1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* memcontrol.h - Memory Controller
4 * Copyright IBM Corporation, 2007
7 * Copyright 2007 OpenVZ SWsoft Inc
11 #ifndef _LINUX_MEMCONTROL_H
12 #define _LINUX_MEMCONTROL_H
13 #include <linux/cgroup.h>
14 #include <linux/vm_event_item.h>
15 #include <linux/hardirq.h>
16 #include <linux/jump_label.h>
17 #include <linux/page_counter.h>
18 #include <linux/vmpressure.h>
19 #include <linux/eventfd.h>
21 #include <linux/vmstat.h>
22 #include <linux/writeback.h>
23 #include <linux/page-flags.h>
30 /* Cgroup-specific page state, on top of universal node page state */
31 enum memcg_stat_item {
32 MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
37 /* XXX: why are these zone and not node counters? */
38 MEMCG_KERNEL_STACK_KB,
42 enum memcg_memory_event {
51 MEMCG_NR_MEMORY_EVENTS,
54 enum mem_cgroup_protection {
60 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,
87 struct memcg_vmstats_percpu {
88 long stat[MEMCG_NR_STAT];
89 unsigned long events[NR_VM_EVENT_ITEMS];
90 unsigned long nr_page_events;
91 unsigned long targets[MEM_CGROUP_NTARGETS];
94 struct mem_cgroup_reclaim_iter {
95 struct mem_cgroup *position;
96 /* scan generation, increased every round-trip */
97 unsigned int generation;
101 long count[NR_VM_NODE_STAT_ITEMS];
105 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
106 * which have elements charged to this memcg.
108 struct memcg_shrinker_map {
114 * per-node information in memory controller.
116 struct mem_cgroup_per_node {
117 struct lruvec lruvec;
119 /* Legacy local VM stats */
120 struct lruvec_stat __percpu *lruvec_stat_local;
122 /* Subtree VM stats (batched updates) */
123 struct lruvec_stat __percpu *lruvec_stat_cpu;
124 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
126 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
128 struct mem_cgroup_reclaim_iter iter;
130 struct memcg_shrinker_map __rcu *shrinker_map;
132 struct rb_node tree_node; /* RB tree node */
133 unsigned long usage_in_excess;/* Set to the value by which */
134 /* the soft limit is exceeded*/
136 struct mem_cgroup *memcg; /* Back pointer, we cannot */
137 /* use container_of */
140 struct mem_cgroup_threshold {
141 struct eventfd_ctx *eventfd;
142 unsigned long threshold;
146 struct mem_cgroup_threshold_ary {
147 /* An array index points to threshold just below or equal to usage. */
148 int current_threshold;
149 /* Size of entries[] */
151 /* Array of thresholds */
152 struct mem_cgroup_threshold entries[];
155 struct mem_cgroup_thresholds {
156 /* Primary thresholds array */
157 struct mem_cgroup_threshold_ary *primary;
159 * Spare threshold array.
160 * This is needed to make mem_cgroup_unregister_event() "never fail".
161 * It must be able to store at least primary->size - 1 entries.
163 struct mem_cgroup_threshold_ary *spare;
166 enum memcg_kmem_state {
172 #if defined(CONFIG_SMP)
173 struct memcg_padding {
175 } ____cacheline_internodealigned_in_smp;
176 #define MEMCG_PADDING(name) struct memcg_padding name;
178 #define MEMCG_PADDING(name)
182 * Remember four most recent foreign writebacks with dirty pages in this
183 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
184 * one in a given round, we're likely to catch it later if it keeps
185 * foreign-dirtying, so a fairly low count should be enough.
187 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
189 #define MEMCG_CGWB_FRN_CNT 4
191 struct memcg_cgwb_frn {
192 u64 bdi_id; /* bdi->id of the foreign inode */
193 int memcg_id; /* memcg->css.id of foreign inode */
194 u64 at; /* jiffies_64 at the time of dirtying */
195 struct wb_completion done; /* tracks in-flight foreign writebacks */
199 * The memory controller data structure. The memory controller controls both
200 * page cache and RSS per cgroup. We would eventually like to provide
201 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
202 * to help the administrator determine what knobs to tune.
205 struct cgroup_subsys_state css;
207 /* Private memcg ID. Used to ID objects that outlive the cgroup */
208 struct mem_cgroup_id id;
210 /* Accounted resources */
211 struct page_counter memory;
212 struct page_counter swap;
214 /* Legacy consumer-oriented counters */
215 struct page_counter memsw;
216 struct page_counter kmem;
217 struct page_counter tcpmem;
219 /* Range enforcement for interrupt charges */
220 struct work_struct high_work;
222 unsigned long soft_limit;
224 /* vmpressure notifications */
225 struct vmpressure vmpressure;
228 * Should the accounting and control be hierarchical, per subtree?
233 * Should the OOM killer kill all belonging tasks, had it kill one?
237 /* protected by memcg_oom_lock */
242 /* OOM-Killer disable */
243 int oom_kill_disable;
245 /* memory.events and memory.events.local */
246 struct cgroup_file events_file;
247 struct cgroup_file events_local_file;
249 /* handle for "memory.swap.events" */
250 struct cgroup_file swap_events_file;
252 /* protect arrays of thresholds */
253 struct mutex thresholds_lock;
255 /* thresholds for memory usage. RCU-protected */
256 struct mem_cgroup_thresholds thresholds;
258 /* thresholds for mem+swap usage. RCU-protected */
259 struct mem_cgroup_thresholds memsw_thresholds;
261 /* For oom notifier event fd */
262 struct list_head oom_notify;
265 * Should we move charges of a task when a task is moved into this
266 * mem_cgroup ? And what type of charges should we move ?
268 unsigned long move_charge_at_immigrate;
269 /* taken only while moving_account > 0 */
270 spinlock_t move_lock;
271 unsigned long move_lock_flags;
273 MEMCG_PADDING(_pad1_);
276 * set > 0 if pages under this cgroup are moving to other cgroup.
278 atomic_t moving_account;
279 struct task_struct *move_lock_task;
281 /* Legacy local VM stats and events */
282 struct memcg_vmstats_percpu __percpu *vmstats_local;
284 /* Subtree VM stats and events (batched updates) */
285 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
287 MEMCG_PADDING(_pad2_);
289 atomic_long_t vmstats[MEMCG_NR_STAT];
290 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
293 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
294 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
296 unsigned long socket_pressure;
298 /* Legacy tcp memory accounting */
302 #ifdef CONFIG_MEMCG_KMEM
303 /* Index in the kmem_cache->memcg_params.memcg_caches array */
305 enum memcg_kmem_state kmem_state;
306 struct list_head kmem_caches;
309 #ifdef CONFIG_CGROUP_WRITEBACK
310 struct list_head cgwb_list;
311 struct wb_domain cgwb_domain;
312 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
315 /* List of events which userspace want to receive */
316 struct list_head event_list;
317 spinlock_t event_list_lock;
319 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
320 struct deferred_split deferred_split_queue;
323 struct mem_cgroup_per_node *nodeinfo[0];
324 /* WARNING: nodeinfo must be the last member here */
328 * size of first charge trial. "32" comes from vmscan.c's magic value.
329 * TODO: maybe necessary to use big numbers in big irons.
331 #define MEMCG_CHARGE_BATCH 32U
333 extern struct mem_cgroup *root_mem_cgroup;
335 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
337 return (memcg == root_mem_cgroup);
340 static inline bool mem_cgroup_disabled(void)
342 return !cgroup_subsys_enabled(memory_cgrp_subsys);
345 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
348 if (mem_cgroup_disabled())
352 return READ_ONCE(memcg->memory.emin);
354 return max(READ_ONCE(memcg->memory.emin),
355 READ_ONCE(memcg->memory.elow));
358 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
359 struct mem_cgroup *memcg);
361 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
362 gfp_t gfp_mask, struct mem_cgroup **memcgp,
364 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
365 gfp_t gfp_mask, struct mem_cgroup **memcgp,
367 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
368 bool lrucare, bool compound);
369 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
371 void mem_cgroup_uncharge(struct page *page);
372 void mem_cgroup_uncharge_list(struct list_head *page_list);
374 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
376 static struct mem_cgroup_per_node *
377 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
379 return memcg->nodeinfo[nid];
383 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
384 * @memcg: memcg of the wanted lruvec
386 * Returns the lru list vector holding pages for a given @memcg &
387 * @node combination. This can be the node lruvec, if the memory
388 * controller is disabled.
390 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
391 struct pglist_data *pgdat)
393 struct mem_cgroup_per_node *mz;
394 struct lruvec *lruvec;
396 if (mem_cgroup_disabled()) {
397 lruvec = &pgdat->__lruvec;
402 memcg = root_mem_cgroup;
404 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
405 lruvec = &mz->lruvec;
408 * Since a node can be onlined after the mem_cgroup was created,
409 * we have to be prepared to initialize lruvec->pgdat here;
410 * and if offlined then reonlined, we need to reinitialize it.
412 if (unlikely(lruvec->pgdat != pgdat))
413 lruvec->pgdat = pgdat;
417 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
419 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
421 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
423 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
426 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
427 return css ? container_of(css, struct mem_cgroup, css) : NULL;
430 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
433 css_put(&memcg->css);
436 #define mem_cgroup_from_counter(counter, member) \
437 container_of(counter, struct mem_cgroup, member)
439 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
441 struct mem_cgroup_reclaim_cookie *);
442 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
443 int mem_cgroup_scan_tasks(struct mem_cgroup *,
444 int (*)(struct task_struct *, void *), void *);
446 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
448 if (mem_cgroup_disabled())
453 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
455 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
457 return mem_cgroup_from_css(seq_css(m));
460 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
462 struct mem_cgroup_per_node *mz;
464 if (mem_cgroup_disabled())
467 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
472 * parent_mem_cgroup - find the accounting parent of a memcg
473 * @memcg: memcg whose parent to find
475 * Returns the parent memcg, or NULL if this is the root or the memory
476 * controller is in legacy no-hierarchy mode.
478 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
480 if (!memcg->memory.parent)
482 return mem_cgroup_from_counter(memcg->memory.parent, memory);
485 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
486 struct mem_cgroup *root)
490 if (!root->use_hierarchy)
492 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
495 static inline bool mm_match_cgroup(struct mm_struct *mm,
496 struct mem_cgroup *memcg)
498 struct mem_cgroup *task_memcg;
502 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
504 match = mem_cgroup_is_descendant(task_memcg, memcg);
509 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
510 ino_t page_cgroup_ino(struct page *page);
512 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
514 if (mem_cgroup_disabled())
516 return !!(memcg->css.flags & CSS_ONLINE);
520 * For memory reclaim.
522 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
524 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
525 int zid, int nr_pages);
528 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
529 enum lru_list lru, int zone_idx)
531 struct mem_cgroup_per_node *mz;
533 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
534 return mz->lru_zone_size[zone_idx][lru];
537 void mem_cgroup_handle_over_high(void);
539 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
541 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
543 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
544 struct task_struct *p);
546 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
548 static inline void mem_cgroup_enter_user_fault(void)
550 WARN_ON(current->in_user_fault);
551 current->in_user_fault = 1;
554 static inline void mem_cgroup_exit_user_fault(void)
556 WARN_ON(!current->in_user_fault);
557 current->in_user_fault = 0;
560 static inline bool task_in_memcg_oom(struct task_struct *p)
562 return p->memcg_in_oom;
565 bool mem_cgroup_oom_synchronize(bool wait);
566 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
567 struct mem_cgroup *oom_domain);
568 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
570 #ifdef CONFIG_MEMCG_SWAP
571 extern int do_swap_account;
574 struct mem_cgroup *lock_page_memcg(struct page *page);
575 void __unlock_page_memcg(struct mem_cgroup *memcg);
576 void unlock_page_memcg(struct page *page);
579 * idx can be of type enum memcg_stat_item or node_stat_item.
580 * Keep in sync with memcg_exact_page_state().
582 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
584 long x = atomic_long_read(&memcg->vmstats[idx]);
593 * idx can be of type enum memcg_stat_item or node_stat_item.
594 * Keep in sync with memcg_exact_page_state().
596 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
602 for_each_possible_cpu(cpu)
603 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
611 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
613 /* idx can be of type enum memcg_stat_item or node_stat_item */
614 static inline void mod_memcg_state(struct mem_cgroup *memcg,
619 local_irq_save(flags);
620 __mod_memcg_state(memcg, idx, val);
621 local_irq_restore(flags);
625 * mod_memcg_page_state - update page state statistics
627 * @idx: page state item to account
628 * @val: number of pages (positive or negative)
630 * The @page must be locked or the caller must use lock_page_memcg()
631 * to prevent double accounting when the page is concurrently being
632 * moved to another memcg:
634 * lock_page(page) or lock_page_memcg(page)
635 * if (TestClearPageState(page))
636 * mod_memcg_page_state(page, state, -1);
637 * unlock_page(page) or unlock_page_memcg(page)
639 * Kernel pages are an exception to this, since they'll never move.
641 static inline void __mod_memcg_page_state(struct page *page,
644 if (page->mem_cgroup)
645 __mod_memcg_state(page->mem_cgroup, idx, val);
648 static inline void mod_memcg_page_state(struct page *page,
651 if (page->mem_cgroup)
652 mod_memcg_state(page->mem_cgroup, idx, val);
655 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
656 enum node_stat_item idx)
658 struct mem_cgroup_per_node *pn;
661 if (mem_cgroup_disabled())
662 return node_page_state(lruvec_pgdat(lruvec), idx);
664 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
665 x = atomic_long_read(&pn->lruvec_stat[idx]);
673 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
674 enum node_stat_item idx)
676 struct mem_cgroup_per_node *pn;
680 if (mem_cgroup_disabled())
681 return node_page_state(lruvec_pgdat(lruvec), idx);
683 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
684 for_each_possible_cpu(cpu)
685 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
693 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
695 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
696 void mod_memcg_obj_state(void *p, int idx, int val);
698 static inline void mod_lruvec_state(struct lruvec *lruvec,
699 enum node_stat_item idx, int val)
703 local_irq_save(flags);
704 __mod_lruvec_state(lruvec, idx, val);
705 local_irq_restore(flags);
708 static inline void __mod_lruvec_page_state(struct page *page,
709 enum node_stat_item idx, int val)
711 pg_data_t *pgdat = page_pgdat(page);
712 struct lruvec *lruvec;
714 /* Untracked pages have no memcg, no lruvec. Update only the node */
715 if (!page->mem_cgroup) {
716 __mod_node_page_state(pgdat, idx, val);
720 lruvec = mem_cgroup_lruvec(page->mem_cgroup, pgdat);
721 __mod_lruvec_state(lruvec, idx, val);
724 static inline void mod_lruvec_page_state(struct page *page,
725 enum node_stat_item idx, int val)
729 local_irq_save(flags);
730 __mod_lruvec_page_state(page, idx, val);
731 local_irq_restore(flags);
734 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
736 unsigned long *total_scanned);
738 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
739 unsigned long count);
741 static inline void count_memcg_events(struct mem_cgroup *memcg,
742 enum vm_event_item idx,
747 local_irq_save(flags);
748 __count_memcg_events(memcg, idx, count);
749 local_irq_restore(flags);
752 static inline void count_memcg_page_event(struct page *page,
753 enum vm_event_item idx)
755 if (page->mem_cgroup)
756 count_memcg_events(page->mem_cgroup, idx, 1);
759 static inline void count_memcg_event_mm(struct mm_struct *mm,
760 enum vm_event_item idx)
762 struct mem_cgroup *memcg;
764 if (mem_cgroup_disabled())
768 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
770 count_memcg_events(memcg, idx, 1);
774 static inline void memcg_memory_event(struct mem_cgroup *memcg,
775 enum memcg_memory_event event)
777 atomic_long_inc(&memcg->memory_events_local[event]);
778 cgroup_file_notify(&memcg->events_local_file);
781 atomic_long_inc(&memcg->memory_events[event]);
782 cgroup_file_notify(&memcg->events_file);
784 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
786 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
788 } while ((memcg = parent_mem_cgroup(memcg)) &&
789 !mem_cgroup_is_root(memcg));
792 static inline void memcg_memory_event_mm(struct mm_struct *mm,
793 enum memcg_memory_event event)
795 struct mem_cgroup *memcg;
797 if (mem_cgroup_disabled())
801 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
803 memcg_memory_event(memcg, event);
807 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
808 void mem_cgroup_split_huge_fixup(struct page *head);
811 #else /* CONFIG_MEMCG */
813 #define MEM_CGROUP_ID_SHIFT 0
814 #define MEM_CGROUP_ID_MAX 0
818 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
823 static inline bool mem_cgroup_disabled(void)
828 static inline void memcg_memory_event(struct mem_cgroup *memcg,
829 enum memcg_memory_event event)
833 static inline void memcg_memory_event_mm(struct mm_struct *mm,
834 enum memcg_memory_event event)
838 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
844 static inline enum mem_cgroup_protection mem_cgroup_protected(
845 struct mem_cgroup *root, struct mem_cgroup *memcg)
847 return MEMCG_PROT_NONE;
850 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
852 struct mem_cgroup **memcgp,
859 static inline int mem_cgroup_try_charge_delay(struct page *page,
860 struct mm_struct *mm,
862 struct mem_cgroup **memcgp,
869 static inline void mem_cgroup_commit_charge(struct page *page,
870 struct mem_cgroup *memcg,
871 bool lrucare, bool compound)
875 static inline void mem_cgroup_cancel_charge(struct page *page,
876 struct mem_cgroup *memcg,
881 static inline void mem_cgroup_uncharge(struct page *page)
885 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
889 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
893 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
894 struct pglist_data *pgdat)
896 return &pgdat->__lruvec;
899 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
900 struct pglist_data *pgdat)
902 return &pgdat->__lruvec;
905 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
910 static inline bool mm_match_cgroup(struct mm_struct *mm,
911 struct mem_cgroup *memcg)
916 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
921 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
926 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
930 static inline struct mem_cgroup *
931 mem_cgroup_iter(struct mem_cgroup *root,
932 struct mem_cgroup *prev,
933 struct mem_cgroup_reclaim_cookie *reclaim)
938 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
939 struct mem_cgroup *prev)
943 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
944 int (*fn)(struct task_struct *, void *), void *arg)
949 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
954 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
957 /* XXX: This should always return root_mem_cgroup */
961 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
966 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
971 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
977 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
978 enum lru_list lru, int zone_idx)
983 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
988 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
994 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
999 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1003 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1008 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1012 static inline void unlock_page_memcg(struct page *page)
1016 static inline void mem_cgroup_handle_over_high(void)
1020 static inline void mem_cgroup_enter_user_fault(void)
1024 static inline void mem_cgroup_exit_user_fault(void)
1028 static inline bool task_in_memcg_oom(struct task_struct *p)
1033 static inline bool mem_cgroup_oom_synchronize(bool wait)
1038 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1039 struct task_struct *victim, struct mem_cgroup *oom_domain)
1044 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1048 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1053 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1059 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1065 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1071 static inline void __mod_memcg_page_state(struct page *page,
1077 static inline void mod_memcg_page_state(struct page *page,
1083 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1084 enum node_stat_item idx)
1086 return node_page_state(lruvec_pgdat(lruvec), idx);
1089 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1090 enum node_stat_item idx)
1092 return node_page_state(lruvec_pgdat(lruvec), idx);
1095 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1096 enum node_stat_item idx, int val)
1098 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1101 static inline void mod_lruvec_state(struct lruvec *lruvec,
1102 enum node_stat_item idx, int val)
1104 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1107 static inline void __mod_lruvec_page_state(struct page *page,
1108 enum node_stat_item idx, int val)
1110 __mod_node_page_state(page_pgdat(page), idx, val);
1113 static inline void mod_lruvec_page_state(struct page *page,
1114 enum node_stat_item idx, int val)
1116 mod_node_page_state(page_pgdat(page), idx, val);
1119 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1122 struct page *page = virt_to_head_page(p);
1124 __mod_node_page_state(page_pgdat(page), idx, val);
1127 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1132 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1134 unsigned long *total_scanned)
1139 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1143 static inline void count_memcg_events(struct mem_cgroup *memcg,
1144 enum vm_event_item idx,
1145 unsigned long count)
1149 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1150 enum vm_event_item idx,
1151 unsigned long count)
1155 static inline void count_memcg_page_event(struct page *page,
1161 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1164 #endif /* CONFIG_MEMCG */
1166 /* idx can be of type enum memcg_stat_item or node_stat_item */
1167 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1170 __mod_memcg_state(memcg, idx, 1);
1173 /* idx can be of type enum memcg_stat_item or node_stat_item */
1174 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1177 __mod_memcg_state(memcg, idx, -1);
1180 /* idx can be of type enum memcg_stat_item or node_stat_item */
1181 static inline void __inc_memcg_page_state(struct page *page,
1184 __mod_memcg_page_state(page, idx, 1);
1187 /* idx can be of type enum memcg_stat_item or node_stat_item */
1188 static inline void __dec_memcg_page_state(struct page *page,
1191 __mod_memcg_page_state(page, idx, -1);
1194 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1195 enum node_stat_item idx)
1197 __mod_lruvec_state(lruvec, idx, 1);
1200 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1201 enum node_stat_item idx)
1203 __mod_lruvec_state(lruvec, idx, -1);
1206 static inline void __inc_lruvec_page_state(struct page *page,
1207 enum node_stat_item idx)
1209 __mod_lruvec_page_state(page, idx, 1);
1212 static inline void __dec_lruvec_page_state(struct page *page,
1213 enum node_stat_item idx)
1215 __mod_lruvec_page_state(page, idx, -1);
1218 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1220 __mod_lruvec_slab_state(p, idx, 1);
1223 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1225 __mod_lruvec_slab_state(p, idx, -1);
1228 /* idx can be of type enum memcg_stat_item or node_stat_item */
1229 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1232 mod_memcg_state(memcg, idx, 1);
1235 /* idx can be of type enum memcg_stat_item or node_stat_item */
1236 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1239 mod_memcg_state(memcg, idx, -1);
1242 /* idx can be of type enum memcg_stat_item or node_stat_item */
1243 static inline void inc_memcg_page_state(struct page *page,
1246 mod_memcg_page_state(page, idx, 1);
1249 /* idx can be of type enum memcg_stat_item or node_stat_item */
1250 static inline void dec_memcg_page_state(struct page *page,
1253 mod_memcg_page_state(page, idx, -1);
1256 static inline void inc_lruvec_state(struct lruvec *lruvec,
1257 enum node_stat_item idx)
1259 mod_lruvec_state(lruvec, idx, 1);
1262 static inline void dec_lruvec_state(struct lruvec *lruvec,
1263 enum node_stat_item idx)
1265 mod_lruvec_state(lruvec, idx, -1);
1268 static inline void inc_lruvec_page_state(struct page *page,
1269 enum node_stat_item idx)
1271 mod_lruvec_page_state(page, idx, 1);
1274 static inline void dec_lruvec_page_state(struct page *page,
1275 enum node_stat_item idx)
1277 mod_lruvec_page_state(page, idx, -1);
1280 #ifdef CONFIG_CGROUP_WRITEBACK
1282 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1283 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1284 unsigned long *pheadroom, unsigned long *pdirty,
1285 unsigned long *pwriteback);
1287 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1288 struct bdi_writeback *wb);
1290 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1291 struct bdi_writeback *wb)
1293 if (mem_cgroup_disabled())
1296 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1297 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1300 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1302 #else /* CONFIG_CGROUP_WRITEBACK */
1304 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1309 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1310 unsigned long *pfilepages,
1311 unsigned long *pheadroom,
1312 unsigned long *pdirty,
1313 unsigned long *pwriteback)
1317 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1318 struct bdi_writeback *wb)
1322 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1326 #endif /* CONFIG_CGROUP_WRITEBACK */
1329 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1330 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1332 extern struct static_key_false memcg_sockets_enabled_key;
1333 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1334 void mem_cgroup_sk_alloc(struct sock *sk);
1335 void mem_cgroup_sk_free(struct sock *sk);
1336 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1338 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1341 if (time_before(jiffies, memcg->socket_pressure))
1343 } while ((memcg = parent_mem_cgroup(memcg)));
1347 extern int memcg_expand_shrinker_maps(int new_id);
1349 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1350 int nid, int shrinker_id);
1352 #define mem_cgroup_sockets_enabled 0
1353 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1354 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1355 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1360 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1361 int nid, int shrinker_id)
1366 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1367 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1369 #ifdef CONFIG_MEMCG_KMEM
1370 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1371 unsigned int nr_pages);
1372 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1373 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1374 void __memcg_kmem_uncharge_page(struct page *page, int order);
1376 extern struct static_key_false memcg_kmem_enabled_key;
1377 extern struct workqueue_struct *memcg_kmem_cache_wq;
1379 extern int memcg_nr_cache_ids;
1380 void memcg_get_cache_ids(void);
1381 void memcg_put_cache_ids(void);
1384 * Helper macro to loop through all memcg-specific caches. Callers must still
1385 * check if the cache is valid (it is either valid or NULL).
1386 * the slab_mutex must be held when looping through those caches
1388 #define for_each_memcg_cache_index(_idx) \
1389 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1391 static inline bool memcg_kmem_enabled(void)
1393 return static_branch_unlikely(&memcg_kmem_enabled_key);
1396 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1399 if (memcg_kmem_enabled())
1400 return __memcg_kmem_charge_page(page, gfp, order);
1404 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1406 if (memcg_kmem_enabled())
1407 __memcg_kmem_uncharge_page(page, order);
1410 static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1411 unsigned int nr_pages)
1413 if (memcg_kmem_enabled())
1414 return __memcg_kmem_charge(memcg, gfp, nr_pages);
1418 static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg,
1419 unsigned int nr_pages)
1421 if (memcg_kmem_enabled())
1422 __memcg_kmem_uncharge(memcg, nr_pages);
1426 * helper for accessing a memcg's index. It will be used as an index in the
1427 * child cache array in kmem_cache, and also to derive its name. This function
1428 * will return -1 when this is not a kmem-limited memcg.
1430 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1432 return memcg ? memcg->kmemcg_id : -1;
1435 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1439 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1445 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1449 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1455 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1459 #define for_each_memcg_cache_index(_idx) \
1462 static inline bool memcg_kmem_enabled(void)
1467 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1472 static inline void memcg_get_cache_ids(void)
1476 static inline void memcg_put_cache_ids(void)
1480 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1485 #endif /* CONFIG_MEMCG_KMEM */
1487 #endif /* _LINUX_MEMCONTROL_H */