scan_pages = sysctl_numa_balancing_scan_size << (20 - PAGE_SHIFT); rss = get_mm_rss(p->mm); if (!rss) rss = nr_scan_pages; rss = round_up(rss, nr_scan_pages); return rss / nr_scan_pages; } /* For sanitys sake, never scan more PTEs than MAX_SCAN_WINDOW MB/sec. */ #define MAX_SCAN_WINDOW 2560 static unsigned int task_scan_min(struct task_struct *p) { unsigned int scan, floor; unsigned int windows = 1; if (sysctl_numa_balancing_scan_size < MAX_SCAN_WINDOW) windows = MAX_SCAN_WINDOW / sysctl_numa_balancing_scan_size; floor = 1000 / windows; scan = sysctl_numa_balancing_scan_period_min / task_nr_scan_windows(p); return max_t(unsigned int, floor, scan); } static unsigned int task_scan_max(struct task_struct *p) { unsigned int smin = task_scan_min(p); unsigned int smax; /* Watch for min being lower than max due to floor calculations */ smax = sysctl_numa_balancing_scan_period_max / task_nr_scan_windows(p); return max(smin, smax); } |
0ec8aa00f sched/numa: Avoid... |
799 800 801 802 803 804 805 806 807 808 809 |
static void account_numa_enqueue(struct rq *rq, struct task_struct *p) { rq->nr_numa_running += (p->numa_preferred_nid != -1); rq->nr_preferred_running += (p->numa_preferred_nid == task_node(p)); } static void account_numa_dequeue(struct rq *rq, struct task_struct *p) { rq->nr_numa_running -= (p->numa_preferred_nid != -1); rq->nr_preferred_running -= (p->numa_preferred_nid == task_node(p)); } |
8c8a743c5 sched/numa: Use {... |
810 811 812 813 814 |
struct numa_group { atomic_t refcount; spinlock_t lock; /* nr_tasks, tasks */ int nr_tasks; |
e29cf08b0 sched/numa: Repor... |
815 |
pid_t gid; |
8c8a743c5 sched/numa: Use {... |
816 817 818 |
struct list_head task_list; struct rcu_head rcu; |
20e07dea2 sched/numa: Build... |
819 |
nodemask_t active_nodes; |
989348b5f sched/numa: Use u... |
820 |
unsigned long total_faults; |
7e2703e60 sched/numa: Norma... |
821 822 823 824 825 |
/* * Faults_cpu is used to decide whether memory should move * towards the CPU. As a consequence, these stats are weighted * more by CPU use than by memory faults. */ |
50ec8a401 sched/numa: Track... |
826 |
unsigned long *faults_cpu; |
989348b5f sched/numa: Use u... |
827 |
unsigned long faults[0]; |
8c8a743c5 sched/numa: Use {... |
828 |
}; |
be1e4e760 sched/numa: Turn ... |
829 830 831 832 833 834 835 836 |
/* Shared or private faults. */ #define NR_NUMA_HINT_FAULT_TYPES 2 /* Memory and CPU locality */ #define NR_NUMA_HINT_FAULT_STATS (NR_NUMA_HINT_FAULT_TYPES * 2) /* Averaged statistics, and temporary buffers. */ #define NR_NUMA_HINT_FAULT_BUCKETS (NR_NUMA_HINT_FAULT_STATS * 2) |
e29cf08b0 sched/numa: Repor... |
837 838 839 840 |
pid_t task_numa_group_id(struct task_struct *p) { return p->numa_group ? p->numa_group->gid : 0; } |
ac8e895bd sched/numa: Add i... |
841 842 |
static inline int task_faults_idx(int nid, int priv) { |
be1e4e760 sched/numa: Turn ... |
843 |
return NR_NUMA_HINT_FAULT_TYPES * nid + priv; |
ac8e895bd sched/numa: Add i... |
844 845 846 847 |
} static inline unsigned long task_faults(struct task_struct *p, int nid) { |
ff1df896a sched/numa: Renam... |
848 |
if (!p->numa_faults_memory) |
ac8e895bd sched/numa: Add i... |
849 |
return 0; |
ff1df896a sched/numa: Renam... |
850 851 |
return p->numa_faults_memory[task_faults_idx(nid, 0)] + p->numa_faults_memory[task_faults_idx(nid, 1)]; |
ac8e895bd sched/numa: Add i... |
852 |
} |
83e1d2cd9 sched/numa: Use g... |
853 854 855 856 |
static inline unsigned long group_faults(struct task_struct *p, int nid) { if (!p->numa_group) return 0; |
82897b4fd sched/numa: Use w... |
857 858 |
return p->numa_group->faults[task_faults_idx(nid, 0)] + p->numa_group->faults[task_faults_idx(nid, 1)]; |
83e1d2cd9 sched/numa: Use g... |
859 |
} |
20e07dea2 sched/numa: Build... |
860 861 862 863 864 |
static inline unsigned long group_faults_cpu(struct numa_group *group, int nid) { return group->faults_cpu[task_faults_idx(nid, 0)] + group->faults_cpu[task_faults_idx(nid, 1)]; } |
83e1d2cd9 sched/numa: Use g... |
865 866 867 868 869 870 871 872 873 |
/* * These return the fraction of accesses done by a particular task, or * task group, on a particular numa node. The group weight is given a * larger multiplier, in order to group tasks together that are almost * evenly spread out between numa nodes. */ static inline unsigned long task_weight(struct task_struct *p, int nid) { unsigned long total_faults; |
ff1df896a sched/numa: Renam... |
874 |
if (!p->numa_faults_memory) |
83e1d2cd9 sched/numa: Use g... |
875 876 877 878 879 880 881 882 883 884 885 886 |
return 0; total_faults = p->total_numa_faults; if (!total_faults) return 0; return 1000 * task_faults(p, nid) / total_faults; } static inline unsigned long group_weight(struct task_struct *p, int nid) { |
989348b5f sched/numa: Use u... |
887 |
if (!p->numa_group || !p->numa_group->total_faults) |
83e1d2cd9 sched/numa: Use g... |
888 |
return 0; |
989348b5f sched/numa: Use u... |
889 |
return 1000 * group_faults(p, nid) / p->numa_group->total_faults; |
83e1d2cd9 sched/numa: Use g... |
890 |
} |
10f390427 sched/numa, mm: U... |
891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 |
bool should_numa_migrate_memory(struct task_struct *p, struct page * page, int src_nid, int dst_cpu) { struct numa_group *ng = p->numa_group; int dst_nid = cpu_to_node(dst_cpu); int last_cpupid, this_cpupid; this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid); /* * Multi-stage node selection is used in conjunction with a periodic * migration fault to build a temporal task<->page relation. By using * a two-stage filter we remove short/unlikely relations. * * Using P(p) ~ n_p / n_t as per frequentist probability, we can equate * a task's usage of a particular page (n_p) per total usage of this * page (n_t) (in a given time-span) to a probability. * * Our periodic faults will sample this probability and getting the * same result twice in a row, given these samples are fully * independent, is then given by P(n)^2, provided our sample period * is sufficiently short compared to the usage pattern. * * This quadric squishes small probabilities, making it less likely we * act on an unlikely task<->page relation. */ last_cpupid = page_cpupid_xchg_last(page, this_cpupid); if (!cpupid_pid_unset(last_cpupid) && cpupid_to_nid(last_cpupid) != dst_nid) return false; /* Always allow migrate on private faults */ if (cpupid_match_pid(p, last_cpupid)) return true; /* A shared fault, but p->numa_group has not been set up yet. */ if (!ng) return true; /* * Do not migrate if the destination is not a node that * is actively used by this numa group. */ if (!node_isset(dst_nid, ng->active_nodes)) return false; /* * Source is a node that is not actively used by this * numa group, while the destination is. Migrate. */ if (!node_isset(src_nid, ng->active_nodes)) return true; /* * Both source and destination are nodes in active * use by this numa group. Maximize memory bandwidth * by migrating from more heavily used groups, to less * heavily used ones, spreading the load around. * Use a 1/4 hysteresis to avoid spurious page movement. */ return group_faults(p, dst_nid) < (group_faults(p, src_nid) * 3 / 4); } |
e6628d5b0 sched/numa: Resch... |
953 |
static unsigned long weighted_cpuload(const int cpu); |
58d081b50 sched/numa: Avoid... |
954 955 956 957 |
static unsigned long source_load(int cpu, int type); static unsigned long target_load(int cpu, int type); static unsigned long power_of(int cpu); static long effective_load(struct task_group *tg, int cpu, long wl, long wg); |
fb13c7ee0 sched/numa: Use a... |
958 |
/* Cached statistics for all CPUs within a node */ |
58d081b50 sched/numa: Avoid... |
959 |
struct numa_stats { |
fb13c7ee0 sched/numa: Use a... |
960 |
unsigned long nr_running; |
58d081b50 sched/numa: Avoid... |
961 |
unsigned long load; |
fb13c7ee0 sched/numa: Use a... |
962 963 964 965 966 967 968 |
/* Total compute capacity of CPUs on a node */ unsigned long power; /* Approximate capacity in terms of runnable tasks on a node */ unsigned long capacity; int has_capacity; |
58d081b50 sched/numa: Avoid... |
969 |
}; |
e6628d5b0 sched/numa: Resch... |
970 |
|
fb13c7ee0 sched/numa: Use a... |
971 972 973 974 975 |
/* * XXX borrowed from update_sg_lb_stats */ static void update_numa_stats(struct numa_stats *ns, int nid) { |
5eca82a9a sched/numa: Cure ... |
976 |
int cpu, cpus = 0; |
fb13c7ee0 sched/numa: Use a... |
977 978 979 980 981 982 983 984 |
memset(ns, 0, sizeof(*ns)); for_each_cpu(cpu, cpumask_of_node(nid)) { struct rq *rq = cpu_rq(cpu); ns->nr_running += rq->nr_running; ns->load += weighted_cpuload(cpu); ns->power += power_of(cpu); |
5eca82a9a sched/numa: Cure ... |
985 986 |
cpus++; |
fb13c7ee0 sched/numa: Use a... |
987 |
} |
5eca82a9a sched/numa: Cure ... |
988 989 990 991 992 993 994 995 996 997 |
/* * If we raced with hotplug and there are no CPUs left in our mask * the @ns structure is NULL'ed and task_numa_compare() will * not find this node attractive. * * We'll either bail at !has_capacity, or we'll detect a huge imbalance * and bail there. */ if (!cpus) return; |
fb13c7ee0 sched/numa: Use a... |
998 999 1000 1001 |
ns->load = (ns->load * SCHED_POWER_SCALE) / ns->power; ns->capacity = DIV_ROUND_CLOSEST(ns->power, SCHED_POWER_SCALE); ns->has_capacity = (ns->nr_running < ns->capacity); } |
58d081b50 sched/numa: Avoid... |
1002 1003 |
struct task_numa_env { struct task_struct *p; |
e6628d5b0 sched/numa: Resch... |
1004 |
|
58d081b50 sched/numa: Avoid... |
1005 1006 |
int src_cpu, src_nid; int dst_cpu, dst_nid; |
e6628d5b0 sched/numa: Resch... |
1007 |
|
58d081b50 sched/numa: Avoid... |
1008 |
struct numa_stats src_stats, dst_stats; |
e6628d5b0 sched/numa: Resch... |
1009 |
|
40ea2b42d sched/numa: Drop ... |
1010 |
int imbalance_pct; |
fb13c7ee0 sched/numa: Use a... |
1011 1012 1013 |
struct task_struct *best_task; long best_imp; |
58d081b50 sched/numa: Avoid... |
1014 1015 |
int best_cpu; }; |
fb13c7ee0 sched/numa: Use a... |
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 |
static void task_numa_assign(struct task_numa_env *env, struct task_struct *p, long imp) { if (env->best_task) put_task_struct(env->best_task); if (p) get_task_struct(p); env->best_task = p; env->best_imp = imp; env->best_cpu = env->dst_cpu; } /* * This checks if the overall compute and NUMA accesses of the system would * be improved if the source tasks was migrated to the target dst_cpu taking * into account that it might be best if task running on the dst_cpu should * be exchanged with the source task */ |
887c290e8 sched/numa: Decid... |
1035 1036 |
static void task_numa_compare(struct task_numa_env *env, long taskimp, long groupimp) |
fb13c7ee0 sched/numa: Use a... |
1037 1038 1039 1040 1041 1042 |
{ struct rq *src_rq = cpu_rq(env->src_cpu); struct rq *dst_rq = cpu_rq(env->dst_cpu); struct task_struct *cur; long dst_load, src_load; long load; |
887c290e8 sched/numa: Decid... |
1043 |
long imp = (groupimp > 0) ? groupimp : taskimp; |
fb13c7ee0 sched/numa: Use a... |
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 |
rcu_read_lock(); cur = ACCESS_ONCE(dst_rq->curr); if (cur->pid == 0) /* idle */ cur = NULL; /* * "imp" is the fault differential for the source task between the * source and destination node. Calculate the total differential for * the source task and potential destination task. The more negative * the value is, the more rmeote accesses that would be expected to * be incurred if the tasks were swapped. */ if (cur) { /* Skip this swap candidate if cannot move to the source cpu */ if (!cpumask_test_cpu(env->src_cpu, tsk_cpus_allowed(cur))) goto unlock; |
887c290e8 sched/numa: Decid... |
1061 1062 |
/* * If dst and source tasks are in the same NUMA group, or not |
ca28aa53d sched/numa: Fix t... |
1063 |
* in any group then look only at task weights. |
887c290e8 sched/numa: Decid... |
1064 |
*/ |
ca28aa53d sched/numa: Fix t... |
1065 |
if (cur->numa_group == env->p->numa_group) { |
887c290e8 sched/numa: Decid... |
1066 1067 |
imp = taskimp + task_weight(cur, env->src_nid) - task_weight(cur, env->dst_nid); |
ca28aa53d sched/numa: Fix t... |
1068 1069 1070 1071 1072 1073 |
/* * Add some hysteresis to prevent swapping the * tasks within a group over tiny differences. */ if (cur->numa_group) imp -= imp/16; |
887c290e8 sched/numa: Decid... |
1074 |
} else { |
ca28aa53d sched/numa: Fix t... |
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 |
/* * Compare the group weights. If a task is all by * itself (not part of a group), use the task weight * instead. */ if (env->p->numa_group) imp = groupimp; else imp = taskimp; if (cur->numa_group) imp += group_weight(cur, env->src_nid) - group_weight(cur, env->dst_nid); else imp += task_weight(cur, env->src_nid) - task_weight(cur, env->dst_nid); |
887c290e8 sched/numa: Decid... |
1091 |
} |
fb13c7ee0 sched/numa: Use a... |
1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 |
} if (imp < env->best_imp) goto unlock; if (!cur) { /* Is there capacity at our destination? */ if (env->src_stats.has_capacity && !env->dst_stats.has_capacity) goto unlock; goto balance; } /* Balance doesn't matter much if we're running a task per cpu */ if (src_rq->nr_running == 1 && dst_rq->nr_running == 1) goto assign; /* * In the overloaded case, try and keep the load balanced. */ balance: dst_load = env->dst_stats.load; src_load = env->src_stats.load; /* XXX missing power terms */ load = task_h_load(env->p); dst_load += load; src_load -= load; if (cur) { load = task_h_load(cur); dst_load -= load; src_load += load; } /* make src_load the smaller */ if (dst_load < src_load) swap(dst_load, src_load); if (src_load * env->imbalance_pct < dst_load * 100) goto unlock; assign: task_numa_assign(env, cur, imp); unlock: rcu_read_unlock(); } |
887c290e8 sched/numa: Decid... |
1140 1141 |
static void task_numa_find_cpu(struct task_numa_env *env, long taskimp, long groupimp) |
2c8a50aa8 sched/numa: Favor... |
1142 1143 1144 1145 1146 1147 1148 1149 1150 |
{ int cpu; for_each_cpu(cpu, cpumask_of_node(env->dst_nid)) { /* Skip this CPU if the source task cannot migrate */ if (!cpumask_test_cpu(cpu, tsk_cpus_allowed(env->p))) continue; env->dst_cpu = cpu; |
887c290e8 sched/numa: Decid... |
1151 |
task_numa_compare(env, taskimp, groupimp); |
2c8a50aa8 sched/numa: Favor... |
1152 1153 |
} } |
58d081b50 sched/numa: Avoid... |
1154 1155 |
static int task_numa_migrate(struct task_struct *p) { |
58d081b50 sched/numa: Avoid... |
1156 1157 |
struct task_numa_env env = { .p = p, |
fb13c7ee0 sched/numa: Use a... |
1158 |
|
58d081b50 sched/numa: Avoid... |
1159 |
.src_cpu = task_cpu(p), |
b32e86b43 sched/numa: Add d... |
1160 |
.src_nid = task_node(p), |
fb13c7ee0 sched/numa: Use a... |
1161 1162 1163 1164 1165 1166 |
.imbalance_pct = 112, .best_task = NULL, .best_imp = 0, .best_cpu = -1 |
58d081b50 sched/numa: Avoid... |
1167 1168 |
}; struct sched_domain *sd; |
887c290e8 sched/numa: Decid... |
1169 |
unsigned long taskweight, groupweight; |
2c8a50aa8 sched/numa: Favor... |
1170 |
int nid, ret; |
887c290e8 sched/numa: Decid... |
1171 |
long taskimp, groupimp; |
e6628d5b0 sched/numa: Resch... |
1172 |
|
58d081b50 sched/numa: Avoid... |
1173 |
/* |
fb13c7ee0 sched/numa: Use a... |
1174 1175 1176 1177 1178 1179 |
* Pick the lowest SD_NUMA domain, as that would have the smallest * imbalance and would be the first to start moving tasks about. * * And we want to avoid any moving of tasks about, as that would create * random movement of tasks -- counter the numa conditions we're trying * to satisfy here. |
58d081b50 sched/numa: Avoid... |
1180 1181 |
*/ rcu_read_lock(); |
fb13c7ee0 sched/numa: Use a... |
1182 |
sd = rcu_dereference(per_cpu(sd_numa, env.src_cpu)); |
46a73e8a1 sched/numa: Fix N... |
1183 1184 |
if (sd) env.imbalance_pct = 100 + (sd->imbalance_pct - 100) / 2; |
e6628d5b0 sched/numa: Resch... |
1185 |
rcu_read_unlock(); |
46a73e8a1 sched/numa: Fix N... |
1186 1187 1188 1189 1190 1191 1192 |
/* * Cpusets can break the scheduler domain tree into smaller * balance domains, some of which do not cross NUMA boundaries. * Tasks that are "trapped" in such domains cannot be migrated * elsewhere, so there is no point in (re)trying. */ if (unlikely(!sd)) { |
de1b301a1 sched/numa: Use w... |
1193 |
p->numa_preferred_nid = task_node(p); |
46a73e8a1 sched/numa: Fix N... |
1194 1195 |
return -EINVAL; } |
887c290e8 sched/numa: Decid... |
1196 1197 |
taskweight = task_weight(p, env.src_nid); groupweight = group_weight(p, env.src_nid); |
fb13c7ee0 sched/numa: Use a... |
1198 |
update_numa_stats(&env.src_stats, env.src_nid); |
2c8a50aa8 sched/numa: Favor... |
1199 |
env.dst_nid = p->numa_preferred_nid; |
887c290e8 sched/numa: Decid... |
1200 1201 |
taskimp = task_weight(p, env.dst_nid) - taskweight; groupimp = group_weight(p, env.dst_nid) - groupweight; |
2c8a50aa8 sched/numa: Favor... |
1202 |
update_numa_stats(&env.dst_stats, env.dst_nid); |
58d081b50 sched/numa: Avoid... |
1203 |
|
e1dda8a79 sched/numa: Fix p... |
1204 1205 |
/* If the preferred nid has capacity, try to use it. */ if (env.dst_stats.has_capacity) |
887c290e8 sched/numa: Decid... |
1206 |
task_numa_find_cpu(&env, taskimp, groupimp); |
e1dda8a79 sched/numa: Fix p... |
1207 1208 1209 |
/* No space available on the preferred nid. Look elsewhere. */ if (env.best_cpu == -1) { |
2c8a50aa8 sched/numa: Favor... |
1210 1211 1212 |
for_each_online_node(nid) { if (nid == env.src_nid || nid == p->numa_preferred_nid) continue; |
58d081b50 sched/numa: Avoid... |
1213 |
|
83e1d2cd9 sched/numa: Use g... |
1214 |
/* Only consider nodes where both task and groups benefit */ |
887c290e8 sched/numa: Decid... |
1215 1216 1217 |
taskimp = task_weight(p, nid) - taskweight; groupimp = group_weight(p, nid) - groupweight; if (taskimp < 0 && groupimp < 0) |
fb13c7ee0 sched/numa: Use a... |
1218 |
continue; |
2c8a50aa8 sched/numa: Favor... |
1219 1220 |
env.dst_nid = nid; update_numa_stats(&env.dst_stats, env.dst_nid); |
887c290e8 sched/numa: Decid... |
1221 |
task_numa_find_cpu(&env, taskimp, groupimp); |
58d081b50 sched/numa: Avoid... |
1222 1223 |
} } |
fb13c7ee0 sched/numa: Use a... |
1224 1225 1226 |
/* No better CPU than the current one was found. */ if (env.best_cpu == -1) return -EAGAIN; |
0ec8aa00f sched/numa: Avoid... |
1227 |
sched_setnuma(p, env.dst_nid); |
04bb2f947 sched/numa: Adjus... |
1228 1229 1230 1231 1232 |
/* * Reset the scan period if the task is being rescheduled on an * alternative node to recheck if the tasks is now properly placed. */ p->numa_scan_period = task_scan_min(p); |
fb13c7ee0 sched/numa: Use a... |
1233 |
if (env.best_task == NULL) { |
286549dca sched: add tracep... |
1234 1235 1236 |
ret = migrate_task_to(p, env.best_cpu); if (ret != 0) trace_sched_stick_numa(p, env.src_cpu, env.best_cpu); |
fb13c7ee0 sched/numa: Use a... |
1237 1238 1239 1240 |
return ret; } ret = migrate_swap(p, env.best_task); |
286549dca sched: add tracep... |
1241 1242 |
if (ret != 0) trace_sched_stick_numa(p, env.src_cpu, task_cpu(env.best_task)); |
fb13c7ee0 sched/numa: Use a... |
1243 1244 |
put_task_struct(env.best_task); return ret; |
e6628d5b0 sched/numa: Resch... |
1245 |
} |
6b9a7460b sched/numa: Retry... |
1246 1247 1248 |
/* Attempt to migrate a task to a CPU on the preferred node. */ static void numa_migrate_preferred(struct task_struct *p) { |
2739d3eef sched/numa: Retry... |
1249 |
/* This task has no NUMA fault statistics yet */ |
ff1df896a sched/numa: Renam... |
1250 |
if (unlikely(p->numa_preferred_nid == -1 || !p->numa_faults_memory)) |
6b9a7460b sched/numa: Retry... |
1251 |
return; |
2739d3eef sched/numa: Retry... |
1252 1253 1254 1255 |
/* Periodically retry migrating the task to the preferred node */ p->numa_migrate_retry = jiffies + HZ; /* Success if task is already running on preferred CPU */ |
de1b301a1 sched/numa: Use w... |
1256 |
if (task_node(p) == p->numa_preferred_nid) |
6b9a7460b sched/numa: Retry... |
1257 1258 1259 |
return; /* Otherwise, try migrate to a CPU on the preferred node */ |
2739d3eef sched/numa: Retry... |
1260 |
task_numa_migrate(p); |
6b9a7460b sched/numa: Retry... |
1261 |
} |
04bb2f947 sched/numa: Adjus... |
1262 |
/* |
20e07dea2 sched/numa: Build... |
1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 |
* Find the nodes on which the workload is actively running. We do this by * tracking the nodes from which NUMA hinting faults are triggered. This can * be different from the set of nodes where the workload's memory is currently * located. * * The bitmask is used to make smarter decisions on when to do NUMA page * migrations, To prevent flip-flopping, and excessive page migrations, nodes * are added when they cause over 6/16 of the maximum number of faults, but * only removed when they drop below 3/16. */ static void update_numa_active_node_mask(struct numa_group *numa_group) { unsigned long faults, max_faults = 0; int nid; for_each_online_node(nid) { faults = group_faults_cpu(numa_group, nid); if (faults > max_faults) max_faults = faults; } for_each_online_node(nid) { faults = group_faults_cpu(numa_group, nid); if (!node_isset(nid, numa_group->active_nodes)) { if (faults > max_faults * 6 / 16) node_set(nid, numa_group->active_nodes); } else if (faults < max_faults * 3 / 16) node_clear(nid, numa_group->active_nodes); } } /* |
04bb2f947 sched/numa: Adjus... |
1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 |
* When adapting the scan rate, the period is divided into NUMA_PERIOD_SLOTS * increments. The more local the fault statistics are, the higher the scan * period will be for the next scan window. If local/remote ratio is below * NUMA_PERIOD_THRESHOLD (where range of ratio is 1..NUMA_PERIOD_SLOTS) the * scan period will decrease */ #define NUMA_PERIOD_SLOTS 10 #define NUMA_PERIOD_THRESHOLD 3 /* * Increase the scan period (slow down scanning) if the majority of * our memory is already on our local node, or if the majority of * the page accesses are shared with other processes. * Otherwise, decrease the scan period. */ static void update_task_scan_period(struct task_struct *p, unsigned long shared, unsigned long private) { unsigned int period_slot; int ratio; int diff; unsigned long remote = p->numa_faults_locality[0]; unsigned long local = p->numa_faults_locality[1]; /* * If there were no record hinting faults then either the task is * completely idle or all activity is areas that are not of interest * to automatic numa balancing. Scan slower */ if (local + shared == 0) { p->numa_scan_period = min(p->numa_scan_period_max, p->numa_scan_period << 1); p->mm->numa_next_scan = jiffies + msecs_to_jiffies(p->numa_scan_period); return; } /* * Prepare to scale scan period relative to the current period. * == NUMA_PERIOD_THRESHOLD scan period stays the same * < NUMA_PERIOD_THRESHOLD scan period decreases (scan faster) * >= NUMA_PERIOD_THRESHOLD scan period increases (scan slower) */ period_slot = DIV_ROUND_UP(p->numa_scan_period, NUMA_PERIOD_SLOTS); ratio = (local * NUMA_PERIOD_SLOTS) / (local + remote); if (ratio >= NUMA_PERIOD_THRESHOLD) { int slot = ratio - NUMA_PERIOD_THRESHOLD; if (!slot) slot = 1; diff = slot * period_slot; } else { diff = -(NUMA_PERIOD_THRESHOLD - ratio) * period_slot; /* * Scale scan rate increases based on sharing. There is an * inverse relationship between the degree of sharing and * the adjustment made to the scanning period. Broadly * speaking the intent is that there is little point * scanning faster if shared accesses dominate as it may * simply bounce migrations uselessly */ |
04bb2f947 sched/numa: Adjus... |
1359 1360 1361 1362 1363 1364 1365 1366 |
ratio = DIV_ROUND_UP(private * NUMA_PERIOD_SLOTS, (private + shared)); diff = (diff * ratio) / NUMA_PERIOD_SLOTS; } p->numa_scan_period = clamp(p->numa_scan_period + diff, task_scan_min(p), task_scan_max(p)); memset(p->numa_faults_locality, 0, sizeof(p->numa_faults_locality)); } |
7e2703e60 sched/numa: Norma... |
1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 |
/* * Get the fraction of time the task has been running since the last * NUMA placement cycle. The scheduler keeps similar statistics, but * decays those on a 32ms period, which is orders of magnitude off * from the dozens-of-seconds NUMA balancing period. Use the scheduler * stats only if the task is so new there are no NUMA statistics yet. */ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period) { u64 runtime, delta, now; /* Use the start of this time slice to avoid calculations. */ now = p->se.exec_start; runtime = p->se.sum_exec_runtime; if (p->last_task_numa_placement) { delta = runtime - p->last_sum_exec_runtime; *period = now - p->last_task_numa_placement; } else { delta = p->se.avg.runnable_avg_sum; *period = p->se.avg.runnable_avg_period; } p->last_sum_exec_runtime = runtime; p->last_task_numa_placement = now; return delta; } |
cbee9f88e mm: numa: Add fau... |
1394 1395 |
static void task_numa_placement(struct task_struct *p) { |
83e1d2cd9 sched/numa: Use g... |
1396 1397 |
int seq, nid, max_nid = -1, max_group_nid = -1; unsigned long max_faults = 0, max_group_faults = 0; |
04bb2f947 sched/numa: Adjus... |
1398 |
unsigned long fault_types[2] = { 0, 0 }; |
7e2703e60 sched/numa: Norma... |
1399 1400 |
unsigned long total_faults; u64 runtime, period; |
7dbd13ed0 sched/numa: Preve... |
1401 |
spinlock_t *group_lock = NULL; |
cbee9f88e mm: numa: Add fau... |
1402 |
|
2832bc19f sched: numa: ksm:... |
1403 |
seq = ACCESS_ONCE(p->mm->numa_scan_seq); |
cbee9f88e mm: numa: Add fau... |
1404 1405 1406 |
if (p->numa_scan_seq == seq) return; p->numa_scan_seq = seq; |
598f0ec0b sched/numa: Set t... |
1407 |
p->numa_scan_period_max = task_scan_max(p); |
cbee9f88e mm: numa: Add fau... |
1408 |
|
7e2703e60 sched/numa: Norma... |
1409 1410 1411 |
total_faults = p->numa_faults_locality[0] + p->numa_faults_locality[1]; runtime = numa_get_avg_runtime(p, &period); |
7dbd13ed0 sched/numa: Preve... |
1412 1413 1414 |
/* If the task is part of a group prevent parallel updates to group stats */ if (p->numa_group) { group_lock = &p->numa_group->lock; |
60e69eed8 sched/numa: Fix t... |
1415 |
spin_lock_irq(group_lock); |
7dbd13ed0 sched/numa: Preve... |
1416 |
} |
688b7585d sched/numa: Selec... |
1417 1418 |
/* Find the node with the highest number of faults */ for_each_online_node(nid) { |
83e1d2cd9 sched/numa: Use g... |
1419 |
unsigned long faults = 0, group_faults = 0; |
ac8e895bd sched/numa: Add i... |
1420 |
int priv, i; |
745d61476 sched/numa: Updat... |
1421 |
|
be1e4e760 sched/numa: Turn ... |
1422 |
for (priv = 0; priv < NR_NUMA_HINT_FAULT_TYPES; priv++) { |
7e2703e60 sched/numa: Norma... |
1423 |
long diff, f_diff, f_weight; |
8c8a743c5 sched/numa: Use {... |
1424 |
|
ac8e895bd sched/numa: Add i... |
1425 |
i = task_faults_idx(nid, priv); |
745d61476 sched/numa: Updat... |
1426 |
|
ac8e895bd sched/numa: Add i... |
1427 |
/* Decay existing window, copy faults since last scan */ |
35664fd41 sched/numa: Do st... |
1428 |
diff = p->numa_faults_buffer_memory[i] - p->numa_faults_memory[i] / 2; |
ff1df896a sched/numa: Renam... |
1429 1430 |
fault_types[priv] += p->numa_faults_buffer_memory[i]; p->numa_faults_buffer_memory[i] = 0; |
fb13c7ee0 sched/numa: Use a... |
1431 |
|
7e2703e60 sched/numa: Norma... |
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 |
/* * Normalize the faults_from, so all tasks in a group * count according to CPU use, instead of by the raw * number of faults. Tasks with little runtime have * little over-all impact on throughput, and thus their * faults are less important. */ f_weight = div64_u64(runtime << 16, period + 1); f_weight = (f_weight * p->numa_faults_buffer_cpu[i]) / (total_faults + 1); |
35664fd41 sched/numa: Do st... |
1442 |
f_diff = f_weight - p->numa_faults_cpu[i] / 2; |
50ec8a401 sched/numa: Track... |
1443 |
p->numa_faults_buffer_cpu[i] = 0; |
35664fd41 sched/numa: Do st... |
1444 1445 |
p->numa_faults_memory[i] += diff; p->numa_faults_cpu[i] += f_diff; |
ff1df896a sched/numa: Renam... |
1446 |
faults += p->numa_faults_memory[i]; |
83e1d2cd9 sched/numa: Use g... |
1447 |
p->total_numa_faults += diff; |
8c8a743c5 sched/numa: Use {... |
1448 1449 |
if (p->numa_group) { /* safe because we can only change our own group */ |
989348b5f sched/numa: Use u... |
1450 |
p->numa_group->faults[i] += diff; |
50ec8a401 sched/numa: Track... |
1451 |
p->numa_group->faults_cpu[i] += f_diff; |
989348b5f sched/numa: Use u... |
1452 1453 |
p->numa_group->total_faults += diff; group_faults += p->numa_group->faults[i]; |
8c8a743c5 sched/numa: Use {... |
1454 |
} |
ac8e895bd sched/numa: Add i... |
1455 |
} |
688b7585d sched/numa: Selec... |
1456 1457 1458 1459 |
if (faults > max_faults) { max_faults = faults; max_nid = nid; } |
83e1d2cd9 sched/numa: Use g... |
1460 1461 1462 1463 1464 1465 |
if (group_faults > max_group_faults) { max_group_faults = group_faults; max_group_nid = nid; } } |
04bb2f947 sched/numa: Adjus... |
1466 |
update_task_scan_period(p, fault_types[0], fault_types[1]); |
7dbd13ed0 sched/numa: Preve... |
1467 |
if (p->numa_group) { |
20e07dea2 sched/numa: Build... |
1468 |
update_numa_active_node_mask(p->numa_group); |
7dbd13ed0 sched/numa: Preve... |
1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 |
/* * If the preferred task and group nids are different, * iterate over the nodes again to find the best place. */ if (max_nid != max_group_nid) { unsigned long weight, max_weight = 0; for_each_online_node(nid) { weight = task_weight(p, nid) + group_weight(p, nid); if (weight > max_weight) { max_weight = weight; max_nid = nid; } |
83e1d2cd9 sched/numa: Use g... |
1482 1483 |
} } |
7dbd13ed0 sched/numa: Preve... |
1484 |
|
60e69eed8 sched/numa: Fix t... |
1485 |
spin_unlock_irq(group_lock); |
688b7585d sched/numa: Selec... |
1486 |
} |
6b9a7460b sched/numa: Retry... |
1487 |
/* Preferred node as the node with the most faults */ |
3a7053b32 sched/numa: Favou... |
1488 |
if (max_faults && max_nid != p->numa_preferred_nid) { |
e6628d5b0 sched/numa: Resch... |
1489 |
/* Update the preferred nid and migrate task if possible */ |
0ec8aa00f sched/numa: Avoid... |
1490 |
sched_setnuma(p, max_nid); |
6b9a7460b sched/numa: Retry... |
1491 |
numa_migrate_preferred(p); |
3a7053b32 sched/numa: Favou... |
1492 |
} |
cbee9f88e mm: numa: Add fau... |
1493 |
} |
8c8a743c5 sched/numa: Use {... |
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 |
static inline int get_numa_group(struct numa_group *grp) { return atomic_inc_not_zero(&grp->refcount); } static inline void put_numa_group(struct numa_group *grp) { if (atomic_dec_and_test(&grp->refcount)) kfree_rcu(grp, rcu); } |
3e6a9418c sched/numa: Take ... |
1504 1505 |
static void task_numa_group(struct task_struct *p, int cpupid, int flags, int *priv) |
8c8a743c5 sched/numa: Use {... |
1506 1507 1508 1509 1510 1511 1512 1513 1514 |
{ struct numa_group *grp, *my_grp; struct task_struct *tsk; bool join = false; int cpu = cpupid_to_cpu(cpupid); int i; if (unlikely(!p->numa_group)) { unsigned int size = sizeof(struct numa_group) + |
50ec8a401 sched/numa: Track... |
1515 |
4*nr_node_ids*sizeof(unsigned long); |
8c8a743c5 sched/numa: Use {... |
1516 1517 1518 1519 1520 1521 1522 1523 |
grp = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); if (!grp) return; atomic_set(&grp->refcount, 1); spin_lock_init(&grp->lock); INIT_LIST_HEAD(&grp->task_list); |
e29cf08b0 sched/numa: Repor... |
1524 |
grp->gid = p->pid; |
50ec8a401 sched/numa: Track... |
1525 |
/* Second half of the array tracks nids where faults happen */ |
be1e4e760 sched/numa: Turn ... |
1526 1527 |
grp->faults_cpu = grp->faults + NR_NUMA_HINT_FAULT_TYPES * nr_node_ids; |
8c8a743c5 sched/numa: Use {... |
1528 |
|
20e07dea2 sched/numa: Build... |
1529 |
node_set(task_node(current), grp->active_nodes); |
be1e4e760 sched/numa: Turn ... |
1530 |
for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) |
ff1df896a sched/numa: Renam... |
1531 |
grp->faults[i] = p->numa_faults_memory[i]; |
8c8a743c5 sched/numa: Use {... |
1532 |
|
989348b5f sched/numa: Use u... |
1533 |
grp->total_faults = p->total_numa_faults; |
83e1d2cd9 sched/numa: Use g... |
1534 |
|
8c8a743c5 sched/numa: Use {... |
1535 1536 1537 1538 1539 1540 1541 1542 1543 |
list_add(&p->numa_entry, &grp->task_list); grp->nr_tasks++; rcu_assign_pointer(p->numa_group, grp); } rcu_read_lock(); tsk = ACCESS_ONCE(cpu_rq(cpu)->curr); if (!cpupid_match_pid(tsk, cpupid)) |
3354781a2 sched/numa: Reflo... |
1544 |
goto no_join; |
8c8a743c5 sched/numa: Use {... |
1545 1546 1547 |
grp = rcu_dereference(tsk->numa_group); if (!grp) |
3354781a2 sched/numa: Reflo... |
1548 |
goto no_join; |
8c8a743c5 sched/numa: Use {... |
1549 1550 1551 |
my_grp = p->numa_group; if (grp == my_grp) |
3354781a2 sched/numa: Reflo... |
1552 |
goto no_join; |
8c8a743c5 sched/numa: Use {... |
1553 1554 1555 1556 1557 1558 |
/* * Only join the other group if its bigger; if we're the bigger group, * the other task will join us. */ if (my_grp->nr_tasks > grp->nr_tasks) |
3354781a2 sched/numa: Reflo... |
1559 |
goto no_join; |
8c8a743c5 sched/numa: Use {... |
1560 1561 1562 1563 1564 |
/* * Tie-break on the grp address. */ if (my_grp->nr_tasks == grp->nr_tasks && my_grp > grp) |
3354781a2 sched/numa: Reflo... |
1565 |
goto no_join; |
8c8a743c5 sched/numa: Use {... |
1566 |
|
dabe1d992 sched/numa: Be mo... |
1567 1568 1569 1570 1571 1572 1573 |
/* Always join threads in the same process. */ if (tsk->mm == current->mm) join = true; /* Simple filter to avoid false positives due to PID collisions */ if (flags & TNF_SHARED) join = true; |
8c8a743c5 sched/numa: Use {... |
1574 |
|
3e6a9418c sched/numa: Take ... |
1575 1576 |
/* Update priv based on whether false sharing was detected */ *priv = !join; |
dabe1d992 sched/numa: Be mo... |
1577 |
if (join && !get_numa_group(grp)) |
3354781a2 sched/numa: Reflo... |
1578 |
goto no_join; |
8c8a743c5 sched/numa: Use {... |
1579 |
|
8c8a743c5 sched/numa: Use {... |
1580 1581 1582 1583 |
rcu_read_unlock(); if (!join) return; |
60e69eed8 sched/numa: Fix t... |
1584 1585 |
BUG_ON(irqs_disabled()); double_lock_irq(&my_grp->lock, &grp->lock); |
989348b5f sched/numa: Use u... |
1586 |
|
be1e4e760 sched/numa: Turn ... |
1587 |
for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) { |
ff1df896a sched/numa: Renam... |
1588 1589 |
my_grp->faults[i] -= p->numa_faults_memory[i]; grp->faults[i] += p->numa_faults_memory[i]; |
8c8a743c5 sched/numa: Use {... |
1590 |
} |
989348b5f sched/numa: Use u... |
1591 1592 |
my_grp->total_faults -= p->total_numa_faults; grp->total_faults += p->total_numa_faults; |
8c8a743c5 sched/numa: Use {... |
1593 1594 1595 1596 1597 1598 |
list_move(&p->numa_entry, &grp->task_list); my_grp->nr_tasks--; grp->nr_tasks++; spin_unlock(&my_grp->lock); |
60e69eed8 sched/numa: Fix t... |
1599 |
spin_unlock_irq(&grp->lock); |
8c8a743c5 sched/numa: Use {... |
1600 1601 1602 1603 |
rcu_assign_pointer(p->numa_group, grp); put_numa_group(my_grp); |
3354781a2 sched/numa: Reflo... |
1604 1605 1606 1607 1608 |
return; no_join: rcu_read_unlock(); return; |
8c8a743c5 sched/numa: Use {... |
1609 1610 1611 1612 1613 |
} void task_numa_free(struct task_struct *p) { struct numa_group *grp = p->numa_group; |
ff1df896a sched/numa: Renam... |
1614 |
void *numa_faults = p->numa_faults_memory; |
e9dd685ce sched/numa: Fix u... |
1615 1616 |
unsigned long flags; int i; |
8c8a743c5 sched/numa: Use {... |
1617 1618 |
if (grp) { |
e9dd685ce sched/numa: Fix u... |
1619 |
spin_lock_irqsave(&grp->lock, flags); |
be1e4e760 sched/numa: Turn ... |
1620 |
for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) |
ff1df896a sched/numa: Renam... |
1621 |
grp->faults[i] -= p->numa_faults_memory[i]; |
989348b5f sched/numa: Use u... |
1622 |
grp->total_faults -= p->total_numa_faults; |
83e1d2cd9 sched/numa: Use g... |
1623 |
|
8c8a743c5 sched/numa: Use {... |
1624 1625 |
list_del(&p->numa_entry); grp->nr_tasks--; |
e9dd685ce sched/numa: Fix u... |
1626 |
spin_unlock_irqrestore(&grp->lock, flags); |
8c8a743c5 sched/numa: Use {... |
1627 1628 1629 |
rcu_assign_pointer(p->numa_group, NULL); put_numa_group(grp); } |
ff1df896a sched/numa: Renam... |
1630 1631 |
p->numa_faults_memory = NULL; p->numa_faults_buffer_memory = NULL; |
50ec8a401 sched/numa: Track... |
1632 1633 |
p->numa_faults_cpu= NULL; p->numa_faults_buffer_cpu = NULL; |
82727018b sched/numa: Call ... |
1634 |
kfree(numa_faults); |
8c8a743c5 sched/numa: Use {... |
1635 |
} |
cbee9f88e mm: numa: Add fau... |
1636 1637 1638 |
/* * Got a PROT_NONE fault for a page on @node. */ |
58b46da33 sched/numa: Renam... |
1639 |
void task_numa_fault(int last_cpupid, int mem_node, int pages, int flags) |
cbee9f88e mm: numa: Add fau... |
1640 1641 |
{ struct task_struct *p = current; |
6688cc054 mm: numa: Do not ... |
1642 |
bool migrated = flags & TNF_MIGRATED; |
58b46da33 sched/numa: Renam... |
1643 |
int cpu_node = task_node(current); |
ac8e895bd sched/numa: Add i... |
1644 |
int priv; |
cbee9f88e mm: numa: Add fau... |
1645 |
|
10e84b97e mm: sched: numa: ... |
1646 |
if (!numabalancing_enabled) |
1a687c2e9 mm: sched: numa: ... |
1647 |
return; |
9ff1d9ff3 sched/numa: Check... |
1648 1649 1650 |
/* for example, ksmd faulting in a user's mm */ if (!p->mm) return; |
82727018b sched/numa: Call ... |
1651 1652 1653 |
/* Do not worry about placement if exiting */ if (p->state == TASK_DEAD) return; |
f809ca9a5 sched/numa: Track... |
1654 |
/* Allocate buffer to track faults on a per-node basis */ |
ff1df896a sched/numa: Renam... |
1655 |
if (unlikely(!p->numa_faults_memory)) { |
be1e4e760 sched/numa: Turn ... |
1656 1657 |
int size = sizeof(*p->numa_faults_memory) * NR_NUMA_HINT_FAULT_BUCKETS * nr_node_ids; |
f809ca9a5 sched/numa: Track... |
1658 |
|
be1e4e760 sched/numa: Turn ... |
1659 |
p->numa_faults_memory = kzalloc(size, GFP_KERNEL|__GFP_NOWARN); |
ff1df896a sched/numa: Renam... |
1660 |
if (!p->numa_faults_memory) |
f809ca9a5 sched/numa: Track... |
1661 |
return; |
745d61476 sched/numa: Updat... |
1662 |
|
ff1df896a sched/numa: Renam... |
1663 |
BUG_ON(p->numa_faults_buffer_memory); |
be1e4e760 sched/numa: Turn ... |
1664 1665 1666 1667 1668 1669 |
/* * The averaged statistics, shared & private, memory & cpu, * occupy the first half of the array. The second half of the * array is for current counters, which are averaged into the * first set by task_numa_placement. */ |
50ec8a401 sched/numa: Track... |
1670 1671 1672 |
p->numa_faults_cpu = p->numa_faults_memory + (2 * nr_node_ids); p->numa_faults_buffer_memory = p->numa_faults_memory + (4 * nr_node_ids); p->numa_faults_buffer_cpu = p->numa_faults_memory + (6 * nr_node_ids); |
83e1d2cd9 sched/numa: Use g... |
1673 |
p->total_numa_faults = 0; |
04bb2f947 sched/numa: Adjus... |
1674 |
memset(p->numa_faults_locality, 0, sizeof(p->numa_faults_locality)); |
f809ca9a5 sched/numa: Track... |
1675 |
} |
cbee9f88e mm: numa: Add fau... |
1676 |
|
fb003b80d sched: numa: Slow... |
1677 |
/* |
8c8a743c5 sched/numa: Use {... |
1678 1679 1680 1681 1682 1683 1684 |
* First accesses are treated as private, otherwise consider accesses * to be private if the accessing pid has not changed */ if (unlikely(last_cpupid == (-1 & LAST_CPUPID_MASK))) { priv = 1; } else { priv = cpupid_match_pid(p, last_cpupid); |
6688cc054 mm: numa: Do not ... |
1685 |
if (!priv && !(flags & TNF_NO_GROUP)) |
3e6a9418c sched/numa: Take ... |
1686 |
task_numa_group(p, last_cpupid, flags, &priv); |
8c8a743c5 sched/numa: Use {... |
1687 |
} |
cbee9f88e mm: numa: Add fau... |
1688 |
task_numa_placement(p); |
f809ca9a5 sched/numa: Track... |
1689 |
|
2739d3eef sched/numa: Retry... |
1690 1691 1692 1693 1694 |
/* * Retry task to preferred node migration periodically, in case it * case it previously failed, or the scheduler moved us. */ if (time_after(jiffies, p->numa_migrate_retry)) |
6b9a7460b sched/numa: Retry... |
1695 |
numa_migrate_preferred(p); |
b32e86b43 sched/numa: Add d... |
1696 1697 |
if (migrated) p->numa_pages_migrated += pages; |
58b46da33 sched/numa: Renam... |
1698 1699 |
p->numa_faults_buffer_memory[task_faults_idx(mem_node, priv)] += pages; p->numa_faults_buffer_cpu[task_faults_idx(cpu_node, priv)] += pages; |
04bb2f947 sched/numa: Adjus... |
1700 |
p->numa_faults_locality[!!(flags & TNF_FAULT_LOCAL)] += pages; |
cbee9f88e mm: numa: Add fau... |
1701 |
} |
6e5fb223e mm: sched: numa: ... |
1702 1703 1704 1705 1706 |
static void reset_ptenuma_scan(struct task_struct *p) { ACCESS_ONCE(p->mm->numa_scan_seq)++; p->mm->numa_scan_offset = 0; } |
cbee9f88e mm: numa: Add fau... |
1707 1708 1709 1710 1711 1712 1713 1714 1715 |
/* * The expensive part of numa migration is done from task_work context. * Triggered from task_tick_numa(). */ void task_numa_work(struct callback_head *work) { unsigned long migrate, next_scan, now = jiffies; struct task_struct *p = current; struct mm_struct *mm = p->mm; |
6e5fb223e mm: sched: numa: ... |
1716 |
struct vm_area_struct *vma; |
9f40604cd sched, numa, mm: ... |
1717 |
unsigned long start, end; |
598f0ec0b sched/numa: Set t... |
1718 |
unsigned long nr_pte_updates = 0; |
9f40604cd sched, numa, mm: ... |
1719 |
long pages; |
cbee9f88e mm: numa: Add fau... |
1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 |
WARN_ON_ONCE(p != container_of(work, struct task_struct, numa_work)); work->next = work; /* protect against double add */ /* * Who cares about NUMA placement when they're dying. * * NOTE: make sure not to dereference p->mm before this check, * exit_task_work() happens _after_ exit_mm() so we could be called * without p->mm even though we still had it when we enqueued this * work. */ if (p->flags & PF_EXITING) return; |
930aa174f sched/numa: Remov... |
1734 |
if (!mm->numa_next_scan) { |
7e8d16b6c sched/numa: Initi... |
1735 1736 |
mm->numa_next_scan = now + msecs_to_jiffies(sysctl_numa_balancing_scan_delay); |
b8593bfda mm: sched: Adapt ... |
1737 1738 1739 |
} /* |
cbee9f88e mm: numa: Add fau... |
1740 1741 1742 1743 1744 |
* Enforce maximal scan/migration frequency.. */ migrate = mm->numa_next_scan; if (time_before(now, migrate)) return; |
598f0ec0b sched/numa: Set t... |
1745 1746 1747 1748 |
if (p->numa_scan_period == 0) { p->numa_scan_period_max = task_scan_max(p); p->numa_scan_period = task_scan_min(p); } |
cbee9f88e mm: numa: Add fau... |
1749 |
|
fb003b80d sched: numa: Slow... |
1750 |
next_scan = now + msecs_to_jiffies(p->numa_scan_period); |
cbee9f88e mm: numa: Add fau... |
1751 1752 |
if (cmpxchg(&mm->numa_next_scan, migrate, next_scan) != migrate) return; |
e14808b49 mm: numa: Rate li... |
1753 |
/* |
19a78d110 sched/numa: Mitig... |
1754 1755 1756 1757 |
* Delay this task enough that another task of this mm will likely win * the next time around. */ p->node_stamp += 2 * TICK_NSEC; |
9f40604cd sched, numa, mm: ... |
1758 1759 1760 1761 1762 |
start = mm->numa_scan_offset; pages = sysctl_numa_balancing_scan_size; pages <<= 20 - PAGE_SHIFT; /* MB in pages */ if (!pages) return; |
cbee9f88e mm: numa: Add fau... |
1763 |
|
6e5fb223e mm: sched: numa: ... |
1764 |
down_read(&mm->mmap_sem); |
9f40604cd sched, numa, mm: ... |
1765 |
vma = find_vma(mm, start); |
6e5fb223e mm: sched: numa: ... |
1766 1767 |
if (!vma) { reset_ptenuma_scan(p); |
9f40604cd sched, numa, mm: ... |
1768 |
start = 0; |
6e5fb223e mm: sched: numa: ... |
1769 1770 |
vma = mm->mmap; } |
9f40604cd sched, numa, mm: ... |
1771 |
for (; vma; vma = vma->vm_next) { |
fc3147245 mm: numa: Limit N... |
1772 |
if (!vma_migratable(vma) || !vma_policy_mof(p, vma)) |
6e5fb223e mm: sched: numa: ... |
1773 |
continue; |
4591ce4f2 sched/numa: Do no... |
1774 1775 1776 1777 1778 1779 1780 1781 1782 |
/* * Shared library pages mapped by multiple processes are not * migrated as it is expected they are cache replicated. Avoid * hinting faults in read-only file-backed mappings or the vdso * as migrating the pages will be of marginal benefit. */ if (!vma->vm_mm || (vma->vm_file && (vma->vm_flags & (VM_READ|VM_WRITE)) == (VM_READ))) continue; |
3c67f4745 sched: numa: skip... |
1783 1784 1785 1786 1787 1788 |
/* * Skip inaccessible VMAs to avoid any confusion between * PROT_NONE and NUMA hinting ptes */ if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))) continue; |
4591ce4f2 sched/numa: Do no... |
1789 |
|
9f40604cd sched, numa, mm: ... |
1790 1791 1792 1793 |
do { start = max(start, vma->vm_start); end = ALIGN(start + (pages << PAGE_SHIFT), HPAGE_SIZE); end = min(end, vma->vm_end); |
598f0ec0b sched/numa: Set t... |
1794 1795 1796 1797 1798 1799 1800 1801 1802 |
nr_pte_updates += change_prot_numa(vma, start, end); /* * Scan sysctl_numa_balancing_scan_size but ensure that * at least one PTE is updated so that unused virtual * address space is quickly skipped. */ if (nr_pte_updates) pages -= (end - start) >> PAGE_SHIFT; |
6e5fb223e mm: sched: numa: ... |
1803 |
|
9f40604cd sched, numa, mm: ... |
1804 1805 1806 |
start = end; if (pages <= 0) goto out; |
3cf1962cd sched,numa: add c... |
1807 1808 |
cond_resched(); |
9f40604cd sched, numa, mm: ... |
1809 |
} while (end != vma->vm_end); |
cbee9f88e mm: numa: Add fau... |
1810 |
} |
6e5fb223e mm: sched: numa: ... |
1811 |
|
9f40604cd sched, numa, mm: ... |
1812 |
out: |
6e5fb223e mm: sched: numa: ... |
1813 |
/* |
c69307d53 sched/numa: Fix c... |
1814 1815 1816 1817 |
* It is possible to reach the end of the VMA list but the last few * VMAs are not guaranteed to the vma_migratable. If they are not, we * would find the !migratable VMA on the next scan but not reset the * scanner to the start so check it now. |
6e5fb223e mm: sched: numa: ... |
1818 1819 |
*/ if (vma) |
9f40604cd sched, numa, mm: ... |
1820 |
mm->numa_scan_offset = start; |
6e5fb223e mm: sched: numa: ... |
1821 1822 1823 |
else reset_ptenuma_scan(p); up_read(&mm->mmap_sem); |
cbee9f88e mm: numa: Add fau... |
1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 |
} /* * Drive the periodic memory faults.. */ void task_tick_numa(struct rq *rq, struct task_struct *curr) { struct callback_head *work = &curr->numa_work; u64 period, now; /* * We don't care about NUMA placement if we don't have memory. */ if (!curr->mm || (curr->flags & PF_EXITING) || work->next != work) return; /* * Using runtime rather than walltime has the dual advantage that * we (mostly) drive the selection from busy threads and that the * task needs to have done some actual work before we bother with * NUMA placement. */ now = curr->se.sum_exec_runtime; period = (u64)curr->numa_scan_period * NSEC_PER_MSEC; if (now - curr->node_stamp > period) { |
4b96a29ba mm: sched: numa: ... |
1850 |
if (!curr->node_stamp) |
598f0ec0b sched/numa: Set t... |
1851 |
curr->numa_scan_period = task_scan_min(curr); |
19a78d110 sched/numa: Mitig... |
1852 |
curr->node_stamp += period; |
cbee9f88e mm: numa: Add fau... |
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 |
if (!time_before(jiffies, curr->mm->numa_next_scan)) { init_task_work(work, task_numa_work); /* TODO: move this into sched_fork() */ task_work_add(curr, work, true); } } } #else static void task_tick_numa(struct rq *rq, struct task_struct *curr) { } |
0ec8aa00f sched/numa: Avoid... |
1864 1865 1866 1867 1868 1869 1870 1871 |
static inline void account_numa_enqueue(struct rq *rq, struct task_struct *p) { } static inline void account_numa_dequeue(struct rq *rq, struct task_struct *p) { } |
cbee9f88e mm: numa: Add fau... |
1872 |
#endif /* CONFIG_NUMA_BALANCING */ |
30cfdcfc5 sched: do not kee... |
1873 1874 1875 1876 |
static void account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) { update_load_add(&cfs_rq->load, se->load.weight); |
c09595f63 sched: revert rev... |
1877 |
if (!parent_entity(se)) |
029632fbb sched: Make separ... |
1878 |
update_load_add(&rq_of(cfs_rq)->load, se->load.weight); |
367456c75 sched: Ditch per ... |
1879 |
#ifdef CONFIG_SMP |
0ec8aa00f sched/numa: Avoid... |
1880 1881 1882 1883 1884 1885 |
if (entity_is_task(se)) { struct rq *rq = rq_of(cfs_rq); account_numa_enqueue(rq, task_of(se)); list_add(&se->group_node, &rq->cfs_tasks); } |
367456c75 sched: Ditch per ... |
1886 |
#endif |
30cfdcfc5 sched: do not kee... |
1887 |
cfs_rq->nr_running++; |
30cfdcfc5 sched: do not kee... |
1888 1889 1890 1891 1892 1893 |
} static void account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) { update_load_sub(&cfs_rq->load, se->load.weight); |
c09595f63 sched: revert rev... |
1894 |
if (!parent_entity(se)) |
029632fbb sched: Make separ... |
1895 |
update_load_sub(&rq_of(cfs_rq)->load, se->load.weight); |
0ec8aa00f sched/numa: Avoid... |
1896 1897 |
if (entity_is_task(se)) { account_numa_dequeue(rq_of(cfs_rq), task_of(se)); |
b87f17242 sched: maintain o... |
1898 |
list_del_init(&se->group_node); |
0ec8aa00f sched/numa: Avoid... |
1899 |
} |
30cfdcfc5 sched: do not kee... |
1900 |
cfs_rq->nr_running--; |
30cfdcfc5 sched: do not kee... |
1901 |
} |
3ff6dcac7 sched: Fix poor i... |
1902 1903 |
#ifdef CONFIG_FAIR_GROUP_SCHED # ifdef CONFIG_SMP |
cf5f0acf3 sched: Add a comm... |
1904 1905 1906 1907 1908 1909 1910 1911 1912 |
static inline long calc_tg_weight(struct task_group *tg, struct cfs_rq *cfs_rq) { long tg_weight; /* * Use this CPU's actual weight instead of the last load_contribution * to gain a more accurate current total weight. See * update_cfs_rq_load_contribution(). */ |
bf5b986ed sched/tg: Use 'un... |
1913 |
tg_weight = atomic_long_read(&tg->load_avg); |
82958366c sched: Replace up... |
1914 |
tg_weight -= cfs_rq->tg_load_contrib; |
cf5f0acf3 sched: Add a comm... |
1915 1916 1917 1918 |
tg_weight += cfs_rq->load.weight; return tg_weight; } |
6d5ab2932 sched: Simplify u... |
1919 |
static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg) |
3ff6dcac7 sched: Fix poor i... |
1920 |
{ |
cf5f0acf3 sched: Add a comm... |
1921 |
long tg_weight, load, shares; |
3ff6dcac7 sched: Fix poor i... |
1922 |
|
cf5f0acf3 sched: Add a comm... |
1923 |
tg_weight = calc_tg_weight(tg, cfs_rq); |
6d5ab2932 sched: Simplify u... |
1924 |
load = cfs_rq->load.weight; |
3ff6dcac7 sched: Fix poor i... |
1925 |
|
3ff6dcac7 sched: Fix poor i... |
1926 |
shares = (tg->shares * load); |
cf5f0acf3 sched: Add a comm... |
1927 1928 |
if (tg_weight) shares /= tg_weight; |
3ff6dcac7 sched: Fix poor i... |
1929 1930 1931 1932 1933 1934 1935 1936 |
if (shares < MIN_SHARES) shares = MIN_SHARES; if (shares > tg->shares) shares = tg->shares; return shares; } |
3ff6dcac7 sched: Fix poor i... |
1937 |
# else /* CONFIG_SMP */ |
6d5ab2932 sched: Simplify u... |
1938 |
static inline long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg) |
3ff6dcac7 sched: Fix poor i... |
1939 1940 1941 |
{ return tg->shares; } |
3ff6dcac7 sched: Fix poor i... |
1942 |
# endif /* CONFIG_SMP */ |
2069dd75c sched: Rewrite tg... |
1943 1944 1945 |
static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, unsigned long weight) { |
19e5eebb8 sched: Fix intera... |
1946 1947 1948 1949 |
if (se->on_rq) { /* commit outstanding execution time */ if (cfs_rq->curr == se) update_curr(cfs_rq); |
2069dd75c sched: Rewrite tg... |
1950 |
account_entity_dequeue(cfs_rq, se); |
19e5eebb8 sched: Fix intera... |
1951 |
} |
2069dd75c sched: Rewrite tg... |
1952 1953 1954 1955 1956 1957 |
update_load_set(&se->load, weight); if (se->on_rq) account_entity_enqueue(cfs_rq, se); } |
82958366c sched: Replace up... |
1958 |
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq); |
6d5ab2932 sched: Simplify u... |
1959 |
static void update_cfs_shares(struct cfs_rq *cfs_rq) |
2069dd75c sched: Rewrite tg... |
1960 1961 1962 |
{ struct task_group *tg; struct sched_entity *se; |
3ff6dcac7 sched: Fix poor i... |
1963 |
long shares; |
2069dd75c sched: Rewrite tg... |
1964 |
|
2069dd75c sched: Rewrite tg... |
1965 1966 |
tg = cfs_rq->tg; se = tg->se[cpu_of(rq_of(cfs_rq))]; |
64660c864 sched: Prevent in... |
1967 |
if (!se || throttled_hierarchy(cfs_rq)) |
2069dd75c sched: Rewrite tg... |
1968 |
return; |
3ff6dcac7 sched: Fix poor i... |
1969 1970 1971 1972 |
#ifndef CONFIG_SMP if (likely(se->load.weight == tg->shares)) return; #endif |
6d5ab2932 sched: Simplify u... |
1973 |
shares = calc_cfs_shares(cfs_rq, tg); |
2069dd75c sched: Rewrite tg... |
1974 1975 1976 1977 |
reweight_entity(cfs_rq_of(se), se, shares); } #else /* CONFIG_FAIR_GROUP_SCHED */ |
6d5ab2932 sched: Simplify u... |
1978 |
static inline void update_cfs_shares(struct cfs_rq *cfs_rq) |
2069dd75c sched: Rewrite tg... |
1979 1980 1981 |
{ } #endif /* CONFIG_FAIR_GROUP_SCHED */ |
141965c74 Revert "sched: In... |
1982 |
#ifdef CONFIG_SMP |
9d85f21c9 sched: Track the ... |
1983 |
/* |
5b51f2f80 sched: Make __upd... |
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 |
* We choose a half-life close to 1 scheduling period. * Note: The tables below are dependent on this value. */ #define LOAD_AVG_PERIOD 32 #define LOAD_AVG_MAX 47742 /* maximum possible load avg */ #define LOAD_AVG_MAX_N 345 /* number of full periods to produce LOAD_MAX_AVG */ /* Precomputed fixed inverse multiplies for multiplication by y^n */ static const u32 runnable_avg_yN_inv[] = { 0xffffffff, 0xfa83b2da, 0xf5257d14, 0xefe4b99a, 0xeac0c6e6, 0xe5b906e6, 0xe0ccdeeb, 0xdbfbb796, 0xd744fcc9, 0xd2a81d91, 0xce248c14, 0xc9b9bd85, 0xc5672a10, 0xc12c4cc9, 0xbd08a39e, 0xb8fbaf46, 0xb504f333, 0xb123f581, 0xad583ee9, 0xa9a15ab4, 0xa5fed6a9, 0xa2704302, 0x9ef5325f, 0x9b8d39b9, 0x9837f050, 0x94f4efa8, 0x91c3d373, 0x8ea4398a, 0x8b95c1e3, 0x88980e80, 0x85aac367, 0x82cd8698, }; /* * Precomputed \Sum y^k { 1<=k<=n }. These are floor(true_value) to prevent * over-estimates when re-combining. */ static const u32 runnable_avg_yN_sum[] = { 0, 1002, 1982, 2941, 3880, 4798, 5697, 6576, 7437, 8279, 9103, 9909,10698,11470,12226,12966,13690,14398,15091,15769,16433,17082, 17718,18340,18949,19545,20128,20698,21256,21802,22336,22859,23371, }; /* |
9d85f21c9 sched: Track the ... |
2012 2013 2014 2015 2016 |
* Approximate: * val * y^n, where y^32 ~= 0.5 (~1 scheduling period) */ static __always_inline u64 decay_load(u64 val, u64 n) { |
5b51f2f80 sched: Make __upd... |
2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 |
unsigned int local_n; if (!n) return val; else if (unlikely(n > LOAD_AVG_PERIOD * 63)) return 0; /* after bounds checking we can collapse to 32-bit */ local_n = n; /* * As y^PERIOD = 1/2, we can combine * y^n = 1/2^(n/PERIOD) * k^(n%PERIOD) * With a look-up table which covers k^n (n<PERIOD) * * To achieve constant time decay_load. */ if (unlikely(local_n >= LOAD_AVG_PERIOD)) { val >>= local_n / LOAD_AVG_PERIOD; local_n %= LOAD_AVG_PERIOD; |
9d85f21c9 sched: Track the ... |
2037 |
} |
5b51f2f80 sched: Make __upd... |
2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 |
val *= runnable_avg_yN_inv[local_n]; /* We don't use SRR here since we always want to round down. */ return val >> 32; } /* * For updates fully spanning n periods, the contribution to runnable * average will be: \Sum 1024*y^n * * We can compute this reasonably efficiently by combining: * y^PERIOD = 1/2 with precomputed \Sum 1024*y^n {for n <PERIOD} */ static u32 __compute_runnable_contrib(u64 n) { u32 contrib = 0; if (likely(n <= LOAD_AVG_PERIOD)) return runnable_avg_yN_sum[n]; else if (unlikely(n >= LOAD_AVG_MAX_N)) return LOAD_AVG_MAX; /* Compute \Sum k^n combining precomputed values for k^i, \Sum k^j */ do { contrib /= 2; /* y^LOAD_AVG_PERIOD = 1/2 */ contrib += runnable_avg_yN_sum[LOAD_AVG_PERIOD]; n -= LOAD_AVG_PERIOD; } while (n > LOAD_AVG_PERIOD); contrib = decay_load(contrib, n); return contrib + runnable_avg_yN_sum[n]; |
9d85f21c9 sched: Track the ... |
2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 |
} /* * We can represent the historical contribution to runnable average as the * coefficients of a geometric series. To do this we sub-divide our runnable * history into segments of approximately 1ms (1024us); label the segment that * occurred N-ms ago p_N, with p_0 corresponding to the current period, e.g. * * [<- 1024us ->|<- 1024us ->|<- 1024us ->| ... * p0 p1 p2 * (now) (~1ms ago) (~2ms ago) * * Let u_i denote the fraction of p_i that the entity was runnable. * * We then designate the fractions u_i as our co-efficients, yielding the * following representation of historical load: * u_0 + u_1*y + u_2*y^2 + u_3*y^3 + ... * * We choose y based on the with of a reasonably scheduling period, fixing: * y^32 = 0.5 * * This means that the contribution to load ~32ms ago (u_32) will be weighted * approximately half as much as the contribution to load within the last ms * (u_0). * * When a period "rolls over" and we have new u_0`, multiplying the previous * sum again by y is sufficient to update: * load_avg = u_0` + y*(u_0 + u_1*y + u_2*y^2 + ... ) * = u_0 + u_1*y + u_2*y^2 + ... [re-labeling u_i --> u_{i+1}] */ static __always_inline int __update_entity_runnable_avg(u64 now, struct sched_avg *sa, int runnable) { |
5b51f2f80 sched: Make __upd... |
2103 2104 |
u64 delta, periods; u32 runnable_contrib; |
9d85f21c9 sched: Track the ... |
2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 |
int delta_w, decayed = 0; delta = now - sa->last_runnable_update; /* * This should only happen when time goes backwards, which it * unfortunately does during sched clock init when we swap over to TSC. */ if ((s64)delta < 0) { sa->last_runnable_update = now; return 0; } /* * Use 1024ns as the unit of measurement since it's a reasonable * approximation of 1us and fast to compute. */ delta >>= 10; if (!delta) return 0; sa->last_runnable_update = now; /* delta_w is the amount already accumulated against our next period */ delta_w = sa->runnable_avg_period % 1024; if (delta + delta_w >= 1024) { /* period roll-over */ decayed = 1; /* * Now that we know we're crossing a period boundary, figure * out how much from delta we need to complete the current * period and accrue it. */ delta_w = 1024 - delta_w; |
5b51f2f80 sched: Make __upd... |
2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 |
if (runnable) sa->runnable_avg_sum += delta_w; sa->runnable_avg_period += delta_w; delta -= delta_w; /* Figure out how many additional periods this update spans */ periods = delta / 1024; delta %= 1024; sa->runnable_avg_sum = decay_load(sa->runnable_avg_sum, periods + 1); sa->runnable_avg_period = decay_load(sa->runnable_avg_period, periods + 1); /* Efficiently calculate \sum (1..n_period) 1024*y^i */ runnable_contrib = __compute_runnable_contrib(periods); if (runnable) sa->runnable_avg_sum += runnable_contrib; sa->runnable_avg_period += runnable_contrib; |
9d85f21c9 sched: Track the ... |
2158 2159 2160 2161 2162 2163 2164 2165 2166 |
} /* Remainder of delta accrued against u_0` */ if (runnable) sa->runnable_avg_sum += delta; sa->runnable_avg_period += delta; return decayed; } |
9ee474f55 sched: Maintain t... |
2167 |
/* Synchronize an entity's decay with its parenting cfs_rq.*/ |
aff3e4988 sched: Account fo... |
2168 |
static inline u64 __synchronize_entity_decay(struct sched_entity *se) |
9ee474f55 sched: Maintain t... |
2169 2170 2171 2172 2173 2174 |
{ struct cfs_rq *cfs_rq = cfs_rq_of(se); u64 decays = atomic64_read(&cfs_rq->decay_counter); decays -= se->avg.decay_count; if (!decays) |
aff3e4988 sched: Account fo... |
2175 |
return 0; |
9ee474f55 sched: Maintain t... |
2176 2177 2178 |
se->avg.load_avg_contrib = decay_load(se->avg.load_avg_contrib, decays); se->avg.decay_count = 0; |
aff3e4988 sched: Account fo... |
2179 2180 |
return decays; |
9ee474f55 sched: Maintain t... |
2181 |
} |
c566e8e9e sched: Aggregate ... |
2182 2183 2184 2185 2186 |
#ifdef CONFIG_FAIR_GROUP_SCHED static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq, int force_update) { struct task_group *tg = cfs_rq->tg; |
bf5b986ed sched/tg: Use 'un... |
2187 |
long tg_contrib; |
c566e8e9e sched: Aggregate ... |
2188 2189 2190 |
tg_contrib = cfs_rq->runnable_load_avg + cfs_rq->blocked_load_avg; tg_contrib -= cfs_rq->tg_load_contrib; |
bf5b986ed sched/tg: Use 'un... |
2191 2192 |
if (force_update || abs(tg_contrib) > cfs_rq->tg_load_contrib / 8) { atomic_long_add(tg_contrib, &tg->load_avg); |
c566e8e9e sched: Aggregate ... |
2193 2194 2195 |
cfs_rq->tg_load_contrib += tg_contrib; } } |
8165e145c sched: Compute lo... |
2196 |
|
bb17f6557 sched: Normalize ... |
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 |
/* * Aggregate cfs_rq runnable averages into an equivalent task_group * representation for computing load contributions. */ static inline void __update_tg_runnable_avg(struct sched_avg *sa, struct cfs_rq *cfs_rq) { struct task_group *tg = cfs_rq->tg; long contrib; /* The fraction of a cpu used by this cfs_rq */ |
85b088e93 sched/fair: Avoid... |
2208 |
contrib = div_u64((u64)sa->runnable_avg_sum << NICE_0_SHIFT, |
bb17f6557 sched: Normalize ... |
2209 2210 2211 2212 2213 2214 2215 2216 |
sa->runnable_avg_period + 1); contrib -= cfs_rq->tg_runnable_contrib; if (abs(contrib) > cfs_rq->tg_runnable_contrib / 64) { atomic_add(contrib, &tg->runnable_avg); cfs_rq->tg_runnable_contrib += contrib; } } |
8165e145c sched: Compute lo... |
2217 2218 2219 2220 |
static inline void __update_group_entity_contrib(struct sched_entity *se) { struct cfs_rq *cfs_rq = group_cfs_rq(se); struct task_group *tg = cfs_rq->tg; |
bb17f6557 sched: Normalize ... |
2221 |
int runnable_avg; |
8165e145c sched: Compute lo... |
2222 2223 2224 |
u64 contrib; contrib = cfs_rq->tg_load_contrib * tg->shares; |
bf5b986ed sched/tg: Use 'un... |
2225 2226 |
se->avg.load_avg_contrib = div_u64(contrib, atomic_long_read(&tg->load_avg) + 1); |
bb17f6557 sched: Normalize ... |
2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 |
/* * For group entities we need to compute a correction term in the case * that they are consuming <1 cpu so that we would contribute the same * load as a task of equal weight. * * Explicitly co-ordinating this measurement would be expensive, but * fortunately the sum of each cpus contribution forms a usable * lower-bound on the true value. * * Consider the aggregate of 2 contributions. Either they are disjoint * (and the sum represents true value) or they are disjoint and we are * understating by the aggregate of their overlap. * * Extending this to N cpus, for a given overlap, the maximum amount we * understand is then n_i(n_i+1)/2 * w_i where n_i is the number of * cpus that overlap for this interval and w_i is the interval width. * * On a small machine; the first term is well-bounded which bounds the * total error since w_i is a subset of the period. Whereas on a * larger machine, while this first term can be larger, if w_i is the * of consequential size guaranteed to see n_i*w_i quickly converge to * our upper bound of 1-cpu. */ runnable_avg = atomic_read(&tg->runnable_avg); if (runnable_avg < NICE_0_LOAD) { se->avg.load_avg_contrib *= runnable_avg; se->avg.load_avg_contrib >>= NICE_0_SHIFT; } |
8165e145c sched: Compute lo... |
2256 |
} |
f5f9739d7 sched: Put rq's s... |
2257 2258 2259 2260 2261 2262 |
static inline void update_rq_runnable_avg(struct rq *rq, int runnable) { __update_entity_runnable_avg(rq_clock_task(rq), &rq->avg, runnable); __update_tg_runnable_avg(&rq->avg, &rq->cfs); } |
6e83125c6 sched/fair: Remov... |
2263 |
#else /* CONFIG_FAIR_GROUP_SCHED */ |
c566e8e9e sched: Aggregate ... |
2264 2265 |
static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq, int force_update) {} |
bb17f6557 sched: Normalize ... |
2266 2267 |
static inline void __update_tg_runnable_avg(struct sched_avg *sa, struct cfs_rq *cfs_rq) {} |
8165e145c sched: Compute lo... |
2268 |
static inline void __update_group_entity_contrib(struct sched_entity *se) {} |
f5f9739d7 sched: Put rq's s... |
2269 |
static inline void update_rq_runnable_avg(struct rq *rq, int runnable) {} |
6e83125c6 sched/fair: Remov... |
2270 |
#endif /* CONFIG_FAIR_GROUP_SCHED */ |
c566e8e9e sched: Aggregate ... |
2271 |
|
8165e145c sched: Compute lo... |
2272 2273 2274 2275 2276 2277 2278 2279 2280 |
static inline void __update_task_entity_contrib(struct sched_entity *se) { u32 contrib; /* avoid overflowing a 32-bit type w/ SCHED_LOAD_SCALE */ contrib = se->avg.runnable_avg_sum * scale_load_down(se->load.weight); contrib /= (se->avg.runnable_avg_period + 1); se->avg.load_avg_contrib = scale_load(contrib); } |
2dac754e1 sched: Aggregate ... |
2281 2282 2283 2284 |
/* Compute the current contribution to load_avg by se, return any delta */ static long __update_entity_load_avg_contrib(struct sched_entity *se) { long old_contrib = se->avg.load_avg_contrib; |
8165e145c sched: Compute lo... |
2285 2286 2287 |
if (entity_is_task(se)) { __update_task_entity_contrib(se); } else { |
bb17f6557 sched: Normalize ... |
2288 |
__update_tg_runnable_avg(&se->avg, group_cfs_rq(se)); |
8165e145c sched: Compute lo... |
2289 2290 |
__update_group_entity_contrib(se); } |
2dac754e1 sched: Aggregate ... |
2291 2292 2293 |
return se->avg.load_avg_contrib - old_contrib; } |
9ee474f55 sched: Maintain t... |
2294 2295 2296 2297 2298 2299 2300 2301 |
static inline void subtract_blocked_load_contrib(struct cfs_rq *cfs_rq, long load_contrib) { if (likely(load_contrib < cfs_rq->blocked_load_avg)) cfs_rq->blocked_load_avg -= load_contrib; else cfs_rq->blocked_load_avg = 0; } |
f1b17280e sched: Maintain r... |
2302 |
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq); |
9d85f21c9 sched: Track the ... |
2303 |
/* Update a sched_entity's runnable average */ |
9ee474f55 sched: Maintain t... |
2304 2305 |
static inline void update_entity_load_avg(struct sched_entity *se, int update_cfs_rq) |
9d85f21c9 sched: Track the ... |
2306 |
{ |
2dac754e1 sched: Aggregate ... |
2307 2308 |
struct cfs_rq *cfs_rq = cfs_rq_of(se); long contrib_delta; |
f1b17280e sched: Maintain r... |
2309 |
u64 now; |
2dac754e1 sched: Aggregate ... |
2310 |
|
f1b17280e sched: Maintain r... |
2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 |
/* * For a group entity we need to use their owned cfs_rq_clock_task() in * case they are the parent of a throttled hierarchy. */ if (entity_is_task(se)) now = cfs_rq_clock_task(cfs_rq); else now = cfs_rq_clock_task(group_cfs_rq(se)); if (!__update_entity_runnable_avg(now, &se->avg, se->on_rq)) |
2dac754e1 sched: Aggregate ... |
2321 2322 2323 |
return; contrib_delta = __update_entity_load_avg_contrib(se); |
9ee474f55 sched: Maintain t... |
2324 2325 2326 |
if (!update_cfs_rq) return; |
2dac754e1 sched: Aggregate ... |
2327 2328 |
if (se->on_rq) cfs_rq->runnable_load_avg += contrib_delta; |
9ee474f55 sched: Maintain t... |
2329 2330 2331 2332 2333 2334 2335 2336 |
else subtract_blocked_load_contrib(cfs_rq, -contrib_delta); } /* * Decay the load contributed by all blocked children and account this so that * their contribution may appropriately discounted when they wake up. */ |
aff3e4988 sched: Account fo... |
2337 |
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update) |
9ee474f55 sched: Maintain t... |
2338 |
{ |
f1b17280e sched: Maintain r... |
2339 |
u64 now = cfs_rq_clock_task(cfs_rq) >> 20; |
9ee474f55 sched: Maintain t... |
2340 2341 2342 |
u64 decays; decays = now - cfs_rq->last_decay; |
aff3e4988 sched: Account fo... |
2343 |
if (!decays && !force_update) |
9ee474f55 sched: Maintain t... |
2344 |
return; |
2509940fd sched/cfs_rq: Cha... |
2345 2346 2347 |
if (atomic_long_read(&cfs_rq->removed_load)) { unsigned long removed_load; removed_load = atomic_long_xchg(&cfs_rq->removed_load, 0); |
aff3e4988 sched: Account fo... |
2348 2349 |
subtract_blocked_load_contrib(cfs_rq, removed_load); } |
9ee474f55 sched: Maintain t... |
2350 |
|
aff3e4988 sched: Account fo... |
2351 2352 2353 2354 2355 2356 |
if (decays) { cfs_rq->blocked_load_avg = decay_load(cfs_rq->blocked_load_avg, decays); atomic64_add(decays, &cfs_rq->decay_counter); cfs_rq->last_decay = now; } |
c566e8e9e sched: Aggregate ... |
2357 2358 |
__update_cfs_rq_tg_load_contrib(cfs_rq, force_update); |
9d85f21c9 sched: Track the ... |
2359 |
} |
18bf2805d sched: Maintain p... |
2360 |
|
2dac754e1 sched: Aggregate ... |
2361 2362 |
/* Add the load generated by se into cfs_rq's child load-average */ static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f55 sched: Maintain t... |
2363 2364 |
struct sched_entity *se, int wakeup) |
2dac754e1 sched: Aggregate ... |
2365 |
{ |
aff3e4988 sched: Account fo... |
2366 2367 2368 2369 |
/* * We track migrations using entity decay_count <= 0, on a wake-up * migration we use a negative decay count to track the remote decays * accumulated while sleeping. |
a75cdaa91 sched: Set an ini... |
2370 2371 2372 2373 |
* * Newly forked tasks are enqueued with se->avg.decay_count == 0, they * are seen by enqueue_entity_load_avg() as a migration with an already * constructed load_avg_contrib. |
aff3e4988 sched: Account fo... |
2374 2375 |
*/ if (unlikely(se->avg.decay_count <= 0)) { |
78becc270 sched: Use an acc... |
2376 |
se->avg.last_runnable_update = rq_clock_task(rq_of(cfs_rq)); |
aff3e4988 sched: Account fo... |
2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 |
if (se->avg.decay_count) { /* * In a wake-up migration we have to approximate the * time sleeping. This is because we can't synchronize * clock_task between the two cpus, and it is not * guaranteed to be read-safe. Instead, we can * approximate this using our carried decays, which are * explicitly atomically readable. */ se->avg.last_runnable_update -= (-se->avg.decay_count) << 20; update_entity_load_avg(se, 0); /* Indicate that we're now synchronized and on-rq */ se->avg.decay_count = 0; } |
9ee474f55 sched: Maintain t... |
2392 2393 |
wakeup = 0; } else { |
9390675af Revert "sched: Fi... |
2394 |
__synchronize_entity_decay(se); |
9ee474f55 sched: Maintain t... |
2395 |
} |
aff3e4988 sched: Account fo... |
2396 2397 |
/* migrated tasks did not contribute to our blocked load */ if (wakeup) { |
9ee474f55 sched: Maintain t... |
2398 |
subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib); |
aff3e4988 sched: Account fo... |
2399 2400 |
update_entity_load_avg(se, 0); } |
9ee474f55 sched: Maintain t... |
2401 |
|
2dac754e1 sched: Aggregate ... |
2402 |
cfs_rq->runnable_load_avg += se->avg.load_avg_contrib; |
aff3e4988 sched: Account fo... |
2403 2404 |
/* we force update consideration on load-balancer moves */ update_cfs_rq_blocked_load(cfs_rq, !wakeup); |
2dac754e1 sched: Aggregate ... |
2405 |
} |
9ee474f55 sched: Maintain t... |
2406 2407 2408 2409 2410 |
/* * Remove se's load from this cfs_rq child load-average, if the entity is * transitioning to a blocked state we track its projected decay using * blocked_load_avg. */ |
2dac754e1 sched: Aggregate ... |
2411 |
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f55 sched: Maintain t... |
2412 2413 |
struct sched_entity *se, int sleep) |
2dac754e1 sched: Aggregate ... |
2414 |
{ |
9ee474f55 sched: Maintain t... |
2415 |
update_entity_load_avg(se, 1); |
aff3e4988 sched: Account fo... |
2416 2417 |
/* we force update consideration on load-balancer moves */ update_cfs_rq_blocked_load(cfs_rq, !sleep); |
9ee474f55 sched: Maintain t... |
2418 |
|
2dac754e1 sched: Aggregate ... |
2419 |
cfs_rq->runnable_load_avg -= se->avg.load_avg_contrib; |
9ee474f55 sched: Maintain t... |
2420 2421 2422 2423 |
if (sleep) { cfs_rq->blocked_load_avg += se->avg.load_avg_contrib; se->avg.decay_count = atomic64_read(&cfs_rq->decay_counter); } /* migrations, e.g. sleep=0 leave decay_count == 0 */ |
2dac754e1 sched: Aggregate ... |
2424 |
} |
642dbc39a sched: Fix wrong ... |
2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 |
/* * Update the rq's load with the elapsed running time before entering * idle. if the last scheduled task is not a CFS task, idle_enter will * be the only way to update the runnable statistic. */ void idle_enter_fair(struct rq *this_rq) { update_rq_runnable_avg(this_rq, 1); } /* * Update the rq's load with the elapsed idle time before a task is * scheduled. if the newly scheduled task is not a CFS task, idle_exit will * be the only way to update the runnable statistic. */ void idle_exit_fair(struct rq *this_rq) { update_rq_runnable_avg(this_rq, 0); } |
6e83125c6 sched/fair: Remov... |
2445 |
static int idle_balance(struct rq *this_rq); |
38033c37f sched: Push down ... |
2446 |
#else /* CONFIG_SMP */ |
9ee474f55 sched: Maintain t... |
2447 2448 |
static inline void update_entity_load_avg(struct sched_entity *se, int update_cfs_rq) {} |
18bf2805d sched: Maintain p... |
2449 |
static inline void update_rq_runnable_avg(struct rq *rq, int runnable) {} |
2dac754e1 sched: Aggregate ... |
2450 |
static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f55 sched: Maintain t... |
2451 2452 |
struct sched_entity *se, int wakeup) {} |
2dac754e1 sched: Aggregate ... |
2453 |
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq, |
9ee474f55 sched: Maintain t... |
2454 2455 |
struct sched_entity *se, int sleep) {} |
aff3e4988 sched: Account fo... |
2456 2457 |
static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update) {} |
6e83125c6 sched/fair: Remov... |
2458 2459 2460 2461 2462 |
static inline int idle_balance(struct rq *rq) { return 0; } |
38033c37f sched: Push down ... |
2463 |
#endif /* CONFIG_SMP */ |
9d85f21c9 sched: Track the ... |
2464 |
|
2396af69b sched: remove the... |
2465 |
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
2466 |
{ |
bf0f6f24a sched: cfs core, ... |
2467 |
#ifdef CONFIG_SCHEDSTATS |
e414314cc sched: Fix latenc... |
2468 2469 2470 2471 |
struct task_struct *tsk = NULL; if (entity_is_task(se)) tsk = task_of(se); |
41acab885 sched: Implement ... |
2472 |
if (se->statistics.sleep_start) { |
78becc270 sched: Use an acc... |
2473 |
u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.sleep_start; |
bf0f6f24a sched: cfs core, ... |
2474 2475 2476 |
if ((s64)delta < 0) delta = 0; |
41acab885 sched: Implement ... |
2477 2478 |
if (unlikely(delta > se->statistics.sleep_max)) se->statistics.sleep_max = delta; |
bf0f6f24a sched: cfs core, ... |
2479 |
|
8c79a045f sched/events: Rev... |
2480 |
se->statistics.sleep_start = 0; |
41acab885 sched: Implement ... |
2481 |
se->statistics.sum_sleep_runtime += delta; |
9745512ce sched: latencytop... |
2482 |
|
768d0c272 sched: Add wait, ... |
2483 |
if (tsk) { |
e414314cc sched: Fix latenc... |
2484 |
account_scheduler_latency(tsk, delta >> 10, 1); |
768d0c272 sched: Add wait, ... |
2485 2486 |
trace_sched_stat_sleep(tsk, delta); } |
bf0f6f24a sched: cfs core, ... |
2487 |
} |
41acab885 sched: Implement ... |
2488 |
if (se->statistics.block_start) { |
78becc270 sched: Use an acc... |
2489 |
u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.block_start; |
bf0f6f24a sched: cfs core, ... |
2490 2491 2492 |
if ((s64)delta < 0) delta = 0; |
41acab885 sched: Implement ... |
2493 2494 |
if (unlikely(delta > se->statistics.block_max)) se->statistics.block_max = delta; |
bf0f6f24a sched: cfs core, ... |
2495 |
|
8c79a045f sched/events: Rev... |
2496 |
se->statistics.block_start = 0; |
41acab885 sched: Implement ... |
2497 |
se->statistics.sum_sleep_runtime += delta; |
30084fbd1 sched: fix profil... |
2498 |
|
e414314cc sched: Fix latenc... |
2499 |
if (tsk) { |
8f0dfc34e sched: Provide io... |
2500 |
if (tsk->in_iowait) { |
41acab885 sched: Implement ... |
2501 2502 |
se->statistics.iowait_sum += delta; se->statistics.iowait_count++; |
768d0c272 sched: Add wait, ... |
2503 |
trace_sched_stat_iowait(tsk, delta); |
8f0dfc34e sched: Provide io... |
2504 |
} |
b781a602a events, sched: Ad... |
2505 |
trace_sched_stat_blocked(tsk, delta); |
e414314cc sched: Fix latenc... |
2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 |
/* * Blocking time is in units of nanosecs, so shift by * 20 to get a milliseconds-range estimation of the * amount of time that the task spent sleeping: */ if (unlikely(prof_on == SLEEP_PROFILING)) { profile_hits(SLEEP_PROFILING, (void *)get_wchan(tsk), delta >> 20); } account_scheduler_latency(tsk, delta >> 10, 0); |
30084fbd1 sched: fix profil... |
2517 |
} |
bf0f6f24a sched: cfs core, ... |
2518 2519 2520 |
} #endif } |
ddc972975 sched debug: chec... |
2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 |
static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se) { #ifdef CONFIG_SCHED_DEBUG s64 d = se->vruntime - cfs_rq->min_vruntime; if (d < 0) d = -d; if (d > 3*sysctl_sched_latency) schedstat_inc(cfs_rq, nr_spread_over); #endif } |
bf0f6f24a sched: cfs core, ... |
2533 |
static void |
aeb73b040 sched: clean up n... |
2534 2535 |
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) { |
1af5f730f sched: more accur... |
2536 |
u64 vruntime = cfs_rq->min_vruntime; |
94dfb5e75 sched: add tree b... |
2537 |
|
2cb8600e6 sched: documentat... |
2538 2539 2540 2541 2542 2543 |
/* * The 'current' period is already promised to the current tasks, * however the extra weight of the new task will slow them down a * little, place the new task so that it fits in the slot that * stays open at the end. */ |
94dfb5e75 sched: add tree b... |
2544 |
if (initial && sched_feat(START_DEBIT)) |
f9c0b0950 sched: revert bac... |
2545 |
vruntime += sched_vslice(cfs_rq, se); |
aeb73b040 sched: clean up n... |
2546 |
|
a2e7a7eb2 sched: Remove unn... |
2547 |
/* sleeps up to a single latency don't count. */ |
5ca9880c6 sched: Remove FAI... |
2548 |
if (!initial) { |
a2e7a7eb2 sched: Remove unn... |
2549 |
unsigned long thresh = sysctl_sched_latency; |
a7be37ac8 sched: revert the... |
2550 |
|
a2e7a7eb2 sched: Remove unn... |
2551 |
/* |
a2e7a7eb2 sched: Remove unn... |
2552 2553 2554 2555 2556 |
* Halve their sleep time's effect, to allow * for a gentler effect of sleepers: */ if (sched_feat(GENTLE_FAIR_SLEEPERS)) thresh >>= 1; |
51e0304ce sched: Implement ... |
2557 |
|
a2e7a7eb2 sched: Remove unn... |
2558 |
vruntime -= thresh; |
aeb73b040 sched: clean up n... |
2559 |
} |
b5d9d734a sched: Ensure tha... |
2560 |
/* ensure we never gain time by being placed backwards. */ |
16c8f1c72 sched/fair: Set s... |
2561 |
se->vruntime = max_vruntime(se->vruntime, vruntime); |
aeb73b040 sched: clean up n... |
2562 |
} |
d3d9dc330 sched: Throttle e... |
2563 |
static void check_enqueue_throttle(struct cfs_rq *cfs_rq); |
aeb73b040 sched: clean up n... |
2564 |
static void |
88ec22d3e sched: Remove the... |
2565 |
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) |
bf0f6f24a sched: cfs core, ... |
2566 2567 |
{ /* |
88ec22d3e sched: Remove the... |
2568 |
* Update the normalized vruntime before updating min_vruntime |
0fc576d59 sched/fair: Fix t... |
2569 |
* through calling update_curr(). |
88ec22d3e sched: Remove the... |
2570 |
*/ |
371fd7e7a sched: Add enqueu... |
2571 |
if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING)) |
88ec22d3e sched: Remove the... |
2572 2573 2574 |
se->vruntime += cfs_rq->min_vruntime; /* |
a2a2d6807 sched: cleanup, m... |
2575 |
* Update run-time statistics of the 'current'. |
bf0f6f24a sched: cfs core, ... |
2576 |
*/ |
b7cc08965 sched: remove the... |
2577 |
update_curr(cfs_rq); |
f269ae046 sched: Update_cfs... |
2578 |
enqueue_entity_load_avg(cfs_rq, se, flags & ENQUEUE_WAKEUP); |
17bc14b76 Revert "sched: Up... |
2579 2580 |
account_entity_enqueue(cfs_rq, se); update_cfs_shares(cfs_rq); |
bf0f6f24a sched: cfs core, ... |
2581 |
|
88ec22d3e sched: Remove the... |
2582 |
if (flags & ENQUEUE_WAKEUP) { |
aeb73b040 sched: clean up n... |
2583 |
place_entity(cfs_rq, se, 0); |
2396af69b sched: remove the... |
2584 |
enqueue_sleeper(cfs_rq, se); |
e9acbff64 sched: introduce ... |
2585 |
} |
bf0f6f24a sched: cfs core, ... |
2586 |
|
d2417e5a3 sched: remove the... |
2587 |
update_stats_enqueue(cfs_rq, se); |
ddc972975 sched debug: chec... |
2588 |
check_spread(cfs_rq, se); |
83b699ed2 sched: revert rec... |
2589 2590 |
if (se != cfs_rq->curr) __enqueue_entity(cfs_rq, se); |
2069dd75c sched: Rewrite tg... |
2591 |
se->on_rq = 1; |
3d4b47b4b sched: Implement ... |
2592 |
|
d3d9dc330 sched: Throttle e... |
2593 |
if (cfs_rq->nr_running == 1) { |
3d4b47b4b sched: Implement ... |
2594 |
list_add_leaf_cfs_rq(cfs_rq); |
d3d9dc330 sched: Throttle e... |
2595 2596 |
check_enqueue_throttle(cfs_rq); } |
bf0f6f24a sched: cfs core, ... |
2597 |
} |
2c13c919d sched: Limit the ... |
2598 |
static void __clear_buddies_last(struct sched_entity *se) |
2002c6959 sched: release bu... |
2599 |
{ |
2c13c919d sched: Limit the ... |
2600 2601 |
for_each_sched_entity(se) { struct cfs_rq *cfs_rq = cfs_rq_of(se); |
f10447998 sched/fair: Clean... |
2602 |
if (cfs_rq->last != se) |
2c13c919d sched: Limit the ... |
2603 |
break; |
f10447998 sched/fair: Clean... |
2604 2605 |
cfs_rq->last = NULL; |
2c13c919d sched: Limit the ... |
2606 2607 |
} } |
2002c6959 sched: release bu... |
2608 |
|
2c13c919d sched: Limit the ... |
2609 2610 2611 2612 |
static void __clear_buddies_next(struct sched_entity *se) { for_each_sched_entity(se) { struct cfs_rq *cfs_rq = cfs_rq_of(se); |
f10447998 sched/fair: Clean... |
2613 |
if (cfs_rq->next != se) |
2c13c919d sched: Limit the ... |
2614 |
break; |
f10447998 sched/fair: Clean... |
2615 2616 |
cfs_rq->next = NULL; |
2c13c919d sched: Limit the ... |
2617 |
} |
2002c6959 sched: release bu... |
2618 |
} |
ac53db596 sched: Use a budd... |
2619 2620 2621 2622 |
static void __clear_buddies_skip(struct sched_entity *se) { for_each_sched_entity(se) { struct cfs_rq *cfs_rq = cfs_rq_of(se); |
f10447998 sched/fair: Clean... |
2623 |
if (cfs_rq->skip != se) |
ac53db596 sched: Use a budd... |
2624 |
break; |
f10447998 sched/fair: Clean... |
2625 2626 |
cfs_rq->skip = NULL; |
ac53db596 sched: Use a budd... |
2627 2628 |
} } |
a571bbeaf sched: fix buddie... |
2629 2630 |
static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se) { |
2c13c919d sched: Limit the ... |
2631 2632 2633 2634 2635 |
if (cfs_rq->last == se) __clear_buddies_last(se); if (cfs_rq->next == se) __clear_buddies_next(se); |
ac53db596 sched: Use a budd... |
2636 2637 2638 |
if (cfs_rq->skip == se) __clear_buddies_skip(se); |
a571bbeaf sched: fix buddie... |
2639 |
} |
6c16a6dcb sched: Fix compil... |
2640 |
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq); |
d8b4986d3 sched: Return unu... |
2641 |
|
bf0f6f24a sched: cfs core, ... |
2642 |
static void |
371fd7e7a sched: Add enqueu... |
2643 |
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) |
bf0f6f24a sched: cfs core, ... |
2644 |
{ |
a2a2d6807 sched: cleanup, m... |
2645 2646 2647 2648 |
/* * Update run-time statistics of the 'current'. */ update_curr(cfs_rq); |
17bc14b76 Revert "sched: Up... |
2649 |
dequeue_entity_load_avg(cfs_rq, se, flags & DEQUEUE_SLEEP); |
a2a2d6807 sched: cleanup, m... |
2650 |
|
19b6a2e37 sched: remove the... |
2651 |
update_stats_dequeue(cfs_rq, se); |
371fd7e7a sched: Add enqueu... |
2652 |
if (flags & DEQUEUE_SLEEP) { |
67e9fb2a3 sched: add vslice |
2653 |
#ifdef CONFIG_SCHEDSTATS |
bf0f6f24a sched: cfs core, ... |
2654 2655 2656 2657 |
if (entity_is_task(se)) { struct task_struct *tsk = task_of(se); if (tsk->state & TASK_INTERRUPTIBLE) |
78becc270 sched: Use an acc... |
2658 |
se->statistics.sleep_start = rq_clock(rq_of(cfs_rq)); |
bf0f6f24a sched: cfs core, ... |
2659 |
if (tsk->state & TASK_UNINTERRUPTIBLE) |
78becc270 sched: Use an acc... |
2660 |
se->statistics.block_start = rq_clock(rq_of(cfs_rq)); |
bf0f6f24a sched: cfs core, ... |
2661 |
} |
db36cc7d6 sched: clean up s... |
2662 |
#endif |
67e9fb2a3 sched: add vslice |
2663 |
} |
2002c6959 sched: release bu... |
2664 |
clear_buddies(cfs_rq, se); |
4793241be sched: backward l... |
2665 |
|
83b699ed2 sched: revert rec... |
2666 |
if (se != cfs_rq->curr) |
30cfdcfc5 sched: do not kee... |
2667 |
__dequeue_entity(cfs_rq, se); |
17bc14b76 Revert "sched: Up... |
2668 |
se->on_rq = 0; |
30cfdcfc5 sched: do not kee... |
2669 |
account_entity_dequeue(cfs_rq, se); |
88ec22d3e sched: Remove the... |
2670 2671 2672 2673 2674 2675 |
/* * Normalize the entity after updating the min_vruntime because the * update can refer to the ->curr item and we need to reflect this * movement in our normalized position. */ |
371fd7e7a sched: Add enqueu... |
2676 |
if (!(flags & DEQUEUE_SLEEP)) |
88ec22d3e sched: Remove the... |
2677 |
se->vruntime -= cfs_rq->min_vruntime; |
1e8762317 sched: Fix ->min_... |
2678 |
|
d8b4986d3 sched: Return unu... |
2679 2680 |
/* return excess runtime on last dequeue */ return_cfs_rq_runtime(cfs_rq); |
1e8762317 sched: Fix ->min_... |
2681 |
update_min_vruntime(cfs_rq); |
17bc14b76 Revert "sched: Up... |
2682 |
update_cfs_shares(cfs_rq); |
bf0f6f24a sched: cfs core, ... |
2683 2684 2685 2686 2687 |
} /* * Preempt the current task with a newly woken task if needed: */ |
7c92e54f6 sched: simplify _... |
2688 |
static void |
2e09bf556 sched: wakeup gra... |
2689 |
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr) |
bf0f6f24a sched: cfs core, ... |
2690 |
{ |
116978308 sched: fix ideal_... |
2691 |
unsigned long ideal_runtime, delta_exec; |
f4cfb33ed sched: Remove red... |
2692 2693 |
struct sched_entity *se; s64 delta; |
116978308 sched: fix ideal_... |
2694 |
|
6d0f0ebd0 sched: simplify a... |
2695 |
ideal_runtime = sched_slice(cfs_rq, curr); |
116978308 sched: fix ideal_... |
2696 |
delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime; |
a9f3e2b54 sched: clear budd... |
2697 |
if (delta_exec > ideal_runtime) { |
bf0f6f24a sched: cfs core, ... |
2698 |
resched_task(rq_of(cfs_rq)->curr); |
a9f3e2b54 sched: clear budd... |
2699 2700 2701 2702 2703 |
/* * The current task ran long enough, ensure it doesn't get * re-elected due to buddy favours. */ clear_buddies(cfs_rq, curr); |
f685ceaca sched: Strengthen... |
2704 2705 2706 2707 2708 2709 2710 2711 |
return; } /* * Ensure that a task that missed wakeup preemption by a * narrow margin doesn't have to wait for a full slice. * This also mitigates buddy induced latencies under load. */ |
f685ceaca sched: Strengthen... |
2712 2713 |
if (delta_exec < sysctl_sched_min_granularity) return; |
f4cfb33ed sched: Remove red... |
2714 2715 |
se = __pick_first_entity(cfs_rq); delta = curr->vruntime - se->vruntime; |
f685ceaca sched: Strengthen... |
2716 |
|
f4cfb33ed sched: Remove red... |
2717 2718 |
if (delta < 0) return; |
d7d829441 sched: Fix signed... |
2719 |
|
f4cfb33ed sched: Remove red... |
2720 2721 |
if (delta > ideal_runtime) resched_task(rq_of(cfs_rq)->curr); |
bf0f6f24a sched: cfs core, ... |
2722 |
} |
83b699ed2 sched: revert rec... |
2723 |
static void |
8494f412e sched: remove the... |
2724 |
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) |
bf0f6f24a sched: cfs core, ... |
2725 |
{ |
83b699ed2 sched: revert rec... |
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 |
/* 'current' is not kept within the tree. */ if (se->on_rq) { /* * Any task has to be enqueued before it get to execute on * a CPU. So account for the time it spent waiting on the * runqueue. */ update_stats_wait_end(cfs_rq, se); __dequeue_entity(cfs_rq, se); } |
79303e9e0 sched: remove the... |
2736 |
update_stats_curr_start(cfs_rq, se); |
429d43bcc sched: cleanup: s... |
2737 |
cfs_rq->curr = se; |
eba1ed4b7 sched: debug: tra... |
2738 2739 2740 2741 2742 2743 |
#ifdef CONFIG_SCHEDSTATS /* * Track our maximum slice length, if the CPU's load is at * least twice that of our own weight (i.e. dont track it * when there are only lesser-weight tasks around): */ |
495eca494 sched: clean up s... |
2744 |
if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) { |
41acab885 sched: Implement ... |
2745 |
se->statistics.slice_max = max(se->statistics.slice_max, |
eba1ed4b7 sched: debug: tra... |
2746 2747 2748 |
se->sum_exec_runtime - se->prev_sum_exec_runtime); } #endif |
4a55b4503 sched: improve pr... |
2749 |
se->prev_sum_exec_runtime = se->sum_exec_runtime; |
bf0f6f24a sched: cfs core, ... |
2750 |
} |
3f3a49048 sched: virtual ti... |
2751 2752 |
static int wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se); |
ac53db596 sched: Use a budd... |
2753 2754 2755 2756 2757 2758 2759 |
/* * Pick the next process, keeping these things in mind, in this order: * 1) keep things fair between processes/task groups * 2) pick the "next" process, since someone really wants that to run * 3) pick the "last" process, for cache locality * 4) do not run the "skip" process, if something else is available */ |
678d5718d sched/fair: Optim... |
2760 2761 |
static struct sched_entity * pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr) |
aa2ac2522 sched: fix overlo... |
2762 |
{ |
678d5718d sched/fair: Optim... |
2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 |
struct sched_entity *left = __pick_first_entity(cfs_rq); struct sched_entity *se; /* * If curr is set we have to see if its left of the leftmost entity * still in the tree, provided there was anything in the tree at all. */ if (!left || (curr && entity_before(curr, left))) left = curr; se = left; /* ideally we run the leftmost entity */ |
f4b6755fb sched: cleanup fa... |
2774 |
|
ac53db596 sched: Use a budd... |
2775 2776 2777 2778 2779 |
/* * Avoid running the skip buddy, if running something else can * be done without getting too unfair. */ if (cfs_rq->skip == se) { |
678d5718d sched/fair: Optim... |
2780 2781 2782 2783 2784 2785 2786 2787 2788 |
struct sched_entity *second; if (se == curr) { second = __pick_first_entity(cfs_rq); } else { second = __pick_next_entity(se); if (!second || (curr && entity_before(curr, second))) second = curr; } |
ac53db596 sched: Use a budd... |
2789 2790 2791 |
if (second && wakeup_preempt_entity(second, left) < 1) se = second; } |
aa2ac2522 sched: fix overlo... |
2792 |
|
f685ceaca sched: Strengthen... |
2793 2794 2795 2796 2797 |
/* * Prefer last buddy, try to return the CPU to a preempted task. */ if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1) se = cfs_rq->last; |
ac53db596 sched: Use a budd... |
2798 2799 2800 2801 2802 |
/* * Someone really wants this to run. If it's not unfair, run it. */ if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1) se = cfs_rq->next; |
f685ceaca sched: Strengthen... |
2803 |
clear_buddies(cfs_rq, se); |
4793241be sched: backward l... |
2804 2805 |
return se; |
aa2ac2522 sched: fix overlo... |
2806 |
} |
678d5718d sched/fair: Optim... |
2807 |
static bool check_cfs_rq_runtime(struct cfs_rq *cfs_rq); |
d3d9dc330 sched: Throttle e... |
2808 |
|
ab6cde269 sched: remove the... |
2809 |
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev) |
bf0f6f24a sched: cfs core, ... |
2810 2811 2812 2813 2814 2815 |
{ /* * If still on the runqueue then deactivate_task() * was not called and update_curr() has to be done: */ if (prev->on_rq) |
b7cc08965 sched: remove the... |
2816 |
update_curr(cfs_rq); |
bf0f6f24a sched: cfs core, ... |
2817 |
|
d3d9dc330 sched: Throttle e... |
2818 2819 |
/* throttle cfs_rqs exceeding runtime */ check_cfs_rq_runtime(cfs_rq); |
ddc972975 sched debug: chec... |
2820 |
check_spread(cfs_rq, prev); |
30cfdcfc5 sched: do not kee... |
2821 |
if (prev->on_rq) { |
5870db5b8 sched: remove the... |
2822 |
update_stats_wait_start(cfs_rq, prev); |
30cfdcfc5 sched: do not kee... |
2823 2824 |
/* Put 'current' back into the tree. */ __enqueue_entity(cfs_rq, prev); |
9d85f21c9 sched: Track the ... |
2825 |
/* in !on_rq case, update occurred at dequeue */ |
9ee474f55 sched: Maintain t... |
2826 |
update_entity_load_avg(prev, 1); |
30cfdcfc5 sched: do not kee... |
2827 |
} |
429d43bcc sched: cleanup: s... |
2828 |
cfs_rq->curr = NULL; |
bf0f6f24a sched: cfs core, ... |
2829 |
} |
8f4d37ec0 sched: high-res p... |
2830 2831 |
static void entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued) |
bf0f6f24a sched: cfs core, ... |
2832 |
{ |
bf0f6f24a sched: cfs core, ... |
2833 |
/* |
30cfdcfc5 sched: do not kee... |
2834 |
* Update run-time statistics of the 'current'. |
bf0f6f24a sched: cfs core, ... |
2835 |
*/ |
30cfdcfc5 sched: do not kee... |
2836 |
update_curr(cfs_rq); |
bf0f6f24a sched: cfs core, ... |
2837 |
|
43365bd7f sched: Move perio... |
2838 |
/* |
9d85f21c9 sched: Track the ... |
2839 2840 |
* Ensure that runnable average is periodically updated. */ |
9ee474f55 sched: Maintain t... |
2841 |
update_entity_load_avg(curr, 1); |
aff3e4988 sched: Account fo... |
2842 |
update_cfs_rq_blocked_load(cfs_rq, 1); |
bf0bd948d sched: Ensure upd... |
2843 |
update_cfs_shares(cfs_rq); |
9d85f21c9 sched: Track the ... |
2844 |
|
8f4d37ec0 sched: high-res p... |
2845 2846 2847 2848 2849 |
#ifdef CONFIG_SCHED_HRTICK /* * queued ticks are scheduled to match the slice, so don't bother * validating it and just reschedule. */ |
983ed7a66 sched: add static... |
2850 2851 2852 2853 |
if (queued) { resched_task(rq_of(cfs_rq)->curr); return; } |
8f4d37ec0 sched: high-res p... |
2854 2855 2856 2857 2858 2859 2860 |
/* * don't let the period tick interfere with the hrtick preemption */ if (!sched_feat(DOUBLE_TICK) && hrtimer_active(&rq_of(cfs_rq)->hrtick_timer)) return; #endif |
2c2efaed9 sched: Kill WAKEU... |
2861 |
if (cfs_rq->nr_running > 1) |
2e09bf556 sched: wakeup gra... |
2862 |
check_preempt_tick(cfs_rq, curr); |
bf0f6f24a sched: cfs core, ... |
2863 |
} |
ab84d31e1 sched: Introduce ... |
2864 2865 2866 2867 2868 2869 |
/************************************************** * CFS bandwidth control machinery */ #ifdef CONFIG_CFS_BANDWIDTH |
029632fbb sched: Make separ... |
2870 2871 |
#ifdef HAVE_JUMP_LABEL |
c5905afb0 static keys: Intr... |
2872 |
static struct static_key __cfs_bandwidth_used; |
029632fbb sched: Make separ... |
2873 2874 2875 |
static inline bool cfs_bandwidth_used(void) { |
c5905afb0 static keys: Intr... |
2876 |
return static_key_false(&__cfs_bandwidth_used); |
029632fbb sched: Make separ... |
2877 |
} |
1ee14e6c8 sched: Fix race o... |
2878 |
void cfs_bandwidth_usage_inc(void) |
029632fbb sched: Make separ... |
2879 |
{ |
1ee14e6c8 sched: Fix race o... |
2880 2881 2882 2883 2884 2885 |
static_key_slow_inc(&__cfs_bandwidth_used); } void cfs_bandwidth_usage_dec(void) { static_key_slow_dec(&__cfs_bandwidth_used); |
029632fbb sched: Make separ... |
2886 2887 2888 2889 2890 2891 |
} #else /* HAVE_JUMP_LABEL */ static bool cfs_bandwidth_used(void) { return true; } |
1ee14e6c8 sched: Fix race o... |
2892 2893 |
void cfs_bandwidth_usage_inc(void) {} void cfs_bandwidth_usage_dec(void) {} |
029632fbb sched: Make separ... |
2894 |
#endif /* HAVE_JUMP_LABEL */ |
ab84d31e1 sched: Introduce ... |
2895 2896 2897 2898 2899 2900 2901 2902 |
/* * default period for cfs group bandwidth. * default: 0.1s, units: nanoseconds */ static inline u64 default_cfs_period(void) { return 100000000ULL; } |
ec12cb7f3 sched: Accumulate... |
2903 2904 2905 2906 2907 |
static inline u64 sched_cfs_bandwidth_slice(void) { return (u64)sysctl_sched_cfs_bandwidth_slice * NSEC_PER_USEC; } |
a9cf55b28 sched: Expire inv... |
2908 2909 2910 2911 2912 2913 2914 |
/* * Replenish runtime according to assigned quota and update expiration time. * We use sched_clock_cpu directly instead of rq->clock to avoid adding * additional synchronization around rq->lock. * * requires cfs_b->lock */ |
029632fbb sched: Make separ... |
2915 |
void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b) |
a9cf55b28 sched: Expire inv... |
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 |
{ u64 now; if (cfs_b->quota == RUNTIME_INF) return; now = sched_clock_cpu(smp_processor_id()); cfs_b->runtime = cfs_b->quota; cfs_b->runtime_expires = now + ktime_to_ns(cfs_b->period); } |
029632fbb sched: Make separ... |
2926 2927 2928 2929 |
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg) { return &tg->cfs_bandwidth; } |
f1b17280e sched: Maintain r... |
2930 2931 2932 2933 2934 |
/* rq->task_clock normalized against any time this cfs_rq has spent throttled */ static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq) { if (unlikely(cfs_rq->throttle_count)) return cfs_rq->throttled_clock_task; |
78becc270 sched: Use an acc... |
2935 |
return rq_clock_task(rq_of(cfs_rq)) - cfs_rq->throttled_clock_task_time; |
f1b17280e sched: Maintain r... |
2936 |
} |
85dac906b sched: Add suppor... |
2937 2938 |
/* returns 0 on failure to allocate runtime */ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq) |
ec12cb7f3 sched: Accumulate... |
2939 2940 2941 |
{ struct task_group *tg = cfs_rq->tg; struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg); |
a9cf55b28 sched: Expire inv... |
2942 |
u64 amount = 0, min_amount, expires; |
ec12cb7f3 sched: Accumulate... |
2943 2944 2945 2946 2947 2948 2949 |
/* note: this is a positive sum as runtime_remaining <= 0 */ min_amount = sched_cfs_bandwidth_slice() - cfs_rq->runtime_remaining; raw_spin_lock(&cfs_b->lock); if (cfs_b->quota == RUNTIME_INF) amount = min_amount; |
58088ad01 sched: Add a time... |
2950 |
else { |
a9cf55b28 sched: Expire inv... |
2951 2952 2953 2954 2955 2956 2957 2958 |
/* * If the bandwidth pool has become inactive, then at least one * period must have elapsed since the last consumption. * Refresh the global state and ensure bandwidth timer becomes * active. */ if (!cfs_b->timer_active) { __refill_cfs_bandwidth_runtime(cfs_b); |
09dc4ab03 sched/fair: Fix t... |
2959 |
__start_cfs_bandwidth(cfs_b, false); |
a9cf55b28 sched: Expire inv... |
2960 |
} |
58088ad01 sched: Add a time... |
2961 2962 2963 2964 2965 2966 |
if (cfs_b->runtime > 0) { amount = min(cfs_b->runtime, min_amount); cfs_b->runtime -= amount; cfs_b->idle = 0; } |
ec12cb7f3 sched: Accumulate... |
2967 |
} |
a9cf55b28 sched: Expire inv... |
2968 |
expires = cfs_b->runtime_expires; |
ec12cb7f3 sched: Accumulate... |
2969 2970 2971 |
raw_spin_unlock(&cfs_b->lock); cfs_rq->runtime_remaining += amount; |
a9cf55b28 sched: Expire inv... |
2972 2973 2974 2975 2976 2977 2978 |
/* * we may have advanced our local expiration to account for allowed * spread between our sched_clock and the one on which runtime was * issued. */ if ((s64)(expires - cfs_rq->runtime_expires) > 0) cfs_rq->runtime_expires = expires; |
85dac906b sched: Add suppor... |
2979 2980 |
return cfs_rq->runtime_remaining > 0; |
ec12cb7f3 sched: Accumulate... |
2981 |
} |
a9cf55b28 sched: Expire inv... |
2982 2983 2984 2985 2986 |
/* * Note: This depends on the synchronization provided by sched_clock and the * fact that rq->clock snapshots this value. */ static void expire_cfs_rq_runtime(struct cfs_rq *cfs_rq) |
ec12cb7f3 sched: Accumulate... |
2987 |
{ |
a9cf55b28 sched: Expire inv... |
2988 |
struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); |
a9cf55b28 sched: Expire inv... |
2989 2990 |
/* if the deadline is ahead of our clock, nothing to do */ |
78becc270 sched: Use an acc... |
2991 |
if (likely((s64)(rq_clock(rq_of(cfs_rq)) - cfs_rq->runtime_expires) < 0)) |
ec12cb7f3 sched: Accumulate... |
2992 |
return; |
a9cf55b28 sched: Expire inv... |
2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 |
if (cfs_rq->runtime_remaining < 0) return; /* * If the local deadline has passed we have to consider the * possibility that our sched_clock is 'fast' and the global deadline * has not truly expired. * * Fortunately we can check determine whether this the case by checking * whether the global deadline has advanced. */ if ((s64)(cfs_rq->runtime_expires - cfs_b->runtime_expires) >= 0) { /* extend local deadline, drift is bounded above by 2 ticks */ cfs_rq->runtime_expires += TICK_NSEC; } else { /* global deadline is ahead, expiration has passed */ cfs_rq->runtime_remaining = 0; } } |
9dbdb1555 sched/fair: Rewor... |
3013 |
static void __account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec) |
a9cf55b28 sched: Expire inv... |
3014 3015 |
{ /* dock delta_exec before expiring quota (as it could span periods) */ |
ec12cb7f3 sched: Accumulate... |
3016 |
cfs_rq->runtime_remaining -= delta_exec; |
a9cf55b28 sched: Expire inv... |
3017 3018 3019 |
expire_cfs_rq_runtime(cfs_rq); if (likely(cfs_rq->runtime_remaining > 0)) |
ec12cb7f3 sched: Accumulate... |
3020 |
return; |
85dac906b sched: Add suppor... |
3021 3022 3023 3024 3025 3026 |
/* * if we're unable to extend our runtime we resched so that the active * hierarchy can be throttled */ if (!assign_cfs_rq_runtime(cfs_rq) && likely(cfs_rq->curr)) resched_task(rq_of(cfs_rq)->curr); |
ec12cb7f3 sched: Accumulate... |
3027 |
} |
6c16a6dcb sched: Fix compil... |
3028 |
static __always_inline |
9dbdb1555 sched/fair: Rewor... |
3029 |
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec) |
ec12cb7f3 sched: Accumulate... |
3030 |
{ |
56f570e51 sched: Use jump l... |
3031 |
if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled) |
ec12cb7f3 sched: Accumulate... |
3032 3033 3034 3035 |
return; __account_cfs_rq_runtime(cfs_rq, delta_exec); } |
85dac906b sched: Add suppor... |
3036 3037 |
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq) { |
56f570e51 sched: Use jump l... |
3038 |
return cfs_bandwidth_used() && cfs_rq->throttled; |
85dac906b sched: Add suppor... |
3039 |
} |
64660c864 sched: Prevent in... |
3040 3041 3042 |
/* check whether cfs_rq, or any parent, is throttled */ static inline int throttled_hierarchy(struct cfs_rq *cfs_rq) { |
56f570e51 sched: Use jump l... |
3043 |
return cfs_bandwidth_used() && cfs_rq->throttle_count; |
64660c864 sched: Prevent in... |
3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 |
} /* * Ensure that neither of the group entities corresponding to src_cpu or * dest_cpu are members of a throttled hierarchy when performing group * load-balance operations. */ static inline int throttled_lb_pair(struct task_group *tg, int src_cpu, int dest_cpu) { struct cfs_rq *src_cfs_rq, *dest_cfs_rq; src_cfs_rq = tg->cfs_rq[src_cpu]; dest_cfs_rq = tg->cfs_rq[dest_cpu]; return throttled_hierarchy(src_cfs_rq) || throttled_hierarchy(dest_cfs_rq); } /* updated child weight may affect parent so we have to do this bottom up */ static int tg_unthrottle_up(struct task_group *tg, void *data) { struct rq *rq = data; struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)]; cfs_rq->throttle_count--; #ifdef CONFIG_SMP if (!cfs_rq->throttle_count) { |
f1b17280e sched: Maintain r... |
3072 |
/* adjust cfs_rq_clock_task() */ |
78becc270 sched: Use an acc... |
3073 |
cfs_rq->throttled_clock_task_time += rq_clock_task(rq) - |
f1b17280e sched: Maintain r... |
3074 |
cfs_rq->throttled_clock_task; |
64660c864 sched: Prevent in... |
3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 |
} #endif return 0; } static int tg_throttle_down(struct task_group *tg, void *data) { struct rq *rq = data; struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)]; |
82958366c sched: Replace up... |
3085 3086 |
/* group is entering throttled state, stop time */ if (!cfs_rq->throttle_count) |
78becc270 sched: Use an acc... |
3087 |
cfs_rq->throttled_clock_task = rq_clock_task(rq); |
64660c864 sched: Prevent in... |
3088 3089 3090 3091 |
cfs_rq->throttle_count++; return 0; } |
d3d9dc330 sched: Throttle e... |
3092 |
static void throttle_cfs_rq(struct cfs_rq *cfs_rq) |
85dac906b sched: Add suppor... |
3093 3094 3095 3096 3097 3098 3099 |
{ struct rq *rq = rq_of(cfs_rq); struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); struct sched_entity *se; long task_delta, dequeue = 1; se = cfs_rq->tg->se[cpu_of(rq_of(cfs_rq))]; |
f1b17280e sched: Maintain r... |
3100 |
/* freeze hierarchy runnable averages while throttled */ |
64660c864 sched: Prevent in... |
3101 3102 3103 |
rcu_read_lock(); walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq); rcu_read_unlock(); |
85dac906b sched: Add suppor... |
3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 |
task_delta = cfs_rq->h_nr_running; for_each_sched_entity(se) { struct cfs_rq *qcfs_rq = cfs_rq_of(se); /* throttled entity or throttle-on-deactivate */ if (!se->on_rq) break; if (dequeue) dequeue_entity(qcfs_rq, se, DEQUEUE_SLEEP); qcfs_rq->h_nr_running -= task_delta; if (qcfs_rq->load.weight) dequeue = 0; } if (!se) rq->nr_running -= task_delta; cfs_rq->throttled = 1; |
78becc270 sched: Use an acc... |
3124 |
cfs_rq->throttled_clock = rq_clock(rq); |
85dac906b sched: Add suppor... |
3125 3126 |
raw_spin_lock(&cfs_b->lock); list_add_tail_rcu(&cfs_rq->throttled_list, &cfs_b->throttled_cfs_rq); |
f9f9ffc23 sched: Avoid thro... |
3127 |
if (!cfs_b->timer_active) |
09dc4ab03 sched/fair: Fix t... |
3128 |
__start_cfs_bandwidth(cfs_b, false); |
85dac906b sched: Add suppor... |
3129 3130 |
raw_spin_unlock(&cfs_b->lock); } |
029632fbb sched: Make separ... |
3131 |
void unthrottle_cfs_rq(struct cfs_rq *cfs_rq) |
671fd9dab sched: Add suppor... |
3132 3133 3134 3135 3136 3137 |
{ struct rq *rq = rq_of(cfs_rq); struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); struct sched_entity *se; int enqueue = 1; long task_delta; |
22b958d8c sched: Refine the... |
3138 |
se = cfs_rq->tg->se[cpu_of(rq)]; |
671fd9dab sched: Add suppor... |
3139 3140 |
cfs_rq->throttled = 0; |
1a55af2e4 sched: Update rq ... |
3141 3142 |
update_rq_clock(rq); |
671fd9dab sched: Add suppor... |
3143 |
raw_spin_lock(&cfs_b->lock); |
78becc270 sched: Use an acc... |
3144 |
cfs_b->throttled_time += rq_clock(rq) - cfs_rq->throttled_clock; |
671fd9dab sched: Add suppor... |
3145 3146 |
list_del_rcu(&cfs_rq->throttled_list); raw_spin_unlock(&cfs_b->lock); |
64660c864 sched: Prevent in... |
3147 3148 |
/* update hierarchical throttle state */ walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq); |
671fd9dab sched: Add suppor... |
3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 |
if (!cfs_rq->load.weight) return; task_delta = cfs_rq->h_nr_running; for_each_sched_entity(se) { if (se->on_rq) enqueue = 0; cfs_rq = cfs_rq_of(se); if (enqueue) enqueue_entity(cfs_rq, se, ENQUEUE_WAKEUP); cfs_rq->h_nr_running += task_delta; if (cfs_rq_throttled(cfs_rq)) break; } if (!se) rq->nr_running += task_delta; /* determine whether we need to wake up potentially idle cpu */ if (rq->curr == rq->idle && rq->cfs.nr_running) resched_task(rq->curr); } static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining, u64 expires) { struct cfs_rq *cfs_rq; u64 runtime = remaining; rcu_read_lock(); list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq, throttled_list) { struct rq *rq = rq_of(cfs_rq); raw_spin_lock(&rq->lock); if (!cfs_rq_throttled(cfs_rq)) goto next; runtime = -cfs_rq->runtime_remaining + 1; if (runtime > remaining) runtime = remaining; remaining -= runtime; cfs_rq->runtime_remaining += runtime; cfs_rq->runtime_expires = expires; /* we check whether we're throttled above */ if (cfs_rq->runtime_remaining > 0) unthrottle_cfs_rq(cfs_rq); next: raw_spin_unlock(&rq->lock); if (!remaining) break; } rcu_read_unlock(); return remaining; } |
58088ad01 sched: Add a time... |
3211 3212 3213 3214 3215 3216 3217 3218 |
/* * Responsible for refilling a task_group's bandwidth and unthrottling its * cfs_rqs as appropriate. If there has been no activity within the last * period the timer is deactivated until scheduling resumes; cfs_b->idle is * used to track this state. */ static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun) { |
671fd9dab sched: Add suppor... |
3219 3220 |
u64 runtime, runtime_expires; int idle = 1, throttled; |
58088ad01 sched: Add a time... |
3221 3222 3223 3224 3225 |
raw_spin_lock(&cfs_b->lock); /* no need to continue the timer with no bandwidth constraint */ if (cfs_b->quota == RUNTIME_INF) goto out_unlock; |
671fd9dab sched: Add suppor... |
3226 3227 3228 |
throttled = !list_empty(&cfs_b->throttled_cfs_rq); /* idle depends on !throttled (for the case of a large deficit) */ idle = cfs_b->idle && !throttled; |
e8da1b18b sched: Add export... |
3229 |
cfs_b->nr_periods += overrun; |
671fd9dab sched: Add suppor... |
3230 |
|
a9cf55b28 sched: Expire inv... |
3231 3232 3233 |
/* if we're going inactive then everything else can be deferred */ if (idle) goto out_unlock; |
927b54fcc sched: Fix hrtime... |
3234 3235 3236 3237 3238 3239 |
/* * if we have relooped after returning idle once, we need to update our * status as actually running, so that other cpus doing * __start_cfs_bandwidth will stop trying to cancel us. */ cfs_b->timer_active = 1; |
a9cf55b28 sched: Expire inv... |
3240 |
__refill_cfs_bandwidth_runtime(cfs_b); |
671fd9dab sched: Add suppor... |
3241 3242 3243 3244 3245 |
if (!throttled) { /* mark as potentially idle for the upcoming period */ cfs_b->idle = 1; goto out_unlock; } |
e8da1b18b sched: Add export... |
3246 3247 |
/* account preceding periods in which throttling occurred */ cfs_b->nr_throttled += overrun; |
671fd9dab sched: Add suppor... |
3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 |
/* * There are throttled entities so we must first use the new bandwidth * to unthrottle them before making it generally available. This * ensures that all existing debts will be paid before a new cfs_rq is * allowed to run. */ runtime = cfs_b->runtime; runtime_expires = cfs_b->runtime_expires; cfs_b->runtime = 0; /* * This check is repeated as we are holding onto the new bandwidth * while we unthrottle. This can potentially race with an unthrottled * group trying to acquire new bandwidth from the global pool. */ while (throttled && runtime > 0) { raw_spin_unlock(&cfs_b->lock); /* we can't nest cfs_b->lock while distributing bandwidth */ runtime = distribute_cfs_runtime(cfs_b, runtime, runtime_expires); raw_spin_lock(&cfs_b->lock); throttled = !list_empty(&cfs_b->throttled_cfs_rq); } |
58088ad01 sched: Add a time... |
3272 |
|
671fd9dab sched: Add suppor... |
3273 3274 3275 3276 3277 3278 3279 3280 3281 |
/* return (any) remaining runtime */ cfs_b->runtime = runtime; /* * While we are ensured activity in the period following an * unthrottle, this also covers the case in which the new bandwidth is * insufficient to cover the existing bandwidth deficit. (Forcing the * timer to remain active while there are any throttled entities.) */ cfs_b->idle = 0; |
58088ad01 sched: Add a time... |
3282 3283 3284 3285 3286 3287 3288 |
out_unlock: if (idle) cfs_b->timer_active = 0; raw_spin_unlock(&cfs_b->lock); return idle; } |
d3d9dc330 sched: Throttle e... |
3289 |
|
d8b4986d3 sched: Return unu... |
3290 3291 3292 3293 3294 3295 |
/* a cfs_rq won't donate quota below this amount */ static const u64 min_cfs_rq_runtime = 1 * NSEC_PER_MSEC; /* minimum remaining period time to redistribute slack quota */ static const u64 min_bandwidth_expiration = 2 * NSEC_PER_MSEC; /* how long we wait to gather additional slack before distributing */ static const u64 cfs_bandwidth_slack_period = 5 * NSEC_PER_MSEC; |
db06e78cc sched: Fix cfs_ba... |
3296 3297 3298 3299 3300 3301 3302 |
/* * Are we near the end of the current quota period? * * Requires cfs_b->lock for hrtimer_expires_remaining to be safe against the * hrtimer base being cleared by __hrtimer_start_range_ns. In the case of * migrate_hrtimers, base is never cleared, so we are fine. */ |
d8b4986d3 sched: Return unu... |
3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 |
static int runtime_refresh_within(struct cfs_bandwidth *cfs_b, u64 min_expire) { struct hrtimer *refresh_timer = &cfs_b->period_timer; u64 remaining; /* if the call-back is running a quota refresh is already occurring */ if (hrtimer_callback_running(refresh_timer)) return 1; /* is a quota refresh about to occur? */ remaining = ktime_to_ns(hrtimer_expires_remaining(refresh_timer)); if (remaining < min_expire) return 1; return 0; } static void start_cfs_slack_bandwidth(struct cfs_bandwidth *cfs_b) { u64 min_left = cfs_bandwidth_slack_period + min_bandwidth_expiration; /* if there's a quota refresh soon don't bother with slack */ if (runtime_refresh_within(cfs_b, min_left)) return; start_bandwidth_timer(&cfs_b->slack_timer, ns_to_ktime(cfs_bandwidth_slack_period)); } /* we know any runtime found here is valid as update_curr() precedes return */ static void __return_cfs_rq_runtime(struct cfs_rq *cfs_rq) { struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); s64 slack_runtime = cfs_rq->runtime_remaining - min_cfs_rq_runtime; if (slack_runtime <= 0) return; raw_spin_lock(&cfs_b->lock); if (cfs_b->quota != RUNTIME_INF && cfs_rq->runtime_expires == cfs_b->runtime_expires) { cfs_b->runtime += slack_runtime; /* we are under rq->lock, defer unthrottling using a timer */ if (cfs_b->runtime > sched_cfs_bandwidth_slice() && !list_empty(&cfs_b->throttled_cfs_rq)) start_cfs_slack_bandwidth(cfs_b); } raw_spin_unlock(&cfs_b->lock); /* even if it's not valid for return we don't want to try again */ cfs_rq->runtime_remaining -= slack_runtime; } static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) { |
56f570e51 sched: Use jump l... |
3359 3360 |
if (!cfs_bandwidth_used()) return; |
fccfdc6f0 sched: Fix buglet... |
3361 |
if (!cfs_rq->runtime_enabled || cfs_rq->nr_running) |
d8b4986d3 sched: Return unu... |
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 |
return; __return_cfs_rq_runtime(cfs_rq); } /* * This is done with a timer (instead of inline with bandwidth return) since * it's necessary to juggle rq->locks to unthrottle their respective cfs_rqs. */ static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b) { u64 runtime = 0, slice = sched_cfs_bandwidth_slice(); u64 expires; /* confirm we're still not at a refresh boundary */ |
db06e78cc sched: Fix cfs_ba... |
3377 3378 3379 |
raw_spin_lock(&cfs_b->lock); if (runtime_refresh_within(cfs_b, min_bandwidth_expiration)) { raw_spin_unlock(&cfs_b->lock); |
d8b4986d3 sched: Return unu... |
3380 |
return; |
db06e78cc sched: Fix cfs_ba... |
3381 |
} |
d8b4986d3 sched: Return unu... |
3382 |
|
d8b4986d3 sched: Return unu... |
3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 |
if (cfs_b->quota != RUNTIME_INF && cfs_b->runtime > slice) { runtime = cfs_b->runtime; cfs_b->runtime = 0; } expires = cfs_b->runtime_expires; raw_spin_unlock(&cfs_b->lock); if (!runtime) return; runtime = distribute_cfs_runtime(cfs_b, runtime, expires); raw_spin_lock(&cfs_b->lock); if (expires == cfs_b->runtime_expires) cfs_b->runtime = runtime; raw_spin_unlock(&cfs_b->lock); } |
d3d9dc330 sched: Throttle e... |
3400 3401 3402 3403 3404 3405 3406 |
/* * When a group wakes up we want to make sure that its quota is not already * expired/exceeded, otherwise it may be allowed to steal additional ticks of * runtime as update_curr() throttling can not not trigger until it's on-rq. */ static void check_enqueue_throttle(struct cfs_rq *cfs_rq) { |
56f570e51 sched: Use jump l... |
3407 3408 |
if (!cfs_bandwidth_used()) return; |
d3d9dc330 sched: Throttle e... |
3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 |
/* an active group must be handled by the update_curr()->put() path */ if (!cfs_rq->runtime_enabled || cfs_rq->curr) return; /* ensure the group is not already throttled */ if (cfs_rq_throttled(cfs_rq)) return; /* update runtime allocation */ account_cfs_rq_runtime(cfs_rq, 0); if (cfs_rq->runtime_remaining <= 0) throttle_cfs_rq(cfs_rq); } /* conditionally throttle active cfs_rq's from put_prev_entity() */ |
678d5718d sched/fair: Optim... |
3424 |
static bool check_cfs_rq_runtime(struct cfs_rq *cfs_rq) |
d3d9dc330 sched: Throttle e... |
3425 |
{ |
56f570e51 sched: Use jump l... |
3426 |
if (!cfs_bandwidth_used()) |
678d5718d sched/fair: Optim... |
3427 |
return false; |
56f570e51 sched: Use jump l... |
3428 |
|
d3d9dc330 sched: Throttle e... |
3429 |
if (likely(!cfs_rq->runtime_enabled || cfs_rq->runtime_remaining > 0)) |
678d5718d sched/fair: Optim... |
3430 |
return false; |
d3d9dc330 sched: Throttle e... |
3431 3432 3433 3434 3435 3436 |
/* * it's possible for a throttled entity to be forced into a running * state (e.g. set_curr_task), in this case we're finished. */ if (cfs_rq_throttled(cfs_rq)) |
678d5718d sched/fair: Optim... |
3437 |
return true; |
d3d9dc330 sched: Throttle e... |
3438 3439 |
throttle_cfs_rq(cfs_rq); |
678d5718d sched/fair: Optim... |
3440 |
return true; |
d3d9dc330 sched: Throttle e... |
3441 |
} |
029632fbb sched: Make separ... |
3442 |
|
029632fbb sched: Make separ... |
3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 |
static enum hrtimer_restart sched_cfs_slack_timer(struct hrtimer *timer) { struct cfs_bandwidth *cfs_b = container_of(timer, struct cfs_bandwidth, slack_timer); do_sched_cfs_slack_timer(cfs_b); return HRTIMER_NORESTART; } static enum hrtimer_restart sched_cfs_period_timer(struct hrtimer *timer) { struct cfs_bandwidth *cfs_b = container_of(timer, struct cfs_bandwidth, period_timer); ktime_t now; int overrun; int idle = 0; for (;;) { now = hrtimer_cb_get_time(timer); overrun = hrtimer_forward(timer, now, cfs_b->period); if (!overrun) break; idle = do_sched_cfs_period_timer(cfs_b, overrun); } return idle ? HRTIMER_NORESTART : HRTIMER_RESTART; } void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b) { raw_spin_lock_init(&cfs_b->lock); cfs_b->runtime = 0; cfs_b->quota = RUNTIME_INF; cfs_b->period = ns_to_ktime(default_cfs_period()); INIT_LIST_HEAD(&cfs_b->throttled_cfs_rq); hrtimer_init(&cfs_b->period_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); cfs_b->period_timer.function = sched_cfs_period_timer; hrtimer_init(&cfs_b->slack_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); cfs_b->slack_timer.function = sched_cfs_slack_timer; } static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq) { cfs_rq->runtime_enabled = 0; INIT_LIST_HEAD(&cfs_rq->throttled_list); } /* requires cfs_b->lock, may release to reprogram timer */ |
09dc4ab03 sched/fair: Fix t... |
3494 |
void __start_cfs_bandwidth(struct cfs_bandwidth *cfs_b, bool force) |
029632fbb sched: Make separ... |
3495 3496 3497 3498 3499 3500 3501 |
{ /* * The timer may be active because we're trying to set a new bandwidth * period or because we're racing with the tear-down path * (timer_active==0 becomes visible before the hrtimer call-back * terminates). In either case we ensure that it's re-programmed */ |
927b54fcc sched: Fix hrtime... |
3502 3503 3504 |
while (unlikely(hrtimer_active(&cfs_b->period_timer)) && hrtimer_try_to_cancel(&cfs_b->period_timer) < 0) { /* bounce the lock to allow do_sched_cfs_period_timer to run */ |
029632fbb sched: Make separ... |
3505 |
raw_spin_unlock(&cfs_b->lock); |
927b54fcc sched: Fix hrtime... |
3506 |
cpu_relax(); |
029632fbb sched: Make separ... |
3507 3508 |
raw_spin_lock(&cfs_b->lock); /* if someone else restarted the timer then we're done */ |
09dc4ab03 sched/fair: Fix t... |
3509 |
if (!force && cfs_b->timer_active) |
029632fbb sched: Make separ... |
3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 |
return; } cfs_b->timer_active = 1; start_bandwidth_timer(&cfs_b->period_timer, cfs_b->period); } static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) { hrtimer_cancel(&cfs_b->period_timer); hrtimer_cancel(&cfs_b->slack_timer); } |
38dc3348e sched: Fix warnin... |
3522 |
static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq) |
029632fbb sched: Make separ... |
3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 |
{ struct cfs_rq *cfs_rq; for_each_leaf_cfs_rq(rq, cfs_rq) { struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg); if (!cfs_rq->runtime_enabled) continue; /* * clock_task is not advancing so we just need to make sure * there's some valid quota amount */ cfs_rq->runtime_remaining = cfs_b->quota; if (cfs_rq_throttled(cfs_rq)) unthrottle_cfs_rq(cfs_rq); } } #else /* CONFIG_CFS_BANDWIDTH */ |
f1b17280e sched: Maintain r... |
3543 3544 |
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq) { |
78becc270 sched: Use an acc... |
3545 |
return rq_clock_task(rq_of(cfs_rq)); |
f1b17280e sched: Maintain r... |
3546 |
} |
9dbdb1555 sched/fair: Rewor... |
3547 |
static void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec) {} |
678d5718d sched/fair: Optim... |
3548 |
static bool check_cfs_rq_runtime(struct cfs_rq *cfs_rq) { return false; } |
d3d9dc330 sched: Throttle e... |
3549 |
static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {} |
6c16a6dcb sched: Fix compil... |
3550 |
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {} |
85dac906b sched: Add suppor... |
3551 3552 3553 3554 3555 |
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq) { return 0; } |
64660c864 sched: Prevent in... |
3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 |
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq) { return 0; } static inline int throttled_lb_pair(struct task_group *tg, int src_cpu, int dest_cpu) { return 0; } |
029632fbb sched: Make separ... |
3567 3568 3569 3570 3571 |
void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {} #ifdef CONFIG_FAIR_GROUP_SCHED static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq) {} |
ab84d31e1 sched: Introduce ... |
3572 |
#endif |
029632fbb sched: Make separ... |
3573 3574 3575 3576 3577 |
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg) { return NULL; } static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {} |
a4c96ae31 sched: Unthrottle... |
3578 |
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {} |
029632fbb sched: Make separ... |
3579 3580 |
#endif /* CONFIG_CFS_BANDWIDTH */ |
bf0f6f24a sched: cfs core, ... |
3581 3582 3583 |
/************************************************** * CFS operations on tasks: */ |
8f4d37ec0 sched: high-res p... |
3584 3585 3586 |
#ifdef CONFIG_SCHED_HRTICK static void hrtick_start_fair(struct rq *rq, struct task_struct *p) { |
8f4d37ec0 sched: high-res p... |
3587 3588 3589 3590 |
struct sched_entity *se = &p->se; struct cfs_rq *cfs_rq = cfs_rq_of(se); WARN_ON(task_rq(p) != rq); |
b39e66eaf sched: Save some ... |
3591 |
if (cfs_rq->nr_running > 1) { |
8f4d37ec0 sched: high-res p... |
3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 |
u64 slice = sched_slice(cfs_rq, se); u64 ran = se->sum_exec_runtime - se->prev_sum_exec_runtime; s64 delta = slice - ran; if (delta < 0) { if (rq->curr == p) resched_task(p); return; } /* * Don't schedule slices shorter than 10000ns, that just * doesn't make sense. Rely on vruntime for fairness. */ |
31656519e sched, x86: clean... |
3606 |
if (rq->curr != p) |
157124c11 sched: fix warnin... |
3607 |
delta = max_t(s64, 10000LL, delta); |
8f4d37ec0 sched: high-res p... |
3608 |
|
31656519e sched, x86: clean... |
3609 |
hrtick_start(rq, delta); |
8f4d37ec0 sched: high-res p... |
3610 3611 |
} } |
a4c2f00f5 sched: fair sched... |
3612 3613 3614 3615 3616 3617 3618 3619 3620 |
/* * called from enqueue/dequeue and updates the hrtick when the * current task is from our class and nr_running is low enough * to matter. */ static void hrtick_update(struct rq *rq) { struct task_struct *curr = rq->curr; |
b39e66eaf sched: Save some ... |
3621 |
if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class) |
a4c2f00f5 sched: fair sched... |
3622 3623 3624 3625 3626 |
return; if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency) hrtick_start_fair(rq, curr); } |
55e12e5e7 sched: make sched... |
3627 |
#else /* !CONFIG_SCHED_HRTICK */ |
8f4d37ec0 sched: high-res p... |
3628 3629 3630 3631 |
static inline void hrtick_start_fair(struct rq *rq, struct task_struct *p) { } |
a4c2f00f5 sched: fair sched... |
3632 3633 3634 3635 |
static inline void hrtick_update(struct rq *rq) { } |
8f4d37ec0 sched: high-res p... |
3636 |
#endif |
bf0f6f24a sched: cfs core, ... |
3637 3638 3639 3640 3641 |
/* * The enqueue_task method is called before nr_running is * increased. Here we update the fair scheduling stats and * then put the task into the rbtree: */ |
ea87bb785 sched: Extend enq... |
3642 |
static void |
371fd7e7a sched: Add enqueu... |
3643 |
enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) |
bf0f6f24a sched: cfs core, ... |
3644 3645 |
{ struct cfs_rq *cfs_rq; |
62fb18513 sched: revert loa... |
3646 |
struct sched_entity *se = &p->se; |
bf0f6f24a sched: cfs core, ... |
3647 3648 |
for_each_sched_entity(se) { |
62fb18513 sched: revert loa... |
3649 |
if (se->on_rq) |
bf0f6f24a sched: cfs core, ... |
3650 3651 |
break; cfs_rq = cfs_rq_of(se); |
88ec22d3e sched: Remove the... |
3652 |
enqueue_entity(cfs_rq, se, flags); |
85dac906b sched: Add suppor... |
3653 3654 3655 3656 3657 3658 3659 3660 3661 |
/* * end evaluation on encountering a throttled cfs_rq * * note: in the case of encountering a throttled cfs_rq we will * post the final h_nr_running increment below. */ if (cfs_rq_throttled(cfs_rq)) break; |
953bfcd10 sched: Implement ... |
3662 |
cfs_rq->h_nr_running++; |
85dac906b sched: Add suppor... |
3663 |
|
88ec22d3e sched: Remove the... |
3664 |
flags = ENQUEUE_WAKEUP; |
bf0f6f24a sched: cfs core, ... |
3665 |
} |
8f4d37ec0 sched: high-res p... |
3666 |
|
2069dd75c sched: Rewrite tg... |
3667 |
for_each_sched_entity(se) { |
0f3171438 sched: Cleanup du... |
3668 |
cfs_rq = cfs_rq_of(se); |
953bfcd10 sched: Implement ... |
3669 |
cfs_rq->h_nr_running++; |
2069dd75c sched: Rewrite tg... |
3670 |
|
85dac906b sched: Add suppor... |
3671 3672 |
if (cfs_rq_throttled(cfs_rq)) break; |
17bc14b76 Revert "sched: Up... |
3673 |
update_cfs_shares(cfs_rq); |
9ee474f55 sched: Maintain t... |
3674 |
update_entity_load_avg(se, 1); |
2069dd75c sched: Rewrite tg... |
3675 |
} |
18bf2805d sched: Maintain p... |
3676 3677 |
if (!se) { update_rq_runnable_avg(rq, rq->nr_running); |
85dac906b sched: Add suppor... |
3678 |
inc_nr_running(rq); |
18bf2805d sched: Maintain p... |
3679 |
} |
a4c2f00f5 sched: fair sched... |
3680 |
hrtick_update(rq); |
bf0f6f24a sched: cfs core, ... |
3681 |
} |
2f36825b1 sched: Next buddy... |
3682 |
static void set_next_buddy(struct sched_entity *se); |
bf0f6f24a sched: cfs core, ... |
3683 3684 3685 3686 3687 |
/* * The dequeue_task method is called before nr_running is * decreased. We remove the task from the rbtree and * update the fair scheduling stats: */ |
371fd7e7a sched: Add enqueu... |
3688 |
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) |
bf0f6f24a sched: cfs core, ... |
3689 3690 |
{ struct cfs_rq *cfs_rq; |
62fb18513 sched: revert loa... |
3691 |
struct sched_entity *se = &p->se; |
2f36825b1 sched: Next buddy... |
3692 |
int task_sleep = flags & DEQUEUE_SLEEP; |
bf0f6f24a sched: cfs core, ... |
3693 3694 3695 |
for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); |
371fd7e7a sched: Add enqueu... |
3696 |
dequeue_entity(cfs_rq, se, flags); |
85dac906b sched: Add suppor... |
3697 3698 3699 3700 3701 3702 3703 3704 3705 |
/* * end evaluation on encountering a throttled cfs_rq * * note: in the case of encountering a throttled cfs_rq we will * post the final h_nr_running decrement below. */ if (cfs_rq_throttled(cfs_rq)) break; |
953bfcd10 sched: Implement ... |
3706 |
cfs_rq->h_nr_running--; |
2069dd75c sched: Rewrite tg... |
3707 |
|
bf0f6f24a sched: cfs core, ... |
3708 |
/* Don't dequeue parent if it has other entities besides us */ |
2f36825b1 sched: Next buddy... |
3709 3710 3711 3712 3713 3714 3715 |
if (cfs_rq->load.weight) { /* * Bias pick_next to pick a task from this cfs_rq, as * p is sleeping when it is within its sched_slice. */ if (task_sleep && parent_entity(se)) set_next_buddy(parent_entity(se)); |
9598c82dc sched: Don't upda... |
3716 3717 3718 |
/* avoid re-evaluating load for this entity */ se = parent_entity(se); |
bf0f6f24a sched: cfs core, ... |
3719 |
break; |
2f36825b1 sched: Next buddy... |
3720 |
} |
371fd7e7a sched: Add enqueu... |
3721 |
flags |= DEQUEUE_SLEEP; |
bf0f6f24a sched: cfs core, ... |
3722 |
} |
8f4d37ec0 sched: high-res p... |
3723 |
|
2069dd75c sched: Rewrite tg... |
3724 |
for_each_sched_entity(se) { |
0f3171438 sched: Cleanup du... |
3725 |
cfs_rq = cfs_rq_of(se); |
953bfcd10 sched: Implement ... |
3726 |
cfs_rq->h_nr_running--; |
2069dd75c sched: Rewrite tg... |
3727 |
|
85dac906b sched: Add suppor... |
3728 3729 |
if (cfs_rq_throttled(cfs_rq)) break; |
17bc14b76 Revert "sched: Up... |
3730 |
update_cfs_shares(cfs_rq); |
9ee474f55 sched: Maintain t... |
3731 |
update_entity_load_avg(se, 1); |
2069dd75c sched: Rewrite tg... |
3732 |
} |
18bf2805d sched: Maintain p... |
3733 |
if (!se) { |
85dac906b sched: Add suppor... |
3734 |
dec_nr_running(rq); |
18bf2805d sched: Maintain p... |
3735 3736 |
update_rq_runnable_avg(rq, 1); } |
a4c2f00f5 sched: fair sched... |
3737 |
hrtick_update(rq); |
bf0f6f24a sched: cfs core, ... |
3738 |
} |
e7693a362 sched: de-SCHED_O... |
3739 |
#ifdef CONFIG_SMP |
029632fbb sched: Make separ... |
3740 3741 3742 |
/* Used instead of source_load when we know the type == 0 */ static unsigned long weighted_cpuload(const int cpu) { |
b92486cbf sched: Compute ru... |
3743 |
return cpu_rq(cpu)->cfs.runnable_load_avg; |
029632fbb sched: Make separ... |
3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 |
} /* * Return a low guess at the load of a migration-source cpu weighted * according to the scheduling class and "nice" value. * * We want to under-estimate the load of migration sources, to * balance conservatively. */ static unsigned long source_load(int cpu, int type) { struct rq *rq = cpu_rq(cpu); unsigned long total = weighted_cpuload(cpu); if (type == 0 || !sched_feat(LB_BIAS)) return total; return min(rq->cpu_load[type-1], total); } /* * Return a high guess at the load of a migration-target cpu weighted * according to the scheduling class and "nice" value. */ static unsigned long target_load(int cpu, int type) { struct rq *rq = cpu_rq(cpu); unsigned long total = weighted_cpuload(cpu); if (type == 0 || !sched_feat(LB_BIAS)) return total; return max(rq->cpu_load[type-1], total); } static unsigned long power_of(int cpu) { return cpu_rq(cpu)->cpu_power; } static unsigned long cpu_avg_load_per_task(int cpu) { struct rq *rq = cpu_rq(cpu); unsigned long nr_running = ACCESS_ONCE(rq->nr_running); |
b92486cbf sched: Compute ru... |
3788 |
unsigned long load_avg = rq->cfs.runnable_load_avg; |
029632fbb sched: Make separ... |
3789 3790 |
if (nr_running) |
b92486cbf sched: Compute ru... |
3791 |
return load_avg / nr_running; |
029632fbb sched: Make separ... |
3792 3793 3794 |
return 0; } |
62470419e sched: Implement ... |
3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 |
static void record_wakee(struct task_struct *p) { /* * Rough decay (wiping) for cost saving, don't worry * about the boundary, really active task won't care * about the loss. */ if (jiffies > current->wakee_flip_decay_ts + HZ) { current->wakee_flips = 0; current->wakee_flip_decay_ts = jiffies; } if (current->last_wakee != p) { current->last_wakee = p; current->wakee_flips++; } } |
098fb9db2 sched: clean up w... |
3812 |
|
74f8e4b23 sched: Remove rq ... |
3813 |
static void task_waking_fair(struct task_struct *p) |
88ec22d3e sched: Remove the... |
3814 3815 3816 |
{ struct sched_entity *se = &p->se; struct cfs_rq *cfs_rq = cfs_rq_of(se); |
3fe1698b7 sched: Deal with ... |
3817 3818 3819 3820 |
u64 min_vruntime; #ifndef CONFIG_64BIT u64 min_vruntime_copy; |
88ec22d3e sched: Remove the... |
3821 |
|
3fe1698b7 sched: Deal with ... |
3822 3823 3824 3825 3826 3827 3828 3829 |
do { min_vruntime_copy = cfs_rq->min_vruntime_copy; smp_rmb(); min_vruntime = cfs_rq->min_vruntime; } while (min_vruntime != min_vruntime_copy); #else min_vruntime = cfs_rq->min_vruntime; #endif |
88ec22d3e sched: Remove the... |
3830 |
|
3fe1698b7 sched: Deal with ... |
3831 |
se->vruntime -= min_vruntime; |
62470419e sched: Implement ... |
3832 |
record_wakee(p); |
88ec22d3e sched: Remove the... |
3833 |
} |
bb3469ac9 sched: hierarchic... |
3834 |
#ifdef CONFIG_FAIR_GROUP_SCHED |
f5bfb7d9f sched: bias effec... |
3835 3836 3837 3838 3839 3840 |
/* * effective_load() calculates the load change as seen from the root_task_group * * Adding load to a group doesn't make a group heavier, but can cause movement * of group shares between cpus. Assuming the shares were perfectly aligned one * can calculate the shift in shares. |
cf5f0acf3 sched: Add a comm... |
3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 |
* * Calculate the effective load difference if @wl is added (subtracted) to @tg * on this @cpu and results in a total addition (subtraction) of @wg to the * total group weight. * * Given a runqueue weight distribution (rw_i) we can compute a shares * distribution (s_i) using: * * s_i = rw_i / \Sum rw_j (1) * * Suppose we have 4 CPUs and our @tg is a direct child of the root group and * has 7 equal weight tasks, distributed as below (rw_i), with the resulting * shares distribution (s_i): * * rw_i = { 2, 4, 1, 0 } * s_i = { 2/7, 4/7, 1/7, 0 } * * As per wake_affine() we're interested in the load of two CPUs (the CPU the * task used to run on and the CPU the waker is running on), we need to * compute the effect of waking a task on either CPU and, in case of a sync * wakeup, compute the effect of the current task going to sleep. * * So for a change of @wl to the local @cpu with an overall group weight change * of @wl we can compute the new shares distribution (s'_i) using: * * s'_i = (rw_i + @wl) / (@wg + \Sum rw_j) (2) * * Suppose we're interested in CPUs 0 and 1, and want to compute the load * differences in waking a task to CPU 0. The additional task changes the * weight and shares distributions like: * * rw'_i = { 3, 4, 1, 0 } * s'_i = { 3/8, 4/8, 1/8, 0 } * * We can then compute the difference in effective weight by using: * * dw_i = S * (s'_i - s_i) (3) * * Where 'S' is the group weight as seen by its parent. * * Therefore the effective change in loads on CPU 0 would be 5/56 (3/8 - 2/7) * times the weight of the group. The effect on CPU 1 would be -4/56 (4/8 - * 4/7) times the weight of the group. |
f5bfb7d9f sched: bias effec... |
3884 |
*/ |
2069dd75c sched: Rewrite tg... |
3885 |
static long effective_load(struct task_group *tg, int cpu, long wl, long wg) |
bb3469ac9 sched: hierarchic... |
3886 |
{ |
4be9daaa1 sched: fix task_h... |
3887 |
struct sched_entity *se = tg->se[cpu]; |
f1d239f73 sched: incrementa... |
3888 |
|
9722c2dac sched: Calculate ... |
3889 |
if (!tg->parent) /* the trivial, non-cgroup case */ |
f1d239f73 sched: incrementa... |
3890 |
return wl; |
4be9daaa1 sched: fix task_h... |
3891 |
for_each_sched_entity(se) { |
cf5f0acf3 sched: Add a comm... |
3892 |
long w, W; |
4be9daaa1 sched: fix task_h... |
3893 |
|
977dda7c9 sched: Update eff... |
3894 |
tg = se->my_q->tg; |
bb3469ac9 sched: hierarchic... |
3895 |
|
cf5f0acf3 sched: Add a comm... |
3896 3897 3898 3899 |
/* * W = @wg + \Sum rw_j */ W = wg + calc_tg_weight(tg, se->my_q); |
4be9daaa1 sched: fix task_h... |
3900 |
|
cf5f0acf3 sched: Add a comm... |
3901 3902 3903 3904 |
/* * w = rw_i + @wl */ w = se->my_q->load.weight + wl; |
940959e93 sched: fixlet for... |
3905 |
|
cf5f0acf3 sched: Add a comm... |
3906 3907 3908 3909 3910 |
/* * wl = S * s'_i; see (2) */ if (W > 0 && w < W) wl = (w * tg->shares) / W; |
977dda7c9 sched: Update eff... |
3911 3912 |
else wl = tg->shares; |
940959e93 sched: fixlet for... |
3913 |
|
cf5f0acf3 sched: Add a comm... |
3914 3915 3916 3917 3918 |
/* * Per the above, wl is the new se->load.weight value; since * those are clipped to [MIN_SHARES, ...) do so now. See * calc_cfs_shares(). */ |
977dda7c9 sched: Update eff... |
3919 3920 |
if (wl < MIN_SHARES) wl = MIN_SHARES; |
cf5f0acf3 sched: Add a comm... |
3921 3922 3923 3924 |
/* * wl = dw_i = S * (s'_i - s_i); see (3) */ |
977dda7c9 sched: Update eff... |
3925 |
wl -= se->load.weight; |
cf5f0acf3 sched: Add a comm... |
3926 3927 3928 3929 3930 3931 3932 3933 |
/* * Recursively apply this logic to all parent groups to compute * the final effective load change on the root group. Since * only the @tg group gets extra weight, all parent groups can * only redistribute existing shares. @wl is the shift in shares * resulting from this level per the above. */ |
4be9daaa1 sched: fix task_h... |
3934 |
wg = 0; |
4be9daaa1 sched: fix task_h... |
3935 |
} |
bb3469ac9 sched: hierarchic... |
3936 |
|
4be9daaa1 sched: fix task_h... |
3937 |
return wl; |
bb3469ac9 sched: hierarchic... |
3938 3939 |
} #else |
4be9daaa1 sched: fix task_h... |
3940 |
|
58d081b50 sched/numa: Avoid... |
3941 |
static long effective_load(struct task_group *tg, int cpu, long wl, long wg) |
4be9daaa1 sched: fix task_h... |
3942 |
{ |
83378269a sched: correct wa... |
3943 |
return wl; |
bb3469ac9 sched: hierarchic... |
3944 |
} |
4be9daaa1 sched: fix task_h... |
3945 |
|
bb3469ac9 sched: hierarchic... |
3946 |
#endif |
62470419e sched: Implement ... |
3947 3948 |
static int wake_wide(struct task_struct *p) { |
7d9ffa896 sched: Micro-opti... |
3949 |
int factor = this_cpu_read(sd_llc_size); |
62470419e sched: Implement ... |
3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 |
/* * Yeah, it's the switching-frequency, could means many wakee or * rapidly switch, use factor here will just help to automatically * adjust the loose-degree, so bigger node will lead to more pull. */ if (p->wakee_flips > factor) { /* * wakee is somewhat hot, it needs certain amount of cpu * resource, so if waker is far more hot, prefer to leave * it alone. */ if (current->wakee_flips > (factor * p->wakee_flips)) return 1; } return 0; } |
c88d59108 sched: Merge sele... |
3968 |
static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) |
098fb9db2 sched: clean up w... |
3969 |
{ |
e37b6a7b2 sched: Fix sign u... |
3970 |
s64 this_load, load; |
c88d59108 sched: Merge sele... |
3971 |
int idx, this_cpu, prev_cpu; |
098fb9db2 sched: clean up w... |
3972 |
unsigned long tl_per_task; |
c88d59108 sched: Merge sele... |
3973 |
struct task_group *tg; |
83378269a sched: correct wa... |
3974 |
unsigned long weight; |
b3137bc8e sched: stop wake_... |
3975 |
int balanced; |
098fb9db2 sched: clean up w... |
3976 |
|
62470419e sched: Implement ... |
3977 3978 3979 3980 3981 3982 |
/* * If we wake multiple tasks be careful to not bounce * ourselves around too much. */ if (wake_wide(p)) return 0; |
c88d59108 sched: Merge sele... |
3983 3984 3985 3986 3987 |
idx = sd->wake_idx; this_cpu = smp_processor_id(); prev_cpu = task_cpu(p); load = source_load(prev_cpu, idx); this_load = target_load(this_cpu, idx); |
098fb9db2 sched: clean up w... |
3988 3989 |
/* |
b3137bc8e sched: stop wake_... |
3990 3991 3992 3993 |
* If sync wakeup then subtract the (maximum possible) * effect of the currently running task from the load * of the current CPU: */ |
83378269a sched: correct wa... |
3994 3995 3996 |
if (sync) { tg = task_group(current); weight = current->se.load.weight; |
c88d59108 sched: Merge sele... |
3997 |
this_load += effective_load(tg, this_cpu, -weight, -weight); |
83378269a sched: correct wa... |
3998 3999 |
load += effective_load(tg, prev_cpu, 0, -weight); } |
b3137bc8e sched: stop wake_... |
4000 |
|
83378269a sched: correct wa... |
4001 4002 |
tg = task_group(p); weight = p->se.load.weight; |
b3137bc8e sched: stop wake_... |
4003 |
|
71a29aa7b sched: Deal with ... |
4004 4005 |
/* * In low-load situations, where prev_cpu is idle and this_cpu is idle |
c88d59108 sched: Merge sele... |
4006 4007 4008 |
* due to the sync cause above having dropped this_load to 0, we'll * always have an imbalance, but there's really nothing you can do * about that, so that's good too. |
71a29aa7b sched: Deal with ... |
4009 4010 4011 4012 |
* * Otherwise check if either cpus are near enough in load to allow this * task to be woken on this_cpu. */ |
e37b6a7b2 sched: Fix sign u... |
4013 4014 |
if (this_load > 0) { s64 this_eff_load, prev_eff_load; |
e51fd5e22 sched: Fix wake_a... |
4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 |
this_eff_load = 100; this_eff_load *= power_of(prev_cpu); this_eff_load *= this_load + effective_load(tg, this_cpu, weight, weight); prev_eff_load = 100 + (sd->imbalance_pct - 100) / 2; prev_eff_load *= power_of(this_cpu); prev_eff_load *= load + effective_load(tg, prev_cpu, 0, weight); balanced = this_eff_load <= prev_eff_load; } else balanced = true; |
b3137bc8e sched: stop wake_... |
4028 4029 |
/* |
4ae7d5cef sched: improve af... |
4030 4031 4032 |
* If the currently running task will sleep within * a reasonable amount of time then attract this newly * woken task: |
098fb9db2 sched: clean up w... |
4033 |
*/ |
2fb7635c4 sched: sync wakeu... |
4034 4035 |
if (sync && balanced) return 1; |
098fb9db2 sched: clean up w... |
4036 |
|
41acab885 sched: Implement ... |
4037 |
schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts); |
098fb9db2 sched: clean up w... |
4038 |
tl_per_task = cpu_avg_load_per_task(this_cpu); |
c88d59108 sched: Merge sele... |
4039 4040 4041 |
if (balanced || (this_load <= load && this_load + target_load(prev_cpu, idx) <= tl_per_task)) { |
098fb9db2 sched: clean up w... |
4042 4043 4044 4045 4046 |
/* * This domain has SD_WAKE_AFFINE and * p is cache cold in this domain, and * there is no bad imbalance. */ |
c88d59108 sched: Merge sele... |
4047 |
schedstat_inc(sd, ttwu_move_affine); |
41acab885 sched: Implement ... |
4048 |
schedstat_inc(p, se.statistics.nr_wakeups_affine); |
098fb9db2 sched: clean up w... |
4049 4050 4051 4052 4053 |
return 1; } return 0; } |
aaee1203c sched: Move sched... |
4054 4055 4056 4057 4058 |
/* * find_idlest_group finds and returns the least busy CPU group within the * domain. */ static struct sched_group * |
78e7ed53c sched: Tweak wake... |
4059 |
find_idlest_group(struct sched_domain *sd, struct task_struct *p, |
c44f2a020 sched/fair: Move ... |
4060 |
int this_cpu, int sd_flag) |
e7693a362 sched: de-SCHED_O... |
4061 |
{ |
b3bd3de66 gcc-4.6: kernel/*... |
4062 |
struct sched_group *idlest = NULL, *group = sd->groups; |
aaee1203c sched: Move sched... |
4063 |
unsigned long min_load = ULONG_MAX, this_load = 0; |
c44f2a020 sched/fair: Move ... |
4064 |
int load_idx = sd->forkexec_idx; |
aaee1203c sched: Move sched... |
4065 |
int imbalance = 100 + (sd->imbalance_pct-100)/2; |
e7693a362 sched: de-SCHED_O... |
4066 |
|
c44f2a020 sched/fair: Move ... |
4067 4068 |
if (sd_flag & SD_BALANCE_WAKE) load_idx = sd->wake_idx; |
aaee1203c sched: Move sched... |
4069 4070 4071 4072 |
do { unsigned long load, avg_load; int local_group; int i; |
e7693a362 sched: de-SCHED_O... |
4073 |
|
aaee1203c sched: Move sched... |
4074 4075 |
/* Skip over this group if it has no CPUs allowed */ if (!cpumask_intersects(sched_group_cpus(group), |
fa17b507f sched: Wrap sched... |
4076 |
tsk_cpus_allowed(p))) |
aaee1203c sched: Move sched... |
4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 |
continue; local_group = cpumask_test_cpu(this_cpu, sched_group_cpus(group)); /* Tally up the load of all CPUs in the group */ avg_load = 0; for_each_cpu(i, sched_group_cpus(group)) { /* Bias balancing toward cpus of our domain */ if (local_group) load = source_load(i, load_idx); else load = target_load(i, load_idx); avg_load += load; } /* Adjust by relative CPU power of the group */ |
9c3f75cbd sched: Break out ... |
4096 |
avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power; |
aaee1203c sched: Move sched... |
4097 4098 4099 |
if (local_group) { this_load = avg_load; |
aaee1203c sched: Move sched... |
4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 |
} else if (avg_load < min_load) { min_load = avg_load; idlest = group; } } while (group = group->next, group != sd->groups); if (!idlest || 100*this_load < imbalance*min_load) return NULL; return idlest; } /* * find_idlest_cpu - find the idlest cpu among the cpus in group. */ static int find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu) { unsigned long load, min_load = ULONG_MAX; int idlest = -1; int i; /* Traverse only the allowed CPUs */ |
fa17b507f sched: Wrap sched... |
4122 |
for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) { |
aaee1203c sched: Move sched... |
4123 4124 4125 4126 4127 |
load = weighted_cpuload(i); if (load < min_load || (load == min_load && i == this_cpu)) { min_load = load; idlest = i; |
e7693a362 sched: de-SCHED_O... |
4128 4129 |
} } |
aaee1203c sched: Move sched... |
4130 4131 |
return idlest; } |
e7693a362 sched: de-SCHED_O... |
4132 |
|
aaee1203c sched: Move sched... |
4133 |
/* |
a50bde513 sched: Cleanup se... |
4134 4135 |
* Try and locate an idle CPU in the sched_domain. */ |
99bd5e2f2 sched: Fix select... |
4136 |
static int select_idle_sibling(struct task_struct *p, int target) |
a50bde513 sched: Cleanup se... |
4137 |
{ |
99bd5e2f2 sched: Fix select... |
4138 |
struct sched_domain *sd; |
37407ea7f Revert "sched: Im... |
4139 |
struct sched_group *sg; |
e0a79f529 sched: Fix select... |
4140 |
int i = task_cpu(p); |
a50bde513 sched: Cleanup se... |
4141 |
|
e0a79f529 sched: Fix select... |
4142 4143 |
if (idle_cpu(target)) return target; |
99bd5e2f2 sched: Fix select... |
4144 4145 |
/* |
e0a79f529 sched: Fix select... |
4146 |
* If the prevous cpu is cache affine and idle, don't be stupid. |
99bd5e2f2 sched: Fix select... |
4147 |
*/ |
e0a79f529 sched: Fix select... |
4148 4149 |
if (i != target && cpus_share_cache(i, target) && idle_cpu(i)) return i; |
a50bde513 sched: Cleanup se... |
4150 4151 |
/* |
37407ea7f Revert "sched: Im... |
4152 |
* Otherwise, iterate the domains and find an elegible idle cpu. |
a50bde513 sched: Cleanup se... |
4153 |
*/ |
518cd6234 sched: Only queue... |
4154 |
sd = rcu_dereference(per_cpu(sd_llc, target)); |
970e17898 sched: Improve sc... |
4155 |
for_each_lower_domain(sd) { |
37407ea7f Revert "sched: Im... |
4156 4157 4158 4159 4160 4161 4162 |
sg = sd->groups; do { if (!cpumask_intersects(sched_group_cpus(sg), tsk_cpus_allowed(p))) goto next; for_each_cpu(i, sched_group_cpus(sg)) { |
e0a79f529 sched: Fix select... |
4163 |
if (i == target || !idle_cpu(i)) |
37407ea7f Revert "sched: Im... |
4164 4165 |
goto next; } |
970e17898 sched: Improve sc... |
4166 |
|
37407ea7f Revert "sched: Im... |
4167 4168 4169 4170 4171 4172 4173 4174 |
target = cpumask_first_and(sched_group_cpus(sg), tsk_cpus_allowed(p)); goto done; next: sg = sg->next; } while (sg != sd->groups); } done: |
a50bde513 sched: Cleanup se... |
4175 4176 4177 4178 |
return target; } /* |
de91b9cb9 sched: Fix select... |
4179 4180 4181 |
* select_task_rq_fair: Select target runqueue for the waking task in domains * that have the 'sd_flag' flag set. In practice, this is SD_BALANCE_WAKE, * SD_BALANCE_FORK, or SD_BALANCE_EXEC. |
aaee1203c sched: Move sched... |
4182 |
* |
de91b9cb9 sched: Fix select... |
4183 4184 |
* Balances load by selecting the idlest cpu in the idlest group, or under * certain conditions an idle sibling cpu if the domain has SD_WAKE_AFFINE set. |
aaee1203c sched: Move sched... |
4185 |
* |
de91b9cb9 sched: Fix select... |
4186 |
* Returns the target cpu number. |
aaee1203c sched: Move sched... |
4187 4188 4189 |
* * preempt must be disabled. */ |
0017d7350 sched: Fix TASK_W... |
4190 |
static int |
ac66f5477 sched/numa: Intro... |
4191 |
select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_flags) |
aaee1203c sched: Move sched... |
4192 |
{ |
29cd8bae3 sched: Fix SD_POW... |
4193 |
struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL; |
c88d59108 sched: Merge sele... |
4194 |
int cpu = smp_processor_id(); |
c88d59108 sched: Merge sele... |
4195 |
int new_cpu = cpu; |
99bd5e2f2 sched: Fix select... |
4196 |
int want_affine = 0; |
5158f4e44 sched: Clean up t... |
4197 |
int sync = wake_flags & WF_SYNC; |
c88d59108 sched: Merge sele... |
4198 |
|
29baa7478 sched: Move nr_cp... |
4199 |
if (p->nr_cpus_allowed == 1) |
76854c7e8 sched: Use rt.nr_... |
4200 |
return prev_cpu; |
0763a660a sched: Rename sel... |
4201 |
if (sd_flag & SD_BALANCE_WAKE) { |
fa17b507f sched: Wrap sched... |
4202 |
if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) |
c88d59108 sched: Merge sele... |
4203 4204 4205 |
want_affine = 1; new_cpu = prev_cpu; } |
aaee1203c sched: Move sched... |
4206 |
|
dce840a08 sched: Dynamicall... |
4207 |
rcu_read_lock(); |
aaee1203c sched: Move sched... |
4208 |
for_each_domain(cpu, tmp) { |
e4f428884 sched: Select_tas... |
4209 4210 |
if (!(tmp->flags & SD_LOAD_BALANCE)) continue; |
aaee1203c sched: Move sched... |
4211 |
/* |
99bd5e2f2 sched: Fix select... |
4212 4213 |
* If both cpu and prev_cpu are part of this domain, * cpu is a valid SD_WAKE_AFFINE target. |
fe3bcfe1f sched: More gener... |
4214 |
*/ |
99bd5e2f2 sched: Fix select... |
4215 4216 4217 |
if (want_affine && (tmp->flags & SD_WAKE_AFFINE) && cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) { affine_sd = tmp; |
29cd8bae3 sched: Fix SD_POW... |
4218 |
break; |
f03542a70 sched: recover SD... |
4219 |
} |
29cd8bae3 sched: Fix SD_POW... |
4220 |
|
f03542a70 sched: recover SD... |
4221 |
if (tmp->flags & sd_flag) |
29cd8bae3 sched: Fix SD_POW... |
4222 4223 |
sd = tmp; } |
8b911acdf sched: Fix select... |
4224 |
if (affine_sd) { |
f03542a70 sched: recover SD... |
4225 |
if (cpu != prev_cpu && wake_affine(affine_sd, p, sync)) |
dce840a08 sched: Dynamicall... |
4226 4227 4228 4229 |
prev_cpu = cpu; new_cpu = select_idle_sibling(p, prev_cpu); goto unlock; |
8b911acdf sched: Fix select... |
4230 |
} |
e7693a362 sched: de-SCHED_O... |
4231 |
|
aaee1203c sched: Move sched... |
4232 4233 |
while (sd) { struct sched_group *group; |
c88d59108 sched: Merge sele... |
4234 |
int weight; |
098fb9db2 sched: clean up w... |
4235 |
|
0763a660a sched: Rename sel... |
4236 |
if (!(sd->flags & sd_flag)) { |
aaee1203c sched: Move sched... |
4237 4238 4239 |
sd = sd->child; continue; } |
098fb9db2 sched: clean up w... |
4240 |
|
c44f2a020 sched/fair: Move ... |
4241 |
group = find_idlest_group(sd, p, cpu, sd_flag); |
aaee1203c sched: Move sched... |
4242 4243 4244 4245 |
if (!group) { sd = sd->child; continue; } |
4ae7d5cef sched: improve af... |
4246 |
|
d7c33c493 sched: Fix task a... |
4247 |
new_cpu = find_idlest_cpu(group, p, cpu); |
aaee1203c sched: Move sched... |
4248 4249 4250 4251 |
if (new_cpu == -1 || new_cpu == cpu) { /* Now try balancing at a lower domain level of cpu */ sd = sd->child; continue; |
e7693a362 sched: de-SCHED_O... |
4252 |
} |
aaee1203c sched: Move sched... |
4253 4254 4255 |
/* Now try balancing at a lower domain level of new_cpu */ cpu = new_cpu; |
669c55e9f sched: Pre-comput... |
4256 |
weight = sd->span_weight; |
aaee1203c sched: Move sched... |
4257 4258 |
sd = NULL; for_each_domain(cpu, tmp) { |
669c55e9f sched: Pre-comput... |
4259 |
if (weight <= tmp->span_weight) |
aaee1203c sched: Move sched... |
4260 |
break; |
0763a660a sched: Rename sel... |
4261 |
if (tmp->flags & sd_flag) |
aaee1203c sched: Move sched... |
4262 4263 4264 |
sd = tmp; } /* while loop will break here if sd == NULL */ |
e7693a362 sched: de-SCHED_O... |
4265 |
} |
dce840a08 sched: Dynamicall... |
4266 4267 |
unlock: rcu_read_unlock(); |
e7693a362 sched: de-SCHED_O... |
4268 |
|
c88d59108 sched: Merge sele... |
4269 |
return new_cpu; |
e7693a362 sched: de-SCHED_O... |
4270 |
} |
0a74bef8b sched: Add an rq ... |
4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 |
/* * Called immediately before a task is migrated to a new cpu; task_cpu(p) and * cfs_rq_of(p) references at time of call are still valid and identify the * previous cpu. However, the caller only guarantees p->pi_lock is held; no * other assumptions, including the state of rq->lock, should be made. */ static void migrate_task_rq_fair(struct task_struct *p, int next_cpu) { |
aff3e4988 sched: Account fo... |
4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 |
struct sched_entity *se = &p->se; struct cfs_rq *cfs_rq = cfs_rq_of(se); /* * Load tracking: accumulate removed load so that it can be processed * when we next update owning cfs_rq under rq->lock. Tasks contribute * to blocked load iff they have a positive decay-count. It can never * be negative here since on-rq tasks have decay-count == 0. */ if (se->avg.decay_count) { se->avg.decay_count = -__synchronize_entity_decay(se); |
2509940fd sched/cfs_rq: Cha... |
4292 4293 |
atomic_long_add(se->avg.load_avg_contrib, &cfs_rq->removed_load); |
aff3e4988 sched: Account fo... |
4294 |
} |
0a74bef8b sched: Add an rq ... |
4295 |
} |
e7693a362 sched: de-SCHED_O... |
4296 |
#endif /* CONFIG_SMP */ |
e52fb7c09 sched: prefer wakers |
4297 4298 |
static unsigned long wakeup_gran(struct sched_entity *curr, struct sched_entity *se) |
0bbd3336e sched: fix wakeup... |
4299 4300 4301 4302 |
{ unsigned long gran = sysctl_sched_wakeup_granularity; /* |
e52fb7c09 sched: prefer wakers |
4303 4304 |
* Since its curr running now, convert the gran from real-time * to virtual-time in his units. |
13814d42e sched: Remove ASY... |
4305 4306 4307 4308 4309 4310 4311 4312 4313 |
* * By using 'se' instead of 'curr' we penalize light tasks, so * they get preempted easier. That is, if 'se' < 'curr' then * the resulting gran will be larger, therefore penalizing the * lighter, if otoh 'se' > 'curr' then the resulting gran will * be smaller, again penalizing the lighter task. * * This is especially important for buddies when the leftmost * task is higher priority than the buddy. |
0bbd3336e sched: fix wakeup... |
4314 |
*/ |
f4ad9bd20 sched: Eliminate ... |
4315 |
return calc_delta_fair(gran, se); |
0bbd3336e sched: fix wakeup... |
4316 4317 4318 |
} /* |
464b75273 sched: re-instate... |
4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 |
* Should 'se' preempt 'curr'. * * |s1 * |s2 * |s3 * g * |<--->|c * * w(c, s1) = -1 * w(c, s2) = 0 * w(c, s3) = 1 * */ static int wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se) { s64 gran, vdiff = curr->vruntime - se->vruntime; if (vdiff <= 0) return -1; |
e52fb7c09 sched: prefer wakers |
4339 |
gran = wakeup_gran(curr, se); |
464b75273 sched: re-instate... |
4340 4341 4342 4343 4344 |
if (vdiff > gran) return 1; return 0; } |
02479099c sched: fix buddie... |
4345 4346 |
static void set_last_buddy(struct sched_entity *se) { |
69c80f3e9 sched: Make set_*... |
4347 4348 4349 4350 4351 |
if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE)) return; for_each_sched_entity(se) cfs_rq_of(se)->last = se; |
02479099c sched: fix buddie... |
4352 4353 4354 4355 |
} static void set_next_buddy(struct sched_entity *se) { |
69c80f3e9 sched: Make set_*... |
4356 4357 4358 4359 4360 |
if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE)) return; for_each_sched_entity(se) cfs_rq_of(se)->next = se; |
02479099c sched: fix buddie... |
4361 |
} |
ac53db596 sched: Use a budd... |
4362 4363 |
static void set_skip_buddy(struct sched_entity *se) { |
69c80f3e9 sched: Make set_*... |
4364 4365 |
for_each_sched_entity(se) cfs_rq_of(se)->skip = se; |
ac53db596 sched: Use a budd... |
4366 |
} |
464b75273 sched: re-instate... |
4367 |
/* |
bf0f6f24a sched: cfs core, ... |
4368 4369 |
* Preempt the current task with a newly woken task if needed: */ |
5a9b86f64 sched: Rename fla... |
4370 |
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) |
bf0f6f24a sched: cfs core, ... |
4371 4372 |
{ struct task_struct *curr = rq->curr; |
8651a86c3 sched: group sche... |
4373 |
struct sched_entity *se = &curr->se, *pse = &p->se; |
03e89e457 sched: fix wakeup... |
4374 |
struct cfs_rq *cfs_rq = task_cfs_rq(curr); |
f685ceaca sched: Strengthen... |
4375 |
int scale = cfs_rq->nr_running >= sched_nr_latency; |
2f36825b1 sched: Next buddy... |
4376 |
int next_buddy_marked = 0; |
bf0f6f24a sched: cfs core, ... |
4377 |
|
4ae7d5cef sched: improve af... |
4378 4379 |
if (unlikely(se == pse)) return; |
5238cdd38 sched: Prevent bu... |
4380 |
/* |
ddcdf6e7d sched: Rename loa... |
4381 |
* This is possible from callers such as move_task(), in which we |
5238cdd38 sched: Prevent bu... |
4382 4383 4384 4385 4386 4387 |
* unconditionally check_prempt_curr() after an enqueue (which may have * lead to a throttle). This both saves work and prevents false * next-buddy nomination below. */ if (unlikely(throttled_hierarchy(cfs_rq_of(pse)))) return; |
2f36825b1 sched: Next buddy... |
4388 |
if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) { |
3cb63d527 sched: Complete b... |
4389 |
set_next_buddy(pse); |
2f36825b1 sched: Next buddy... |
4390 4391 |
next_buddy_marked = 1; } |
57fdc26d4 sched: fixup budd... |
4392 |
|
aec0a5142 sched: call resch... |
4393 4394 4395 |
/* * We can come here with TIF_NEED_RESCHED already set from new task * wake up path. |
5238cdd38 sched: Prevent bu... |
4396 4397 4398 4399 4400 4401 |
* * Note: this also catches the edge-case of curr being in a throttled * group (e.g. via set_curr_task), since update_curr() (in the * enqueue of curr) will have resulted in resched being set. This * prevents us from potentially nominating it as a false LAST_BUDDY * below. |
aec0a5142 sched: call resch... |
4402 4403 4404 |
*/ if (test_tsk_need_resched(curr)) return; |
a2f5c9ab7 sched: Allow SCHE... |
4405 4406 4407 4408 |
/* Idle tasks are by definition preempted by non-idle tasks. */ if (unlikely(curr->policy == SCHED_IDLE) && likely(p->policy != SCHED_IDLE)) goto preempt; |
91c234b4e sched: do not wak... |
4409 |
/* |
a2f5c9ab7 sched: Allow SCHE... |
4410 4411 |
* Batch and idle tasks do not preempt non-idle tasks (their preemption * is driven by the tick): |
91c234b4e sched: do not wak... |
4412 |
*/ |
8ed92e51f sched: Add WAKEUP... |
4413 |
if (unlikely(p->policy != SCHED_NORMAL) || !sched_feat(WAKEUP_PREEMPTION)) |
91c234b4e sched: do not wak... |
4414 |
return; |
bf0f6f24a sched: cfs core, ... |
4415 |
|
464b75273 sched: re-instate... |
4416 |
find_matching_se(&se, &pse); |
9bbd73743 sched: update cor... |
4417 |
update_curr(cfs_rq_of(se)); |
002f128b4 sched: remove red... |
4418 |
BUG_ON(!pse); |
2f36825b1 sched: Next buddy... |
4419 4420 4421 4422 4423 4424 4425 |
if (wakeup_preempt_entity(se, pse) == 1) { /* * Bias pick_next to pick the sched entity that is * triggering this preemption. */ if (!next_buddy_marked) set_next_buddy(pse); |
3a7e73a2e sched: Clean up c... |
4426 |
goto preempt; |
2f36825b1 sched: Next buddy... |
4427 |
} |
464b75273 sched: re-instate... |
4428 |
|
3a7e73a2e sched: Clean up c... |
4429 |
return; |
a65ac745e sched: Move updat... |
4430 |
|
3a7e73a2e sched: Clean up c... |
4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 |
preempt: resched_task(curr); /* * Only set the backward buddy when the current task is still * on the rq. This can happen when a wakeup gets interleaved * with schedule on the ->pre_schedule() or idle_balance() * point, either of which can * drop the rq lock. * * Also, during early boot the idle thread is in the fair class, * for obvious reasons its a bad idea to schedule back to it. */ if (unlikely(!se->on_rq || curr == rq->idle)) return; if (sched_feat(LAST_BUDDY) && scale && entity_is_task(se)) set_last_buddy(se); |
bf0f6f24a sched: cfs core, ... |
4447 |
} |
606dba2e2 sched: Push put_p... |
4448 4449 |
static struct task_struct * pick_next_task_fair(struct rq *rq, struct task_struct *prev) |
bf0f6f24a sched: cfs core, ... |
4450 4451 4452 |
{ struct cfs_rq *cfs_rq = &rq->cfs; struct sched_entity *se; |
678d5718d sched/fair: Optim... |
4453 |
struct task_struct *p; |
37e117c07 sched: Guarantee ... |
4454 |
int new_tasks; |
678d5718d sched/fair: Optim... |
4455 |
|
6e83125c6 sched/fair: Remov... |
4456 |
again: |
678d5718d sched/fair: Optim... |
4457 4458 |
#ifdef CONFIG_FAIR_GROUP_SCHED if (!cfs_rq->nr_running) |
38033c37f sched: Push down ... |
4459 |
goto idle; |
678d5718d sched/fair: Optim... |
4460 |
|
3f1d2a318 sched: Fix hotplu... |
4461 |
if (prev->sched_class != &fair_sched_class) |
678d5718d sched/fair: Optim... |
4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 |
goto simple; /* * Because of the set_next_buddy() in dequeue_task_fair() it is rather * likely that a next task is from the same cgroup as the current. * * Therefore attempt to avoid putting and setting the entire cgroup * hierarchy, only change the part that actually changes. */ do { struct sched_entity *curr = cfs_rq->curr; /* * Since we got here without doing put_prev_entity() we also * have to consider cfs_rq->curr. If it is still a runnable * entity, update_curr() will update its vruntime, otherwise * forget we've ever seen it. */ if (curr && curr->on_rq) update_curr(cfs_rq); else curr = NULL; /* * This call to check_cfs_rq_runtime() will do the throttle and * dequeue its entity in the parent(s). Therefore the 'simple' * nr_running test will indeed be correct. */ if (unlikely(check_cfs_rq_runtime(cfs_rq))) goto simple; se = pick_next_entity(cfs_rq, curr); cfs_rq = group_cfs_rq(se); } while (cfs_rq); p = task_of(se); /* * Since we haven't yet done put_prev_entity and if the selected task * is a different task than we started out with, try and touch the * least amount of cfs_rqs. */ if (prev != p) { struct sched_entity *pse = &prev->se; while (!(cfs_rq = is_same_group(se, pse))) { int se_depth = se->depth; int pse_depth = pse->depth; if (se_depth <= pse_depth) { put_prev_entity(cfs_rq_of(pse), pse); pse = parent_entity(pse); } if (se_depth >= pse_depth) { set_next_entity(cfs_rq_of(se), se); se = parent_entity(se); } } put_prev_entity(cfs_rq, pse); set_next_entity(cfs_rq, se); } if (hrtick_enabled(rq)) hrtick_start_fair(rq, p); return p; simple: cfs_rq = &rq->cfs; #endif |
bf0f6f24a sched: cfs core, ... |
4533 |
|
36ace27e3 sched: Optimize b... |
4534 |
if (!cfs_rq->nr_running) |
38033c37f sched: Push down ... |
4535 |
goto idle; |
bf0f6f24a sched: cfs core, ... |
4536 |
|
3f1d2a318 sched: Fix hotplu... |
4537 |
put_prev_task(rq, prev); |
606dba2e2 sched: Push put_p... |
4538 |
|
bf0f6f24a sched: cfs core, ... |
4539 |
do { |
678d5718d sched/fair: Optim... |
4540 |
se = pick_next_entity(cfs_rq, NULL); |
f4b6755fb sched: cleanup fa... |
4541 |
set_next_entity(cfs_rq, se); |
bf0f6f24a sched: cfs core, ... |
4542 4543 |
cfs_rq = group_cfs_rq(se); } while (cfs_rq); |
8f4d37ec0 sched: high-res p... |
4544 |
p = task_of(se); |
678d5718d sched/fair: Optim... |
4545 |
|
b39e66eaf sched: Save some ... |
4546 4547 |
if (hrtick_enabled(rq)) hrtick_start_fair(rq, p); |
8f4d37ec0 sched: high-res p... |
4548 4549 |
return p; |
38033c37f sched: Push down ... |
4550 4551 |
idle: |
e4aa358b6 sched/fair: Push ... |
4552 |
new_tasks = idle_balance(rq); |
37e117c07 sched: Guarantee ... |
4553 4554 4555 4556 4557 |
/* * Because idle_balance() releases (and re-acquires) rq->lock, it is * possible for any higher priority task to appear. In that case we * must re-start the pick_next_entity() loop. */ |
e4aa358b6 sched/fair: Push ... |
4558 |
if (new_tasks < 0) |
37e117c07 sched: Guarantee ... |
4559 |
return RETRY_TASK; |
e4aa358b6 sched/fair: Push ... |
4560 |
if (new_tasks > 0) |
38033c37f sched: Push down ... |
4561 |
goto again; |
38033c37f sched: Push down ... |
4562 4563 |
return NULL; |
bf0f6f24a sched: cfs core, ... |
4564 4565 4566 4567 4568 |
} /* * Account for a descheduled task: */ |
31ee529cc sched: remove the... |
4569 |
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev) |
bf0f6f24a sched: cfs core, ... |
4570 4571 4572 4573 4574 4575 |
{ struct sched_entity *se = &prev->se; struct cfs_rq *cfs_rq; for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); |
ab6cde269 sched: remove the... |
4576 |
put_prev_entity(cfs_rq, se); |
bf0f6f24a sched: cfs core, ... |
4577 4578 |
} } |
ac53db596 sched: Use a budd... |
4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 |
/* * sched_yield() is very simple * * The magic of dealing with the ->skip buddy is in pick_next_entity. */ static void yield_task_fair(struct rq *rq) { struct task_struct *curr = rq->curr; struct cfs_rq *cfs_rq = task_cfs_rq(curr); struct sched_entity *se = &curr->se; /* * Are we the only task in the tree? */ if (unlikely(rq->nr_running == 1)) return; clear_buddies(cfs_rq, se); if (curr->policy != SCHED_BATCH) { update_rq_clock(rq); /* * Update run-time statistics of the 'current'. */ update_curr(cfs_rq); |
916671c08 sched: Set skip_c... |
4604 4605 4606 4607 4608 4609 |
/* * Tell update_rq_clock() that we've just updated, * so we don't do microscopic update in schedule() * and double the fastpath cost. */ rq->skip_clock_update = 1; |
ac53db596 sched: Use a budd... |
4610 4611 4612 4613 |
} set_skip_buddy(se); } |
d95f41220 sched: Add yield_... |
4614 4615 4616 |
static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt) { struct sched_entity *se = &p->se; |
5238cdd38 sched: Prevent bu... |
4617 4618 |
/* throttled hierarchies are not runnable */ if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se))) |
d95f41220 sched: Add yield_... |
4619 4620 4621 4622 |
return false; /* Tell the scheduler that we'd really like pse to run next. */ set_next_buddy(se); |
d95f41220 sched: Add yield_... |
4623 4624 4625 4626 |
yield_task_fair(rq); return true; } |
681f3e685 sched: isolate SM... |
4627 |
#ifdef CONFIG_SMP |
bf0f6f24a sched: cfs core, ... |
4628 |
/************************************************** |
e9c84cb8d sched: Describe C... |
4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 |
* Fair scheduling class load-balancing methods. * * BASICS * * The purpose of load-balancing is to achieve the same basic fairness the * per-cpu scheduler provides, namely provide a proportional amount of compute * time to each task. This is expressed in the following equation: * * W_i,n/P_i == W_j,n/P_j for all i,j (1) * * Where W_i,n is the n-th weight average for cpu i. The instantaneous weight * W_i,0 is defined as: * * W_i,0 = \Sum_j w_i,j (2) * * Where w_i,j is the weight of the j-th runnable task on cpu i. This weight * is derived from the nice value as per prio_to_weight[]. * * The weight average is an exponential decay average of the instantaneous * weight: * * W'_i,n = (2^n - 1) / 2^n * W_i,n + 1 / 2^n * W_i,0 (3) * * P_i is the cpu power (or compute capacity) of cpu i, typically it is the * fraction of 'recent' time available for SCHED_OTHER task execution. But it * can also include other factors [XXX]. * * To achieve this balance we define a measure of imbalance which follows * directly from (1): * * imb_i,j = max{ avg(W/P), W_i/P_i } - min{ avg(W/P), W_j/P_j } (4) * * We them move tasks around to minimize the imbalance. In the continuous * function space it is obvious this converges, in the discrete case we get * a few fun cases generally called infeasible weight scenarios. * * [XXX expand on: * - infeasible weights; * - local vs global optima in the discrete case. ] * * * SCHED DOMAINS * * In order to solve the imbalance equation (4), and avoid the obvious O(n^2) * for all i,j solution, we create a tree of cpus that follows the hardware * topology where each level pairs two lower groups (or better). This results * in O(log n) layers. Furthermore we reduce the number of cpus going up the * tree to only the first of the previous level and we decrease the frequency * of load-balance at each level inv. proportional to the number of cpus in * the groups. * * This yields: * * log_2 n 1 n * \Sum { --- * --- * 2^i } = O(n) (5) * i = 0 2^i 2^i * `- size of each group * | | `- number of cpus doing load-balance * | `- freq * `- sum over all levels * * Coupled with a limit on how many tasks we can migrate every balance pass, * this makes (5) the runtime complexity of the balancer. * * An important property here is that each CPU is still (indirectly) connected * to every other cpu in at most O(log n) steps: * * The adjacency matrix of the resulting graph is given by: * * log_2 n * A_i,j = \Union (i % 2^k == 0) && i / 2^(k+1) == j / 2^(k+1) (6) * k = 0 * * And you'll find that: * * A^(log_2 n)_i,j != 0 for all i,j (7) * * Showing there's indeed a path between every cpu in at most O(log n) steps. * The task movement gives a factor of O(m), giving a convergence complexity * of: * * O(nm log n), n := nr_cpus, m := nr_tasks (8) * * * WORK CONSERVING * * In order to avoid CPUs going idle while there's still work to do, new idle * balancing is more aggressive and has the newly idle cpu iterate up the domain * tree itself instead of relying on other CPUs to bring it work. * * This adds some complexity to both (5) and (8) but it reduces the total idle * time. * * [XXX more?] * * * CGROUPS * * Cgroups make a horror show out of (2), instead of a simple sum we get: * * s_k,i * W_i,0 = \Sum_j \Prod_k w_k * ----- (9) * S_k * * Where * * s_k,i = \Sum_j w_i,j,k and S_k = \Sum_i s_k,i (10) * * w_i,j,k is the weight of the j-th runnable task in the k-th cgroup on cpu i. * * The big problem is S_k, its a global sum needed to compute a local (W_i) * property. * * [XXX write more on how we solve this.. _after_ merging pjt's patches that * rewrite all of this once again.] */ |
bf0f6f24a sched: cfs core, ... |
4745 |
|
ed387b781 sched: Move SMP-o... |
4746 |
static unsigned long __read_mostly max_load_balance_interval = HZ/10; |
0ec8aa00f sched/numa: Avoid... |
4747 |
enum fbq_type { regular, remote, all }; |
ddcdf6e7d sched: Rename loa... |
4748 |
#define LBF_ALL_PINNED 0x01 |
367456c75 sched: Ditch per ... |
4749 |
#define LBF_NEED_BREAK 0x02 |
6263322c5 sched/fair: Rewri... |
4750 4751 |
#define LBF_DST_PINNED 0x04 #define LBF_SOME_PINNED 0x08 |
ddcdf6e7d sched: Rename loa... |
4752 4753 4754 |
struct lb_env { struct sched_domain *sd; |
ddcdf6e7d sched: Rename loa... |
4755 |
struct rq *src_rq; |
85c1e7dae sched: Reorder 's... |
4756 |
int src_cpu; |
ddcdf6e7d sched: Rename loa... |
4757 4758 4759 |
int dst_cpu; struct rq *dst_rq; |
88b8dac0a sched: Improve ba... |
4760 4761 |
struct cpumask *dst_grpmask; int new_dst_cpu; |
ddcdf6e7d sched: Rename loa... |
4762 |
enum cpu_idle_type idle; |
bd939f45d sched/fair: Propa... |
4763 |
long imbalance; |
b9403130a sched/cleanups: A... |
4764 4765 |
/* The set of CPUs under consideration for load-balancing */ struct cpumask *cpus; |
ddcdf6e7d sched: Rename loa... |
4766 |
unsigned int flags; |
367456c75 sched: Ditch per ... |
4767 4768 4769 4770 |
unsigned int loop; unsigned int loop_break; unsigned int loop_max; |
0ec8aa00f sched/numa: Avoid... |
4771 4772 |
enum fbq_type fbq_type; |
ddcdf6e7d sched: Rename loa... |
4773 |
}; |
1e3c88bde sched: Move load ... |
4774 |
/* |
ddcdf6e7d sched: Rename loa... |
4775 |
* move_task - move a task from one runqueue to another runqueue. |
1e3c88bde sched: Move load ... |
4776 4777 |
* Both runqueues must be locked. */ |
ddcdf6e7d sched: Rename loa... |
4778 |
static void move_task(struct task_struct *p, struct lb_env *env) |
1e3c88bde sched: Move load ... |
4779 |
{ |
ddcdf6e7d sched: Rename loa... |
4780 4781 4782 4783 |
deactivate_task(env->src_rq, p, 0); set_task_cpu(p, env->dst_cpu); activate_task(env->dst_rq, p, 0); check_preempt_curr(env->dst_rq, p, 0); |
1e3c88bde sched: Move load ... |
4784 4785 4786 |
} /* |
029632fbb sched: Make separ... |
4787 4788 4789 |
* Is this task likely cache-hot: */ static int |
6037dd1a4 sched: Clean up t... |
4790 |
task_hot(struct task_struct *p, u64 now) |
029632fbb sched: Make separ... |
4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 |
{ s64 delta; if (p->sched_class != &fair_sched_class) return 0; if (unlikely(p->policy == SCHED_IDLE)) return 0; /* * Buddy candidates are cache hot: */ if (sched_feat(CACHE_HOT_BUDDY) && this_rq()->nr_running && (&p->se == cfs_rq_of(&p->se)->next || &p->se == cfs_rq_of(&p->se)->last)) return 1; if (sysctl_sched_migration_cost == -1) return 1; if (sysctl_sched_migration_cost == 0) return 0; delta = now - p->se.exec_start; return delta < (s64)sysctl_sched_migration_cost; } |
3a7053b32 sched/numa: Favou... |
4817 4818 4819 4820 4821 |
#ifdef CONFIG_NUMA_BALANCING /* Returns true if the destination node has incurred more faults */ static bool migrate_improves_locality(struct task_struct *p, struct lb_env *env) { int src_nid, dst_nid; |
ff1df896a sched/numa: Renam... |
4822 |
if (!sched_feat(NUMA_FAVOUR_HIGHER) || !p->numa_faults_memory || |
3a7053b32 sched/numa: Favou... |
4823 4824 4825 4826 4827 4828 |
!(env->sd->flags & SD_NUMA)) { return false; } src_nid = cpu_to_node(env->src_cpu); dst_nid = cpu_to_node(env->dst_cpu); |
83e1d2cd9 sched/numa: Use g... |
4829 |
if (src_nid == dst_nid) |
3a7053b32 sched/numa: Favou... |
4830 |
return false; |
83e1d2cd9 sched/numa: Use g... |
4831 4832 4833 |
/* Always encourage migration to the preferred node. */ if (dst_nid == p->numa_preferred_nid) return true; |
887c290e8 sched/numa: Decid... |
4834 4835 4836 |
/* If both task and group weight improve, this move is a winner. */ if (task_weight(p, dst_nid) > task_weight(p, src_nid) && group_weight(p, dst_nid) > group_weight(p, src_nid)) |
3a7053b32 sched/numa: Favou... |
4837 4838 4839 4840 |
return true; return false; } |
7a0f30833 sched/numa: Resis... |
4841 4842 4843 4844 4845 4846 4847 4848 |
static bool migrate_degrades_locality(struct task_struct *p, struct lb_env *env) { int src_nid, dst_nid; if (!sched_feat(NUMA) || !sched_feat(NUMA_RESIST_LOWER)) return false; |
ff1df896a sched/numa: Renam... |
4849 |
if (!p->numa_faults_memory || !(env->sd->flags & SD_NUMA)) |
7a0f30833 sched/numa: Resis... |
4850 4851 4852 4853 |
return false; src_nid = cpu_to_node(env->src_cpu); dst_nid = cpu_to_node(env->dst_cpu); |
83e1d2cd9 sched/numa: Use g... |
4854 |
if (src_nid == dst_nid) |
7a0f30833 sched/numa: Resis... |
4855 |
return false; |
83e1d2cd9 sched/numa: Use g... |
4856 4857 4858 |
/* Migrating away from the preferred node is always bad. */ if (src_nid == p->numa_preferred_nid) return true; |
887c290e8 sched/numa: Decid... |
4859 4860 4861 |
/* If either task or group weight get worse, don't do it. */ if (task_weight(p, dst_nid) < task_weight(p, src_nid) || group_weight(p, dst_nid) < group_weight(p, src_nid)) |
7a0f30833 sched/numa: Resis... |
4862 4863 4864 4865 |
return true; return false; } |
3a7053b32 sched/numa: Favou... |
4866 4867 4868 4869 4870 4871 |
#else static inline bool migrate_improves_locality(struct task_struct *p, struct lb_env *env) { return false; } |
7a0f30833 sched/numa: Resis... |
4872 4873 4874 4875 4876 4877 |
static inline bool migrate_degrades_locality(struct task_struct *p, struct lb_env *env) { return false; } |
3a7053b32 sched/numa: Favou... |
4878 |
#endif |
029632fbb sched: Make separ... |
4879 |
/* |
1e3c88bde sched: Move load ... |
4880 4881 4882 |
* can_migrate_task - may task p from runqueue rq be migrated to this_cpu? */ static |
8e45cb545 sched: Move load-... |
4883 |
int can_migrate_task(struct task_struct *p, struct lb_env *env) |
1e3c88bde sched: Move load ... |
4884 4885 4886 4887 |
{ int tsk_cache_hot = 0; /* * We do not migrate tasks that are: |
d31980846 sched: Move up af... |
4888 |
* 1) throttled_lb_pair, or |
1e3c88bde sched: Move load ... |
4889 |
* 2) cannot be migrated to this CPU due to cpus_allowed, or |
d31980846 sched: Move up af... |
4890 4891 |
* 3) running (obviously), or * 4) are cache-hot on their current CPU. |
1e3c88bde sched: Move load ... |
4892 |
*/ |
d31980846 sched: Move up af... |
4893 4894 |
if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu)) return 0; |
ddcdf6e7d sched: Rename loa... |
4895 |
if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) { |
e02e60c10 sched: Prevent to... |
4896 |
int cpu; |
88b8dac0a sched: Improve ba... |
4897 |
|
41acab885 sched: Implement ... |
4898 |
schedstat_inc(p, se.statistics.nr_failed_migrations_affine); |
88b8dac0a sched: Improve ba... |
4899 |
|
6263322c5 sched/fair: Rewri... |
4900 |
env->flags |= LBF_SOME_PINNED; |
88b8dac0a sched: Improve ba... |
4901 4902 4903 4904 4905 4906 4907 4908 |
/* * Remember if this task can be migrated to any other cpu in * our sched_group. We may want to revisit it if we couldn't * meet load balance goals by pulling other tasks on src_cpu. * * Also avoid computing new_dst_cpu if we have already computed * one in current iteration. */ |
6263322c5 sched/fair: Rewri... |
4909 |
if (!env->dst_grpmask || (env->flags & LBF_DST_PINNED)) |
88b8dac0a sched: Improve ba... |
4910 |
return 0; |
e02e60c10 sched: Prevent to... |
4911 4912 4913 |
/* Prevent to re-select dst_cpu via env's cpus */ for_each_cpu_and(cpu, env->dst_grpmask, env->cpus) { if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) { |
6263322c5 sched/fair: Rewri... |
4914 |
env->flags |= LBF_DST_PINNED; |
e02e60c10 sched: Prevent to... |
4915 4916 4917 |
env->new_dst_cpu = cpu; break; } |
88b8dac0a sched: Improve ba... |
4918 |
} |
e02e60c10 sched: Prevent to... |
4919 |
|
1e3c88bde sched: Move load ... |
4920 4921 |
return 0; } |
88b8dac0a sched: Improve ba... |
4922 4923 |
/* Record that we found atleast one task that could run on dst_cpu */ |
8e45cb545 sched: Move load-... |
4924 |
env->flags &= ~LBF_ALL_PINNED; |
1e3c88bde sched: Move load ... |
4925 |
|
ddcdf6e7d sched: Rename loa... |
4926 |
if (task_running(env->src_rq, p)) { |
41acab885 sched: Implement ... |
4927 |
schedstat_inc(p, se.statistics.nr_failed_migrations_running); |
1e3c88bde sched: Move load ... |
4928 4929 4930 4931 4932 |
return 0; } /* * Aggressive migration if: |
3a7053b32 sched/numa: Favou... |
4933 4934 4935 |
* 1) destination numa is preferred * 2) task is cache cold, or * 3) too many balance attempts have failed. |
1e3c88bde sched: Move load ... |
4936 |
*/ |
6037dd1a4 sched: Clean up t... |
4937 |
tsk_cache_hot = task_hot(p, rq_clock_task(env->src_rq)); |
7a0f30833 sched/numa: Resis... |
4938 4939 |
if (!tsk_cache_hot) tsk_cache_hot = migrate_degrades_locality(p, env); |
3a7053b32 sched/numa: Favou... |
4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 |
if (migrate_improves_locality(p, env)) { #ifdef CONFIG_SCHEDSTATS if (tsk_cache_hot) { schedstat_inc(env->sd, lb_hot_gained[env->idle]); schedstat_inc(p, se.statistics.nr_forced_migrations); } #endif return 1; } |
1e3c88bde sched: Move load ... |
4950 |
if (!tsk_cache_hot || |
8e45cb545 sched: Move load-... |
4951 |
env->sd->nr_balance_failed > env->sd->cache_nice_tries) { |
4e2dcb73a sched: Simplify c... |
4952 |
|
1e3c88bde sched: Move load ... |
4953 |
if (tsk_cache_hot) { |
8e45cb545 sched: Move load-... |
4954 |
schedstat_inc(env->sd, lb_hot_gained[env->idle]); |
41acab885 sched: Implement ... |
4955 |
schedstat_inc(p, se.statistics.nr_forced_migrations); |
1e3c88bde sched: Move load ... |
4956 |
} |
4e2dcb73a sched: Simplify c... |
4957 |
|
1e3c88bde sched: Move load ... |
4958 4959 |
return 1; } |
4e2dcb73a sched: Simplify c... |
4960 4961 |
schedstat_inc(p, se.statistics.nr_failed_migrations_hot); return 0; |
1e3c88bde sched: Move load ... |
4962 |
} |
897c395f4 sched: Remove rq_... |
4963 4964 4965 4966 4967 4968 4969 |
/* * move_one_task tries to move exactly one task from busiest to this_rq, as * part of active balancing operations within "domain". * Returns 1 if successful and 0 otherwise. * * Called with both runqueues locked. */ |
8e45cb545 sched: Move load-... |
4970 |
static int move_one_task(struct lb_env *env) |
897c395f4 sched: Remove rq_... |
4971 4972 |
{ struct task_struct *p, *n; |
897c395f4 sched: Remove rq_... |
4973 |
|
367456c75 sched: Ditch per ... |
4974 |
list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) { |
367456c75 sched: Ditch per ... |
4975 4976 |
if (!can_migrate_task(p, env)) continue; |
897c395f4 sched: Remove rq_... |
4977 |
|
367456c75 sched: Ditch per ... |
4978 4979 4980 4981 4982 4983 4984 4985 |
move_task(p, env); /* * Right now, this is only the second place move_task() * is called, so we can safely collect move_task() * stats here rather than inside move_task(). */ schedstat_inc(env->sd, lb_gained[env->idle]); return 1; |
897c395f4 sched: Remove rq_... |
4986 |
} |
897c395f4 sched: Remove rq_... |
4987 4988 |
return 0; } |
eb95308ee sched: Fix more l... |
4989 |
static const unsigned int sched_nr_migrate_break = 32; |
5d6523ebd sched: Fix load-b... |
4990 |
/* |
bd939f45d sched/fair: Propa... |
4991 |
* move_tasks tries to move up to imbalance weighted load from busiest to |
5d6523ebd sched: Fix load-b... |
4992 4993 4994 4995 4996 4997 |
* this_rq, as part of a balancing operation within domain "sd". * Returns 1 if successful and 0 otherwise. * * Called with both runqueues locked. */ static int move_tasks(struct lb_env *env) |
1e3c88bde sched: Move load ... |
4998 |
{ |
5d6523ebd sched: Fix load-b... |
4999 5000 |
struct list_head *tasks = &env->src_rq->cfs_tasks; struct task_struct *p; |
367456c75 sched: Ditch per ... |
5001 5002 |
unsigned long load; int pulled = 0; |
1e3c88bde sched: Move load ... |
5003 |
|
bd939f45d sched/fair: Propa... |
5004 |
if (env->imbalance <= 0) |
5d6523ebd sched: Fix load-b... |
5005 |
return 0; |
1e3c88bde sched: Move load ... |
5006 |
|
5d6523ebd sched: Fix load-b... |
5007 5008 |
while (!list_empty(tasks)) { p = list_first_entry(tasks, struct task_struct, se.group_node); |
1e3c88bde sched: Move load ... |
5009 |
|
367456c75 sched: Ditch per ... |
5010 5011 |
env->loop++; /* We've more or less seen every task there is, call it quits */ |
5d6523ebd sched: Fix load-b... |
5012 |
if (env->loop > env->loop_max) |
367456c75 sched: Ditch per ... |
5013 |
break; |
5d6523ebd sched: Fix load-b... |
5014 5015 |
/* take a breather every nr_migrate tasks */ |
367456c75 sched: Ditch per ... |
5016 |
if (env->loop > env->loop_break) { |
eb95308ee sched: Fix more l... |
5017 |
env->loop_break += sched_nr_migrate_break; |
8e45cb545 sched: Move load-... |
5018 |
env->flags |= LBF_NEED_BREAK; |
ee00e66ff sched: Remove rq_... |
5019 |
break; |
a195f004e sched: Fix load-b... |
5020 |
} |
1e3c88bde sched: Move load ... |
5021 |
|
d31980846 sched: Move up af... |
5022 |
if (!can_migrate_task(p, env)) |
367456c75 sched: Ditch per ... |
5023 5024 5025 |
goto next; load = task_h_load(p); |
5d6523ebd sched: Fix load-b... |
5026 |
|
eb95308ee sched: Fix more l... |
5027 |
if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed) |
367456c75 sched: Ditch per ... |
5028 |
goto next; |
bd939f45d sched/fair: Propa... |
5029 |
if ((load / 2) > env->imbalance) |
367456c75 sched: Ditch per ... |
5030 |
goto next; |
1e3c88bde sched: Move load ... |
5031 |
|
ddcdf6e7d sched: Rename loa... |
5032 |
move_task(p, env); |
ee00e66ff sched: Remove rq_... |
5033 |
pulled++; |
bd939f45d sched/fair: Propa... |
5034 |
env->imbalance -= load; |
1e3c88bde sched: Move load ... |
5035 5036 |
#ifdef CONFIG_PREEMPT |
ee00e66ff sched: Remove rq_... |
5037 5038 5039 5040 5041 |
/* * NEWIDLE balancing is a source of latency, so preemptible * kernels will stop after the first task is pulled to minimize * the critical section. */ |
5d6523ebd sched: Fix load-b... |
5042 |
if (env->idle == CPU_NEWLY_IDLE) |
ee00e66ff sched: Remove rq_... |
5043 |
break; |
1e3c88bde sched: Move load ... |
5044 |
#endif |
ee00e66ff sched: Remove rq_... |
5045 5046 5047 5048 |
/* * We only want to steal up to the prescribed amount of * weighted load. */ |
bd939f45d sched/fair: Propa... |
5049 |
if (env->imbalance <= 0) |
ee00e66ff sched: Remove rq_... |
5050 |
break; |
367456c75 sched: Ditch per ... |
5051 5052 5053 |
continue; next: |
5d6523ebd sched: Fix load-b... |
5054 |
list_move_tail(&p->se.group_node, tasks); |
1e3c88bde sched: Move load ... |
5055 |
} |
5d6523ebd sched: Fix load-b... |
5056 |
|
1e3c88bde sched: Move load ... |
5057 |
/* |
ddcdf6e7d sched: Rename loa... |
5058 5059 5060 |
* Right now, this is one of only two places move_task() is called, * so we can safely collect move_task() stats here rather than * inside move_task(). |
1e3c88bde sched: Move load ... |
5061 |
*/ |
8e45cb545 sched: Move load-... |
5062 |
schedstat_add(env->sd, lb_gained[env->idle], pulled); |
1e3c88bde sched: Move load ... |
5063 |
|
5d6523ebd sched: Fix load-b... |
5064 |
return pulled; |
1e3c88bde sched: Move load ... |
5065 |
} |
230059de7 sched: Remove fro... |
5066 |
#ifdef CONFIG_FAIR_GROUP_SCHED |
9e3081ca6 sched: Make tg_sh... |
5067 5068 5069 |
/* * update tg->load_weight by folding this cpu's load_avg */ |
48a167532 sched: Refactor u... |
5070 |
static void __update_blocked_averages_cpu(struct task_group *tg, int cpu) |
9e3081ca6 sched: Make tg_sh... |
5071 |
{ |
48a167532 sched: Refactor u... |
5072 5073 |
struct sched_entity *se = tg->se[cpu]; struct cfs_rq *cfs_rq = tg->cfs_rq[cpu]; |
9e3081ca6 sched: Make tg_sh... |
5074 |
|
48a167532 sched: Refactor u... |
5075 5076 5077 |
/* throttled entities do not contribute to load */ if (throttled_hierarchy(cfs_rq)) return; |
9e3081ca6 sched: Make tg_sh... |
5078 |
|
aff3e4988 sched: Account fo... |
5079 |
update_cfs_rq_blocked_load(cfs_rq, 1); |
9e3081ca6 sched: Make tg_sh... |
5080 |
|
82958366c sched: Replace up... |
5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 |
if (se) { update_entity_load_avg(se, 1); /* * We pivot on our runnable average having decayed to zero for * list removal. This generally implies that all our children * have also been removed (modulo rounding error or bandwidth * control); however, such cases are rare and we can fix these * at enqueue. * * TODO: fix up out-of-order children on enqueue. */ if (!se->avg.runnable_avg_sum && !cfs_rq->nr_running) list_del_leaf_cfs_rq(cfs_rq); } else { |
48a167532 sched: Refactor u... |
5095 |
struct rq *rq = rq_of(cfs_rq); |
82958366c sched: Replace up... |
5096 5097 |
update_rq_runnable_avg(rq, rq->nr_running); } |
9e3081ca6 sched: Make tg_sh... |
5098 |
} |
48a167532 sched: Refactor u... |
5099 |
static void update_blocked_averages(int cpu) |
9e3081ca6 sched: Make tg_sh... |
5100 |
{ |
9e3081ca6 sched: Make tg_sh... |
5101 |
struct rq *rq = cpu_rq(cpu); |
48a167532 sched: Refactor u... |
5102 5103 |
struct cfs_rq *cfs_rq; unsigned long flags; |
9e3081ca6 sched: Make tg_sh... |
5104 |
|
48a167532 sched: Refactor u... |
5105 5106 |
raw_spin_lock_irqsave(&rq->lock, flags); update_rq_clock(rq); |
9763b67fb sched, cgroup: Op... |
5107 5108 5109 5110 |
/* * Iterates the task_group tree in a bottom up fashion, see * list_add_leaf_cfs_rq() for details. */ |
64660c864 sched: Prevent in... |
5111 |
for_each_leaf_cfs_rq(rq, cfs_rq) { |
48a167532 sched: Refactor u... |
5112 5113 5114 5115 5116 5117 |
/* * Note: We may want to consider periodically releasing * rq->lock about these updates so that creating many task * groups does not result in continually extending hold time. */ __update_blocked_averages_cpu(cfs_rq->tg, rq->cpu); |
64660c864 sched: Prevent in... |
5118 |
} |
48a167532 sched: Refactor u... |
5119 5120 |
raw_spin_unlock_irqrestore(&rq->lock, flags); |
9e3081ca6 sched: Make tg_sh... |
5121 |
} |
9763b67fb sched, cgroup: Op... |
5122 |
/* |
685207963 sched: Move h_loa... |
5123 |
* Compute the hierarchical load factor for cfs_rq and all its ascendants. |
9763b67fb sched, cgroup: Op... |
5124 5125 5126 |
* This needs to be done in a top-down fashion because the load of a child * group is a fraction of its parents load. */ |
685207963 sched: Move h_loa... |
5127 |
static void update_cfs_rq_h_load(struct cfs_rq *cfs_rq) |
9763b67fb sched, cgroup: Op... |
5128 |
{ |
685207963 sched: Move h_loa... |
5129 5130 |
struct rq *rq = rq_of(cfs_rq); struct sched_entity *se = cfs_rq->tg->se[cpu_of(rq)]; |
a35b6466a sched, cgroup: Re... |
5131 |
unsigned long now = jiffies; |
685207963 sched: Move h_loa... |
5132 |
unsigned long load; |
a35b6466a sched, cgroup: Re... |
5133 |
|
685207963 sched: Move h_loa... |
5134 |
if (cfs_rq->last_h_load_update == now) |
a35b6466a sched, cgroup: Re... |
5135 |
return; |
685207963 sched: Move h_loa... |
5136 5137 5138 5139 5140 5141 5142 |
cfs_rq->h_load_next = NULL; for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); cfs_rq->h_load_next = se; if (cfs_rq->last_h_load_update == now) break; } |
a35b6466a sched, cgroup: Re... |
5143 |
|
685207963 sched: Move h_loa... |
5144 |
if (!se) { |
7e3115ef5 sched/balancing: ... |
5145 |
cfs_rq->h_load = cfs_rq->runnable_load_avg; |
685207963 sched: Move h_loa... |
5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 |
cfs_rq->last_h_load_update = now; } while ((se = cfs_rq->h_load_next) != NULL) { load = cfs_rq->h_load; load = div64_ul(load * se->avg.load_avg_contrib, cfs_rq->runnable_load_avg + 1); cfs_rq = group_cfs_rq(se); cfs_rq->h_load = load; cfs_rq->last_h_load_update = now; } |
9763b67fb sched, cgroup: Op... |
5157 |
} |
367456c75 sched: Ditch per ... |
5158 |
static unsigned long task_h_load(struct task_struct *p) |
230059de7 sched: Remove fro... |
5159 |
{ |
367456c75 sched: Ditch per ... |
5160 |
struct cfs_rq *cfs_rq = task_cfs_rq(p); |
230059de7 sched: Remove fro... |
5161 |
|
685207963 sched: Move h_loa... |
5162 |
update_cfs_rq_h_load(cfs_rq); |
a003a25b2 sched: Consider r... |
5163 5164 |
return div64_ul(p->se.avg.load_avg_contrib * cfs_rq->h_load, cfs_rq->runnable_load_avg + 1); |
230059de7 sched: Remove fro... |
5165 5166 |
} #else |
48a167532 sched: Refactor u... |
5167 |
static inline void update_blocked_averages(int cpu) |
9e3081ca6 sched: Make tg_sh... |
5168 5169 |
{ } |
367456c75 sched: Ditch per ... |
5170 |
static unsigned long task_h_load(struct task_struct *p) |
1e3c88bde sched: Move load ... |
5171 |
{ |
a003a25b2 sched: Consider r... |
5172 |
return p->se.avg.load_avg_contrib; |
1e3c88bde sched: Move load ... |
5173 |
} |
230059de7 sched: Remove fro... |
5174 |
#endif |
1e3c88bde sched: Move load ... |
5175 |
|
1e3c88bde sched: Move load ... |
5176 5177 |
/********** Helpers for find_busiest_group ************************/ /* |
1e3c88bde sched: Move load ... |
5178 5179 5180 5181 5182 |
* sg_lb_stats - stats of a sched_group required for load_balancing */ struct sg_lb_stats { unsigned long avg_load; /*Avg load across the CPUs of the group */ unsigned long group_load; /* Total load over the CPUs of the group */ |
1e3c88bde sched: Move load ... |
5183 |
unsigned long sum_weighted_load; /* Weighted load of group's tasks */ |
56cf515b4 sched: Clean-up s... |
5184 |
unsigned long load_per_task; |
3ae11c90f sched/fair: Make ... |
5185 |
unsigned long group_power; |
147c5fc2b sched/fair: Shrin... |
5186 5187 5188 5189 |
unsigned int sum_nr_running; /* Nr tasks running in the group */ unsigned int group_capacity; unsigned int idle_cpus; unsigned int group_weight; |
1e3c88bde sched: Move load ... |
5190 |
int group_imb; /* Is there an imbalance in the group ? */ |
fab476228 sched: Force bala... |
5191 |
int group_has_capacity; /* Is there extra capacity in the group? */ |
0ec8aa00f sched/numa: Avoid... |
5192 5193 5194 5195 |
#ifdef CONFIG_NUMA_BALANCING unsigned int nr_numa_running; unsigned int nr_preferred_running; #endif |
1e3c88bde sched: Move load ... |
5196 |
}; |
56cf515b4 sched: Clean-up s... |
5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 |
/* * sd_lb_stats - Structure to store the statistics of a sched_domain * during load balancing. */ struct sd_lb_stats { struct sched_group *busiest; /* Busiest group in this sd */ struct sched_group *local; /* Local group in this sd */ unsigned long total_load; /* Total load of all groups in sd */ unsigned long total_pwr; /* Total power of all groups in sd */ unsigned long avg_load; /* Average load across all groups in sd */ |
56cf515b4 sched: Clean-up s... |
5207 |
struct sg_lb_stats busiest_stat;/* Statistics of the busiest group */ |
147c5fc2b sched/fair: Shrin... |
5208 |
struct sg_lb_stats local_stat; /* Statistics of the local group */ |
56cf515b4 sched: Clean-up s... |
5209 |
}; |
147c5fc2b sched/fair: Shrin... |
5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 |
static inline void init_sd_lb_stats(struct sd_lb_stats *sds) { /* * Skimp on the clearing to avoid duplicate work. We can avoid clearing * local_stat because update_sg_lb_stats() does a full clear/assignment. * We must however clear busiest_stat::avg_load because * update_sd_pick_busiest() reads this before assignment. */ *sds = (struct sd_lb_stats){ .busiest = NULL, .local = NULL, .total_load = 0UL, .total_pwr = 0UL, .busiest_stat = { .avg_load = 0UL, }, }; } |
1e3c88bde sched: Move load ... |
5228 |
/** |
1e3c88bde sched: Move load ... |
5229 5230 |
* get_sd_load_idx - Obtain the load index for a given sched domain. * @sd: The sched_domain whose load_idx is to be obtained. |
ed1b77328 sched/fair: Fix t... |
5231 |
* @idle: The idle status of the CPU for whose sd load_idx is obtained. |
e69f61862 sched: Fix some k... |
5232 5233 |
* * Return: The load index. |
1e3c88bde sched: Move load ... |
5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 |
*/ static inline int get_sd_load_idx(struct sched_domain *sd, enum cpu_idle_type idle) { int load_idx; switch (idle) { case CPU_NOT_IDLE: load_idx = sd->busy_idx; break; case CPU_NEWLY_IDLE: load_idx = sd->newidle_idx; break; default: load_idx = sd->idle_idx; break; } return load_idx; } |
15f803c94 sched: Make defau... |
5255 |
static unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu) |
1e3c88bde sched: Move load ... |
5256 |
{ |
1399fa780 sched: Introduce ... |
5257 |
return SCHED_POWER_SCALE; |
1e3c88bde sched: Move load ... |
5258 5259 5260 5261 5262 5263 |
} unsigned long __weak arch_scale_freq_power(struct sched_domain *sd, int cpu) { return default_scale_freq_power(sd, cpu); } |
15f803c94 sched: Make defau... |
5264 |
static unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu) |
1e3c88bde sched: Move load ... |
5265 |
{ |
669c55e9f sched: Pre-comput... |
5266 |
unsigned long weight = sd->span_weight; |
1e3c88bde sched: Move load ... |
5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 |
unsigned long smt_gain = sd->smt_gain; smt_gain /= weight; return smt_gain; } unsigned long __weak arch_scale_smt_power(struct sched_domain *sd, int cpu) { return default_scale_smt_power(sd, cpu); } |
15f803c94 sched: Make defau... |
5278 |
static unsigned long scale_rt_power(int cpu) |
1e3c88bde sched: Move load ... |
5279 5280 |
{ struct rq *rq = cpu_rq(cpu); |
b654f7de4 sched: Make sure ... |
5281 |
u64 total, available, age_stamp, avg; |
1e3c88bde sched: Move load ... |
5282 |
|
b654f7de4 sched: Make sure ... |
5283 5284 5285 5286 5287 5288 |
/* * Since we're reading these variables without serialization make sure * we read them once before doing sanity checks on them. */ age_stamp = ACCESS_ONCE(rq->age_stamp); avg = ACCESS_ONCE(rq->rt_avg); |
78becc270 sched: Use an acc... |
5289 |
total = sched_avg_period() + (rq_clock(rq) - age_stamp); |
aa4838085 sched: Remove irq... |
5290 |
|
b654f7de4 sched: Make sure ... |
5291 |
if (unlikely(total < avg)) { |
aa4838085 sched: Remove irq... |
5292 5293 5294 |
/* Ensures that power won't end up being negative */ available = 0; } else { |
b654f7de4 sched: Make sure ... |
5295 |
available = total - avg; |
aa4838085 sched: Remove irq... |
5296 |
} |
1e3c88bde sched: Move load ... |
5297 |
|
1399fa780 sched: Introduce ... |
5298 5299 |
if (unlikely((s64)total < SCHED_POWER_SCALE)) total = SCHED_POWER_SCALE; |
1e3c88bde sched: Move load ... |
5300 |
|
1399fa780 sched: Introduce ... |
5301 |
total >>= SCHED_POWER_SHIFT; |
1e3c88bde sched: Move load ... |
5302 5303 5304 5305 5306 5307 |
return div_u64(available, total); } static void update_cpu_power(struct sched_domain *sd, int cpu) { |
669c55e9f sched: Pre-comput... |
5308 |
unsigned long weight = sd->span_weight; |
1399fa780 sched: Introduce ... |
5309 |
unsigned long power = SCHED_POWER_SCALE; |
1e3c88bde sched: Move load ... |
5310 |
struct sched_group *sdg = sd->groups; |
1e3c88bde sched: Move load ... |
5311 5312 5313 5314 5315 |
if ((sd->flags & SD_SHARE_CPUPOWER) && weight > 1) { if (sched_feat(ARCH_POWER)) power *= arch_scale_smt_power(sd, cpu); else power *= default_scale_smt_power(sd, cpu); |
1399fa780 sched: Introduce ... |
5316 |
power >>= SCHED_POWER_SHIFT; |
1e3c88bde sched: Move load ... |
5317 |
} |
9c3f75cbd sched: Break out ... |
5318 |
sdg->sgp->power_orig = power; |
9d5efe05e sched: Fix capaci... |
5319 5320 5321 5322 5323 |
if (sched_feat(ARCH_POWER)) power *= arch_scale_freq_power(sd, cpu); else power *= default_scale_freq_power(sd, cpu); |
1399fa780 sched: Introduce ... |
5324 |
power >>= SCHED_POWER_SHIFT; |
9d5efe05e sched: Fix capaci... |
5325 |
|
1e3c88bde sched: Move load ... |
5326 |
power *= scale_rt_power(cpu); |
1399fa780 sched: Introduce ... |
5327 |
power >>= SCHED_POWER_SHIFT; |
1e3c88bde sched: Move load ... |
5328 5329 5330 |
if (!power) power = 1; |
e51fd5e22 sched: Fix wake_a... |
5331 |
cpu_rq(cpu)->cpu_power = power; |
9c3f75cbd sched: Break out ... |
5332 |
sdg->sgp->power = power; |
1e3c88bde sched: Move load ... |
5333 |
} |
029632fbb sched: Make separ... |
5334 |
void update_group_power(struct sched_domain *sd, int cpu) |
1e3c88bde sched: Move load ... |
5335 5336 5337 |
{ struct sched_domain *child = sd->child; struct sched_group *group, *sdg = sd->groups; |
863bffc80 sched/fair: Fix g... |
5338 |
unsigned long power, power_orig; |
4ec4412e1 sched: Ensure cpu... |
5339 5340 5341 5342 5343 |
unsigned long interval; interval = msecs_to_jiffies(sd->balance_interval); interval = clamp(interval, 1UL, max_load_balance_interval); sdg->sgp->next_update = jiffies + interval; |
1e3c88bde sched: Move load ... |
5344 5345 5346 5347 5348 |
if (!child) { update_cpu_power(sd, cpu); return; } |
863bffc80 sched/fair: Fix g... |
5349 |
power_orig = power = 0; |
1e3c88bde sched: Move load ... |
5350 |
|
74a5ce20e sched: Fix SD_OVE... |
5351 5352 5353 5354 5355 |
if (child->flags & SD_OVERLAP) { /* * SD_OVERLAP domains cannot assume that child groups * span the current group. */ |
863bffc80 sched/fair: Fix g... |
5356 |
for_each_cpu(cpu, sched_group_cpus(sdg)) { |
9abf24d46 sched: Check sche... |
5357 5358 |
struct sched_group_power *sgp; struct rq *rq = cpu_rq(cpu); |
863bffc80 sched/fair: Fix g... |
5359 |
|
9abf24d46 sched: Check sche... |
5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 |
/* * build_sched_domains() -> init_sched_groups_power() * gets here before we've attached the domains to the * runqueues. * * Use power_of(), which is set irrespective of domains * in update_cpu_power(). * * This avoids power/power_orig from being 0 and * causing divide-by-zero issues on boot. * * Runtime updates will correct power_orig. */ if (unlikely(!rq->sd)) { power_orig += power_of(cpu); power += power_of(cpu); continue; } |
863bffc80 sched/fair: Fix g... |
5378 |
|
9abf24d46 sched: Check sche... |
5379 5380 5381 |
sgp = rq->sd->groups->sgp; power_orig += sgp->power_orig; power += sgp->power; |
863bffc80 sched/fair: Fix g... |
5382 |
} |
74a5ce20e sched: Fix SD_OVE... |
5383 5384 5385 5386 5387 5388 5389 5390 |
} else { /* * !SD_OVERLAP domains can assume that child groups * span the current group. */ group = child->groups; do { |
863bffc80 sched/fair: Fix g... |
5391 |
power_orig += group->sgp->power_orig; |
74a5ce20e sched: Fix SD_OVE... |
5392 5393 5394 5395 |
power += group->sgp->power; group = group->next; } while (group != child->groups); } |
1e3c88bde sched: Move load ... |
5396 |
|
863bffc80 sched/fair: Fix g... |
5397 5398 |
sdg->sgp->power_orig = power_orig; sdg->sgp->power = power; |
1e3c88bde sched: Move load ... |
5399 |
} |
9d5efe05e sched: Fix capaci... |
5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 |
/* * Try and fix up capacity for tiny siblings, this is needed when * things like SD_ASYM_PACKING need f_b_g to select another sibling * which on its own isn't powerful enough. * * See update_sd_pick_busiest() and check_asym_packing(). */ static inline int fix_small_capacity(struct sched_domain *sd, struct sched_group *group) { /* |
1399fa780 sched: Introduce ... |
5411 |
* Only siblings can have significantly less than SCHED_POWER_SCALE |
9d5efe05e sched: Fix capaci... |
5412 |
*/ |
a6c75f2f8 sched: Avoid usin... |
5413 |
if (!(sd->flags & SD_SHARE_CPUPOWER)) |
9d5efe05e sched: Fix capaci... |
5414 5415 5416 5417 5418 |
return 0; /* * If ~90% of the cpu_power is still there, we're good. */ |
9c3f75cbd sched: Break out ... |
5419 |
if (group->sgp->power * 32 > group->sgp->power_orig * 29) |
9d5efe05e sched: Fix capaci... |
5420 5421 5422 5423 |
return 1; return 0; } |
30ce5dabc sched/fair: Rewor... |
5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 |
/* * Group imbalance indicates (and tries to solve) the problem where balancing * groups is inadequate due to tsk_cpus_allowed() constraints. * * Imagine a situation of two groups of 4 cpus each and 4 tasks each with a * cpumask covering 1 cpu of the first group and 3 cpus of the second group. * Something like: * * { 0 1 2 3 } { 4 5 6 7 } * * * * * * * If we were to balance group-wise we'd place two tasks in the first group and * two tasks in the second group. Clearly this is undesired as it will overload * cpu 3 and leave one of the cpus in the second group unused. * * The current solution to this issue is detecting the skew in the first group |
6263322c5 sched/fair: Rewri... |
5440 5441 |
* by noticing the lower domain failed to reach balance and had difficulty * moving tasks due to affinity constraints. |
30ce5dabc sched/fair: Rewor... |
5442 5443 |
* * When this is so detected; this group becomes a candidate for busiest; see |
ed1b77328 sched/fair: Fix t... |
5444 |
* update_sd_pick_busiest(). And calculate_imbalance() and |
6263322c5 sched/fair: Rewri... |
5445 |
* find_busiest_group() avoid some of the usual balance conditions to allow it |
30ce5dabc sched/fair: Rewor... |
5446 5447 5448 5449 5450 5451 |
* to create an effective group imbalance. * * This is a somewhat tricky proposition since the next run might not find the * group imbalance and decide the groups need to be balanced again. A most * subtle and fragile situation. */ |
6263322c5 sched/fair: Rewri... |
5452 |
static inline int sg_imbalanced(struct sched_group *group) |
30ce5dabc sched/fair: Rewor... |
5453 |
{ |
6263322c5 sched/fair: Rewri... |
5454 |
return group->sgp->imbalance; |
30ce5dabc sched/fair: Rewor... |
5455 |
} |
b37d93168 sched/fair: Rewor... |
5456 5457 5458 |
/* * Compute the group capacity. * |
c61037e90 sched/fair: Fix t... |
5459 5460 5461 |
* Avoid the issue where N*frac(smt_power) >= 1 creates 'phantom' cores by * first dividing out the smt factor and computing the actual number of cores * and limit power unit capacity with that. |
b37d93168 sched/fair: Rewor... |
5462 5463 5464 |
*/ static inline int sg_capacity(struct lb_env *env, struct sched_group *group) { |
c61037e90 sched/fair: Fix t... |
5465 5466 5467 5468 5469 5470 |
unsigned int capacity, smt, cpus; unsigned int power, power_orig; power = group->sgp->power; power_orig = group->sgp->power_orig; cpus = group->group_weight; |
b37d93168 sched/fair: Rewor... |
5471 |
|
c61037e90 sched/fair: Fix t... |
5472 5473 5474 |
/* smt := ceil(cpus / power), assumes: 1 < smt_power < 2 */ smt = DIV_ROUND_UP(SCHED_POWER_SCALE * cpus, power_orig); capacity = cpus / smt; /* cores */ |
b37d93168 sched/fair: Rewor... |
5475 |
|
c61037e90 sched/fair: Fix t... |
5476 |
capacity = min_t(unsigned, capacity, DIV_ROUND_CLOSEST(power, SCHED_POWER_SCALE)); |
b37d93168 sched/fair: Rewor... |
5477 5478 5479 5480 5481 |
if (!capacity) capacity = fix_small_capacity(env->sd, group); return capacity; } |
1e3c88bde sched: Move load ... |
5482 5483 |
/** * update_sg_lb_stats - Update sched_group's statistics for load balancing. |
cd96891d4 sched/fair: fix l... |
5484 |
* @env: The load balancing environment. |
1e3c88bde sched: Move load ... |
5485 |
* @group: sched_group whose statistics are to be updated. |
1e3c88bde sched: Move load ... |
5486 |
* @load_idx: Load index of sched_domain of this_cpu for load calc. |
1e3c88bde sched: Move load ... |
5487 |
* @local_group: Does group contain this_cpu. |
1e3c88bde sched: Move load ... |
5488 5489 |
* @sgs: variable to hold the statistics for this group. */ |
bd939f45d sched/fair: Propa... |
5490 5491 |
static inline void update_sg_lb_stats(struct lb_env *env, struct sched_group *group, int load_idx, |
23f0d2093 sched: Factor out... |
5492 |
int local_group, struct sg_lb_stats *sgs) |
1e3c88bde sched: Move load ... |
5493 |
{ |
30ce5dabc sched/fair: Rewor... |
5494 |
unsigned long load; |
bd939f45d sched/fair: Propa... |
5495 |
int i; |
1e3c88bde sched: Move load ... |
5496 |
|
b72ff13ce sched/fair: Reduc... |
5497 |
memset(sgs, 0, sizeof(*sgs)); |
b9403130a sched/cleanups: A... |
5498 |
for_each_cpu_and(i, sched_group_cpus(group), env->cpus) { |
1e3c88bde sched: Move load ... |
5499 |
struct rq *rq = cpu_rq(i); |
1e3c88bde sched: Move load ... |
5500 |
/* Bias balancing toward cpus of our domain */ |
6263322c5 sched/fair: Rewri... |
5501 |
if (local_group) |
04f733b4a sched/fair: Rever... |
5502 |
load = target_load(i, load_idx); |
6263322c5 sched/fair: Rewri... |
5503 |
else |
1e3c88bde sched: Move load ... |
5504 |
load = source_load(i, load_idx); |
1e3c88bde sched: Move load ... |
5505 5506 |
sgs->group_load += load; |
380c9077b sched/fair: Clean... |
5507 |
sgs->sum_nr_running += rq->nr_running; |
0ec8aa00f sched/numa: Avoid... |
5508 5509 5510 5511 |
#ifdef CONFIG_NUMA_BALANCING sgs->nr_numa_running += rq->nr_numa_running; sgs->nr_preferred_running += rq->nr_preferred_running; #endif |
1e3c88bde sched: Move load ... |
5512 |
sgs->sum_weighted_load += weighted_cpuload(i); |
aae6d3ddd sched: Use group ... |
5513 5514 |
if (idle_cpu(i)) sgs->idle_cpus++; |
1e3c88bde sched: Move load ... |
5515 |
} |
1e3c88bde sched: Move load ... |
5516 |
/* Adjust by relative CPU power of the group */ |
3ae11c90f sched/fair: Make ... |
5517 5518 |
sgs->group_power = group->sgp->power; sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / sgs->group_power; |
1e3c88bde sched: Move load ... |
5519 |
|
dd5feea14 sched: Fix SCHED_... |
5520 |
if (sgs->sum_nr_running) |
38d0f7708 sched/fair: Remov... |
5521 |
sgs->load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running; |
1e3c88bde sched: Move load ... |
5522 |
|
aae6d3ddd sched: Use group ... |
5523 |
sgs->group_weight = group->group_weight; |
fab476228 sched: Force bala... |
5524 |
|
b37d93168 sched/fair: Rewor... |
5525 5526 |
sgs->group_imb = sg_imbalanced(group); sgs->group_capacity = sg_capacity(env, group); |
fab476228 sched: Force bala... |
5527 5528 |
if (sgs->group_capacity > sgs->sum_nr_running) sgs->group_has_capacity = 1; |
1e3c88bde sched: Move load ... |
5529 5530 5531 |
} /** |
532cb4c40 sched: Add asymme... |
5532 |
* update_sd_pick_busiest - return 1 on busiest group |
cd96891d4 sched/fair: fix l... |
5533 |
* @env: The load balancing environment. |
532cb4c40 sched: Add asymme... |
5534 5535 |
* @sds: sched_domain statistics * @sg: sched_group candidate to be checked for being the busiest |
b6b122944 sched: Fix commen... |
5536 |
* @sgs: sched_group statistics |
532cb4c40 sched: Add asymme... |
5537 5538 5539 |
* * Determine if @sg is a busier group than the previously selected * busiest group. |
e69f61862 sched: Fix some k... |
5540 5541 5542 |
* * Return: %true if @sg is a busier group than the previously selected * busiest group. %false otherwise. |
532cb4c40 sched: Add asymme... |
5543 |
*/ |
bd939f45d sched/fair: Propa... |
5544 |
static bool update_sd_pick_busiest(struct lb_env *env, |
532cb4c40 sched: Add asymme... |
5545 5546 |
struct sd_lb_stats *sds, struct sched_group *sg, |
bd939f45d sched/fair: Propa... |
5547 |
struct sg_lb_stats *sgs) |
532cb4c40 sched: Add asymme... |
5548 |
{ |
56cf515b4 sched: Clean-up s... |
5549 |
if (sgs->avg_load <= sds->busiest_stat.avg_load) |
532cb4c40 sched: Add asymme... |
5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 |
return false; if (sgs->sum_nr_running > sgs->group_capacity) return true; if (sgs->group_imb) return true; /* * ASYM_PACKING needs to move all the work to the lowest * numbered CPUs in the group, therefore mark all groups * higher than ourself as busy. */ |
bd939f45d sched/fair: Propa... |
5563 5564 |
if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running && env->dst_cpu < group_first_cpu(sg)) { |
532cb4c40 sched: Add asymme... |
5565 5566 5567 5568 5569 5570 5571 5572 5573 |
if (!sds->busiest) return true; if (group_first_cpu(sds->busiest) > group_first_cpu(sg)) return true; } return false; } |
0ec8aa00f sched/numa: Avoid... |
5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 |
#ifdef CONFIG_NUMA_BALANCING static inline enum fbq_type fbq_classify_group(struct sg_lb_stats *sgs) { if (sgs->sum_nr_running > sgs->nr_numa_running) return regular; if (sgs->sum_nr_running > sgs->nr_preferred_running) return remote; return all; } static inline enum fbq_type fbq_classify_rq(struct rq *rq) { if (rq->nr_running > rq->nr_numa_running) return regular; if (rq->nr_running > rq->nr_preferred_running) return remote; return all; } #else static inline enum fbq_type fbq_classify_group(struct sg_lb_stats *sgs) { return all; } static inline enum fbq_type fbq_classify_rq(struct rq *rq) { return regular; } #endif /* CONFIG_NUMA_BALANCING */ |
532cb4c40 sched: Add asymme... |
5603 |
/** |
461819ac8 sched_fair: Fix a... |
5604 |
* update_sd_lb_stats - Update sched_domain's statistics for load balancing. |
cd96891d4 sched/fair: fix l... |
5605 |
* @env: The load balancing environment. |
1e3c88bde sched: Move load ... |
5606 5607 |
* @sds: variable to hold the statistics for this sched_domain. */ |
0ec8aa00f sched/numa: Avoid... |
5608 |
static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sds) |
1e3c88bde sched: Move load ... |
5609 |
{ |
bd939f45d sched/fair: Propa... |
5610 5611 |
struct sched_domain *child = env->sd->child; struct sched_group *sg = env->sd->groups; |
56cf515b4 sched: Clean-up s... |
5612 |
struct sg_lb_stats tmp_sgs; |
1e3c88bde sched: Move load ... |
5613 5614 5615 5616 |
int load_idx, prefer_sibling = 0; if (child && child->flags & SD_PREFER_SIBLING) prefer_sibling = 1; |
bd939f45d sched/fair: Propa... |
5617 |
load_idx = get_sd_load_idx(env->sd, env->idle); |
1e3c88bde sched: Move load ... |
5618 5619 |
do { |
56cf515b4 sched: Clean-up s... |
5620 |
struct sg_lb_stats *sgs = &tmp_sgs; |
1e3c88bde sched: Move load ... |
5621 |
int local_group; |
bd939f45d sched/fair: Propa... |
5622 |
local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg)); |
56cf515b4 sched: Clean-up s... |
5623 5624 5625 |
if (local_group) { sds->local = sg; sgs = &sds->local_stat; |
b72ff13ce sched/fair: Reduc... |
5626 5627 5628 5629 |
if (env->idle != CPU_NEWLY_IDLE || time_after_eq(jiffies, sg->sgp->next_update)) update_group_power(env->sd, env->dst_cpu); |
56cf515b4 sched: Clean-up s... |
5630 |
} |
1e3c88bde sched: Move load ... |
5631 |
|
56cf515b4 sched: Clean-up s... |
5632 |
update_sg_lb_stats(env, sg, load_idx, local_group, sgs); |
1e3c88bde sched: Move load ... |
5633 |
|
b72ff13ce sched/fair: Reduc... |
5634 5635 |
if (local_group) goto next_group; |
1e3c88bde sched: Move load ... |
5636 5637 |
/* * In case the child domain prefers tasks go to siblings |
532cb4c40 sched: Add asymme... |
5638 |
* first, lower the sg capacity to one so that we'll try |
75dd321d7 sched: Drop group... |
5639 5640 5641 5642 5643 5644 |
* and move all the excess tasks away. We lower the capacity * of a group only if the local group has the capacity to fit * these excess tasks, i.e. nr_running < group_capacity. The * extra check prevents the case where you always pull from the * heaviest group when it is already under-utilized (possible * with a large weight task outweighs the tasks on the system). |
1e3c88bde sched: Move load ... |
5645 |
*/ |
b72ff13ce sched/fair: Reduc... |
5646 5647 |
if (prefer_sibling && sds->local && sds->local_stat.group_has_capacity) |
147c5fc2b sched/fair: Shrin... |
5648 |
sgs->group_capacity = min(sgs->group_capacity, 1U); |
1e3c88bde sched: Move load ... |
5649 |
|
b72ff13ce sched/fair: Reduc... |
5650 |
if (update_sd_pick_busiest(env, sds, sg, sgs)) { |
532cb4c40 sched: Add asymme... |
5651 |
sds->busiest = sg; |
56cf515b4 sched: Clean-up s... |
5652 |
sds->busiest_stat = *sgs; |
1e3c88bde sched: Move load ... |
5653 |
} |
b72ff13ce sched/fair: Reduc... |
5654 5655 5656 5657 |
next_group: /* Now, start updating sd_lb_stats */ sds->total_load += sgs->group_load; sds->total_pwr += sgs->group_power; |
532cb4c40 sched: Add asymme... |
5658 |
sg = sg->next; |
bd939f45d sched/fair: Propa... |
5659 |
} while (sg != env->sd->groups); |
0ec8aa00f sched/numa: Avoid... |
5660 5661 5662 |
if (env->sd->flags & SD_NUMA) env->fbq_type = fbq_classify_group(&sds->busiest_stat); |
532cb4c40 sched: Add asymme... |
5663 |
} |
532cb4c40 sched: Add asymme... |
5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 |
/** * check_asym_packing - Check to see if the group is packed into the * sched doman. * * This is primarily intended to used at the sibling level. Some * cores like POWER7 prefer to use lower numbered SMT threads. In the * case of POWER7, it can move to lower SMT modes only when higher * threads are idle. When in lower SMT modes, the threads will * perform better since they share less core resources. Hence when we * have idle threads, we want them to be the higher ones. * * This packing function is run on idle threads. It checks to see if * the busiest CPU in this domain (core in the P7 case) has a higher * CPU number than the packing function is being run on. Here we are * assuming lower CPU number will be equivalent to lower a SMT thread * number. * |
e69f61862 sched: Fix some k... |
5681 |
* Return: 1 when packing is required and a task should be moved to |
b6b122944 sched: Fix commen... |
5682 5683 |
* this CPU. The amount of the imbalance is returned in *imbalance. * |
cd96891d4 sched/fair: fix l... |
5684 |
* @env: The load balancing environment. |
532cb4c40 sched: Add asymme... |
5685 |
* @sds: Statistics of the sched_domain which is to be packed |
532cb4c40 sched: Add asymme... |
5686 |
*/ |
bd939f45d sched/fair: Propa... |
5687 |
static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds) |
532cb4c40 sched: Add asymme... |
5688 5689 |
{ int busiest_cpu; |
bd939f45d sched/fair: Propa... |
5690 |
if (!(env->sd->flags & SD_ASYM_PACKING)) |
532cb4c40 sched: Add asymme... |
5691 5692 5693 5694 5695 5696 |
return 0; if (!sds->busiest) return 0; busiest_cpu = group_first_cpu(sds->busiest); |
bd939f45d sched/fair: Propa... |
5697 |
if (env->dst_cpu > busiest_cpu) |
532cb4c40 sched: Add asymme... |
5698 |
return 0; |
bd939f45d sched/fair: Propa... |
5699 |
env->imbalance = DIV_ROUND_CLOSEST( |
3ae11c90f sched/fair: Make ... |
5700 5701 |
sds->busiest_stat.avg_load * sds->busiest_stat.group_power, SCHED_POWER_SCALE); |
bd939f45d sched/fair: Propa... |
5702 |
|
532cb4c40 sched: Add asymme... |
5703 |
return 1; |
1e3c88bde sched: Move load ... |
5704 5705 5706 5707 5708 5709 |
} /** * fix_small_imbalance - Calculate the minor imbalance that exists * amongst the groups of a sched_domain, during * load balancing. |
cd96891d4 sched/fair: fix l... |
5710 |
* @env: The load balancing environment. |
1e3c88bde sched: Move load ... |
5711 |
* @sds: Statistics of the sched_domain whose imbalance is to be calculated. |
1e3c88bde sched: Move load ... |
5712 |
*/ |
bd939f45d sched/fair: Propa... |
5713 5714 |
static inline void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds) |
1e3c88bde sched: Move load ... |
5715 5716 5717 |
{ unsigned long tmp, pwr_now = 0, pwr_move = 0; unsigned int imbn = 2; |
dd5feea14 sched: Fix SCHED_... |
5718 |
unsigned long scaled_busy_load_per_task; |
56cf515b4 sched: Clean-up s... |
5719 |
struct sg_lb_stats *local, *busiest; |
1e3c88bde sched: Move load ... |
5720 |
|
56cf515b4 sched: Clean-up s... |
5721 5722 |
local = &sds->local_stat; busiest = &sds->busiest_stat; |
1e3c88bde sched: Move load ... |
5723 |
|
56cf515b4 sched: Clean-up s... |
5724 5725 5726 5727 |
if (!local->sum_nr_running) local->load_per_task = cpu_avg_load_per_task(env->dst_cpu); else if (busiest->load_per_task > local->load_per_task) imbn = 1; |
dd5feea14 sched: Fix SCHED_... |
5728 |
|
56cf515b4 sched: Clean-up s... |
5729 5730 |
scaled_busy_load_per_task = (busiest->load_per_task * SCHED_POWER_SCALE) / |
3ae11c90f sched/fair: Make ... |
5731 |
busiest->group_power; |
56cf515b4 sched: Clean-up s... |
5732 |
|
3029ede39 sched/balancing: ... |
5733 5734 |
if (busiest->avg_load + scaled_busy_load_per_task >= local->avg_load + (scaled_busy_load_per_task * imbn)) { |
56cf515b4 sched: Clean-up s... |
5735 |
env->imbalance = busiest->load_per_task; |
1e3c88bde sched: Move load ... |
5736 5737 5738 5739 5740 5741 5742 5743 |
return; } /* * OK, we don't have enough imbalance to justify moving tasks, * however we may be able to increase total CPU power used by * moving them. */ |
3ae11c90f sched/fair: Make ... |
5744 |
pwr_now += busiest->group_power * |
56cf515b4 sched: Clean-up s... |
5745 |
min(busiest->load_per_task, busiest->avg_load); |
3ae11c90f sched/fair: Make ... |
5746 |
pwr_now += local->group_power * |
56cf515b4 sched: Clean-up s... |
5747 |
min(local->load_per_task, local->avg_load); |
1399fa780 sched: Introduce ... |
5748 |
pwr_now /= SCHED_POWER_SCALE; |
1e3c88bde sched: Move load ... |
5749 5750 |
/* Amount of load we'd subtract */ |
a2cd42601 sched: Remove dou... |
5751 |
if (busiest->avg_load > scaled_busy_load_per_task) { |
3ae11c90f sched/fair: Make ... |
5752 |
pwr_move += busiest->group_power * |
56cf515b4 sched: Clean-up s... |
5753 |
min(busiest->load_per_task, |
a2cd42601 sched: Remove dou... |
5754 |
busiest->avg_load - scaled_busy_load_per_task); |
56cf515b4 sched: Clean-up s... |
5755 |
} |
1e3c88bde sched: Move load ... |
5756 5757 |
/* Amount of load we'd add */ |
3ae11c90f sched/fair: Make ... |
5758 |
if (busiest->avg_load * busiest->group_power < |
56cf515b4 sched: Clean-up s... |
5759 |
busiest->load_per_task * SCHED_POWER_SCALE) { |
3ae11c90f sched/fair: Make ... |
5760 5761 |
tmp = (busiest->avg_load * busiest->group_power) / local->group_power; |
56cf515b4 sched: Clean-up s... |
5762 5763 |
} else { tmp = (busiest->load_per_task * SCHED_POWER_SCALE) / |
3ae11c90f sched/fair: Make ... |
5764 |
local->group_power; |
56cf515b4 sched: Clean-up s... |
5765 |
} |
3ae11c90f sched/fair: Make ... |
5766 5767 |
pwr_move += local->group_power * min(local->load_per_task, local->avg_load + tmp); |
1399fa780 sched: Introduce ... |
5768 |
pwr_move /= SCHED_POWER_SCALE; |
1e3c88bde sched: Move load ... |
5769 5770 5771 |
/* Move if we gain throughput */ if (pwr_move > pwr_now) |
56cf515b4 sched: Clean-up s... |
5772 |
env->imbalance = busiest->load_per_task; |
1e3c88bde sched: Move load ... |
5773 5774 5775 5776 5777 |
} /** * calculate_imbalance - Calculate the amount of imbalance present within the * groups of a given sched_domain during load balance. |
bd939f45d sched/fair: Propa... |
5778 |
* @env: load balance environment |
1e3c88bde sched: Move load ... |
5779 |
* @sds: statistics of the sched_domain whose imbalance is to be calculated. |
1e3c88bde sched: Move load ... |
5780 |
*/ |
bd939f45d sched/fair: Propa... |
5781 |
static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds) |
1e3c88bde sched: Move load ... |
5782 |
{ |
dd5feea14 sched: Fix SCHED_... |
5783 |
unsigned long max_pull, load_above_capacity = ~0UL; |
56cf515b4 sched: Clean-up s... |
5784 5785 5786 |
struct sg_lb_stats *local, *busiest; local = &sds->local_stat; |
56cf515b4 sched: Clean-up s... |
5787 |
busiest = &sds->busiest_stat; |
dd5feea14 sched: Fix SCHED_... |
5788 |
|
56cf515b4 sched: Clean-up s... |
5789 |
if (busiest->group_imb) { |
30ce5dabc sched/fair: Rewor... |
5790 5791 5792 5793 |
/* * In the group_imb case we cannot rely on group-wide averages * to ensure cpu-load equilibrium, look at wider averages. XXX */ |
56cf515b4 sched: Clean-up s... |
5794 5795 |
busiest->load_per_task = min(busiest->load_per_task, sds->avg_load); |
dd5feea14 sched: Fix SCHED_... |
5796 |
} |
1e3c88bde sched: Move load ... |
5797 5798 5799 5800 5801 |
/* * In the presence of smp nice balancing, certain scenarios can have * max load less than avg load(as we skip the groups at or below * its cpu_power, while calculating max_load..) */ |
b18855500 sched/balancing: ... |
5802 5803 |
if (busiest->avg_load <= sds->avg_load || local->avg_load >= sds->avg_load) { |
bd939f45d sched/fair: Propa... |
5804 5805 |
env->imbalance = 0; return fix_small_imbalance(env, sds); |
1e3c88bde sched: Move load ... |
5806 |
} |
56cf515b4 sched: Clean-up s... |
5807 |
if (!busiest->group_imb) { |
dd5feea14 sched: Fix SCHED_... |
5808 5809 |
/* * Don't want to pull so many tasks that a group would go idle. |
30ce5dabc sched/fair: Rewor... |
5810 5811 |
* Except of course for the group_imb case, since then we might * have to drop below capacity to reach cpu-load equilibrium. |
dd5feea14 sched: Fix SCHED_... |
5812 |
*/ |
56cf515b4 sched: Clean-up s... |
5813 5814 |
load_above_capacity = (busiest->sum_nr_running - busiest->group_capacity); |
dd5feea14 sched: Fix SCHED_... |
5815 |
|
1399fa780 sched: Introduce ... |
5816 |
load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE); |
3ae11c90f sched/fair: Make ... |
5817 |
load_above_capacity /= busiest->group_power; |
dd5feea14 sched: Fix SCHED_... |
5818 5819 5820 5821 5822 5823 5824 5825 5826 |
} /* * We're trying to get all the cpus to the average_load, so we don't * want to push ourselves above the average load, nor do we wish to * reduce the max loaded cpu below the average load. At the same time, * we also don't want to reduce the group load below the group capacity * (so that we can implement power-savings policies etc). Thus we look * for the minimum possible imbalance. |
dd5feea14 sched: Fix SCHED_... |
5827 |
*/ |
30ce5dabc sched/fair: Rewor... |
5828 |
max_pull = min(busiest->avg_load - sds->avg_load, load_above_capacity); |
1e3c88bde sched: Move load ... |
5829 5830 |
/* How much load to actually move to equalise the imbalance */ |
56cf515b4 sched: Clean-up s... |
5831 |
env->imbalance = min( |
3ae11c90f sched/fair: Make ... |
5832 5833 |
max_pull * busiest->group_power, (sds->avg_load - local->avg_load) * local->group_power |
56cf515b4 sched: Clean-up s... |
5834 |
) / SCHED_POWER_SCALE; |
1e3c88bde sched: Move load ... |
5835 5836 5837 |
/* * if *imbalance is less than the average load per runnable task |
25985edce Fix common misspe... |
5838 |
* there is no guarantee that any tasks will be moved so we'll have |
1e3c88bde sched: Move load ... |
5839 5840 5841 |
* a think about bumping its value to force at least one task to be * moved */ |
56cf515b4 sched: Clean-up s... |
5842 |
if (env->imbalance < busiest->load_per_task) |
bd939f45d sched/fair: Propa... |
5843 |
return fix_small_imbalance(env, sds); |
1e3c88bde sched: Move load ... |
5844 |
} |
fab476228 sched: Force bala... |
5845 |
|
1e3c88bde sched: Move load ... |
5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 |
/******* find_busiest_group() helpers end here *********************/ /** * find_busiest_group - Returns the busiest group within the sched_domain * if there is an imbalance. If there isn't an imbalance, and * the user has opted for power-savings, it returns a group whose * CPUs can be put to idle by rebalancing those tasks elsewhere, if * such a group exists. * * Also calculates the amount of weighted load which should be moved * to restore balance. * |
cd96891d4 sched/fair: fix l... |
5858 |
* @env: The load balancing environment. |
1e3c88bde sched: Move load ... |
5859 |
* |
e69f61862 sched: Fix some k... |
5860 |
* Return: - The busiest group if imbalance exists. |
1e3c88bde sched: Move load ... |
5861 5862 5863 5864 |
* - If no imbalance and user has opted for power-savings balance, * return the least loaded group whose CPUs can be * put to idle by rebalancing its tasks onto our group. */ |
56cf515b4 sched: Clean-up s... |
5865 |
static struct sched_group *find_busiest_group(struct lb_env *env) |
1e3c88bde sched: Move load ... |
5866 |
{ |
56cf515b4 sched: Clean-up s... |
5867 |
struct sg_lb_stats *local, *busiest; |
1e3c88bde sched: Move load ... |
5868 |
struct sd_lb_stats sds; |
147c5fc2b sched/fair: Shrin... |
5869 |
init_sd_lb_stats(&sds); |
1e3c88bde sched: Move load ... |
5870 5871 5872 5873 5874 |
/* * Compute the various statistics relavent for load balancing at * this level. */ |
23f0d2093 sched: Factor out... |
5875 |
update_sd_lb_stats(env, &sds); |
56cf515b4 sched: Clean-up s... |
5876 5877 |
local = &sds.local_stat; busiest = &sds.busiest_stat; |
1e3c88bde sched: Move load ... |
5878 |
|
bd939f45d sched/fair: Propa... |
5879 5880 |
if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) && check_asym_packing(env, &sds)) |
532cb4c40 sched: Add asymme... |
5881 |
return sds.busiest; |
cc57aa8f4 sched: Clean up s... |
5882 |
/* There is no busy sibling group to pull tasks from */ |
56cf515b4 sched: Clean-up s... |
5883 |
if (!sds.busiest || busiest->sum_nr_running == 0) |
1e3c88bde sched: Move load ... |
5884 |
goto out_balanced; |
1399fa780 sched: Introduce ... |
5885 |
sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr; |
b0432d8f1 sched: Fix sched-... |
5886 |
|
866ab43ef sched: Fix the gr... |
5887 5888 |
/* * If the busiest group is imbalanced the below checks don't |
30ce5dabc sched/fair: Rewor... |
5889 |
* work because they assume all things are equal, which typically |
866ab43ef sched: Fix the gr... |
5890 5891 |
* isn't true due to cpus_allowed constraints and the like. */ |
56cf515b4 sched: Clean-up s... |
5892 |
if (busiest->group_imb) |
866ab43ef sched: Fix the gr... |
5893 |
goto force_balance; |
cc57aa8f4 sched: Clean up s... |
5894 |
/* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */ |
56cf515b4 sched: Clean-up s... |
5895 5896 |
if (env->idle == CPU_NEWLY_IDLE && local->group_has_capacity && !busiest->group_has_capacity) |
fab476228 sched: Force bala... |
5897 |
goto force_balance; |
cc57aa8f4 sched: Clean up s... |
5898 5899 5900 5901 |
/* * If the local group is more busy than the selected busiest group * don't try and pull any tasks. */ |
56cf515b4 sched: Clean-up s... |
5902 |
if (local->avg_load >= busiest->avg_load) |
1e3c88bde sched: Move load ... |
5903 |
goto out_balanced; |
cc57aa8f4 sched: Clean up s... |
5904 5905 5906 5907 |
/* * Don't pull any tasks if this group is already above the domain * average load. */ |
56cf515b4 sched: Clean-up s... |
5908 |
if (local->avg_load >= sds.avg_load) |
1e3c88bde sched: Move load ... |
5909 |
goto out_balanced; |
bd939f45d sched/fair: Propa... |
5910 |
if (env->idle == CPU_IDLE) { |
aae6d3ddd sched: Use group ... |
5911 5912 5913 5914 5915 5916 |
/* * This cpu is idle. If the busiest group load doesn't * have more tasks than the number of available cpu's and * there is no imbalance between this and busiest group * wrt to idle cpu's, it is balanced. */ |
56cf515b4 sched: Clean-up s... |
5917 5918 |
if ((local->idle_cpus < busiest->idle_cpus) && busiest->sum_nr_running <= busiest->group_weight) |
aae6d3ddd sched: Use group ... |
5919 |
goto out_balanced; |
c186fafe9 sched: Clean up r... |
5920 5921 5922 5923 5924 |
} else { /* * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use * imbalance_pct to be conservative. */ |
56cf515b4 sched: Clean-up s... |
5925 5926 |
if (100 * busiest->avg_load <= env->sd->imbalance_pct * local->avg_load) |
c186fafe9 sched: Clean up r... |
5927 |
goto out_balanced; |
aae6d3ddd sched: Use group ... |
5928 |
} |
1e3c88bde sched: Move load ... |
5929 |
|
fab476228 sched: Force bala... |
5930 |
force_balance: |
1e3c88bde sched: Move load ... |
5931 |
/* Looks like there is an imbalance. Compute it */ |
bd939f45d sched/fair: Propa... |
5932 |
calculate_imbalance(env, &sds); |
1e3c88bde sched: Move load ... |
5933 5934 5935 |
return sds.busiest; out_balanced: |
bd939f45d sched/fair: Propa... |
5936 |
env->imbalance = 0; |
1e3c88bde sched: Move load ... |
5937 5938 5939 5940 5941 5942 |
return NULL; } /* * find_busiest_queue - find the busiest runqueue among the cpus in group. */ |
bd939f45d sched/fair: Propa... |
5943 |
static struct rq *find_busiest_queue(struct lb_env *env, |
b9403130a sched/cleanups: A... |
5944 |
struct sched_group *group) |
1e3c88bde sched: Move load ... |
5945 5946 |
{ struct rq *busiest = NULL, *rq; |
95a79b805 sched: Remove one... |
5947 |
unsigned long busiest_load = 0, busiest_power = 1; |
1e3c88bde sched: Move load ... |
5948 |
int i; |
6906a4083 sched/fair: Optim... |
5949 |
for_each_cpu_and(i, sched_group_cpus(group), env->cpus) { |
0ec8aa00f sched/numa: Avoid... |
5950 5951 5952 5953 5954 |
unsigned long power, capacity, wl; enum fbq_type rt; rq = cpu_rq(i); rt = fbq_classify_rq(rq); |
1e3c88bde sched: Move load ... |
5955 |
|
0ec8aa00f sched/numa: Avoid... |
5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 |
/* * We classify groups/runqueues into three groups: * - regular: there are !numa tasks * - remote: there are numa tasks that run on the 'wrong' node * - all: there is no distinction * * In order to avoid migrating ideally placed numa tasks, * ignore those when there's better options. * * If we ignore the actual busiest queue to migrate another * task, the next balance pass can still reduce the busiest * queue by moving tasks around inside the node. * * If we cannot move enough load due to this classification * the next pass will adjust the group classification and * allow migration of more tasks. * * Both cases only affect the total convergence complexity. */ if (rt > env->fbq_type) continue; power = power_of(i); capacity = DIV_ROUND_CLOSEST(power, SCHED_POWER_SCALE); |
9d5efe05e sched: Fix capaci... |
5980 |
if (!capacity) |
bd939f45d sched/fair: Propa... |
5981 |
capacity = fix_small_capacity(env->sd, group); |
9d5efe05e sched: Fix capaci... |
5982 |
|
6e40f5bbb Merge branch 'sch... |
5983 |
wl = weighted_cpuload(i); |
1e3c88bde sched: Move load ... |
5984 |
|
6e40f5bbb Merge branch 'sch... |
5985 5986 5987 5988 |
/* * When comparing with imbalance, use weighted_cpuload() * which is not scaled with the cpu power. */ |
bd939f45d sched/fair: Propa... |
5989 |
if (capacity && rq->nr_running == 1 && wl > env->imbalance) |
1e3c88bde sched: Move load ... |
5990 |
continue; |
6e40f5bbb Merge branch 'sch... |
5991 5992 5993 5994 5995 |
/* * For the load comparisons with the other cpu's, consider * the weighted_cpuload() scaled with the cpu power, so that * the load can be moved away from the cpu that is potentially * running at a lower capacity. |
95a79b805 sched: Remove one... |
5996 5997 5998 5999 6000 |
* * Thus we're looking for max(wl_i / power_i), crosswise * multiplication to rid ourselves of the division works out * to: wl_i * power_j > wl_j * power_i; where j is our * previous maximum. |
6e40f5bbb Merge branch 'sch... |
6001 |
*/ |
95a79b805 sched: Remove one... |
6002 6003 6004 |
if (wl * busiest_power > busiest_load * power) { busiest_load = wl; busiest_power = power; |
1e3c88bde sched: Move load ... |
6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 |
busiest = rq; } } return busiest; } /* * Max backoff if we encounter pinned tasks. Pretty arbitrary value, but * so long as it is large enough. */ #define MAX_PINNED_INTERVAL 512 /* Working cpumask for load_balance and load_balance_newidle. */ |
e6252c3ef sched: Rename loa... |
6019 |
DEFINE_PER_CPU(cpumask_var_t, load_balance_mask); |
1e3c88bde sched: Move load ... |
6020 |
|
bd939f45d sched/fair: Propa... |
6021 |
static int need_active_balance(struct lb_env *env) |
1af3ed3dd sched: Unify load... |
6022 |
{ |
bd939f45d sched/fair: Propa... |
6023 6024 6025 |
struct sched_domain *sd = env->sd; if (env->idle == CPU_NEWLY_IDLE) { |
532cb4c40 sched: Add asymme... |
6026 6027 6028 6029 6030 6031 |
/* * ASYM_PACKING needs to force migrate tasks from busy but * higher numbered CPUs in order to pack all tasks in the * lowest numbered CPUs. */ |
bd939f45d sched/fair: Propa... |
6032 |
if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu) |
532cb4c40 sched: Add asymme... |
6033 |
return 1; |
1af3ed3dd sched: Unify load... |
6034 6035 6036 6037 |
} return unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2); } |
969c79215 sched: replace mi... |
6038 |
static int active_load_balance_cpu_stop(void *data); |
23f0d2093 sched: Factor out... |
6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 |
static int should_we_balance(struct lb_env *env) { struct sched_group *sg = env->sd->groups; struct cpumask *sg_cpus, *sg_mask; int cpu, balance_cpu = -1; /* * In the newly idle case, we will allow all the cpu's * to do the newly idle load balance. */ if (env->idle == CPU_NEWLY_IDLE) return 1; sg_cpus = sched_group_cpus(sg); sg_mask = sched_group_mask(sg); /* Try to find first idle cpu */ for_each_cpu_and(cpu, sg_cpus, env->cpus) { if (!cpumask_test_cpu(cpu, sg_mask) || !idle_cpu(cpu)) continue; balance_cpu = cpu; break; } if (balance_cpu == -1) balance_cpu = group_balance_cpu(sg); /* * First idle cpu or the first cpu(busiest) in this sched group * is eligible for doing load balancing at this and above domains. */ |
b0cff9d88 sched: Fix load b... |
6070 |
return balance_cpu == env->dst_cpu; |
23f0d2093 sched: Factor out... |
6071 |
} |
1e3c88bde sched: Move load ... |
6072 6073 6074 6075 6076 6077 |
/* * Check this_cpu to ensure it is balanced within domain. Attempt to move * tasks if there is an imbalance. */ static int load_balance(int this_cpu, struct rq *this_rq, struct sched_domain *sd, enum cpu_idle_type idle, |
23f0d2093 sched: Factor out... |
6078 |
int *continue_balancing) |
1e3c88bde sched: Move load ... |
6079 |
{ |
88b8dac0a sched: Improve ba... |
6080 |
int ld_moved, cur_ld_moved, active_balance = 0; |
6263322c5 sched/fair: Rewri... |
6081 |
struct sched_domain *sd_parent = sd->parent; |
1e3c88bde sched: Move load ... |
6082 |
struct sched_group *group; |
1e3c88bde sched: Move load ... |
6083 6084 |
struct rq *busiest; unsigned long flags; |
e6252c3ef sched: Rename loa... |
6085 |
struct cpumask *cpus = __get_cpu_var(load_balance_mask); |
1e3c88bde sched: Move load ... |
6086 |
|
8e45cb545 sched: Move load-... |
6087 6088 |
struct lb_env env = { .sd = sd, |
ddcdf6e7d sched: Rename loa... |
6089 6090 |
.dst_cpu = this_cpu, .dst_rq = this_rq, |
88b8dac0a sched: Improve ba... |
6091 |
.dst_grpmask = sched_group_cpus(sd->groups), |
8e45cb545 sched: Move load-... |
6092 |
.idle = idle, |
eb95308ee sched: Fix more l... |
6093 |
.loop_break = sched_nr_migrate_break, |
b9403130a sched/cleanups: A... |
6094 |
.cpus = cpus, |
0ec8aa00f sched/numa: Avoid... |
6095 |
.fbq_type = all, |
8e45cb545 sched: Move load-... |
6096 |
}; |
cfc031180 sched: Don't cons... |
6097 6098 6099 6100 |
/* * For NEWLY_IDLE load_balancing, we don't need to consider * other cpus in our group */ |
e02e60c10 sched: Prevent to... |
6101 |
if (idle == CPU_NEWLY_IDLE) |
cfc031180 sched: Don't cons... |
6102 |
env.dst_grpmask = NULL; |
cfc031180 sched: Don't cons... |
6103 |
|
1e3c88bde sched: Move load ... |
6104 |
cpumask_copy(cpus, cpu_active_mask); |
1e3c88bde sched: Move load ... |
6105 6106 6107 |
schedstat_inc(sd, lb_count[idle]); redo: |
23f0d2093 sched: Factor out... |
6108 6109 |
if (!should_we_balance(&env)) { *continue_balancing = 0; |
1e3c88bde sched: Move load ... |
6110 |
goto out_balanced; |
23f0d2093 sched: Factor out... |
6111 |
} |
1e3c88bde sched: Move load ... |
6112 |
|
23f0d2093 sched: Factor out... |
6113 |
group = find_busiest_group(&env); |
1e3c88bde sched: Move load ... |
6114 6115 6116 6117 |
if (!group) { schedstat_inc(sd, lb_nobusyg[idle]); goto out_balanced; } |
b9403130a sched/cleanups: A... |
6118 |
busiest = find_busiest_queue(&env, group); |
1e3c88bde sched: Move load ... |
6119 6120 6121 6122 |
if (!busiest) { schedstat_inc(sd, lb_nobusyq[idle]); goto out_balanced; } |
78feefc51 sched: using dst_... |
6123 |
BUG_ON(busiest == env.dst_rq); |
1e3c88bde sched: Move load ... |
6124 |
|
bd939f45d sched/fair: Propa... |
6125 |
schedstat_add(sd, lb_imbalance[idle], env.imbalance); |
1e3c88bde sched: Move load ... |
6126 6127 6128 6129 6130 6131 6132 6133 6134 |
ld_moved = 0; if (busiest->nr_running > 1) { /* * Attempt to move tasks. If find_busiest_group has found * an imbalance but busiest->nr_running <= 1, the group is * still unbalanced. ld_moved simply stays zero, so it is * correctly treated as an imbalance. */ |
8e45cb545 sched: Move load-... |
6135 |
env.flags |= LBF_ALL_PINNED; |
c82513e51 sched: Change rq-... |
6136 6137 6138 |
env.src_cpu = busiest->cpu; env.src_rq = busiest; env.loop_max = min(sysctl_sched_nr_migrate, busiest->nr_running); |
8e45cb545 sched: Move load-... |
6139 |
|
5d6523ebd sched: Fix load-b... |
6140 |
more_balance: |
1e3c88bde sched: Move load ... |
6141 |
local_irq_save(flags); |
78feefc51 sched: using dst_... |
6142 |
double_rq_lock(env.dst_rq, busiest); |
88b8dac0a sched: Improve ba... |
6143 6144 6145 6146 6147 6148 6149 |
/* * cur_ld_moved - load moved in current iteration * ld_moved - cumulative load moved across iterations */ cur_ld_moved = move_tasks(&env); ld_moved += cur_ld_moved; |
78feefc51 sched: using dst_... |
6150 |
double_rq_unlock(env.dst_rq, busiest); |
1e3c88bde sched: Move load ... |
6151 6152 6153 6154 6155 |
local_irq_restore(flags); /* * some other cpu did the load balance for us. */ |
88b8dac0a sched: Improve ba... |
6156 6157 |
if (cur_ld_moved && env.dst_cpu != smp_processor_id()) resched_cpu(env.dst_cpu); |
f1cd08581 sched: Change pos... |
6158 6159 6160 6161 |
if (env.flags & LBF_NEED_BREAK) { env.flags &= ~LBF_NEED_BREAK; goto more_balance; } |
88b8dac0a sched: Improve ba... |
6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178 6179 6180 |
/* * Revisit (affine) tasks on src_cpu that couldn't be moved to * us and move them to an alternate dst_cpu in our sched_group * where they can run. The upper limit on how many times we * iterate on same src_cpu is dependent on number of cpus in our * sched_group. * * This changes load balance semantics a bit on who can move * load to a given_cpu. In addition to the given_cpu itself * (or a ilb_cpu acting on its behalf where given_cpu is * nohz-idle), we now have balance_cpu in a position to move * load to given_cpu. In rare situations, this may cause * conflicts (balance_cpu and given_cpu/ilb_cpu deciding * _independently_ and at _same_ time to move some load to * given_cpu) causing exceess load to be moved to given_cpu. * This however should not happen so much in practice and * moreover subsequent load balance cycles should correct the * excess load moved. */ |
6263322c5 sched/fair: Rewri... |
6181 |
if ((env.flags & LBF_DST_PINNED) && env.imbalance > 0) { |
88b8dac0a sched: Improve ba... |
6182 |
|
7aff2e3a5 sched/balancing: ... |
6183 6184 |
/* Prevent to re-select dst_cpu via env's cpus */ cpumask_clear_cpu(env.dst_cpu, env.cpus); |
78feefc51 sched: using dst_... |
6185 |
env.dst_rq = cpu_rq(env.new_dst_cpu); |
88b8dac0a sched: Improve ba... |
6186 |
env.dst_cpu = env.new_dst_cpu; |
6263322c5 sched/fair: Rewri... |
6187 |
env.flags &= ~LBF_DST_PINNED; |
88b8dac0a sched: Improve ba... |
6188 6189 |
env.loop = 0; env.loop_break = sched_nr_migrate_break; |
e02e60c10 sched: Prevent to... |
6190 |
|
88b8dac0a sched: Improve ba... |
6191 6192 6193 6194 6195 6196 |
/* * Go back to "more_balance" rather than "redo" since we * need to continue with same src_cpu. */ goto more_balance; } |
1e3c88bde sched: Move load ... |
6197 |
|
6263322c5 sched/fair: Rewri... |
6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 |
/* * We failed to reach balance because of affinity. */ if (sd_parent) { int *group_imbalance = &sd_parent->groups->sgp->imbalance; if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0) { *group_imbalance = 1; } else if (*group_imbalance) *group_imbalance = 0; } |
1e3c88bde sched: Move load ... |
6209 |
/* All tasks on this runqueue were pinned by CPU affinity */ |
8e45cb545 sched: Move load-... |
6210 |
if (unlikely(env.flags & LBF_ALL_PINNED)) { |
1e3c88bde sched: Move load ... |
6211 |
cpumask_clear_cpu(cpu_of(busiest), cpus); |
bbf18b194 sched: Reset loop... |
6212 6213 6214 |
if (!cpumask_empty(cpus)) { env.loop = 0; env.loop_break = sched_nr_migrate_break; |
1e3c88bde sched: Move load ... |
6215 |
goto redo; |
bbf18b194 sched: Reset loop... |
6216 |
} |
1e3c88bde sched: Move load ... |
6217 6218 6219 6220 6221 6222 |
goto out_balanced; } } if (!ld_moved) { schedstat_inc(sd, lb_failed[idle]); |
58b26c4c0 sched: Increment ... |
6223 6224 6225 6226 6227 6228 6229 6230 |
/* * Increment the failure counter only on periodic balance. * We do not want newidle balance, which can be very * frequent, pollute the failure counter causing * excessive cache_hot migrations and active balances. */ if (idle != CPU_NEWLY_IDLE) sd->nr_balance_failed++; |
1e3c88bde sched: Move load ... |
6231 |
|
bd939f45d sched/fair: Propa... |
6232 |
if (need_active_balance(&env)) { |
1e3c88bde sched: Move load ... |
6233 |
raw_spin_lock_irqsave(&busiest->lock, flags); |
969c79215 sched: replace mi... |
6234 6235 6236 |
/* don't kick the active_load_balance_cpu_stop, * if the curr task on busiest cpu can't be * moved to this_cpu |
1e3c88bde sched: Move load ... |
6237 6238 |
*/ if (!cpumask_test_cpu(this_cpu, |
fa17b507f sched: Wrap sched... |
6239 |
tsk_cpus_allowed(busiest->curr))) { |
1e3c88bde sched: Move load ... |
6240 6241 |
raw_spin_unlock_irqrestore(&busiest->lock, flags); |
8e45cb545 sched: Move load-... |
6242 |
env.flags |= LBF_ALL_PINNED; |
1e3c88bde sched: Move load ... |
6243 6244 |
goto out_one_pinned; } |
969c79215 sched: replace mi... |
6245 6246 6247 6248 6249 |
/* * ->active_balance synchronizes accesses to * ->active_balance_work. Once set, it's cleared * only after active load balance is finished. */ |
1e3c88bde sched: Move load ... |
6250 6251 6252 6253 6254 6255 |
if (!busiest->active_balance) { busiest->active_balance = 1; busiest->push_cpu = this_cpu; active_balance = 1; } raw_spin_unlock_irqrestore(&busiest->lock, flags); |
969c79215 sched: replace mi... |
6256 |
|
bd939f45d sched/fair: Propa... |
6257 |
if (active_balance) { |
969c79215 sched: replace mi... |
6258 6259 6260 |
stop_one_cpu_nowait(cpu_of(busiest), active_load_balance_cpu_stop, busiest, &busiest->active_balance_work); |
bd939f45d sched/fair: Propa... |
6261 |
} |
1e3c88bde sched: Move load ... |
6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 |
/* * We've kicked active balancing, reset the failure * counter. */ sd->nr_balance_failed = sd->cache_nice_tries+1; } } else sd->nr_balance_failed = 0; if (likely(!active_balance)) { /* We were unbalanced, so reset the balancing interval */ sd->balance_interval = sd->min_interval; } else { /* * If we've begun active balancing, start to back off. This * case may not be covered by the all_pinned logic if there * is only 1 task on the busy runqueue (because we don't call * move_tasks). */ if (sd->balance_interval < sd->max_interval) sd->balance_interval *= 2; } |
1e3c88bde sched: Move load ... |
6285 6286 6287 6288 6289 6290 6291 6292 6293 |
goto out; out_balanced: schedstat_inc(sd, lb_balanced[idle]); sd->nr_balance_failed = 0; out_one_pinned: /* tune up the balancing interval */ |
8e45cb545 sched: Move load-... |
6294 |
if (((env.flags & LBF_ALL_PINNED) && |
5b54b56be sched: Replace al... |
6295 |
sd->balance_interval < MAX_PINNED_INTERVAL) || |
1e3c88bde sched: Move load ... |
6296 6297 |
(sd->balance_interval < sd->max_interval)) sd->balance_interval *= 2; |
46e49b383 sched: Wholesale ... |
6298 |
ld_moved = 0; |
1e3c88bde sched: Move load ... |
6299 |
out: |
1e3c88bde sched: Move load ... |
6300 6301 6302 6303 |
return ld_moved; } /* |
1e3c88bde sched: Move load ... |
6304 6305 6306 |
* idle_balance is called by schedule() if this_cpu is about to become * idle. Attempts to pull tasks from other CPUs. */ |
6e83125c6 sched/fair: Remov... |
6307 |
static int idle_balance(struct rq *this_rq) |
1e3c88bde sched: Move load ... |
6308 6309 6310 6311 |
{ struct sched_domain *sd; int pulled_task = 0; unsigned long next_balance = jiffies + HZ; |
9bd721c55 sched/balancing: ... |
6312 |
u64 curr_cost = 0; |
b4f2ab436 sched: Remove 'cp... |
6313 |
int this_cpu = this_rq->cpu; |
1e3c88bde sched: Move load ... |
6314 |
|
6e83125c6 sched/fair: Remov... |
6315 |
idle_enter_fair(this_rq); |
0e5b5337f sched: Fix updati... |
6316 |
|
6e83125c6 sched/fair: Remov... |
6317 6318 6319 6320 6321 |
/* * We must set idle_stamp _before_ calling idle_balance(), such that we * measure the duration of idle_balance() as idle time. */ this_rq->idle_stamp = rq_clock(this_rq); |
1e3c88bde sched: Move load ... |
6322 |
if (this_rq->avg_idle < sysctl_sched_migration_cost) |
6e83125c6 sched/fair: Remov... |
6323 |
goto out; |
1e3c88bde sched: Move load ... |
6324 |
|
f492e12ef sched: Remove loa... |
6325 6326 6327 6328 |
/* * Drop the rq->lock, but keep IRQ/preempt disabled. */ raw_spin_unlock(&this_rq->lock); |
48a167532 sched: Refactor u... |
6329 |
update_blocked_averages(this_cpu); |
dce840a08 sched: Dynamicall... |
6330 |
rcu_read_lock(); |
1e3c88bde sched: Move load ... |
6331 6332 |
for_each_domain(this_cpu, sd) { unsigned long interval; |
23f0d2093 sched: Factor out... |
6333 |
int continue_balancing = 1; |
9bd721c55 sched/balancing: ... |
6334 |
u64 t0, domain_cost; |
1e3c88bde sched: Move load ... |
6335 6336 6337 |
if (!(sd->flags & SD_LOAD_BALANCE)) continue; |
9bd721c55 sched/balancing: ... |
6338 6339 |
if (this_rq->avg_idle < curr_cost + sd->max_newidle_lb_cost) break; |
f492e12ef sched: Remove loa... |
6340 |
if (sd->flags & SD_BALANCE_NEWIDLE) { |
9bd721c55 sched/balancing: ... |
6341 |
t0 = sched_clock_cpu(this_cpu); |
1e3c88bde sched: Move load ... |
6342 |
/* If we've pulled tasks over stop searching: */ |
f492e12ef sched: Remove loa... |
6343 |
pulled_task = load_balance(this_cpu, this_rq, |
23f0d2093 sched: Factor out... |
6344 6345 |
sd, CPU_NEWLY_IDLE, &continue_balancing); |
9bd721c55 sched/balancing: ... |
6346 6347 6348 6349 6350 6351 |
domain_cost = sched_clock_cpu(this_cpu) - t0; if (domain_cost > sd->max_newidle_lb_cost) sd->max_newidle_lb_cost = domain_cost; curr_cost += domain_cost; |
f492e12ef sched: Remove loa... |
6352 |
} |
1e3c88bde sched: Move load ... |
6353 6354 6355 6356 |
interval = msecs_to_jiffies(sd->balance_interval); if (time_after(next_balance, sd->last_balance + interval)) next_balance = sd->last_balance + interval; |
3c4017c13 sched: Move rq->i... |
6357 |
if (pulled_task) |
1e3c88bde sched: Move load ... |
6358 |
break; |
1e3c88bde sched: Move load ... |
6359 |
} |
dce840a08 sched: Dynamicall... |
6360 |
rcu_read_unlock(); |
f492e12ef sched: Remove loa... |
6361 6362 |
raw_spin_lock(&this_rq->lock); |
0e5b5337f sched: Fix updati... |
6363 6364 |
if (curr_cost > this_rq->max_idle_balance_cost) this_rq->max_idle_balance_cost = curr_cost; |
e5fc66119 sched: Fix race i... |
6365 |
/* |
0e5b5337f sched: Fix updati... |
6366 6367 6368 |
* While browsing the domains, we released the rq lock, a task could * have been enqueued in the meantime. Since we're not going idle, * pretend we pulled a task. |
e5fc66119 sched: Fix race i... |
6369 |
*/ |
0e5b5337f sched: Fix updati... |
6370 |
if (this_rq->cfs.h_nr_running && !pulled_task) |
6e83125c6 sched/fair: Remov... |
6371 |
pulled_task = 1; |
e5fc66119 sched: Fix race i... |
6372 |
|
1e3c88bde sched: Move load ... |
6373 6374 6375 6376 6377 6378 6379 |
if (pulled_task || time_after(jiffies, this_rq->next_balance)) { /* * We are going idle. next_balance may be set based on * a busy processor. So reset next_balance. */ this_rq->next_balance = next_balance; } |
9bd721c55 sched/balancing: ... |
6380 |
|
6e83125c6 sched/fair: Remov... |
6381 |
out: |
e4aa358b6 sched/fair: Push ... |
6382 |
/* Is there a task of a high priority class? */ |
4c6c4e38c sched/core: Fix e... |
6383 |
if (this_rq->nr_running != this_rq->cfs.h_nr_running && |
a1d9a3231 sched: Check for ... |
6384 6385 |
((this_rq->stop && this_rq->stop->on_rq) || this_rq->dl.dl_nr_running || |
4c6c4e38c sched/core: Fix e... |
6386 |
(this_rq->rt.rt_nr_running && !rt_rq_throttled(&this_rq->rt)))) |
e4aa358b6 sched/fair: Push ... |
6387 6388 6389 6390 |
pulled_task = -1; if (pulled_task) { idle_exit_fair(this_rq); |
6e83125c6 sched/fair: Remov... |
6391 |
this_rq->idle_stamp = 0; |
e4aa358b6 sched/fair: Push ... |
6392 |
} |
6e83125c6 sched/fair: Remov... |
6393 |
|
3c4017c13 sched: Move rq->i... |
6394 |
return pulled_task; |
1e3c88bde sched: Move load ... |
6395 6396 6397 |
} /* |
969c79215 sched: replace mi... |
6398 6399 6400 6401 |
* active_load_balance_cpu_stop is run by cpu stopper. It pushes * running tasks off the busiest CPU onto idle CPUs. It requires at * least 1 task to be running on each physical CPU where possible, and * avoids physical / logical imbalances. |
1e3c88bde sched: Move load ... |
6402 |
*/ |
969c79215 sched: replace mi... |
6403 |
static int active_load_balance_cpu_stop(void *data) |
1e3c88bde sched: Move load ... |
6404 |
{ |
969c79215 sched: replace mi... |
6405 6406 |
struct rq *busiest_rq = data; int busiest_cpu = cpu_of(busiest_rq); |
1e3c88bde sched: Move load ... |
6407 |
int target_cpu = busiest_rq->push_cpu; |
969c79215 sched: replace mi... |
6408 |
struct rq *target_rq = cpu_rq(target_cpu); |
1e3c88bde sched: Move load ... |
6409 |
struct sched_domain *sd; |
969c79215 sched: replace mi... |
6410 6411 6412 6413 6414 6415 6416 |
raw_spin_lock_irq(&busiest_rq->lock); /* make sure the requested cpu hasn't gone down in the meantime */ if (unlikely(busiest_cpu != smp_processor_id() || !busiest_rq->active_balance)) goto out_unlock; |
1e3c88bde sched: Move load ... |
6417 6418 6419 |
/* Is there any task to move? */ if (busiest_rq->nr_running <= 1) |
969c79215 sched: replace mi... |
6420 |
goto out_unlock; |
1e3c88bde sched: Move load ... |
6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 |
/* * This condition is "impossible", if it occurs * we need to fix it. Originally reported by * Bjorn Helgaas on a 128-cpu setup. */ BUG_ON(busiest_rq == target_rq); /* move a task from busiest_rq to target_rq */ double_lock_balance(busiest_rq, target_rq); |
1e3c88bde sched: Move load ... |
6431 6432 |
/* Search for an sd spanning us and the target CPU. */ |
dce840a08 sched: Dynamicall... |
6433 |
rcu_read_lock(); |
1e3c88bde sched: Move load ... |
6434 6435 6436 6437 6438 6439 6440 |
for_each_domain(target_cpu, sd) { if ((sd->flags & SD_LOAD_BALANCE) && cpumask_test_cpu(busiest_cpu, sched_domain_span(sd))) break; } if (likely(sd)) { |
8e45cb545 sched: Move load-... |
6441 6442 |
struct lb_env env = { .sd = sd, |
ddcdf6e7d sched: Rename loa... |
6443 6444 6445 6446 |
.dst_cpu = target_cpu, .dst_rq = target_rq, .src_cpu = busiest_rq->cpu, .src_rq = busiest_rq, |
8e45cb545 sched: Move load-... |
6447 6448 |
.idle = CPU_IDLE, }; |
1e3c88bde sched: Move load ... |
6449 |
schedstat_inc(sd, alb_count); |
8e45cb545 sched: Move load-... |
6450 |
if (move_one_task(&env)) |
1e3c88bde sched: Move load ... |
6451 6452 6453 6454 |
schedstat_inc(sd, alb_pushed); else schedstat_inc(sd, alb_failed); } |
dce840a08 sched: Dynamicall... |
6455 |
rcu_read_unlock(); |
1e3c88bde sched: Move load ... |
6456 |
double_unlock_balance(busiest_rq, target_rq); |
969c79215 sched: replace mi... |
6457 6458 6459 6460 |
out_unlock: busiest_rq->active_balance = 0; raw_spin_unlock_irq(&busiest_rq->lock); return 0; |
1e3c88bde sched: Move load ... |
6461 |
} |
d987fc7f3 sched, nohz: Excl... |
6462 6463 6464 6465 |
static inline int on_null_domain(struct rq *rq) { return unlikely(!rcu_dereference_sched(rq->sd)); } |
3451d0243 nohz: Rename CONF... |
6466 |
#ifdef CONFIG_NO_HZ_COMMON |
83cd4fe27 sched: Change noh... |
6467 6468 |
/* * idle load balancing details |
83cd4fe27 sched: Change noh... |
6469 6470 6471 6472 |
* - When one of the busy CPUs notice that there may be an idle rebalancing * needed, they will kick the idle load balancer, which then does idle * load balancing for all the idle CPUs. */ |
1e3c88bde sched: Move load ... |
6473 |
static struct { |
83cd4fe27 sched: Change noh... |
6474 |
cpumask_var_t idle_cpus_mask; |
0b005cf54 sched, nohz: Impl... |
6475 |
atomic_t nr_cpus; |
83cd4fe27 sched: Change noh... |
6476 6477 |
unsigned long next_balance; /* in jiffy units */ } nohz ____cacheline_aligned; |
1e3c88bde sched: Move load ... |
6478 |
|
3dd0337d6 sched: Remove unu... |
6479 |
static inline int find_new_ilb(void) |
1e3c88bde sched: Move load ... |
6480 |
{ |
0b005cf54 sched, nohz: Impl... |
6481 |
int ilb = cpumask_first(nohz.idle_cpus_mask); |
1e3c88bde sched: Move load ... |
6482 |
|
786d6dc7a sched, nohz: Clea... |
6483 6484 6485 6486 |
if (ilb < nr_cpu_ids && idle_cpu(ilb)) return ilb; return nr_cpu_ids; |
1e3c88bde sched: Move load ... |
6487 |
} |
1e3c88bde sched: Move load ... |
6488 6489 |
/* |
83cd4fe27 sched: Change noh... |
6490 6491 6492 6493 |
* Kick a CPU to do the nohz balancing, if it is time for it. We pick the * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle * CPU (if there is one). */ |
0aeeeebac sched: Remove unu... |
6494 |
static void nohz_balancer_kick(void) |
83cd4fe27 sched: Change noh... |
6495 6496 6497 6498 |
{ int ilb_cpu; nohz.next_balance++; |
3dd0337d6 sched: Remove unu... |
6499 |
ilb_cpu = find_new_ilb(); |
83cd4fe27 sched: Change noh... |
6500 |
|
0b005cf54 sched, nohz: Impl... |
6501 6502 |
if (ilb_cpu >= nr_cpu_ids) return; |
83cd4fe27 sched: Change noh... |
6503 |
|
cd490c5b2 sched, nohz: Set ... |
6504 |
if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu))) |
1c792db7f sched, nohz: Intr... |
6505 6506 6507 6508 6509 6510 6511 6512 |
return; /* * Use smp_send_reschedule() instead of resched_cpu(). * This way we generate a sched IPI on the target cpu which * is idle. And the softirq performing nohz idle load balance * will be run before returning from the IPI. */ smp_send_reschedule(ilb_cpu); |
83cd4fe27 sched: Change noh... |
6513 6514 |
return; } |
c1cc017c5 sched/nohz: Clean... |
6515 |
static inline void nohz_balance_exit_idle(int cpu) |
71325960d sched/nohz: Fix n... |
6516 6517 |
{ if (unlikely(test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))) { |
d987fc7f3 sched, nohz: Excl... |
6518 6519 6520 6521 6522 6523 6524 |
/* * Completely isolated CPUs don't ever set, so we must test. */ if (likely(cpumask_test_cpu(cpu, nohz.idle_cpus_mask))) { cpumask_clear_cpu(cpu, nohz.idle_cpus_mask); atomic_dec(&nohz.nr_cpus); } |
71325960d sched/nohz: Fix n... |
6525 6526 6527 |
clear_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)); } } |
69e1e811d sched, nohz: Trac... |
6528 6529 6530 |
static inline void set_cpu_sd_state_busy(void) { struct sched_domain *sd; |
37dc6b50c sched: Remove unn... |
6531 |
int cpu = smp_processor_id(); |
69e1e811d sched, nohz: Trac... |
6532 |
|
69e1e811d sched, nohz: Trac... |
6533 |
rcu_read_lock(); |
37dc6b50c sched: Remove unn... |
6534 |
sd = rcu_dereference(per_cpu(sd_busy, cpu)); |
25f55d9d0 sched: Fix init N... |
6535 6536 6537 6538 |
if (!sd || !sd->nohz_idle) goto unlock; sd->nohz_idle = 0; |
37dc6b50c sched: Remove unn... |
6539 |
atomic_inc(&sd->groups->sgp->nr_busy_cpus); |
25f55d9d0 sched: Fix init N... |
6540 |
unlock: |
69e1e811d sched, nohz: Trac... |
6541 6542 6543 6544 6545 6546 |
rcu_read_unlock(); } void set_cpu_sd_state_idle(void) { struct sched_domain *sd; |
37dc6b50c sched: Remove unn... |
6547 |
int cpu = smp_processor_id(); |
69e1e811d sched, nohz: Trac... |
6548 |
|
69e1e811d sched, nohz: Trac... |
6549 |
rcu_read_lock(); |
37dc6b50c sched: Remove unn... |
6550 |
sd = rcu_dereference(per_cpu(sd_busy, cpu)); |
25f55d9d0 sched: Fix init N... |
6551 6552 6553 6554 |
if (!sd || sd->nohz_idle) goto unlock; sd->nohz_idle = 1; |
37dc6b50c sched: Remove unn... |
6555 |
atomic_dec(&sd->groups->sgp->nr_busy_cpus); |
25f55d9d0 sched: Fix init N... |
6556 |
unlock: |
69e1e811d sched, nohz: Trac... |
6557 6558 |
rcu_read_unlock(); } |
83cd4fe27 sched: Change noh... |
6559 |
/* |
c1cc017c5 sched/nohz: Clean... |
6560 |
* This routine will record that the cpu is going idle with tick stopped. |
0b005cf54 sched, nohz: Impl... |
6561 |
* This info will be used in performing idle load balancing in the future. |
1e3c88bde sched: Move load ... |
6562 |
*/ |
c1cc017c5 sched/nohz: Clean... |
6563 |
void nohz_balance_enter_idle(int cpu) |
1e3c88bde sched: Move load ... |
6564 |
{ |
71325960d sched/nohz: Fix n... |
6565 6566 6567 6568 6569 |
/* * If this cpu is going down, then nothing needs to be done. */ if (!cpu_active(cpu)) return; |
c1cc017c5 sched/nohz: Clean... |
6570 6571 |
if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu))) return; |
1e3c88bde sched: Move load ... |
6572 |
|
d987fc7f3 sched, nohz: Excl... |
6573 6574 6575 6576 6577 |
/* * If we're a completely isolated CPU, we don't play. */ if (on_null_domain(cpu_rq(cpu))) return; |
c1cc017c5 sched/nohz: Clean... |
6578 6579 6580 |
cpumask_set_cpu(cpu, nohz.idle_cpus_mask); atomic_inc(&nohz.nr_cpus); set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)); |
1e3c88bde sched: Move load ... |
6581 |
} |
71325960d sched/nohz: Fix n... |
6582 |
|
0db0628d9 kernel: delete __... |
6583 |
static int sched_ilb_notifier(struct notifier_block *nfb, |
71325960d sched/nohz: Fix n... |
6584 6585 6586 6587 |
unsigned long action, void *hcpu) { switch (action & ~CPU_TASKS_FROZEN) { case CPU_DYING: |
c1cc017c5 sched/nohz: Clean... |
6588 |
nohz_balance_exit_idle(smp_processor_id()); |
71325960d sched/nohz: Fix n... |
6589 6590 6591 6592 6593 |
return NOTIFY_OK; default: return NOTIFY_DONE; } } |
1e3c88bde sched: Move load ... |
6594 6595 6596 |
#endif static DEFINE_SPINLOCK(balancing); |
49c022e65 sched: Clean up r... |
6597 6598 6599 6600 |
/* * Scale the max load_balance interval with the number of CPUs in the system. * This trades load-balance latency on larger machines for less cross talk. */ |
029632fbb sched: Make separ... |
6601 |
void update_max_interval(void) |
49c022e65 sched: Clean up r... |
6602 6603 6604 |
{ max_load_balance_interval = HZ*num_online_cpus()/10; } |
1e3c88bde sched: Move load ... |
6605 6606 6607 6608 |
/* * It checks each scheduling domain to see if it is due to be balanced, * and initiates a balancing operation if so. * |
b9b0853a4 sched: Fix commen... |
6609 |
* Balancing parameters are set up in init_sched_domains. |
1e3c88bde sched: Move load ... |
6610 |
*/ |
f7ed0a895 sched: Pass 'stru... |
6611 |
static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle) |
1e3c88bde sched: Move load ... |
6612 |
{ |
23f0d2093 sched: Factor out... |
6613 |
int continue_balancing = 1; |
f7ed0a895 sched: Pass 'stru... |
6614 |
int cpu = rq->cpu; |
1e3c88bde sched: Move load ... |
6615 |
unsigned long interval; |
04f733b4a sched/fair: Rever... |
6616 |
struct sched_domain *sd; |
1e3c88bde sched: Move load ... |
6617 6618 6619 |
/* Earliest time when we have to do rebalance again */ unsigned long next_balance = jiffies + 60*HZ; int update_next_balance = 0; |
f48627e68 sched/balancing: ... |
6620 6621 |
int need_serialize, need_decay = 0; u64 max_cost = 0; |
1e3c88bde sched: Move load ... |
6622 |
|
48a167532 sched: Refactor u... |
6623 |
update_blocked_averages(cpu); |
2069dd75c sched: Rewrite tg... |
6624 |
|
dce840a08 sched: Dynamicall... |
6625 |
rcu_read_lock(); |
1e3c88bde sched: Move load ... |
6626 |
for_each_domain(cpu, sd) { |
f48627e68 sched/balancing: ... |
6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 |
/* * Decay the newidle max times here because this is a regular * visit to all the domains. Decay ~1% per second. */ if (time_after(jiffies, sd->next_decay_max_lb_cost)) { sd->max_newidle_lb_cost = (sd->max_newidle_lb_cost * 253) / 256; sd->next_decay_max_lb_cost = jiffies + HZ; need_decay = 1; } max_cost += sd->max_newidle_lb_cost; |
1e3c88bde sched: Move load ... |
6638 6639 |
if (!(sd->flags & SD_LOAD_BALANCE)) continue; |
f48627e68 sched/balancing: ... |
6640 6641 6642 6643 6644 6645 6646 6647 6648 6649 |
/* * Stop the load balance at this level. There is another * CPU in our sched group which is doing load balancing more * actively. */ if (!continue_balancing) { if (need_decay) continue; break; } |
1e3c88bde sched: Move load ... |
6650 6651 6652 6653 6654 6655 |
interval = sd->balance_interval; if (idle != CPU_IDLE) interval *= sd->busy_factor; /* scale ms to jiffies */ interval = msecs_to_jiffies(interval); |
49c022e65 sched: Clean up r... |
6656 |
interval = clamp(interval, 1UL, max_load_balance_interval); |
1e3c88bde sched: Move load ... |
6657 6658 6659 6660 6661 6662 6663 6664 6665 |
need_serialize = sd->flags & SD_SERIALIZE; if (need_serialize) { if (!spin_trylock(&balancing)) goto out; } if (time_after_eq(jiffies, sd->last_balance + interval)) { |
23f0d2093 sched: Factor out... |
6666 |
if (load_balance(cpu, rq, sd, idle, &continue_balancing)) { |
1e3c88bde sched: Move load ... |
6667 |
/* |
6263322c5 sched/fair: Rewri... |
6668 |
* The LBF_DST_PINNED logic could have changed |
de5eb2dd7 sched: Explicitly... |
6669 6670 |
* env->dst_cpu, so we can't know our idle * state even if we migrated tasks. Update it. |
1e3c88bde sched: Move load ... |
6671 |
*/ |
de5eb2dd7 sched: Explicitly... |
6672 |
idle = idle_cpu(cpu) ? CPU_IDLE : CPU_NOT_IDLE; |
1e3c88bde sched: Move load ... |
6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 |
} sd->last_balance = jiffies; } if (need_serialize) spin_unlock(&balancing); out: if (time_after(next_balance, sd->last_balance + interval)) { next_balance = sd->last_balance + interval; update_next_balance = 1; } |
f48627e68 sched/balancing: ... |
6683 6684 |
} if (need_decay) { |
1e3c88bde sched: Move load ... |
6685 |
/* |
f48627e68 sched/balancing: ... |
6686 6687 |
* Ensure the rq-wide value also decays but keep it at a * reasonable floor to avoid funnies with rq->avg_idle. |
1e3c88bde sched: Move load ... |
6688 |
*/ |
f48627e68 sched/balancing: ... |
6689 6690 |
rq->max_idle_balance_cost = max((u64)sysctl_sched_migration_cost, max_cost); |
1e3c88bde sched: Move load ... |
6691 |
} |
dce840a08 sched: Dynamicall... |
6692 |
rcu_read_unlock(); |
1e3c88bde sched: Move load ... |
6693 6694 6695 6696 6697 6698 6699 6700 6701 |
/* * next_balance will be updated only when there is a need. * When the cpu is attached to null domain for ex, it will not be * updated. */ if (likely(update_next_balance)) rq->next_balance = next_balance; } |
3451d0243 nohz: Rename CONF... |
6702 |
#ifdef CONFIG_NO_HZ_COMMON |
1e3c88bde sched: Move load ... |
6703 |
/* |
3451d0243 nohz: Rename CONF... |
6704 |
* In CONFIG_NO_HZ_COMMON case, the idle balance kickee will do the |
1e3c88bde sched: Move load ... |
6705 6706 |
* rebalancing for all the cpus for whom scheduler ticks are stopped. */ |
208cb16ba sched: Pass 'stru... |
6707 |
static void nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle) |
83cd4fe27 sched: Change noh... |
6708 |
{ |
208cb16ba sched: Pass 'stru... |
6709 |
int this_cpu = this_rq->cpu; |
83cd4fe27 sched: Change noh... |
6710 6711 |
struct rq *rq; int balance_cpu; |
1c792db7f sched, nohz: Intr... |
6712 6713 6714 |
if (idle != CPU_IDLE || !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu))) goto end; |
83cd4fe27 sched: Change noh... |
6715 6716 |
for_each_cpu(balance_cpu, nohz.idle_cpus_mask) { |
8a6d42d1b sched, nohz: Fix ... |
6717 |
if (balance_cpu == this_cpu || !idle_cpu(balance_cpu)) |
83cd4fe27 sched: Change noh... |
6718 6719 6720 6721 6722 6723 6724 |
continue; /* * If this cpu gets work to do, stop the load balancing * work being done for other cpus. Next load * balancing owner will pick it up. */ |
1c792db7f sched, nohz: Intr... |
6725 |
if (need_resched()) |
83cd4fe27 sched: Change noh... |
6726 |
break; |
83cd4fe27 sched: Change noh... |
6727 |
|
5ed4f1d96 sched: Fix nohz_i... |
6728 6729 6730 6731 6732 6733 |
rq = cpu_rq(balance_cpu); raw_spin_lock_irq(&rq->lock); update_rq_clock(rq); update_idle_cpu_load(rq); raw_spin_unlock_irq(&rq->lock); |
83cd4fe27 sched: Change noh... |
6734 |
|
f7ed0a895 sched: Pass 'stru... |
6735 |
rebalance_domains(rq, CPU_IDLE); |
83cd4fe27 sched: Change noh... |
6736 |
|
83cd4fe27 sched: Change noh... |
6737 6738 6739 6740 |
if (time_after(this_rq->next_balance, rq->next_balance)) this_rq->next_balance = rq->next_balance; } nohz.next_balance = this_rq->next_balance; |
1c792db7f sched, nohz: Intr... |
6741 6742 |
end: clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)); |
83cd4fe27 sched: Change noh... |
6743 6744 6745 |
} /* |
0b005cf54 sched, nohz: Impl... |
6746 6747 6748 6749 6750 6751 6752 |
* Current heuristic for kicking the idle load balancer in the presence * of an idle cpu is the system. * - This rq has more than one task. * - At any scheduler domain level, this cpu's scheduler group has multiple * busy cpu's exceeding the group's power. * - For SD_ASYM_PACKING, if the lower numbered cpu's in the scheduler * domain span are idle. |
83cd4fe27 sched: Change noh... |
6753 |
*/ |
4a725627f sched: Reduce noh... |
6754 |
static inline int nohz_kick_needed(struct rq *rq) |
83cd4fe27 sched: Change noh... |
6755 6756 |
{ unsigned long now = jiffies; |
0b005cf54 sched, nohz: Impl... |
6757 |
struct sched_domain *sd; |
37dc6b50c sched: Remove unn... |
6758 |
struct sched_group_power *sgp; |
4a725627f sched: Reduce noh... |
6759 |
int nr_busy, cpu = rq->cpu; |
83cd4fe27 sched: Change noh... |
6760 |
|
4a725627f sched: Reduce noh... |
6761 |
if (unlikely(rq->idle_balance)) |
83cd4fe27 sched: Change noh... |
6762 |
return 0; |
1c792db7f sched, nohz: Intr... |
6763 6764 6765 6766 |
/* * We may be recently in ticked or tickless idle mode. At the first * busy tick after returning from idle, we will update the busy stats. */ |
69e1e811d sched, nohz: Trac... |
6767 |
set_cpu_sd_state_busy(); |
c1cc017c5 sched/nohz: Clean... |
6768 |
nohz_balance_exit_idle(cpu); |
0b005cf54 sched, nohz: Impl... |
6769 6770 6771 6772 6773 6774 6775 |
/* * None are in tickless mode and hence no need for NOHZ idle load * balancing. */ if (likely(!atomic_read(&nohz.nr_cpus))) return 0; |
1c792db7f sched, nohz: Intr... |
6776 6777 |
if (time_before(now, nohz.next_balance)) |
83cd4fe27 sched: Change noh... |
6778 |
return 0; |
0b005cf54 sched, nohz: Impl... |
6779 6780 |
if (rq->nr_running >= 2) goto need_kick; |
83cd4fe27 sched: Change noh... |
6781 |
|
067491b73 sched, nohz: Fix ... |
6782 |
rcu_read_lock(); |
37dc6b50c sched: Remove unn... |
6783 |
sd = rcu_dereference(per_cpu(sd_busy, cpu)); |
83cd4fe27 sched: Change noh... |
6784 |
|
37dc6b50c sched: Remove unn... |
6785 6786 6787 |
if (sd) { sgp = sd->groups->sgp; nr_busy = atomic_read(&sgp->nr_busy_cpus); |
0b005cf54 sched, nohz: Impl... |
6788 |
|
37dc6b50c sched: Remove unn... |
6789 |
if (nr_busy > 1) |
067491b73 sched, nohz: Fix ... |
6790 |
goto need_kick_unlock; |
83cd4fe27 sched: Change noh... |
6791 |
} |
37dc6b50c sched: Remove unn... |
6792 6793 6794 6795 6796 6797 |
sd = rcu_dereference(per_cpu(sd_asym, cpu)); if (sd && (cpumask_first_and(nohz.idle_cpus_mask, sched_domain_span(sd)) < cpu)) goto need_kick_unlock; |
067491b73 sched, nohz: Fix ... |
6798 |
rcu_read_unlock(); |
83cd4fe27 sched: Change noh... |
6799 |
return 0; |
067491b73 sched, nohz: Fix ... |
6800 6801 6802 |
need_kick_unlock: rcu_read_unlock(); |
0b005cf54 sched, nohz: Impl... |
6803 6804 |
need_kick: return 1; |
83cd4fe27 sched: Change noh... |
6805 6806 |
} #else |
208cb16ba sched: Pass 'stru... |
6807 |
static void nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle) { } |
83cd4fe27 sched: Change noh... |
6808 6809 6810 6811 6812 6813 |
#endif /* * run_rebalance_domains is triggered when needed from the scheduler tick. * Also triggered for nohz idle balancing (with nohz_balancing_kick set). */ |
1e3c88bde sched: Move load ... |
6814 6815 |
static void run_rebalance_domains(struct softirq_action *h) { |
208cb16ba sched: Pass 'stru... |
6816 |
struct rq *this_rq = this_rq(); |
6eb57e0d6 sched: Request fo... |
6817 |
enum cpu_idle_type idle = this_rq->idle_balance ? |
1e3c88bde sched: Move load ... |
6818 |
CPU_IDLE : CPU_NOT_IDLE; |
f7ed0a895 sched: Pass 'stru... |
6819 |
rebalance_domains(this_rq, idle); |
1e3c88bde sched: Move load ... |
6820 |
|
1e3c88bde sched: Move load ... |
6821 |
/* |
83cd4fe27 sched: Change noh... |
6822 |
* If this cpu has a pending nohz_balance_kick, then do the |
1e3c88bde sched: Move load ... |
6823 6824 6825 |
* balancing on behalf of the other idle cpus whose ticks are * stopped. */ |
208cb16ba sched: Pass 'stru... |
6826 |
nohz_idle_balance(this_rq, idle); |
1e3c88bde sched: Move load ... |
6827 |
} |
1e3c88bde sched: Move load ... |
6828 6829 |
/* * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. |
1e3c88bde sched: Move load ... |
6830 |
*/ |
7caff66f3 sched: Reduce tri... |
6831 |
void trigger_load_balance(struct rq *rq) |
1e3c88bde sched: Move load ... |
6832 |
{ |
1e3c88bde sched: Move load ... |
6833 |
/* Don't need to rebalance while attached to NULL domain */ |
c726099ec sched: Factor out... |
6834 6835 6836 6837 |
if (unlikely(on_null_domain(rq))) return; if (time_after_eq(jiffies, rq->next_balance)) |
1e3c88bde sched: Move load ... |
6838 |
raise_softirq(SCHED_SOFTIRQ); |
3451d0243 nohz: Rename CONF... |
6839 |
#ifdef CONFIG_NO_HZ_COMMON |
c726099ec sched: Factor out... |
6840 |
if (nohz_kick_needed(rq)) |
0aeeeebac sched: Remove unu... |
6841 |
nohz_balancer_kick(); |
83cd4fe27 sched: Change noh... |
6842 |
#endif |
1e3c88bde sched: Move load ... |
6843 |
} |
0bcdcf28c sched: Fix missin... |
6844 6845 6846 6847 6848 6849 6850 6851 |
static void rq_online_fair(struct rq *rq) { update_sysctl(); } static void rq_offline_fair(struct rq *rq) { update_sysctl(); |
a4c96ae31 sched: Unthrottle... |
6852 6853 6854 |
/* Ensure any throttled groups are reachable by pick_next_task */ unthrottle_offline_cfs_rqs(rq); |
0bcdcf28c sched: Fix missin... |
6855 |
} |
55e12e5e7 sched: make sched... |
6856 |
#endif /* CONFIG_SMP */ |
e1d1484f7 sched: reduce bal... |
6857 |
|
bf0f6f24a sched: cfs core, ... |
6858 6859 6860 |
/* * scheduler tick hitting a task of our scheduling class: */ |
8f4d37ec0 sched: high-res p... |
6861 |
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued) |
bf0f6f24a sched: cfs core, ... |
6862 6863 6864 6865 6866 6867 |
{ struct cfs_rq *cfs_rq; struct sched_entity *se = &curr->se; for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); |
8f4d37ec0 sched: high-res p... |
6868 |
entity_tick(cfs_rq, se, queued); |
bf0f6f24a sched: cfs core, ... |
6869 |
} |
18bf2805d sched: Maintain p... |
6870 |
|
10e84b97e mm: sched: numa: ... |
6871 |
if (numabalancing_enabled) |
cbee9f88e mm: numa: Add fau... |
6872 |
task_tick_numa(rq, curr); |
3d59eebc5 Merge tag 'balanc... |
6873 |
|
18bf2805d sched: Maintain p... |
6874 |
update_rq_runnable_avg(rq, 1); |
bf0f6f24a sched: cfs core, ... |
6875 6876 6877 |
} /* |
cd29fe6f2 sched: Sanitize f... |
6878 6879 6880 |
* called on fork with the child task as argument from the parent's context * - child not yet on the tasklist * - preemption disabled |
bf0f6f24a sched: cfs core, ... |
6881 |
*/ |
cd29fe6f2 sched: Sanitize f... |
6882 |
static void task_fork_fair(struct task_struct *p) |
bf0f6f24a sched: cfs core, ... |
6883 |
{ |
4fc420c91 sched: Fix cgroup... |
6884 6885 |
struct cfs_rq *cfs_rq; struct sched_entity *se = &p->se, *curr; |
00bf7bfc2 sched: fix: move ... |
6886 |
int this_cpu = smp_processor_id(); |
cd29fe6f2 sched: Sanitize f... |
6887 6888 |
struct rq *rq = this_rq(); unsigned long flags; |
05fa785cf sched: Convert rq... |
6889 |
raw_spin_lock_irqsave(&rq->lock, flags); |
bf0f6f24a sched: cfs core, ... |
6890 |
|
861d034ee sched: Fix rq->cl... |
6891 |
update_rq_clock(rq); |
4fc420c91 sched: Fix cgroup... |
6892 6893 |
cfs_rq = task_cfs_rq(current); curr = cfs_rq->curr; |
6c9a27f5d sched/fair: Fix s... |
6894 6895 6896 6897 6898 6899 6900 6901 6902 |
/* * Not only the cpu but also the task_group of the parent might have * been changed after parent->se.parent,cfs_rq were copied to * child->se.parent,cfs_rq. So call __set_task_cpu() to make those * of child point to valid ones. */ rcu_read_lock(); __set_task_cpu(p, this_cpu); rcu_read_unlock(); |
bf0f6f24a sched: cfs core, ... |
6903 |
|
7109c4429 sched: call updat... |
6904 |
update_curr(cfs_rq); |
cd29fe6f2 sched: Sanitize f... |
6905 |
|
b5d9d734a sched: Ensure tha... |
6906 6907 |
if (curr) se->vruntime = curr->vruntime; |
aeb73b040 sched: clean up n... |
6908 |
place_entity(cfs_rq, se, 1); |
4d78e7b65 sched: new task p... |
6909 |
|
cd29fe6f2 sched: Sanitize f... |
6910 |
if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) { |
87fefa381 sched: optimize t... |
6911 |
/* |
edcb60a30 sched: kernel/sch... |
6912 6913 6914 |
* Upon rescheduling, sched_class::put_prev_task() will place * 'current' within the tree based on its new key value. */ |
4d78e7b65 sched: new task p... |
6915 |
swap(curr->vruntime, se->vruntime); |
aec0a5142 sched: call resch... |
6916 |
resched_task(rq->curr); |
4d78e7b65 sched: new task p... |
6917 |
} |
bf0f6f24a sched: cfs core, ... |
6918 |
|
88ec22d3e sched: Remove the... |
6919 |
se->vruntime -= cfs_rq->min_vruntime; |
05fa785cf sched: Convert rq... |
6920 |
raw_spin_unlock_irqrestore(&rq->lock, flags); |
bf0f6f24a sched: cfs core, ... |
6921 |
} |
cb4698450 sched: RT-balance... |
6922 6923 6924 6925 |
/* * Priority of the task has changed. Check to see if we preempt * the current task. */ |
da7a735e5 sched: Fix switch... |
6926 6927 |
static void prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio) |
cb4698450 sched: RT-balance... |
6928 |
{ |
da7a735e5 sched: Fix switch... |
6929 6930 |
if (!p->se.on_rq) return; |
cb4698450 sched: RT-balance... |
6931 6932 6933 6934 6935 |
/* * Reschedule if we are currently running on this runqueue and * our priority decreased, or if we are not currently running on * this runqueue and our priority is higher than the current's */ |
da7a735e5 sched: Fix switch... |
6936 |
if (rq->curr == p) { |
cb4698450 sched: RT-balance... |
6937 6938 6939 |
if (p->prio > oldprio) resched_task(rq->curr); } else |
15afe09bf sched: wakeup pre... |
6940 |
check_preempt_curr(rq, p, 0); |
cb4698450 sched: RT-balance... |
6941 |
} |
da7a735e5 sched: Fix switch... |
6942 6943 6944 6945 6946 6947 |
static void switched_from_fair(struct rq *rq, struct task_struct *p) { struct sched_entity *se = &p->se; struct cfs_rq *cfs_rq = cfs_rq_of(se); /* |
791c9e029 sched: Fix double... |
6948 |
* Ensure the task's vruntime is normalized, so that when it's |
da7a735e5 sched: Fix switch... |
6949 6950 6951 |
* switched back to the fair class the enqueue_entity(.flags=0) will * do the right thing. * |
791c9e029 sched: Fix double... |
6952 6953 |
* If it's on_rq, then the dequeue_entity(.flags=0) will already * have normalized the vruntime, if it's !on_rq, then only when |
da7a735e5 sched: Fix switch... |
6954 6955 |
* the task is sleeping will it still have non-normalized vruntime. */ |
791c9e029 sched: Fix double... |
6956 |
if (!p->on_rq && p->state != TASK_RUNNING) { |
da7a735e5 sched: Fix switch... |
6957 6958 6959 6960 6961 6962 6963 |
/* * Fix up our vruntime so that the current sleep doesn't * cause 'unlimited' sleep bonus. */ place_entity(cfs_rq, se, 0); se->vruntime -= cfs_rq->min_vruntime; } |
9ee474f55 sched: Maintain t... |
6964 |
|
141965c74 Revert "sched: In... |
6965 |
#ifdef CONFIG_SMP |
9ee474f55 sched: Maintain t... |
6966 6967 6968 6969 6970 |
/* * Remove our load from contribution when we leave sched_fair * and ensure we don't carry in an old decay_count if we * switch back. */ |
87e3c8ae1 sched/fair: Clean... |
6971 6972 6973 |
if (se->avg.decay_count) { __synchronize_entity_decay(se); subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib); |
9ee474f55 sched: Maintain t... |
6974 6975 |
} #endif |
da7a735e5 sched: Fix switch... |
6976 |
} |
cb4698450 sched: RT-balance... |
6977 6978 6979 |
/* * We switched to the sched_fair class. */ |
da7a735e5 sched: Fix switch... |
6980 |
static void switched_to_fair(struct rq *rq, struct task_struct *p) |
cb4698450 sched: RT-balance... |
6981 |
{ |
eb7a59b2c sched/fair: Reset... |
6982 6983 6984 6985 6986 6987 6988 6989 6990 |
struct sched_entity *se = &p->se; #ifdef CONFIG_FAIR_GROUP_SCHED /* * Since the real-depth could have been changed (only FAIR * class maintain depth value), reset depth properly. */ se->depth = se->parent ? se->parent->depth + 1 : 0; #endif if (!se->on_rq) |
da7a735e5 sched: Fix switch... |
6991 |
return; |
cb4698450 sched: RT-balance... |
6992 6993 6994 6995 6996 |
/* * We were most likely switched from sched_rt, so * kick off the schedule if running, otherwise just see * if we can still preempt the current task. */ |
da7a735e5 sched: Fix switch... |
6997 |
if (rq->curr == p) |
cb4698450 sched: RT-balance... |
6998 6999 |
resched_task(rq->curr); else |
15afe09bf sched: wakeup pre... |
7000 |
check_preempt_curr(rq, p, 0); |
cb4698450 sched: RT-balance... |
7001 |
} |
83b699ed2 sched: revert rec... |
7002 7003 7004 7005 7006 7007 7008 7009 |
/* Account for a task changing its policy or group. * * This routine is mostly called to set cfs_rq->curr field when a task * migrates between groups/classes. */ static void set_curr_task_fair(struct rq *rq) { struct sched_entity *se = &rq->curr->se; |
ec12cb7f3 sched: Accumulate... |
7010 7011 7012 7013 7014 7015 7016 |
for_each_sched_entity(se) { struct cfs_rq *cfs_rq = cfs_rq_of(se); set_next_entity(cfs_rq, se); /* ensure bandwidth has been allocated on our new cfs_rq */ account_cfs_rq_runtime(cfs_rq, 0); } |
83b699ed2 sched: revert rec... |
7017 |
} |
029632fbb sched: Make separ... |
7018 7019 7020 |
void init_cfs_rq(struct cfs_rq *cfs_rq) { cfs_rq->tasks_timeline = RB_ROOT; |
029632fbb sched: Make separ... |
7021 7022 7023 7024 |
cfs_rq->min_vruntime = (u64)(-(1LL << 20)); #ifndef CONFIG_64BIT cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime; #endif |
141965c74 Revert "sched: In... |
7025 |
#ifdef CONFIG_SMP |
9ee474f55 sched: Maintain t... |
7026 |
atomic64_set(&cfs_rq->decay_counter, 1); |
2509940fd sched/cfs_rq: Cha... |
7027 |
atomic_long_set(&cfs_rq->removed_load, 0); |
9ee474f55 sched: Maintain t... |
7028 |
#endif |
029632fbb sched: Make separ... |
7029 |
} |
810b38179 sched: retain vru... |
7030 |
#ifdef CONFIG_FAIR_GROUP_SCHED |
b2b5ce022 sched, cgroup: Fi... |
7031 |
static void task_move_group_fair(struct task_struct *p, int on_rq) |
810b38179 sched: retain vru... |
7032 |
{ |
fed14d45f sched/fair: Track... |
7033 |
struct sched_entity *se = &p->se; |
aff3e4988 sched: Account fo... |
7034 |
struct cfs_rq *cfs_rq; |
fed14d45f sched/fair: Track... |
7035 |
|
b2b5ce022 sched, cgroup: Fi... |
7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046 7047 7048 |
/* * If the task was not on the rq at the time of this cgroup movement * it must have been asleep, sleeping tasks keep their ->vruntime * absolute on their old rq until wakeup (needed for the fair sleeper * bonus in place_entity()). * * If it was on the rq, we've just 'preempted' it, which does convert * ->vruntime to a relative base. * * Make sure both cases convert their relative position when migrating * to another cgroup's rq. This does somewhat interfere with the * fair sleeper stuff for the first placement, but who cares. */ |
7ceff013c sched: Fix cgroup... |
7049 7050 7051 7052 7053 7054 |
/* * When !on_rq, vruntime of the task has usually NOT been normalized. * But there are some cases where it has already been normalized: * * - Moving a forked child which is waiting for being woken up by * wake_up_new_task(). |
62af3783e sched: Fix cgroup... |
7055 7056 |
* - Moving a task which has been woken up by try_to_wake_up() and * waiting for actually being woken up by sched_ttwu_pending(). |
7ceff013c sched: Fix cgroup... |
7057 7058 7059 7060 |
* * To prevent boost or penalty in the new cfs_rq caused by delta * min_vruntime between the two cfs_rqs, we skip vruntime adjustment. */ |
fed14d45f sched/fair: Track... |
7061 |
if (!on_rq && (!se->sum_exec_runtime || p->state == TASK_WAKING)) |
7ceff013c sched: Fix cgroup... |
7062 |
on_rq = 1; |
b2b5ce022 sched, cgroup: Fi... |
7063 |
if (!on_rq) |
fed14d45f sched/fair: Track... |
7064 |
se->vruntime -= cfs_rq_of(se)->min_vruntime; |
b2b5ce022 sched, cgroup: Fi... |
7065 |
set_task_rq(p, task_cpu(p)); |
fed14d45f sched/fair: Track... |
7066 |
se->depth = se->parent ? se->parent->depth + 1 : 0; |
aff3e4988 sched: Account fo... |
7067 |
if (!on_rq) { |
fed14d45f sched/fair: Track... |
7068 7069 |
cfs_rq = cfs_rq_of(se); se->vruntime += cfs_rq->min_vruntime; |
aff3e4988 sched: Account fo... |
7070 7071 7072 7073 7074 7075 |
#ifdef CONFIG_SMP /* * migrate_task_rq_fair() will have removed our previous * contribution, but we must synchronize for ongoing future * decay. */ |
fed14d45f sched/fair: Track... |
7076 7077 |
se->avg.decay_count = atomic64_read(&cfs_rq->decay_counter); cfs_rq->blocked_load_avg += se->avg.load_avg_contrib; |
aff3e4988 sched: Account fo... |
7078 7079 |
#endif } |
810b38179 sched: retain vru... |
7080 |
} |
029632fbb sched: Make separ... |
7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 |
void free_fair_sched_group(struct task_group *tg) { int i; destroy_cfs_bandwidth(tg_cfs_bandwidth(tg)); for_each_possible_cpu(i) { if (tg->cfs_rq) kfree(tg->cfs_rq[i]); if (tg->se) kfree(tg->se[i]); } kfree(tg->cfs_rq); kfree(tg->se); } int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) { struct cfs_rq *cfs_rq; struct sched_entity *se; int i; tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL); if (!tg->cfs_rq) goto err; tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL); if (!tg->se) goto err; tg->shares = NICE_0_LOAD; init_cfs_bandwidth(tg_cfs_bandwidth(tg)); for_each_possible_cpu(i) { cfs_rq = kzalloc_node(sizeof(struct cfs_rq), GFP_KERNEL, cpu_to_node(i)); if (!cfs_rq) goto err; se = kzalloc_node(sizeof(struct sched_entity), GFP_KERNEL, cpu_to_node(i)); if (!se) goto err_free_rq; init_cfs_rq(cfs_rq); init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]); } return 1; err_free_rq: kfree(cfs_rq); err: return 0; } void unregister_fair_sched_group(struct task_group *tg, int cpu) { struct rq *rq = cpu_rq(cpu); unsigned long flags; /* * Only empty task groups can be destroyed; so we can speculatively * check on_list without danger of it being re-added. */ if (!tg->cfs_rq[cpu]->on_list) return; raw_spin_lock_irqsave(&rq->lock, flags); list_del_leaf_cfs_rq(tg->cfs_rq[cpu]); raw_spin_unlock_irqrestore(&rq->lock, flags); } void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq, struct sched_entity *se, int cpu, struct sched_entity *parent) { struct rq *rq = cpu_rq(cpu); cfs_rq->tg = tg; cfs_rq->rq = rq; |
029632fbb sched: Make separ... |
7164 7165 7166 7167 7168 7169 7170 7171 |
init_cfs_rq_runtime(cfs_rq); tg->cfs_rq[cpu] = cfs_rq; tg->se[cpu] = se; /* se could be NULL for root_task_group */ if (!se) return; |
fed14d45f sched/fair: Track... |
7172 |
if (!parent) { |
029632fbb sched: Make separ... |
7173 |
se->cfs_rq = &rq->cfs; |
fed14d45f sched/fair: Track... |
7174 7175 |
se->depth = 0; } else { |
029632fbb sched: Make separ... |
7176 |
se->cfs_rq = parent->my_q; |
fed14d45f sched/fair: Track... |
7177 7178 |
se->depth = parent->depth + 1; } |
029632fbb sched: Make separ... |
7179 7180 |
se->my_q = cfs_rq; |
0ac9b1c21 sched: Guarantee ... |
7181 7182 |
/* guarantee group entities always have weight */ update_load_set(&se->load, NICE_0_LOAD); |
029632fbb sched: Make separ... |
7183 7184 7185 7186 7187 7188 7189 7190 7191 7192 7193 7194 7195 7196 7197 7198 7199 7200 7201 7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 |
se->parent = parent; } static DEFINE_MUTEX(shares_mutex); int sched_group_set_shares(struct task_group *tg, unsigned long shares) { int i; unsigned long flags; /* * We can't change the weight of the root cgroup. */ if (!tg->se[0]) return -EINVAL; shares = clamp(shares, scale_load(MIN_SHARES), scale_load(MAX_SHARES)); mutex_lock(&shares_mutex); if (tg->shares == shares) goto done; tg->shares = shares; for_each_possible_cpu(i) { struct rq *rq = cpu_rq(i); struct sched_entity *se; se = tg->se[i]; /* Propagate contribution to hierarchy */ raw_spin_lock_irqsave(&rq->lock, flags); |
71b1da46f sched: Update rq ... |
7213 7214 7215 |
/* Possible calls to update_curr() need rq clock */ update_rq_clock(rq); |
17bc14b76 Revert "sched: Up... |
7216 |
for_each_sched_entity(se) |
029632fbb sched: Make separ... |
7217 7218 7219 7220 7221 7222 7223 7224 7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235 7236 |
update_cfs_shares(group_cfs_rq(se)); raw_spin_unlock_irqrestore(&rq->lock, flags); } done: mutex_unlock(&shares_mutex); return 0; } #else /* CONFIG_FAIR_GROUP_SCHED */ void free_fair_sched_group(struct task_group *tg) { } int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) { return 1; } void unregister_fair_sched_group(struct task_group *tg, int cpu) { } #endif /* CONFIG_FAIR_GROUP_SCHED */ |
810b38179 sched: retain vru... |
7237 |
|
6d686f456 sched: Don't expo... |
7238 |
static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task) |
0d721cead sched: Simplify s... |
7239 7240 |
{ struct sched_entity *se = &task->se; |
0d721cead sched: Simplify s... |
7241 7242 7243 7244 7245 7246 |
unsigned int rr_interval = 0; /* * Time slice is 0 for SCHED_OTHER tasks that are on an otherwise * idle runqueue: */ |
0d721cead sched: Simplify s... |
7247 |
if (rq->cfs.load.weight) |
a59f4e079 sched: Fix the br... |
7248 |
rr_interval = NS_TO_JIFFIES(sched_slice(cfs_rq_of(se), se)); |
0d721cead sched: Simplify s... |
7249 7250 7251 |
return rr_interval; } |
bf0f6f24a sched: cfs core, ... |
7252 7253 7254 |
/* * All the scheduling class methods: */ |
029632fbb sched: Make separ... |
7255 |
const struct sched_class fair_sched_class = { |
5522d5d5f sched: mark sched... |
7256 |
.next = &idle_sched_class, |
bf0f6f24a sched: cfs core, ... |
7257 7258 7259 |
.enqueue_task = enqueue_task_fair, .dequeue_task = dequeue_task_fair, .yield_task = yield_task_fair, |
d95f41220 sched: Add yield_... |
7260 |
.yield_to_task = yield_to_task_fair, |
bf0f6f24a sched: cfs core, ... |
7261 |
|
2e09bf556 sched: wakeup gra... |
7262 |
.check_preempt_curr = check_preempt_wakeup, |
bf0f6f24a sched: cfs core, ... |
7263 7264 7265 |
.pick_next_task = pick_next_task_fair, .put_prev_task = put_prev_task_fair, |
681f3e685 sched: isolate SM... |
7266 |
#ifdef CONFIG_SMP |
4ce72a2c0 sched: add CONFIG... |
7267 |
.select_task_rq = select_task_rq_fair, |
0a74bef8b sched: Add an rq ... |
7268 |
.migrate_task_rq = migrate_task_rq_fair, |
141965c74 Revert "sched: In... |
7269 |
|
0bcdcf28c sched: Fix missin... |
7270 7271 |
.rq_online = rq_online_fair, .rq_offline = rq_offline_fair, |
88ec22d3e sched: Remove the... |
7272 7273 |
.task_waking = task_waking_fair, |
681f3e685 sched: isolate SM... |
7274 |
#endif |
bf0f6f24a sched: cfs core, ... |
7275 |
|
83b699ed2 sched: revert rec... |
7276 |
.set_curr_task = set_curr_task_fair, |
bf0f6f24a sched: cfs core, ... |
7277 |
.task_tick = task_tick_fair, |
cd29fe6f2 sched: Sanitize f... |
7278 |
.task_fork = task_fork_fair, |
cb4698450 sched: RT-balance... |
7279 7280 |
.prio_changed = prio_changed_fair, |
da7a735e5 sched: Fix switch... |
7281 |
.switched_from = switched_from_fair, |
cb4698450 sched: RT-balance... |
7282 |
.switched_to = switched_to_fair, |
810b38179 sched: retain vru... |
7283 |
|
0d721cead sched: Simplify s... |
7284 |
.get_rr_interval = get_rr_interval_fair, |
810b38179 sched: retain vru... |
7285 |
#ifdef CONFIG_FAIR_GROUP_SCHED |
b2b5ce022 sched, cgroup: Fi... |
7286 |
.task_move_group = task_move_group_fair, |
810b38179 sched: retain vru... |
7287 |
#endif |
bf0f6f24a sched: cfs core, ... |
7288 7289 7290 |
}; #ifdef CONFIG_SCHED_DEBUG |
029632fbb sched: Make separ... |
7291 |
void print_cfs_stats(struct seq_file *m, int cpu) |
bf0f6f24a sched: cfs core, ... |
7292 |
{ |
bf0f6f24a sched: cfs core, ... |
7293 |
struct cfs_rq *cfs_rq; |
5973e5b95 sched: fix: don't... |
7294 |
rcu_read_lock(); |
c3b64f1e4 sched: clean up s... |
7295 |
for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq) |
5cef9eca3 sched: remove the... |
7296 |
print_cfs_rq(m, cpu, cfs_rq); |
5973e5b95 sched: fix: don't... |
7297 |
rcu_read_unlock(); |
bf0f6f24a sched: cfs core, ... |
7298 7299 |
} #endif |
029632fbb sched: Make separ... |
7300 7301 7302 7303 7304 |
__init void init_sched_fair_class(void) { #ifdef CONFIG_SMP open_softirq(SCHED_SOFTIRQ, run_rebalance_domains); |
3451d0243 nohz: Rename CONF... |
7305 |
#ifdef CONFIG_NO_HZ_COMMON |
554cecaf7 sched/nohz: Corre... |
7306 |
nohz.next_balance = jiffies; |
029632fbb sched: Make separ... |
7307 |
zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT); |
71325960d sched/nohz: Fix n... |
7308 |
cpu_notifier(sched_ilb_notifier, 0); |
029632fbb sched: Make separ... |
7309 7310 7311 7312 |
#endif #endif /* SMP */ } |