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_SWAP = NR_VM_NODE_STAT_ITEMS,
34 /* XXX: why are these zone and not node counters? */
35 MEMCG_KERNEL_STACK_KB,
39 enum memcg_memory_event {
48 MEMCG_NR_MEMORY_EVENTS,
51 enum mem_cgroup_protection {
57 struct mem_cgroup_reclaim_cookie {
59 unsigned int generation;
64 #define MEM_CGROUP_ID_SHIFT 16
65 #define MEM_CGROUP_ID_MAX USHRT_MAX
67 struct mem_cgroup_id {
73 * Per memcg event counter is incremented at every pagein/pageout. With THP,
74 * it will be incremated by the number of pages. This counter is used for
75 * for trigger some periodic events. This is straightforward and better
76 * than using jiffies etc. to handle periodic memcg event.
78 enum mem_cgroup_events_target {
79 MEM_CGROUP_TARGET_THRESH,
80 MEM_CGROUP_TARGET_SOFTLIMIT,
84 struct memcg_vmstats_percpu {
85 long stat[MEMCG_NR_STAT];
86 unsigned long events[NR_VM_EVENT_ITEMS];
87 unsigned long nr_page_events;
88 unsigned long targets[MEM_CGROUP_NTARGETS];
91 struct mem_cgroup_reclaim_iter {
92 struct mem_cgroup *position;
93 /* scan generation, increased every round-trip */
94 unsigned int generation;
98 long count[NR_VM_NODE_STAT_ITEMS];
102 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
103 * which have elements charged to this memcg.
105 struct memcg_shrinker_map {
111 * per-node information in memory controller.
113 struct mem_cgroup_per_node {
114 struct lruvec lruvec;
116 /* Legacy local VM stats */
117 struct lruvec_stat __percpu *lruvec_stat_local;
119 /* Subtree VM stats (batched updates) */
120 struct lruvec_stat __percpu *lruvec_stat_cpu;
121 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
123 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
125 struct mem_cgroup_reclaim_iter iter;
127 struct memcg_shrinker_map __rcu *shrinker_map;
129 struct rb_node tree_node; /* RB tree node */
130 unsigned long usage_in_excess;/* Set to the value by which */
131 /* the soft limit is exceeded*/
133 struct mem_cgroup *memcg; /* Back pointer, we cannot */
134 /* use container_of */
137 struct mem_cgroup_threshold {
138 struct eventfd_ctx *eventfd;
139 unsigned long threshold;
143 struct mem_cgroup_threshold_ary {
144 /* An array index points to threshold just below or equal to usage. */
145 int current_threshold;
146 /* Size of entries[] */
148 /* Array of thresholds */
149 struct mem_cgroup_threshold entries[];
152 struct mem_cgroup_thresholds {
153 /* Primary thresholds array */
154 struct mem_cgroup_threshold_ary *primary;
156 * Spare threshold array.
157 * This is needed to make mem_cgroup_unregister_event() "never fail".
158 * It must be able to store at least primary->size - 1 entries.
160 struct mem_cgroup_threshold_ary *spare;
163 enum memcg_kmem_state {
169 #if defined(CONFIG_SMP)
170 struct memcg_padding {
172 } ____cacheline_internodealigned_in_smp;
173 #define MEMCG_PADDING(name) struct memcg_padding name;
175 #define MEMCG_PADDING(name)
179 * Remember four most recent foreign writebacks with dirty pages in this
180 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
181 * one in a given round, we're likely to catch it later if it keeps
182 * foreign-dirtying, so a fairly low count should be enough.
184 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
186 #define MEMCG_CGWB_FRN_CNT 4
188 struct memcg_cgwb_frn {
189 u64 bdi_id; /* bdi->id of the foreign inode */
190 int memcg_id; /* memcg->css.id of foreign inode */
191 u64 at; /* jiffies_64 at the time of dirtying */
192 struct wb_completion done; /* tracks in-flight foreign writebacks */
196 * The memory controller data structure. The memory controller controls both
197 * page cache and RSS per cgroup. We would eventually like to provide
198 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
199 * to help the administrator determine what knobs to tune.
202 struct cgroup_subsys_state css;
204 /* Private memcg ID. Used to ID objects that outlive the cgroup */
205 struct mem_cgroup_id id;
207 /* Accounted resources */
208 struct page_counter memory;
209 struct page_counter swap;
211 /* Legacy consumer-oriented counters */
212 struct page_counter memsw;
213 struct page_counter kmem;
214 struct page_counter tcpmem;
216 /* Range enforcement for interrupt charges */
217 struct work_struct high_work;
219 unsigned long soft_limit;
221 /* vmpressure notifications */
222 struct vmpressure vmpressure;
225 * Should the accounting and control be hierarchical, per subtree?
230 * Should the OOM killer kill all belonging tasks, had it kill one?
234 /* protected by memcg_oom_lock */
239 /* OOM-Killer disable */
240 int oom_kill_disable;
242 /* memory.events and memory.events.local */
243 struct cgroup_file events_file;
244 struct cgroup_file events_local_file;
246 /* handle for "memory.swap.events" */
247 struct cgroup_file swap_events_file;
249 /* protect arrays of thresholds */
250 struct mutex thresholds_lock;
252 /* thresholds for memory usage. RCU-protected */
253 struct mem_cgroup_thresholds thresholds;
255 /* thresholds for mem+swap usage. RCU-protected */
256 struct mem_cgroup_thresholds memsw_thresholds;
258 /* For oom notifier event fd */
259 struct list_head oom_notify;
262 * Should we move charges of a task when a task is moved into this
263 * mem_cgroup ? And what type of charges should we move ?
265 unsigned long move_charge_at_immigrate;
266 /* taken only while moving_account > 0 */
267 spinlock_t move_lock;
268 unsigned long move_lock_flags;
270 MEMCG_PADDING(_pad1_);
273 * set > 0 if pages under this cgroup are moving to other cgroup.
275 atomic_t moving_account;
276 struct task_struct *move_lock_task;
278 /* Legacy local VM stats and events */
279 struct memcg_vmstats_percpu __percpu *vmstats_local;
281 /* Subtree VM stats and events (batched updates) */
282 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
284 MEMCG_PADDING(_pad2_);
286 atomic_long_t vmstats[MEMCG_NR_STAT];
287 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
290 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
291 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
293 unsigned long socket_pressure;
295 /* Legacy tcp memory accounting */
299 #ifdef CONFIG_MEMCG_KMEM
300 /* Index in the kmem_cache->memcg_params.memcg_caches array */
302 enum memcg_kmem_state kmem_state;
303 struct list_head kmem_caches;
306 #ifdef CONFIG_CGROUP_WRITEBACK
307 struct list_head cgwb_list;
308 struct wb_domain cgwb_domain;
309 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
312 /* List of events which userspace want to receive */
313 struct list_head event_list;
314 spinlock_t event_list_lock;
316 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
317 struct deferred_split deferred_split_queue;
320 struct mem_cgroup_per_node *nodeinfo[0];
321 /* WARNING: nodeinfo must be the last member here */
325 * size of first charge trial. "32" comes from vmscan.c's magic value.
326 * TODO: maybe necessary to use big numbers in big irons.
328 #define MEMCG_CHARGE_BATCH 32U
330 extern struct mem_cgroup *root_mem_cgroup;
332 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
334 return (memcg == root_mem_cgroup);
337 static inline bool mem_cgroup_disabled(void)
339 return !cgroup_subsys_enabled(memory_cgrp_subsys);
342 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
345 if (mem_cgroup_disabled())
349 return READ_ONCE(memcg->memory.emin);
351 return max(READ_ONCE(memcg->memory.emin),
352 READ_ONCE(memcg->memory.elow));
355 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
356 struct mem_cgroup *memcg);
358 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask);
360 void mem_cgroup_uncharge(struct page *page);
361 void mem_cgroup_uncharge_list(struct list_head *page_list);
363 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
365 static struct mem_cgroup_per_node *
366 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
368 return memcg->nodeinfo[nid];
372 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
373 * @memcg: memcg of the wanted lruvec
375 * Returns the lru list vector holding pages for a given @memcg &
376 * @node combination. This can be the node lruvec, if the memory
377 * controller is disabled.
379 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
380 struct pglist_data *pgdat)
382 struct mem_cgroup_per_node *mz;
383 struct lruvec *lruvec;
385 if (mem_cgroup_disabled()) {
386 lruvec = &pgdat->__lruvec;
391 memcg = root_mem_cgroup;
393 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
394 lruvec = &mz->lruvec;
397 * Since a node can be onlined after the mem_cgroup was created,
398 * we have to be prepared to initialize lruvec->pgdat here;
399 * and if offlined then reonlined, we need to reinitialize it.
401 if (unlikely(lruvec->pgdat != pgdat))
402 lruvec->pgdat = pgdat;
406 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
408 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
410 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
412 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
415 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
416 return css ? container_of(css, struct mem_cgroup, css) : NULL;
419 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
422 css_put(&memcg->css);
425 #define mem_cgroup_from_counter(counter, member) \
426 container_of(counter, struct mem_cgroup, member)
428 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
430 struct mem_cgroup_reclaim_cookie *);
431 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
432 int mem_cgroup_scan_tasks(struct mem_cgroup *,
433 int (*)(struct task_struct *, void *), void *);
435 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
437 if (mem_cgroup_disabled())
442 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
444 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
446 return mem_cgroup_from_css(seq_css(m));
449 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
451 struct mem_cgroup_per_node *mz;
453 if (mem_cgroup_disabled())
456 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
461 * parent_mem_cgroup - find the accounting parent of a memcg
462 * @memcg: memcg whose parent to find
464 * Returns the parent memcg, or NULL if this is the root or the memory
465 * controller is in legacy no-hierarchy mode.
467 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
469 if (!memcg->memory.parent)
471 return mem_cgroup_from_counter(memcg->memory.parent, memory);
474 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
475 struct mem_cgroup *root)
479 if (!root->use_hierarchy)
481 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
484 static inline bool mm_match_cgroup(struct mm_struct *mm,
485 struct mem_cgroup *memcg)
487 struct mem_cgroup *task_memcg;
491 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
493 match = mem_cgroup_is_descendant(task_memcg, memcg);
498 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
499 ino_t page_cgroup_ino(struct page *page);
501 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
503 if (mem_cgroup_disabled())
505 return !!(memcg->css.flags & CSS_ONLINE);
509 * For memory reclaim.
511 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
513 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
514 int zid, int nr_pages);
517 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
518 enum lru_list lru, int zone_idx)
520 struct mem_cgroup_per_node *mz;
522 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
523 return mz->lru_zone_size[zone_idx][lru];
526 void mem_cgroup_handle_over_high(void);
528 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
530 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
532 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
533 struct task_struct *p);
535 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
537 static inline void mem_cgroup_enter_user_fault(void)
539 WARN_ON(current->in_user_fault);
540 current->in_user_fault = 1;
543 static inline void mem_cgroup_exit_user_fault(void)
545 WARN_ON(!current->in_user_fault);
546 current->in_user_fault = 0;
549 static inline bool task_in_memcg_oom(struct task_struct *p)
551 return p->memcg_in_oom;
554 bool mem_cgroup_oom_synchronize(bool wait);
555 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
556 struct mem_cgroup *oom_domain);
557 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
559 #ifdef CONFIG_MEMCG_SWAP
560 extern bool cgroup_memory_noswap;
563 struct mem_cgroup *lock_page_memcg(struct page *page);
564 void __unlock_page_memcg(struct mem_cgroup *memcg);
565 void unlock_page_memcg(struct page *page);
568 * idx can be of type enum memcg_stat_item or node_stat_item.
569 * Keep in sync with memcg_exact_page_state().
571 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
573 long x = atomic_long_read(&memcg->vmstats[idx]);
582 * idx can be of type enum memcg_stat_item or node_stat_item.
583 * Keep in sync with memcg_exact_page_state().
585 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
591 for_each_possible_cpu(cpu)
592 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
600 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
602 /* idx can be of type enum memcg_stat_item or node_stat_item */
603 static inline void mod_memcg_state(struct mem_cgroup *memcg,
608 local_irq_save(flags);
609 __mod_memcg_state(memcg, idx, val);
610 local_irq_restore(flags);
614 * mod_memcg_page_state - update page state statistics
616 * @idx: page state item to account
617 * @val: number of pages (positive or negative)
619 * The @page must be locked or the caller must use lock_page_memcg()
620 * to prevent double accounting when the page is concurrently being
621 * moved to another memcg:
623 * lock_page(page) or lock_page_memcg(page)
624 * if (TestClearPageState(page))
625 * mod_memcg_page_state(page, state, -1);
626 * unlock_page(page) or unlock_page_memcg(page)
628 * Kernel pages are an exception to this, since they'll never move.
630 static inline void __mod_memcg_page_state(struct page *page,
633 if (page->mem_cgroup)
634 __mod_memcg_state(page->mem_cgroup, idx, val);
637 static inline void mod_memcg_page_state(struct page *page,
640 if (page->mem_cgroup)
641 mod_memcg_state(page->mem_cgroup, idx, val);
644 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
645 enum node_stat_item idx)
647 struct mem_cgroup_per_node *pn;
650 if (mem_cgroup_disabled())
651 return node_page_state(lruvec_pgdat(lruvec), idx);
653 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
654 x = atomic_long_read(&pn->lruvec_stat[idx]);
662 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
663 enum node_stat_item idx)
665 struct mem_cgroup_per_node *pn;
669 if (mem_cgroup_disabled())
670 return node_page_state(lruvec_pgdat(lruvec), idx);
672 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
673 for_each_possible_cpu(cpu)
674 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
682 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
684 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
685 void mod_memcg_obj_state(void *p, int idx, int val);
687 static inline void mod_lruvec_state(struct lruvec *lruvec,
688 enum node_stat_item idx, int val)
692 local_irq_save(flags);
693 __mod_lruvec_state(lruvec, idx, val);
694 local_irq_restore(flags);
697 static inline void __mod_lruvec_page_state(struct page *page,
698 enum node_stat_item idx, int val)
700 struct page *head = compound_head(page); /* rmap on tail pages */
701 pg_data_t *pgdat = page_pgdat(page);
702 struct lruvec *lruvec;
704 /* Untracked pages have no memcg, no lruvec. Update only the node */
705 if (!head->mem_cgroup) {
706 __mod_node_page_state(pgdat, idx, val);
710 lruvec = mem_cgroup_lruvec(head->mem_cgroup, pgdat);
711 __mod_lruvec_state(lruvec, idx, val);
714 static inline void mod_lruvec_page_state(struct page *page,
715 enum node_stat_item idx, int val)
719 local_irq_save(flags);
720 __mod_lruvec_page_state(page, idx, val);
721 local_irq_restore(flags);
724 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
726 unsigned long *total_scanned);
728 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
729 unsigned long count);
731 static inline void count_memcg_events(struct mem_cgroup *memcg,
732 enum vm_event_item idx,
737 local_irq_save(flags);
738 __count_memcg_events(memcg, idx, count);
739 local_irq_restore(flags);
742 static inline void count_memcg_page_event(struct page *page,
743 enum vm_event_item idx)
745 if (page->mem_cgroup)
746 count_memcg_events(page->mem_cgroup, idx, 1);
749 static inline void count_memcg_event_mm(struct mm_struct *mm,
750 enum vm_event_item idx)
752 struct mem_cgroup *memcg;
754 if (mem_cgroup_disabled())
758 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
760 count_memcg_events(memcg, idx, 1);
764 static inline void memcg_memory_event(struct mem_cgroup *memcg,
765 enum memcg_memory_event event)
767 atomic_long_inc(&memcg->memory_events_local[event]);
768 cgroup_file_notify(&memcg->events_local_file);
771 atomic_long_inc(&memcg->memory_events[event]);
772 cgroup_file_notify(&memcg->events_file);
774 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
776 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
778 } while ((memcg = parent_mem_cgroup(memcg)) &&
779 !mem_cgroup_is_root(memcg));
782 static inline void memcg_memory_event_mm(struct mm_struct *mm,
783 enum memcg_memory_event event)
785 struct mem_cgroup *memcg;
787 if (mem_cgroup_disabled())
791 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
793 memcg_memory_event(memcg, event);
797 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
798 void mem_cgroup_split_huge_fixup(struct page *head);
801 #else /* CONFIG_MEMCG */
803 #define MEM_CGROUP_ID_SHIFT 0
804 #define MEM_CGROUP_ID_MAX 0
808 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
813 static inline bool mem_cgroup_disabled(void)
818 static inline void memcg_memory_event(struct mem_cgroup *memcg,
819 enum memcg_memory_event event)
823 static inline void memcg_memory_event_mm(struct mm_struct *mm,
824 enum memcg_memory_event event)
828 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
834 static inline enum mem_cgroup_protection mem_cgroup_protected(
835 struct mem_cgroup *root, struct mem_cgroup *memcg)
837 return MEMCG_PROT_NONE;
840 static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
846 static inline void mem_cgroup_uncharge(struct page *page)
850 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
854 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
858 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
859 struct pglist_data *pgdat)
861 return &pgdat->__lruvec;
864 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
865 struct pglist_data *pgdat)
867 return &pgdat->__lruvec;
870 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
875 static inline bool mm_match_cgroup(struct mm_struct *mm,
876 struct mem_cgroup *memcg)
881 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
886 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
891 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
895 static inline struct mem_cgroup *
896 mem_cgroup_iter(struct mem_cgroup *root,
897 struct mem_cgroup *prev,
898 struct mem_cgroup_reclaim_cookie *reclaim)
903 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
904 struct mem_cgroup *prev)
908 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
909 int (*fn)(struct task_struct *, void *), void *arg)
914 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
919 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
922 /* XXX: This should always return root_mem_cgroup */
926 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
931 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
936 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
942 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
943 enum lru_list lru, int zone_idx)
948 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
953 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
959 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
964 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
968 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
973 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
977 static inline void unlock_page_memcg(struct page *page)
981 static inline void mem_cgroup_handle_over_high(void)
985 static inline void mem_cgroup_enter_user_fault(void)
989 static inline void mem_cgroup_exit_user_fault(void)
993 static inline bool task_in_memcg_oom(struct task_struct *p)
998 static inline bool mem_cgroup_oom_synchronize(bool wait)
1003 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1004 struct task_struct *victim, struct mem_cgroup *oom_domain)
1009 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1013 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1018 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1024 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1030 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1036 static inline void __mod_memcg_page_state(struct page *page,
1042 static inline void mod_memcg_page_state(struct page *page,
1048 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1049 enum node_stat_item idx)
1051 return node_page_state(lruvec_pgdat(lruvec), idx);
1054 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1055 enum node_stat_item idx)
1057 return node_page_state(lruvec_pgdat(lruvec), idx);
1060 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1061 enum node_stat_item idx, int val)
1063 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1066 static inline void mod_lruvec_state(struct lruvec *lruvec,
1067 enum node_stat_item idx, int val)
1069 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1072 static inline void __mod_lruvec_page_state(struct page *page,
1073 enum node_stat_item idx, int val)
1075 __mod_node_page_state(page_pgdat(page), idx, val);
1078 static inline void mod_lruvec_page_state(struct page *page,
1079 enum node_stat_item idx, int val)
1081 mod_node_page_state(page_pgdat(page), idx, val);
1084 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1087 struct page *page = virt_to_head_page(p);
1089 __mod_node_page_state(page_pgdat(page), idx, val);
1092 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1097 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1099 unsigned long *total_scanned)
1104 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1108 static inline void count_memcg_events(struct mem_cgroup *memcg,
1109 enum vm_event_item idx,
1110 unsigned long count)
1114 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1115 enum vm_event_item idx,
1116 unsigned long count)
1120 static inline void count_memcg_page_event(struct page *page,
1126 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1129 #endif /* CONFIG_MEMCG */
1131 /* idx can be of type enum memcg_stat_item or node_stat_item */
1132 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1135 __mod_memcg_state(memcg, idx, 1);
1138 /* idx can be of type enum memcg_stat_item or node_stat_item */
1139 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1142 __mod_memcg_state(memcg, idx, -1);
1145 /* idx can be of type enum memcg_stat_item or node_stat_item */
1146 static inline void __inc_memcg_page_state(struct page *page,
1149 __mod_memcg_page_state(page, idx, 1);
1152 /* idx can be of type enum memcg_stat_item or node_stat_item */
1153 static inline void __dec_memcg_page_state(struct page *page,
1156 __mod_memcg_page_state(page, idx, -1);
1159 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1160 enum node_stat_item idx)
1162 __mod_lruvec_state(lruvec, idx, 1);
1165 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1166 enum node_stat_item idx)
1168 __mod_lruvec_state(lruvec, idx, -1);
1171 static inline void __inc_lruvec_page_state(struct page *page,
1172 enum node_stat_item idx)
1174 __mod_lruvec_page_state(page, idx, 1);
1177 static inline void __dec_lruvec_page_state(struct page *page,
1178 enum node_stat_item idx)
1180 __mod_lruvec_page_state(page, idx, -1);
1183 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1185 __mod_lruvec_slab_state(p, idx, 1);
1188 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1190 __mod_lruvec_slab_state(p, idx, -1);
1193 /* idx can be of type enum memcg_stat_item or node_stat_item */
1194 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1197 mod_memcg_state(memcg, idx, 1);
1200 /* idx can be of type enum memcg_stat_item or node_stat_item */
1201 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1204 mod_memcg_state(memcg, idx, -1);
1207 /* idx can be of type enum memcg_stat_item or node_stat_item */
1208 static inline void inc_memcg_page_state(struct page *page,
1211 mod_memcg_page_state(page, idx, 1);
1214 /* idx can be of type enum memcg_stat_item or node_stat_item */
1215 static inline void dec_memcg_page_state(struct page *page,
1218 mod_memcg_page_state(page, idx, -1);
1221 static inline void inc_lruvec_state(struct lruvec *lruvec,
1222 enum node_stat_item idx)
1224 mod_lruvec_state(lruvec, idx, 1);
1227 static inline void dec_lruvec_state(struct lruvec *lruvec,
1228 enum node_stat_item idx)
1230 mod_lruvec_state(lruvec, idx, -1);
1233 static inline void inc_lruvec_page_state(struct page *page,
1234 enum node_stat_item idx)
1236 mod_lruvec_page_state(page, idx, 1);
1239 static inline void dec_lruvec_page_state(struct page *page,
1240 enum node_stat_item idx)
1242 mod_lruvec_page_state(page, idx, -1);
1245 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1247 struct mem_cgroup *memcg;
1249 memcg = lruvec_memcg(lruvec);
1252 memcg = parent_mem_cgroup(memcg);
1255 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1258 #ifdef CONFIG_CGROUP_WRITEBACK
1260 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1261 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1262 unsigned long *pheadroom, unsigned long *pdirty,
1263 unsigned long *pwriteback);
1265 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1266 struct bdi_writeback *wb);
1268 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1269 struct bdi_writeback *wb)
1271 if (mem_cgroup_disabled())
1274 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1275 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1278 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1280 #else /* CONFIG_CGROUP_WRITEBACK */
1282 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1287 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1288 unsigned long *pfilepages,
1289 unsigned long *pheadroom,
1290 unsigned long *pdirty,
1291 unsigned long *pwriteback)
1295 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1296 struct bdi_writeback *wb)
1300 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1304 #endif /* CONFIG_CGROUP_WRITEBACK */
1307 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1308 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1310 extern struct static_key_false memcg_sockets_enabled_key;
1311 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1312 void mem_cgroup_sk_alloc(struct sock *sk);
1313 void mem_cgroup_sk_free(struct sock *sk);
1314 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1316 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1319 if (time_before(jiffies, memcg->socket_pressure))
1321 } while ((memcg = parent_mem_cgroup(memcg)));
1325 extern int memcg_expand_shrinker_maps(int new_id);
1327 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1328 int nid, int shrinker_id);
1330 #define mem_cgroup_sockets_enabled 0
1331 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1332 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1333 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1338 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1339 int nid, int shrinker_id)
1344 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1345 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1347 #ifdef CONFIG_MEMCG_KMEM
1348 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1349 unsigned int nr_pages);
1350 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1351 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1352 void __memcg_kmem_uncharge_page(struct page *page, int order);
1354 extern struct static_key_false memcg_kmem_enabled_key;
1355 extern struct workqueue_struct *memcg_kmem_cache_wq;
1357 extern int memcg_nr_cache_ids;
1358 void memcg_get_cache_ids(void);
1359 void memcg_put_cache_ids(void);
1362 * Helper macro to loop through all memcg-specific caches. Callers must still
1363 * check if the cache is valid (it is either valid or NULL).
1364 * the slab_mutex must be held when looping through those caches
1366 #define for_each_memcg_cache_index(_idx) \
1367 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1369 static inline bool memcg_kmem_enabled(void)
1371 return static_branch_unlikely(&memcg_kmem_enabled_key);
1374 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1377 if (memcg_kmem_enabled())
1378 return __memcg_kmem_charge_page(page, gfp, order);
1382 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1384 if (memcg_kmem_enabled())
1385 __memcg_kmem_uncharge_page(page, order);
1388 static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1389 unsigned int nr_pages)
1391 if (memcg_kmem_enabled())
1392 return __memcg_kmem_charge(memcg, gfp, nr_pages);
1396 static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg,
1397 unsigned int nr_pages)
1399 if (memcg_kmem_enabled())
1400 __memcg_kmem_uncharge(memcg, nr_pages);
1404 * helper for accessing a memcg's index. It will be used as an index in the
1405 * child cache array in kmem_cache, and also to derive its name. This function
1406 * will return -1 when this is not a kmem-limited memcg.
1408 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1410 return memcg ? memcg->kmemcg_id : -1;
1413 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1417 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1423 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1427 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1433 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1437 #define for_each_memcg_cache_index(_idx) \
1440 static inline bool memcg_kmem_enabled(void)
1445 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1450 static inline void memcg_get_cache_ids(void)
1454 static inline void memcg_put_cache_ids(void)
1458 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1463 #endif /* CONFIG_MEMCG_KMEM */
1465 #endif /* _LINUX_MEMCONTROL_H */