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>
33 /* Stats that can be updated by kernel. */
34 enum mem_cgroup_page_stat_item {
35 MEMCG_NR_FILE_MAPPED, /* # of pages charged as file rss */
38 struct mem_cgroup_reclaim_cookie {
41 unsigned int generation;
46 * All "charge" functions with gfp_mask should use GFP_KERNEL or
47 * (gfp_mask & GFP_RECLAIM_MASK). In current implementatin, memcg doesn't
48 * alloc memory but reclaims memory from all available zones. So, "where I want
49 * memory from" bits of gfp_mask has no meaning. So any bits of that field is
50 * available but adding a rule is better. charge functions' gfp_mask should
51 * be set to GFP_KERNEL or gfp_mask & GFP_RECLAIM_MASK for avoiding ambiguous
53 * (Of course, if memcg does memory allocation in future, GFP_KERNEL is sane.)
56 extern int mem_cgroup_newpage_charge(struct page *page, struct mm_struct *mm,
58 /* for swap handling */
59 extern int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
60 struct page *page, gfp_t mask, struct mem_cgroup **memcgp);
61 extern void mem_cgroup_commit_charge_swapin(struct page *page,
62 struct mem_cgroup *memcg);
63 extern void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg);
65 extern int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
68 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
69 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
71 /* For coalescing uncharge for reducing memcg' overhead*/
72 extern void mem_cgroup_uncharge_start(void);
73 extern void mem_cgroup_uncharge_end(void);
75 extern void mem_cgroup_uncharge_page(struct page *page);
76 extern void mem_cgroup_uncharge_cache_page(struct page *page);
78 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
79 struct mem_cgroup *memcg);
80 int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg);
82 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
83 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
84 extern struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm);
86 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
87 extern struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont);
90 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
92 struct mem_cgroup *task_memcg;
96 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
97 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
102 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
105 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
106 struct mem_cgroup **memcgp);
107 extern void mem_cgroup_end_migration(struct mem_cgroup *memcg,
108 struct page *oldpage, struct page *newpage, bool migration_ok);
110 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
112 struct mem_cgroup_reclaim_cookie *);
113 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
116 * For memory reclaim.
118 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
119 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
120 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
121 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
122 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
123 struct task_struct *p);
124 extern void mem_cgroup_replace_page_cache(struct page *oldpage,
125 struct page *newpage);
127 #ifdef CONFIG_MEMCG_SWAP
128 extern int do_swap_account;
131 static inline bool mem_cgroup_disabled(void)
133 if (mem_cgroup_subsys.disabled)
138 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
139 unsigned long *flags);
141 extern atomic_t memcg_moving;
143 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
144 bool *locked, unsigned long *flags)
146 if (mem_cgroup_disabled())
150 if (atomic_read(&memcg_moving))
151 __mem_cgroup_begin_update_page_stat(page, locked, flags);
154 void __mem_cgroup_end_update_page_stat(struct page *page,
155 unsigned long *flags);
156 static inline void mem_cgroup_end_update_page_stat(struct page *page,
157 bool *locked, unsigned long *flags)
159 if (mem_cgroup_disabled())
162 __mem_cgroup_end_update_page_stat(page, flags);
166 void mem_cgroup_update_page_stat(struct page *page,
167 enum mem_cgroup_page_stat_item idx,
170 static inline void mem_cgroup_inc_page_stat(struct page *page,
171 enum mem_cgroup_page_stat_item idx)
173 mem_cgroup_update_page_stat(page, idx, 1);
176 static inline void mem_cgroup_dec_page_stat(struct page *page,
177 enum mem_cgroup_page_stat_item idx)
179 mem_cgroup_update_page_stat(page, idx, -1);
182 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
184 unsigned long *total_scanned);
186 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
187 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
188 enum vm_event_item idx)
190 if (mem_cgroup_disabled())
192 __mem_cgroup_count_vm_event(mm, idx);
194 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
195 void mem_cgroup_split_huge_fixup(struct page *head);
198 #ifdef CONFIG_DEBUG_VM
199 bool mem_cgroup_bad_page_check(struct page *page);
200 void mem_cgroup_print_bad_page(struct page *page);
202 #else /* CONFIG_MEMCG */
205 static inline int mem_cgroup_newpage_charge(struct page *page,
206 struct mm_struct *mm, gfp_t gfp_mask)
211 static inline int mem_cgroup_cache_charge(struct page *page,
212 struct mm_struct *mm, gfp_t gfp_mask)
217 static inline int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
218 struct page *page, gfp_t gfp_mask, struct mem_cgroup **memcgp)
223 static inline void mem_cgroup_commit_charge_swapin(struct page *page,
224 struct mem_cgroup *memcg)
228 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
232 static inline void mem_cgroup_uncharge_start(void)
236 static inline void mem_cgroup_uncharge_end(void)
240 static inline void mem_cgroup_uncharge_page(struct page *page)
244 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
248 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
249 struct mem_cgroup *memcg)
251 return &zone->lruvec;
254 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
257 return &zone->lruvec;
260 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
265 static inline struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
270 static inline bool mm_match_cgroup(struct mm_struct *mm,
271 struct mem_cgroup *memcg)
276 static inline int task_in_mem_cgroup(struct task_struct *task,
277 const struct mem_cgroup *memcg)
282 static inline struct cgroup_subsys_state
283 *mem_cgroup_css(struct mem_cgroup *memcg)
289 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
290 struct mem_cgroup **memcgp)
294 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
295 struct page *oldpage, struct page *newpage, bool migration_ok)
299 static inline struct mem_cgroup *
300 mem_cgroup_iter(struct mem_cgroup *root,
301 struct mem_cgroup *prev,
302 struct mem_cgroup_reclaim_cookie *reclaim)
307 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
308 struct mem_cgroup *prev)
312 static inline bool mem_cgroup_disabled(void)
318 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
323 static inline unsigned long
324 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
330 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
336 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
340 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
341 bool *locked, unsigned long *flags)
345 static inline void mem_cgroup_end_update_page_stat(struct page *page,
346 bool *locked, unsigned long *flags)
350 static inline void mem_cgroup_inc_page_stat(struct page *page,
351 enum mem_cgroup_page_stat_item idx)
355 static inline void mem_cgroup_dec_page_stat(struct page *page,
356 enum mem_cgroup_page_stat_item idx)
361 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
363 unsigned long *total_scanned)
368 static inline void mem_cgroup_split_huge_fixup(struct page *head)
373 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
376 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
377 struct page *newpage)
380 #endif /* CONFIG_MEMCG */
382 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
384 mem_cgroup_bad_page_check(struct page *page)
390 mem_cgroup_print_bad_page(struct page *page)
402 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
403 void sock_update_memcg(struct sock *sk);
404 void sock_release_memcg(struct sock *sk);
406 static inline void sock_update_memcg(struct sock *sk)
409 static inline void sock_release_memcg(struct sock *sk)
412 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
414 #ifdef CONFIG_MEMCG_KMEM
415 extern struct static_key memcg_kmem_enabled_key;
417 extern int memcg_limited_groups_array_size;
420 * Helper macro to loop through all memcg-specific caches. Callers must still
421 * check if the cache is valid (it is either valid or NULL).
422 * the slab_mutex must be held when looping through those caches
424 #define for_each_memcg_cache_index(_idx) \
425 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
427 static inline bool memcg_kmem_enabled(void)
429 return static_key_false(&memcg_kmem_enabled_key);
433 * In general, we'll do everything in our power to not incur in any overhead
434 * for non-memcg users for the kmem functions. Not even a function call, if we
437 * Therefore, we'll inline all those functions so that in the best case, we'll
438 * see that kmemcg is off for everybody and proceed quickly. If it is on,
439 * we'll still do most of the flag checking inline. We check a lot of
440 * conditions, but because they are pretty simple, they are expected to be
443 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
445 void __memcg_kmem_commit_charge(struct page *page,
446 struct mem_cgroup *memcg, int order);
447 void __memcg_kmem_uncharge_pages(struct page *page, int order);
449 int memcg_cache_id(struct mem_cgroup *memcg);
450 int memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
451 struct kmem_cache *root_cache);
452 void memcg_release_cache(struct kmem_cache *cachep);
453 void memcg_cache_list_add(struct mem_cgroup *memcg, struct kmem_cache *cachep);
455 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
456 void memcg_update_array_size(int num_groups);
459 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
461 void mem_cgroup_destroy_cache(struct kmem_cache *cachep);
462 void kmem_cache_destroy_memcg_children(struct kmem_cache *s);
465 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
466 * @gfp: the gfp allocation flags.
467 * @memcg: a pointer to the memcg this was charged against.
468 * @order: allocation order.
470 * returns true if the memcg where the current task belongs can hold this
473 * We return true automatically if this allocation is not to be accounted to
477 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
479 if (!memcg_kmem_enabled())
483 * __GFP_NOFAIL allocations will move on even if charging is not
484 * possible. Therefore we don't even try, and have this allocation
485 * unaccounted. We could in theory charge it with
486 * res_counter_charge_nofail, but we hope those allocations are rare,
487 * and won't be worth the trouble.
489 if (!(gfp & __GFP_KMEMCG) || (gfp & __GFP_NOFAIL))
491 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
494 /* If the test is dying, just let it go. */
495 if (unlikely(fatal_signal_pending(current)))
498 return __memcg_kmem_newpage_charge(gfp, memcg, order);
502 * memcg_kmem_uncharge_pages: uncharge pages from memcg
503 * @page: pointer to struct page being freed
504 * @order: allocation order.
506 * there is no need to specify memcg here, since it is embedded in page_cgroup
509 memcg_kmem_uncharge_pages(struct page *page, int order)
511 if (memcg_kmem_enabled())
512 __memcg_kmem_uncharge_pages(page, order);
516 * memcg_kmem_commit_charge: embeds correct memcg in a page
517 * @page: pointer to struct page recently allocated
518 * @memcg: the memcg structure we charged against
519 * @order: allocation order.
521 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
522 * failure of the allocation. if @page is NULL, this function will revert the
523 * charges. Otherwise, it will commit the memcg given by @memcg to the
524 * corresponding page_cgroup.
527 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
529 if (memcg_kmem_enabled() && memcg)
530 __memcg_kmem_commit_charge(page, memcg, order);
534 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
535 * @cachep: the original global kmem cache
536 * @gfp: allocation flags.
538 * This function assumes that the task allocating, which determines the memcg
539 * in the page allocator, belongs to the same cgroup throughout the whole
540 * process. Misacounting can happen if the task calls memcg_kmem_get_cache()
541 * while belonging to a cgroup, and later on changes. This is considered
542 * acceptable, and should only happen upon task migration.
544 * Before the cache is created by the memcg core, there is also a possible
545 * imbalance: the task belongs to a memcg, but the cache being allocated from
546 * is the global cache, since the child cache is not yet guaranteed to be
547 * ready. This case is also fine, since in this case the GFP_KMEMCG will not be
548 * passed and the page allocator will not attempt any cgroup accounting.
550 static __always_inline struct kmem_cache *
551 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
553 if (!memcg_kmem_enabled())
555 if (gfp & __GFP_NOFAIL)
557 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
559 if (unlikely(fatal_signal_pending(current)))
562 return __memcg_kmem_get_cache(cachep, gfp);
565 #define for_each_memcg_cache_index(_idx) \
568 static inline bool memcg_kmem_enabled(void)
574 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
579 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
584 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
588 static inline int memcg_cache_id(struct mem_cgroup *memcg)
594 memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
595 struct kmem_cache *root_cache)
600 static inline void memcg_release_cache(struct kmem_cache *cachep)
604 static inline void memcg_cache_list_add(struct mem_cgroup *memcg,
605 struct kmem_cache *s)
609 static inline struct kmem_cache *
610 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
615 static inline void kmem_cache_destroy_memcg_children(struct kmem_cache *s)
618 #endif /* CONFIG_MEMCG_KMEM */
619 #endif /* _LINUX_MEMCONTROL_H */