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 /* Cgroup-specific events, on top of universal VM events */
52 enum memcg_event_item {
53 MEMCG_LOW = NR_VM_EVENT_ITEMS,
60 struct mem_cgroup_reclaim_cookie {
63 unsigned int generation;
68 #define MEM_CGROUP_ID_SHIFT 16
69 #define MEM_CGROUP_ID_MAX USHRT_MAX
71 struct mem_cgroup_id {
77 * Per memcg event counter is incremented at every pagein/pageout. With THP,
78 * it will be incremated by the number of pages. This counter is used for
79 * for trigger some periodic events. This is straightforward and better
80 * than using jiffies etc. to handle periodic memcg event.
82 enum mem_cgroup_events_target {
83 MEM_CGROUP_TARGET_THRESH,
84 MEM_CGROUP_TARGET_SOFTLIMIT,
85 MEM_CGROUP_TARGET_NUMAINFO,
89 struct mem_cgroup_stat_cpu {
90 long count[MEMCG_NR_STAT];
91 unsigned long events[MEMCG_NR_EVENTS];
92 unsigned long nr_page_events;
93 unsigned long targets[MEM_CGROUP_NTARGETS];
96 struct mem_cgroup_reclaim_iter {
97 struct mem_cgroup *position;
98 /* scan generation, increased every round-trip */
99 unsigned int generation;
103 long count[NR_VM_NODE_STAT_ITEMS];
107 * per-zone information in memory controller.
109 struct mem_cgroup_per_node {
110 struct lruvec lruvec;
112 struct lruvec_stat __percpu *lruvec_stat_cpu;
113 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
115 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
117 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
119 struct rb_node tree_node; /* RB tree node */
120 unsigned long usage_in_excess;/* Set to the value by which */
121 /* the soft limit is exceeded*/
123 bool congested; /* memcg has many dirty pages */
124 /* backed by a congested BDI */
126 struct mem_cgroup *memcg; /* Back pointer, we cannot */
127 /* use container_of */
130 struct mem_cgroup_threshold {
131 struct eventfd_ctx *eventfd;
132 unsigned long threshold;
136 struct mem_cgroup_threshold_ary {
137 /* An array index points to threshold just below or equal to usage. */
138 int current_threshold;
139 /* Size of entries[] */
141 /* Array of thresholds */
142 struct mem_cgroup_threshold entries[0];
145 struct mem_cgroup_thresholds {
146 /* Primary thresholds array */
147 struct mem_cgroup_threshold_ary *primary;
149 * Spare threshold array.
150 * This is needed to make mem_cgroup_unregister_event() "never fail".
151 * It must be able to store at least primary->size - 1 entries.
153 struct mem_cgroup_threshold_ary *spare;
156 enum memcg_kmem_state {
163 * The memory controller data structure. The memory controller controls both
164 * page cache and RSS per cgroup. We would eventually like to provide
165 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
166 * to help the administrator determine what knobs to tune.
169 struct cgroup_subsys_state css;
171 /* Private memcg ID. Used to ID objects that outlive the cgroup */
172 struct mem_cgroup_id id;
174 /* Accounted resources */
175 struct page_counter memory;
176 struct page_counter swap;
178 /* Legacy consumer-oriented counters */
179 struct page_counter memsw;
180 struct page_counter kmem;
181 struct page_counter tcpmem;
183 /* 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;
208 /* handle for "memory.events" */
209 struct cgroup_file events_file;
211 /* protect arrays of thresholds */
212 struct mutex thresholds_lock;
214 /* thresholds for memory usage. RCU-protected */
215 struct mem_cgroup_thresholds thresholds;
217 /* thresholds for mem+swap usage. RCU-protected */
218 struct mem_cgroup_thresholds memsw_thresholds;
220 /* For oom notifier event fd */
221 struct list_head oom_notify;
224 * Should we move charges of a task when a task is moved into this
225 * mem_cgroup ? And what type of charges should we move ?
227 unsigned long move_charge_at_immigrate;
229 * set > 0 if pages under this cgroup are moving to other cgroup.
231 atomic_t moving_account;
232 /* taken only while moving_account > 0 */
233 spinlock_t move_lock;
234 struct task_struct *move_lock_task;
235 unsigned long move_lock_flags;
237 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
238 atomic_long_t stat[MEMCG_NR_STAT];
239 atomic_long_t events[MEMCG_NR_EVENTS];
241 unsigned long socket_pressure;
243 /* Legacy tcp memory accounting */
248 /* Index in the kmem_cache->memcg_params.memcg_caches array */
250 enum memcg_kmem_state kmem_state;
251 struct list_head kmem_caches;
254 int last_scanned_node;
256 nodemask_t scan_nodes;
257 atomic_t numainfo_events;
258 atomic_t numainfo_updating;
261 #ifdef CONFIG_CGROUP_WRITEBACK
262 struct list_head cgwb_list;
263 struct wb_domain cgwb_domain;
266 /* List of events which userspace want to receive */
267 struct list_head event_list;
268 spinlock_t event_list_lock;
270 struct mem_cgroup_per_node *nodeinfo[0];
271 /* WARNING: nodeinfo must be the last member here */
275 * size of first charge trial. "32" comes from vmscan.c's magic value.
276 * TODO: maybe necessary to use big numbers in big irons.
278 #define MEMCG_CHARGE_BATCH 32U
280 extern struct mem_cgroup *root_mem_cgroup;
282 static inline bool mem_cgroup_disabled(void)
284 return !cgroup_subsys_enabled(memory_cgrp_subsys);
287 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
289 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
290 gfp_t gfp_mask, struct mem_cgroup **memcgp,
292 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
293 bool lrucare, bool compound);
294 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
296 void mem_cgroup_uncharge(struct page *page);
297 void mem_cgroup_uncharge_list(struct list_head *page_list);
299 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
301 static struct mem_cgroup_per_node *
302 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
304 return memcg->nodeinfo[nid];
308 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
309 * @node: node of the wanted lruvec
310 * @memcg: memcg of the wanted lruvec
312 * Returns the lru list vector holding pages for a given @node or a given
313 * @memcg and @zone. This can be the node lruvec, if the memory controller
316 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
317 struct mem_cgroup *memcg)
319 struct mem_cgroup_per_node *mz;
320 struct lruvec *lruvec;
322 if (mem_cgroup_disabled()) {
323 lruvec = node_lruvec(pgdat);
327 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
328 lruvec = &mz->lruvec;
331 * Since a node can be onlined after the mem_cgroup was created,
332 * we have to be prepared to initialize lruvec->pgdat here;
333 * and if offlined then reonlined, we need to reinitialize it.
335 if (unlikely(lruvec->pgdat != pgdat))
336 lruvec->pgdat = pgdat;
340 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
342 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
343 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
346 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
347 return css ? container_of(css, struct mem_cgroup, css) : NULL;
350 #define mem_cgroup_from_counter(counter, member) \
351 container_of(counter, struct mem_cgroup, member)
353 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
355 struct mem_cgroup_reclaim_cookie *);
356 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
357 int mem_cgroup_scan_tasks(struct mem_cgroup *,
358 int (*)(struct task_struct *, void *), void *);
360 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
362 if (mem_cgroup_disabled())
367 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
369 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
371 struct mem_cgroup_per_node *mz;
373 if (mem_cgroup_disabled())
376 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
381 * parent_mem_cgroup - find the accounting parent of a memcg
382 * @memcg: memcg whose parent to find
384 * Returns the parent memcg, or NULL if this is the root or the memory
385 * controller is in legacy no-hierarchy mode.
387 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
389 if (!memcg->memory.parent)
391 return mem_cgroup_from_counter(memcg->memory.parent, memory);
394 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
395 struct mem_cgroup *root)
399 if (!root->use_hierarchy)
401 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
404 static inline bool mm_match_cgroup(struct mm_struct *mm,
405 struct mem_cgroup *memcg)
407 struct mem_cgroup *task_memcg;
411 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
413 match = mem_cgroup_is_descendant(task_memcg, memcg);
418 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
419 ino_t page_cgroup_ino(struct page *page);
421 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
423 if (mem_cgroup_disabled())
425 return !!(memcg->css.flags & CSS_ONLINE);
429 * For memory reclaim.
431 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
433 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
434 int zid, int nr_pages);
436 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
437 int nid, unsigned int lru_mask);
440 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
442 struct mem_cgroup_per_node *mz;
443 unsigned long nr_pages = 0;
446 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
447 for (zid = 0; zid < MAX_NR_ZONES; zid++)
448 nr_pages += mz->lru_zone_size[zid][lru];
453 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
454 enum lru_list lru, int zone_idx)
456 struct mem_cgroup_per_node *mz;
458 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
459 return mz->lru_zone_size[zone_idx][lru];
462 void mem_cgroup_handle_over_high(void);
464 unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg);
466 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
467 struct task_struct *p);
469 static inline void mem_cgroup_oom_enable(void)
471 WARN_ON(current->memcg_may_oom);
472 current->memcg_may_oom = 1;
475 static inline void mem_cgroup_oom_disable(void)
477 WARN_ON(!current->memcg_may_oom);
478 current->memcg_may_oom = 0;
481 static inline bool task_in_memcg_oom(struct task_struct *p)
483 return p->memcg_in_oom;
486 bool mem_cgroup_oom_synchronize(bool wait);
488 #ifdef CONFIG_MEMCG_SWAP
489 extern int do_swap_account;
492 struct mem_cgroup *lock_page_memcg(struct page *page);
493 void __unlock_page_memcg(struct mem_cgroup *memcg);
494 void unlock_page_memcg(struct page *page);
496 /* idx can be of type enum memcg_stat_item or node_stat_item */
497 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
500 long x = atomic_long_read(&memcg->stat[idx]);
508 /* idx can be of type enum memcg_stat_item or node_stat_item */
509 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
514 if (mem_cgroup_disabled())
517 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
518 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
519 atomic_long_add(x, &memcg->stat[idx]);
522 __this_cpu_write(memcg->stat_cpu->count[idx], x);
525 /* idx can be of type enum memcg_stat_item or node_stat_item */
526 static inline void mod_memcg_state(struct mem_cgroup *memcg,
531 local_irq_save(flags);
532 __mod_memcg_state(memcg, idx, val);
533 local_irq_restore(flags);
537 * mod_memcg_page_state - update page state statistics
539 * @idx: page state item to account
540 * @val: number of pages (positive or negative)
542 * The @page must be locked or the caller must use lock_page_memcg()
543 * to prevent double accounting when the page is concurrently being
544 * moved to another memcg:
546 * lock_page(page) or lock_page_memcg(page)
547 * if (TestClearPageState(page))
548 * mod_memcg_page_state(page, state, -1);
549 * unlock_page(page) or unlock_page_memcg(page)
551 * Kernel pages are an exception to this, since they'll never move.
553 static inline void __mod_memcg_page_state(struct page *page,
556 if (page->mem_cgroup)
557 __mod_memcg_state(page->mem_cgroup, idx, val);
560 static inline void mod_memcg_page_state(struct page *page,
563 if (page->mem_cgroup)
564 mod_memcg_state(page->mem_cgroup, idx, val);
567 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
568 enum node_stat_item idx)
570 struct mem_cgroup_per_node *pn;
573 if (mem_cgroup_disabled())
574 return node_page_state(lruvec_pgdat(lruvec), idx);
576 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
577 x = atomic_long_read(&pn->lruvec_stat[idx]);
585 static inline void __mod_lruvec_state(struct lruvec *lruvec,
586 enum node_stat_item idx, int val)
588 struct mem_cgroup_per_node *pn;
592 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
594 if (mem_cgroup_disabled())
597 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
600 __mod_memcg_state(pn->memcg, idx, val);
603 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
604 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
605 atomic_long_add(x, &pn->lruvec_stat[idx]);
608 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
611 static inline void mod_lruvec_state(struct lruvec *lruvec,
612 enum node_stat_item idx, int val)
616 local_irq_save(flags);
617 __mod_lruvec_state(lruvec, idx, val);
618 local_irq_restore(flags);
621 static inline void __mod_lruvec_page_state(struct page *page,
622 enum node_stat_item idx, int val)
624 pg_data_t *pgdat = page_pgdat(page);
625 struct lruvec *lruvec;
627 /* Untracked pages have no memcg, no lruvec. Update only the node */
628 if (!page->mem_cgroup) {
629 __mod_node_page_state(pgdat, idx, val);
633 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
634 __mod_lruvec_state(lruvec, idx, val);
637 static inline void mod_lruvec_page_state(struct page *page,
638 enum node_stat_item idx, int val)
642 local_irq_save(flags);
643 __mod_lruvec_page_state(page, idx, val);
644 local_irq_restore(flags);
647 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
649 unsigned long *total_scanned);
651 /* idx can be of type enum memcg_event_item or vm_event_item */
652 static inline void __count_memcg_events(struct mem_cgroup *memcg,
653 int idx, unsigned long count)
657 if (mem_cgroup_disabled())
660 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
661 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
662 atomic_long_add(x, &memcg->events[idx]);
665 __this_cpu_write(memcg->stat_cpu->events[idx], x);
668 static inline void count_memcg_events(struct mem_cgroup *memcg,
669 int idx, unsigned long count)
673 local_irq_save(flags);
674 __count_memcg_events(memcg, idx, count);
675 local_irq_restore(flags);
678 /* idx can be of type enum memcg_event_item or vm_event_item */
679 static inline void count_memcg_page_event(struct page *page,
682 if (page->mem_cgroup)
683 count_memcg_events(page->mem_cgroup, idx, 1);
686 static inline void count_memcg_event_mm(struct mm_struct *mm,
687 enum vm_event_item idx)
689 struct mem_cgroup *memcg;
691 if (mem_cgroup_disabled())
695 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
697 count_memcg_events(memcg, idx, 1);
699 cgroup_file_notify(&memcg->events_file);
704 static inline void mem_cgroup_event(struct mem_cgroup *memcg,
705 enum memcg_event_item event)
707 count_memcg_events(memcg, event, 1);
708 cgroup_file_notify(&memcg->events_file);
711 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
712 void mem_cgroup_split_huge_fixup(struct page *head);
715 #else /* CONFIG_MEMCG */
717 #define MEM_CGROUP_ID_SHIFT 0
718 #define MEM_CGROUP_ID_MAX 0
722 static inline bool mem_cgroup_disabled(void)
727 static inline void mem_cgroup_event(struct mem_cgroup *memcg,
728 enum memcg_event_item event)
732 static inline bool mem_cgroup_low(struct mem_cgroup *root,
733 struct mem_cgroup *memcg)
738 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
740 struct mem_cgroup **memcgp,
747 static inline void mem_cgroup_commit_charge(struct page *page,
748 struct mem_cgroup *memcg,
749 bool lrucare, bool compound)
753 static inline void mem_cgroup_cancel_charge(struct page *page,
754 struct mem_cgroup *memcg,
759 static inline void mem_cgroup_uncharge(struct page *page)
763 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
767 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
771 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
772 struct mem_cgroup *memcg)
774 return node_lruvec(pgdat);
777 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
778 struct pglist_data *pgdat)
780 return &pgdat->lruvec;
783 static inline bool mm_match_cgroup(struct mm_struct *mm,
784 struct mem_cgroup *memcg)
789 static inline bool task_in_mem_cgroup(struct task_struct *task,
790 const struct mem_cgroup *memcg)
795 static inline struct mem_cgroup *
796 mem_cgroup_iter(struct mem_cgroup *root,
797 struct mem_cgroup *prev,
798 struct mem_cgroup_reclaim_cookie *reclaim)
803 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
804 struct mem_cgroup *prev)
808 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
809 int (*fn)(struct task_struct *, void *), void *arg)
814 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
819 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
822 /* XXX: This should always return root_mem_cgroup */
826 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
831 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
836 static inline unsigned long
837 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
842 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
843 enum lru_list lru, int zone_idx)
848 static inline unsigned long
849 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
850 int nid, unsigned int lru_mask)
855 static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
861 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
865 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
870 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
874 static inline void unlock_page_memcg(struct page *page)
878 static inline void mem_cgroup_handle_over_high(void)
882 static inline void mem_cgroup_oom_enable(void)
886 static inline void mem_cgroup_oom_disable(void)
890 static inline bool task_in_memcg_oom(struct task_struct *p)
895 static inline bool mem_cgroup_oom_synchronize(bool wait)
900 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
906 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
912 static inline void mod_memcg_state(struct mem_cgroup *memcg,
918 static inline void __mod_memcg_page_state(struct page *page,
924 static inline void mod_memcg_page_state(struct page *page,
930 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
931 enum node_stat_item idx)
933 return node_page_state(lruvec_pgdat(lruvec), idx);
936 static inline void __mod_lruvec_state(struct lruvec *lruvec,
937 enum node_stat_item idx, int val)
939 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
942 static inline void mod_lruvec_state(struct lruvec *lruvec,
943 enum node_stat_item idx, int val)
945 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
948 static inline void __mod_lruvec_page_state(struct page *page,
949 enum node_stat_item idx, int val)
951 __mod_node_page_state(page_pgdat(page), idx, val);
954 static inline void mod_lruvec_page_state(struct page *page,
955 enum node_stat_item idx, int val)
957 mod_node_page_state(page_pgdat(page), idx, val);
961 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
963 unsigned long *total_scanned)
968 static inline void mem_cgroup_split_huge_fixup(struct page *head)
972 static inline void count_memcg_events(struct mem_cgroup *memcg,
973 enum vm_event_item idx,
978 static inline void count_memcg_page_event(struct page *page,
984 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
987 #endif /* CONFIG_MEMCG */
989 /* idx can be of type enum memcg_stat_item or node_stat_item */
990 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
993 __mod_memcg_state(memcg, idx, 1);
996 /* idx can be of type enum memcg_stat_item or node_stat_item */
997 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1000 __mod_memcg_state(memcg, idx, -1);
1003 /* idx can be of type enum memcg_stat_item or node_stat_item */
1004 static inline void __inc_memcg_page_state(struct page *page,
1007 __mod_memcg_page_state(page, idx, 1);
1010 /* idx can be of type enum memcg_stat_item or node_stat_item */
1011 static inline void __dec_memcg_page_state(struct page *page,
1014 __mod_memcg_page_state(page, idx, -1);
1017 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1018 enum node_stat_item idx)
1020 __mod_lruvec_state(lruvec, idx, 1);
1023 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1024 enum node_stat_item idx)
1026 __mod_lruvec_state(lruvec, idx, -1);
1029 static inline void __inc_lruvec_page_state(struct page *page,
1030 enum node_stat_item idx)
1032 __mod_lruvec_page_state(page, idx, 1);
1035 static inline void __dec_lruvec_page_state(struct page *page,
1036 enum node_stat_item idx)
1038 __mod_lruvec_page_state(page, idx, -1);
1041 /* idx can be of type enum memcg_stat_item or node_stat_item */
1042 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1045 mod_memcg_state(memcg, idx, 1);
1048 /* idx can be of type enum memcg_stat_item or node_stat_item */
1049 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1052 mod_memcg_state(memcg, idx, -1);
1055 /* idx can be of type enum memcg_stat_item or node_stat_item */
1056 static inline void inc_memcg_page_state(struct page *page,
1059 mod_memcg_page_state(page, idx, 1);
1062 /* idx can be of type enum memcg_stat_item or node_stat_item */
1063 static inline void dec_memcg_page_state(struct page *page,
1066 mod_memcg_page_state(page, idx, -1);
1069 static inline void inc_lruvec_state(struct lruvec *lruvec,
1070 enum node_stat_item idx)
1072 mod_lruvec_state(lruvec, idx, 1);
1075 static inline void dec_lruvec_state(struct lruvec *lruvec,
1076 enum node_stat_item idx)
1078 mod_lruvec_state(lruvec, idx, -1);
1081 static inline void inc_lruvec_page_state(struct page *page,
1082 enum node_stat_item idx)
1084 mod_lruvec_page_state(page, idx, 1);
1087 static inline void dec_lruvec_page_state(struct page *page,
1088 enum node_stat_item idx)
1090 mod_lruvec_page_state(page, idx, -1);
1093 #ifdef CONFIG_CGROUP_WRITEBACK
1095 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
1096 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1097 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1098 unsigned long *pheadroom, unsigned long *pdirty,
1099 unsigned long *pwriteback);
1101 #else /* CONFIG_CGROUP_WRITEBACK */
1103 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1108 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1109 unsigned long *pfilepages,
1110 unsigned long *pheadroom,
1111 unsigned long *pdirty,
1112 unsigned long *pwriteback)
1116 #endif /* CONFIG_CGROUP_WRITEBACK */
1119 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1120 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1122 extern struct static_key_false memcg_sockets_enabled_key;
1123 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1124 void mem_cgroup_sk_alloc(struct sock *sk);
1125 void mem_cgroup_sk_free(struct sock *sk);
1126 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1128 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1131 if (time_before(jiffies, memcg->socket_pressure))
1133 } while ((memcg = parent_mem_cgroup(memcg)));
1137 #define mem_cgroup_sockets_enabled 0
1138 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1139 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1140 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1146 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1147 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1148 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1149 struct mem_cgroup *memcg);
1150 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1151 void memcg_kmem_uncharge(struct page *page, int order);
1153 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1154 extern struct static_key_false memcg_kmem_enabled_key;
1155 extern struct workqueue_struct *memcg_kmem_cache_wq;
1157 extern int memcg_nr_cache_ids;
1158 void memcg_get_cache_ids(void);
1159 void memcg_put_cache_ids(void);
1162 * Helper macro to loop through all memcg-specific caches. Callers must still
1163 * check if the cache is valid (it is either valid or NULL).
1164 * the slab_mutex must be held when looping through those caches
1166 #define for_each_memcg_cache_index(_idx) \
1167 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1169 static inline bool memcg_kmem_enabled(void)
1171 return static_branch_unlikely(&memcg_kmem_enabled_key);
1175 * helper for accessing a memcg's index. It will be used as an index in the
1176 * child cache array in kmem_cache, and also to derive its name. This function
1177 * will return -1 when this is not a kmem-limited memcg.
1179 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1181 return memcg ? memcg->kmemcg_id : -1;
1185 #define for_each_memcg_cache_index(_idx) \
1188 static inline bool memcg_kmem_enabled(void)
1193 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1198 static inline void memcg_get_cache_ids(void)
1202 static inline void memcg_put_cache_ids(void)
1206 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1208 #endif /* _LINUX_MEMCONTROL_H */