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>
26 #include <linux/page_counter.h>
27 #include <linux/vmpressure.h>
28 #include <linux/eventfd.h>
30 #include <linux/vmstat.h>
31 #include <linux/writeback.h>
32 #include <linux/page-flags.h>
39 /* Cgroup-specific page state, on top of universal node page state */
40 enum memcg_stat_item {
41 MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
46 /* XXX: why are these zone and not node counters? */
47 MEMCG_KERNEL_STACK_KB,
51 enum memcg_memory_event {
58 MEMCG_NR_MEMORY_EVENTS,
61 struct mem_cgroup_reclaim_cookie {
64 unsigned int generation;
69 #define MEM_CGROUP_ID_SHIFT 16
70 #define MEM_CGROUP_ID_MAX USHRT_MAX
72 struct mem_cgroup_id {
78 * Per memcg event counter is incremented at every pagein/pageout. With THP,
79 * it will be incremated by the number of pages. This counter is used for
80 * for trigger some periodic events. This is straightforward and better
81 * than using jiffies etc. to handle periodic memcg event.
83 enum mem_cgroup_events_target {
84 MEM_CGROUP_TARGET_THRESH,
85 MEM_CGROUP_TARGET_SOFTLIMIT,
86 MEM_CGROUP_TARGET_NUMAINFO,
90 struct mem_cgroup_stat_cpu {
91 long count[MEMCG_NR_STAT];
92 unsigned long events[NR_VM_EVENT_ITEMS];
93 unsigned long nr_page_events;
94 unsigned long targets[MEM_CGROUP_NTARGETS];
97 struct mem_cgroup_reclaim_iter {
98 struct mem_cgroup *position;
99 /* scan generation, increased every round-trip */
100 unsigned int generation;
104 long count[NR_VM_NODE_STAT_ITEMS];
108 * per-zone information in memory controller.
110 struct mem_cgroup_per_node {
111 struct lruvec lruvec;
113 struct lruvec_stat __percpu *lruvec_stat_cpu;
114 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
116 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
118 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
120 struct rb_node tree_node; /* RB tree node */
121 unsigned long usage_in_excess;/* Set to the value by which */
122 /* the soft limit is exceeded*/
124 bool congested; /* memcg has many dirty pages */
125 /* backed by a congested BDI */
127 struct mem_cgroup *memcg; /* Back pointer, we cannot */
128 /* use container_of */
131 struct mem_cgroup_threshold {
132 struct eventfd_ctx *eventfd;
133 unsigned long threshold;
137 struct mem_cgroup_threshold_ary {
138 /* An array index points to threshold just below or equal to usage. */
139 int current_threshold;
140 /* Size of entries[] */
142 /* Array of thresholds */
143 struct mem_cgroup_threshold entries[0];
146 struct mem_cgroup_thresholds {
147 /* Primary thresholds array */
148 struct mem_cgroup_threshold_ary *primary;
150 * Spare threshold array.
151 * This is needed to make mem_cgroup_unregister_event() "never fail".
152 * It must be able to store at least primary->size - 1 entries.
154 struct mem_cgroup_threshold_ary *spare;
157 enum memcg_kmem_state {
164 * The memory controller data structure. The memory controller controls both
165 * page cache and RSS per cgroup. We would eventually like to provide
166 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
167 * to help the administrator determine what knobs to tune.
170 struct cgroup_subsys_state css;
172 /* Private memcg ID. Used to ID objects that outlive the cgroup */
173 struct mem_cgroup_id id;
175 /* Accounted resources */
176 struct page_counter memory;
177 struct page_counter swap;
179 /* Legacy consumer-oriented counters */
180 struct page_counter memsw;
181 struct page_counter kmem;
182 struct page_counter tcpmem;
184 /* Upper bound of normal memory consumption range */
187 /* Range enforcement for interrupt charges */
188 struct work_struct high_work;
190 unsigned long soft_limit;
192 /* vmpressure notifications */
193 struct vmpressure vmpressure;
196 * Should the accounting and control be hierarchical, per subtree?
200 /* protected by memcg_oom_lock */
205 /* OOM-Killer disable */
206 int oom_kill_disable;
209 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
210 struct cgroup_file events_file;
212 /* handle for "memory.swap.events" */
213 struct cgroup_file swap_events_file;
215 /* protect arrays of thresholds */
216 struct mutex thresholds_lock;
218 /* thresholds for memory usage. RCU-protected */
219 struct mem_cgroup_thresholds thresholds;
221 /* thresholds for mem+swap usage. RCU-protected */
222 struct mem_cgroup_thresholds memsw_thresholds;
224 /* For oom notifier event fd */
225 struct list_head oom_notify;
228 * Should we move charges of a task when a task is moved into this
229 * mem_cgroup ? And what type of charges should we move ?
231 unsigned long move_charge_at_immigrate;
233 * set > 0 if pages under this cgroup are moving to other cgroup.
235 atomic_t moving_account;
236 /* taken only while moving_account > 0 */
237 spinlock_t move_lock;
238 struct task_struct *move_lock_task;
239 unsigned long move_lock_flags;
242 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
243 atomic_long_t stat[MEMCG_NR_STAT];
244 atomic_long_t events[NR_VM_EVENT_ITEMS];
246 unsigned long socket_pressure;
248 /* Legacy tcp memory accounting */
253 /* Index in the kmem_cache->memcg_params.memcg_caches array */
255 enum memcg_kmem_state kmem_state;
256 struct list_head kmem_caches;
259 int last_scanned_node;
261 nodemask_t scan_nodes;
262 atomic_t numainfo_events;
263 atomic_t numainfo_updating;
266 #ifdef CONFIG_CGROUP_WRITEBACK
267 struct list_head cgwb_list;
268 struct wb_domain cgwb_domain;
271 /* List of events which userspace want to receive */
272 struct list_head event_list;
273 spinlock_t event_list_lock;
275 struct mem_cgroup_per_node *nodeinfo[0];
276 /* WARNING: nodeinfo must be the last member here */
280 * size of first charge trial. "32" comes from vmscan.c's magic value.
281 * TODO: maybe necessary to use big numbers in big irons.
283 #define MEMCG_CHARGE_BATCH 32U
285 extern struct mem_cgroup *root_mem_cgroup;
287 static inline bool mem_cgroup_disabled(void)
289 return !cgroup_subsys_enabled(memory_cgrp_subsys);
292 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
294 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
295 gfp_t gfp_mask, struct mem_cgroup **memcgp,
297 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
298 bool lrucare, bool compound);
299 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
301 void mem_cgroup_uncharge(struct page *page);
302 void mem_cgroup_uncharge_list(struct list_head *page_list);
304 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
306 static struct mem_cgroup_per_node *
307 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
309 return memcg->nodeinfo[nid];
313 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
314 * @node: node of the wanted lruvec
315 * @memcg: memcg of the wanted lruvec
317 * Returns the lru list vector holding pages for a given @node or a given
318 * @memcg and @zone. This can be the node lruvec, if the memory controller
321 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
322 struct mem_cgroup *memcg)
324 struct mem_cgroup_per_node *mz;
325 struct lruvec *lruvec;
327 if (mem_cgroup_disabled()) {
328 lruvec = node_lruvec(pgdat);
332 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
333 lruvec = &mz->lruvec;
336 * Since a node can be onlined after the mem_cgroup was created,
337 * we have to be prepared to initialize lruvec->pgdat here;
338 * and if offlined then reonlined, we need to reinitialize it.
340 if (unlikely(lruvec->pgdat != pgdat))
341 lruvec->pgdat = pgdat;
345 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
347 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
348 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
351 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
352 return css ? container_of(css, struct mem_cgroup, css) : NULL;
355 #define mem_cgroup_from_counter(counter, member) \
356 container_of(counter, struct mem_cgroup, member)
358 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
360 struct mem_cgroup_reclaim_cookie *);
361 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
362 int mem_cgroup_scan_tasks(struct mem_cgroup *,
363 int (*)(struct task_struct *, void *), void *);
365 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
367 if (mem_cgroup_disabled())
372 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
374 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
376 struct mem_cgroup_per_node *mz;
378 if (mem_cgroup_disabled())
381 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
386 * parent_mem_cgroup - find the accounting parent of a memcg
387 * @memcg: memcg whose parent to find
389 * Returns the parent memcg, or NULL if this is the root or the memory
390 * controller is in legacy no-hierarchy mode.
392 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
394 if (!memcg->memory.parent)
396 return mem_cgroup_from_counter(memcg->memory.parent, memory);
399 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
400 struct mem_cgroup *root)
404 if (!root->use_hierarchy)
406 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
409 static inline bool mm_match_cgroup(struct mm_struct *mm,
410 struct mem_cgroup *memcg)
412 struct mem_cgroup *task_memcg;
416 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
418 match = mem_cgroup_is_descendant(task_memcg, memcg);
423 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
424 ino_t page_cgroup_ino(struct page *page);
426 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
428 if (mem_cgroup_disabled())
430 return !!(memcg->css.flags & CSS_ONLINE);
434 * For memory reclaim.
436 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
438 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
439 int zid, int nr_pages);
441 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
442 int nid, unsigned int lru_mask);
445 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
447 struct mem_cgroup_per_node *mz;
448 unsigned long nr_pages = 0;
451 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
452 for (zid = 0; zid < MAX_NR_ZONES; zid++)
453 nr_pages += mz->lru_zone_size[zid][lru];
458 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
459 enum lru_list lru, int zone_idx)
461 struct mem_cgroup_per_node *mz;
463 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
464 return mz->lru_zone_size[zone_idx][lru];
467 void mem_cgroup_handle_over_high(void);
469 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
471 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
472 struct task_struct *p);
474 static inline void mem_cgroup_oom_enable(void)
476 WARN_ON(current->memcg_may_oom);
477 current->memcg_may_oom = 1;
480 static inline void mem_cgroup_oom_disable(void)
482 WARN_ON(!current->memcg_may_oom);
483 current->memcg_may_oom = 0;
486 static inline bool task_in_memcg_oom(struct task_struct *p)
488 return p->memcg_in_oom;
491 bool mem_cgroup_oom_synchronize(bool wait);
493 #ifdef CONFIG_MEMCG_SWAP
494 extern int do_swap_account;
497 struct mem_cgroup *lock_page_memcg(struct page *page);
498 void __unlock_page_memcg(struct mem_cgroup *memcg);
499 void unlock_page_memcg(struct page *page);
501 /* idx can be of type enum memcg_stat_item or node_stat_item */
502 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
505 long x = atomic_long_read(&memcg->stat[idx]);
513 /* idx can be of type enum memcg_stat_item or node_stat_item */
514 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
519 if (mem_cgroup_disabled())
522 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
523 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
524 atomic_long_add(x, &memcg->stat[idx]);
527 __this_cpu_write(memcg->stat_cpu->count[idx], x);
530 /* idx can be of type enum memcg_stat_item or node_stat_item */
531 static inline void mod_memcg_state(struct mem_cgroup *memcg,
536 local_irq_save(flags);
537 __mod_memcg_state(memcg, idx, val);
538 local_irq_restore(flags);
542 * mod_memcg_page_state - update page state statistics
544 * @idx: page state item to account
545 * @val: number of pages (positive or negative)
547 * The @page must be locked or the caller must use lock_page_memcg()
548 * to prevent double accounting when the page is concurrently being
549 * moved to another memcg:
551 * lock_page(page) or lock_page_memcg(page)
552 * if (TestClearPageState(page))
553 * mod_memcg_page_state(page, state, -1);
554 * unlock_page(page) or unlock_page_memcg(page)
556 * Kernel pages are an exception to this, since they'll never move.
558 static inline void __mod_memcg_page_state(struct page *page,
561 if (page->mem_cgroup)
562 __mod_memcg_state(page->mem_cgroup, idx, val);
565 static inline void mod_memcg_page_state(struct page *page,
568 if (page->mem_cgroup)
569 mod_memcg_state(page->mem_cgroup, idx, val);
572 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
573 enum node_stat_item idx)
575 struct mem_cgroup_per_node *pn;
578 if (mem_cgroup_disabled())
579 return node_page_state(lruvec_pgdat(lruvec), idx);
581 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
582 x = atomic_long_read(&pn->lruvec_stat[idx]);
590 static inline void __mod_lruvec_state(struct lruvec *lruvec,
591 enum node_stat_item idx, int val)
593 struct mem_cgroup_per_node *pn;
597 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
599 if (mem_cgroup_disabled())
602 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
605 __mod_memcg_state(pn->memcg, idx, val);
608 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
609 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
610 atomic_long_add(x, &pn->lruvec_stat[idx]);
613 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
616 static inline void mod_lruvec_state(struct lruvec *lruvec,
617 enum node_stat_item idx, int val)
621 local_irq_save(flags);
622 __mod_lruvec_state(lruvec, idx, val);
623 local_irq_restore(flags);
626 static inline void __mod_lruvec_page_state(struct page *page,
627 enum node_stat_item idx, int val)
629 pg_data_t *pgdat = page_pgdat(page);
630 struct lruvec *lruvec;
632 /* Untracked pages have no memcg, no lruvec. Update only the node */
633 if (!page->mem_cgroup) {
634 __mod_node_page_state(pgdat, idx, val);
638 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
639 __mod_lruvec_state(lruvec, idx, val);
642 static inline void mod_lruvec_page_state(struct page *page,
643 enum node_stat_item idx, int val)
647 local_irq_save(flags);
648 __mod_lruvec_page_state(page, idx, val);
649 local_irq_restore(flags);
652 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
654 unsigned long *total_scanned);
656 static inline void __count_memcg_events(struct mem_cgroup *memcg,
657 enum vm_event_item idx,
662 if (mem_cgroup_disabled())
665 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
666 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
667 atomic_long_add(x, &memcg->events[idx]);
670 __this_cpu_write(memcg->stat_cpu->events[idx], x);
673 static inline void count_memcg_events(struct mem_cgroup *memcg,
674 enum vm_event_item idx,
679 local_irq_save(flags);
680 __count_memcg_events(memcg, idx, count);
681 local_irq_restore(flags);
684 static inline void count_memcg_page_event(struct page *page,
685 enum vm_event_item idx)
687 if (page->mem_cgroup)
688 count_memcg_events(page->mem_cgroup, idx, 1);
691 static inline void count_memcg_event_mm(struct mm_struct *mm,
692 enum vm_event_item idx)
694 struct mem_cgroup *memcg;
696 if (mem_cgroup_disabled())
700 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
702 count_memcg_events(memcg, idx, 1);
704 cgroup_file_notify(&memcg->events_file);
709 static inline void memcg_memory_event(struct mem_cgroup *memcg,
710 enum memcg_memory_event event)
712 atomic_long_inc(&memcg->memory_events[event]);
713 cgroup_file_notify(&memcg->events_file);
716 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
717 void mem_cgroup_split_huge_fixup(struct page *head);
720 #else /* CONFIG_MEMCG */
722 #define MEM_CGROUP_ID_SHIFT 0
723 #define MEM_CGROUP_ID_MAX 0
727 static inline bool mem_cgroup_disabled(void)
732 static inline void memcg_memory_event(struct mem_cgroup *memcg,
733 enum memcg_memory_event event)
737 static inline bool mem_cgroup_low(struct mem_cgroup *root,
738 struct mem_cgroup *memcg)
743 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
745 struct mem_cgroup **memcgp,
752 static inline void mem_cgroup_commit_charge(struct page *page,
753 struct mem_cgroup *memcg,
754 bool lrucare, bool compound)
758 static inline void mem_cgroup_cancel_charge(struct page *page,
759 struct mem_cgroup *memcg,
764 static inline void mem_cgroup_uncharge(struct page *page)
768 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
772 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
776 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
777 struct mem_cgroup *memcg)
779 return node_lruvec(pgdat);
782 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
783 struct pglist_data *pgdat)
785 return &pgdat->lruvec;
788 static inline bool mm_match_cgroup(struct mm_struct *mm,
789 struct mem_cgroup *memcg)
794 static inline bool task_in_mem_cgroup(struct task_struct *task,
795 const struct mem_cgroup *memcg)
800 static inline struct mem_cgroup *
801 mem_cgroup_iter(struct mem_cgroup *root,
802 struct mem_cgroup *prev,
803 struct mem_cgroup_reclaim_cookie *reclaim)
808 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
809 struct mem_cgroup *prev)
813 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
814 int (*fn)(struct task_struct *, void *), void *arg)
819 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
824 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
827 /* XXX: This should always return root_mem_cgroup */
831 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
836 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
841 static inline unsigned long
842 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
847 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
848 enum lru_list lru, int zone_idx)
853 static inline unsigned long
854 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
855 int nid, unsigned int lru_mask)
860 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
866 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
870 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
875 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
879 static inline void unlock_page_memcg(struct page *page)
883 static inline void mem_cgroup_handle_over_high(void)
887 static inline void mem_cgroup_oom_enable(void)
891 static inline void mem_cgroup_oom_disable(void)
895 static inline bool task_in_memcg_oom(struct task_struct *p)
900 static inline bool mem_cgroup_oom_synchronize(bool wait)
905 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
911 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
917 static inline void mod_memcg_state(struct mem_cgroup *memcg,
923 static inline void __mod_memcg_page_state(struct page *page,
929 static inline void mod_memcg_page_state(struct page *page,
935 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
936 enum node_stat_item idx)
938 return node_page_state(lruvec_pgdat(lruvec), idx);
941 static inline void __mod_lruvec_state(struct lruvec *lruvec,
942 enum node_stat_item idx, int val)
944 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
947 static inline void mod_lruvec_state(struct lruvec *lruvec,
948 enum node_stat_item idx, int val)
950 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
953 static inline void __mod_lruvec_page_state(struct page *page,
954 enum node_stat_item idx, int val)
956 __mod_node_page_state(page_pgdat(page), idx, val);
959 static inline void mod_lruvec_page_state(struct page *page,
960 enum node_stat_item idx, int val)
962 mod_node_page_state(page_pgdat(page), idx, val);
966 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
968 unsigned long *total_scanned)
973 static inline void mem_cgroup_split_huge_fixup(struct page *head)
977 static inline void count_memcg_events(struct mem_cgroup *memcg,
978 enum vm_event_item idx,
983 static inline void count_memcg_page_event(struct page *page,
989 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
992 #endif /* CONFIG_MEMCG */
994 /* idx can be of type enum memcg_stat_item or node_stat_item */
995 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
998 __mod_memcg_state(memcg, idx, 1);
1001 /* idx can be of type enum memcg_stat_item or node_stat_item */
1002 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1005 __mod_memcg_state(memcg, idx, -1);
1008 /* idx can be of type enum memcg_stat_item or node_stat_item */
1009 static inline void __inc_memcg_page_state(struct page *page,
1012 __mod_memcg_page_state(page, idx, 1);
1015 /* idx can be of type enum memcg_stat_item or node_stat_item */
1016 static inline void __dec_memcg_page_state(struct page *page,
1019 __mod_memcg_page_state(page, idx, -1);
1022 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1023 enum node_stat_item idx)
1025 __mod_lruvec_state(lruvec, idx, 1);
1028 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1029 enum node_stat_item idx)
1031 __mod_lruvec_state(lruvec, idx, -1);
1034 static inline void __inc_lruvec_page_state(struct page *page,
1035 enum node_stat_item idx)
1037 __mod_lruvec_page_state(page, idx, 1);
1040 static inline void __dec_lruvec_page_state(struct page *page,
1041 enum node_stat_item idx)
1043 __mod_lruvec_page_state(page, idx, -1);
1046 /* idx can be of type enum memcg_stat_item or node_stat_item */
1047 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1050 mod_memcg_state(memcg, idx, 1);
1053 /* idx can be of type enum memcg_stat_item or node_stat_item */
1054 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1057 mod_memcg_state(memcg, idx, -1);
1060 /* idx can be of type enum memcg_stat_item or node_stat_item */
1061 static inline void inc_memcg_page_state(struct page *page,
1064 mod_memcg_page_state(page, idx, 1);
1067 /* idx can be of type enum memcg_stat_item or node_stat_item */
1068 static inline void dec_memcg_page_state(struct page *page,
1071 mod_memcg_page_state(page, idx, -1);
1074 static inline void inc_lruvec_state(struct lruvec *lruvec,
1075 enum node_stat_item idx)
1077 mod_lruvec_state(lruvec, idx, 1);
1080 static inline void dec_lruvec_state(struct lruvec *lruvec,
1081 enum node_stat_item idx)
1083 mod_lruvec_state(lruvec, idx, -1);
1086 static inline void inc_lruvec_page_state(struct page *page,
1087 enum node_stat_item idx)
1089 mod_lruvec_page_state(page, idx, 1);
1092 static inline void dec_lruvec_page_state(struct page *page,
1093 enum node_stat_item idx)
1095 mod_lruvec_page_state(page, idx, -1);
1098 #ifdef CONFIG_CGROUP_WRITEBACK
1100 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
1101 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1102 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1103 unsigned long *pheadroom, unsigned long *pdirty,
1104 unsigned long *pwriteback);
1106 #else /* CONFIG_CGROUP_WRITEBACK */
1108 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1113 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1114 unsigned long *pfilepages,
1115 unsigned long *pheadroom,
1116 unsigned long *pdirty,
1117 unsigned long *pwriteback)
1121 #endif /* CONFIG_CGROUP_WRITEBACK */
1124 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1125 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1127 extern struct static_key_false memcg_sockets_enabled_key;
1128 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1129 void mem_cgroup_sk_alloc(struct sock *sk);
1130 void mem_cgroup_sk_free(struct sock *sk);
1131 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1133 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1136 if (time_before(jiffies, memcg->socket_pressure))
1138 } while ((memcg = parent_mem_cgroup(memcg)));
1142 #define mem_cgroup_sockets_enabled 0
1143 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1144 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1145 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1151 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1152 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1153 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1154 struct mem_cgroup *memcg);
1155 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1156 void memcg_kmem_uncharge(struct page *page, int order);
1158 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1159 extern struct static_key_false memcg_kmem_enabled_key;
1160 extern struct workqueue_struct *memcg_kmem_cache_wq;
1162 extern int memcg_nr_cache_ids;
1163 void memcg_get_cache_ids(void);
1164 void memcg_put_cache_ids(void);
1167 * Helper macro to loop through all memcg-specific caches. Callers must still
1168 * check if the cache is valid (it is either valid or NULL).
1169 * the slab_mutex must be held when looping through those caches
1171 #define for_each_memcg_cache_index(_idx) \
1172 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1174 static inline bool memcg_kmem_enabled(void)
1176 return static_branch_unlikely(&memcg_kmem_enabled_key);
1180 * helper for accessing a memcg's index. It will be used as an index in the
1181 * child cache array in kmem_cache, and also to derive its name. This function
1182 * will return -1 when this is not a kmem-limited memcg.
1184 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1186 return memcg ? memcg->kmemcg_id : -1;
1190 #define for_each_memcg_cache_index(_idx) \
1193 static inline bool memcg_kmem_enabled(void)
1198 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1203 static inline void memcg_get_cache_ids(void)
1207 static inline void memcg_put_cache_ids(void)
1211 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1213 #endif /* _LINUX_MEMCONTROL_H */