1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* memcontrol.h - Memory Controller
4 * Copyright IBM Corporation, 2007
5 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
7 * Copyright 2007 OpenVZ SWsoft Inc
8 * Author: Pavel Emelianov <xemul@openvz.org>
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 struct mem_cgroup_reclaim_cookie {
56 unsigned int generation;
61 #define MEM_CGROUP_ID_SHIFT 16
62 #define MEM_CGROUP_ID_MAX USHRT_MAX
64 struct mem_cgroup_id {
70 * Per memcg event counter is incremented at every pagein/pageout. With THP,
71 * it will be incremated by the number of pages. This counter is used for
72 * for trigger some periodic events. This is straightforward and better
73 * than using jiffies etc. to handle periodic memcg event.
75 enum mem_cgroup_events_target {
76 MEM_CGROUP_TARGET_THRESH,
77 MEM_CGROUP_TARGET_SOFTLIMIT,
78 MEM_CGROUP_TARGET_NUMAINFO,
82 struct memcg_vmstats_percpu {
83 long stat[MEMCG_NR_STAT];
84 unsigned long events[NR_VM_EVENT_ITEMS];
85 unsigned long nr_page_events;
86 unsigned long targets[MEM_CGROUP_NTARGETS];
89 struct mem_cgroup_reclaim_iter {
90 struct mem_cgroup *position;
91 /* scan generation, increased every round-trip */
92 unsigned int generation;
96 long count[NR_VM_NODE_STAT_ITEMS];
100 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
101 * which have elements charged to this memcg.
103 struct memcg_shrinker_map {
105 unsigned long map[0];
109 * per-zone information in memory controller.
111 struct mem_cgroup_per_node {
112 struct lruvec lruvec;
114 /* Legacy local VM stats */
115 struct lruvec_stat __percpu *lruvec_stat_local;
117 /* Subtree VM stats (batched updates) */
118 struct lruvec_stat __percpu *lruvec_stat_cpu;
119 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
121 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
123 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
125 struct memcg_shrinker_map __rcu *shrinker_map;
127 struct rb_node tree_node; /* RB tree node */
128 unsigned long usage_in_excess;/* Set to the value by which */
129 /* the soft limit is exceeded*/
131 struct mem_cgroup *memcg; /* Back pointer, we cannot */
132 /* use container_of */
135 struct mem_cgroup_threshold {
136 struct eventfd_ctx *eventfd;
137 unsigned long threshold;
141 struct mem_cgroup_threshold_ary {
142 /* An array index points to threshold just below or equal to usage. */
143 int current_threshold;
144 /* Size of entries[] */
146 /* Array of thresholds */
147 struct mem_cgroup_threshold entries[0];
150 struct mem_cgroup_thresholds {
151 /* Primary thresholds array */
152 struct mem_cgroup_threshold_ary *primary;
154 * Spare threshold array.
155 * This is needed to make mem_cgroup_unregister_event() "never fail".
156 * It must be able to store at least primary->size - 1 entries.
158 struct mem_cgroup_threshold_ary *spare;
161 enum memcg_kmem_state {
167 #if defined(CONFIG_SMP)
168 struct memcg_padding {
170 } ____cacheline_internodealigned_in_smp;
171 #define MEMCG_PADDING(name) struct memcg_padding name;
173 #define MEMCG_PADDING(name)
177 * Remember four most recent foreign writebacks with dirty pages in this
178 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
179 * one in a given round, we're likely to catch it later if it keeps
180 * foreign-dirtying, so a fairly low count should be enough.
182 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
184 #define MEMCG_CGWB_FRN_CNT 4
186 struct memcg_cgwb_frn {
187 u64 bdi_id; /* bdi->id of the foreign inode */
188 int memcg_id; /* memcg->css.id of foreign inode */
189 u64 at; /* jiffies_64 at the time of dirtying */
190 struct wb_completion done; /* tracks in-flight foreign writebacks */
194 * The memory controller data structure. The memory controller controls both
195 * page cache and RSS per cgroup. We would eventually like to provide
196 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
197 * to help the administrator determine what knobs to tune.
200 struct cgroup_subsys_state css;
202 /* Private memcg ID. Used to ID objects that outlive the cgroup */
203 struct mem_cgroup_id id;
205 /* Accounted resources */
206 struct page_counter memory;
207 struct page_counter swap;
209 /* Legacy consumer-oriented counters */
210 struct page_counter memsw;
211 struct page_counter kmem;
212 struct page_counter tcpmem;
214 /* Upper bound of normal memory consumption range */
217 /* Range enforcement for interrupt charges */
218 struct work_struct high_work;
220 unsigned long soft_limit;
222 /* vmpressure notifications */
223 struct vmpressure vmpressure;
226 * Should the accounting and control be hierarchical, per subtree?
231 * Should the OOM killer kill all belonging tasks, had it kill one?
235 /* protected by memcg_oom_lock */
240 /* OOM-Killer disable */
241 int oom_kill_disable;
243 /* memory.events and memory.events.local */
244 struct cgroup_file events_file;
245 struct cgroup_file events_local_file;
247 /* handle for "memory.swap.events" */
248 struct cgroup_file swap_events_file;
250 /* protect arrays of thresholds */
251 struct mutex thresholds_lock;
253 /* thresholds for memory usage. RCU-protected */
254 struct mem_cgroup_thresholds thresholds;
256 /* thresholds for mem+swap usage. RCU-protected */
257 struct mem_cgroup_thresholds memsw_thresholds;
259 /* For oom notifier event fd */
260 struct list_head oom_notify;
263 * Should we move charges of a task when a task is moved into this
264 * mem_cgroup ? And what type of charges should we move ?
266 unsigned long move_charge_at_immigrate;
267 /* taken only while moving_account > 0 */
268 spinlock_t move_lock;
269 unsigned long move_lock_flags;
271 MEMCG_PADDING(_pad1_);
274 * set > 0 if pages under this cgroup are moving to other cgroup.
276 atomic_t moving_account;
277 struct task_struct *move_lock_task;
279 /* Legacy local VM stats and events */
280 struct memcg_vmstats_percpu __percpu *vmstats_local;
282 /* Subtree VM stats and events (batched updates) */
283 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
285 MEMCG_PADDING(_pad2_);
287 atomic_long_t vmstats[MEMCG_NR_STAT];
288 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
291 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
292 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
294 unsigned long socket_pressure;
296 /* Legacy tcp memory accounting */
300 #ifdef CONFIG_MEMCG_KMEM
301 /* Index in the kmem_cache->memcg_params.memcg_caches array */
303 enum memcg_kmem_state kmem_state;
304 struct list_head kmem_caches;
307 int last_scanned_node;
309 nodemask_t scan_nodes;
310 atomic_t numainfo_events;
311 atomic_t numainfo_updating;
314 #ifdef CONFIG_CGROUP_WRITEBACK
315 struct list_head cgwb_list;
316 struct wb_domain cgwb_domain;
317 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
320 /* List of events which userspace want to receive */
321 struct list_head event_list;
322 spinlock_t event_list_lock;
324 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
325 struct deferred_split deferred_split_queue;
328 #ifdef CONFIG_LRU_GEN
329 /* per-memcg mm_struct list */
330 struct lru_gen_mm_list mm_list;
333 struct mem_cgroup_per_node *nodeinfo[0];
334 /* WARNING: nodeinfo must be the last member here */
338 * size of first charge trial. "32" comes from vmscan.c's magic value.
339 * TODO: maybe necessary to use big numbers in big irons.
341 #define MEMCG_CHARGE_BATCH 32U
343 extern struct mem_cgroup *root_mem_cgroup;
345 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
347 return (memcg == root_mem_cgroup);
350 static inline bool mem_cgroup_disabled(void)
352 return !cgroup_subsys_enabled(memory_cgrp_subsys);
355 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
358 if (mem_cgroup_disabled())
362 return READ_ONCE(memcg->memory.emin);
364 return max(READ_ONCE(memcg->memory.emin),
365 READ_ONCE(memcg->memory.elow));
368 void mem_cgroup_calculate_protection(struct mem_cgroup *root,
369 struct mem_cgroup *memcg);
371 static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg)
374 * The root memcg doesn't account charges, and doesn't support
377 return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg);
381 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
383 if (!mem_cgroup_supports_protection(memcg))
386 return READ_ONCE(memcg->memory.elow) >=
387 page_counter_read(&memcg->memory);
390 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
392 if (!mem_cgroup_supports_protection(memcg))
395 return READ_ONCE(memcg->memory.emin) >=
396 page_counter_read(&memcg->memory);
399 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
400 gfp_t gfp_mask, struct mem_cgroup **memcgp,
402 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
403 gfp_t gfp_mask, struct mem_cgroup **memcgp,
405 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
406 bool lrucare, bool compound);
407 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
409 void mem_cgroup_uncharge(struct page *page);
410 void mem_cgroup_uncharge_list(struct list_head *page_list);
412 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
414 static struct mem_cgroup_per_node *
415 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
417 return memcg->nodeinfo[nid];
421 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
422 * @memcg: memcg of the wanted lruvec
424 * Returns the lru list vector holding pages for a given @memcg &
425 * @node combination. This can be the node lruvec, if the memory
426 * controller is disabled.
428 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
429 struct pglist_data *pgdat)
431 struct mem_cgroup_per_node *mz;
432 struct lruvec *lruvec;
434 if (mem_cgroup_disabled()) {
435 lruvec = &pgdat->__lruvec;
440 memcg = root_mem_cgroup;
442 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
443 lruvec = &mz->lruvec;
446 * Since a node can be onlined after the mem_cgroup was created,
447 * we have to be prepared to initialize lruvec->pgdat here;
448 * and if offlined then reonlined, we need to reinitialize it.
450 if (unlikely(lruvec->pgdat != pgdat))
451 lruvec->pgdat = pgdat;
455 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
457 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
459 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
461 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
464 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
465 return css ? container_of(css, struct mem_cgroup, css) : NULL;
468 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
471 css_put(&memcg->css);
474 #define mem_cgroup_from_counter(counter, member) \
475 container_of(counter, struct mem_cgroup, member)
477 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
479 struct mem_cgroup_reclaim_cookie *);
480 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
481 int mem_cgroup_scan_tasks(struct mem_cgroup *,
482 int (*)(struct task_struct *, void *), void *);
484 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
486 if (mem_cgroup_disabled())
491 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
493 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
495 return mem_cgroup_from_css(seq_css(m));
498 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
500 struct mem_cgroup_per_node *mz;
502 if (mem_cgroup_disabled())
505 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
510 * parent_mem_cgroup - find the accounting parent of a memcg
511 * @memcg: memcg whose parent to find
513 * Returns the parent memcg, or NULL if this is the root or the memory
514 * controller is in legacy no-hierarchy mode.
516 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
518 if (!memcg->memory.parent)
520 return mem_cgroup_from_counter(memcg->memory.parent, memory);
523 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
524 struct mem_cgroup *root)
528 if (!root->use_hierarchy)
530 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
533 static inline bool mm_match_cgroup(struct mm_struct *mm,
534 struct mem_cgroup *memcg)
536 struct mem_cgroup *task_memcg;
540 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
542 match = mem_cgroup_is_descendant(task_memcg, memcg);
547 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
548 ino_t page_cgroup_ino(struct page *page);
550 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
552 if (mem_cgroup_disabled())
554 return !!(memcg->css.flags & CSS_ONLINE);
558 * For memory reclaim.
560 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
562 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
563 int zid, int nr_pages);
566 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
567 enum lru_list lru, int zone_idx)
569 struct mem_cgroup_per_node *mz;
571 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
572 return mz->lru_zone_size[zone_idx][lru];
575 void mem_cgroup_handle_over_high(void);
577 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
579 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
581 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
582 struct task_struct *p);
584 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
586 static inline void mem_cgroup_enter_user_fault(void)
588 WARN_ON(current->in_user_fault);
589 current->in_user_fault = 1;
592 static inline void mem_cgroup_exit_user_fault(void)
594 WARN_ON(!current->in_user_fault);
595 current->in_user_fault = 0;
598 static inline bool task_in_memcg_oom(struct task_struct *p)
600 return p->memcg_in_oom;
603 bool mem_cgroup_oom_synchronize(bool wait);
604 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
605 struct mem_cgroup *oom_domain);
606 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
608 #ifdef CONFIG_MEMCG_SWAP
609 extern int do_swap_account;
612 struct mem_cgroup *lock_page_memcg(struct page *page);
613 void __unlock_page_memcg(struct mem_cgroup *memcg);
614 void unlock_page_memcg(struct page *page);
617 * idx can be of type enum memcg_stat_item or node_stat_item.
618 * Keep in sync with memcg_exact_page_state().
620 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
622 long x = atomic_long_read(&memcg->vmstats[idx]);
631 * idx can be of type enum memcg_stat_item or node_stat_item.
632 * Keep in sync with memcg_exact_page_state().
634 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
640 for_each_possible_cpu(cpu)
641 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
649 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
651 /* try to stablize page_memcg() for all the pages in a memcg */
652 static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
656 if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account))
663 static inline void mem_cgroup_unlock_pages(void)
668 /* idx can be of type enum memcg_stat_item or node_stat_item */
669 static inline void mod_memcg_state(struct mem_cgroup *memcg,
674 local_irq_save(flags);
675 __mod_memcg_state(memcg, idx, val);
676 local_irq_restore(flags);
680 * mod_memcg_page_state - update page state statistics
682 * @idx: page state item to account
683 * @val: number of pages (positive or negative)
685 * The @page must be locked or the caller must use lock_page_memcg()
686 * to prevent double accounting when the page is concurrently being
687 * moved to another memcg:
689 * lock_page(page) or lock_page_memcg(page)
690 * if (TestClearPageState(page))
691 * mod_memcg_page_state(page, state, -1);
692 * unlock_page(page) or unlock_page_memcg(page)
694 * Kernel pages are an exception to this, since they'll never move.
696 static inline void __mod_memcg_page_state(struct page *page,
699 if (page->mem_cgroup)
700 __mod_memcg_state(page->mem_cgroup, idx, val);
703 static inline void mod_memcg_page_state(struct page *page,
706 if (page->mem_cgroup)
707 mod_memcg_state(page->mem_cgroup, idx, val);
710 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
711 enum node_stat_item idx)
713 struct mem_cgroup_per_node *pn;
716 if (mem_cgroup_disabled())
717 return node_page_state(lruvec_pgdat(lruvec), idx);
719 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
720 x = atomic_long_read(&pn->lruvec_stat[idx]);
728 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
729 enum node_stat_item idx)
731 struct mem_cgroup_per_node *pn;
735 if (mem_cgroup_disabled())
736 return node_page_state(lruvec_pgdat(lruvec), idx);
738 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
739 for_each_possible_cpu(cpu)
740 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
748 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
750 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
751 void mod_memcg_obj_state(void *p, int idx, int val);
753 static inline void mod_lruvec_state(struct lruvec *lruvec,
754 enum node_stat_item idx, int val)
758 local_irq_save(flags);
759 __mod_lruvec_state(lruvec, idx, val);
760 local_irq_restore(flags);
763 static inline void __mod_lruvec_page_state(struct page *page,
764 enum node_stat_item idx, int val)
766 pg_data_t *pgdat = page_pgdat(page);
767 struct lruvec *lruvec;
769 /* Untracked pages have no memcg, no lruvec. Update only the node */
770 if (!page->mem_cgroup) {
771 __mod_node_page_state(pgdat, idx, val);
775 lruvec = mem_cgroup_lruvec(page->mem_cgroup, pgdat);
776 __mod_lruvec_state(lruvec, idx, val);
779 static inline void mod_lruvec_page_state(struct page *page,
780 enum node_stat_item idx, int val)
784 local_irq_save(flags);
785 __mod_lruvec_page_state(page, idx, val);
786 local_irq_restore(flags);
789 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
791 unsigned long *total_scanned);
793 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
794 unsigned long count);
796 static inline void count_memcg_events(struct mem_cgroup *memcg,
797 enum vm_event_item idx,
802 local_irq_save(flags);
803 __count_memcg_events(memcg, idx, count);
804 local_irq_restore(flags);
807 static inline void count_memcg_page_event(struct page *page,
808 enum vm_event_item idx)
810 if (page->mem_cgroup)
811 count_memcg_events(page->mem_cgroup, idx, 1);
814 static inline void count_memcg_event_mm(struct mm_struct *mm,
815 enum vm_event_item idx)
817 struct mem_cgroup *memcg;
819 if (mem_cgroup_disabled())
823 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
825 count_memcg_events(memcg, idx, 1);
829 static inline void memcg_memory_event(struct mem_cgroup *memcg,
830 enum memcg_memory_event event)
832 atomic_long_inc(&memcg->memory_events_local[event]);
833 cgroup_file_notify(&memcg->events_local_file);
836 atomic_long_inc(&memcg->memory_events[event]);
837 cgroup_file_notify(&memcg->events_file);
839 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
841 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
843 } while ((memcg = parent_mem_cgroup(memcg)) &&
844 !mem_cgroup_is_root(memcg));
847 static inline void memcg_memory_event_mm(struct mm_struct *mm,
848 enum memcg_memory_event event)
850 struct mem_cgroup *memcg;
852 if (mem_cgroup_disabled())
856 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
858 memcg_memory_event(memcg, event);
862 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
863 void mem_cgroup_split_huge_fixup(struct page *head);
866 #else /* CONFIG_MEMCG */
868 #define MEM_CGROUP_ID_SHIFT 0
869 #define MEM_CGROUP_ID_MAX 0
873 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
878 static inline bool mem_cgroup_disabled(void)
883 static inline void memcg_memory_event(struct mem_cgroup *memcg,
884 enum memcg_memory_event event)
888 static inline void memcg_memory_event_mm(struct mm_struct *mm,
889 enum memcg_memory_event event)
893 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
899 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
900 struct mem_cgroup *memcg)
904 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
909 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
914 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
916 struct mem_cgroup **memcgp,
923 static inline int mem_cgroup_try_charge_delay(struct page *page,
924 struct mm_struct *mm,
926 struct mem_cgroup **memcgp,
933 static inline void mem_cgroup_commit_charge(struct page *page,
934 struct mem_cgroup *memcg,
935 bool lrucare, bool compound)
939 static inline void mem_cgroup_cancel_charge(struct page *page,
940 struct mem_cgroup *memcg,
945 static inline void mem_cgroup_uncharge(struct page *page)
949 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
953 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
957 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
958 struct pglist_data *pgdat)
960 return &pgdat->__lruvec;
963 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
964 struct pglist_data *pgdat)
966 return &pgdat->__lruvec;
969 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
974 static inline bool mm_match_cgroup(struct mm_struct *mm,
975 struct mem_cgroup *memcg)
980 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
985 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
990 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
994 static inline struct mem_cgroup *
995 mem_cgroup_iter(struct mem_cgroup *root,
996 struct mem_cgroup *prev,
997 struct mem_cgroup_reclaim_cookie *reclaim)
1002 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1003 struct mem_cgroup *prev)
1007 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1008 int (*fn)(struct task_struct *, void *), void *arg)
1013 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1018 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1021 /* XXX: This should always return root_mem_cgroup */
1025 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1030 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1035 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1041 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1042 enum lru_list lru, int zone_idx)
1047 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1052 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1058 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1063 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1067 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1072 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1076 static inline void unlock_page_memcg(struct page *page)
1080 static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
1082 /* to match page_memcg_rcu() */
1087 static inline void mem_cgroup_unlock_pages(void)
1092 static inline void mem_cgroup_handle_over_high(void)
1096 static inline void mem_cgroup_enter_user_fault(void)
1100 static inline void mem_cgroup_exit_user_fault(void)
1104 static inline bool task_in_memcg_oom(struct task_struct *p)
1109 static inline bool mem_cgroup_oom_synchronize(bool wait)
1114 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1115 struct task_struct *victim, struct mem_cgroup *oom_domain)
1120 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1124 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1129 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1135 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1141 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1147 static inline void __mod_memcg_page_state(struct page *page,
1153 static inline void mod_memcg_page_state(struct page *page,
1159 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1160 enum node_stat_item idx)
1162 return node_page_state(lruvec_pgdat(lruvec), idx);
1165 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1166 enum node_stat_item idx)
1168 return node_page_state(lruvec_pgdat(lruvec), idx);
1171 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1172 enum node_stat_item idx, int val)
1174 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1177 static inline void mod_lruvec_state(struct lruvec *lruvec,
1178 enum node_stat_item idx, int val)
1180 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1183 static inline void __mod_lruvec_page_state(struct page *page,
1184 enum node_stat_item idx, int val)
1186 __mod_node_page_state(page_pgdat(page), idx, val);
1189 static inline void mod_lruvec_page_state(struct page *page,
1190 enum node_stat_item idx, int val)
1192 mod_node_page_state(page_pgdat(page), idx, val);
1195 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1198 struct page *page = virt_to_head_page(p);
1200 __mod_node_page_state(page_pgdat(page), idx, val);
1203 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1208 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1210 unsigned long *total_scanned)
1215 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1219 static inline void count_memcg_events(struct mem_cgroup *memcg,
1220 enum vm_event_item idx,
1221 unsigned long count)
1225 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1226 enum vm_event_item idx,
1227 unsigned long count)
1231 static inline void count_memcg_page_event(struct page *page,
1237 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1240 #endif /* CONFIG_MEMCG */
1242 /* idx can be of type enum memcg_stat_item or node_stat_item */
1243 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1246 __mod_memcg_state(memcg, idx, 1);
1249 /* idx can be of type enum memcg_stat_item or node_stat_item */
1250 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1253 __mod_memcg_state(memcg, idx, -1);
1256 /* idx can be of type enum memcg_stat_item or node_stat_item */
1257 static inline void __inc_memcg_page_state(struct page *page,
1260 __mod_memcg_page_state(page, idx, 1);
1263 /* idx can be of type enum memcg_stat_item or node_stat_item */
1264 static inline void __dec_memcg_page_state(struct page *page,
1267 __mod_memcg_page_state(page, idx, -1);
1270 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1271 enum node_stat_item idx)
1273 __mod_lruvec_state(lruvec, idx, 1);
1276 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1277 enum node_stat_item idx)
1279 __mod_lruvec_state(lruvec, idx, -1);
1282 static inline void __inc_lruvec_page_state(struct page *page,
1283 enum node_stat_item idx)
1285 __mod_lruvec_page_state(page, idx, 1);
1288 static inline void __dec_lruvec_page_state(struct page *page,
1289 enum node_stat_item idx)
1291 __mod_lruvec_page_state(page, idx, -1);
1294 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1296 __mod_lruvec_slab_state(p, idx, 1);
1299 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1301 __mod_lruvec_slab_state(p, idx, -1);
1304 /* idx can be of type enum memcg_stat_item or node_stat_item */
1305 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1308 mod_memcg_state(memcg, idx, 1);
1311 /* idx can be of type enum memcg_stat_item or node_stat_item */
1312 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1315 mod_memcg_state(memcg, idx, -1);
1318 /* idx can be of type enum memcg_stat_item or node_stat_item */
1319 static inline void inc_memcg_page_state(struct page *page,
1322 mod_memcg_page_state(page, idx, 1);
1325 /* idx can be of type enum memcg_stat_item or node_stat_item */
1326 static inline void dec_memcg_page_state(struct page *page,
1329 mod_memcg_page_state(page, idx, -1);
1332 static inline void inc_lruvec_state(struct lruvec *lruvec,
1333 enum node_stat_item idx)
1335 mod_lruvec_state(lruvec, idx, 1);
1338 static inline void dec_lruvec_state(struct lruvec *lruvec,
1339 enum node_stat_item idx)
1341 mod_lruvec_state(lruvec, idx, -1);
1344 static inline void inc_lruvec_page_state(struct page *page,
1345 enum node_stat_item idx)
1347 mod_lruvec_page_state(page, idx, 1);
1350 static inline void dec_lruvec_page_state(struct page *page,
1351 enum node_stat_item idx)
1353 mod_lruvec_page_state(page, idx, -1);
1356 #ifdef CONFIG_CGROUP_WRITEBACK
1358 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1359 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1360 unsigned long *pheadroom, unsigned long *pdirty,
1361 unsigned long *pwriteback);
1363 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1364 struct bdi_writeback *wb);
1366 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1367 struct bdi_writeback *wb)
1369 if (mem_cgroup_disabled())
1372 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1373 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1376 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1378 #else /* CONFIG_CGROUP_WRITEBACK */
1380 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1385 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1386 unsigned long *pfilepages,
1387 unsigned long *pheadroom,
1388 unsigned long *pdirty,
1389 unsigned long *pwriteback)
1393 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1394 struct bdi_writeback *wb)
1398 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1402 #endif /* CONFIG_CGROUP_WRITEBACK */
1405 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1406 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1408 extern struct static_key_false memcg_sockets_enabled_key;
1409 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1410 void mem_cgroup_sk_alloc(struct sock *sk);
1411 void mem_cgroup_sk_free(struct sock *sk);
1412 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1414 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1417 if (time_before(jiffies, memcg->socket_pressure))
1419 } while ((memcg = parent_mem_cgroup(memcg)));
1423 extern int memcg_expand_shrinker_maps(int new_id);
1425 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1426 int nid, int shrinker_id);
1428 #define mem_cgroup_sockets_enabled 0
1429 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1430 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1431 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1436 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1437 int nid, int shrinker_id)
1442 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1443 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1445 #ifdef CONFIG_MEMCG_KMEM
1446 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1447 void __memcg_kmem_uncharge(struct page *page, int order);
1448 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1449 struct mem_cgroup *memcg);
1450 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1451 unsigned int nr_pages);
1453 extern struct static_key_false memcg_kmem_enabled_key;
1454 extern struct workqueue_struct *memcg_kmem_cache_wq;
1456 extern int memcg_nr_cache_ids;
1457 void memcg_get_cache_ids(void);
1458 void memcg_put_cache_ids(void);
1461 * Helper macro to loop through all memcg-specific caches. Callers must still
1462 * check if the cache is valid (it is either valid or NULL).
1463 * the slab_mutex must be held when looping through those caches
1465 #define for_each_memcg_cache_index(_idx) \
1466 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1468 static inline bool memcg_kmem_enabled(void)
1470 return static_branch_unlikely(&memcg_kmem_enabled_key);
1473 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1475 if (memcg_kmem_enabled())
1476 return __memcg_kmem_charge(page, gfp, order);
1480 static inline void memcg_kmem_uncharge(struct page *page, int order)
1482 if (memcg_kmem_enabled())
1483 __memcg_kmem_uncharge(page, order);
1486 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1487 int order, struct mem_cgroup *memcg)
1489 if (memcg_kmem_enabled())
1490 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1494 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1495 struct mem_cgroup *memcg)
1497 if (memcg_kmem_enabled())
1498 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1502 * helper for accessing a memcg's index. It will be used as an index in the
1503 * child cache array in kmem_cache, and also to derive its name. This function
1504 * will return -1 when this is not a kmem-limited memcg.
1506 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1508 return memcg ? memcg->kmemcg_id : -1;
1511 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1515 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1520 static inline void memcg_kmem_uncharge(struct page *page, int order)
1524 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1529 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1533 #define for_each_memcg_cache_index(_idx) \
1536 static inline bool memcg_kmem_enabled(void)
1541 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1546 static inline void memcg_get_cache_ids(void)
1550 static inline void memcg_put_cache_ids(void)
1554 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1559 #endif /* CONFIG_MEMCG_KMEM */
1561 #endif /* _LINUX_MEMCONTROL_H */