1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* memcontrol.h - Memory Controller
4 * Copyright IBM Corporation, 2007
7 * Copyright 2007 OpenVZ SWsoft Inc
11 #ifndef _LINUX_MEMCONTROL_H
12 #define _LINUX_MEMCONTROL_H
13 #include <linux/cgroup.h>
14 #include <linux/vm_event_item.h>
15 #include <linux/hardirq.h>
16 #include <linux/jump_label.h>
17 #include <linux/page_counter.h>
18 #include <linux/vmpressure.h>
19 #include <linux/eventfd.h>
21 #include <linux/vmstat.h>
22 #include <linux/writeback.h>
23 #include <linux/page-flags.h>
30 /* Cgroup-specific page state, on top of universal node page state */
31 enum memcg_stat_item {
32 MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
37 /* XXX: why are these zone and not node counters? */
38 MEMCG_KERNEL_STACK_KB,
42 enum memcg_memory_event {
50 MEMCG_NR_MEMORY_EVENTS,
53 enum mem_cgroup_protection {
59 struct mem_cgroup_reclaim_cookie {
62 unsigned int generation;
67 #define MEM_CGROUP_ID_SHIFT 16
68 #define MEM_CGROUP_ID_MAX USHRT_MAX
70 struct mem_cgroup_id {
76 * Per memcg event counter is incremented at every pagein/pageout. With THP,
77 * it will be incremated by the number of pages. This counter is used for
78 * for trigger some periodic events. This is straightforward and better
79 * than using jiffies etc. to handle periodic memcg event.
81 enum mem_cgroup_events_target {
82 MEM_CGROUP_TARGET_THRESH,
83 MEM_CGROUP_TARGET_SOFTLIMIT,
84 MEM_CGROUP_TARGET_NUMAINFO,
88 struct memcg_vmstats_percpu {
89 long stat[MEMCG_NR_STAT];
90 unsigned long events[NR_VM_EVENT_ITEMS];
91 unsigned long nr_page_events;
92 unsigned long targets[MEM_CGROUP_NTARGETS];
95 struct mem_cgroup_reclaim_iter {
96 struct mem_cgroup *position;
97 /* scan generation, increased every round-trip */
98 unsigned int generation;
102 long count[NR_VM_NODE_STAT_ITEMS];
106 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
107 * which have elements charged to this memcg.
109 struct memcg_shrinker_map {
111 unsigned long map[0];
115 * per-zone information in memory controller.
117 struct mem_cgroup_per_node {
118 struct lruvec lruvec;
120 /* Legacy local VM stats */
121 struct lruvec_stat __percpu *lruvec_stat_local;
123 /* Subtree VM stats (batched updates) */
124 struct lruvec_stat __percpu *lruvec_stat_cpu;
125 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
127 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
129 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
131 #ifdef CONFIG_MEMCG_KMEM
132 struct memcg_shrinker_map __rcu *shrinker_map;
134 struct rb_node tree_node; /* RB tree node */
135 unsigned long usage_in_excess;/* Set to the value by which */
136 /* the soft limit is exceeded*/
138 bool congested; /* memcg has many dirty pages */
139 /* backed by a congested BDI */
141 struct mem_cgroup *memcg; /* Back pointer, we cannot */
142 /* use container_of */
145 struct mem_cgroup_threshold {
146 struct eventfd_ctx *eventfd;
147 unsigned long threshold;
151 struct mem_cgroup_threshold_ary {
152 /* An array index points to threshold just below or equal to usage. */
153 int current_threshold;
154 /* Size of entries[] */
156 /* Array of thresholds */
157 struct mem_cgroup_threshold entries[0];
160 struct mem_cgroup_thresholds {
161 /* Primary thresholds array */
162 struct mem_cgroup_threshold_ary *primary;
164 * Spare threshold array.
165 * This is needed to make mem_cgroup_unregister_event() "never fail".
166 * It must be able to store at least primary->size - 1 entries.
168 struct mem_cgroup_threshold_ary *spare;
171 enum memcg_kmem_state {
177 #if defined(CONFIG_SMP)
178 struct memcg_padding {
180 } ____cacheline_internodealigned_in_smp;
181 #define MEMCG_PADDING(name) struct memcg_padding name;
183 #define MEMCG_PADDING(name)
187 * The memory controller data structure. The memory controller controls both
188 * page cache and RSS per cgroup. We would eventually like to provide
189 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
190 * to help the administrator determine what knobs to tune.
193 struct cgroup_subsys_state css;
195 /* Private memcg ID. Used to ID objects that outlive the cgroup */
196 struct mem_cgroup_id id;
198 /* Accounted resources */
199 struct page_counter memory;
200 struct page_counter swap;
202 /* Legacy consumer-oriented counters */
203 struct page_counter memsw;
204 struct page_counter kmem;
205 struct page_counter tcpmem;
207 /* Upper bound of normal memory consumption range */
210 /* Range enforcement for interrupt charges */
211 struct work_struct high_work;
213 unsigned long soft_limit;
215 /* vmpressure notifications */
216 struct vmpressure vmpressure;
219 * Should the accounting and control be hierarchical, per subtree?
224 * Should the OOM killer kill all belonging tasks, had it kill one?
228 /* protected by memcg_oom_lock */
233 /* OOM-Killer disable */
234 int oom_kill_disable;
236 /* memory.events and memory.events.local */
237 struct cgroup_file events_file;
238 struct cgroup_file events_local_file;
240 /* handle for "memory.swap.events" */
241 struct cgroup_file swap_events_file;
243 /* protect arrays of thresholds */
244 struct mutex thresholds_lock;
246 /* thresholds for memory usage. RCU-protected */
247 struct mem_cgroup_thresholds thresholds;
249 /* thresholds for mem+swap usage. RCU-protected */
250 struct mem_cgroup_thresholds memsw_thresholds;
252 /* For oom notifier event fd */
253 struct list_head oom_notify;
256 * Should we move charges of a task when a task is moved into this
257 * mem_cgroup ? And what type of charges should we move ?
259 unsigned long move_charge_at_immigrate;
260 /* taken only while moving_account > 0 */
261 spinlock_t move_lock;
262 unsigned long move_lock_flags;
264 MEMCG_PADDING(_pad1_);
267 * set > 0 if pages under this cgroup are moving to other cgroup.
269 atomic_t moving_account;
270 struct task_struct *move_lock_task;
272 /* Legacy local VM stats and events */
273 struct memcg_vmstats_percpu __percpu *vmstats_local;
275 /* Subtree VM stats and events (batched updates) */
276 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
278 MEMCG_PADDING(_pad2_);
280 atomic_long_t vmstats[MEMCG_NR_STAT];
281 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
284 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
285 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
287 unsigned long socket_pressure;
289 /* Legacy tcp memory accounting */
293 #ifdef CONFIG_MEMCG_KMEM
294 /* Index in the kmem_cache->memcg_params.memcg_caches array */
296 enum memcg_kmem_state kmem_state;
297 struct list_head kmem_caches;
300 int last_scanned_node;
302 nodemask_t scan_nodes;
303 atomic_t numainfo_events;
304 atomic_t numainfo_updating;
307 #ifdef CONFIG_CGROUP_WRITEBACK
308 struct list_head cgwb_list;
309 struct wb_domain cgwb_domain;
312 /* List of events which userspace want to receive */
313 struct list_head event_list;
314 spinlock_t event_list_lock;
316 struct mem_cgroup_per_node *nodeinfo[0];
317 /* WARNING: nodeinfo must be the last member here */
321 * size of first charge trial. "32" comes from vmscan.c's magic value.
322 * TODO: maybe necessary to use big numbers in big irons.
324 #define MEMCG_CHARGE_BATCH 32U
326 extern struct mem_cgroup *root_mem_cgroup;
328 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
330 return (memcg == root_mem_cgroup);
333 static inline bool mem_cgroup_disabled(void)
335 return !cgroup_subsys_enabled(memory_cgrp_subsys);
338 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
339 struct mem_cgroup *memcg);
341 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
342 gfp_t gfp_mask, struct mem_cgroup **memcgp,
344 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
345 gfp_t gfp_mask, struct mem_cgroup **memcgp,
347 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
348 bool lrucare, bool compound);
349 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
351 void mem_cgroup_uncharge(struct page *page);
352 void mem_cgroup_uncharge_list(struct list_head *page_list);
354 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
356 static struct mem_cgroup_per_node *
357 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
359 return memcg->nodeinfo[nid];
363 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
364 * @node: node of the wanted lruvec
365 * @memcg: memcg of the wanted lruvec
367 * Returns the lru list vector holding pages for a given @node or a given
368 * @memcg and @zone. This can be the node lruvec, if the memory controller
371 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
372 struct mem_cgroup *memcg)
374 struct mem_cgroup_per_node *mz;
375 struct lruvec *lruvec;
377 if (mem_cgroup_disabled()) {
378 lruvec = node_lruvec(pgdat);
382 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
383 lruvec = &mz->lruvec;
386 * Since a node can be onlined after the mem_cgroup was created,
387 * we have to be prepared to initialize lruvec->pgdat here;
388 * and if offlined then reonlined, we need to reinitialize it.
390 if (unlikely(lruvec->pgdat != pgdat))
391 lruvec->pgdat = pgdat;
395 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
397 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
399 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
401 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
404 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
405 return css ? container_of(css, struct mem_cgroup, css) : NULL;
408 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
411 css_put(&memcg->css);
414 #define mem_cgroup_from_counter(counter, member) \
415 container_of(counter, struct mem_cgroup, member)
417 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
419 struct mem_cgroup_reclaim_cookie *);
420 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
421 int mem_cgroup_scan_tasks(struct mem_cgroup *,
422 int (*)(struct task_struct *, void *), void *);
424 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
426 if (mem_cgroup_disabled())
431 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
433 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
435 return mem_cgroup_from_css(seq_css(m));
438 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
440 struct mem_cgroup_per_node *mz;
442 if (mem_cgroup_disabled())
445 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
450 * parent_mem_cgroup - find the accounting parent of a memcg
451 * @memcg: memcg whose parent to find
453 * Returns the parent memcg, or NULL if this is the root or the memory
454 * controller is in legacy no-hierarchy mode.
456 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
458 if (!memcg->memory.parent)
460 return mem_cgroup_from_counter(memcg->memory.parent, memory);
463 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
464 struct mem_cgroup *root)
468 if (!root->use_hierarchy)
470 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
473 static inline bool mm_match_cgroup(struct mm_struct *mm,
474 struct mem_cgroup *memcg)
476 struct mem_cgroup *task_memcg;
480 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
482 match = mem_cgroup_is_descendant(task_memcg, memcg);
487 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
488 ino_t page_cgroup_ino(struct page *page);
490 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
492 if (mem_cgroup_disabled())
494 return !!(memcg->css.flags & CSS_ONLINE);
498 * For memory reclaim.
500 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
502 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
503 int zid, int nr_pages);
506 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
507 enum lru_list lru, int zone_idx)
509 struct mem_cgroup_per_node *mz;
511 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
512 return mz->lru_zone_size[zone_idx][lru];
515 void mem_cgroup_handle_over_high(void);
517 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
519 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
520 struct task_struct *p);
522 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
524 static inline void mem_cgroup_enter_user_fault(void)
526 WARN_ON(current->in_user_fault);
527 current->in_user_fault = 1;
530 static inline void mem_cgroup_exit_user_fault(void)
532 WARN_ON(!current->in_user_fault);
533 current->in_user_fault = 0;
536 static inline bool task_in_memcg_oom(struct task_struct *p)
538 return p->memcg_in_oom;
541 bool mem_cgroup_oom_synchronize(bool wait);
542 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
543 struct mem_cgroup *oom_domain);
544 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
546 #ifdef CONFIG_MEMCG_SWAP
547 extern int do_swap_account;
550 struct mem_cgroup *lock_page_memcg(struct page *page);
551 void __unlock_page_memcg(struct mem_cgroup *memcg);
552 void unlock_page_memcg(struct page *page);
555 * idx can be of type enum memcg_stat_item or node_stat_item.
556 * Keep in sync with memcg_exact_page_state().
558 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
560 long x = atomic_long_read(&memcg->vmstats[idx]);
569 * idx can be of type enum memcg_stat_item or node_stat_item.
570 * Keep in sync with memcg_exact_page_state().
572 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
578 for_each_possible_cpu(cpu)
579 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
587 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
589 /* idx can be of type enum memcg_stat_item or node_stat_item */
590 static inline void mod_memcg_state(struct mem_cgroup *memcg,
595 local_irq_save(flags);
596 __mod_memcg_state(memcg, idx, val);
597 local_irq_restore(flags);
601 * mod_memcg_page_state - update page state statistics
603 * @idx: page state item to account
604 * @val: number of pages (positive or negative)
606 * The @page must be locked or the caller must use lock_page_memcg()
607 * to prevent double accounting when the page is concurrently being
608 * moved to another memcg:
610 * lock_page(page) or lock_page_memcg(page)
611 * if (TestClearPageState(page))
612 * mod_memcg_page_state(page, state, -1);
613 * unlock_page(page) or unlock_page_memcg(page)
615 * Kernel pages are an exception to this, since they'll never move.
617 static inline void __mod_memcg_page_state(struct page *page,
620 if (page->mem_cgroup)
621 __mod_memcg_state(page->mem_cgroup, idx, val);
624 static inline void mod_memcg_page_state(struct page *page,
627 if (page->mem_cgroup)
628 mod_memcg_state(page->mem_cgroup, idx, val);
631 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
632 enum node_stat_item idx)
634 struct mem_cgroup_per_node *pn;
637 if (mem_cgroup_disabled())
638 return node_page_state(lruvec_pgdat(lruvec), idx);
640 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
641 x = atomic_long_read(&pn->lruvec_stat[idx]);
649 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
650 enum node_stat_item idx)
652 struct mem_cgroup_per_node *pn;
656 if (mem_cgroup_disabled())
657 return node_page_state(lruvec_pgdat(lruvec), idx);
659 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
660 for_each_possible_cpu(cpu)
661 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
669 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
671 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
673 static inline void mod_lruvec_state(struct lruvec *lruvec,
674 enum node_stat_item idx, int val)
678 local_irq_save(flags);
679 __mod_lruvec_state(lruvec, idx, val);
680 local_irq_restore(flags);
683 static inline void __mod_lruvec_page_state(struct page *page,
684 enum node_stat_item idx, int val)
686 pg_data_t *pgdat = page_pgdat(page);
687 struct lruvec *lruvec;
689 /* Untracked pages have no memcg, no lruvec. Update only the node */
690 if (!page->mem_cgroup) {
691 __mod_node_page_state(pgdat, idx, val);
695 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
696 __mod_lruvec_state(lruvec, idx, val);
699 static inline void mod_lruvec_page_state(struct page *page,
700 enum node_stat_item idx, int val)
704 local_irq_save(flags);
705 __mod_lruvec_page_state(page, idx, val);
706 local_irq_restore(flags);
709 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
711 unsigned long *total_scanned);
713 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
714 unsigned long count);
716 static inline void count_memcg_events(struct mem_cgroup *memcg,
717 enum vm_event_item idx,
722 local_irq_save(flags);
723 __count_memcg_events(memcg, idx, count);
724 local_irq_restore(flags);
727 static inline void count_memcg_page_event(struct page *page,
728 enum vm_event_item idx)
730 if (page->mem_cgroup)
731 count_memcg_events(page->mem_cgroup, idx, 1);
734 static inline void count_memcg_event_mm(struct mm_struct *mm,
735 enum vm_event_item idx)
737 struct mem_cgroup *memcg;
739 if (mem_cgroup_disabled())
743 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
745 count_memcg_events(memcg, idx, 1);
749 static inline void memcg_memory_event(struct mem_cgroup *memcg,
750 enum memcg_memory_event event)
752 atomic_long_inc(&memcg->memory_events_local[event]);
753 cgroup_file_notify(&memcg->events_local_file);
756 atomic_long_inc(&memcg->memory_events[event]);
757 cgroup_file_notify(&memcg->events_file);
759 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
761 } while ((memcg = parent_mem_cgroup(memcg)) &&
762 !mem_cgroup_is_root(memcg));
765 static inline void memcg_memory_event_mm(struct mm_struct *mm,
766 enum memcg_memory_event event)
768 struct mem_cgroup *memcg;
770 if (mem_cgroup_disabled())
774 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
776 memcg_memory_event(memcg, event);
780 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
781 void mem_cgroup_split_huge_fixup(struct page *head);
784 #else /* CONFIG_MEMCG */
786 #define MEM_CGROUP_ID_SHIFT 0
787 #define MEM_CGROUP_ID_MAX 0
791 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
796 static inline bool mem_cgroup_disabled(void)
801 static inline void memcg_memory_event(struct mem_cgroup *memcg,
802 enum memcg_memory_event event)
806 static inline void memcg_memory_event_mm(struct mm_struct *mm,
807 enum memcg_memory_event event)
811 static inline enum mem_cgroup_protection mem_cgroup_protected(
812 struct mem_cgroup *root, struct mem_cgroup *memcg)
814 return MEMCG_PROT_NONE;
817 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
819 struct mem_cgroup **memcgp,
826 static inline int mem_cgroup_try_charge_delay(struct page *page,
827 struct mm_struct *mm,
829 struct mem_cgroup **memcgp,
836 static inline void mem_cgroup_commit_charge(struct page *page,
837 struct mem_cgroup *memcg,
838 bool lrucare, bool compound)
842 static inline void mem_cgroup_cancel_charge(struct page *page,
843 struct mem_cgroup *memcg,
848 static inline void mem_cgroup_uncharge(struct page *page)
852 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
856 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
860 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
861 struct mem_cgroup *memcg)
863 return node_lruvec(pgdat);
866 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
867 struct pglist_data *pgdat)
869 return &pgdat->lruvec;
872 static inline bool mm_match_cgroup(struct mm_struct *mm,
873 struct mem_cgroup *memcg)
878 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
883 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
888 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
892 static inline struct mem_cgroup *
893 mem_cgroup_iter(struct mem_cgroup *root,
894 struct mem_cgroup *prev,
895 struct mem_cgroup_reclaim_cookie *reclaim)
900 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
901 struct mem_cgroup *prev)
905 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
906 int (*fn)(struct task_struct *, void *), void *arg)
911 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
916 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
919 /* XXX: This should always return root_mem_cgroup */
923 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
928 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
933 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
939 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
940 enum lru_list lru, int zone_idx)
945 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
951 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
956 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
960 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
965 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
969 static inline void unlock_page_memcg(struct page *page)
973 static inline void mem_cgroup_handle_over_high(void)
977 static inline void mem_cgroup_enter_user_fault(void)
981 static inline void mem_cgroup_exit_user_fault(void)
985 static inline bool task_in_memcg_oom(struct task_struct *p)
990 static inline bool mem_cgroup_oom_synchronize(bool wait)
995 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
996 struct task_struct *victim, struct mem_cgroup *oom_domain)
1001 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1005 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1010 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1016 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1022 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1028 static inline void __mod_memcg_page_state(struct page *page,
1034 static inline void mod_memcg_page_state(struct page *page,
1040 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1041 enum node_stat_item idx)
1043 return node_page_state(lruvec_pgdat(lruvec), idx);
1046 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1047 enum node_stat_item idx)
1049 return node_page_state(lruvec_pgdat(lruvec), idx);
1052 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1053 enum node_stat_item idx, int val)
1055 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1058 static inline void mod_lruvec_state(struct lruvec *lruvec,
1059 enum node_stat_item idx, int val)
1061 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1064 static inline void __mod_lruvec_page_state(struct page *page,
1065 enum node_stat_item idx, int val)
1067 __mod_node_page_state(page_pgdat(page), idx, val);
1070 static inline void mod_lruvec_page_state(struct page *page,
1071 enum node_stat_item idx, int val)
1073 mod_node_page_state(page_pgdat(page), idx, val);
1076 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1079 struct page *page = virt_to_head_page(p);
1081 __mod_node_page_state(page_pgdat(page), idx, val);
1085 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1087 unsigned long *total_scanned)
1092 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1096 static inline void count_memcg_events(struct mem_cgroup *memcg,
1097 enum vm_event_item idx,
1098 unsigned long count)
1102 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1103 enum vm_event_item idx,
1104 unsigned long count)
1108 static inline void count_memcg_page_event(struct page *page,
1114 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1117 #endif /* CONFIG_MEMCG */
1119 /* idx can be of type enum memcg_stat_item or node_stat_item */
1120 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1123 __mod_memcg_state(memcg, idx, 1);
1126 /* idx can be of type enum memcg_stat_item or node_stat_item */
1127 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1130 __mod_memcg_state(memcg, idx, -1);
1133 /* idx can be of type enum memcg_stat_item or node_stat_item */
1134 static inline void __inc_memcg_page_state(struct page *page,
1137 __mod_memcg_page_state(page, idx, 1);
1140 /* idx can be of type enum memcg_stat_item or node_stat_item */
1141 static inline void __dec_memcg_page_state(struct page *page,
1144 __mod_memcg_page_state(page, idx, -1);
1147 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1148 enum node_stat_item idx)
1150 __mod_lruvec_state(lruvec, idx, 1);
1153 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1154 enum node_stat_item idx)
1156 __mod_lruvec_state(lruvec, idx, -1);
1159 static inline void __inc_lruvec_page_state(struct page *page,
1160 enum node_stat_item idx)
1162 __mod_lruvec_page_state(page, idx, 1);
1165 static inline void __dec_lruvec_page_state(struct page *page,
1166 enum node_stat_item idx)
1168 __mod_lruvec_page_state(page, idx, -1);
1171 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1173 __mod_lruvec_slab_state(p, idx, 1);
1176 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1178 __mod_lruvec_slab_state(p, idx, -1);
1181 /* idx can be of type enum memcg_stat_item or node_stat_item */
1182 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1185 mod_memcg_state(memcg, idx, 1);
1188 /* idx can be of type enum memcg_stat_item or node_stat_item */
1189 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1192 mod_memcg_state(memcg, idx, -1);
1195 /* idx can be of type enum memcg_stat_item or node_stat_item */
1196 static inline void inc_memcg_page_state(struct page *page,
1199 mod_memcg_page_state(page, idx, 1);
1202 /* idx can be of type enum memcg_stat_item or node_stat_item */
1203 static inline void dec_memcg_page_state(struct page *page,
1206 mod_memcg_page_state(page, idx, -1);
1209 static inline void inc_lruvec_state(struct lruvec *lruvec,
1210 enum node_stat_item idx)
1212 mod_lruvec_state(lruvec, idx, 1);
1215 static inline void dec_lruvec_state(struct lruvec *lruvec,
1216 enum node_stat_item idx)
1218 mod_lruvec_state(lruvec, idx, -1);
1221 static inline void inc_lruvec_page_state(struct page *page,
1222 enum node_stat_item idx)
1224 mod_lruvec_page_state(page, idx, 1);
1227 static inline void dec_lruvec_page_state(struct page *page,
1228 enum node_stat_item idx)
1230 mod_lruvec_page_state(page, idx, -1);
1233 #ifdef CONFIG_CGROUP_WRITEBACK
1235 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1236 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1237 unsigned long *pheadroom, unsigned long *pdirty,
1238 unsigned long *pwriteback);
1240 #else /* CONFIG_CGROUP_WRITEBACK */
1242 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1247 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1248 unsigned long *pfilepages,
1249 unsigned long *pheadroom,
1250 unsigned long *pdirty,
1251 unsigned long *pwriteback)
1255 #endif /* CONFIG_CGROUP_WRITEBACK */
1258 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1259 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1261 extern struct static_key_false memcg_sockets_enabled_key;
1262 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1263 void mem_cgroup_sk_alloc(struct sock *sk);
1264 void mem_cgroup_sk_free(struct sock *sk);
1265 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1267 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1270 if (time_before(jiffies, memcg->socket_pressure))
1272 } while ((memcg = parent_mem_cgroup(memcg)));
1276 #define mem_cgroup_sockets_enabled 0
1277 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1278 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1279 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1285 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1286 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1288 #ifdef CONFIG_MEMCG_KMEM
1289 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1290 void __memcg_kmem_uncharge(struct page *page, int order);
1291 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1292 struct mem_cgroup *memcg);
1293 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1294 unsigned int nr_pages);
1296 extern struct static_key_false memcg_kmem_enabled_key;
1297 extern struct workqueue_struct *memcg_kmem_cache_wq;
1299 extern int memcg_nr_cache_ids;
1300 void memcg_get_cache_ids(void);
1301 void memcg_put_cache_ids(void);
1304 * Helper macro to loop through all memcg-specific caches. Callers must still
1305 * check if the cache is valid (it is either valid or NULL).
1306 * the slab_mutex must be held when looping through those caches
1308 #define for_each_memcg_cache_index(_idx) \
1309 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1311 static inline bool memcg_kmem_enabled(void)
1313 return static_branch_unlikely(&memcg_kmem_enabled_key);
1316 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1318 if (memcg_kmem_enabled())
1319 return __memcg_kmem_charge(page, gfp, order);
1323 static inline void memcg_kmem_uncharge(struct page *page, int order)
1325 if (memcg_kmem_enabled())
1326 __memcg_kmem_uncharge(page, order);
1329 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1330 int order, struct mem_cgroup *memcg)
1332 if (memcg_kmem_enabled())
1333 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1337 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1338 struct mem_cgroup *memcg)
1340 if (memcg_kmem_enabled())
1341 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1345 * helper for accessing a memcg's index. It will be used as an index in the
1346 * child cache array in kmem_cache, and also to derive its name. This function
1347 * will return -1 when this is not a kmem-limited memcg.
1349 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1351 return memcg ? memcg->kmemcg_id : -1;
1354 extern int memcg_expand_shrinker_maps(int new_id);
1356 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1357 int nid, int shrinker_id);
1360 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1365 static inline void memcg_kmem_uncharge(struct page *page, int order)
1369 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1374 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1378 #define for_each_memcg_cache_index(_idx) \
1381 static inline bool memcg_kmem_enabled(void)
1386 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1391 static inline void memcg_get_cache_ids(void)
1395 static inline void memcg_put_cache_ids(void)
1399 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1400 int nid, int shrinker_id) { }
1401 #endif /* CONFIG_MEMCG_KMEM */
1403 #endif /* _LINUX_MEMCONTROL_H */