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 {
61 unsigned int generation;
66 #define MEM_CGROUP_ID_SHIFT 16
67 #define MEM_CGROUP_ID_MAX USHRT_MAX
69 struct mem_cgroup_id {
75 * Per memcg event counter is incremented at every pagein/pageout. With THP,
76 * it will be incremated by the number of pages. This counter is used for
77 * for trigger some periodic events. This is straightforward and better
78 * than using jiffies etc. to handle periodic memcg event.
80 enum mem_cgroup_events_target {
81 MEM_CGROUP_TARGET_THRESH,
82 MEM_CGROUP_TARGET_SOFTLIMIT,
86 struct memcg_vmstats_percpu {
87 long stat[MEMCG_NR_STAT];
88 unsigned long events[NR_VM_EVENT_ITEMS];
89 unsigned long nr_page_events;
90 unsigned long targets[MEM_CGROUP_NTARGETS];
93 struct mem_cgroup_reclaim_iter {
94 struct mem_cgroup *position;
95 /* scan generation, increased every round-trip */
96 unsigned int generation;
100 long count[NR_VM_NODE_STAT_ITEMS];
104 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
105 * which have elements charged to this memcg.
107 struct memcg_shrinker_map {
109 unsigned long map[0];
113 * per-node information in memory controller.
115 struct mem_cgroup_per_node {
116 struct lruvec lruvec;
118 /* Legacy local VM stats */
119 struct lruvec_stat __percpu *lruvec_stat_local;
121 /* Subtree VM stats (batched updates) */
122 struct lruvec_stat __percpu *lruvec_stat_cpu;
123 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
125 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
127 struct mem_cgroup_reclaim_iter iter;
129 struct memcg_shrinker_map __rcu *shrinker_map;
131 struct rb_node tree_node; /* RB tree node */
132 unsigned long usage_in_excess;/* Set to the value by which */
133 /* the soft limit is exceeded*/
135 struct mem_cgroup *memcg; /* Back pointer, we cannot */
136 /* use container_of */
139 struct mem_cgroup_threshold {
140 struct eventfd_ctx *eventfd;
141 unsigned long threshold;
145 struct mem_cgroup_threshold_ary {
146 /* An array index points to threshold just below or equal to usage. */
147 int current_threshold;
148 /* Size of entries[] */
150 /* Array of thresholds */
151 struct mem_cgroup_threshold entries[0];
154 struct mem_cgroup_thresholds {
155 /* Primary thresholds array */
156 struct mem_cgroup_threshold_ary *primary;
158 * Spare threshold array.
159 * This is needed to make mem_cgroup_unregister_event() "never fail".
160 * It must be able to store at least primary->size - 1 entries.
162 struct mem_cgroup_threshold_ary *spare;
165 enum memcg_kmem_state {
171 #if defined(CONFIG_SMP)
172 struct memcg_padding {
174 } ____cacheline_internodealigned_in_smp;
175 #define MEMCG_PADDING(name) struct memcg_padding name;
177 #define MEMCG_PADDING(name)
181 * Remember four most recent foreign writebacks with dirty pages in this
182 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
183 * one in a given round, we're likely to catch it later if it keeps
184 * foreign-dirtying, so a fairly low count should be enough.
186 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
188 #define MEMCG_CGWB_FRN_CNT 4
190 struct memcg_cgwb_frn {
191 u64 bdi_id; /* bdi->id of the foreign inode */
192 int memcg_id; /* memcg->css.id of foreign inode */
193 u64 at; /* jiffies_64 at the time of dirtying */
194 struct wb_completion done; /* tracks in-flight foreign writebacks */
198 * The memory controller data structure. The memory controller controls both
199 * page cache and RSS per cgroup. We would eventually like to provide
200 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
201 * to help the administrator determine what knobs to tune.
204 struct cgroup_subsys_state css;
206 /* Private memcg ID. Used to ID objects that outlive the cgroup */
207 struct mem_cgroup_id id;
209 /* Accounted resources */
210 struct page_counter memory;
211 struct page_counter swap;
213 /* Legacy consumer-oriented counters */
214 struct page_counter memsw;
215 struct page_counter kmem;
216 struct page_counter tcpmem;
218 /* Upper bound of normal memory consumption range */
221 /* Range enforcement for interrupt charges */
222 struct work_struct high_work;
224 unsigned long soft_limit;
226 /* vmpressure notifications */
227 struct vmpressure vmpressure;
230 * Should the accounting and control be hierarchical, per subtree?
235 * Should the OOM killer kill all belonging tasks, had it kill one?
239 /* protected by memcg_oom_lock */
244 /* OOM-Killer disable */
245 int oom_kill_disable;
247 /* memory.events and memory.events.local */
248 struct cgroup_file events_file;
249 struct cgroup_file events_local_file;
251 /* handle for "memory.swap.events" */
252 struct cgroup_file swap_events_file;
254 /* protect arrays of thresholds */
255 struct mutex thresholds_lock;
257 /* thresholds for memory usage. RCU-protected */
258 struct mem_cgroup_thresholds thresholds;
260 /* thresholds for mem+swap usage. RCU-protected */
261 struct mem_cgroup_thresholds memsw_thresholds;
263 /* For oom notifier event fd */
264 struct list_head oom_notify;
267 * Should we move charges of a task when a task is moved into this
268 * mem_cgroup ? And what type of charges should we move ?
270 unsigned long move_charge_at_immigrate;
271 /* taken only while moving_account > 0 */
272 spinlock_t move_lock;
273 unsigned long move_lock_flags;
275 MEMCG_PADDING(_pad1_);
278 * set > 0 if pages under this cgroup are moving to other cgroup.
280 atomic_t moving_account;
281 struct task_struct *move_lock_task;
283 /* Legacy local VM stats and events */
284 struct memcg_vmstats_percpu __percpu *vmstats_local;
286 /* Subtree VM stats and events (batched updates) */
287 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
289 MEMCG_PADDING(_pad2_);
291 atomic_long_t vmstats[MEMCG_NR_STAT];
292 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
295 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
296 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
298 unsigned long socket_pressure;
300 /* Legacy tcp memory accounting */
304 #ifdef CONFIG_MEMCG_KMEM
305 /* Index in the kmem_cache->memcg_params.memcg_caches array */
307 enum memcg_kmem_state kmem_state;
308 struct list_head kmem_caches;
311 #ifdef CONFIG_CGROUP_WRITEBACK
312 struct list_head cgwb_list;
313 struct wb_domain cgwb_domain;
314 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
317 /* List of events which userspace want to receive */
318 struct list_head event_list;
319 spinlock_t event_list_lock;
321 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
322 struct deferred_split deferred_split_queue;
325 struct mem_cgroup_per_node *nodeinfo[0];
326 /* WARNING: nodeinfo must be the last member here */
330 * size of first charge trial. "32" comes from vmscan.c's magic value.
331 * TODO: maybe necessary to use big numbers in big irons.
333 #define MEMCG_CHARGE_BATCH 32U
335 extern struct mem_cgroup *root_mem_cgroup;
337 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
339 return (memcg == root_mem_cgroup);
342 static inline bool mem_cgroup_disabled(void)
344 return !cgroup_subsys_enabled(memory_cgrp_subsys);
347 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
350 if (mem_cgroup_disabled())
354 return READ_ONCE(memcg->memory.emin);
356 return max(READ_ONCE(memcg->memory.emin),
357 READ_ONCE(memcg->memory.elow));
360 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
361 struct mem_cgroup *memcg);
363 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
364 gfp_t gfp_mask, struct mem_cgroup **memcgp,
366 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
367 gfp_t gfp_mask, struct mem_cgroup **memcgp,
369 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
370 bool lrucare, bool compound);
371 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
373 void mem_cgroup_uncharge(struct page *page);
374 void mem_cgroup_uncharge_list(struct list_head *page_list);
376 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
378 static struct mem_cgroup_per_node *
379 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
381 return memcg->nodeinfo[nid];
385 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
386 * @memcg: memcg of the wanted lruvec
388 * Returns the lru list vector holding pages for a given @memcg &
389 * @node combination. This can be the node lruvec, if the memory
390 * controller is disabled.
392 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
393 struct pglist_data *pgdat)
395 struct mem_cgroup_per_node *mz;
396 struct lruvec *lruvec;
398 if (mem_cgroup_disabled()) {
399 lruvec = &pgdat->__lruvec;
404 memcg = root_mem_cgroup;
406 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
407 lruvec = &mz->lruvec;
410 * Since a node can be onlined after the mem_cgroup was created,
411 * we have to be prepared to initialize lruvec->pgdat here;
412 * and if offlined then reonlined, we need to reinitialize it.
414 if (unlikely(lruvec->pgdat != pgdat))
415 lruvec->pgdat = pgdat;
419 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
421 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
423 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
425 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
428 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
429 return css ? container_of(css, struct mem_cgroup, css) : NULL;
432 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
435 css_put(&memcg->css);
438 #define mem_cgroup_from_counter(counter, member) \
439 container_of(counter, struct mem_cgroup, member)
441 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
443 struct mem_cgroup_reclaim_cookie *);
444 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
445 int mem_cgroup_scan_tasks(struct mem_cgroup *,
446 int (*)(struct task_struct *, void *), void *);
448 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
450 if (mem_cgroup_disabled())
455 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
457 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
459 return mem_cgroup_from_css(seq_css(m));
462 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
464 struct mem_cgroup_per_node *mz;
466 if (mem_cgroup_disabled())
469 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
474 * parent_mem_cgroup - find the accounting parent of a memcg
475 * @memcg: memcg whose parent to find
477 * Returns the parent memcg, or NULL if this is the root or the memory
478 * controller is in legacy no-hierarchy mode.
480 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
482 if (!memcg->memory.parent)
484 return mem_cgroup_from_counter(memcg->memory.parent, memory);
487 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
488 struct mem_cgroup *root)
492 if (!root->use_hierarchy)
494 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
497 static inline bool mm_match_cgroup(struct mm_struct *mm,
498 struct mem_cgroup *memcg)
500 struct mem_cgroup *task_memcg;
504 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
506 match = mem_cgroup_is_descendant(task_memcg, memcg);
511 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
512 ino_t page_cgroup_ino(struct page *page);
514 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
516 if (mem_cgroup_disabled())
518 return !!(memcg->css.flags & CSS_ONLINE);
522 * For memory reclaim.
524 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
526 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
527 int zid, int nr_pages);
530 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
531 enum lru_list lru, int zone_idx)
533 struct mem_cgroup_per_node *mz;
535 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
536 return mz->lru_zone_size[zone_idx][lru];
539 void mem_cgroup_handle_over_high(void);
541 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
543 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
545 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
546 struct task_struct *p);
548 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
550 static inline void mem_cgroup_enter_user_fault(void)
552 WARN_ON(current->in_user_fault);
553 current->in_user_fault = 1;
556 static inline void mem_cgroup_exit_user_fault(void)
558 WARN_ON(!current->in_user_fault);
559 current->in_user_fault = 0;
562 static inline bool task_in_memcg_oom(struct task_struct *p)
564 return p->memcg_in_oom;
567 bool mem_cgroup_oom_synchronize(bool wait);
568 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
569 struct mem_cgroup *oom_domain);
570 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
572 #ifdef CONFIG_MEMCG_SWAP
573 extern int do_swap_account;
576 struct mem_cgroup *lock_page_memcg(struct page *page);
577 void __unlock_page_memcg(struct mem_cgroup *memcg);
578 void unlock_page_memcg(struct page *page);
581 * idx can be of type enum memcg_stat_item or node_stat_item.
582 * Keep in sync with memcg_exact_page_state().
584 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
586 long x = atomic_long_read(&memcg->vmstats[idx]);
595 * idx can be of type enum memcg_stat_item or node_stat_item.
596 * Keep in sync with memcg_exact_page_state().
598 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
604 for_each_possible_cpu(cpu)
605 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
613 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
615 /* idx can be of type enum memcg_stat_item or node_stat_item */
616 static inline void mod_memcg_state(struct mem_cgroup *memcg,
621 local_irq_save(flags);
622 __mod_memcg_state(memcg, idx, val);
623 local_irq_restore(flags);
627 * mod_memcg_page_state - update page state statistics
629 * @idx: page state item to account
630 * @val: number of pages (positive or negative)
632 * The @page must be locked or the caller must use lock_page_memcg()
633 * to prevent double accounting when the page is concurrently being
634 * moved to another memcg:
636 * lock_page(page) or lock_page_memcg(page)
637 * if (TestClearPageState(page))
638 * mod_memcg_page_state(page, state, -1);
639 * unlock_page(page) or unlock_page_memcg(page)
641 * Kernel pages are an exception to this, since they'll never move.
643 static inline void __mod_memcg_page_state(struct page *page,
646 if (page->mem_cgroup)
647 __mod_memcg_state(page->mem_cgroup, idx, val);
650 static inline void mod_memcg_page_state(struct page *page,
653 if (page->mem_cgroup)
654 mod_memcg_state(page->mem_cgroup, idx, val);
657 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
658 enum node_stat_item idx)
660 struct mem_cgroup_per_node *pn;
663 if (mem_cgroup_disabled())
664 return node_page_state(lruvec_pgdat(lruvec), idx);
666 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
667 x = atomic_long_read(&pn->lruvec_stat[idx]);
675 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
676 enum node_stat_item idx)
678 struct mem_cgroup_per_node *pn;
682 if (mem_cgroup_disabled())
683 return node_page_state(lruvec_pgdat(lruvec), idx);
685 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
686 for_each_possible_cpu(cpu)
687 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
695 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
697 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
699 static inline void mod_lruvec_state(struct lruvec *lruvec,
700 enum node_stat_item idx, int val)
704 local_irq_save(flags);
705 __mod_lruvec_state(lruvec, idx, val);
706 local_irq_restore(flags);
709 static inline void __mod_lruvec_page_state(struct page *page,
710 enum node_stat_item idx, int val)
712 pg_data_t *pgdat = page_pgdat(page);
713 struct lruvec *lruvec;
715 /* Untracked pages have no memcg, no lruvec. Update only the node */
716 if (!page->mem_cgroup) {
717 __mod_node_page_state(pgdat, idx, val);
721 lruvec = mem_cgroup_lruvec(page->mem_cgroup, pgdat);
722 __mod_lruvec_state(lruvec, idx, val);
725 static inline void mod_lruvec_page_state(struct page *page,
726 enum node_stat_item idx, int val)
730 local_irq_save(flags);
731 __mod_lruvec_page_state(page, idx, val);
732 local_irq_restore(flags);
735 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
737 unsigned long *total_scanned);
739 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
740 unsigned long count);
742 static inline void count_memcg_events(struct mem_cgroup *memcg,
743 enum vm_event_item idx,
748 local_irq_save(flags);
749 __count_memcg_events(memcg, idx, count);
750 local_irq_restore(flags);
753 static inline void count_memcg_page_event(struct page *page,
754 enum vm_event_item idx)
756 if (page->mem_cgroup)
757 count_memcg_events(page->mem_cgroup, idx, 1);
760 static inline void count_memcg_event_mm(struct mm_struct *mm,
761 enum vm_event_item idx)
763 struct mem_cgroup *memcg;
765 if (mem_cgroup_disabled())
769 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
771 count_memcg_events(memcg, idx, 1);
775 static inline void memcg_memory_event(struct mem_cgroup *memcg,
776 enum memcg_memory_event event)
778 atomic_long_inc(&memcg->memory_events_local[event]);
779 cgroup_file_notify(&memcg->events_local_file);
782 atomic_long_inc(&memcg->memory_events[event]);
783 cgroup_file_notify(&memcg->events_file);
785 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
787 } while ((memcg = parent_mem_cgroup(memcg)) &&
788 !mem_cgroup_is_root(memcg));
791 static inline void memcg_memory_event_mm(struct mm_struct *mm,
792 enum memcg_memory_event event)
794 struct mem_cgroup *memcg;
796 if (mem_cgroup_disabled())
800 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
802 memcg_memory_event(memcg, event);
806 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
807 void mem_cgroup_split_huge_fixup(struct page *head);
810 #else /* CONFIG_MEMCG */
812 #define MEM_CGROUP_ID_SHIFT 0
813 #define MEM_CGROUP_ID_MAX 0
817 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
822 static inline bool mem_cgroup_disabled(void)
827 static inline void memcg_memory_event(struct mem_cgroup *memcg,
828 enum memcg_memory_event event)
832 static inline void memcg_memory_event_mm(struct mm_struct *mm,
833 enum memcg_memory_event event)
837 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
843 static inline enum mem_cgroup_protection mem_cgroup_protected(
844 struct mem_cgroup *root, struct mem_cgroup *memcg)
846 return MEMCG_PROT_NONE;
849 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
851 struct mem_cgroup **memcgp,
858 static inline int mem_cgroup_try_charge_delay(struct page *page,
859 struct mm_struct *mm,
861 struct mem_cgroup **memcgp,
868 static inline void mem_cgroup_commit_charge(struct page *page,
869 struct mem_cgroup *memcg,
870 bool lrucare, bool compound)
874 static inline void mem_cgroup_cancel_charge(struct page *page,
875 struct mem_cgroup *memcg,
880 static inline void mem_cgroup_uncharge(struct page *page)
884 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
888 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
892 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
893 struct pglist_data *pgdat)
895 return &pgdat->__lruvec;
898 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
899 struct pglist_data *pgdat)
901 return &pgdat->__lruvec;
904 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
909 static inline bool mm_match_cgroup(struct mm_struct *mm,
910 struct mem_cgroup *memcg)
915 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
920 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
925 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
929 static inline struct mem_cgroup *
930 mem_cgroup_iter(struct mem_cgroup *root,
931 struct mem_cgroup *prev,
932 struct mem_cgroup_reclaim_cookie *reclaim)
937 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
938 struct mem_cgroup *prev)
942 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
943 int (*fn)(struct task_struct *, void *), void *arg)
948 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
953 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
956 /* XXX: This should always return root_mem_cgroup */
960 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
965 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
970 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
976 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
977 enum lru_list lru, int zone_idx)
982 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
987 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
993 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
998 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1002 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1007 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1011 static inline void unlock_page_memcg(struct page *page)
1015 static inline void mem_cgroup_handle_over_high(void)
1019 static inline void mem_cgroup_enter_user_fault(void)
1023 static inline void mem_cgroup_exit_user_fault(void)
1027 static inline bool task_in_memcg_oom(struct task_struct *p)
1032 static inline bool mem_cgroup_oom_synchronize(bool wait)
1037 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1038 struct task_struct *victim, struct mem_cgroup *oom_domain)
1043 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1047 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1052 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1058 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1064 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1070 static inline void __mod_memcg_page_state(struct page *page,
1076 static inline void mod_memcg_page_state(struct page *page,
1082 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1083 enum node_stat_item idx)
1085 return node_page_state(lruvec_pgdat(lruvec), idx);
1088 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1089 enum node_stat_item idx)
1091 return node_page_state(lruvec_pgdat(lruvec), idx);
1094 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1095 enum node_stat_item idx, int val)
1097 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1100 static inline void mod_lruvec_state(struct lruvec *lruvec,
1101 enum node_stat_item idx, int val)
1103 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1106 static inline void __mod_lruvec_page_state(struct page *page,
1107 enum node_stat_item idx, int val)
1109 __mod_node_page_state(page_pgdat(page), idx, val);
1112 static inline void mod_lruvec_page_state(struct page *page,
1113 enum node_stat_item idx, int val)
1115 mod_node_page_state(page_pgdat(page), idx, val);
1118 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1121 struct page *page = virt_to_head_page(p);
1123 __mod_node_page_state(page_pgdat(page), idx, val);
1127 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1129 unsigned long *total_scanned)
1134 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1138 static inline void count_memcg_events(struct mem_cgroup *memcg,
1139 enum vm_event_item idx,
1140 unsigned long count)
1144 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1145 enum vm_event_item idx,
1146 unsigned long count)
1150 static inline void count_memcg_page_event(struct page *page,
1156 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1159 #endif /* CONFIG_MEMCG */
1161 /* idx can be of type enum memcg_stat_item or node_stat_item */
1162 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1165 __mod_memcg_state(memcg, idx, 1);
1168 /* idx can be of type enum memcg_stat_item or node_stat_item */
1169 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1172 __mod_memcg_state(memcg, idx, -1);
1175 /* idx can be of type enum memcg_stat_item or node_stat_item */
1176 static inline void __inc_memcg_page_state(struct page *page,
1179 __mod_memcg_page_state(page, idx, 1);
1182 /* idx can be of type enum memcg_stat_item or node_stat_item */
1183 static inline void __dec_memcg_page_state(struct page *page,
1186 __mod_memcg_page_state(page, idx, -1);
1189 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1190 enum node_stat_item idx)
1192 __mod_lruvec_state(lruvec, idx, 1);
1195 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1196 enum node_stat_item idx)
1198 __mod_lruvec_state(lruvec, idx, -1);
1201 static inline void __inc_lruvec_page_state(struct page *page,
1202 enum node_stat_item idx)
1204 __mod_lruvec_page_state(page, idx, 1);
1207 static inline void __dec_lruvec_page_state(struct page *page,
1208 enum node_stat_item idx)
1210 __mod_lruvec_page_state(page, idx, -1);
1213 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1215 __mod_lruvec_slab_state(p, idx, 1);
1218 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1220 __mod_lruvec_slab_state(p, idx, -1);
1223 /* idx can be of type enum memcg_stat_item or node_stat_item */
1224 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1227 mod_memcg_state(memcg, idx, 1);
1230 /* idx can be of type enum memcg_stat_item or node_stat_item */
1231 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1234 mod_memcg_state(memcg, idx, -1);
1237 /* idx can be of type enum memcg_stat_item or node_stat_item */
1238 static inline void inc_memcg_page_state(struct page *page,
1241 mod_memcg_page_state(page, idx, 1);
1244 /* idx can be of type enum memcg_stat_item or node_stat_item */
1245 static inline void dec_memcg_page_state(struct page *page,
1248 mod_memcg_page_state(page, idx, -1);
1251 static inline void inc_lruvec_state(struct lruvec *lruvec,
1252 enum node_stat_item idx)
1254 mod_lruvec_state(lruvec, idx, 1);
1257 static inline void dec_lruvec_state(struct lruvec *lruvec,
1258 enum node_stat_item idx)
1260 mod_lruvec_state(lruvec, idx, -1);
1263 static inline void inc_lruvec_page_state(struct page *page,
1264 enum node_stat_item idx)
1266 mod_lruvec_page_state(page, idx, 1);
1269 static inline void dec_lruvec_page_state(struct page *page,
1270 enum node_stat_item idx)
1272 mod_lruvec_page_state(page, idx, -1);
1275 #ifdef CONFIG_CGROUP_WRITEBACK
1277 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1278 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1279 unsigned long *pheadroom, unsigned long *pdirty,
1280 unsigned long *pwriteback);
1282 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1283 struct bdi_writeback *wb);
1285 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1286 struct bdi_writeback *wb)
1288 if (mem_cgroup_disabled())
1291 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1292 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1295 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1297 #else /* CONFIG_CGROUP_WRITEBACK */
1299 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1304 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1305 unsigned long *pfilepages,
1306 unsigned long *pheadroom,
1307 unsigned long *pdirty,
1308 unsigned long *pwriteback)
1312 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1313 struct bdi_writeback *wb)
1317 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1321 #endif /* CONFIG_CGROUP_WRITEBACK */
1324 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1325 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1327 extern struct static_key_false memcg_sockets_enabled_key;
1328 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1329 void mem_cgroup_sk_alloc(struct sock *sk);
1330 void mem_cgroup_sk_free(struct sock *sk);
1331 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1333 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1336 if (time_before(jiffies, memcg->socket_pressure))
1338 } while ((memcg = parent_mem_cgroup(memcg)));
1342 extern int memcg_expand_shrinker_maps(int new_id);
1344 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1345 int nid, int shrinker_id);
1347 #define mem_cgroup_sockets_enabled 0
1348 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1349 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1350 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1355 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1356 int nid, int shrinker_id)
1361 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1362 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1364 #ifdef CONFIG_MEMCG_KMEM
1365 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1366 void __memcg_kmem_uncharge(struct page *page, int order);
1367 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1368 struct mem_cgroup *memcg);
1369 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1370 unsigned int nr_pages);
1372 extern struct static_key_false memcg_kmem_enabled_key;
1373 extern struct workqueue_struct *memcg_kmem_cache_wq;
1375 extern int memcg_nr_cache_ids;
1376 void memcg_get_cache_ids(void);
1377 void memcg_put_cache_ids(void);
1380 * Helper macro to loop through all memcg-specific caches. Callers must still
1381 * check if the cache is valid (it is either valid or NULL).
1382 * the slab_mutex must be held when looping through those caches
1384 #define for_each_memcg_cache_index(_idx) \
1385 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1387 static inline bool memcg_kmem_enabled(void)
1389 return static_branch_unlikely(&memcg_kmem_enabled_key);
1392 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1394 if (memcg_kmem_enabled())
1395 return __memcg_kmem_charge(page, gfp, order);
1399 static inline void memcg_kmem_uncharge(struct page *page, int order)
1401 if (memcg_kmem_enabled())
1402 __memcg_kmem_uncharge(page, order);
1405 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1406 int order, struct mem_cgroup *memcg)
1408 if (memcg_kmem_enabled())
1409 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1413 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1414 struct mem_cgroup *memcg)
1416 if (memcg_kmem_enabled())
1417 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1421 * helper for accessing a memcg's index. It will be used as an index in the
1422 * child cache array in kmem_cache, and also to derive its name. This function
1423 * will return -1 when this is not a kmem-limited memcg.
1425 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1427 return memcg ? memcg->kmemcg_id : -1;
1432 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1437 static inline void memcg_kmem_uncharge(struct page *page, int order)
1441 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1446 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1450 #define for_each_memcg_cache_index(_idx) \
1453 static inline bool memcg_kmem_enabled(void)
1458 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1463 static inline void memcg_get_cache_ids(void)
1467 static inline void memcg_put_cache_ids(void)
1471 #endif /* CONFIG_MEMCG_KMEM */
1473 #endif /* _LINUX_MEMCONTROL_H */