1 /* memcontrol.h - Memory Controller
3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
6 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org>
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.
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.
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>
34 * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
35 * These two lists should keep in accord with each other.
37 enum mem_cgroup_stat_index {
39 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
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,
50 struct mem_cgroup_reclaim_cookie {
53 unsigned int generation;
57 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
58 gfp_t gfp_mask, struct mem_cgroup **memcgp);
59 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
61 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
63 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
64 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
66 /* For coalescing uncharge for reducing memcg' overhead*/
67 extern void mem_cgroup_uncharge_start(void);
68 extern void mem_cgroup_uncharge_end(void);
70 extern void mem_cgroup_uncharge_page(struct page *page);
71 extern void mem_cgroup_uncharge_cache_page(struct page *page);
73 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
74 struct mem_cgroup *memcg);
75 bool task_in_mem_cgroup(struct task_struct *task,
76 const struct mem_cgroup *memcg);
78 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
79 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
81 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
82 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
85 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
87 struct mem_cgroup *task_memcg;
91 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
92 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
97 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
100 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
101 struct mem_cgroup **memcgp);
102 extern void mem_cgroup_end_migration(struct mem_cgroup *memcg,
103 struct page *oldpage, struct page *newpage, bool migration_ok);
105 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
107 struct mem_cgroup_reclaim_cookie *);
108 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
111 * For memory reclaim.
113 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
114 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
115 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
116 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
117 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
118 struct task_struct *p);
119 extern void mem_cgroup_replace_page_cache(struct page *oldpage,
120 struct page *newpage);
122 static inline void mem_cgroup_oom_enable(void)
124 WARN_ON(current->memcg_oom.may_oom);
125 current->memcg_oom.may_oom = 1;
128 static inline void mem_cgroup_oom_disable(void)
130 WARN_ON(!current->memcg_oom.may_oom);
131 current->memcg_oom.may_oom = 0;
134 static inline bool task_in_memcg_oom(struct task_struct *p)
136 return p->memcg_oom.memcg;
139 bool mem_cgroup_oom_synchronize(bool wait);
141 #ifdef CONFIG_MEMCG_SWAP
142 extern int do_swap_account;
145 static inline bool mem_cgroup_disabled(void)
147 if (memory_cgrp_subsys.disabled)
152 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
153 unsigned long *flags);
155 extern atomic_t memcg_moving;
157 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
158 bool *locked, unsigned long *flags)
160 if (mem_cgroup_disabled())
164 if (atomic_read(&memcg_moving))
165 __mem_cgroup_begin_update_page_stat(page, locked, flags);
168 void __mem_cgroup_end_update_page_stat(struct page *page,
169 unsigned long *flags);
170 static inline void mem_cgroup_end_update_page_stat(struct page *page,
171 bool *locked, unsigned long *flags)
173 if (mem_cgroup_disabled())
176 __mem_cgroup_end_update_page_stat(page, flags);
180 void mem_cgroup_update_page_stat(struct page *page,
181 enum mem_cgroup_stat_index idx,
184 static inline void mem_cgroup_inc_page_stat(struct page *page,
185 enum mem_cgroup_stat_index idx)
187 mem_cgroup_update_page_stat(page, idx, 1);
190 static inline void mem_cgroup_dec_page_stat(struct page *page,
191 enum mem_cgroup_stat_index idx)
193 mem_cgroup_update_page_stat(page, idx, -1);
196 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
198 unsigned long *total_scanned);
200 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
201 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
202 enum vm_event_item idx)
204 if (mem_cgroup_disabled())
206 __mem_cgroup_count_vm_event(mm, idx);
208 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
209 void mem_cgroup_split_huge_fixup(struct page *head);
212 #ifdef CONFIG_DEBUG_VM
213 bool mem_cgroup_bad_page_check(struct page *page);
214 void mem_cgroup_print_bad_page(struct page *page);
216 #else /* CONFIG_MEMCG */
219 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
221 struct mem_cgroup **memcgp)
227 static inline void mem_cgroup_commit_charge(struct page *page,
228 struct mem_cgroup *memcg,
233 static inline void mem_cgroup_cancel_charge(struct page *page,
234 struct mem_cgroup *memcg)
238 static inline void mem_cgroup_uncharge_start(void)
242 static inline void mem_cgroup_uncharge_end(void)
246 static inline void mem_cgroup_uncharge_page(struct page *page)
250 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
254 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
255 struct mem_cgroup *memcg)
257 return &zone->lruvec;
260 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
263 return &zone->lruvec;
266 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
271 static inline bool mm_match_cgroup(struct mm_struct *mm,
272 struct mem_cgroup *memcg)
277 static inline bool task_in_mem_cgroup(struct task_struct *task,
278 const struct mem_cgroup *memcg)
283 static inline struct cgroup_subsys_state
284 *mem_cgroup_css(struct mem_cgroup *memcg)
290 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
291 struct mem_cgroup **memcgp)
295 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
296 struct page *oldpage, struct page *newpage, bool migration_ok)
300 static inline struct mem_cgroup *
301 mem_cgroup_iter(struct mem_cgroup *root,
302 struct mem_cgroup *prev,
303 struct mem_cgroup_reclaim_cookie *reclaim)
308 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
309 struct mem_cgroup *prev)
313 static inline bool mem_cgroup_disabled(void)
319 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
324 static inline unsigned long
325 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
331 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
337 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
341 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
342 bool *locked, unsigned long *flags)
346 static inline void mem_cgroup_end_update_page_stat(struct page *page,
347 bool *locked, unsigned long *flags)
351 static inline void mem_cgroup_oom_enable(void)
355 static inline void mem_cgroup_oom_disable(void)
359 static inline bool task_in_memcg_oom(struct task_struct *p)
364 static inline bool mem_cgroup_oom_synchronize(bool wait)
369 static inline void mem_cgroup_inc_page_stat(struct page *page,
370 enum mem_cgroup_stat_index idx)
374 static inline void mem_cgroup_dec_page_stat(struct page *page,
375 enum mem_cgroup_stat_index idx)
380 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
382 unsigned long *total_scanned)
387 static inline void mem_cgroup_split_huge_fixup(struct page *head)
392 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
395 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
396 struct page *newpage)
399 #endif /* CONFIG_MEMCG */
401 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
403 mem_cgroup_bad_page_check(struct page *page)
409 mem_cgroup_print_bad_page(struct page *page)
421 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
422 void sock_update_memcg(struct sock *sk);
423 void sock_release_memcg(struct sock *sk);
425 static inline void sock_update_memcg(struct sock *sk)
428 static inline void sock_release_memcg(struct sock *sk)
431 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
433 #ifdef CONFIG_MEMCG_KMEM
434 extern struct static_key memcg_kmem_enabled_key;
436 extern int memcg_limited_groups_array_size;
439 * Helper macro to loop through all memcg-specific caches. Callers must still
440 * check if the cache is valid (it is either valid or NULL).
441 * the slab_mutex must be held when looping through those caches
443 #define for_each_memcg_cache_index(_idx) \
444 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
446 static inline bool memcg_kmem_enabled(void)
448 return static_key_false(&memcg_kmem_enabled_key);
452 * In general, we'll do everything in our power to not incur in any overhead
453 * for non-memcg users for the kmem functions. Not even a function call, if we
456 * Therefore, we'll inline all those functions so that in the best case, we'll
457 * see that kmemcg is off for everybody and proceed quickly. If it is on,
458 * we'll still do most of the flag checking inline. We check a lot of
459 * conditions, but because they are pretty simple, they are expected to be
462 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
464 void __memcg_kmem_commit_charge(struct page *page,
465 struct mem_cgroup *memcg, int order);
466 void __memcg_kmem_uncharge_pages(struct page *page, int order);
468 int memcg_cache_id(struct mem_cgroup *memcg);
470 int memcg_alloc_cache_params(struct mem_cgroup *memcg, struct kmem_cache *s,
471 struct kmem_cache *root_cache);
472 void memcg_free_cache_params(struct kmem_cache *s);
474 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
475 void memcg_update_array_size(int num_groups);
478 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
480 int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order);
481 void __memcg_uncharge_slab(struct kmem_cache *cachep, int order);
483 int __memcg_cleanup_cache_params(struct kmem_cache *s);
486 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
487 * @gfp: the gfp allocation flags.
488 * @memcg: a pointer to the memcg this was charged against.
489 * @order: allocation order.
491 * returns true if the memcg where the current task belongs can hold this
494 * We return true automatically if this allocation is not to be accounted to
498 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
500 if (!memcg_kmem_enabled())
504 * __GFP_NOFAIL allocations will move on even if charging is not
505 * possible. Therefore we don't even try, and have this allocation
506 * unaccounted. We could in theory charge it with
507 * res_counter_charge_nofail, but we hope those allocations are rare,
508 * and won't be worth the trouble.
510 if (gfp & __GFP_NOFAIL)
512 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
515 /* If the test is dying, just let it go. */
516 if (unlikely(fatal_signal_pending(current)))
519 return __memcg_kmem_newpage_charge(gfp, memcg, order);
523 * memcg_kmem_uncharge_pages: uncharge pages from memcg
524 * @page: pointer to struct page being freed
525 * @order: allocation order.
527 * there is no need to specify memcg here, since it is embedded in page_cgroup
530 memcg_kmem_uncharge_pages(struct page *page, int order)
532 if (memcg_kmem_enabled())
533 __memcg_kmem_uncharge_pages(page, order);
537 * memcg_kmem_commit_charge: embeds correct memcg in a page
538 * @page: pointer to struct page recently allocated
539 * @memcg: the memcg structure we charged against
540 * @order: allocation order.
542 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
543 * failure of the allocation. if @page is NULL, this function will revert the
544 * charges. Otherwise, it will commit the memcg given by @memcg to the
545 * corresponding page_cgroup.
548 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
550 if (memcg_kmem_enabled() && memcg)
551 __memcg_kmem_commit_charge(page, memcg, order);
555 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
556 * @cachep: the original global kmem cache
557 * @gfp: allocation flags.
559 * All memory allocated from a per-memcg cache is charged to the owner memcg.
561 static __always_inline struct kmem_cache *
562 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
564 if (!memcg_kmem_enabled())
566 if (gfp & __GFP_NOFAIL)
568 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
570 if (unlikely(fatal_signal_pending(current)))
573 return __memcg_kmem_get_cache(cachep, gfp);
576 #define for_each_memcg_cache_index(_idx) \
579 static inline bool memcg_kmem_enabled(void)
585 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
590 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
595 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
599 static inline int memcg_cache_id(struct mem_cgroup *memcg)
604 static inline int memcg_alloc_cache_params(struct mem_cgroup *memcg,
605 struct kmem_cache *s, struct kmem_cache *root_cache)
610 static inline void memcg_free_cache_params(struct kmem_cache *s)
614 static inline struct kmem_cache *
615 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
619 #endif /* CONFIG_MEMCG_KMEM */
620 #endif /* _LINUX_MEMCONTROL_H */