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;
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
65 * (Of course, if memcg does memory allocation in future, GFP_KERNEL is sane.)
68 extern int mem_cgroup_charge_anon(struct page *page, struct mm_struct *mm,
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);
77 extern int mem_cgroup_charge_file(struct page *page, struct mm_struct *mm,
80 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
81 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
83 /* For coalescing uncharge for reducing memcg' overhead*/
84 extern void mem_cgroup_uncharge_start(void);
85 extern void mem_cgroup_uncharge_end(void);
87 extern void mem_cgroup_uncharge_page(struct page *page);
88 extern void mem_cgroup_uncharge_cache_page(struct page *page);
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);
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);
98 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
99 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
102 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
104 struct mem_cgroup *task_memcg;
108 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
109 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
114 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
117 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
118 struct mem_cgroup **memcgp);
119 extern void mem_cgroup_end_migration(struct mem_cgroup *memcg,
120 struct page *oldpage, struct page *newpage, bool migration_ok);
122 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
124 struct mem_cgroup_reclaim_cookie *);
125 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
128 * For memory reclaim.
130 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
131 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
132 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
133 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
134 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
135 struct task_struct *p);
136 extern void mem_cgroup_replace_page_cache(struct page *oldpage,
137 struct page *newpage);
139 static inline void mem_cgroup_oom_enable(void)
141 WARN_ON(current->memcg_oom.may_oom);
142 current->memcg_oom.may_oom = 1;
145 static inline void mem_cgroup_oom_disable(void)
147 WARN_ON(!current->memcg_oom.may_oom);
148 current->memcg_oom.may_oom = 0;
151 static inline bool task_in_memcg_oom(struct task_struct *p)
153 return p->memcg_oom.memcg;
156 bool mem_cgroup_oom_synchronize(bool wait);
158 #ifdef CONFIG_MEMCG_SWAP
159 extern int do_swap_account;
162 static inline bool mem_cgroup_disabled(void)
164 if (memory_cgrp_subsys.disabled)
169 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
170 unsigned long *flags);
172 extern atomic_t memcg_moving;
174 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
175 bool *locked, unsigned long *flags)
177 if (mem_cgroup_disabled())
181 if (atomic_read(&memcg_moving))
182 __mem_cgroup_begin_update_page_stat(page, locked, flags);
185 void __mem_cgroup_end_update_page_stat(struct page *page,
186 unsigned long *flags);
187 static inline void mem_cgroup_end_update_page_stat(struct page *page,
188 bool *locked, unsigned long *flags)
190 if (mem_cgroup_disabled())
193 __mem_cgroup_end_update_page_stat(page, flags);
197 void mem_cgroup_update_page_stat(struct page *page,
198 enum mem_cgroup_stat_index idx,
201 static inline void mem_cgroup_inc_page_stat(struct page *page,
202 enum mem_cgroup_stat_index idx)
204 mem_cgroup_update_page_stat(page, idx, 1);
207 static inline void mem_cgroup_dec_page_stat(struct page *page,
208 enum mem_cgroup_stat_index idx)
210 mem_cgroup_update_page_stat(page, idx, -1);
213 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
215 unsigned long *total_scanned);
217 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
218 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
219 enum vm_event_item idx)
221 if (mem_cgroup_disabled())
223 __mem_cgroup_count_vm_event(mm, idx);
225 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
226 void mem_cgroup_split_huge_fixup(struct page *head);
229 #ifdef CONFIG_DEBUG_VM
230 bool mem_cgroup_bad_page_check(struct page *page);
231 void mem_cgroup_print_bad_page(struct page *page);
233 #else /* CONFIG_MEMCG */
236 static inline int mem_cgroup_charge_anon(struct page *page,
237 struct mm_struct *mm, gfp_t gfp_mask)
242 static inline int mem_cgroup_charge_file(struct page *page,
243 struct mm_struct *mm, gfp_t gfp_mask)
248 static inline int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
249 struct page *page, gfp_t gfp_mask, struct mem_cgroup **memcgp)
254 static inline void mem_cgroup_commit_charge_swapin(struct page *page,
255 struct mem_cgroup *memcg)
259 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
263 static inline void mem_cgroup_uncharge_start(void)
267 static inline void mem_cgroup_uncharge_end(void)
271 static inline void mem_cgroup_uncharge_page(struct page *page)
275 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
279 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
280 struct mem_cgroup *memcg)
282 return &zone->lruvec;
285 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
288 return &zone->lruvec;
291 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
296 static inline bool mm_match_cgroup(struct mm_struct *mm,
297 struct mem_cgroup *memcg)
302 static inline bool task_in_mem_cgroup(struct task_struct *task,
303 const struct mem_cgroup *memcg)
308 static inline struct cgroup_subsys_state
309 *mem_cgroup_css(struct mem_cgroup *memcg)
315 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
316 struct mem_cgroup **memcgp)
320 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
321 struct page *oldpage, struct page *newpage, bool migration_ok)
325 static inline struct mem_cgroup *
326 mem_cgroup_iter(struct mem_cgroup *root,
327 struct mem_cgroup *prev,
328 struct mem_cgroup_reclaim_cookie *reclaim)
333 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
334 struct mem_cgroup *prev)
338 static inline bool mem_cgroup_disabled(void)
344 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
349 static inline unsigned long
350 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
356 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
362 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
366 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
367 bool *locked, unsigned long *flags)
371 static inline void mem_cgroup_end_update_page_stat(struct page *page,
372 bool *locked, unsigned long *flags)
376 static inline void mem_cgroup_oom_enable(void)
380 static inline void mem_cgroup_oom_disable(void)
384 static inline bool task_in_memcg_oom(struct task_struct *p)
389 static inline bool mem_cgroup_oom_synchronize(bool wait)
394 static inline void mem_cgroup_inc_page_stat(struct page *page,
395 enum mem_cgroup_stat_index idx)
399 static inline void mem_cgroup_dec_page_stat(struct page *page,
400 enum mem_cgroup_stat_index idx)
405 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
407 unsigned long *total_scanned)
412 static inline void mem_cgroup_split_huge_fixup(struct page *head)
417 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
420 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
421 struct page *newpage)
424 #endif /* CONFIG_MEMCG */
426 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
428 mem_cgroup_bad_page_check(struct page *page)
434 mem_cgroup_print_bad_page(struct page *page)
446 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
447 void sock_update_memcg(struct sock *sk);
448 void sock_release_memcg(struct sock *sk);
450 static inline void sock_update_memcg(struct sock *sk)
453 static inline void sock_release_memcg(struct sock *sk)
456 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
458 #ifdef CONFIG_MEMCG_KMEM
459 extern struct static_key memcg_kmem_enabled_key;
461 extern int memcg_limited_groups_array_size;
464 * Helper macro to loop through all memcg-specific caches. Callers must still
465 * check if the cache is valid (it is either valid or NULL).
466 * the slab_mutex must be held when looping through those caches
468 #define for_each_memcg_cache_index(_idx) \
469 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
471 static inline bool memcg_kmem_enabled(void)
473 return static_key_false(&memcg_kmem_enabled_key);
477 * In general, we'll do everything in our power to not incur in any overhead
478 * for non-memcg users for the kmem functions. Not even a function call, if we
481 * Therefore, we'll inline all those functions so that in the best case, we'll
482 * see that kmemcg is off for everybody and proceed quickly. If it is on,
483 * we'll still do most of the flag checking inline. We check a lot of
484 * conditions, but because they are pretty simple, they are expected to be
487 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
489 void __memcg_kmem_commit_charge(struct page *page,
490 struct mem_cgroup *memcg, int order);
491 void __memcg_kmem_uncharge_pages(struct page *page, int order);
493 int memcg_cache_id(struct mem_cgroup *memcg);
495 char *memcg_create_cache_name(struct mem_cgroup *memcg,
496 struct kmem_cache *root_cache);
497 int memcg_alloc_cache_params(struct mem_cgroup *memcg, struct kmem_cache *s,
498 struct kmem_cache *root_cache);
499 void memcg_free_cache_params(struct kmem_cache *s);
500 void memcg_register_cache(struct kmem_cache *s);
501 void memcg_unregister_cache(struct kmem_cache *s);
503 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
504 void memcg_update_array_size(int num_groups);
507 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
509 void mem_cgroup_destroy_cache(struct kmem_cache *cachep);
510 int __kmem_cache_destroy_memcg_children(struct kmem_cache *s);
513 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
514 * @gfp: the gfp allocation flags.
515 * @memcg: a pointer to the memcg this was charged against.
516 * @order: allocation order.
518 * returns true if the memcg where the current task belongs can hold this
521 * We return true automatically if this allocation is not to be accounted to
525 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
527 if (!memcg_kmem_enabled())
531 * __GFP_NOFAIL allocations will move on even if charging is not
532 * possible. Therefore we don't even try, and have this allocation
533 * unaccounted. We could in theory charge it with
534 * res_counter_charge_nofail, but we hope those allocations are rare,
535 * and won't be worth the trouble.
537 if (!(gfp & __GFP_KMEMCG) || (gfp & __GFP_NOFAIL))
539 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
542 /* If the test is dying, just let it go. */
543 if (unlikely(fatal_signal_pending(current)))
546 return __memcg_kmem_newpage_charge(gfp, memcg, order);
550 * memcg_kmem_uncharge_pages: uncharge pages from memcg
551 * @page: pointer to struct page being freed
552 * @order: allocation order.
554 * there is no need to specify memcg here, since it is embedded in page_cgroup
557 memcg_kmem_uncharge_pages(struct page *page, int order)
559 if (memcg_kmem_enabled())
560 __memcg_kmem_uncharge_pages(page, order);
564 * memcg_kmem_commit_charge: embeds correct memcg in a page
565 * @page: pointer to struct page recently allocated
566 * @memcg: the memcg structure we charged against
567 * @order: allocation order.
569 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
570 * failure of the allocation. if @page is NULL, this function will revert the
571 * charges. Otherwise, it will commit the memcg given by @memcg to the
572 * corresponding page_cgroup.
575 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
577 if (memcg_kmem_enabled() && memcg)
578 __memcg_kmem_commit_charge(page, memcg, order);
582 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
583 * @cachep: the original global kmem cache
584 * @gfp: allocation flags.
586 * This function assumes that the task allocating, which determines the memcg
587 * in the page allocator, belongs to the same cgroup throughout the whole
588 * process. Misacounting can happen if the task calls memcg_kmem_get_cache()
589 * while belonging to a cgroup, and later on changes. This is considered
590 * acceptable, and should only happen upon task migration.
592 * Before the cache is created by the memcg core, there is also a possible
593 * imbalance: the task belongs to a memcg, but the cache being allocated from
594 * is the global cache, since the child cache is not yet guaranteed to be
595 * ready. This case is also fine, since in this case the GFP_KMEMCG will not be
596 * passed and the page allocator will not attempt any cgroup accounting.
598 static __always_inline struct kmem_cache *
599 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
601 if (!memcg_kmem_enabled())
603 if (gfp & __GFP_NOFAIL)
605 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
607 if (unlikely(fatal_signal_pending(current)))
610 return __memcg_kmem_get_cache(cachep, gfp);
613 #define for_each_memcg_cache_index(_idx) \
616 static inline bool memcg_kmem_enabled(void)
622 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
627 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
632 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
636 static inline int memcg_cache_id(struct mem_cgroup *memcg)
641 static inline int memcg_alloc_cache_params(struct mem_cgroup *memcg,
642 struct kmem_cache *s, struct kmem_cache *root_cache)
647 static inline void memcg_free_cache_params(struct kmem_cache *s)
651 static inline void memcg_register_cache(struct kmem_cache *s)
655 static inline void memcg_unregister_cache(struct kmem_cache *s)
659 static inline struct kmem_cache *
660 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
664 #endif /* CONFIG_MEMCG_KMEM */
665 #endif /* _LINUX_MEMCONTROL_H */