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 struct mem_cgroup *memcg; /* Back pointer, we cannot */
138 /* use container_of */
141 struct mem_cgroup_threshold {
142 struct eventfd_ctx *eventfd;
143 unsigned long threshold;
147 struct mem_cgroup_threshold_ary {
148 /* An array index points to threshold just below or equal to usage. */
149 int current_threshold;
150 /* Size of entries[] */
152 /* Array of thresholds */
153 struct mem_cgroup_threshold entries[0];
156 struct mem_cgroup_thresholds {
157 /* Primary thresholds array */
158 struct mem_cgroup_threshold_ary *primary;
160 * Spare threshold array.
161 * This is needed to make mem_cgroup_unregister_event() "never fail".
162 * It must be able to store at least primary->size - 1 entries.
164 struct mem_cgroup_threshold_ary *spare;
167 enum memcg_kmem_state {
173 #if defined(CONFIG_SMP)
174 struct memcg_padding {
176 } ____cacheline_internodealigned_in_smp;
177 #define MEMCG_PADDING(name) struct memcg_padding name;
179 #define MEMCG_PADDING(name)
183 * Remember four most recent foreign writebacks with dirty pages in this
184 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
185 * one in a given round, we're likely to catch it later if it keeps
186 * foreign-dirtying, so a fairly low count should be enough.
188 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
190 #define MEMCG_CGWB_FRN_CNT 4
192 struct memcg_cgwb_frn {
193 u64 bdi_id; /* bdi->id of the foreign inode */
194 int memcg_id; /* memcg->css.id of foreign inode */
195 u64 at; /* jiffies_64 at the time of dirtying */
196 struct wb_completion done; /* tracks in-flight foreign writebacks */
200 * The memory controller data structure. The memory controller controls both
201 * page cache and RSS per cgroup. We would eventually like to provide
202 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
203 * to help the administrator determine what knobs to tune.
206 struct cgroup_subsys_state css;
208 /* Private memcg ID. Used to ID objects that outlive the cgroup */
209 struct mem_cgroup_id id;
211 /* Accounted resources */
212 struct page_counter memory;
213 struct page_counter swap;
215 /* Legacy consumer-oriented counters */
216 struct page_counter memsw;
217 struct page_counter kmem;
218 struct page_counter tcpmem;
220 /* Upper bound of normal memory consumption range */
223 /* Range enforcement for interrupt charges */
224 struct work_struct high_work;
226 unsigned long soft_limit;
228 /* vmpressure notifications */
229 struct vmpressure vmpressure;
232 * Should the accounting and control be hierarchical, per subtree?
237 * Should the OOM killer kill all belonging tasks, had it kill one?
241 /* protected by memcg_oom_lock */
246 /* OOM-Killer disable */
247 int oom_kill_disable;
249 /* memory.events and memory.events.local */
250 struct cgroup_file events_file;
251 struct cgroup_file events_local_file;
253 /* handle for "memory.swap.events" */
254 struct cgroup_file swap_events_file;
256 /* protect arrays of thresholds */
257 struct mutex thresholds_lock;
259 /* thresholds for memory usage. RCU-protected */
260 struct mem_cgroup_thresholds thresholds;
262 /* thresholds for mem+swap usage. RCU-protected */
263 struct mem_cgroup_thresholds memsw_thresholds;
265 /* For oom notifier event fd */
266 struct list_head oom_notify;
269 * Should we move charges of a task when a task is moved into this
270 * mem_cgroup ? And what type of charges should we move ?
272 unsigned long move_charge_at_immigrate;
273 /* taken only while moving_account > 0 */
274 spinlock_t move_lock;
275 unsigned long move_lock_flags;
277 MEMCG_PADDING(_pad1_);
280 * set > 0 if pages under this cgroup are moving to other cgroup.
282 atomic_t moving_account;
283 struct task_struct *move_lock_task;
285 /* Legacy local VM stats and events */
286 struct memcg_vmstats_percpu __percpu *vmstats_local;
288 /* Subtree VM stats and events (batched updates) */
289 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
291 MEMCG_PADDING(_pad2_);
293 atomic_long_t vmstats[MEMCG_NR_STAT];
294 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
297 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
298 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
300 unsigned long socket_pressure;
302 /* Legacy tcp memory accounting */
306 #ifdef CONFIG_MEMCG_KMEM
307 /* Index in the kmem_cache->memcg_params.memcg_caches array */
309 enum memcg_kmem_state kmem_state;
310 struct list_head kmem_caches;
313 int last_scanned_node;
315 nodemask_t scan_nodes;
316 atomic_t numainfo_events;
317 atomic_t numainfo_updating;
320 #ifdef CONFIG_CGROUP_WRITEBACK
321 struct list_head cgwb_list;
322 struct wb_domain cgwb_domain;
323 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
326 /* List of events which userspace want to receive */
327 struct list_head event_list;
328 spinlock_t event_list_lock;
330 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
331 struct deferred_split deferred_split_queue;
334 struct mem_cgroup_per_node *nodeinfo[0];
335 /* WARNING: nodeinfo must be the last member here */
339 * size of first charge trial. "32" comes from vmscan.c's magic value.
340 * TODO: maybe necessary to use big numbers in big irons.
342 #define MEMCG_CHARGE_BATCH 32U
344 extern struct mem_cgroup *root_mem_cgroup;
346 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
348 return (memcg == root_mem_cgroup);
351 static inline bool mem_cgroup_disabled(void)
353 return !cgroup_subsys_enabled(memory_cgrp_subsys);
356 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
359 if (mem_cgroup_disabled())
363 return READ_ONCE(memcg->memory.emin);
365 return max(READ_ONCE(memcg->memory.emin),
366 READ_ONCE(memcg->memory.elow));
369 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
370 struct mem_cgroup *memcg);
372 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
373 gfp_t gfp_mask, struct mem_cgroup **memcgp,
375 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
376 gfp_t gfp_mask, struct mem_cgroup **memcgp,
378 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
379 bool lrucare, bool compound);
380 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
382 void mem_cgroup_uncharge(struct page *page);
383 void mem_cgroup_uncharge_list(struct list_head *page_list);
385 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
387 static struct mem_cgroup_per_node *
388 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
390 return memcg->nodeinfo[nid];
394 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
395 * @memcg: memcg of the wanted lruvec
397 * Returns the lru list vector holding pages for a given @memcg &
398 * @node combination. This can be the node lruvec, if the memory
399 * controller is disabled.
401 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
402 struct pglist_data *pgdat)
404 struct mem_cgroup_per_node *mz;
405 struct lruvec *lruvec;
407 if (mem_cgroup_disabled()) {
408 lruvec = &pgdat->__lruvec;
413 memcg = root_mem_cgroup;
415 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
416 lruvec = &mz->lruvec;
419 * Since a node can be onlined after the mem_cgroup was created,
420 * we have to be prepared to initialize lruvec->pgdat here;
421 * and if offlined then reonlined, we need to reinitialize it.
423 if (unlikely(lruvec->pgdat != pgdat))
424 lruvec->pgdat = pgdat;
428 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
430 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
432 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
434 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
437 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
438 return css ? container_of(css, struct mem_cgroup, css) : NULL;
441 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
444 css_put(&memcg->css);
447 #define mem_cgroup_from_counter(counter, member) \
448 container_of(counter, struct mem_cgroup, member)
450 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
452 struct mem_cgroup_reclaim_cookie *);
453 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
454 int mem_cgroup_scan_tasks(struct mem_cgroup *,
455 int (*)(struct task_struct *, void *), void *);
457 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
459 if (mem_cgroup_disabled())
464 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
466 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
468 return mem_cgroup_from_css(seq_css(m));
471 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
473 struct mem_cgroup_per_node *mz;
475 if (mem_cgroup_disabled())
478 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
483 * parent_mem_cgroup - find the accounting parent of a memcg
484 * @memcg: memcg whose parent to find
486 * Returns the parent memcg, or NULL if this is the root or the memory
487 * controller is in legacy no-hierarchy mode.
489 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
491 if (!memcg->memory.parent)
493 return mem_cgroup_from_counter(memcg->memory.parent, memory);
496 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
497 struct mem_cgroup *root)
501 if (!root->use_hierarchy)
503 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
506 static inline bool mm_match_cgroup(struct mm_struct *mm,
507 struct mem_cgroup *memcg)
509 struct mem_cgroup *task_memcg;
513 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
515 match = mem_cgroup_is_descendant(task_memcg, memcg);
520 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
521 ino_t page_cgroup_ino(struct page *page);
523 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
525 if (mem_cgroup_disabled())
527 return !!(memcg->css.flags & CSS_ONLINE);
531 * For memory reclaim.
533 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
535 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
536 int zid, int nr_pages);
539 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
540 enum lru_list lru, int zone_idx)
542 struct mem_cgroup_per_node *mz;
544 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
545 return mz->lru_zone_size[zone_idx][lru];
548 void mem_cgroup_handle_over_high(void);
550 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
552 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
554 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
555 struct task_struct *p);
557 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
559 static inline void mem_cgroup_enter_user_fault(void)
561 WARN_ON(current->in_user_fault);
562 current->in_user_fault = 1;
565 static inline void mem_cgroup_exit_user_fault(void)
567 WARN_ON(!current->in_user_fault);
568 current->in_user_fault = 0;
571 static inline bool task_in_memcg_oom(struct task_struct *p)
573 return p->memcg_in_oom;
576 bool mem_cgroup_oom_synchronize(bool wait);
577 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
578 struct mem_cgroup *oom_domain);
579 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
581 #ifdef CONFIG_MEMCG_SWAP
582 extern int do_swap_account;
585 struct mem_cgroup *lock_page_memcg(struct page *page);
586 void __unlock_page_memcg(struct mem_cgroup *memcg);
587 void unlock_page_memcg(struct page *page);
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(struct mem_cgroup *memcg, int idx)
595 long x = atomic_long_read(&memcg->vmstats[idx]);
604 * idx can be of type enum memcg_stat_item or node_stat_item.
605 * Keep in sync with memcg_exact_page_state().
607 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
613 for_each_possible_cpu(cpu)
614 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
622 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
624 /* idx can be of type enum memcg_stat_item or node_stat_item */
625 static inline void mod_memcg_state(struct mem_cgroup *memcg,
630 local_irq_save(flags);
631 __mod_memcg_state(memcg, idx, val);
632 local_irq_restore(flags);
636 * mod_memcg_page_state - update page state statistics
638 * @idx: page state item to account
639 * @val: number of pages (positive or negative)
641 * The @page must be locked or the caller must use lock_page_memcg()
642 * to prevent double accounting when the page is concurrently being
643 * moved to another memcg:
645 * lock_page(page) or lock_page_memcg(page)
646 * if (TestClearPageState(page))
647 * mod_memcg_page_state(page, state, -1);
648 * unlock_page(page) or unlock_page_memcg(page)
650 * Kernel pages are an exception to this, since they'll never move.
652 static inline void __mod_memcg_page_state(struct page *page,
655 if (page->mem_cgroup)
656 __mod_memcg_state(page->mem_cgroup, idx, val);
659 static inline void mod_memcg_page_state(struct page *page,
662 if (page->mem_cgroup)
663 mod_memcg_state(page->mem_cgroup, idx, val);
666 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
667 enum node_stat_item idx)
669 struct mem_cgroup_per_node *pn;
672 if (mem_cgroup_disabled())
673 return node_page_state(lruvec_pgdat(lruvec), idx);
675 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
676 x = atomic_long_read(&pn->lruvec_stat[idx]);
684 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
685 enum node_stat_item idx)
687 struct mem_cgroup_per_node *pn;
691 if (mem_cgroup_disabled())
692 return node_page_state(lruvec_pgdat(lruvec), idx);
694 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
695 for_each_possible_cpu(cpu)
696 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
704 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
706 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
707 void mod_memcg_obj_state(void *p, int idx, int val);
709 static inline void mod_lruvec_state(struct lruvec *lruvec,
710 enum node_stat_item idx, int val)
714 local_irq_save(flags);
715 __mod_lruvec_state(lruvec, idx, val);
716 local_irq_restore(flags);
719 static inline void __mod_lruvec_page_state(struct page *page,
720 enum node_stat_item idx, int val)
722 pg_data_t *pgdat = page_pgdat(page);
723 struct lruvec *lruvec;
725 /* Untracked pages have no memcg, no lruvec. Update only the node */
726 if (!page->mem_cgroup) {
727 __mod_node_page_state(pgdat, idx, val);
731 lruvec = mem_cgroup_lruvec(page->mem_cgroup, pgdat);
732 __mod_lruvec_state(lruvec, idx, val);
735 static inline void mod_lruvec_page_state(struct page *page,
736 enum node_stat_item idx, int val)
740 local_irq_save(flags);
741 __mod_lruvec_page_state(page, idx, val);
742 local_irq_restore(flags);
745 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
747 unsigned long *total_scanned);
749 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
750 unsigned long count);
752 static inline void count_memcg_events(struct mem_cgroup *memcg,
753 enum vm_event_item idx,
758 local_irq_save(flags);
759 __count_memcg_events(memcg, idx, count);
760 local_irq_restore(flags);
763 static inline void count_memcg_page_event(struct page *page,
764 enum vm_event_item idx)
766 if (page->mem_cgroup)
767 count_memcg_events(page->mem_cgroup, idx, 1);
770 static inline void count_memcg_event_mm(struct mm_struct *mm,
771 enum vm_event_item idx)
773 struct mem_cgroup *memcg;
775 if (mem_cgroup_disabled())
779 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
781 count_memcg_events(memcg, idx, 1);
785 static inline void memcg_memory_event(struct mem_cgroup *memcg,
786 enum memcg_memory_event event)
788 atomic_long_inc(&memcg->memory_events_local[event]);
789 cgroup_file_notify(&memcg->events_local_file);
792 atomic_long_inc(&memcg->memory_events[event]);
793 cgroup_file_notify(&memcg->events_file);
795 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
797 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
799 } while ((memcg = parent_mem_cgroup(memcg)) &&
800 !mem_cgroup_is_root(memcg));
803 static inline void memcg_memory_event_mm(struct mm_struct *mm,
804 enum memcg_memory_event event)
806 struct mem_cgroup *memcg;
808 if (mem_cgroup_disabled())
812 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
814 memcg_memory_event(memcg, event);
818 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
819 void mem_cgroup_split_huge_fixup(struct page *head);
822 #else /* CONFIG_MEMCG */
824 #define MEM_CGROUP_ID_SHIFT 0
825 #define MEM_CGROUP_ID_MAX 0
829 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
834 static inline bool mem_cgroup_disabled(void)
839 static inline void memcg_memory_event(struct mem_cgroup *memcg,
840 enum memcg_memory_event event)
844 static inline void memcg_memory_event_mm(struct mm_struct *mm,
845 enum memcg_memory_event event)
849 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
855 static inline enum mem_cgroup_protection mem_cgroup_protected(
856 struct mem_cgroup *root, struct mem_cgroup *memcg)
858 return MEMCG_PROT_NONE;
861 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
863 struct mem_cgroup **memcgp,
870 static inline int mem_cgroup_try_charge_delay(struct page *page,
871 struct mm_struct *mm,
873 struct mem_cgroup **memcgp,
880 static inline void mem_cgroup_commit_charge(struct page *page,
881 struct mem_cgroup *memcg,
882 bool lrucare, bool compound)
886 static inline void mem_cgroup_cancel_charge(struct page *page,
887 struct mem_cgroup *memcg,
892 static inline void mem_cgroup_uncharge(struct page *page)
896 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
900 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
904 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
905 struct pglist_data *pgdat)
907 return &pgdat->__lruvec;
910 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
911 struct pglist_data *pgdat)
913 return &pgdat->__lruvec;
916 static inline bool mm_match_cgroup(struct mm_struct *mm,
917 struct mem_cgroup *memcg)
922 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
927 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
932 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
936 static inline struct mem_cgroup *
937 mem_cgroup_iter(struct mem_cgroup *root,
938 struct mem_cgroup *prev,
939 struct mem_cgroup_reclaim_cookie *reclaim)
944 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
945 struct mem_cgroup *prev)
949 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
950 int (*fn)(struct task_struct *, void *), void *arg)
955 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
960 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
963 /* XXX: This should always return root_mem_cgroup */
967 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
972 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
977 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
983 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
984 enum lru_list lru, int zone_idx)
989 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
994 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1000 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1005 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1009 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1014 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1018 static inline void unlock_page_memcg(struct page *page)
1022 static inline void mem_cgroup_handle_over_high(void)
1026 static inline void mem_cgroup_enter_user_fault(void)
1030 static inline void mem_cgroup_exit_user_fault(void)
1034 static inline bool task_in_memcg_oom(struct task_struct *p)
1039 static inline bool mem_cgroup_oom_synchronize(bool wait)
1044 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1045 struct task_struct *victim, struct mem_cgroup *oom_domain)
1050 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1054 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1059 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1065 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1071 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1077 static inline void __mod_memcg_page_state(struct page *page,
1083 static inline void mod_memcg_page_state(struct page *page,
1089 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1090 enum node_stat_item idx)
1092 return node_page_state(lruvec_pgdat(lruvec), idx);
1095 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1096 enum node_stat_item idx)
1098 return node_page_state(lruvec_pgdat(lruvec), idx);
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_state(struct lruvec *lruvec,
1108 enum node_stat_item idx, int val)
1110 mod_node_page_state(lruvec_pgdat(lruvec), 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_page_state(struct page *page,
1120 enum node_stat_item idx, int val)
1122 mod_node_page_state(page_pgdat(page), idx, val);
1125 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1128 struct page *page = virt_to_head_page(p);
1130 __mod_node_page_state(page_pgdat(page), idx, val);
1133 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1138 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1140 unsigned long *total_scanned)
1145 static inline void mem_cgroup_split_huge_fixup(struct page *head)
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_events(struct mem_cgroup *memcg,
1156 enum vm_event_item idx,
1157 unsigned long count)
1161 static inline void count_memcg_page_event(struct page *page,
1167 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1170 #endif /* CONFIG_MEMCG */
1172 /* idx can be of type enum memcg_stat_item or node_stat_item */
1173 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1176 __mod_memcg_state(memcg, idx, 1);
1179 /* idx can be of type enum memcg_stat_item or node_stat_item */
1180 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1183 __mod_memcg_state(memcg, idx, -1);
1186 /* idx can be of type enum memcg_stat_item or node_stat_item */
1187 static inline void __inc_memcg_page_state(struct page *page,
1190 __mod_memcg_page_state(page, idx, 1);
1193 /* idx can be of type enum memcg_stat_item or node_stat_item */
1194 static inline void __dec_memcg_page_state(struct page *page,
1197 __mod_memcg_page_state(page, idx, -1);
1200 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1201 enum node_stat_item idx)
1203 __mod_lruvec_state(lruvec, idx, 1);
1206 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1207 enum node_stat_item idx)
1209 __mod_lruvec_state(lruvec, idx, -1);
1212 static inline void __inc_lruvec_page_state(struct page *page,
1213 enum node_stat_item idx)
1215 __mod_lruvec_page_state(page, idx, 1);
1218 static inline void __dec_lruvec_page_state(struct page *page,
1219 enum node_stat_item idx)
1221 __mod_lruvec_page_state(page, idx, -1);
1224 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1226 __mod_lruvec_slab_state(p, idx, 1);
1229 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1231 __mod_lruvec_slab_state(p, idx, -1);
1234 /* idx can be of type enum memcg_stat_item or node_stat_item */
1235 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1238 mod_memcg_state(memcg, idx, 1);
1241 /* idx can be of type enum memcg_stat_item or node_stat_item */
1242 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1245 mod_memcg_state(memcg, idx, -1);
1248 /* idx can be of type enum memcg_stat_item or node_stat_item */
1249 static inline void inc_memcg_page_state(struct page *page,
1252 mod_memcg_page_state(page, idx, 1);
1255 /* idx can be of type enum memcg_stat_item or node_stat_item */
1256 static inline void dec_memcg_page_state(struct page *page,
1259 mod_memcg_page_state(page, idx, -1);
1262 static inline void inc_lruvec_state(struct lruvec *lruvec,
1263 enum node_stat_item idx)
1265 mod_lruvec_state(lruvec, idx, 1);
1268 static inline void dec_lruvec_state(struct lruvec *lruvec,
1269 enum node_stat_item idx)
1271 mod_lruvec_state(lruvec, idx, -1);
1274 static inline void inc_lruvec_page_state(struct page *page,
1275 enum node_stat_item idx)
1277 mod_lruvec_page_state(page, idx, 1);
1280 static inline void dec_lruvec_page_state(struct page *page,
1281 enum node_stat_item idx)
1283 mod_lruvec_page_state(page, idx, -1);
1286 #ifdef CONFIG_CGROUP_WRITEBACK
1288 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1289 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1290 unsigned long *pheadroom, unsigned long *pdirty,
1291 unsigned long *pwriteback);
1293 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1294 struct bdi_writeback *wb);
1296 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1297 struct bdi_writeback *wb)
1299 if (mem_cgroup_disabled())
1302 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1303 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1306 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1308 #else /* CONFIG_CGROUP_WRITEBACK */
1310 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1315 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1316 unsigned long *pfilepages,
1317 unsigned long *pheadroom,
1318 unsigned long *pdirty,
1319 unsigned long *pwriteback)
1323 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1324 struct bdi_writeback *wb)
1328 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1332 #endif /* CONFIG_CGROUP_WRITEBACK */
1335 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1336 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1338 extern struct static_key_false memcg_sockets_enabled_key;
1339 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1340 void mem_cgroup_sk_alloc(struct sock *sk);
1341 void mem_cgroup_sk_free(struct sock *sk);
1342 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1344 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1347 if (time_before(jiffies, memcg->socket_pressure))
1349 } while ((memcg = parent_mem_cgroup(memcg)));
1353 extern int memcg_expand_shrinker_maps(int new_id);
1355 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1356 int nid, int shrinker_id);
1358 #define mem_cgroup_sockets_enabled 0
1359 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1360 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1361 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1366 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1367 int nid, int shrinker_id)
1372 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1373 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1375 #ifdef CONFIG_MEMCG_KMEM
1376 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1377 void __memcg_kmem_uncharge(struct page *page, int order);
1378 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1379 struct mem_cgroup *memcg);
1380 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1381 unsigned int nr_pages);
1383 extern struct static_key_false memcg_kmem_enabled_key;
1384 extern struct workqueue_struct *memcg_kmem_cache_wq;
1386 extern int memcg_nr_cache_ids;
1387 void memcg_get_cache_ids(void);
1388 void memcg_put_cache_ids(void);
1391 * Helper macro to loop through all memcg-specific caches. Callers must still
1392 * check if the cache is valid (it is either valid or NULL).
1393 * the slab_mutex must be held when looping through those caches
1395 #define for_each_memcg_cache_index(_idx) \
1396 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1398 static inline bool memcg_kmem_enabled(void)
1400 return static_branch_unlikely(&memcg_kmem_enabled_key);
1403 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1405 if (memcg_kmem_enabled())
1406 return __memcg_kmem_charge(page, gfp, order);
1410 static inline void memcg_kmem_uncharge(struct page *page, int order)
1412 if (memcg_kmem_enabled())
1413 __memcg_kmem_uncharge(page, order);
1416 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1417 int order, struct mem_cgroup *memcg)
1419 if (memcg_kmem_enabled())
1420 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1424 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1425 struct mem_cgroup *memcg)
1427 if (memcg_kmem_enabled())
1428 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1432 * helper for accessing a memcg's index. It will be used as an index in the
1433 * child cache array in kmem_cache, and also to derive its name. This function
1434 * will return -1 when this is not a kmem-limited memcg.
1436 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1438 return memcg ? memcg->kmemcg_id : -1;
1441 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1445 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1450 static inline void memcg_kmem_uncharge(struct page *page, int order)
1454 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1459 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1463 #define for_each_memcg_cache_index(_idx) \
1466 static inline bool memcg_kmem_enabled(void)
1471 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1476 static inline void memcg_get_cache_ids(void)
1480 static inline void memcg_put_cache_ids(void)
1484 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1489 #endif /* CONFIG_MEMCG_KMEM */
1491 #endif /* _LINUX_MEMCONTROL_H */