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 struct mem_cgroup_per_node *nodeinfo[0];
334 /* WARNING: nodeinfo must be the last member here */
338 * size of first charge trial. "32" comes from vmscan.c's magic value.
339 * TODO: maybe necessary to use big numbers in big irons.
341 #define MEMCG_CHARGE_BATCH 32U
343 extern struct mem_cgroup *root_mem_cgroup;
345 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
347 return (memcg == root_mem_cgroup);
350 static inline bool mem_cgroup_disabled(void)
352 return !cgroup_subsys_enabled(memory_cgrp_subsys);
355 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
356 struct mem_cgroup *memcg);
358 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
359 gfp_t gfp_mask, struct mem_cgroup **memcgp,
361 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
362 gfp_t gfp_mask, struct mem_cgroup **memcgp,
364 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
365 bool lrucare, bool compound);
366 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
368 void mem_cgroup_uncharge(struct page *page);
369 void mem_cgroup_uncharge_list(struct list_head *page_list);
371 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
373 static struct mem_cgroup_per_node *
374 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
376 return memcg->nodeinfo[nid];
380 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
381 * @node: node of the wanted lruvec
382 * @memcg: memcg of the wanted lruvec
384 * Returns the lru list vector holding pages for a given @node or a given
385 * @memcg and @zone. This can be the node lruvec, if the memory controller
388 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
389 struct mem_cgroup *memcg)
391 struct mem_cgroup_per_node *mz;
392 struct lruvec *lruvec;
394 if (mem_cgroup_disabled()) {
395 lruvec = node_lruvec(pgdat);
399 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
400 lruvec = &mz->lruvec;
403 * Since a node can be onlined after the mem_cgroup was created,
404 * we have to be prepared to initialize lruvec->pgdat here;
405 * and if offlined then reonlined, we need to reinitialize it.
407 if (unlikely(lruvec->pgdat != pgdat))
408 lruvec->pgdat = pgdat;
412 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
414 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
416 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
418 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
421 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
422 return css ? container_of(css, struct mem_cgroup, css) : NULL;
425 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
428 css_put(&memcg->css);
431 #define mem_cgroup_from_counter(counter, member) \
432 container_of(counter, struct mem_cgroup, member)
434 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
436 struct mem_cgroup_reclaim_cookie *);
437 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
438 int mem_cgroup_scan_tasks(struct mem_cgroup *,
439 int (*)(struct task_struct *, void *), void *);
441 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
443 if (mem_cgroup_disabled())
448 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
450 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
452 return mem_cgroup_from_css(seq_css(m));
455 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
457 struct mem_cgroup_per_node *mz;
459 if (mem_cgroup_disabled())
462 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
467 * parent_mem_cgroup - find the accounting parent of a memcg
468 * @memcg: memcg whose parent to find
470 * Returns the parent memcg, or NULL if this is the root or the memory
471 * controller is in legacy no-hierarchy mode.
473 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
475 if (!memcg->memory.parent)
477 return mem_cgroup_from_counter(memcg->memory.parent, memory);
480 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
481 struct mem_cgroup *root)
485 if (!root->use_hierarchy)
487 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
490 static inline bool mm_match_cgroup(struct mm_struct *mm,
491 struct mem_cgroup *memcg)
493 struct mem_cgroup *task_memcg;
497 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
499 match = mem_cgroup_is_descendant(task_memcg, memcg);
504 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
505 ino_t page_cgroup_ino(struct page *page);
507 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
509 if (mem_cgroup_disabled())
511 return !!(memcg->css.flags & CSS_ONLINE);
515 * For memory reclaim.
517 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
519 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
520 int zid, int nr_pages);
523 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
524 enum lru_list lru, int zone_idx)
526 struct mem_cgroup_per_node *mz;
528 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
529 return mz->lru_zone_size[zone_idx][lru];
532 void mem_cgroup_handle_over_high(void);
534 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
536 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
537 struct task_struct *p);
539 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
541 static inline void mem_cgroup_enter_user_fault(void)
543 WARN_ON(current->in_user_fault);
544 current->in_user_fault = 1;
547 static inline void mem_cgroup_exit_user_fault(void)
549 WARN_ON(!current->in_user_fault);
550 current->in_user_fault = 0;
553 static inline bool task_in_memcg_oom(struct task_struct *p)
555 return p->memcg_in_oom;
558 bool mem_cgroup_oom_synchronize(bool wait);
559 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
560 struct mem_cgroup *oom_domain);
561 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
563 #ifdef CONFIG_MEMCG_SWAP
564 extern int do_swap_account;
567 struct mem_cgroup *lock_page_memcg(struct page *page);
568 void __unlock_page_memcg(struct mem_cgroup *memcg);
569 void unlock_page_memcg(struct page *page);
572 * idx can be of type enum memcg_stat_item or node_stat_item.
573 * Keep in sync with memcg_exact_page_state().
575 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
577 long x = atomic_long_read(&memcg->vmstats[idx]);
586 * idx can be of type enum memcg_stat_item or node_stat_item.
587 * Keep in sync with memcg_exact_page_state().
589 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
595 for_each_possible_cpu(cpu)
596 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
604 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
606 /* idx can be of type enum memcg_stat_item or node_stat_item */
607 static inline void mod_memcg_state(struct mem_cgroup *memcg,
612 local_irq_save(flags);
613 __mod_memcg_state(memcg, idx, val);
614 local_irq_restore(flags);
618 * mod_memcg_page_state - update page state statistics
620 * @idx: page state item to account
621 * @val: number of pages (positive or negative)
623 * The @page must be locked or the caller must use lock_page_memcg()
624 * to prevent double accounting when the page is concurrently being
625 * moved to another memcg:
627 * lock_page(page) or lock_page_memcg(page)
628 * if (TestClearPageState(page))
629 * mod_memcg_page_state(page, state, -1);
630 * unlock_page(page) or unlock_page_memcg(page)
632 * Kernel pages are an exception to this, since they'll never move.
634 static inline void __mod_memcg_page_state(struct page *page,
637 if (page->mem_cgroup)
638 __mod_memcg_state(page->mem_cgroup, idx, val);
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 unsigned long lruvec_page_state(struct lruvec *lruvec,
649 enum node_stat_item idx)
651 struct mem_cgroup_per_node *pn;
654 if (mem_cgroup_disabled())
655 return node_page_state(lruvec_pgdat(lruvec), idx);
657 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
658 x = atomic_long_read(&pn->lruvec_stat[idx]);
666 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
667 enum node_stat_item idx)
669 struct mem_cgroup_per_node *pn;
673 if (mem_cgroup_disabled())
674 return node_page_state(lruvec_pgdat(lruvec), idx);
676 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
677 for_each_possible_cpu(cpu)
678 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
686 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
688 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
690 static inline void mod_lruvec_state(struct lruvec *lruvec,
691 enum node_stat_item idx, int val)
695 local_irq_save(flags);
696 __mod_lruvec_state(lruvec, idx, val);
697 local_irq_restore(flags);
700 static inline void __mod_lruvec_page_state(struct page *page,
701 enum node_stat_item idx, int val)
703 pg_data_t *pgdat = page_pgdat(page);
704 struct lruvec *lruvec;
706 /* Untracked pages have no memcg, no lruvec. Update only the node */
707 if (!page->mem_cgroup) {
708 __mod_node_page_state(pgdat, idx, val);
712 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
713 __mod_lruvec_state(lruvec, idx, val);
716 static inline void mod_lruvec_page_state(struct page *page,
717 enum node_stat_item idx, int val)
721 local_irq_save(flags);
722 __mod_lruvec_page_state(page, idx, val);
723 local_irq_restore(flags);
726 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
728 unsigned long *total_scanned);
730 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
731 unsigned long count);
733 static inline void count_memcg_events(struct mem_cgroup *memcg,
734 enum vm_event_item idx,
739 local_irq_save(flags);
740 __count_memcg_events(memcg, idx, count);
741 local_irq_restore(flags);
744 static inline void count_memcg_page_event(struct page *page,
745 enum vm_event_item idx)
747 if (page->mem_cgroup)
748 count_memcg_events(page->mem_cgroup, idx, 1);
751 static inline void count_memcg_event_mm(struct mm_struct *mm,
752 enum vm_event_item idx)
754 struct mem_cgroup *memcg;
756 if (mem_cgroup_disabled())
760 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
762 count_memcg_events(memcg, idx, 1);
766 static inline void memcg_memory_event(struct mem_cgroup *memcg,
767 enum memcg_memory_event event)
769 atomic_long_inc(&memcg->memory_events_local[event]);
770 cgroup_file_notify(&memcg->events_local_file);
773 atomic_long_inc(&memcg->memory_events[event]);
774 cgroup_file_notify(&memcg->events_file);
776 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
778 } while ((memcg = parent_mem_cgroup(memcg)) &&
779 !mem_cgroup_is_root(memcg));
782 static inline void memcg_memory_event_mm(struct mm_struct *mm,
783 enum memcg_memory_event event)
785 struct mem_cgroup *memcg;
787 if (mem_cgroup_disabled())
791 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
793 memcg_memory_event(memcg, event);
797 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
798 void mem_cgroup_split_huge_fixup(struct page *head);
801 #else /* CONFIG_MEMCG */
803 #define MEM_CGROUP_ID_SHIFT 0
804 #define MEM_CGROUP_ID_MAX 0
808 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
813 static inline bool mem_cgroup_disabled(void)
818 static inline void memcg_memory_event(struct mem_cgroup *memcg,
819 enum memcg_memory_event event)
823 static inline void memcg_memory_event_mm(struct mm_struct *mm,
824 enum memcg_memory_event event)
828 static inline enum mem_cgroup_protection mem_cgroup_protected(
829 struct mem_cgroup *root, struct mem_cgroup *memcg)
831 return MEMCG_PROT_NONE;
834 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
836 struct mem_cgroup **memcgp,
843 static inline int mem_cgroup_try_charge_delay(struct page *page,
844 struct mm_struct *mm,
846 struct mem_cgroup **memcgp,
853 static inline void mem_cgroup_commit_charge(struct page *page,
854 struct mem_cgroup *memcg,
855 bool lrucare, bool compound)
859 static inline void mem_cgroup_cancel_charge(struct page *page,
860 struct mem_cgroup *memcg,
865 static inline void mem_cgroup_uncharge(struct page *page)
869 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
873 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
877 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
878 struct mem_cgroup *memcg)
880 return node_lruvec(pgdat);
883 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
884 struct pglist_data *pgdat)
886 return &pgdat->lruvec;
889 static inline bool mm_match_cgroup(struct mm_struct *mm,
890 struct mem_cgroup *memcg)
895 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
900 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
905 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
909 static inline struct mem_cgroup *
910 mem_cgroup_iter(struct mem_cgroup *root,
911 struct mem_cgroup *prev,
912 struct mem_cgroup_reclaim_cookie *reclaim)
917 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
918 struct mem_cgroup *prev)
922 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
923 int (*fn)(struct task_struct *, void *), void *arg)
928 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
933 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
936 /* XXX: This should always return root_mem_cgroup */
940 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
945 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
950 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
956 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
957 enum lru_list lru, int zone_idx)
962 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
968 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
973 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
977 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
982 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
986 static inline void unlock_page_memcg(struct page *page)
990 static inline void mem_cgroup_handle_over_high(void)
994 static inline void mem_cgroup_enter_user_fault(void)
998 static inline void mem_cgroup_exit_user_fault(void)
1002 static inline bool task_in_memcg_oom(struct task_struct *p)
1007 static inline bool mem_cgroup_oom_synchronize(bool wait)
1012 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1013 struct task_struct *victim, struct mem_cgroup *oom_domain)
1018 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1022 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1027 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1033 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1039 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1045 static inline void __mod_memcg_page_state(struct page *page,
1051 static inline void mod_memcg_page_state(struct page *page,
1057 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1058 enum node_stat_item idx)
1060 return node_page_state(lruvec_pgdat(lruvec), idx);
1063 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1064 enum node_stat_item idx)
1066 return node_page_state(lruvec_pgdat(lruvec), idx);
1069 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1070 enum node_stat_item idx, int val)
1072 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1075 static inline void mod_lruvec_state(struct lruvec *lruvec,
1076 enum node_stat_item idx, int val)
1078 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1081 static inline void __mod_lruvec_page_state(struct page *page,
1082 enum node_stat_item idx, int val)
1084 __mod_node_page_state(page_pgdat(page), idx, val);
1087 static inline void mod_lruvec_page_state(struct page *page,
1088 enum node_stat_item idx, int val)
1090 mod_node_page_state(page_pgdat(page), idx, val);
1093 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1096 struct page *page = virt_to_head_page(p);
1098 __mod_node_page_state(page_pgdat(page), idx, val);
1102 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1104 unsigned long *total_scanned)
1109 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1113 static inline void count_memcg_events(struct mem_cgroup *memcg,
1114 enum vm_event_item idx,
1115 unsigned long count)
1119 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1120 enum vm_event_item idx,
1121 unsigned long count)
1125 static inline void count_memcg_page_event(struct page *page,
1131 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1134 #endif /* CONFIG_MEMCG */
1136 /* idx can be of type enum memcg_stat_item or node_stat_item */
1137 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1140 __mod_memcg_state(memcg, idx, 1);
1143 /* idx can be of type enum memcg_stat_item or node_stat_item */
1144 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1147 __mod_memcg_state(memcg, idx, -1);
1150 /* idx can be of type enum memcg_stat_item or node_stat_item */
1151 static inline void __inc_memcg_page_state(struct page *page,
1154 __mod_memcg_page_state(page, idx, 1);
1157 /* idx can be of type enum memcg_stat_item or node_stat_item */
1158 static inline void __dec_memcg_page_state(struct page *page,
1161 __mod_memcg_page_state(page, idx, -1);
1164 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1165 enum node_stat_item idx)
1167 __mod_lruvec_state(lruvec, idx, 1);
1170 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1171 enum node_stat_item idx)
1173 __mod_lruvec_state(lruvec, idx, -1);
1176 static inline void __inc_lruvec_page_state(struct page *page,
1177 enum node_stat_item idx)
1179 __mod_lruvec_page_state(page, idx, 1);
1182 static inline void __dec_lruvec_page_state(struct page *page,
1183 enum node_stat_item idx)
1185 __mod_lruvec_page_state(page, idx, -1);
1188 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1190 __mod_lruvec_slab_state(p, idx, 1);
1193 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1195 __mod_lruvec_slab_state(p, idx, -1);
1198 /* idx can be of type enum memcg_stat_item or node_stat_item */
1199 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1202 mod_memcg_state(memcg, idx, 1);
1205 /* idx can be of type enum memcg_stat_item or node_stat_item */
1206 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1209 mod_memcg_state(memcg, idx, -1);
1212 /* idx can be of type enum memcg_stat_item or node_stat_item */
1213 static inline void inc_memcg_page_state(struct page *page,
1216 mod_memcg_page_state(page, idx, 1);
1219 /* idx can be of type enum memcg_stat_item or node_stat_item */
1220 static inline void dec_memcg_page_state(struct page *page,
1223 mod_memcg_page_state(page, idx, -1);
1226 static inline void inc_lruvec_state(struct lruvec *lruvec,
1227 enum node_stat_item idx)
1229 mod_lruvec_state(lruvec, idx, 1);
1232 static inline void dec_lruvec_state(struct lruvec *lruvec,
1233 enum node_stat_item idx)
1235 mod_lruvec_state(lruvec, idx, -1);
1238 static inline void inc_lruvec_page_state(struct page *page,
1239 enum node_stat_item idx)
1241 mod_lruvec_page_state(page, idx, 1);
1244 static inline void dec_lruvec_page_state(struct page *page,
1245 enum node_stat_item idx)
1247 mod_lruvec_page_state(page, idx, -1);
1250 #ifdef CONFIG_CGROUP_WRITEBACK
1252 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1253 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1254 unsigned long *pheadroom, unsigned long *pdirty,
1255 unsigned long *pwriteback);
1257 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1258 struct bdi_writeback *wb);
1260 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1261 struct bdi_writeback *wb)
1263 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1264 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1267 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1269 #else /* CONFIG_CGROUP_WRITEBACK */
1271 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1276 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1277 unsigned long *pfilepages,
1278 unsigned long *pheadroom,
1279 unsigned long *pdirty,
1280 unsigned long *pwriteback)
1284 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1285 struct bdi_writeback *wb)
1289 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1293 #endif /* CONFIG_CGROUP_WRITEBACK */
1296 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1297 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1299 extern struct static_key_false memcg_sockets_enabled_key;
1300 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1301 void mem_cgroup_sk_alloc(struct sock *sk);
1302 void mem_cgroup_sk_free(struct sock *sk);
1303 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1305 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1308 if (time_before(jiffies, memcg->socket_pressure))
1310 } while ((memcg = parent_mem_cgroup(memcg)));
1314 extern int memcg_expand_shrinker_maps(int new_id);
1316 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1317 int nid, int shrinker_id);
1319 #define mem_cgroup_sockets_enabled 0
1320 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1321 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1322 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1327 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1328 int nid, int shrinker_id)
1333 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1334 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1336 #ifdef CONFIG_MEMCG_KMEM
1337 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1338 void __memcg_kmem_uncharge(struct page *page, int order);
1339 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1340 struct mem_cgroup *memcg);
1341 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1342 unsigned int nr_pages);
1344 extern struct static_key_false memcg_kmem_enabled_key;
1345 extern struct workqueue_struct *memcg_kmem_cache_wq;
1347 extern int memcg_nr_cache_ids;
1348 void memcg_get_cache_ids(void);
1349 void memcg_put_cache_ids(void);
1352 * Helper macro to loop through all memcg-specific caches. Callers must still
1353 * check if the cache is valid (it is either valid or NULL).
1354 * the slab_mutex must be held when looping through those caches
1356 #define for_each_memcg_cache_index(_idx) \
1357 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1359 static inline bool memcg_kmem_enabled(void)
1361 return static_branch_unlikely(&memcg_kmem_enabled_key);
1364 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1366 if (memcg_kmem_enabled())
1367 return __memcg_kmem_charge(page, gfp, order);
1371 static inline void memcg_kmem_uncharge(struct page *page, int order)
1373 if (memcg_kmem_enabled())
1374 __memcg_kmem_uncharge(page, order);
1377 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1378 int order, struct mem_cgroup *memcg)
1380 if (memcg_kmem_enabled())
1381 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1385 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1386 struct mem_cgroup *memcg)
1388 if (memcg_kmem_enabled())
1389 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1393 * helper for accessing a memcg's index. It will be used as an index in the
1394 * child cache array in kmem_cache, and also to derive its name. This function
1395 * will return -1 when this is not a kmem-limited memcg.
1397 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1399 return memcg ? memcg->kmemcg_id : -1;
1404 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1409 static inline void memcg_kmem_uncharge(struct page *page, int order)
1413 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1418 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1422 #define for_each_memcg_cache_index(_idx) \
1425 static inline bool memcg_kmem_enabled(void)
1430 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1435 static inline void memcg_get_cache_ids(void)
1439 static inline void memcg_put_cache_ids(void)
1443 #endif /* CONFIG_MEMCG_KMEM */
1445 #endif /* _LINUX_MEMCONTROL_H */