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 enum mem_cgroup_protection {
59 struct mem_cgroup_reclaim_cookie {
62 unsigned int generation;
67 #define MEM_CGROUP_ID_SHIFT 16
68 #define MEM_CGROUP_ID_MAX USHRT_MAX
70 struct mem_cgroup_id {
76 * Per memcg event counter is incremented at every pagein/pageout. With THP,
77 * it will be incremated by the number of pages. This counter is used for
78 * for trigger some periodic events. This is straightforward and better
79 * than using jiffies etc. to handle periodic memcg event.
81 enum mem_cgroup_events_target {
82 MEM_CGROUP_TARGET_THRESH,
83 MEM_CGROUP_TARGET_SOFTLIMIT,
84 MEM_CGROUP_TARGET_NUMAINFO,
88 struct memcg_vmstats_percpu {
89 long stat[MEMCG_NR_STAT];
90 unsigned long events[NR_VM_EVENT_ITEMS];
91 unsigned long nr_page_events;
92 unsigned long targets[MEM_CGROUP_NTARGETS];
95 struct mem_cgroup_reclaim_iter {
96 struct mem_cgroup *position;
97 /* scan generation, increased every round-trip */
98 unsigned int generation;
102 long count[NR_VM_NODE_STAT_ITEMS];
106 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
107 * which have elements charged to this memcg.
109 struct memcg_shrinker_map {
111 unsigned long map[0];
115 * per-zone information in memory controller.
117 struct mem_cgroup_per_node {
118 struct lruvec lruvec;
120 /* Legacy local VM stats */
121 struct lruvec_stat __percpu *lruvec_stat_local;
123 /* Subtree VM stats (batched updates) */
124 struct lruvec_stat __percpu *lruvec_stat_cpu;
125 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
127 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
129 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
131 #ifdef CONFIG_MEMCG_KMEM
132 struct memcg_shrinker_map __rcu *shrinker_map;
134 struct rb_node tree_node; /* RB tree node */
135 unsigned long usage_in_excess;/* Set to the value by which */
136 /* the soft limit is exceeded*/
138 bool congested; /* memcg has many dirty pages */
139 /* backed by a congested BDI */
141 struct mem_cgroup *memcg; /* Back pointer, we cannot */
142 /* use container_of */
145 struct mem_cgroup_threshold {
146 struct eventfd_ctx *eventfd;
147 unsigned long threshold;
151 struct mem_cgroup_threshold_ary {
152 /* An array index points to threshold just below or equal to usage. */
153 int current_threshold;
154 /* Size of entries[] */
156 /* Array of thresholds */
157 struct mem_cgroup_threshold entries[0];
160 struct mem_cgroup_thresholds {
161 /* Primary thresholds array */
162 struct mem_cgroup_threshold_ary *primary;
164 * Spare threshold array.
165 * This is needed to make mem_cgroup_unregister_event() "never fail".
166 * It must be able to store at least primary->size - 1 entries.
168 struct mem_cgroup_threshold_ary *spare;
171 enum memcg_kmem_state {
177 #if defined(CONFIG_SMP)
178 struct memcg_padding {
180 } ____cacheline_internodealigned_in_smp;
181 #define MEMCG_PADDING(name) struct memcg_padding name;
183 #define MEMCG_PADDING(name)
187 * Remember four most recent foreign writebacks with dirty pages in this
188 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
189 * one in a given round, we're likely to catch it later if it keeps
190 * foreign-dirtying, so a fairly low count should be enough.
192 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
194 #define MEMCG_CGWB_FRN_CNT 4
196 struct memcg_cgwb_frn {
197 u64 bdi_id; /* bdi->id of the foreign inode */
198 int memcg_id; /* memcg->css.id of foreign inode */
199 u64 at; /* jiffies_64 at the time of dirtying */
200 struct wb_completion done; /* tracks in-flight foreign writebacks */
204 * The memory controller data structure. The memory controller controls both
205 * page cache and RSS per cgroup. We would eventually like to provide
206 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
207 * to help the administrator determine what knobs to tune.
210 struct cgroup_subsys_state css;
212 /* Private memcg ID. Used to ID objects that outlive the cgroup */
213 struct mem_cgroup_id id;
215 /* Accounted resources */
216 struct page_counter memory;
217 struct page_counter swap;
219 /* Legacy consumer-oriented counters */
220 struct page_counter memsw;
221 struct page_counter kmem;
222 struct page_counter tcpmem;
224 /* Upper bound of normal memory consumption range */
227 /* Range enforcement for interrupt charges */
228 struct work_struct high_work;
230 unsigned long soft_limit;
232 /* vmpressure notifications */
233 struct vmpressure vmpressure;
236 * Should the accounting and control be hierarchical, per subtree?
241 * Should the OOM killer kill all belonging tasks, had it kill one?
245 /* protected by memcg_oom_lock */
250 /* OOM-Killer disable */
251 int oom_kill_disable;
253 /* memory.events and memory.events.local */
254 struct cgroup_file events_file;
255 struct cgroup_file events_local_file;
257 /* handle for "memory.swap.events" */
258 struct cgroup_file swap_events_file;
260 /* protect arrays of thresholds */
261 struct mutex thresholds_lock;
263 /* thresholds for memory usage. RCU-protected */
264 struct mem_cgroup_thresholds thresholds;
266 /* thresholds for mem+swap usage. RCU-protected */
267 struct mem_cgroup_thresholds memsw_thresholds;
269 /* For oom notifier event fd */
270 struct list_head oom_notify;
273 * Should we move charges of a task when a task is moved into this
274 * mem_cgroup ? And what type of charges should we move ?
276 unsigned long move_charge_at_immigrate;
277 /* taken only while moving_account > 0 */
278 spinlock_t move_lock;
279 unsigned long move_lock_flags;
281 MEMCG_PADDING(_pad1_);
284 * set > 0 if pages under this cgroup are moving to other cgroup.
286 atomic_t moving_account;
287 struct task_struct *move_lock_task;
289 /* Legacy local VM stats and events */
290 struct memcg_vmstats_percpu __percpu *vmstats_local;
292 /* Subtree VM stats and events (batched updates) */
293 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
295 MEMCG_PADDING(_pad2_);
297 atomic_long_t vmstats[MEMCG_NR_STAT];
298 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
301 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
302 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
304 unsigned long socket_pressure;
306 /* Legacy tcp memory accounting */
310 #ifdef CONFIG_MEMCG_KMEM
311 /* Index in the kmem_cache->memcg_params.memcg_caches array */
313 enum memcg_kmem_state kmem_state;
314 struct list_head kmem_caches;
317 int last_scanned_node;
319 nodemask_t scan_nodes;
320 atomic_t numainfo_events;
321 atomic_t numainfo_updating;
324 #ifdef CONFIG_CGROUP_WRITEBACK
325 struct list_head cgwb_list;
326 struct wb_domain cgwb_domain;
327 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
330 /* List of events which userspace want to receive */
331 struct list_head event_list;
332 spinlock_t event_list_lock;
334 struct mem_cgroup_per_node *nodeinfo[0];
335 /* WARNING: nodeinfo must be the last member here */
339 * size of first charge trial. "32" comes from vmscan.c's magic value.
340 * TODO: maybe necessary to use big numbers in big irons.
342 #define MEMCG_CHARGE_BATCH 32U
344 extern struct mem_cgroup *root_mem_cgroup;
346 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
348 return (memcg == root_mem_cgroup);
351 static inline bool mem_cgroup_disabled(void)
353 return !cgroup_subsys_enabled(memory_cgrp_subsys);
356 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
357 struct mem_cgroup *memcg);
359 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
360 gfp_t gfp_mask, struct mem_cgroup **memcgp,
362 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
363 gfp_t gfp_mask, struct mem_cgroup **memcgp,
365 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
366 bool lrucare, bool compound);
367 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
369 void mem_cgroup_uncharge(struct page *page);
370 void mem_cgroup_uncharge_list(struct list_head *page_list);
372 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
374 static struct mem_cgroup_per_node *
375 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
377 return memcg->nodeinfo[nid];
381 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
382 * @node: node of the wanted lruvec
383 * @memcg: memcg of the wanted lruvec
385 * Returns the lru list vector holding pages for a given @node or a given
386 * @memcg and @zone. This can be the node lruvec, if the memory controller
389 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
390 struct mem_cgroup *memcg)
392 struct mem_cgroup_per_node *mz;
393 struct lruvec *lruvec;
395 if (mem_cgroup_disabled()) {
396 lruvec = node_lruvec(pgdat);
400 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
401 lruvec = &mz->lruvec;
404 * Since a node can be onlined after the mem_cgroup was created,
405 * we have to be prepared to initialize lruvec->pgdat here;
406 * and if offlined then reonlined, we need to reinitialize it.
408 if (unlikely(lruvec->pgdat != pgdat))
409 lruvec->pgdat = pgdat;
413 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
415 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
417 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
419 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
422 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
423 return css ? container_of(css, struct mem_cgroup, css) : NULL;
426 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
429 css_put(&memcg->css);
432 #define mem_cgroup_from_counter(counter, member) \
433 container_of(counter, struct mem_cgroup, member)
435 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
437 struct mem_cgroup_reclaim_cookie *);
438 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
439 int mem_cgroup_scan_tasks(struct mem_cgroup *,
440 int (*)(struct task_struct *, void *), void *);
442 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
444 if (mem_cgroup_disabled())
449 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
451 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
453 return mem_cgroup_from_css(seq_css(m));
456 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
458 struct mem_cgroup_per_node *mz;
460 if (mem_cgroup_disabled())
463 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
468 * parent_mem_cgroup - find the accounting parent of a memcg
469 * @memcg: memcg whose parent to find
471 * Returns the parent memcg, or NULL if this is the root or the memory
472 * controller is in legacy no-hierarchy mode.
474 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
476 if (!memcg->memory.parent)
478 return mem_cgroup_from_counter(memcg->memory.parent, memory);
481 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
482 struct mem_cgroup *root)
486 if (!root->use_hierarchy)
488 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
491 static inline bool mm_match_cgroup(struct mm_struct *mm,
492 struct mem_cgroup *memcg)
494 struct mem_cgroup *task_memcg;
498 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
500 match = mem_cgroup_is_descendant(task_memcg, memcg);
505 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
506 ino_t page_cgroup_ino(struct page *page);
508 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
510 if (mem_cgroup_disabled())
512 return !!(memcg->css.flags & CSS_ONLINE);
516 * For memory reclaim.
518 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
520 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
521 int zid, int nr_pages);
524 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
525 enum lru_list lru, int zone_idx)
527 struct mem_cgroup_per_node *mz;
529 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
530 return mz->lru_zone_size[zone_idx][lru];
533 void mem_cgroup_handle_over_high(void);
535 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
537 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
538 struct task_struct *p);
540 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
542 static inline void mem_cgroup_enter_user_fault(void)
544 WARN_ON(current->in_user_fault);
545 current->in_user_fault = 1;
548 static inline void mem_cgroup_exit_user_fault(void)
550 WARN_ON(!current->in_user_fault);
551 current->in_user_fault = 0;
554 static inline bool task_in_memcg_oom(struct task_struct *p)
556 return p->memcg_in_oom;
559 bool mem_cgroup_oom_synchronize(bool wait);
560 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
561 struct mem_cgroup *oom_domain);
562 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
564 #ifdef CONFIG_MEMCG_SWAP
565 extern int do_swap_account;
568 struct mem_cgroup *lock_page_memcg(struct page *page);
569 void __unlock_page_memcg(struct mem_cgroup *memcg);
570 void unlock_page_memcg(struct page *page);
573 * idx can be of type enum memcg_stat_item or node_stat_item.
574 * Keep in sync with memcg_exact_page_state().
576 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
578 long x = atomic_long_read(&memcg->vmstats[idx]);
587 * idx can be of type enum memcg_stat_item or node_stat_item.
588 * Keep in sync with memcg_exact_page_state().
590 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
596 for_each_possible_cpu(cpu)
597 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
605 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
607 /* idx can be of type enum memcg_stat_item or node_stat_item */
608 static inline void mod_memcg_state(struct mem_cgroup *memcg,
613 local_irq_save(flags);
614 __mod_memcg_state(memcg, idx, val);
615 local_irq_restore(flags);
619 * mod_memcg_page_state - update page state statistics
621 * @idx: page state item to account
622 * @val: number of pages (positive or negative)
624 * The @page must be locked or the caller must use lock_page_memcg()
625 * to prevent double accounting when the page is concurrently being
626 * moved to another memcg:
628 * lock_page(page) or lock_page_memcg(page)
629 * if (TestClearPageState(page))
630 * mod_memcg_page_state(page, state, -1);
631 * unlock_page(page) or unlock_page_memcg(page)
633 * Kernel pages are an exception to this, since they'll never move.
635 static inline void __mod_memcg_page_state(struct page *page,
638 if (page->mem_cgroup)
639 __mod_memcg_state(page->mem_cgroup, idx, val);
642 static inline void mod_memcg_page_state(struct page *page,
645 if (page->mem_cgroup)
646 mod_memcg_state(page->mem_cgroup, idx, val);
649 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
650 enum node_stat_item idx)
652 struct mem_cgroup_per_node *pn;
655 if (mem_cgroup_disabled())
656 return node_page_state(lruvec_pgdat(lruvec), idx);
658 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
659 x = atomic_long_read(&pn->lruvec_stat[idx]);
667 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
668 enum node_stat_item idx)
670 struct mem_cgroup_per_node *pn;
674 if (mem_cgroup_disabled())
675 return node_page_state(lruvec_pgdat(lruvec), idx);
677 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
678 for_each_possible_cpu(cpu)
679 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
687 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
689 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
691 static inline void mod_lruvec_state(struct lruvec *lruvec,
692 enum node_stat_item idx, int val)
696 local_irq_save(flags);
697 __mod_lruvec_state(lruvec, idx, val);
698 local_irq_restore(flags);
701 static inline void __mod_lruvec_page_state(struct page *page,
702 enum node_stat_item idx, int val)
704 pg_data_t *pgdat = page_pgdat(page);
705 struct lruvec *lruvec;
707 /* Untracked pages have no memcg, no lruvec. Update only the node */
708 if (!page->mem_cgroup) {
709 __mod_node_page_state(pgdat, idx, val);
713 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
714 __mod_lruvec_state(lruvec, idx, val);
717 static inline void mod_lruvec_page_state(struct page *page,
718 enum node_stat_item idx, int val)
722 local_irq_save(flags);
723 __mod_lruvec_page_state(page, idx, val);
724 local_irq_restore(flags);
727 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
729 unsigned long *total_scanned);
731 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
732 unsigned long count);
734 static inline void count_memcg_events(struct mem_cgroup *memcg,
735 enum vm_event_item idx,
740 local_irq_save(flags);
741 __count_memcg_events(memcg, idx, count);
742 local_irq_restore(flags);
745 static inline void count_memcg_page_event(struct page *page,
746 enum vm_event_item idx)
748 if (page->mem_cgroup)
749 count_memcg_events(page->mem_cgroup, idx, 1);
752 static inline void count_memcg_event_mm(struct mm_struct *mm,
753 enum vm_event_item idx)
755 struct mem_cgroup *memcg;
757 if (mem_cgroup_disabled())
761 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
763 count_memcg_events(memcg, idx, 1);
767 static inline void memcg_memory_event(struct mem_cgroup *memcg,
768 enum memcg_memory_event event)
770 atomic_long_inc(&memcg->memory_events_local[event]);
771 cgroup_file_notify(&memcg->events_local_file);
774 atomic_long_inc(&memcg->memory_events[event]);
775 cgroup_file_notify(&memcg->events_file);
777 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
779 } while ((memcg = parent_mem_cgroup(memcg)) &&
780 !mem_cgroup_is_root(memcg));
783 static inline void memcg_memory_event_mm(struct mm_struct *mm,
784 enum memcg_memory_event event)
786 struct mem_cgroup *memcg;
788 if (mem_cgroup_disabled())
792 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
794 memcg_memory_event(memcg, event);
798 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
799 void mem_cgroup_split_huge_fixup(struct page *head);
802 #else /* CONFIG_MEMCG */
804 #define MEM_CGROUP_ID_SHIFT 0
805 #define MEM_CGROUP_ID_MAX 0
809 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
814 static inline bool mem_cgroup_disabled(void)
819 static inline void memcg_memory_event(struct mem_cgroup *memcg,
820 enum memcg_memory_event event)
824 static inline void memcg_memory_event_mm(struct mm_struct *mm,
825 enum memcg_memory_event event)
829 static inline enum mem_cgroup_protection mem_cgroup_protected(
830 struct mem_cgroup *root, struct mem_cgroup *memcg)
832 return MEMCG_PROT_NONE;
835 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
837 struct mem_cgroup **memcgp,
844 static inline int mem_cgroup_try_charge_delay(struct page *page,
845 struct mm_struct *mm,
847 struct mem_cgroup **memcgp,
854 static inline void mem_cgroup_commit_charge(struct page *page,
855 struct mem_cgroup *memcg,
856 bool lrucare, bool compound)
860 static inline void mem_cgroup_cancel_charge(struct page *page,
861 struct mem_cgroup *memcg,
866 static inline void mem_cgroup_uncharge(struct page *page)
870 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
874 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
878 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
879 struct mem_cgroup *memcg)
881 return node_lruvec(pgdat);
884 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
885 struct pglist_data *pgdat)
887 return &pgdat->lruvec;
890 static inline bool mm_match_cgroup(struct mm_struct *mm,
891 struct mem_cgroup *memcg)
896 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
901 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
906 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
910 static inline struct mem_cgroup *
911 mem_cgroup_iter(struct mem_cgroup *root,
912 struct mem_cgroup *prev,
913 struct mem_cgroup_reclaim_cookie *reclaim)
918 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
919 struct mem_cgroup *prev)
923 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
924 int (*fn)(struct task_struct *, void *), void *arg)
929 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
934 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
937 /* XXX: This should always return root_mem_cgroup */
941 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
946 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
951 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
957 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
958 enum lru_list lru, int zone_idx)
963 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
969 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
974 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
978 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
983 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
987 static inline void unlock_page_memcg(struct page *page)
991 static inline void mem_cgroup_handle_over_high(void)
995 static inline void mem_cgroup_enter_user_fault(void)
999 static inline void mem_cgroup_exit_user_fault(void)
1003 static inline bool task_in_memcg_oom(struct task_struct *p)
1008 static inline bool mem_cgroup_oom_synchronize(bool wait)
1013 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1014 struct task_struct *victim, struct mem_cgroup *oom_domain)
1019 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1023 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1028 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1034 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1040 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1046 static inline void __mod_memcg_page_state(struct page *page,
1052 static inline void mod_memcg_page_state(struct page *page,
1058 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1059 enum node_stat_item idx)
1061 return node_page_state(lruvec_pgdat(lruvec), idx);
1064 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1065 enum node_stat_item idx)
1067 return node_page_state(lruvec_pgdat(lruvec), idx);
1070 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1071 enum node_stat_item idx, int val)
1073 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1076 static inline void mod_lruvec_state(struct lruvec *lruvec,
1077 enum node_stat_item idx, int val)
1079 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1082 static inline void __mod_lruvec_page_state(struct page *page,
1083 enum node_stat_item idx, int val)
1085 __mod_node_page_state(page_pgdat(page), idx, val);
1088 static inline void mod_lruvec_page_state(struct page *page,
1089 enum node_stat_item idx, int val)
1091 mod_node_page_state(page_pgdat(page), idx, val);
1094 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1097 struct page *page = virt_to_head_page(p);
1099 __mod_node_page_state(page_pgdat(page), idx, val);
1103 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1105 unsigned long *total_scanned)
1110 static inline void mem_cgroup_split_huge_fixup(struct page *head)
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_events(struct mem_cgroup *memcg,
1121 enum vm_event_item idx,
1122 unsigned long count)
1126 static inline void count_memcg_page_event(struct page *page,
1132 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1135 #endif /* CONFIG_MEMCG */
1137 /* idx can be of type enum memcg_stat_item or node_stat_item */
1138 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1141 __mod_memcg_state(memcg, idx, 1);
1144 /* idx can be of type enum memcg_stat_item or node_stat_item */
1145 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1148 __mod_memcg_state(memcg, idx, -1);
1151 /* idx can be of type enum memcg_stat_item or node_stat_item */
1152 static inline void __inc_memcg_page_state(struct page *page,
1155 __mod_memcg_page_state(page, idx, 1);
1158 /* idx can be of type enum memcg_stat_item or node_stat_item */
1159 static inline void __dec_memcg_page_state(struct page *page,
1162 __mod_memcg_page_state(page, idx, -1);
1165 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1166 enum node_stat_item idx)
1168 __mod_lruvec_state(lruvec, idx, 1);
1171 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1172 enum node_stat_item idx)
1174 __mod_lruvec_state(lruvec, idx, -1);
1177 static inline void __inc_lruvec_page_state(struct page *page,
1178 enum node_stat_item idx)
1180 __mod_lruvec_page_state(page, idx, 1);
1183 static inline void __dec_lruvec_page_state(struct page *page,
1184 enum node_stat_item idx)
1186 __mod_lruvec_page_state(page, idx, -1);
1189 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1191 __mod_lruvec_slab_state(p, idx, 1);
1194 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1196 __mod_lruvec_slab_state(p, idx, -1);
1199 /* idx can be of type enum memcg_stat_item or node_stat_item */
1200 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1203 mod_memcg_state(memcg, idx, 1);
1206 /* idx can be of type enum memcg_stat_item or node_stat_item */
1207 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1210 mod_memcg_state(memcg, idx, -1);
1213 /* idx can be of type enum memcg_stat_item or node_stat_item */
1214 static inline void inc_memcg_page_state(struct page *page,
1217 mod_memcg_page_state(page, idx, 1);
1220 /* idx can be of type enum memcg_stat_item or node_stat_item */
1221 static inline void dec_memcg_page_state(struct page *page,
1224 mod_memcg_page_state(page, idx, -1);
1227 static inline void inc_lruvec_state(struct lruvec *lruvec,
1228 enum node_stat_item idx)
1230 mod_lruvec_state(lruvec, idx, 1);
1233 static inline void dec_lruvec_state(struct lruvec *lruvec,
1234 enum node_stat_item idx)
1236 mod_lruvec_state(lruvec, idx, -1);
1239 static inline void inc_lruvec_page_state(struct page *page,
1240 enum node_stat_item idx)
1242 mod_lruvec_page_state(page, idx, 1);
1245 static inline void dec_lruvec_page_state(struct page *page,
1246 enum node_stat_item idx)
1248 mod_lruvec_page_state(page, idx, -1);
1251 #ifdef CONFIG_CGROUP_WRITEBACK
1253 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1254 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1255 unsigned long *pheadroom, unsigned long *pdirty,
1256 unsigned long *pwriteback);
1258 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1259 struct bdi_writeback *wb);
1261 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1262 struct bdi_writeback *wb)
1264 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1265 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1268 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1270 #else /* CONFIG_CGROUP_WRITEBACK */
1272 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1277 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1278 unsigned long *pfilepages,
1279 unsigned long *pheadroom,
1280 unsigned long *pdirty,
1281 unsigned long *pwriteback)
1285 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1286 struct bdi_writeback *wb)
1290 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1294 #endif /* CONFIG_CGROUP_WRITEBACK */
1297 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1298 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1300 extern struct static_key_false memcg_sockets_enabled_key;
1301 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1302 void mem_cgroup_sk_alloc(struct sock *sk);
1303 void mem_cgroup_sk_free(struct sock *sk);
1304 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1306 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1309 if (time_before(jiffies, memcg->socket_pressure))
1311 } while ((memcg = parent_mem_cgroup(memcg)));
1315 #define mem_cgroup_sockets_enabled 0
1316 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1317 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1318 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1324 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1325 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1327 #ifdef CONFIG_MEMCG_KMEM
1328 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1329 void __memcg_kmem_uncharge(struct page *page, int order);
1330 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1331 struct mem_cgroup *memcg);
1332 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1333 unsigned int nr_pages);
1335 extern struct static_key_false memcg_kmem_enabled_key;
1336 extern struct workqueue_struct *memcg_kmem_cache_wq;
1338 extern int memcg_nr_cache_ids;
1339 void memcg_get_cache_ids(void);
1340 void memcg_put_cache_ids(void);
1343 * Helper macro to loop through all memcg-specific caches. Callers must still
1344 * check if the cache is valid (it is either valid or NULL).
1345 * the slab_mutex must be held when looping through those caches
1347 #define for_each_memcg_cache_index(_idx) \
1348 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1350 static inline bool memcg_kmem_enabled(void)
1352 return static_branch_unlikely(&memcg_kmem_enabled_key);
1355 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1357 if (memcg_kmem_enabled())
1358 return __memcg_kmem_charge(page, gfp, order);
1362 static inline void memcg_kmem_uncharge(struct page *page, int order)
1364 if (memcg_kmem_enabled())
1365 __memcg_kmem_uncharge(page, order);
1368 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1369 int order, struct mem_cgroup *memcg)
1371 if (memcg_kmem_enabled())
1372 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1376 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1377 struct mem_cgroup *memcg)
1379 if (memcg_kmem_enabled())
1380 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1384 * helper for accessing a memcg's index. It will be used as an index in the
1385 * child cache array in kmem_cache, and also to derive its name. This function
1386 * will return -1 when this is not a kmem-limited memcg.
1388 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1390 return memcg ? memcg->kmemcg_id : -1;
1393 extern int memcg_expand_shrinker_maps(int new_id);
1395 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1396 int nid, int shrinker_id);
1399 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1404 static inline void memcg_kmem_uncharge(struct page *page, int order)
1408 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1413 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1417 #define for_each_memcg_cache_index(_idx) \
1420 static inline bool memcg_kmem_enabled(void)
1425 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1430 static inline void memcg_get_cache_ids(void)
1434 static inline void memcg_put_cache_ids(void)
1438 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1439 int nid, int shrinker_id) { }
1440 #endif /* CONFIG_MEMCG_KMEM */
1442 #endif /* _LINUX_MEMCONTROL_H */