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_is_descendant(struct mem_cgroup
*memcg
,
72 struct mem_cgroup
*root
);
73 bool task_in_mem_cgroup(struct task_struct
*task
, struct mem_cgroup
*memcg
);
75 extern struct mem_cgroup
*try_get_mem_cgroup_from_page(struct page
*page
);
76 extern struct mem_cgroup
*mem_cgroup_from_task(struct task_struct
*p
);
78 extern struct mem_cgroup
*parent_mem_cgroup(struct mem_cgroup
*memcg
);
79 extern struct mem_cgroup
*mem_cgroup_from_css(struct cgroup_subsys_state
*css
);
81 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
82 struct mem_cgroup
*memcg
)
84 struct mem_cgroup
*task_memcg
;
88 task_memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
90 match
= mem_cgroup_is_descendant(task_memcg
, 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 struct mem_cgroup
*mem_cgroup_begin_page_stat(struct page
*page
, bool *locked
,
143 unsigned long *flags
);
144 void mem_cgroup_end_page_stat(struct mem_cgroup
*memcg
, bool *locked
,
145 unsigned long *flags
);
146 void mem_cgroup_update_page_stat(struct mem_cgroup
*memcg
,
147 enum mem_cgroup_stat_index idx
, int val
);
149 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup
*memcg
,
150 enum mem_cgroup_stat_index idx
)
152 mem_cgroup_update_page_stat(memcg
, idx
, 1);
155 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup
*memcg
,
156 enum mem_cgroup_stat_index idx
)
158 mem_cgroup_update_page_stat(memcg
, idx
, -1);
161 unsigned long mem_cgroup_soft_limit_reclaim(struct zone
*zone
, int order
,
163 unsigned long *total_scanned
);
165 void __mem_cgroup_count_vm_event(struct mm_struct
*mm
, enum vm_event_item idx
);
166 static inline void mem_cgroup_count_vm_event(struct mm_struct
*mm
,
167 enum vm_event_item idx
)
169 if (mem_cgroup_disabled())
171 __mem_cgroup_count_vm_event(mm
, idx
);
173 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
174 void mem_cgroup_split_huge_fixup(struct page
*head
);
177 #ifdef CONFIG_DEBUG_VM
178 bool mem_cgroup_bad_page_check(struct page
*page
);
179 void mem_cgroup_print_bad_page(struct page
*page
);
181 #else /* CONFIG_MEMCG */
184 static inline int mem_cgroup_try_charge(struct page
*page
, struct mm_struct
*mm
,
186 struct mem_cgroup
**memcgp
)
192 static inline void mem_cgroup_commit_charge(struct page
*page
,
193 struct mem_cgroup
*memcg
,
198 static inline void mem_cgroup_cancel_charge(struct page
*page
,
199 struct mem_cgroup
*memcg
)
203 static inline void mem_cgroup_uncharge(struct page
*page
)
207 static inline void mem_cgroup_uncharge_list(struct list_head
*page_list
)
211 static inline void mem_cgroup_migrate(struct page
*oldpage
,
212 struct page
*newpage
,
217 static inline struct lruvec
*mem_cgroup_zone_lruvec(struct zone
*zone
,
218 struct mem_cgroup
*memcg
)
220 return &zone
->lruvec
;
223 static inline struct lruvec
*mem_cgroup_page_lruvec(struct page
*page
,
226 return &zone
->lruvec
;
229 static inline struct mem_cgroup
*try_get_mem_cgroup_from_page(struct page
*page
)
234 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
235 struct mem_cgroup
*memcg
)
240 static inline bool task_in_mem_cgroup(struct task_struct
*task
,
241 const struct mem_cgroup
*memcg
)
246 static inline struct cgroup_subsys_state
247 *mem_cgroup_css(struct mem_cgroup
*memcg
)
252 static inline struct mem_cgroup
*
253 mem_cgroup_iter(struct mem_cgroup
*root
,
254 struct mem_cgroup
*prev
,
255 struct mem_cgroup_reclaim_cookie
*reclaim
)
260 static inline void mem_cgroup_iter_break(struct mem_cgroup
*root
,
261 struct mem_cgroup
*prev
)
265 static inline bool mem_cgroup_disabled(void)
271 mem_cgroup_inactive_anon_is_low(struct lruvec
*lruvec
)
276 static inline unsigned long
277 mem_cgroup_get_lru_size(struct lruvec
*lruvec
, enum lru_list lru
)
283 mem_cgroup_update_lru_size(struct lruvec
*lruvec
, enum lru_list lru
,
289 mem_cgroup_print_oom_info(struct mem_cgroup
*memcg
, struct task_struct
*p
)
293 static inline struct mem_cgroup
*mem_cgroup_begin_page_stat(struct page
*page
,
294 bool *locked
, unsigned long *flags
)
299 static inline void mem_cgroup_end_page_stat(struct mem_cgroup
*memcg
,
300 bool *locked
, unsigned long *flags
)
304 static inline void mem_cgroup_oom_enable(void)
308 static inline void mem_cgroup_oom_disable(void)
312 static inline bool task_in_memcg_oom(struct task_struct
*p
)
317 static inline bool mem_cgroup_oom_synchronize(bool wait
)
322 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup
*memcg
,
323 enum mem_cgroup_stat_index idx
)
327 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup
*memcg
,
328 enum mem_cgroup_stat_index idx
)
333 unsigned long mem_cgroup_soft_limit_reclaim(struct zone
*zone
, int order
,
335 unsigned long *total_scanned
)
340 static inline void mem_cgroup_split_huge_fixup(struct page
*head
)
345 void mem_cgroup_count_vm_event(struct mm_struct
*mm
, enum vm_event_item idx
)
348 #endif /* CONFIG_MEMCG */
350 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
352 mem_cgroup_bad_page_check(struct page
*page
)
358 mem_cgroup_print_bad_page(struct page
*page
)
370 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
371 void sock_update_memcg(struct sock
*sk
);
372 void sock_release_memcg(struct sock
*sk
);
374 static inline void sock_update_memcg(struct sock
*sk
)
377 static inline void sock_release_memcg(struct sock
*sk
)
380 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
382 #ifdef CONFIG_MEMCG_KMEM
383 extern struct static_key memcg_kmem_enabled_key
;
385 extern int memcg_limited_groups_array_size
;
388 * Helper macro to loop through all memcg-specific caches. Callers must still
389 * check if the cache is valid (it is either valid or NULL).
390 * the slab_mutex must be held when looping through those caches
392 #define for_each_memcg_cache_index(_idx) \
393 for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
395 static inline bool memcg_kmem_enabled(void)
397 return static_key_false(&memcg_kmem_enabled_key
);
401 * In general, we'll do everything in our power to not incur in any overhead
402 * for non-memcg users for the kmem functions. Not even a function call, if we
405 * Therefore, we'll inline all those functions so that in the best case, we'll
406 * see that kmemcg is off for everybody and proceed quickly. If it is on,
407 * we'll still do most of the flag checking inline. We check a lot of
408 * conditions, but because they are pretty simple, they are expected to be
411 bool __memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
,
413 void __memcg_kmem_commit_charge(struct page
*page
,
414 struct mem_cgroup
*memcg
, int order
);
415 void __memcg_kmem_uncharge_pages(struct page
*page
, int order
);
417 int memcg_cache_id(struct mem_cgroup
*memcg
);
419 void memcg_update_array_size(int num_groups
);
422 __memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
);
424 int __memcg_charge_slab(struct kmem_cache
*cachep
, gfp_t gfp
, int order
);
425 void __memcg_uncharge_slab(struct kmem_cache
*cachep
, int order
);
427 int __memcg_cleanup_cache_params(struct kmem_cache
*s
);
430 * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
431 * @gfp: the gfp allocation flags.
432 * @memcg: a pointer to the memcg this was charged against.
433 * @order: allocation order.
435 * returns true if the memcg where the current task belongs can hold this
438 * We return true automatically if this allocation is not to be accounted to
442 memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
, int order
)
444 if (!memcg_kmem_enabled())
448 * __GFP_NOFAIL allocations will move on even if charging is not
449 * possible. Therefore we don't even try, and have this allocation
450 * unaccounted. We could in theory charge it forcibly, but we hope
451 * those allocations are rare, and won't be worth the trouble.
453 if (gfp
& __GFP_NOFAIL
)
455 if (in_interrupt() || (!current
->mm
) || (current
->flags
& PF_KTHREAD
))
458 /* If the test is dying, just let it go. */
459 if (unlikely(fatal_signal_pending(current
)))
462 return __memcg_kmem_newpage_charge(gfp
, memcg
, order
);
466 * memcg_kmem_uncharge_pages: uncharge pages from memcg
467 * @page: pointer to struct page being freed
468 * @order: allocation order.
470 * there is no need to specify memcg here, since it is embedded in page_cgroup
473 memcg_kmem_uncharge_pages(struct page
*page
, int order
)
475 if (memcg_kmem_enabled())
476 __memcg_kmem_uncharge_pages(page
, order
);
480 * memcg_kmem_commit_charge: embeds correct memcg in a page
481 * @page: pointer to struct page recently allocated
482 * @memcg: the memcg structure we charged against
483 * @order: allocation order.
485 * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
486 * failure of the allocation. if @page is NULL, this function will revert the
487 * charges. Otherwise, it will commit the memcg given by @memcg to the
488 * corresponding page_cgroup.
491 memcg_kmem_commit_charge(struct page
*page
, struct mem_cgroup
*memcg
, int order
)
493 if (memcg_kmem_enabled() && memcg
)
494 __memcg_kmem_commit_charge(page
, memcg
, order
);
498 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
499 * @cachep: the original global kmem cache
500 * @gfp: allocation flags.
502 * All memory allocated from a per-memcg cache is charged to the owner memcg.
504 static __always_inline
struct kmem_cache
*
505 memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
)
507 if (!memcg_kmem_enabled())
509 if (gfp
& __GFP_NOFAIL
)
511 if (in_interrupt() || (!current
->mm
) || (current
->flags
& PF_KTHREAD
))
513 if (unlikely(fatal_signal_pending(current
)))
516 return __memcg_kmem_get_cache(cachep
, gfp
);
519 #define for_each_memcg_cache_index(_idx) \
522 static inline bool memcg_kmem_enabled(void)
528 memcg_kmem_newpage_charge(gfp_t gfp
, struct mem_cgroup
**memcg
, int order
)
533 static inline void memcg_kmem_uncharge_pages(struct page
*page
, int order
)
538 memcg_kmem_commit_charge(struct page
*page
, struct mem_cgroup
*memcg
, int order
)
542 static inline int memcg_cache_id(struct mem_cgroup
*memcg
)
547 static inline struct kmem_cache
*
548 memcg_kmem_get_cache(struct kmem_cache
*cachep
, gfp_t gfp
)
552 #endif /* CONFIG_MEMCG_KMEM */
553 #endif /* _LINUX_MEMCONTROL_H */