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 bool mem_cgroup_lruvec_online(struct lruvec *lruvec);
106 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
107 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
108 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
109 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
110 struct task_struct *p);
112 static inline void mem_cgroup_oom_enable(void)
114 WARN_ON(current->memcg_oom.may_oom);
115 current->memcg_oom.may_oom = 1;
118 static inline void mem_cgroup_oom_disable(void)
120 WARN_ON(!current->memcg_oom.may_oom);
121 current->memcg_oom.may_oom = 0;
124 static inline bool task_in_memcg_oom(struct task_struct *p)
126 return p->memcg_oom.memcg;
129 bool mem_cgroup_oom_synchronize(bool wait);
131 #ifdef CONFIG_MEMCG_SWAP
132 extern int do_swap_account;
135 static inline bool mem_cgroup_disabled(void)
137 if (memory_cgrp_subsys.disabled)
142 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
143 void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
144 enum mem_cgroup_stat_index idx, int val);
145 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
147 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
148 enum mem_cgroup_stat_index idx)
150 mem_cgroup_update_page_stat(memcg, idx, 1);
153 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
154 enum mem_cgroup_stat_index idx)
156 mem_cgroup_update_page_stat(memcg, idx, -1);
159 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
161 unsigned long *total_scanned);
163 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
164 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
165 enum vm_event_item idx)
167 if (mem_cgroup_disabled())
169 __mem_cgroup_count_vm_event(mm, idx);
171 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
172 void mem_cgroup_split_huge_fixup(struct page *head);
175 #else /* CONFIG_MEMCG */
178 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
180 struct mem_cgroup **memcgp)
186 static inline void mem_cgroup_commit_charge(struct page *page,
187 struct mem_cgroup *memcg,
192 static inline void mem_cgroup_cancel_charge(struct page *page,
193 struct mem_cgroup *memcg)
197 static inline void mem_cgroup_uncharge(struct page *page)
201 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
205 static inline void mem_cgroup_migrate(struct page *oldpage,
206 struct page *newpage,
211 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
212 struct mem_cgroup *memcg)
214 return &zone->lruvec;
217 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
220 return &zone->lruvec;
223 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
228 static inline bool mm_match_cgroup(struct mm_struct *mm,
229 struct mem_cgroup *memcg)
234 static inline bool task_in_mem_cgroup(struct task_struct *task,
235 const struct mem_cgroup *memcg)
240 static inline struct cgroup_subsys_state
241 *mem_cgroup_css(struct mem_cgroup *memcg)
246 static inline struct mem_cgroup *
247 mem_cgroup_iter(struct mem_cgroup *root,
248 struct mem_cgroup *prev,
249 struct mem_cgroup_reclaim_cookie *reclaim)
254 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
255 struct mem_cgroup *prev)
259 static inline bool mem_cgroup_disabled(void)
265 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
270 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
275 static inline unsigned long
276 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
282 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
288 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
292 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
297 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
301 static inline void mem_cgroup_oom_enable(void)
305 static inline void mem_cgroup_oom_disable(void)
309 static inline bool task_in_memcg_oom(struct task_struct *p)
314 static inline bool mem_cgroup_oom_synchronize(bool wait)
319 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
320 enum mem_cgroup_stat_index idx)
324 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
325 enum mem_cgroup_stat_index idx)
330 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
332 unsigned long *total_scanned)
337 static inline void mem_cgroup_split_huge_fixup(struct page *head)
342 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
345 #endif /* CONFIG_MEMCG */
354 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
355 void sock_update_memcg(struct sock *sk);
356 void sock_release_memcg(struct sock *sk);
358 static inline void sock_update_memcg(struct sock *sk)
361 static inline void sock_release_memcg(struct sock *sk)
364 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
366 #ifdef CONFIG_MEMCG_KMEM
367 extern struct static_key memcg_kmem_enabled_key;
369 extern int memcg_limited_groups_array_size;
372 * Helper macro to loop through all memcg-specific caches. Callers must still
373 * check if the cache is valid (it is either valid or NULL).
374 * the slab_mutex must be held when looping through those caches
376 #define for_each_memcg_cache_index(_idx) \
377 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
379 static inline bool memcg_kmem_enabled(void)
381 return static_key_false(&memcg_kmem_enabled_key);
385 * In general, we'll do everything in our power to not incur in any overhead
386 * for non-memcg users for the kmem functions. Not even a function call, if we
389 * Therefore, we'll inline all those functions so that in the best case, we'll
390 * see that kmemcg is off for everybody and proceed quickly. If it is on,
391 * we'll still do most of the flag checking inline. We check a lot of
392 * conditions, but because they are pretty simple, they are expected to be
395 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
397 void __memcg_kmem_commit_charge(struct page *page,
398 struct mem_cgroup *memcg, int order);
399 void __memcg_kmem_uncharge_pages(struct page *page, int order);
401 int memcg_cache_id(struct mem_cgroup *memcg);
403 void memcg_update_array_size(int num_groups);
405 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep);
406 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
408 int memcg_charge_kmem(struct mem_cgroup *memcg, gfp_t gfp,
409 unsigned long nr_pages);
410 void memcg_uncharge_kmem(struct mem_cgroup *memcg, unsigned long nr_pages);
413 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
414 * @gfp: the gfp allocation flags.
415 * @memcg: a pointer to the memcg this was charged against.
416 * @order: allocation order.
418 * returns true if the memcg where the current task belongs can hold this
421 * We return true automatically if this allocation is not to be accounted to
425 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
427 if (!memcg_kmem_enabled())
431 * __GFP_NOFAIL allocations will move on even if charging is not
432 * possible. Therefore we don't even try, and have this allocation
433 * unaccounted. We could in theory charge it forcibly, but we hope
434 * those allocations are rare, and won't be worth the trouble.
436 if (gfp & __GFP_NOFAIL)
438 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
441 /* If the test is dying, just let it go. */
442 if (unlikely(fatal_signal_pending(current)))
445 return __memcg_kmem_newpage_charge(gfp, memcg, order);
449 * memcg_kmem_uncharge_pages: uncharge pages from memcg
450 * @page: pointer to struct page being freed
451 * @order: allocation order.
454 memcg_kmem_uncharge_pages(struct page *page, int order)
456 if (memcg_kmem_enabled())
457 __memcg_kmem_uncharge_pages(page, order);
461 * memcg_kmem_commit_charge: embeds correct memcg in a page
462 * @page: pointer to struct page recently allocated
463 * @memcg: the memcg structure we charged against
464 * @order: allocation order.
466 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
467 * failure of the allocation. if @page is NULL, this function will revert the
468 * charges. Otherwise, it will commit @page to @memcg.
471 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
473 if (memcg_kmem_enabled() && memcg)
474 __memcg_kmem_commit_charge(page, memcg, order);
478 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
479 * @cachep: the original global kmem cache
480 * @gfp: allocation flags.
482 * All memory allocated from a per-memcg cache is charged to the owner memcg.
484 static __always_inline struct kmem_cache *
485 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
487 if (!memcg_kmem_enabled())
489 if (gfp & __GFP_NOFAIL)
491 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
493 if (unlikely(fatal_signal_pending(current)))
496 return __memcg_kmem_get_cache(cachep);
499 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
501 if (memcg_kmem_enabled())
502 __memcg_kmem_put_cache(cachep);
505 #define for_each_memcg_cache_index(_idx) \
508 static inline bool memcg_kmem_enabled(void)
514 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
519 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
524 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
528 static inline int memcg_cache_id(struct mem_cgroup *memcg)
533 static inline struct kmem_cache *
534 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
539 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
542 #endif /* CONFIG_MEMCG_KMEM */
543 #endif /* _LINUX_MEMCONTROL_H */