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
));
90 match
= __mem_cgroup_same_or_subtree(memcg
, task_memcg
);
95 extern struct cgroup_subsys_state
*mem_cgroup_css(struct mem_cgroup
*memcg
);
97 struct mem_cgroup
*mem_cgroup_iter(struct mem_cgroup
*,
99 struct mem_cgroup_reclaim_cookie
*);
100 void mem_cgroup_iter_break(struct mem_cgroup
*, struct mem_cgroup
*);
103 * For memory reclaim.
105 int mem_cgroup_inactive_anon_is_low(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 void __mem_cgroup_begin_update_page_stat(struct page
*page
, bool *locked
,
143 unsigned long *flags
);
145 extern atomic_t memcg_moving
;
147 static inline void mem_cgroup_begin_update_page_stat(struct page
*page
,
148 bool *locked
, unsigned long *flags
)
150 if (mem_cgroup_disabled())
154 if (atomic_read(&memcg_moving
))
155 __mem_cgroup_begin_update_page_stat(page
, locked
, flags
);
158 void __mem_cgroup_end_update_page_stat(struct page
*page
,
159 unsigned long *flags
);
160 static inline void mem_cgroup_end_update_page_stat(struct page
*page
,
161 bool *locked
, unsigned long *flags
)
163 if (mem_cgroup_disabled())
166 __mem_cgroup_end_update_page_stat(page
, flags
);
170 void mem_cgroup_update_page_stat(struct page
*page
,
171 enum mem_cgroup_stat_index idx
,
174 static inline void mem_cgroup_inc_page_stat(struct page
*page
,
175 enum mem_cgroup_stat_index idx
)
177 mem_cgroup_update_page_stat(page
, idx
, 1);
180 static inline void mem_cgroup_dec_page_stat(struct page
*page
,
181 enum mem_cgroup_stat_index idx
)
183 mem_cgroup_update_page_stat(page
, idx
, -1);
186 unsigned long mem_cgroup_soft_limit_reclaim(struct zone
*zone
, int order
,
188 unsigned long *total_scanned
);
190 void __mem_cgroup_count_vm_event(struct mm_struct
*mm
, enum vm_event_item idx
);
191 static inline void mem_cgroup_count_vm_event(struct mm_struct
*mm
,
192 enum vm_event_item idx
)
194 if (mem_cgroup_disabled())
196 __mem_cgroup_count_vm_event(mm
, idx
);
198 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
199 void mem_cgroup_split_huge_fixup(struct page
*head
);
202 #ifdef CONFIG_DEBUG_VM
203 bool mem_cgroup_bad_page_check(struct page
*page
);
204 void mem_cgroup_print_bad_page(struct page
*page
);
206 #else /* CONFIG_MEMCG */
209 static inline int mem_cgroup_try_charge(struct page
*page
, struct mm_struct
*mm
,
211 struct mem_cgroup
**memcgp
)
217 static inline void mem_cgroup_commit_charge(struct page
*page
,
218 struct mem_cgroup
*memcg
,
223 static inline void mem_cgroup_cancel_charge(struct page
*page
,
224 struct mem_cgroup
*memcg
)
228 static inline void mem_cgroup_uncharge(struct page
*page
)
232 static inline void mem_cgroup_uncharge_list(struct list_head
*page_list
)
236 static inline void mem_cgroup_migrate(struct page
*oldpage
,
237 struct page
*newpage
,
242 static inline struct lruvec
*mem_cgroup_zone_lruvec(struct zone
*zone
,
243 struct mem_cgroup
*memcg
)
245 return &zone
->lruvec
;
248 static inline struct lruvec
*mem_cgroup_page_lruvec(struct page
*page
,
251 return &zone
->lruvec
;
254 static inline struct mem_cgroup
*try_get_mem_cgroup_from_page(struct page
*page
)
259 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
260 struct mem_cgroup
*memcg
)
265 static inline bool task_in_mem_cgroup(struct task_struct
*task
,
266 const struct mem_cgroup
*memcg
)
271 static inline struct cgroup_subsys_state
272 *mem_cgroup_css(struct mem_cgroup
*memcg
)
277 static inline struct mem_cgroup
*
278 mem_cgroup_iter(struct mem_cgroup
*root
,
279 struct mem_cgroup
*prev
,
280 struct mem_cgroup_reclaim_cookie
*reclaim
)
285 static inline void mem_cgroup_iter_break(struct mem_cgroup
*root
,
286 struct mem_cgroup
*prev
)
290 static inline bool mem_cgroup_disabled(void)
296 mem_cgroup_inactive_anon_is_low(struct lruvec
*lruvec
)
301 static inline unsigned long
302 mem_cgroup_get_lru_size(struct lruvec
*lruvec
, enum lru_list lru
)
308 mem_cgroup_update_lru_size(struct lruvec
*lruvec
, enum lru_list lru
,
314 mem_cgroup_print_oom_info(struct mem_cgroup
*memcg
, struct task_struct
*p
)
318 static inline void mem_cgroup_begin_update_page_stat(struct page
*page
,
319 bool *locked
, unsigned long *flags
)
323 static inline void mem_cgroup_end_update_page_stat(struct page
*page
,
324 bool *locked
, unsigned long *flags
)
328 static inline void mem_cgroup_oom_enable(void)
332 static inline void mem_cgroup_oom_disable(void)
336 static inline bool task_in_memcg_oom(struct task_struct
*p
)
341 static inline bool mem_cgroup_oom_synchronize(bool wait
)
346 static inline void mem_cgroup_inc_page_stat(struct page
*page
,
347 enum mem_cgroup_stat_index idx
)
351 static inline void mem_cgroup_dec_page_stat(struct page
*page
,
352 enum mem_cgroup_stat_index idx
)
357 unsigned long mem_cgroup_soft_limit_reclaim(struct zone
*zone
, int order
,
359 unsigned long *total_scanned
)
364 static inline void mem_cgroup_split_huge_fixup(struct page
*head
)
369 void mem_cgroup_count_vm_event(struct mm_struct
*mm
, enum vm_event_item idx
)
372 #endif /* CONFIG_MEMCG */
374 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
376 mem_cgroup_bad_page_check(struct page
*page
)
382 mem_cgroup_print_bad_page(struct page
*page
)
394 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
395 void sock_update_memcg(struct sock
*sk
);
396 void sock_release_memcg(struct sock
*sk
);
398 static inline void sock_update_memcg(struct sock
*sk
)
401 static inline void sock_release_memcg(struct sock
*sk
)
404 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
406 #ifdef CONFIG_MEMCG_KMEM
407 extern struct static_key memcg_kmem_enabled_key
;
409 extern int memcg_limited_groups_array_size
;
412 * Helper macro to loop through all memcg-specific caches. Callers must still
413 * check if the cache is valid (it is either valid or NULL).
414 * the slab_mutex must be held when looping through those caches
416 #define for_each_memcg_cache_index(_idx) \
417 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
419 static inline bool memcg_kmem_enabled(void)
421 return static_key_false(&memcg_kmem_enabled_key
);
425 * In general, we'll do everything in our power to not incur in any overhead
426 * for non-memcg users for the kmem functions. Not even a function call, if we
429 * Therefore, we'll inline all those functions so that in the best case, we'll
430 * see that kmemcg is off for everybody and proceed quickly. If it is on,
431 * we'll still do most of the flag checking inline. We check a lot of
432 * conditions, but because they are pretty simple, they are expected to be
435 bool __memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
,
437 void __memcg_kmem_commit_charge(struct page
*page
,
438 struct mem_cgroup
*memcg
, int order
);
439 void __memcg_kmem_uncharge_pages(struct page
*page
, int order
);
441 int memcg_cache_id(struct mem_cgroup
*memcg
);
443 int memcg_alloc_cache_params(struct mem_cgroup
*memcg
, struct kmem_cache
*s
,
444 struct kmem_cache
*root_cache
);
445 void memcg_free_cache_params(struct kmem_cache
*s
);
447 int memcg_update_cache_size(struct kmem_cache
*s
, int num_groups
);
448 void memcg_update_array_size(int num_groups
);
451 __memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
);
453 int __memcg_charge_slab(struct kmem_cache
*cachep
, gfp_t gfp
, int order
);
454 void __memcg_uncharge_slab(struct kmem_cache
*cachep
, int order
);
456 int __memcg_cleanup_cache_params(struct kmem_cache
*s
);
459 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
460 * @gfp: the gfp allocation flags.
461 * @memcg: a pointer to the memcg this was charged against.
462 * @order: allocation order.
464 * returns true if the memcg where the current task belongs can hold this
467 * We return true automatically if this allocation is not to be accounted to
471 memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
, int order
)
473 if (!memcg_kmem_enabled())
477 * __GFP_NOFAIL allocations will move on even if charging is not
478 * possible. Therefore we don't even try, and have this allocation
479 * unaccounted. We could in theory charge it with
480 * res_counter_charge_nofail, but we hope those allocations are rare,
481 * and won't be worth the trouble.
483 if (gfp
& __GFP_NOFAIL
)
485 if (in_interrupt() || (!current
->mm
) || (current
->flags
& PF_KTHREAD
))
488 /* If the test is dying, just let it go. */
489 if (unlikely(fatal_signal_pending(current
)))
492 return __memcg_kmem_newpage_charge(gfp
, memcg
, order
);
496 * memcg_kmem_uncharge_pages: uncharge pages from memcg
497 * @page: pointer to struct page being freed
498 * @order: allocation order.
500 * there is no need to specify memcg here, since it is embedded in page_cgroup
503 memcg_kmem_uncharge_pages(struct page
*page
, int order
)
505 if (memcg_kmem_enabled())
506 __memcg_kmem_uncharge_pages(page
, order
);
510 * memcg_kmem_commit_charge: embeds correct memcg in a page
511 * @page: pointer to struct page recently allocated
512 * @memcg: the memcg structure we charged against
513 * @order: allocation order.
515 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
516 * failure of the allocation. if @page is NULL, this function will revert the
517 * charges. Otherwise, it will commit the memcg given by @memcg to the
518 * corresponding page_cgroup.
521 memcg_kmem_commit_charge(struct page
*page
, struct mem_cgroup
*memcg
, int order
)
523 if (memcg_kmem_enabled() && memcg
)
524 __memcg_kmem_commit_charge(page
, memcg
, order
);
528 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
529 * @cachep: the original global kmem cache
530 * @gfp: allocation flags.
532 * All memory allocated from a per-memcg cache is charged to the owner memcg.
534 static __always_inline
struct kmem_cache
*
535 memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
)
537 if (!memcg_kmem_enabled())
539 if (gfp
& __GFP_NOFAIL
)
541 if (in_interrupt() || (!current
->mm
) || (current
->flags
& PF_KTHREAD
))
543 if (unlikely(fatal_signal_pending(current
)))
546 return __memcg_kmem_get_cache(cachep
, gfp
);
549 #define for_each_memcg_cache_index(_idx) \
552 static inline bool memcg_kmem_enabled(void)
558 memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
, int order
)
563 static inline void memcg_kmem_uncharge_pages(struct page
*page
, int order
)
568 memcg_kmem_commit_charge(struct page
*page
, struct mem_cgroup
*memcg
, int order
)
572 static inline int memcg_cache_id(struct mem_cgroup
*memcg
)
577 static inline int memcg_alloc_cache_params(struct mem_cgroup
*memcg
,
578 struct kmem_cache
*s
, struct kmem_cache
*root_cache
)
583 static inline void memcg_free_cache_params(struct kmem_cache
*s
)
587 static inline struct kmem_cache
*
588 memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
)
592 #endif /* CONFIG_MEMCG_KMEM */
593 #endif /* _LINUX_MEMCONTROL_H */