Merge tag 'sound-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
[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 /**
141 * mem_cgroup_toggle_oom - toggle the memcg OOM killer for the current task
142 * @new: true to enable, false to disable
143 *
144 * Toggle whether a failed memcg charge should invoke the OOM killer
145 * or just return -ENOMEM. Returns the previous toggle state.
146 *
147 * NOTE: Any path that enables the OOM killer before charging must
148 * call mem_cgroup_oom_synchronize() afterward to finalize the
149 * OOM handling and clean up.
150 */
151 static inline bool mem_cgroup_toggle_oom(bool new)
152 {
153 bool old;
154
155 old = current->memcg_oom.may_oom;
156 current->memcg_oom.may_oom = new;
157
158 return old;
159 }
160
161 static inline void mem_cgroup_enable_oom(void)
162 {
163 bool old = mem_cgroup_toggle_oom(true);
164
165 WARN_ON(old == true);
166 }
167
168 static inline void mem_cgroup_disable_oom(void)
169 {
170 bool old = mem_cgroup_toggle_oom(false);
171
172 WARN_ON(old == false);
173 }
174
175 static inline bool task_in_memcg_oom(struct task_struct *p)
176 {
177 return p->memcg_oom.in_memcg_oom;
178 }
179
180 bool mem_cgroup_oom_synchronize(void);
181
182 #ifdef CONFIG_MEMCG_SWAP
183 extern int do_swap_account;
184 #endif
185
186 static inline bool mem_cgroup_disabled(void)
187 {
188 if (mem_cgroup_subsys.disabled)
189 return true;
190 return false;
191 }
192
193 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
194 unsigned long *flags);
195
196 extern atomic_t memcg_moving;
197
198 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
199 bool *locked, unsigned long *flags)
200 {
201 if (mem_cgroup_disabled())
202 return;
203 rcu_read_lock();
204 *locked = false;
205 if (atomic_read(&memcg_moving))
206 __mem_cgroup_begin_update_page_stat(page, locked, flags);
207 }
208
209 void __mem_cgroup_end_update_page_stat(struct page *page,
210 unsigned long *flags);
211 static inline void mem_cgroup_end_update_page_stat(struct page *page,
212 bool *locked, unsigned long *flags)
213 {
214 if (mem_cgroup_disabled())
215 return;
216 if (*locked)
217 __mem_cgroup_end_update_page_stat(page, flags);
218 rcu_read_unlock();
219 }
220
221 void mem_cgroup_update_page_stat(struct page *page,
222 enum mem_cgroup_stat_index idx,
223 int val);
224
225 static inline void mem_cgroup_inc_page_stat(struct page *page,
226 enum mem_cgroup_stat_index idx)
227 {
228 mem_cgroup_update_page_stat(page, idx, 1);
229 }
230
231 static inline void mem_cgroup_dec_page_stat(struct page *page,
232 enum mem_cgroup_stat_index idx)
233 {
234 mem_cgroup_update_page_stat(page, idx, -1);
235 }
236
237 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
238 gfp_t gfp_mask,
239 unsigned long *total_scanned);
240
241 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
242 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
243 enum vm_event_item idx)
244 {
245 if (mem_cgroup_disabled())
246 return;
247 __mem_cgroup_count_vm_event(mm, idx);
248 }
249 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
250 void mem_cgroup_split_huge_fixup(struct page *head);
251 #endif
252
253 #ifdef CONFIG_DEBUG_VM
254 bool mem_cgroup_bad_page_check(struct page *page);
255 void mem_cgroup_print_bad_page(struct page *page);
256 #endif
257 #else /* CONFIG_MEMCG */
258 struct mem_cgroup;
259
260 static inline int mem_cgroup_newpage_charge(struct page *page,
261 struct mm_struct *mm, gfp_t gfp_mask)
262 {
263 return 0;
264 }
265
266 static inline int mem_cgroup_cache_charge(struct page *page,
267 struct mm_struct *mm, gfp_t gfp_mask)
268 {
269 return 0;
270 }
271
272 static inline int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
273 struct page *page, gfp_t gfp_mask, struct mem_cgroup **memcgp)
274 {
275 return 0;
276 }
277
278 static inline void mem_cgroup_commit_charge_swapin(struct page *page,
279 struct mem_cgroup *memcg)
280 {
281 }
282
283 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
284 {
285 }
286
287 static inline void mem_cgroup_uncharge_start(void)
288 {
289 }
290
291 static inline void mem_cgroup_uncharge_end(void)
292 {
293 }
294
295 static inline void mem_cgroup_uncharge_page(struct page *page)
296 {
297 }
298
299 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
300 {
301 }
302
303 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
304 struct mem_cgroup *memcg)
305 {
306 return &zone->lruvec;
307 }
308
309 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
310 struct zone *zone)
311 {
312 return &zone->lruvec;
313 }
314
315 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
316 {
317 return NULL;
318 }
319
320 static inline struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
321 {
322 return NULL;
323 }
324
325 static inline bool mm_match_cgroup(struct mm_struct *mm,
326 struct mem_cgroup *memcg)
327 {
328 return true;
329 }
330
331 static inline bool task_in_mem_cgroup(struct task_struct *task,
332 const struct mem_cgroup *memcg)
333 {
334 return true;
335 }
336
337 static inline struct cgroup_subsys_state
338 *mem_cgroup_css(struct mem_cgroup *memcg)
339 {
340 return NULL;
341 }
342
343 static inline void
344 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
345 struct mem_cgroup **memcgp)
346 {
347 }
348
349 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
350 struct page *oldpage, struct page *newpage, bool migration_ok)
351 {
352 }
353
354 static inline struct mem_cgroup *
355 mem_cgroup_iter(struct mem_cgroup *root,
356 struct mem_cgroup *prev,
357 struct mem_cgroup_reclaim_cookie *reclaim)
358 {
359 return NULL;
360 }
361
362 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
363 struct mem_cgroup *prev)
364 {
365 }
366
367 static inline bool mem_cgroup_disabled(void)
368 {
369 return true;
370 }
371
372 static inline int
373 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
374 {
375 return 1;
376 }
377
378 static inline unsigned long
379 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
380 {
381 return 0;
382 }
383
384 static inline void
385 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
386 int increment)
387 {
388 }
389
390 static inline void
391 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
392 {
393 }
394
395 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
396 bool *locked, unsigned long *flags)
397 {
398 }
399
400 static inline void mem_cgroup_end_update_page_stat(struct page *page,
401 bool *locked, unsigned long *flags)
402 {
403 }
404
405 static inline bool mem_cgroup_toggle_oom(bool new)
406 {
407 return false;
408 }
409
410 static inline void mem_cgroup_enable_oom(void)
411 {
412 }
413
414 static inline void mem_cgroup_disable_oom(void)
415 {
416 }
417
418 static inline bool task_in_memcg_oom(struct task_struct *p)
419 {
420 return false;
421 }
422
423 static inline bool mem_cgroup_oom_synchronize(void)
424 {
425 return false;
426 }
427
428 static inline void mem_cgroup_inc_page_stat(struct page *page,
429 enum mem_cgroup_stat_index idx)
430 {
431 }
432
433 static inline void mem_cgroup_dec_page_stat(struct page *page,
434 enum mem_cgroup_stat_index idx)
435 {
436 }
437
438 static inline
439 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
440 gfp_t gfp_mask,
441 unsigned long *total_scanned)
442 {
443 return 0;
444 }
445
446 static inline void mem_cgroup_split_huge_fixup(struct page *head)
447 {
448 }
449
450 static inline
451 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
452 {
453 }
454 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
455 struct page *newpage)
456 {
457 }
458 #endif /* CONFIG_MEMCG */
459
460 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
461 static inline bool
462 mem_cgroup_bad_page_check(struct page *page)
463 {
464 return false;
465 }
466
467 static inline void
468 mem_cgroup_print_bad_page(struct page *page)
469 {
470 }
471 #endif
472
473 enum {
474 UNDER_LIMIT,
475 SOFT_LIMIT,
476 OVER_LIMIT,
477 };
478
479 struct sock;
480 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
481 void sock_update_memcg(struct sock *sk);
482 void sock_release_memcg(struct sock *sk);
483 #else
484 static inline void sock_update_memcg(struct sock *sk)
485 {
486 }
487 static inline void sock_release_memcg(struct sock *sk)
488 {
489 }
490 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
491
492 #ifdef CONFIG_MEMCG_KMEM
493 extern struct static_key memcg_kmem_enabled_key;
494
495 extern int memcg_limited_groups_array_size;
496
497 /*
498 * Helper macro to loop through all memcg-specific caches. Callers must still
499 * check if the cache is valid (it is either valid or NULL).
500 * the slab_mutex must be held when looping through those caches
501 */
502 #define for_each_memcg_cache_index(_idx) \
503 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
504
505 static inline bool memcg_kmem_enabled(void)
506 {
507 return static_key_false(&memcg_kmem_enabled_key);
508 }
509
510 /*
511 * In general, we'll do everything in our power to not incur in any overhead
512 * for non-memcg users for the kmem functions. Not even a function call, if we
513 * can avoid it.
514 *
515 * Therefore, we'll inline all those functions so that in the best case, we'll
516 * see that kmemcg is off for everybody and proceed quickly. If it is on,
517 * we'll still do most of the flag checking inline. We check a lot of
518 * conditions, but because they are pretty simple, they are expected to be
519 * fast.
520 */
521 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
522 int order);
523 void __memcg_kmem_commit_charge(struct page *page,
524 struct mem_cgroup *memcg, int order);
525 void __memcg_kmem_uncharge_pages(struct page *page, int order);
526
527 int memcg_cache_id(struct mem_cgroup *memcg);
528 int memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
529 struct kmem_cache *root_cache);
530 void memcg_release_cache(struct kmem_cache *cachep);
531 void memcg_cache_list_add(struct mem_cgroup *memcg, struct kmem_cache *cachep);
532
533 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
534 void memcg_update_array_size(int num_groups);
535
536 struct kmem_cache *
537 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
538
539 void mem_cgroup_destroy_cache(struct kmem_cache *cachep);
540 void kmem_cache_destroy_memcg_children(struct kmem_cache *s);
541
542 /**
543 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
544 * @gfp: the gfp allocation flags.
545 * @memcg: a pointer to the memcg this was charged against.
546 * @order: allocation order.
547 *
548 * returns true if the memcg where the current task belongs can hold this
549 * allocation.
550 *
551 * We return true automatically if this allocation is not to be accounted to
552 * any memcg.
553 */
554 static inline bool
555 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
556 {
557 if (!memcg_kmem_enabled())
558 return true;
559
560 /*
561 * __GFP_NOFAIL allocations will move on even if charging is not
562 * possible. Therefore we don't even try, and have this allocation
563 * unaccounted. We could in theory charge it with
564 * res_counter_charge_nofail, but we hope those allocations are rare,
565 * and won't be worth the trouble.
566 */
567 if (!(gfp & __GFP_KMEMCG) || (gfp & __GFP_NOFAIL))
568 return true;
569 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
570 return true;
571
572 /* If the test is dying, just let it go. */
573 if (unlikely(fatal_signal_pending(current)))
574 return true;
575
576 return __memcg_kmem_newpage_charge(gfp, memcg, order);
577 }
578
579 /**
580 * memcg_kmem_uncharge_pages: uncharge pages from memcg
581 * @page: pointer to struct page being freed
582 * @order: allocation order.
583 *
584 * there is no need to specify memcg here, since it is embedded in page_cgroup
585 */
586 static inline void
587 memcg_kmem_uncharge_pages(struct page *page, int order)
588 {
589 if (memcg_kmem_enabled())
590 __memcg_kmem_uncharge_pages(page, order);
591 }
592
593 /**
594 * memcg_kmem_commit_charge: embeds correct memcg in a page
595 * @page: pointer to struct page recently allocated
596 * @memcg: the memcg structure we charged against
597 * @order: allocation order.
598 *
599 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
600 * failure of the allocation. if @page is NULL, this function will revert the
601 * charges. Otherwise, it will commit the memcg given by @memcg to the
602 * corresponding page_cgroup.
603 */
604 static inline void
605 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
606 {
607 if (memcg_kmem_enabled() && memcg)
608 __memcg_kmem_commit_charge(page, memcg, order);
609 }
610
611 /**
612 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
613 * @cachep: the original global kmem cache
614 * @gfp: allocation flags.
615 *
616 * This function assumes that the task allocating, which determines the memcg
617 * in the page allocator, belongs to the same cgroup throughout the whole
618 * process. Misacounting can happen if the task calls memcg_kmem_get_cache()
619 * while belonging to a cgroup, and later on changes. This is considered
620 * acceptable, and should only happen upon task migration.
621 *
622 * Before the cache is created by the memcg core, there is also a possible
623 * imbalance: the task belongs to a memcg, but the cache being allocated from
624 * is the global cache, since the child cache is not yet guaranteed to be
625 * ready. This case is also fine, since in this case the GFP_KMEMCG will not be
626 * passed and the page allocator will not attempt any cgroup accounting.
627 */
628 static __always_inline struct kmem_cache *
629 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
630 {
631 if (!memcg_kmem_enabled())
632 return cachep;
633 if (gfp & __GFP_NOFAIL)
634 return cachep;
635 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
636 return cachep;
637 if (unlikely(fatal_signal_pending(current)))
638 return cachep;
639
640 return __memcg_kmem_get_cache(cachep, gfp);
641 }
642 #else
643 #define for_each_memcg_cache_index(_idx) \
644 for (; NULL; )
645
646 static inline bool memcg_kmem_enabled(void)
647 {
648 return false;
649 }
650
651 static inline bool
652 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
653 {
654 return true;
655 }
656
657 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
658 {
659 }
660
661 static inline void
662 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
663 {
664 }
665
666 static inline int memcg_cache_id(struct mem_cgroup *memcg)
667 {
668 return -1;
669 }
670
671 static inline int
672 memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
673 struct kmem_cache *root_cache)
674 {
675 return 0;
676 }
677
678 static inline void memcg_release_cache(struct kmem_cache *cachep)
679 {
680 }
681
682 static inline void memcg_cache_list_add(struct mem_cgroup *memcg,
683 struct kmem_cache *s)
684 {
685 }
686
687 static inline struct kmem_cache *
688 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
689 {
690 return cachep;
691 }
692
693 static inline void kmem_cache_destroy_memcg_children(struct kmem_cache *s)
694 {
695 }
696 #endif /* CONFIG_MEMCG_KMEM */
697 #endif /* _LINUX_MEMCONTROL_H */
698
This page took 0.060192 seconds and 6 git commands to generate.