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 * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
34 * These two lists should keep in accord with each other.
36 enum mem_cgroup_stat_index {
38 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
40 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
41 MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */
42 MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */
43 MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */
44 MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
45 MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
46 MEM_CGROUP_STAT_NSTATS,
49 struct mem_cgroup_reclaim_cookie {
52 unsigned int generation;
56 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
57 gfp_t gfp_mask, struct mem_cgroup **memcgp);
58 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
60 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
61 void mem_cgroup_uncharge(struct page *page);
62 void mem_cgroup_uncharge_list(struct list_head *page_list);
64 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage,
67 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
68 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
70 bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
71 struct mem_cgroup *root);
72 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
74 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
75 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
77 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
78 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
80 static inline bool mm_match_cgroup(struct mm_struct *mm,
81 struct mem_cgroup *memcg)
83 struct mem_cgroup *task_memcg;
87 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
89 match = mem_cgroup_is_descendant(task_memcg, memcg);
94 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
96 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
98 struct mem_cgroup_reclaim_cookie *);
99 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
102 * For memory reclaim.
104 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
105 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
106 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
107 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
108 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
109 struct task_struct *p);
111 static inline void mem_cgroup_oom_enable(void)
113 WARN_ON(current->memcg_oom.may_oom);
114 current->memcg_oom.may_oom = 1;
117 static inline void mem_cgroup_oom_disable(void)
119 WARN_ON(!current->memcg_oom.may_oom);
120 current->memcg_oom.may_oom = 0;
123 static inline bool task_in_memcg_oom(struct task_struct *p)
125 return p->memcg_oom.memcg;
128 bool mem_cgroup_oom_synchronize(bool wait);
130 #ifdef CONFIG_MEMCG_SWAP
131 extern int do_swap_account;
134 static inline bool mem_cgroup_disabled(void)
136 if (memory_cgrp_subsys.disabled)
141 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page, bool *locked,
142 unsigned long *flags);
143 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool *locked,
144 unsigned long *flags);
145 void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
146 enum mem_cgroup_stat_index idx, int val);
148 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
149 enum mem_cgroup_stat_index idx)
151 mem_cgroup_update_page_stat(memcg, idx, 1);
154 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
155 enum mem_cgroup_stat_index idx)
157 mem_cgroup_update_page_stat(memcg, idx, -1);
160 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
162 unsigned long *total_scanned);
164 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
165 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
166 enum vm_event_item idx)
168 if (mem_cgroup_disabled())
170 __mem_cgroup_count_vm_event(mm, idx);
172 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
173 void mem_cgroup_split_huge_fixup(struct page *head);
176 #else /* CONFIG_MEMCG */
179 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
181 struct mem_cgroup **memcgp)
187 static inline void mem_cgroup_commit_charge(struct page *page,
188 struct mem_cgroup *memcg,
193 static inline void mem_cgroup_cancel_charge(struct page *page,
194 struct mem_cgroup *memcg)
198 static inline void mem_cgroup_uncharge(struct page *page)
202 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
206 static inline void mem_cgroup_migrate(struct page *oldpage,
207 struct page *newpage,
212 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
213 struct mem_cgroup *memcg)
215 return &zone->lruvec;
218 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
221 return &zone->lruvec;
224 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
229 static inline bool mm_match_cgroup(struct mm_struct *mm,
230 struct mem_cgroup *memcg)
235 static inline bool task_in_mem_cgroup(struct task_struct *task,
236 const struct mem_cgroup *memcg)
241 static inline struct cgroup_subsys_state
242 *mem_cgroup_css(struct mem_cgroup *memcg)
247 static inline struct mem_cgroup *
248 mem_cgroup_iter(struct mem_cgroup *root,
249 struct mem_cgroup *prev,
250 struct mem_cgroup_reclaim_cookie *reclaim)
255 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
256 struct mem_cgroup *prev)
260 static inline bool mem_cgroup_disabled(void)
266 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
271 static inline unsigned long
272 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
278 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
284 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
288 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page,
289 bool *locked, unsigned long *flags)
294 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg,
295 bool *locked, unsigned long *flags)
299 static inline void mem_cgroup_oom_enable(void)
303 static inline void mem_cgroup_oom_disable(void)
307 static inline bool task_in_memcg_oom(struct task_struct *p)
312 static inline bool mem_cgroup_oom_synchronize(bool wait)
317 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
318 enum mem_cgroup_stat_index idx)
322 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
323 enum mem_cgroup_stat_index idx)
328 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
330 unsigned long *total_scanned)
335 static inline void mem_cgroup_split_huge_fixup(struct page *head)
340 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
343 #endif /* CONFIG_MEMCG */
352 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
353 void sock_update_memcg(struct sock *sk);
354 void sock_release_memcg(struct sock *sk);
356 static inline void sock_update_memcg(struct sock *sk)
359 static inline void sock_release_memcg(struct sock *sk)
362 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
364 #ifdef CONFIG_MEMCG_KMEM
365 extern struct static_key memcg_kmem_enabled_key;
367 extern int memcg_limited_groups_array_size;
370 * Helper macro to loop through all memcg-specific caches. Callers must still
371 * check if the cache is valid (it is either valid or NULL).
372 * the slab_mutex must be held when looping through those caches
374 #define for_each_memcg_cache_index(_idx) \
375 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
377 static inline bool memcg_kmem_enabled(void)
379 return static_key_false(&memcg_kmem_enabled_key);
383 * In general, we'll do everything in our power to not incur in any overhead
384 * for non-memcg users for the kmem functions. Not even a function call, if we
387 * Therefore, we'll inline all those functions so that in the best case, we'll
388 * see that kmemcg is off for everybody and proceed quickly. If it is on,
389 * we'll still do most of the flag checking inline. We check a lot of
390 * conditions, but because they are pretty simple, they are expected to be
393 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
395 void __memcg_kmem_commit_charge(struct page *page,
396 struct mem_cgroup *memcg, int order);
397 void __memcg_kmem_uncharge_pages(struct page *page, int order);
399 int memcg_cache_id(struct mem_cgroup *memcg);
401 void memcg_update_array_size(int num_groups);
404 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
406 int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order);
407 void __memcg_uncharge_slab(struct kmem_cache *cachep, int order);
409 int __memcg_cleanup_cache_params(struct kmem_cache *s);
412 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
413 * @gfp: the gfp allocation flags.
414 * @memcg: a pointer to the memcg this was charged against.
415 * @order: allocation order.
417 * returns true if the memcg where the current task belongs can hold this
420 * We return true automatically if this allocation is not to be accounted to
424 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
426 if (!memcg_kmem_enabled())
430 * __GFP_NOFAIL allocations will move on even if charging is not
431 * possible. Therefore we don't even try, and have this allocation
432 * unaccounted. We could in theory charge it forcibly, but we hope
433 * those allocations are rare, and won't be worth the trouble.
435 if (gfp & __GFP_NOFAIL)
437 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
440 /* If the test is dying, just let it go. */
441 if (unlikely(fatal_signal_pending(current)))
444 return __memcg_kmem_newpage_charge(gfp, memcg, order);
448 * memcg_kmem_uncharge_pages: uncharge pages from memcg
449 * @page: pointer to struct page being freed
450 * @order: allocation order.
453 memcg_kmem_uncharge_pages(struct page *page, int order)
455 if (memcg_kmem_enabled())
456 __memcg_kmem_uncharge_pages(page, order);
460 * memcg_kmem_commit_charge: embeds correct memcg in a page
461 * @page: pointer to struct page recently allocated
462 * @memcg: the memcg structure we charged against
463 * @order: allocation order.
465 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
466 * failure of the allocation. if @page is NULL, this function will revert the
467 * charges. Otherwise, it will commit @page to @memcg.
470 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
472 if (memcg_kmem_enabled() && memcg)
473 __memcg_kmem_commit_charge(page, memcg, order);
477 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
478 * @cachep: the original global kmem cache
479 * @gfp: allocation flags.
481 * All memory allocated from a per-memcg cache is charged to the owner memcg.
483 static __always_inline struct kmem_cache *
484 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
486 if (!memcg_kmem_enabled())
488 if (gfp & __GFP_NOFAIL)
490 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
492 if (unlikely(fatal_signal_pending(current)))
495 return __memcg_kmem_get_cache(cachep, gfp);
498 #define for_each_memcg_cache_index(_idx) \
501 static inline bool memcg_kmem_enabled(void)
507 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
512 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
517 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
521 static inline int memcg_cache_id(struct mem_cgroup *memcg)
526 static inline struct kmem_cache *
527 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
531 #endif /* CONFIG_MEMCG_KMEM */
532 #endif /* _LINUX_MEMCONTROL_H */