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);
62 void mem_cgroup_uncharge(struct page *page);
63 void mem_cgroup_uncharge_list(struct list_head *page_list);
65 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage,
68 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
69 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
71 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
72 struct mem_cgroup *memcg);
73 bool task_in_mem_cgroup(struct task_struct *task,
74 const struct mem_cgroup *memcg);
76 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
77 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
79 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
80 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
83 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
85 struct mem_cgroup *task_memcg;
89 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
91 match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
96 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
98 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
100 struct mem_cgroup_reclaim_cookie *);
101 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
104 * For memory reclaim.
106 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
107 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
108 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
109 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
110 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
111 struct task_struct *p);
113 static inline void mem_cgroup_oom_enable(void)
115 WARN_ON(current->memcg_oom.may_oom);
116 current->memcg_oom.may_oom = 1;
119 static inline void mem_cgroup_oom_disable(void)
121 WARN_ON(!current->memcg_oom.may_oom);
122 current->memcg_oom.may_oom = 0;
125 static inline bool task_in_memcg_oom(struct task_struct *p)
127 return p->memcg_oom.memcg;
130 bool mem_cgroup_oom_synchronize(bool wait);
132 #ifdef CONFIG_MEMCG_SWAP
133 extern int do_swap_account;
136 static inline bool mem_cgroup_disabled(void)
138 if (memory_cgrp_subsys.disabled)
143 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page, bool *locked,
144 unsigned long *flags);
145 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool locked,
146 unsigned long flags);
147 void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
148 enum mem_cgroup_stat_index idx, int val);
150 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
151 enum mem_cgroup_stat_index idx)
153 mem_cgroup_update_page_stat(memcg, idx, 1);
156 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
157 enum mem_cgroup_stat_index idx)
159 mem_cgroup_update_page_stat(memcg, idx, -1);
162 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
164 unsigned long *total_scanned);
166 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
167 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
168 enum vm_event_item idx)
170 if (mem_cgroup_disabled())
172 __mem_cgroup_count_vm_event(mm, idx);
174 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
175 void mem_cgroup_split_huge_fixup(struct page *head);
178 #ifdef CONFIG_DEBUG_VM
179 bool mem_cgroup_bad_page_check(struct page *page);
180 void mem_cgroup_print_bad_page(struct page *page);
182 #else /* CONFIG_MEMCG */
185 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
187 struct mem_cgroup **memcgp)
193 static inline void mem_cgroup_commit_charge(struct page *page,
194 struct mem_cgroup *memcg,
199 static inline void mem_cgroup_cancel_charge(struct page *page,
200 struct mem_cgroup *memcg)
204 static inline void mem_cgroup_uncharge(struct page *page)
208 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
212 static inline void mem_cgroup_migrate(struct page *oldpage,
213 struct page *newpage,
218 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
219 struct mem_cgroup *memcg)
221 return &zone->lruvec;
224 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
227 return &zone->lruvec;
230 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
235 static inline bool mm_match_cgroup(struct mm_struct *mm,
236 struct mem_cgroup *memcg)
241 static inline bool task_in_mem_cgroup(struct task_struct *task,
242 const struct mem_cgroup *memcg)
247 static inline struct cgroup_subsys_state
248 *mem_cgroup_css(struct mem_cgroup *memcg)
253 static inline struct mem_cgroup *
254 mem_cgroup_iter(struct mem_cgroup *root,
255 struct mem_cgroup *prev,
256 struct mem_cgroup_reclaim_cookie *reclaim)
261 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
262 struct mem_cgroup *prev)
266 static inline bool mem_cgroup_disabled(void)
272 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
277 static inline unsigned long
278 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
284 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
290 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
294 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page,
295 bool *locked, unsigned long *flags)
300 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg,
301 bool locked, unsigned long flags)
305 static inline void mem_cgroup_oom_enable(void)
309 static inline void mem_cgroup_oom_disable(void)
313 static inline bool task_in_memcg_oom(struct task_struct *p)
318 static inline bool mem_cgroup_oom_synchronize(bool wait)
323 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
324 enum mem_cgroup_stat_index idx)
328 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
329 enum mem_cgroup_stat_index idx)
334 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
336 unsigned long *total_scanned)
341 static inline void mem_cgroup_split_huge_fixup(struct page *head)
346 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
349 #endif /* CONFIG_MEMCG */
351 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
353 mem_cgroup_bad_page_check(struct page *page)
359 mem_cgroup_print_bad_page(struct page *page)
371 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
372 void sock_update_memcg(struct sock *sk);
373 void sock_release_memcg(struct sock *sk);
375 static inline void sock_update_memcg(struct sock *sk)
378 static inline void sock_release_memcg(struct sock *sk)
381 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
383 #ifdef CONFIG_MEMCG_KMEM
384 extern struct static_key memcg_kmem_enabled_key;
386 extern int memcg_limited_groups_array_size;
389 * Helper macro to loop through all memcg-specific caches. Callers must still
390 * check if the cache is valid (it is either valid or NULL).
391 * the slab_mutex must be held when looping through those caches
393 #define for_each_memcg_cache_index(_idx) \
394 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
396 static inline bool memcg_kmem_enabled(void)
398 return static_key_false(&memcg_kmem_enabled_key);
402 * In general, we'll do everything in our power to not incur in any overhead
403 * for non-memcg users for the kmem functions. Not even a function call, if we
406 * Therefore, we'll inline all those functions so that in the best case, we'll
407 * see that kmemcg is off for everybody and proceed quickly. If it is on,
408 * we'll still do most of the flag checking inline. We check a lot of
409 * conditions, but because they are pretty simple, they are expected to be
412 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
414 void __memcg_kmem_commit_charge(struct page *page,
415 struct mem_cgroup *memcg, int order);
416 void __memcg_kmem_uncharge_pages(struct page *page, int order);
418 int memcg_cache_id(struct mem_cgroup *memcg);
420 void memcg_update_array_size(int num_groups);
423 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
425 int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order);
426 void __memcg_uncharge_slab(struct kmem_cache *cachep, int order);
428 int __memcg_cleanup_cache_params(struct kmem_cache *s);
431 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
432 * @gfp: the gfp allocation flags.
433 * @memcg: a pointer to the memcg this was charged against.
434 * @order: allocation order.
436 * returns true if the memcg where the current task belongs can hold this
439 * We return true automatically if this allocation is not to be accounted to
443 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
445 if (!memcg_kmem_enabled())
449 * __GFP_NOFAIL allocations will move on even if charging is not
450 * possible. Therefore we don't even try, and have this allocation
451 * unaccounted. We could in theory charge it forcibly, but we hope
452 * those allocations are rare, and won't be worth the trouble.
454 if (gfp & __GFP_NOFAIL)
456 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
459 /* If the test is dying, just let it go. */
460 if (unlikely(fatal_signal_pending(current)))
463 return __memcg_kmem_newpage_charge(gfp, memcg, order);
467 * memcg_kmem_uncharge_pages: uncharge pages from memcg
468 * @page: pointer to struct page being freed
469 * @order: allocation order.
471 * there is no need to specify memcg here, since it is embedded in page_cgroup
474 memcg_kmem_uncharge_pages(struct page *page, int order)
476 if (memcg_kmem_enabled())
477 __memcg_kmem_uncharge_pages(page, order);
481 * memcg_kmem_commit_charge: embeds correct memcg in a page
482 * @page: pointer to struct page recently allocated
483 * @memcg: the memcg structure we charged against
484 * @order: allocation order.
486 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
487 * failure of the allocation. if @page is NULL, this function will revert the
488 * charges. Otherwise, it will commit the memcg given by @memcg to the
489 * corresponding page_cgroup.
492 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
494 if (memcg_kmem_enabled() && memcg)
495 __memcg_kmem_commit_charge(page, memcg, order);
499 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
500 * @cachep: the original global kmem cache
501 * @gfp: allocation flags.
503 * All memory allocated from a per-memcg cache is charged to the owner memcg.
505 static __always_inline struct kmem_cache *
506 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
508 if (!memcg_kmem_enabled())
510 if (gfp & __GFP_NOFAIL)
512 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
514 if (unlikely(fatal_signal_pending(current)))
517 return __memcg_kmem_get_cache(cachep, gfp);
520 #define for_each_memcg_cache_index(_idx) \
523 static inline bool memcg_kmem_enabled(void)
529 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
534 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
539 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
543 static inline int memcg_cache_id(struct mem_cgroup *memcg)
548 static inline struct kmem_cache *
549 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
553 #endif /* CONFIG_MEMCG_KMEM */
554 #endif /* _LINUX_MEMCONTROL_H */