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 /* Stats that can be updated by kernel. */
34 enum mem_cgroup_page_stat_item
{
35 MEMCG_NR_FILE_MAPPED
, /* # of pages charged as file rss */
38 struct mem_cgroup_reclaim_cookie
{
41 unsigned int generation
;
46 * All "charge" functions with gfp_mask should use GFP_KERNEL or
47 * (gfp_mask & GFP_RECLAIM_MASK). In current implementatin, memcg doesn't
48 * alloc memory but reclaims memory from all available zones. So, "where I want
49 * memory from" bits of gfp_mask has no meaning. So any bits of that field is
50 * available but adding a rule is better. charge functions' gfp_mask should
51 * be set to GFP_KERNEL or gfp_mask & GFP_RECLAIM_MASK for avoiding ambiguous
53 * (Of course, if memcg does memory allocation in future, GFP_KERNEL is sane.)
56 extern int mem_cgroup_newpage_charge(struct page
*page
, struct mm_struct
*mm
,
58 /* for swap handling */
59 extern int mem_cgroup_try_charge_swapin(struct mm_struct
*mm
,
60 struct page
*page
, gfp_t mask
, struct mem_cgroup
**memcgp
);
61 extern void mem_cgroup_commit_charge_swapin(struct page
*page
,
62 struct mem_cgroup
*memcg
);
63 extern void mem_cgroup_cancel_charge_swapin(struct mem_cgroup
*memcg
);
65 extern int mem_cgroup_cache_charge(struct page
*page
, struct mm_struct
*mm
,
68 struct lruvec
*mem_cgroup_zone_lruvec(struct zone
*, struct mem_cgroup
*);
69 struct lruvec
*mem_cgroup_page_lruvec(struct page
*, struct zone
*);
71 /* For coalescing uncharge for reducing memcg' overhead*/
72 extern void mem_cgroup_uncharge_start(void);
73 extern void mem_cgroup_uncharge_end(void);
75 extern void mem_cgroup_uncharge_page(struct page
*page
);
76 extern void mem_cgroup_uncharge_cache_page(struct page
*page
);
78 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup
*root_memcg
,
79 struct mem_cgroup
*memcg
);
80 int task_in_mem_cgroup(struct task_struct
*task
, const struct mem_cgroup
*memcg
);
82 extern struct mem_cgroup
*try_get_mem_cgroup_from_page(struct page
*page
);
83 extern struct mem_cgroup
*mem_cgroup_from_task(struct task_struct
*p
);
84 extern struct mem_cgroup
*try_get_mem_cgroup_from_mm(struct mm_struct
*mm
);
86 extern struct mem_cgroup
*parent_mem_cgroup(struct mem_cgroup
*memcg
);
87 extern struct mem_cgroup
*mem_cgroup_from_cont(struct cgroup
*cont
);
90 bool mm_match_cgroup(const struct mm_struct
*mm
, const struct mem_cgroup
*memcg
)
92 struct mem_cgroup
*task_memcg
;
96 task_memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
97 match
= __mem_cgroup_same_or_subtree(memcg
, task_memcg
);
102 extern struct cgroup_subsys_state
*mem_cgroup_css(struct mem_cgroup
*memcg
);
105 mem_cgroup_prepare_migration(struct page
*page
, struct page
*newpage
,
106 struct mem_cgroup
**memcgp
);
107 extern void mem_cgroup_end_migration(struct mem_cgroup
*memcg
,
108 struct page
*oldpage
, struct page
*newpage
, bool migration_ok
);
110 struct mem_cgroup
*mem_cgroup_iter(struct mem_cgroup
*,
112 struct mem_cgroup_reclaim_cookie
*);
113 void mem_cgroup_iter_break(struct mem_cgroup
*, struct mem_cgroup
*);
116 * For memory reclaim.
118 int mem_cgroup_inactive_anon_is_low(struct lruvec
*lruvec
);
119 int mem_cgroup_inactive_file_is_low(struct lruvec
*lruvec
);
120 int mem_cgroup_select_victim_node(struct mem_cgroup
*memcg
);
121 unsigned long mem_cgroup_get_lru_size(struct lruvec
*lruvec
, enum lru_list
);
122 void mem_cgroup_update_lru_size(struct lruvec
*, enum lru_list
, int);
123 extern void mem_cgroup_print_oom_info(struct mem_cgroup
*memcg
,
124 struct task_struct
*p
);
125 extern void mem_cgroup_replace_page_cache(struct page
*oldpage
,
126 struct page
*newpage
);
128 #ifdef CONFIG_MEMCG_SWAP
129 extern int do_swap_account
;
132 static inline bool mem_cgroup_disabled(void)
134 if (mem_cgroup_subsys
.disabled
)
139 void __mem_cgroup_begin_update_page_stat(struct page
*page
, bool *locked
,
140 unsigned long *flags
);
142 extern atomic_t memcg_moving
;
144 static inline void mem_cgroup_begin_update_page_stat(struct page
*page
,
145 bool *locked
, unsigned long *flags
)
147 if (mem_cgroup_disabled())
151 if (atomic_read(&memcg_moving
))
152 __mem_cgroup_begin_update_page_stat(page
, locked
, flags
);
155 void __mem_cgroup_end_update_page_stat(struct page
*page
,
156 unsigned long *flags
);
157 static inline void mem_cgroup_end_update_page_stat(struct page
*page
,
158 bool *locked
, unsigned long *flags
)
160 if (mem_cgroup_disabled())
163 __mem_cgroup_end_update_page_stat(page
, flags
);
167 void mem_cgroup_update_page_stat(struct page
*page
,
168 enum mem_cgroup_page_stat_item idx
,
171 static inline void mem_cgroup_inc_page_stat(struct page
*page
,
172 enum mem_cgroup_page_stat_item idx
)
174 mem_cgroup_update_page_stat(page
, idx
, 1);
177 static inline void mem_cgroup_dec_page_stat(struct page
*page
,
178 enum mem_cgroup_page_stat_item idx
)
180 mem_cgroup_update_page_stat(page
, idx
, -1);
183 unsigned long mem_cgroup_soft_limit_reclaim(struct zone
*zone
, int order
,
185 unsigned long *total_scanned
);
187 void __mem_cgroup_count_vm_event(struct mm_struct
*mm
, enum vm_event_item idx
);
188 static inline void mem_cgroup_count_vm_event(struct mm_struct
*mm
,
189 enum vm_event_item idx
)
191 if (mem_cgroup_disabled())
193 __mem_cgroup_count_vm_event(mm
, idx
);
195 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
196 void mem_cgroup_split_huge_fixup(struct page
*head
);
199 #ifdef CONFIG_DEBUG_VM
200 bool mem_cgroup_bad_page_check(struct page
*page
);
201 void mem_cgroup_print_bad_page(struct page
*page
);
203 #else /* CONFIG_MEMCG */
206 static inline int mem_cgroup_newpage_charge(struct page
*page
,
207 struct mm_struct
*mm
, gfp_t gfp_mask
)
212 static inline int mem_cgroup_cache_charge(struct page
*page
,
213 struct mm_struct
*mm
, gfp_t gfp_mask
)
218 static inline int mem_cgroup_try_charge_swapin(struct mm_struct
*mm
,
219 struct page
*page
, gfp_t gfp_mask
, struct mem_cgroup
**memcgp
)
224 static inline void mem_cgroup_commit_charge_swapin(struct page
*page
,
225 struct mem_cgroup
*memcg
)
229 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup
*memcg
)
233 static inline void mem_cgroup_uncharge_start(void)
237 static inline void mem_cgroup_uncharge_end(void)
241 static inline void mem_cgroup_uncharge_page(struct page
*page
)
245 static inline void mem_cgroup_uncharge_cache_page(struct page
*page
)
249 static inline struct lruvec
*mem_cgroup_zone_lruvec(struct zone
*zone
,
250 struct mem_cgroup
*memcg
)
252 return &zone
->lruvec
;
255 static inline struct lruvec
*mem_cgroup_page_lruvec(struct page
*page
,
258 return &zone
->lruvec
;
261 static inline struct mem_cgroup
*try_get_mem_cgroup_from_page(struct page
*page
)
266 static inline struct mem_cgroup
*try_get_mem_cgroup_from_mm(struct mm_struct
*mm
)
271 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
272 struct mem_cgroup
*memcg
)
277 static inline int task_in_mem_cgroup(struct task_struct
*task
,
278 const struct mem_cgroup
*memcg
)
283 static inline struct cgroup_subsys_state
284 *mem_cgroup_css(struct mem_cgroup
*memcg
)
290 mem_cgroup_prepare_migration(struct page
*page
, struct page
*newpage
,
291 struct mem_cgroup
**memcgp
)
295 static inline void mem_cgroup_end_migration(struct mem_cgroup
*memcg
,
296 struct page
*oldpage
, struct page
*newpage
, bool migration_ok
)
300 static inline struct mem_cgroup
*
301 mem_cgroup_iter(struct mem_cgroup
*root
,
302 struct mem_cgroup
*prev
,
303 struct mem_cgroup_reclaim_cookie
*reclaim
)
308 static inline void mem_cgroup_iter_break(struct mem_cgroup
*root
,
309 struct mem_cgroup
*prev
)
313 static inline bool mem_cgroup_disabled(void)
319 mem_cgroup_inactive_anon_is_low(struct lruvec
*lruvec
)
325 mem_cgroup_inactive_file_is_low(struct lruvec
*lruvec
)
330 static inline unsigned long
331 mem_cgroup_get_lru_size(struct lruvec
*lruvec
, enum lru_list lru
)
337 mem_cgroup_update_lru_size(struct lruvec
*lruvec
, enum lru_list lru
,
343 mem_cgroup_print_oom_info(struct mem_cgroup
*memcg
, struct task_struct
*p
)
347 static inline void mem_cgroup_begin_update_page_stat(struct page
*page
,
348 bool *locked
, unsigned long *flags
)
352 static inline void mem_cgroup_end_update_page_stat(struct page
*page
,
353 bool *locked
, unsigned long *flags
)
357 static inline void mem_cgroup_inc_page_stat(struct page
*page
,
358 enum mem_cgroup_page_stat_item idx
)
362 static inline void mem_cgroup_dec_page_stat(struct page
*page
,
363 enum mem_cgroup_page_stat_item idx
)
368 unsigned long mem_cgroup_soft_limit_reclaim(struct zone
*zone
, int order
,
370 unsigned long *total_scanned
)
375 static inline void mem_cgroup_split_huge_fixup(struct page
*head
)
380 void mem_cgroup_count_vm_event(struct mm_struct
*mm
, enum vm_event_item idx
)
383 static inline void mem_cgroup_replace_page_cache(struct page
*oldpage
,
384 struct page
*newpage
)
387 #endif /* CONFIG_MEMCG */
389 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
391 mem_cgroup_bad_page_check(struct page
*page
)
397 mem_cgroup_print_bad_page(struct page
*page
)
409 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
410 void sock_update_memcg(struct sock
*sk
);
411 void sock_release_memcg(struct sock
*sk
);
413 static inline void sock_update_memcg(struct sock
*sk
)
416 static inline void sock_release_memcg(struct sock
*sk
)
419 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
421 #ifdef CONFIG_MEMCG_KMEM
422 extern struct static_key memcg_kmem_enabled_key
;
423 static inline bool memcg_kmem_enabled(void)
425 return static_key_false(&memcg_kmem_enabled_key
);
429 * In general, we'll do everything in our power to not incur in any overhead
430 * for non-memcg users for the kmem functions. Not even a function call, if we
433 * Therefore, we'll inline all those functions so that in the best case, we'll
434 * see that kmemcg is off for everybody and proceed quickly. If it is on,
435 * we'll still do most of the flag checking inline. We check a lot of
436 * conditions, but because they are pretty simple, they are expected to be
439 bool __memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
,
441 void __memcg_kmem_commit_charge(struct page
*page
,
442 struct mem_cgroup
*memcg
, int order
);
443 void __memcg_kmem_uncharge_pages(struct page
*page
, int order
);
445 int memcg_cache_id(struct mem_cgroup
*memcg
);
446 int memcg_register_cache(struct mem_cgroup
*memcg
, struct kmem_cache
*s
);
447 void memcg_release_cache(struct kmem_cache
*cachep
);
448 void memcg_cache_list_add(struct mem_cgroup
*memcg
, struct kmem_cache
*cachep
);
450 int memcg_update_cache_size(struct kmem_cache
*s
, int num_groups
);
451 void memcg_update_array_size(int num_groups
);
454 __memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
);
456 void mem_cgroup_destroy_cache(struct kmem_cache
*cachep
);
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_KMEMCG
) || (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 * This function assumes that the task allocating, which determines the memcg
533 * in the page allocator, belongs to the same cgroup throughout the whole
534 * process. Misacounting can happen if the task calls memcg_kmem_get_cache()
535 * while belonging to a cgroup, and later on changes. This is considered
536 * acceptable, and should only happen upon task migration.
538 * Before the cache is created by the memcg core, there is also a possible
539 * imbalance: the task belongs to a memcg, but the cache being allocated from
540 * is the global cache, since the child cache is not yet guaranteed to be
541 * ready. This case is also fine, since in this case the GFP_KMEMCG will not be
542 * passed and the page allocator will not attempt any cgroup accounting.
544 static __always_inline
struct kmem_cache
*
545 memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
)
547 if (!memcg_kmem_enabled())
549 if (gfp
& __GFP_NOFAIL
)
551 if (in_interrupt() || (!current
->mm
) || (current
->flags
& PF_KTHREAD
))
553 if (unlikely(fatal_signal_pending(current
)))
556 return __memcg_kmem_get_cache(cachep
, gfp
);
559 static inline bool memcg_kmem_enabled(void)
565 memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
, int order
)
570 static inline void memcg_kmem_uncharge_pages(struct page
*page
, int order
)
575 memcg_kmem_commit_charge(struct page
*page
, struct mem_cgroup
*memcg
, int order
)
579 static inline int memcg_cache_id(struct mem_cgroup
*memcg
)
584 static inline int memcg_register_cache(struct mem_cgroup
*memcg
,
585 struct kmem_cache
*s
)
590 static inline void memcg_release_cache(struct kmem_cache
*cachep
)
594 static inline void memcg_cache_list_add(struct mem_cgroup
*memcg
,
595 struct kmem_cache
*s
)
599 static inline struct kmem_cache
*
600 memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
)
604 #endif /* CONFIG_MEMCG_KMEM */
605 #endif /* _LINUX_MEMCONTROL_H */