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 struct memcg_shrinker_map __rcu *shrinker_map;
133 struct rb_node tree_node; /* RB tree node */
134 unsigned long usage_in_excess;/* Set to the value by which */
135 /* the soft limit is exceeded*/
137 bool congested; /* memcg has many dirty pages */
138 /* backed by a congested BDI */
140 struct mem_cgroup *memcg; /* Back pointer, we cannot */
141 /* use container_of */
144 struct mem_cgroup_threshold {
145 struct eventfd_ctx *eventfd;
146 unsigned long threshold;
150 struct mem_cgroup_threshold_ary {
151 /* An array index points to threshold just below or equal to usage. */
152 int current_threshold;
153 /* Size of entries[] */
155 /* Array of thresholds */
156 struct mem_cgroup_threshold entries[0];
159 struct mem_cgroup_thresholds {
160 /* Primary thresholds array */
161 struct mem_cgroup_threshold_ary *primary;
163 * Spare threshold array.
164 * This is needed to make mem_cgroup_unregister_event() "never fail".
165 * It must be able to store at least primary->size - 1 entries.
167 struct mem_cgroup_threshold_ary *spare;
170 enum memcg_kmem_state {
176 #if defined(CONFIG_SMP)
177 struct memcg_padding {
179 } ____cacheline_internodealigned_in_smp;
180 #define MEMCG_PADDING(name) struct memcg_padding name;
182 #define MEMCG_PADDING(name)
186 * Remember four most recent foreign writebacks with dirty pages in this
187 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
188 * one in a given round, we're likely to catch it later if it keeps
189 * foreign-dirtying, so a fairly low count should be enough.
191 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
193 #define MEMCG_CGWB_FRN_CNT 4
195 struct memcg_cgwb_frn {
196 u64 bdi_id; /* bdi->id of the foreign inode */
197 int memcg_id; /* memcg->css.id of foreign inode */
198 u64 at; /* jiffies_64 at the time of dirtying */
199 struct wb_completion done; /* tracks in-flight foreign writebacks */
203 * The memory controller data structure. The memory controller controls both
204 * page cache and RSS per cgroup. We would eventually like to provide
205 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
206 * to help the administrator determine what knobs to tune.
209 struct cgroup_subsys_state css;
211 /* Private memcg ID. Used to ID objects that outlive the cgroup */
212 struct mem_cgroup_id id;
214 /* Accounted resources */
215 struct page_counter memory;
216 struct page_counter swap;
218 /* Legacy consumer-oriented counters */
219 struct page_counter memsw;
220 struct page_counter kmem;
221 struct page_counter tcpmem;
223 /* Upper bound of normal memory consumption range */
226 /* Range enforcement for interrupt charges */
227 struct work_struct high_work;
229 unsigned long soft_limit;
231 /* vmpressure notifications */
232 struct vmpressure vmpressure;
235 * Should the accounting and control be hierarchical, per subtree?
240 * Should the OOM killer kill all belonging tasks, had it kill one?
244 /* protected by memcg_oom_lock */
249 /* OOM-Killer disable */
250 int oom_kill_disable;
252 /* memory.events and memory.events.local */
253 struct cgroup_file events_file;
254 struct cgroup_file events_local_file;
256 /* handle for "memory.swap.events" */
257 struct cgroup_file swap_events_file;
259 /* protect arrays of thresholds */
260 struct mutex thresholds_lock;
262 /* thresholds for memory usage. RCU-protected */
263 struct mem_cgroup_thresholds thresholds;
265 /* thresholds for mem+swap usage. RCU-protected */
266 struct mem_cgroup_thresholds memsw_thresholds;
268 /* For oom notifier event fd */
269 struct list_head oom_notify;
272 * Should we move charges of a task when a task is moved into this
273 * mem_cgroup ? And what type of charges should we move ?
275 unsigned long move_charge_at_immigrate;
276 /* taken only while moving_account > 0 */
277 spinlock_t move_lock;
278 unsigned long move_lock_flags;
280 MEMCG_PADDING(_pad1_);
283 * set > 0 if pages under this cgroup are moving to other cgroup.
285 atomic_t moving_account;
286 struct task_struct *move_lock_task;
288 /* Legacy local VM stats and events */
289 struct memcg_vmstats_percpu __percpu *vmstats_local;
291 /* Subtree VM stats and events (batched updates) */
292 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
294 MEMCG_PADDING(_pad2_);
296 atomic_long_t vmstats[MEMCG_NR_STAT];
297 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
300 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
301 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
303 unsigned long socket_pressure;
305 /* Legacy tcp memory accounting */
309 #ifdef CONFIG_MEMCG_KMEM
310 /* Index in the kmem_cache->memcg_params.memcg_caches array */
312 enum memcg_kmem_state kmem_state;
313 struct list_head kmem_caches;
316 int last_scanned_node;
318 nodemask_t scan_nodes;
319 atomic_t numainfo_events;
320 atomic_t numainfo_updating;
323 #ifdef CONFIG_CGROUP_WRITEBACK
324 struct list_head cgwb_list;
325 struct wb_domain cgwb_domain;
326 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
329 /* List of events which userspace want to receive */
330 struct list_head event_list;
331 spinlock_t event_list_lock;
333 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
334 struct deferred_split deferred_split_queue;
337 struct mem_cgroup_per_node *nodeinfo[0];
338 /* WARNING: nodeinfo must be the last member here */
342 * size of first charge trial. "32" comes from vmscan.c's magic value.
343 * TODO: maybe necessary to use big numbers in big irons.
345 #define MEMCG_CHARGE_BATCH 32U
347 extern struct mem_cgroup *root_mem_cgroup;
349 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
351 return (memcg == root_mem_cgroup);
354 static inline bool mem_cgroup_disabled(void)
356 return !cgroup_subsys_enabled(memory_cgrp_subsys);
359 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
360 struct mem_cgroup *memcg);
362 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
363 gfp_t gfp_mask, struct mem_cgroup **memcgp,
365 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
366 gfp_t gfp_mask, struct mem_cgroup **memcgp,
368 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
369 bool lrucare, bool compound);
370 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
372 void mem_cgroup_uncharge(struct page *page);
373 void mem_cgroup_uncharge_list(struct list_head *page_list);
375 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
377 static struct mem_cgroup_per_node *
378 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
380 return memcg->nodeinfo[nid];
384 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
385 * @node: node of the wanted lruvec
386 * @memcg: memcg of the wanted lruvec
388 * Returns the lru list vector holding pages for a given @node or a given
389 * @memcg and @zone. This can be the node lruvec, if the memory controller
392 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
393 struct mem_cgroup *memcg)
395 struct mem_cgroup_per_node *mz;
396 struct lruvec *lruvec;
398 if (mem_cgroup_disabled()) {
399 lruvec = node_lruvec(pgdat);
403 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
404 lruvec = &mz->lruvec;
407 * Since a node can be onlined after the mem_cgroup was created,
408 * we have to be prepared to initialize lruvec->pgdat here;
409 * and if offlined then reonlined, we need to reinitialize it.
411 if (unlikely(lruvec->pgdat != pgdat))
412 lruvec->pgdat = pgdat;
416 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
418 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
420 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
422 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
425 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
426 return css ? container_of(css, struct mem_cgroup, css) : NULL;
429 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
432 css_put(&memcg->css);
435 #define mem_cgroup_from_counter(counter, member) \
436 container_of(counter, struct mem_cgroup, member)
438 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
440 struct mem_cgroup_reclaim_cookie *);
441 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
442 int mem_cgroup_scan_tasks(struct mem_cgroup *,
443 int (*)(struct task_struct *, void *), void *);
445 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
447 if (mem_cgroup_disabled())
452 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
454 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
456 return mem_cgroup_from_css(seq_css(m));
459 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
461 struct mem_cgroup_per_node *mz;
463 if (mem_cgroup_disabled())
466 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
471 * parent_mem_cgroup - find the accounting parent of a memcg
472 * @memcg: memcg whose parent to find
474 * Returns the parent memcg, or NULL if this is the root or the memory
475 * controller is in legacy no-hierarchy mode.
477 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
479 if (!memcg->memory.parent)
481 return mem_cgroup_from_counter(memcg->memory.parent, memory);
484 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
485 struct mem_cgroup *root)
489 if (!root->use_hierarchy)
491 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
494 static inline bool mm_match_cgroup(struct mm_struct *mm,
495 struct mem_cgroup *memcg)
497 struct mem_cgroup *task_memcg;
501 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
503 match = mem_cgroup_is_descendant(task_memcg, memcg);
508 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
509 ino_t page_cgroup_ino(struct page *page);
511 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
513 if (mem_cgroup_disabled())
515 return !!(memcg->css.flags & CSS_ONLINE);
519 * For memory reclaim.
521 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
523 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
524 int zid, int nr_pages);
527 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
528 enum lru_list lru, int zone_idx)
530 struct mem_cgroup_per_node *mz;
532 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
533 return mz->lru_zone_size[zone_idx][lru];
536 void mem_cgroup_handle_over_high(void);
538 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
540 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
541 struct task_struct *p);
543 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
545 static inline void mem_cgroup_enter_user_fault(void)
547 WARN_ON(current->in_user_fault);
548 current->in_user_fault = 1;
551 static inline void mem_cgroup_exit_user_fault(void)
553 WARN_ON(!current->in_user_fault);
554 current->in_user_fault = 0;
557 static inline bool task_in_memcg_oom(struct task_struct *p)
559 return p->memcg_in_oom;
562 bool mem_cgroup_oom_synchronize(bool wait);
563 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
564 struct mem_cgroup *oom_domain);
565 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
567 #ifdef CONFIG_MEMCG_SWAP
568 extern int do_swap_account;
571 struct mem_cgroup *lock_page_memcg(struct page *page);
572 void __unlock_page_memcg(struct mem_cgroup *memcg);
573 void unlock_page_memcg(struct page *page);
576 * idx can be of type enum memcg_stat_item or node_stat_item.
577 * Keep in sync with memcg_exact_page_state().
579 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
581 long x = atomic_long_read(&memcg->vmstats[idx]);
590 * idx can be of type enum memcg_stat_item or node_stat_item.
591 * Keep in sync with memcg_exact_page_state().
593 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
599 for_each_possible_cpu(cpu)
600 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
608 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
610 /* idx can be of type enum memcg_stat_item or node_stat_item */
611 static inline void mod_memcg_state(struct mem_cgroup *memcg,
616 local_irq_save(flags);
617 __mod_memcg_state(memcg, idx, val);
618 local_irq_restore(flags);
622 * mod_memcg_page_state - update page state statistics
624 * @idx: page state item to account
625 * @val: number of pages (positive or negative)
627 * The @page must be locked or the caller must use lock_page_memcg()
628 * to prevent double accounting when the page is concurrently being
629 * moved to another memcg:
631 * lock_page(page) or lock_page_memcg(page)
632 * if (TestClearPageState(page))
633 * mod_memcg_page_state(page, state, -1);
634 * unlock_page(page) or unlock_page_memcg(page)
636 * Kernel pages are an exception to this, since they'll never move.
638 static inline void __mod_memcg_page_state(struct page *page,
641 if (page->mem_cgroup)
642 __mod_memcg_state(page->mem_cgroup, idx, val);
645 static inline void mod_memcg_page_state(struct page *page,
648 if (page->mem_cgroup)
649 mod_memcg_state(page->mem_cgroup, idx, val);
652 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
653 enum node_stat_item idx)
655 struct mem_cgroup_per_node *pn;
658 if (mem_cgroup_disabled())
659 return node_page_state(lruvec_pgdat(lruvec), idx);
661 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
662 x = atomic_long_read(&pn->lruvec_stat[idx]);
670 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
671 enum node_stat_item idx)
673 struct mem_cgroup_per_node *pn;
677 if (mem_cgroup_disabled())
678 return node_page_state(lruvec_pgdat(lruvec), idx);
680 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
681 for_each_possible_cpu(cpu)
682 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
690 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
692 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
694 static inline void mod_lruvec_state(struct lruvec *lruvec,
695 enum node_stat_item idx, int val)
699 local_irq_save(flags);
700 __mod_lruvec_state(lruvec, idx, val);
701 local_irq_restore(flags);
704 static inline void __mod_lruvec_page_state(struct page *page,
705 enum node_stat_item idx, int val)
707 pg_data_t *pgdat = page_pgdat(page);
708 struct lruvec *lruvec;
710 /* Untracked pages have no memcg, no lruvec. Update only the node */
711 if (!page->mem_cgroup) {
712 __mod_node_page_state(pgdat, idx, val);
716 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
717 __mod_lruvec_state(lruvec, idx, val);
720 static inline void mod_lruvec_page_state(struct page *page,
721 enum node_stat_item idx, int val)
725 local_irq_save(flags);
726 __mod_lruvec_page_state(page, idx, val);
727 local_irq_restore(flags);
730 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
732 unsigned long *total_scanned);
734 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
735 unsigned long count);
737 static inline void count_memcg_events(struct mem_cgroup *memcg,
738 enum vm_event_item idx,
743 local_irq_save(flags);
744 __count_memcg_events(memcg, idx, count);
745 local_irq_restore(flags);
748 static inline void count_memcg_page_event(struct page *page,
749 enum vm_event_item idx)
751 if (page->mem_cgroup)
752 count_memcg_events(page->mem_cgroup, idx, 1);
755 static inline void count_memcg_event_mm(struct mm_struct *mm,
756 enum vm_event_item idx)
758 struct mem_cgroup *memcg;
760 if (mem_cgroup_disabled())
764 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
766 count_memcg_events(memcg, idx, 1);
770 static inline void memcg_memory_event(struct mem_cgroup *memcg,
771 enum memcg_memory_event event)
773 atomic_long_inc(&memcg->memory_events_local[event]);
774 cgroup_file_notify(&memcg->events_local_file);
777 atomic_long_inc(&memcg->memory_events[event]);
778 cgroup_file_notify(&memcg->events_file);
780 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
782 } while ((memcg = parent_mem_cgroup(memcg)) &&
783 !mem_cgroup_is_root(memcg));
786 static inline void memcg_memory_event_mm(struct mm_struct *mm,
787 enum memcg_memory_event event)
789 struct mem_cgroup *memcg;
791 if (mem_cgroup_disabled())
795 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
797 memcg_memory_event(memcg, event);
801 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
802 void mem_cgroup_split_huge_fixup(struct page *head);
805 #else /* CONFIG_MEMCG */
807 #define MEM_CGROUP_ID_SHIFT 0
808 #define MEM_CGROUP_ID_MAX 0
812 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
817 static inline bool mem_cgroup_disabled(void)
822 static inline void memcg_memory_event(struct mem_cgroup *memcg,
823 enum memcg_memory_event event)
827 static inline void memcg_memory_event_mm(struct mm_struct *mm,
828 enum memcg_memory_event event)
832 static inline enum mem_cgroup_protection mem_cgroup_protected(
833 struct mem_cgroup *root, struct mem_cgroup *memcg)
835 return MEMCG_PROT_NONE;
838 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
840 struct mem_cgroup **memcgp,
847 static inline int mem_cgroup_try_charge_delay(struct page *page,
848 struct mm_struct *mm,
850 struct mem_cgroup **memcgp,
857 static inline void mem_cgroup_commit_charge(struct page *page,
858 struct mem_cgroup *memcg,
859 bool lrucare, bool compound)
863 static inline void mem_cgroup_cancel_charge(struct page *page,
864 struct mem_cgroup *memcg,
869 static inline void mem_cgroup_uncharge(struct page *page)
873 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
877 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
881 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
882 struct mem_cgroup *memcg)
884 return node_lruvec(pgdat);
887 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
888 struct pglist_data *pgdat)
890 return &pgdat->lruvec;
893 static inline bool mm_match_cgroup(struct mm_struct *mm,
894 struct mem_cgroup *memcg)
899 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
904 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
909 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
913 static inline struct mem_cgroup *
914 mem_cgroup_iter(struct mem_cgroup *root,
915 struct mem_cgroup *prev,
916 struct mem_cgroup_reclaim_cookie *reclaim)
921 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
922 struct mem_cgroup *prev)
926 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
927 int (*fn)(struct task_struct *, void *), void *arg)
932 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
937 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
940 /* XXX: This should always return root_mem_cgroup */
944 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
949 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
954 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
960 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
961 enum lru_list lru, int zone_idx)
966 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
972 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
977 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
981 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
986 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
990 static inline void unlock_page_memcg(struct page *page)
994 static inline void mem_cgroup_handle_over_high(void)
998 static inline void mem_cgroup_enter_user_fault(void)
1002 static inline void mem_cgroup_exit_user_fault(void)
1006 static inline bool task_in_memcg_oom(struct task_struct *p)
1011 static inline bool mem_cgroup_oom_synchronize(bool wait)
1016 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1017 struct task_struct *victim, struct mem_cgroup *oom_domain)
1022 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1026 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1031 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1037 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1043 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1049 static inline void __mod_memcg_page_state(struct page *page,
1055 static inline void mod_memcg_page_state(struct page *page,
1061 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1062 enum node_stat_item idx)
1064 return node_page_state(lruvec_pgdat(lruvec), idx);
1067 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1068 enum node_stat_item idx)
1070 return node_page_state(lruvec_pgdat(lruvec), idx);
1073 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1074 enum node_stat_item idx, int val)
1076 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1079 static inline void mod_lruvec_state(struct lruvec *lruvec,
1080 enum node_stat_item idx, int val)
1082 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1085 static inline void __mod_lruvec_page_state(struct page *page,
1086 enum node_stat_item idx, int val)
1088 __mod_node_page_state(page_pgdat(page), idx, val);
1091 static inline void mod_lruvec_page_state(struct page *page,
1092 enum node_stat_item idx, int val)
1094 mod_node_page_state(page_pgdat(page), idx, val);
1097 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1100 struct page *page = virt_to_head_page(p);
1102 __mod_node_page_state(page_pgdat(page), idx, val);
1106 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1108 unsigned long *total_scanned)
1113 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1117 static inline void count_memcg_events(struct mem_cgroup *memcg,
1118 enum vm_event_item idx,
1119 unsigned long count)
1123 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1124 enum vm_event_item idx,
1125 unsigned long count)
1129 static inline void count_memcg_page_event(struct page *page,
1135 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1138 #endif /* CONFIG_MEMCG */
1140 /* idx can be of type enum memcg_stat_item or node_stat_item */
1141 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1144 __mod_memcg_state(memcg, idx, 1);
1147 /* idx can be of type enum memcg_stat_item or node_stat_item */
1148 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1151 __mod_memcg_state(memcg, idx, -1);
1154 /* idx can be of type enum memcg_stat_item or node_stat_item */
1155 static inline void __inc_memcg_page_state(struct page *page,
1158 __mod_memcg_page_state(page, idx, 1);
1161 /* idx can be of type enum memcg_stat_item or node_stat_item */
1162 static inline void __dec_memcg_page_state(struct page *page,
1165 __mod_memcg_page_state(page, idx, -1);
1168 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1169 enum node_stat_item idx)
1171 __mod_lruvec_state(lruvec, idx, 1);
1174 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1175 enum node_stat_item idx)
1177 __mod_lruvec_state(lruvec, idx, -1);
1180 static inline void __inc_lruvec_page_state(struct page *page,
1181 enum node_stat_item idx)
1183 __mod_lruvec_page_state(page, idx, 1);
1186 static inline void __dec_lruvec_page_state(struct page *page,
1187 enum node_stat_item idx)
1189 __mod_lruvec_page_state(page, idx, -1);
1192 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1194 __mod_lruvec_slab_state(p, idx, 1);
1197 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1199 __mod_lruvec_slab_state(p, idx, -1);
1202 /* idx can be of type enum memcg_stat_item or node_stat_item */
1203 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1206 mod_memcg_state(memcg, idx, 1);
1209 /* idx can be of type enum memcg_stat_item or node_stat_item */
1210 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1213 mod_memcg_state(memcg, idx, -1);
1216 /* idx can be of type enum memcg_stat_item or node_stat_item */
1217 static inline void inc_memcg_page_state(struct page *page,
1220 mod_memcg_page_state(page, idx, 1);
1223 /* idx can be of type enum memcg_stat_item or node_stat_item */
1224 static inline void dec_memcg_page_state(struct page *page,
1227 mod_memcg_page_state(page, idx, -1);
1230 static inline void inc_lruvec_state(struct lruvec *lruvec,
1231 enum node_stat_item idx)
1233 mod_lruvec_state(lruvec, idx, 1);
1236 static inline void dec_lruvec_state(struct lruvec *lruvec,
1237 enum node_stat_item idx)
1239 mod_lruvec_state(lruvec, idx, -1);
1242 static inline void inc_lruvec_page_state(struct page *page,
1243 enum node_stat_item idx)
1245 mod_lruvec_page_state(page, idx, 1);
1248 static inline void dec_lruvec_page_state(struct page *page,
1249 enum node_stat_item idx)
1251 mod_lruvec_page_state(page, idx, -1);
1254 #ifdef CONFIG_CGROUP_WRITEBACK
1256 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1257 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1258 unsigned long *pheadroom, unsigned long *pdirty,
1259 unsigned long *pwriteback);
1261 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1262 struct bdi_writeback *wb);
1264 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1265 struct bdi_writeback *wb)
1267 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1268 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1271 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1273 #else /* CONFIG_CGROUP_WRITEBACK */
1275 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1280 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1281 unsigned long *pfilepages,
1282 unsigned long *pheadroom,
1283 unsigned long *pdirty,
1284 unsigned long *pwriteback)
1288 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1289 struct bdi_writeback *wb)
1293 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1297 #endif /* CONFIG_CGROUP_WRITEBACK */
1300 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1301 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1303 extern struct static_key_false memcg_sockets_enabled_key;
1304 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1305 void mem_cgroup_sk_alloc(struct sock *sk);
1306 void mem_cgroup_sk_free(struct sock *sk);
1307 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1309 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1312 if (time_before(jiffies, memcg->socket_pressure))
1314 } while ((memcg = parent_mem_cgroup(memcg)));
1318 extern int memcg_expand_shrinker_maps(int new_id);
1320 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1321 int nid, int shrinker_id);
1323 #define mem_cgroup_sockets_enabled 0
1324 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1325 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1326 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1331 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1332 int nid, int shrinker_id)
1337 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1338 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1340 #ifdef CONFIG_MEMCG_KMEM
1341 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1342 void __memcg_kmem_uncharge(struct page *page, int order);
1343 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1344 struct mem_cgroup *memcg);
1345 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1346 unsigned int nr_pages);
1348 extern struct static_key_false memcg_kmem_enabled_key;
1349 extern struct workqueue_struct *memcg_kmem_cache_wq;
1351 extern int memcg_nr_cache_ids;
1352 void memcg_get_cache_ids(void);
1353 void memcg_put_cache_ids(void);
1356 * Helper macro to loop through all memcg-specific caches. Callers must still
1357 * check if the cache is valid (it is either valid or NULL).
1358 * the slab_mutex must be held when looping through those caches
1360 #define for_each_memcg_cache_index(_idx) \
1361 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1363 static inline bool memcg_kmem_enabled(void)
1365 return static_branch_unlikely(&memcg_kmem_enabled_key);
1368 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1370 if (memcg_kmem_enabled())
1371 return __memcg_kmem_charge(page, gfp, order);
1375 static inline void memcg_kmem_uncharge(struct page *page, int order)
1377 if (memcg_kmem_enabled())
1378 __memcg_kmem_uncharge(page, order);
1381 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1382 int order, struct mem_cgroup *memcg)
1384 if (memcg_kmem_enabled())
1385 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1389 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1390 struct mem_cgroup *memcg)
1392 if (memcg_kmem_enabled())
1393 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1397 * helper for accessing a memcg's index. It will be used as an index in the
1398 * child cache array in kmem_cache, and also to derive its name. This function
1399 * will return -1 when this is not a kmem-limited memcg.
1401 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1403 return memcg ? memcg->kmemcg_id : -1;
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 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1422 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1426 #define for_each_memcg_cache_index(_idx) \
1429 static inline bool memcg_kmem_enabled(void)
1434 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1439 static inline void memcg_get_cache_ids(void)
1443 static inline void memcg_put_cache_ids(void)
1447 #endif /* CONFIG_MEMCG_KMEM */
1449 #endif /* _LINUX_MEMCONTROL_H */