Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[deliverable/linux.git] / include / linux / memcontrol.h
1 /* memcontrol.h - Memory Controller
2 *
3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
5 *
6 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org>
8 *
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
18 */
19
20 #ifndef _LINUX_MEMCONTROL_H
21 #define _LINUX_MEMCONTROL_H
22 #include <linux/cgroup.h>
23 #include <linux/vm_event_item.h>
24 #include <linux/hardirq.h>
25 #include <linux/jump_label.h>
26
27 struct mem_cgroup;
28 struct page_cgroup;
29 struct page;
30 struct mm_struct;
31 struct kmem_cache;
32
33 /*
34 * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
35 * These two lists should keep in accord with each other.
36 */
37 enum mem_cgroup_stat_index {
38 /*
39 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
40 */
41 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
42 MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */
43 MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */
44 MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */
45 MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
46 MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
47 MEM_CGROUP_STAT_NSTATS,
48 };
49
50 struct mem_cgroup_reclaim_cookie {
51 struct zone *zone;
52 int priority;
53 unsigned int generation;
54 };
55
56 #ifdef CONFIG_MEMCG
57 /*
58 * All "charge" functions with gfp_mask should use GFP_KERNEL or
59 * (gfp_mask & GFP_RECLAIM_MASK). In current implementatin, memcg doesn't
60 * alloc memory but reclaims memory from all available zones. So, "where I want
61 * memory from" bits of gfp_mask has no meaning. So any bits of that field is
62 * available but adding a rule is better. charge functions' gfp_mask should
63 * be set to GFP_KERNEL or gfp_mask & GFP_RECLAIM_MASK for avoiding ambiguous
64 * codes.
65 * (Of course, if memcg does memory allocation in future, GFP_KERNEL is sane.)
66 */
67
68 extern int mem_cgroup_newpage_charge(struct page *page, struct mm_struct *mm,
69 gfp_t gfp_mask);
70 /* for swap handling */
71 extern int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
72 struct page *page, gfp_t mask, struct mem_cgroup **memcgp);
73 extern void mem_cgroup_commit_charge_swapin(struct page *page,
74 struct mem_cgroup *memcg);
75 extern void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg);
76
77 extern int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
78 gfp_t gfp_mask);
79
80 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
81 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
82
83 /* For coalescing uncharge for reducing memcg' overhead*/
84 extern void mem_cgroup_uncharge_start(void);
85 extern void mem_cgroup_uncharge_end(void);
86
87 extern void mem_cgroup_uncharge_page(struct page *page);
88 extern void mem_cgroup_uncharge_cache_page(struct page *page);
89
90 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
91 struct mem_cgroup *memcg);
92 bool task_in_mem_cgroup(struct task_struct *task,
93 const struct mem_cgroup *memcg);
94
95 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
96 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
97 extern struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm);
98
99 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
100 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
101
102 static inline
103 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
104 {
105 struct mem_cgroup *task_memcg;
106 bool match;
107
108 rcu_read_lock();
109 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
110 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
111 rcu_read_unlock();
112 return match;
113 }
114
115 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
116
117 extern void
118 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
119 struct mem_cgroup **memcgp);
120 extern void mem_cgroup_end_migration(struct mem_cgroup *memcg,
121 struct page *oldpage, struct page *newpage, bool migration_ok);
122
123 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
124 struct mem_cgroup *,
125 struct mem_cgroup_reclaim_cookie *);
126 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
127
128 /*
129 * For memory reclaim.
130 */
131 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
132 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
133 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
134 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
135 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
136 struct task_struct *p);
137 extern void mem_cgroup_replace_page_cache(struct page *oldpage,
138 struct page *newpage);
139
140 static inline void mem_cgroup_oom_enable(void)
141 {
142 WARN_ON(current->memcg_oom.may_oom);
143 current->memcg_oom.may_oom = 1;
144 }
145
146 static inline void mem_cgroup_oom_disable(void)
147 {
148 WARN_ON(!current->memcg_oom.may_oom);
149 current->memcg_oom.may_oom = 0;
150 }
151
152 static inline bool task_in_memcg_oom(struct task_struct *p)
153 {
154 return p->memcg_oom.memcg;
155 }
156
157 bool mem_cgroup_oom_synchronize(bool wait);
158
159 #ifdef CONFIG_MEMCG_SWAP
160 extern int do_swap_account;
161 #endif
162
163 static inline bool mem_cgroup_disabled(void)
164 {
165 if (mem_cgroup_subsys.disabled)
166 return true;
167 return false;
168 }
169
170 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
171 unsigned long *flags);
172
173 extern atomic_t memcg_moving;
174
175 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
176 bool *locked, unsigned long *flags)
177 {
178 if (mem_cgroup_disabled())
179 return;
180 rcu_read_lock();
181 *locked = false;
182 if (atomic_read(&memcg_moving))
183 __mem_cgroup_begin_update_page_stat(page, locked, flags);
184 }
185
186 void __mem_cgroup_end_update_page_stat(struct page *page,
187 unsigned long *flags);
188 static inline void mem_cgroup_end_update_page_stat(struct page *page,
189 bool *locked, unsigned long *flags)
190 {
191 if (mem_cgroup_disabled())
192 return;
193 if (*locked)
194 __mem_cgroup_end_update_page_stat(page, flags);
195 rcu_read_unlock();
196 }
197
198 void mem_cgroup_update_page_stat(struct page *page,
199 enum mem_cgroup_stat_index idx,
200 int val);
201
202 static inline void mem_cgroup_inc_page_stat(struct page *page,
203 enum mem_cgroup_stat_index idx)
204 {
205 mem_cgroup_update_page_stat(page, idx, 1);
206 }
207
208 static inline void mem_cgroup_dec_page_stat(struct page *page,
209 enum mem_cgroup_stat_index idx)
210 {
211 mem_cgroup_update_page_stat(page, idx, -1);
212 }
213
214 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
215 gfp_t gfp_mask,
216 unsigned long *total_scanned);
217
218 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
219 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
220 enum vm_event_item idx)
221 {
222 if (mem_cgroup_disabled())
223 return;
224 __mem_cgroup_count_vm_event(mm, idx);
225 }
226 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
227 void mem_cgroup_split_huge_fixup(struct page *head);
228 #endif
229
230 #ifdef CONFIG_DEBUG_VM
231 bool mem_cgroup_bad_page_check(struct page *page);
232 void mem_cgroup_print_bad_page(struct page *page);
233 #endif
234 #else /* CONFIG_MEMCG */
235 struct mem_cgroup;
236
237 static inline int mem_cgroup_newpage_charge(struct page *page,
238 struct mm_struct *mm, gfp_t gfp_mask)
239 {
240 return 0;
241 }
242
243 static inline int mem_cgroup_cache_charge(struct page *page,
244 struct mm_struct *mm, gfp_t gfp_mask)
245 {
246 return 0;
247 }
248
249 static inline int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
250 struct page *page, gfp_t gfp_mask, struct mem_cgroup **memcgp)
251 {
252 return 0;
253 }
254
255 static inline void mem_cgroup_commit_charge_swapin(struct page *page,
256 struct mem_cgroup *memcg)
257 {
258 }
259
260 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
261 {
262 }
263
264 static inline void mem_cgroup_uncharge_start(void)
265 {
266 }
267
268 static inline void mem_cgroup_uncharge_end(void)
269 {
270 }
271
272 static inline void mem_cgroup_uncharge_page(struct page *page)
273 {
274 }
275
276 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
277 {
278 }
279
280 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
281 struct mem_cgroup *memcg)
282 {
283 return &zone->lruvec;
284 }
285
286 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
287 struct zone *zone)
288 {
289 return &zone->lruvec;
290 }
291
292 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
293 {
294 return NULL;
295 }
296
297 static inline struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
298 {
299 return NULL;
300 }
301
302 static inline bool mm_match_cgroup(struct mm_struct *mm,
303 struct mem_cgroup *memcg)
304 {
305 return true;
306 }
307
308 static inline bool task_in_mem_cgroup(struct task_struct *task,
309 const struct mem_cgroup *memcg)
310 {
311 return true;
312 }
313
314 static inline struct cgroup_subsys_state
315 *mem_cgroup_css(struct mem_cgroup *memcg)
316 {
317 return NULL;
318 }
319
320 static inline void
321 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
322 struct mem_cgroup **memcgp)
323 {
324 }
325
326 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
327 struct page *oldpage, struct page *newpage, bool migration_ok)
328 {
329 }
330
331 static inline struct mem_cgroup *
332 mem_cgroup_iter(struct mem_cgroup *root,
333 struct mem_cgroup *prev,
334 struct mem_cgroup_reclaim_cookie *reclaim)
335 {
336 return NULL;
337 }
338
339 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
340 struct mem_cgroup *prev)
341 {
342 }
343
344 static inline bool mem_cgroup_disabled(void)
345 {
346 return true;
347 }
348
349 static inline int
350 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
351 {
352 return 1;
353 }
354
355 static inline unsigned long
356 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
357 {
358 return 0;
359 }
360
361 static inline void
362 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
363 int increment)
364 {
365 }
366
367 static inline void
368 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
369 {
370 }
371
372 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
373 bool *locked, unsigned long *flags)
374 {
375 }
376
377 static inline void mem_cgroup_end_update_page_stat(struct page *page,
378 bool *locked, unsigned long *flags)
379 {
380 }
381
382 static inline void mem_cgroup_oom_enable(void)
383 {
384 }
385
386 static inline void mem_cgroup_oom_disable(void)
387 {
388 }
389
390 static inline bool task_in_memcg_oom(struct task_struct *p)
391 {
392 return false;
393 }
394
395 static inline bool mem_cgroup_oom_synchronize(bool wait)
396 {
397 return false;
398 }
399
400 static inline void mem_cgroup_inc_page_stat(struct page *page,
401 enum mem_cgroup_stat_index idx)
402 {
403 }
404
405 static inline void mem_cgroup_dec_page_stat(struct page *page,
406 enum mem_cgroup_stat_index idx)
407 {
408 }
409
410 static inline
411 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
412 gfp_t gfp_mask,
413 unsigned long *total_scanned)
414 {
415 return 0;
416 }
417
418 static inline void mem_cgroup_split_huge_fixup(struct page *head)
419 {
420 }
421
422 static inline
423 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
424 {
425 }
426 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
427 struct page *newpage)
428 {
429 }
430 #endif /* CONFIG_MEMCG */
431
432 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
433 static inline bool
434 mem_cgroup_bad_page_check(struct page *page)
435 {
436 return false;
437 }
438
439 static inline void
440 mem_cgroup_print_bad_page(struct page *page)
441 {
442 }
443 #endif
444
445 enum {
446 UNDER_LIMIT,
447 SOFT_LIMIT,
448 OVER_LIMIT,
449 };
450
451 struct sock;
452 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
453 void sock_update_memcg(struct sock *sk);
454 void sock_release_memcg(struct sock *sk);
455 #else
456 static inline void sock_update_memcg(struct sock *sk)
457 {
458 }
459 static inline void sock_release_memcg(struct sock *sk)
460 {
461 }
462 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
463
464 #ifdef CONFIG_MEMCG_KMEM
465 extern struct static_key memcg_kmem_enabled_key;
466
467 extern int memcg_limited_groups_array_size;
468
469 /*
470 * Helper macro to loop through all memcg-specific caches. Callers must still
471 * check if the cache is valid (it is either valid or NULL).
472 * the slab_mutex must be held when looping through those caches
473 */
474 #define for_each_memcg_cache_index(_idx) \
475 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
476
477 static inline bool memcg_kmem_enabled(void)
478 {
479 return static_key_false(&memcg_kmem_enabled_key);
480 }
481
482 /*
483 * In general, we'll do everything in our power to not incur in any overhead
484 * for non-memcg users for the kmem functions. Not even a function call, if we
485 * can avoid it.
486 *
487 * Therefore, we'll inline all those functions so that in the best case, we'll
488 * see that kmemcg is off for everybody and proceed quickly. If it is on,
489 * we'll still do most of the flag checking inline. We check a lot of
490 * conditions, but because they are pretty simple, they are expected to be
491 * fast.
492 */
493 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
494 int order);
495 void __memcg_kmem_commit_charge(struct page *page,
496 struct mem_cgroup *memcg, int order);
497 void __memcg_kmem_uncharge_pages(struct page *page, int order);
498
499 int memcg_cache_id(struct mem_cgroup *memcg);
500 int memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
501 struct kmem_cache *root_cache);
502 void memcg_release_cache(struct kmem_cache *cachep);
503 void memcg_cache_list_add(struct mem_cgroup *memcg, struct kmem_cache *cachep);
504
505 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
506 void memcg_update_array_size(int num_groups);
507
508 struct kmem_cache *
509 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
510
511 void mem_cgroup_destroy_cache(struct kmem_cache *cachep);
512 void kmem_cache_destroy_memcg_children(struct kmem_cache *s);
513
514 /**
515 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
516 * @gfp: the gfp allocation flags.
517 * @memcg: a pointer to the memcg this was charged against.
518 * @order: allocation order.
519 *
520 * returns true if the memcg where the current task belongs can hold this
521 * allocation.
522 *
523 * We return true automatically if this allocation is not to be accounted to
524 * any memcg.
525 */
526 static inline bool
527 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
528 {
529 if (!memcg_kmem_enabled())
530 return true;
531
532 /*
533 * __GFP_NOFAIL allocations will move on even if charging is not
534 * possible. Therefore we don't even try, and have this allocation
535 * unaccounted. We could in theory charge it with
536 * res_counter_charge_nofail, but we hope those allocations are rare,
537 * and won't be worth the trouble.
538 */
539 if (!(gfp & __GFP_KMEMCG) || (gfp & __GFP_NOFAIL))
540 return true;
541 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
542 return true;
543
544 /* If the test is dying, just let it go. */
545 if (unlikely(fatal_signal_pending(current)))
546 return true;
547
548 return __memcg_kmem_newpage_charge(gfp, memcg, order);
549 }
550
551 /**
552 * memcg_kmem_uncharge_pages: uncharge pages from memcg
553 * @page: pointer to struct page being freed
554 * @order: allocation order.
555 *
556 * there is no need to specify memcg here, since it is embedded in page_cgroup
557 */
558 static inline void
559 memcg_kmem_uncharge_pages(struct page *page, int order)
560 {
561 if (memcg_kmem_enabled())
562 __memcg_kmem_uncharge_pages(page, order);
563 }
564
565 /**
566 * memcg_kmem_commit_charge: embeds correct memcg in a page
567 * @page: pointer to struct page recently allocated
568 * @memcg: the memcg structure we charged against
569 * @order: allocation order.
570 *
571 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
572 * failure of the allocation. if @page is NULL, this function will revert the
573 * charges. Otherwise, it will commit the memcg given by @memcg to the
574 * corresponding page_cgroup.
575 */
576 static inline void
577 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
578 {
579 if (memcg_kmem_enabled() && memcg)
580 __memcg_kmem_commit_charge(page, memcg, order);
581 }
582
583 /**
584 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
585 * @cachep: the original global kmem cache
586 * @gfp: allocation flags.
587 *
588 * This function assumes that the task allocating, which determines the memcg
589 * in the page allocator, belongs to the same cgroup throughout the whole
590 * process. Misacounting can happen if the task calls memcg_kmem_get_cache()
591 * while belonging to a cgroup, and later on changes. This is considered
592 * acceptable, and should only happen upon task migration.
593 *
594 * Before the cache is created by the memcg core, there is also a possible
595 * imbalance: the task belongs to a memcg, but the cache being allocated from
596 * is the global cache, since the child cache is not yet guaranteed to be
597 * ready. This case is also fine, since in this case the GFP_KMEMCG will not be
598 * passed and the page allocator will not attempt any cgroup accounting.
599 */
600 static __always_inline struct kmem_cache *
601 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
602 {
603 if (!memcg_kmem_enabled())
604 return cachep;
605 if (gfp & __GFP_NOFAIL)
606 return cachep;
607 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
608 return cachep;
609 if (unlikely(fatal_signal_pending(current)))
610 return cachep;
611
612 return __memcg_kmem_get_cache(cachep, gfp);
613 }
614 #else
615 #define for_each_memcg_cache_index(_idx) \
616 for (; NULL; )
617
618 static inline bool memcg_kmem_enabled(void)
619 {
620 return false;
621 }
622
623 static inline bool
624 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
625 {
626 return true;
627 }
628
629 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
630 {
631 }
632
633 static inline void
634 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
635 {
636 }
637
638 static inline int memcg_cache_id(struct mem_cgroup *memcg)
639 {
640 return -1;
641 }
642
643 static inline int
644 memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
645 struct kmem_cache *root_cache)
646 {
647 return 0;
648 }
649
650 static inline void memcg_release_cache(struct kmem_cache *cachep)
651 {
652 }
653
654 static inline void memcg_cache_list_add(struct mem_cgroup *memcg,
655 struct kmem_cache *s)
656 {
657 }
658
659 static inline struct kmem_cache *
660 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
661 {
662 return cachep;
663 }
664
665 static inline void kmem_cache_destroy_memcg_children(struct kmem_cache *s)
666 {
667 }
668 #endif /* CONFIG_MEMCG_KMEM */
669 #endif /* _LINUX_MEMCONTROL_H */
670
This page took 0.045071 seconds and 5 git commands to generate.