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_newpage_charge(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_cache_charge(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);
97 extern struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm);
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);
103 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
105 struct mem_cgroup *task_memcg;
109 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
110 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
115 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
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);
123 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
125 struct mem_cgroup_reclaim_cookie *);
126 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
129 * For memory reclaim.
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);
140 static inline void mem_cgroup_oom_enable(void)
142 WARN_ON(current->memcg_oom.may_oom);
143 current->memcg_oom.may_oom = 1;
146 static inline void mem_cgroup_oom_disable(void)
148 WARN_ON(!current->memcg_oom.may_oom);
149 current->memcg_oom.may_oom = 0;
152 static inline bool task_in_memcg_oom(struct task_struct *p)
154 return p->memcg_oom.memcg;
157 bool mem_cgroup_oom_synchronize(bool wait);
159 #ifdef CONFIG_MEMCG_SWAP
160 extern int do_swap_account;
163 static inline bool mem_cgroup_disabled(void)
165 if (mem_cgroup_subsys.disabled)
170 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
171 unsigned long *flags);
173 extern atomic_t memcg_moving;
175 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
176 bool *locked, unsigned long *flags)
178 if (mem_cgroup_disabled())
182 if (atomic_read(&memcg_moving))
183 __mem_cgroup_begin_update_page_stat(page, locked, flags);
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)
191 if (mem_cgroup_disabled())
194 __mem_cgroup_end_update_page_stat(page, flags);
198 void mem_cgroup_update_page_stat(struct page *page,
199 enum mem_cgroup_stat_index idx,
202 static inline void mem_cgroup_inc_page_stat(struct page *page,
203 enum mem_cgroup_stat_index idx)
205 mem_cgroup_update_page_stat(page, idx, 1);
208 static inline void mem_cgroup_dec_page_stat(struct page *page,
209 enum mem_cgroup_stat_index idx)
211 mem_cgroup_update_page_stat(page, idx, -1);
214 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
216 unsigned long *total_scanned);
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)
222 if (mem_cgroup_disabled())
224 __mem_cgroup_count_vm_event(mm, idx);
226 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
227 void mem_cgroup_split_huge_fixup(struct page *head);
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);
234 #else /* CONFIG_MEMCG */
237 static inline int mem_cgroup_newpage_charge(struct page *page,
238 struct mm_struct *mm, gfp_t gfp_mask)
243 static inline int mem_cgroup_cache_charge(struct page *page,
244 struct mm_struct *mm, gfp_t gfp_mask)
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)
255 static inline void mem_cgroup_commit_charge_swapin(struct page *page,
256 struct mem_cgroup *memcg)
260 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
264 static inline void mem_cgroup_uncharge_start(void)
268 static inline void mem_cgroup_uncharge_end(void)
272 static inline void mem_cgroup_uncharge_page(struct page *page)
276 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
280 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
281 struct mem_cgroup *memcg)
283 return &zone->lruvec;
286 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
289 return &zone->lruvec;
292 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
297 static inline struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
302 static inline bool mm_match_cgroup(struct mm_struct *mm,
303 struct mem_cgroup *memcg)
308 static inline bool task_in_mem_cgroup(struct task_struct *task,
309 const struct mem_cgroup *memcg)
314 static inline struct cgroup_subsys_state
315 *mem_cgroup_css(struct mem_cgroup *memcg)
321 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
322 struct mem_cgroup **memcgp)
326 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
327 struct page *oldpage, struct page *newpage, bool migration_ok)
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)
339 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
340 struct mem_cgroup *prev)
344 static inline bool mem_cgroup_disabled(void)
350 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
355 static inline unsigned long
356 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
362 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
368 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
372 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
373 bool *locked, unsigned long *flags)
377 static inline void mem_cgroup_end_update_page_stat(struct page *page,
378 bool *locked, unsigned long *flags)
382 static inline void mem_cgroup_oom_enable(void)
386 static inline void mem_cgroup_oom_disable(void)
390 static inline bool task_in_memcg_oom(struct task_struct *p)
395 static inline bool mem_cgroup_oom_synchronize(bool wait)
400 static inline void mem_cgroup_inc_page_stat(struct page *page,
401 enum mem_cgroup_stat_index idx)
405 static inline void mem_cgroup_dec_page_stat(struct page *page,
406 enum mem_cgroup_stat_index idx)
411 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
413 unsigned long *total_scanned)
418 static inline void mem_cgroup_split_huge_fixup(struct page *head)
423 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
426 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
427 struct page *newpage)
430 #endif /* CONFIG_MEMCG */
432 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
434 mem_cgroup_bad_page_check(struct page *page)
440 mem_cgroup_print_bad_page(struct page *page)
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);
456 static inline void sock_update_memcg(struct sock *sk)
459 static inline void sock_release_memcg(struct sock *sk)
462 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
464 #ifdef CONFIG_MEMCG_KMEM
465 extern struct static_key memcg_kmem_enabled_key;
467 extern int memcg_limited_groups_array_size;
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
474 #define for_each_memcg_cache_index(_idx) \
475 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
477 static inline bool memcg_kmem_enabled(void)
479 return static_key_false(&memcg_kmem_enabled_key);
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
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
493 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
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);
499 int memcg_cache_id(struct mem_cgroup *memcg);
500 int memcg_alloc_cache_params(struct mem_cgroup *memcg, struct kmem_cache *s,
501 struct kmem_cache *root_cache);
502 void memcg_free_cache_params(struct kmem_cache *s);
503 void memcg_register_cache(struct kmem_cache *s);
504 void memcg_unregister_cache(struct kmem_cache *s);
506 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
507 void memcg_update_array_size(int num_groups);
510 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
512 void mem_cgroup_destroy_cache(struct kmem_cache *cachep);
513 void kmem_cache_destroy_memcg_children(struct kmem_cache *s);
516 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
517 * @gfp: the gfp allocation flags.
518 * @memcg: a pointer to the memcg this was charged against.
519 * @order: allocation order.
521 * returns true if the memcg where the current task belongs can hold this
524 * We return true automatically if this allocation is not to be accounted to
528 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
530 if (!memcg_kmem_enabled())
534 * __GFP_NOFAIL allocations will move on even if charging is not
535 * possible. Therefore we don't even try, and have this allocation
536 * unaccounted. We could in theory charge it with
537 * res_counter_charge_nofail, but we hope those allocations are rare,
538 * and won't be worth the trouble.
540 if (!(gfp & __GFP_KMEMCG) || (gfp & __GFP_NOFAIL))
542 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
545 /* If the test is dying, just let it go. */
546 if (unlikely(fatal_signal_pending(current)))
549 return __memcg_kmem_newpage_charge(gfp, memcg, order);
553 * memcg_kmem_uncharge_pages: uncharge pages from memcg
554 * @page: pointer to struct page being freed
555 * @order: allocation order.
557 * there is no need to specify memcg here, since it is embedded in page_cgroup
560 memcg_kmem_uncharge_pages(struct page *page, int order)
562 if (memcg_kmem_enabled())
563 __memcg_kmem_uncharge_pages(page, order);
567 * memcg_kmem_commit_charge: embeds correct memcg in a page
568 * @page: pointer to struct page recently allocated
569 * @memcg: the memcg structure we charged against
570 * @order: allocation order.
572 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
573 * failure of the allocation. if @page is NULL, this function will revert the
574 * charges. Otherwise, it will commit the memcg given by @memcg to the
575 * corresponding page_cgroup.
578 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
580 if (memcg_kmem_enabled() && memcg)
581 __memcg_kmem_commit_charge(page, memcg, order);
585 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
586 * @cachep: the original global kmem cache
587 * @gfp: allocation flags.
589 * This function assumes that the task allocating, which determines the memcg
590 * in the page allocator, belongs to the same cgroup throughout the whole
591 * process. Misacounting can happen if the task calls memcg_kmem_get_cache()
592 * while belonging to a cgroup, and later on changes. This is considered
593 * acceptable, and should only happen upon task migration.
595 * Before the cache is created by the memcg core, there is also a possible
596 * imbalance: the task belongs to a memcg, but the cache being allocated from
597 * is the global cache, since the child cache is not yet guaranteed to be
598 * ready. This case is also fine, since in this case the GFP_KMEMCG will not be
599 * passed and the page allocator will not attempt any cgroup accounting.
601 static __always_inline struct kmem_cache *
602 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
604 if (!memcg_kmem_enabled())
606 if (gfp & __GFP_NOFAIL)
608 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
610 if (unlikely(fatal_signal_pending(current)))
613 return __memcg_kmem_get_cache(cachep, gfp);
616 #define for_each_memcg_cache_index(_idx) \
619 static inline bool memcg_kmem_enabled(void)
625 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
630 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
635 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
639 static inline int memcg_cache_id(struct mem_cgroup *memcg)
644 static inline int memcg_alloc_cache_params(struct mem_cgroup *memcg,
645 struct kmem_cache *s, struct kmem_cache *root_cache)
650 static inline void memcg_free_cache_params(struct kmem_cache *s)
654 static inline void memcg_register_cache(struct kmem_cache *s)
658 static inline void memcg_unregister_cache(struct kmem_cache *s)
662 static inline struct kmem_cache *
663 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
668 static inline void kmem_cache_destroy_memcg_children(struct kmem_cache *s)
671 #endif /* CONFIG_MEMCG_KMEM */
672 #endif /* _LINUX_MEMCONTROL_H */