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 void mem_cgroup_uncharge(struct page *page);
65 /* Batched uncharging */
66 void mem_cgroup_uncharge_start(void);
67 void mem_cgroup_uncharge_end(void);
69 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage,
72 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
73 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
75 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
76 struct mem_cgroup *memcg);
77 bool task_in_mem_cgroup(struct task_struct *task,
78 const struct mem_cgroup *memcg);
80 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
81 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
83 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
84 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
87 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
89 struct mem_cgroup *task_memcg;
93 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
94 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
99 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
101 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
103 struct mem_cgroup_reclaim_cookie *);
104 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
107 * For memory reclaim.
109 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
110 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
111 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
112 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
113 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
114 struct task_struct *p);
116 static inline void mem_cgroup_oom_enable(void)
118 WARN_ON(current->memcg_oom.may_oom);
119 current->memcg_oom.may_oom = 1;
122 static inline void mem_cgroup_oom_disable(void)
124 WARN_ON(!current->memcg_oom.may_oom);
125 current->memcg_oom.may_oom = 0;
128 static inline bool task_in_memcg_oom(struct task_struct *p)
130 return p->memcg_oom.memcg;
133 bool mem_cgroup_oom_synchronize(bool wait);
135 #ifdef CONFIG_MEMCG_SWAP
136 extern int do_swap_account;
139 static inline bool mem_cgroup_disabled(void)
141 if (memory_cgrp_subsys.disabled)
146 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
147 unsigned long *flags);
149 extern atomic_t memcg_moving;
151 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
152 bool *locked, unsigned long *flags)
154 if (mem_cgroup_disabled())
158 if (atomic_read(&memcg_moving))
159 __mem_cgroup_begin_update_page_stat(page, locked, flags);
162 void __mem_cgroup_end_update_page_stat(struct page *page,
163 unsigned long *flags);
164 static inline void mem_cgroup_end_update_page_stat(struct page *page,
165 bool *locked, unsigned long *flags)
167 if (mem_cgroup_disabled())
170 __mem_cgroup_end_update_page_stat(page, flags);
174 void mem_cgroup_update_page_stat(struct page *page,
175 enum mem_cgroup_stat_index idx,
178 static inline void mem_cgroup_inc_page_stat(struct page *page,
179 enum mem_cgroup_stat_index idx)
181 mem_cgroup_update_page_stat(page, idx, 1);
184 static inline void mem_cgroup_dec_page_stat(struct page *page,
185 enum mem_cgroup_stat_index idx)
187 mem_cgroup_update_page_stat(page, idx, -1);
190 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
192 unsigned long *total_scanned);
194 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
195 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
196 enum vm_event_item idx)
198 if (mem_cgroup_disabled())
200 __mem_cgroup_count_vm_event(mm, idx);
202 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
203 void mem_cgroup_split_huge_fixup(struct page *head);
206 #ifdef CONFIG_DEBUG_VM
207 bool mem_cgroup_bad_page_check(struct page *page);
208 void mem_cgroup_print_bad_page(struct page *page);
210 #else /* CONFIG_MEMCG */
213 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
215 struct mem_cgroup **memcgp)
221 static inline void mem_cgroup_commit_charge(struct page *page,
222 struct mem_cgroup *memcg,
227 static inline void mem_cgroup_cancel_charge(struct page *page,
228 struct mem_cgroup *memcg)
232 static inline void mem_cgroup_uncharge(struct page *page)
236 static inline void mem_cgroup_uncharge_start(void)
240 static inline void mem_cgroup_uncharge_end(void)
244 static inline void mem_cgroup_migrate(struct page *oldpage,
245 struct page *newpage,
250 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
251 struct mem_cgroup *memcg)
253 return &zone->lruvec;
256 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
259 return &zone->lruvec;
262 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
267 static inline bool mm_match_cgroup(struct mm_struct *mm,
268 struct mem_cgroup *memcg)
273 static inline bool task_in_mem_cgroup(struct task_struct *task,
274 const struct mem_cgroup *memcg)
279 static inline struct cgroup_subsys_state
280 *mem_cgroup_css(struct mem_cgroup *memcg)
285 static inline struct mem_cgroup *
286 mem_cgroup_iter(struct mem_cgroup *root,
287 struct mem_cgroup *prev,
288 struct mem_cgroup_reclaim_cookie *reclaim)
293 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
294 struct mem_cgroup *prev)
298 static inline bool mem_cgroup_disabled(void)
304 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
309 static inline unsigned long
310 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
316 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
322 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
326 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
327 bool *locked, unsigned long *flags)
331 static inline void mem_cgroup_end_update_page_stat(struct page *page,
332 bool *locked, unsigned long *flags)
336 static inline void mem_cgroup_oom_enable(void)
340 static inline void mem_cgroup_oom_disable(void)
344 static inline bool task_in_memcg_oom(struct task_struct *p)
349 static inline bool mem_cgroup_oom_synchronize(bool wait)
354 static inline void mem_cgroup_inc_page_stat(struct page *page,
355 enum mem_cgroup_stat_index idx)
359 static inline void mem_cgroup_dec_page_stat(struct page *page,
360 enum mem_cgroup_stat_index idx)
365 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
367 unsigned long *total_scanned)
372 static inline void mem_cgroup_split_huge_fixup(struct page *head)
377 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
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);
451 int memcg_alloc_cache_params(struct mem_cgroup *memcg, struct kmem_cache *s,
452 struct kmem_cache *root_cache);
453 void memcg_free_cache_params(struct kmem_cache *s);
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 int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order);
462 void __memcg_uncharge_slab(struct kmem_cache *cachep, int order);
464 int __memcg_cleanup_cache_params(struct kmem_cache *s);
467 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
468 * @gfp: the gfp allocation flags.
469 * @memcg: a pointer to the memcg this was charged against.
470 * @order: allocation order.
472 * returns true if the memcg where the current task belongs can hold this
475 * We return true automatically if this allocation is not to be accounted to
479 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
481 if (!memcg_kmem_enabled())
485 * __GFP_NOFAIL allocations will move on even if charging is not
486 * possible. Therefore we don't even try, and have this allocation
487 * unaccounted. We could in theory charge it with
488 * res_counter_charge_nofail, but we hope those allocations are rare,
489 * and won't be worth the trouble.
491 if (gfp & __GFP_NOFAIL)
493 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
496 /* If the test is dying, just let it go. */
497 if (unlikely(fatal_signal_pending(current)))
500 return __memcg_kmem_newpage_charge(gfp, memcg, order);
504 * memcg_kmem_uncharge_pages: uncharge pages from memcg
505 * @page: pointer to struct page being freed
506 * @order: allocation order.
508 * there is no need to specify memcg here, since it is embedded in page_cgroup
511 memcg_kmem_uncharge_pages(struct page *page, int order)
513 if (memcg_kmem_enabled())
514 __memcg_kmem_uncharge_pages(page, order);
518 * memcg_kmem_commit_charge: embeds correct memcg in a page
519 * @page: pointer to struct page recently allocated
520 * @memcg: the memcg structure we charged against
521 * @order: allocation order.
523 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
524 * failure of the allocation. if @page is NULL, this function will revert the
525 * charges. Otherwise, it will commit the memcg given by @memcg to the
526 * corresponding page_cgroup.
529 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
531 if (memcg_kmem_enabled() && memcg)
532 __memcg_kmem_commit_charge(page, memcg, order);
536 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
537 * @cachep: the original global kmem cache
538 * @gfp: allocation flags.
540 * All memory allocated from a per-memcg cache is charged to the owner memcg.
542 static __always_inline struct kmem_cache *
543 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
545 if (!memcg_kmem_enabled())
547 if (gfp & __GFP_NOFAIL)
549 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
551 if (unlikely(fatal_signal_pending(current)))
554 return __memcg_kmem_get_cache(cachep, gfp);
557 #define for_each_memcg_cache_index(_idx) \
560 static inline bool memcg_kmem_enabled(void)
566 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
571 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
576 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
580 static inline int memcg_cache_id(struct mem_cgroup *memcg)
585 static inline int memcg_alloc_cache_params(struct mem_cgroup *memcg,
586 struct kmem_cache *s, struct kmem_cache *root_cache)
591 static inline void memcg_free_cache_params(struct kmem_cache *s)
595 static inline struct kmem_cache *
596 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
600 #endif /* CONFIG_MEMCG_KMEM */
601 #endif /* _LINUX_MEMCONTROL_H */