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>
31 /* Cgroup-specific page state, on top of universal node page state */
32 enum memcg_stat_item {
33 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS,
35 /* XXX: why are these zone and not node counters? */
36 MEMCG_KERNEL_STACK_KB,
40 enum memcg_memory_event {
49 MEMCG_NR_MEMORY_EVENTS,
52 enum mem_cgroup_protection {
58 struct mem_cgroup_reclaim_cookie {
60 unsigned int generation;
65 #define MEM_CGROUP_ID_SHIFT 16
66 #define MEM_CGROUP_ID_MAX USHRT_MAX
68 struct mem_cgroup_id {
74 * Per memcg event counter is incremented at every pagein/pageout. With THP,
75 * it will be incremated by the number of pages. This counter is used for
76 * for trigger some periodic events. This is straightforward and better
77 * than using jiffies etc. to handle periodic memcg event.
79 enum mem_cgroup_events_target {
80 MEM_CGROUP_TARGET_THRESH,
81 MEM_CGROUP_TARGET_SOFTLIMIT,
85 struct memcg_vmstats_percpu {
86 long stat[MEMCG_NR_STAT];
87 unsigned long events[NR_VM_EVENT_ITEMS];
88 unsigned long nr_page_events;
89 unsigned long targets[MEM_CGROUP_NTARGETS];
92 struct mem_cgroup_reclaim_iter {
93 struct mem_cgroup *position;
94 /* scan generation, increased every round-trip */
95 unsigned int generation;
99 long count[NR_VM_NODE_STAT_ITEMS];
103 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
104 * which have elements charged to this memcg.
106 struct memcg_shrinker_map {
112 * per-node information in memory controller.
114 struct mem_cgroup_per_node {
115 struct lruvec lruvec;
117 /* Legacy local VM stats */
118 struct lruvec_stat __percpu *lruvec_stat_local;
120 /* Subtree VM stats (batched updates) */
121 struct lruvec_stat __percpu *lruvec_stat_cpu;
122 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
124 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
126 struct mem_cgroup_reclaim_iter iter;
128 struct memcg_shrinker_map __rcu *shrinker_map;
130 struct rb_node tree_node; /* RB tree node */
131 unsigned long usage_in_excess;/* Set to the value by which */
132 /* the soft limit is exceeded*/
134 struct mem_cgroup *memcg; /* Back pointer, we cannot */
135 /* use container_of */
138 struct mem_cgroup_threshold {
139 struct eventfd_ctx *eventfd;
140 unsigned long threshold;
144 struct mem_cgroup_threshold_ary {
145 /* An array index points to threshold just below or equal to usage. */
146 int current_threshold;
147 /* Size of entries[] */
149 /* Array of thresholds */
150 struct mem_cgroup_threshold entries[];
153 struct mem_cgroup_thresholds {
154 /* Primary thresholds array */
155 struct mem_cgroup_threshold_ary *primary;
157 * Spare threshold array.
158 * This is needed to make mem_cgroup_unregister_event() "never fail".
159 * It must be able to store at least primary->size - 1 entries.
161 struct mem_cgroup_threshold_ary *spare;
164 enum memcg_kmem_state {
170 #if defined(CONFIG_SMP)
171 struct memcg_padding {
173 } ____cacheline_internodealigned_in_smp;
174 #define MEMCG_PADDING(name) struct memcg_padding name;
176 #define MEMCG_PADDING(name)
180 * Remember four most recent foreign writebacks with dirty pages in this
181 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
182 * one in a given round, we're likely to catch it later if it keeps
183 * foreign-dirtying, so a fairly low count should be enough.
185 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
187 #define MEMCG_CGWB_FRN_CNT 4
189 struct memcg_cgwb_frn {
190 u64 bdi_id; /* bdi->id of the foreign inode */
191 int memcg_id; /* memcg->css.id of foreign inode */
192 u64 at; /* jiffies_64 at the time of dirtying */
193 struct wb_completion done; /* tracks in-flight foreign writebacks */
197 * Bucket for arbitrarily byte-sized objects charged to a memory
198 * cgroup. The bucket can be reparented in one piece when the cgroup
199 * is destroyed, without having to round up the individual references
200 * of all live memory objects in the wild.
203 struct percpu_ref refcnt;
204 struct mem_cgroup *memcg;
205 atomic_t nr_charged_bytes;
207 struct list_head list;
213 * The memory controller data structure. The memory controller controls both
214 * page cache and RSS per cgroup. We would eventually like to provide
215 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
216 * to help the administrator determine what knobs to tune.
219 struct cgroup_subsys_state css;
221 /* Private memcg ID. Used to ID objects that outlive the cgroup */
222 struct mem_cgroup_id id;
224 /* Accounted resources */
225 struct page_counter memory;
226 struct page_counter swap;
228 /* Legacy consumer-oriented counters */
229 struct page_counter memsw;
230 struct page_counter kmem;
231 struct page_counter tcpmem;
233 /* Range enforcement for interrupt charges */
234 struct work_struct high_work;
236 unsigned long soft_limit;
238 /* vmpressure notifications */
239 struct vmpressure vmpressure;
242 * Should the accounting and control be hierarchical, per subtree?
247 * Should the OOM killer kill all belonging tasks, had it kill one?
251 /* protected by memcg_oom_lock */
256 /* OOM-Killer disable */
257 int oom_kill_disable;
259 /* memory.events and memory.events.local */
260 struct cgroup_file events_file;
261 struct cgroup_file events_local_file;
263 /* handle for "memory.swap.events" */
264 struct cgroup_file swap_events_file;
266 /* protect arrays of thresholds */
267 struct mutex thresholds_lock;
269 /* thresholds for memory usage. RCU-protected */
270 struct mem_cgroup_thresholds thresholds;
272 /* thresholds for mem+swap usage. RCU-protected */
273 struct mem_cgroup_thresholds memsw_thresholds;
275 /* For oom notifier event fd */
276 struct list_head oom_notify;
279 * Should we move charges of a task when a task is moved into this
280 * mem_cgroup ? And what type of charges should we move ?
282 unsigned long move_charge_at_immigrate;
283 /* taken only while moving_account > 0 */
284 spinlock_t move_lock;
285 unsigned long move_lock_flags;
287 MEMCG_PADDING(_pad1_);
290 * set > 0 if pages under this cgroup are moving to other cgroup.
292 atomic_t moving_account;
293 struct task_struct *move_lock_task;
295 /* Legacy local VM stats and events */
296 struct memcg_vmstats_percpu __percpu *vmstats_local;
298 /* Subtree VM stats and events (batched updates) */
299 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
301 MEMCG_PADDING(_pad2_);
303 atomic_long_t vmstats[MEMCG_NR_STAT];
304 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
307 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
308 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
310 unsigned long socket_pressure;
312 /* Legacy tcp memory accounting */
316 #ifdef CONFIG_MEMCG_KMEM
317 /* Index in the kmem_cache->memcg_params.memcg_caches array */
319 enum memcg_kmem_state kmem_state;
320 struct list_head kmem_caches;
321 struct obj_cgroup __rcu *objcg;
322 struct list_head objcg_list; /* list of inherited objcgs */
325 #ifdef CONFIG_CGROUP_WRITEBACK
326 struct list_head cgwb_list;
327 struct wb_domain cgwb_domain;
328 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
331 /* List of events which userspace want to receive */
332 struct list_head event_list;
333 spinlock_t event_list_lock;
335 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
336 struct deferred_split deferred_split_queue;
339 struct mem_cgroup_per_node *nodeinfo[0];
340 /* WARNING: nodeinfo must be the last member here */
344 * size of first charge trial. "32" comes from vmscan.c's magic value.
345 * TODO: maybe necessary to use big numbers in big irons.
347 #define MEMCG_CHARGE_BATCH 32U
349 extern struct mem_cgroup *root_mem_cgroup;
351 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
353 return (memcg == root_mem_cgroup);
356 static inline bool mem_cgroup_disabled(void)
358 return !cgroup_subsys_enabled(memory_cgrp_subsys);
361 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
364 if (mem_cgroup_disabled())
368 return READ_ONCE(memcg->memory.emin);
370 return max(READ_ONCE(memcg->memory.emin),
371 READ_ONCE(memcg->memory.elow));
374 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
375 struct mem_cgroup *memcg);
377 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask);
379 void mem_cgroup_uncharge(struct page *page);
380 void mem_cgroup_uncharge_list(struct list_head *page_list);
382 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
384 static struct mem_cgroup_per_node *
385 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
387 return memcg->nodeinfo[nid];
391 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
392 * @memcg: memcg of the wanted lruvec
394 * Returns the lru list vector holding pages for a given @memcg &
395 * @node combination. This can be the node lruvec, if the memory
396 * controller is disabled.
398 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
399 struct pglist_data *pgdat)
401 struct mem_cgroup_per_node *mz;
402 struct lruvec *lruvec;
404 if (mem_cgroup_disabled()) {
405 lruvec = &pgdat->__lruvec;
410 memcg = root_mem_cgroup;
412 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
413 lruvec = &mz->lruvec;
416 * Since a node can be onlined after the mem_cgroup was created,
417 * we have to be prepared to initialize lruvec->pgdat here;
418 * and if offlined then reonlined, we need to reinitialize it.
420 if (unlikely(lruvec->pgdat != pgdat))
421 lruvec->pgdat = pgdat;
425 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
427 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
429 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
431 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
434 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
435 return css ? container_of(css, struct mem_cgroup, css) : NULL;
438 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
440 return percpu_ref_tryget(&objcg->refcnt);
443 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
445 percpu_ref_get(&objcg->refcnt);
448 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
450 percpu_ref_put(&objcg->refcnt);
454 * After the initialization objcg->memcg is always pointing at
455 * a valid memcg, but can be atomically swapped to the parent memcg.
457 * The caller must ensure that the returned memcg won't be released:
458 * e.g. acquire the rcu_read_lock or css_set_lock.
460 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
462 return READ_ONCE(objcg->memcg);
465 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
468 css_put(&memcg->css);
471 #define mem_cgroup_from_counter(counter, member) \
472 container_of(counter, struct mem_cgroup, member)
474 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
476 struct mem_cgroup_reclaim_cookie *);
477 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
478 int mem_cgroup_scan_tasks(struct mem_cgroup *,
479 int (*)(struct task_struct *, void *), void *);
481 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
483 if (mem_cgroup_disabled())
488 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
490 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
492 return mem_cgroup_from_css(seq_css(m));
495 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
497 struct mem_cgroup_per_node *mz;
499 if (mem_cgroup_disabled())
502 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
507 * parent_mem_cgroup - find the accounting parent of a memcg
508 * @memcg: memcg whose parent to find
510 * Returns the parent memcg, or NULL if this is the root or the memory
511 * controller is in legacy no-hierarchy mode.
513 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
515 if (!memcg->memory.parent)
517 return mem_cgroup_from_counter(memcg->memory.parent, memory);
520 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
521 struct mem_cgroup *root)
525 if (!root->use_hierarchy)
527 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
530 static inline bool mm_match_cgroup(struct mm_struct *mm,
531 struct mem_cgroup *memcg)
533 struct mem_cgroup *task_memcg;
537 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
539 match = mem_cgroup_is_descendant(task_memcg, memcg);
544 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
545 ino_t page_cgroup_ino(struct page *page);
547 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
549 if (mem_cgroup_disabled())
551 return !!(memcg->css.flags & CSS_ONLINE);
555 * For memory reclaim.
557 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
559 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
560 int zid, int nr_pages);
563 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
564 enum lru_list lru, int zone_idx)
566 struct mem_cgroup_per_node *mz;
568 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
569 return mz->lru_zone_size[zone_idx][lru];
572 void mem_cgroup_handle_over_high(void);
574 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
576 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
578 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
579 struct task_struct *p);
581 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
583 static inline void mem_cgroup_enter_user_fault(void)
585 WARN_ON(current->in_user_fault);
586 current->in_user_fault = 1;
589 static inline void mem_cgroup_exit_user_fault(void)
591 WARN_ON(!current->in_user_fault);
592 current->in_user_fault = 0;
595 static inline bool task_in_memcg_oom(struct task_struct *p)
597 return p->memcg_in_oom;
600 bool mem_cgroup_oom_synchronize(bool wait);
601 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
602 struct mem_cgroup *oom_domain);
603 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
605 #ifdef CONFIG_MEMCG_SWAP
606 extern bool cgroup_memory_noswap;
609 struct mem_cgroup *lock_page_memcg(struct page *page);
610 void __unlock_page_memcg(struct mem_cgroup *memcg);
611 void unlock_page_memcg(struct page *page);
614 * idx can be of type enum memcg_stat_item or node_stat_item.
615 * Keep in sync with memcg_exact_page_state().
617 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
619 long x = atomic_long_read(&memcg->vmstats[idx]);
628 * idx can be of type enum memcg_stat_item or node_stat_item.
629 * Keep in sync with memcg_exact_page_state().
631 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
637 for_each_possible_cpu(cpu)
638 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
646 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
648 /* idx can be of type enum memcg_stat_item or node_stat_item */
649 static inline void mod_memcg_state(struct mem_cgroup *memcg,
654 local_irq_save(flags);
655 __mod_memcg_state(memcg, idx, val);
656 local_irq_restore(flags);
660 * mod_memcg_page_state - update page state statistics
662 * @idx: page state item to account
663 * @val: number of pages (positive or negative)
665 * The @page must be locked or the caller must use lock_page_memcg()
666 * to prevent double accounting when the page is concurrently being
667 * moved to another memcg:
669 * lock_page(page) or lock_page_memcg(page)
670 * if (TestClearPageState(page))
671 * mod_memcg_page_state(page, state, -1);
672 * unlock_page(page) or unlock_page_memcg(page)
674 * Kernel pages are an exception to this, since they'll never move.
676 static inline void __mod_memcg_page_state(struct page *page,
679 if (page->mem_cgroup)
680 __mod_memcg_state(page->mem_cgroup, idx, val);
683 static inline void mod_memcg_page_state(struct page *page,
686 if (page->mem_cgroup)
687 mod_memcg_state(page->mem_cgroup, idx, val);
690 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
691 enum node_stat_item idx)
693 struct mem_cgroup_per_node *pn;
696 if (mem_cgroup_disabled())
697 return node_page_state(lruvec_pgdat(lruvec), idx);
699 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
700 x = atomic_long_read(&pn->lruvec_stat[idx]);
708 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
709 enum node_stat_item idx)
711 struct mem_cgroup_per_node *pn;
715 if (mem_cgroup_disabled())
716 return node_page_state(lruvec_pgdat(lruvec), idx);
718 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
719 for_each_possible_cpu(cpu)
720 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
728 void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
730 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
732 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
733 void mod_memcg_obj_state(void *p, int idx, int val);
735 static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
736 enum node_stat_item idx, int val)
740 local_irq_save(flags);
741 __mod_memcg_lruvec_state(lruvec, idx, val);
742 local_irq_restore(flags);
745 static inline void mod_lruvec_state(struct lruvec *lruvec,
746 enum node_stat_item idx, int val)
750 local_irq_save(flags);
751 __mod_lruvec_state(lruvec, idx, val);
752 local_irq_restore(flags);
755 static inline void __mod_lruvec_page_state(struct page *page,
756 enum node_stat_item idx, int val)
758 struct page *head = compound_head(page); /* rmap on tail pages */
759 pg_data_t *pgdat = page_pgdat(page);
760 struct lruvec *lruvec;
762 /* Untracked pages have no memcg, no lruvec. Update only the node */
763 if (!head->mem_cgroup) {
764 __mod_node_page_state(pgdat, idx, val);
768 lruvec = mem_cgroup_lruvec(head->mem_cgroup, pgdat);
769 __mod_lruvec_state(lruvec, idx, val);
772 static inline void mod_lruvec_page_state(struct page *page,
773 enum node_stat_item idx, int val)
777 local_irq_save(flags);
778 __mod_lruvec_page_state(page, idx, val);
779 local_irq_restore(flags);
782 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
784 unsigned long *total_scanned);
786 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
787 unsigned long count);
789 static inline void count_memcg_events(struct mem_cgroup *memcg,
790 enum vm_event_item idx,
795 local_irq_save(flags);
796 __count_memcg_events(memcg, idx, count);
797 local_irq_restore(flags);
800 static inline void count_memcg_page_event(struct page *page,
801 enum vm_event_item idx)
803 if (page->mem_cgroup)
804 count_memcg_events(page->mem_cgroup, idx, 1);
807 static inline void count_memcg_event_mm(struct mm_struct *mm,
808 enum vm_event_item idx)
810 struct mem_cgroup *memcg;
812 if (mem_cgroup_disabled())
816 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
818 count_memcg_events(memcg, idx, 1);
822 static inline void memcg_memory_event(struct mem_cgroup *memcg,
823 enum memcg_memory_event event)
825 atomic_long_inc(&memcg->memory_events_local[event]);
826 cgroup_file_notify(&memcg->events_local_file);
829 atomic_long_inc(&memcg->memory_events[event]);
830 cgroup_file_notify(&memcg->events_file);
832 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
834 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
836 } while ((memcg = parent_mem_cgroup(memcg)) &&
837 !mem_cgroup_is_root(memcg));
840 static inline void memcg_memory_event_mm(struct mm_struct *mm,
841 enum memcg_memory_event event)
843 struct mem_cgroup *memcg;
845 if (mem_cgroup_disabled())
849 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
851 memcg_memory_event(memcg, event);
855 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
856 void mem_cgroup_split_huge_fixup(struct page *head);
859 #else /* CONFIG_MEMCG */
861 #define MEM_CGROUP_ID_SHIFT 0
862 #define MEM_CGROUP_ID_MAX 0
866 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
871 static inline bool mem_cgroup_disabled(void)
876 static inline void memcg_memory_event(struct mem_cgroup *memcg,
877 enum memcg_memory_event event)
881 static inline void memcg_memory_event_mm(struct mm_struct *mm,
882 enum memcg_memory_event event)
886 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
892 static inline enum mem_cgroup_protection mem_cgroup_protected(
893 struct mem_cgroup *root, struct mem_cgroup *memcg)
895 return MEMCG_PROT_NONE;
898 static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
904 static inline void mem_cgroup_uncharge(struct page *page)
908 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
912 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
916 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
917 struct pglist_data *pgdat)
919 return &pgdat->__lruvec;
922 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
923 struct pglist_data *pgdat)
925 return &pgdat->__lruvec;
928 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
933 static inline bool mm_match_cgroup(struct mm_struct *mm,
934 struct mem_cgroup *memcg)
939 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
944 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
949 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
953 static inline struct mem_cgroup *
954 mem_cgroup_iter(struct mem_cgroup *root,
955 struct mem_cgroup *prev,
956 struct mem_cgroup_reclaim_cookie *reclaim)
961 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
962 struct mem_cgroup *prev)
966 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
967 int (*fn)(struct task_struct *, void *), void *arg)
972 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
977 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
980 /* XXX: This should always return root_mem_cgroup */
984 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
989 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
994 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1000 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1001 enum lru_list lru, int zone_idx)
1006 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1011 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1017 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1022 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1026 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1031 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1035 static inline void unlock_page_memcg(struct page *page)
1039 static inline void mem_cgroup_handle_over_high(void)
1043 static inline void mem_cgroup_enter_user_fault(void)
1047 static inline void mem_cgroup_exit_user_fault(void)
1051 static inline bool task_in_memcg_oom(struct task_struct *p)
1056 static inline bool mem_cgroup_oom_synchronize(bool wait)
1061 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1062 struct task_struct *victim, struct mem_cgroup *oom_domain)
1067 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1071 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1076 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1082 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1088 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1094 static inline void __mod_memcg_page_state(struct page *page,
1100 static inline void mod_memcg_page_state(struct page *page,
1106 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1107 enum node_stat_item idx)
1109 return node_page_state(lruvec_pgdat(lruvec), idx);
1112 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1113 enum node_stat_item idx)
1115 return node_page_state(lruvec_pgdat(lruvec), idx);
1118 static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1119 enum node_stat_item idx, int val)
1123 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1124 enum node_stat_item idx, int val)
1126 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1129 static inline void mod_lruvec_state(struct lruvec *lruvec,
1130 enum node_stat_item idx, int val)
1132 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1135 static inline void __mod_lruvec_page_state(struct page *page,
1136 enum node_stat_item idx, int val)
1138 __mod_node_page_state(page_pgdat(page), idx, val);
1141 static inline void mod_lruvec_page_state(struct page *page,
1142 enum node_stat_item idx, int val)
1144 mod_node_page_state(page_pgdat(page), idx, val);
1147 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1150 struct page *page = virt_to_head_page(p);
1152 __mod_node_page_state(page_pgdat(page), idx, val);
1155 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1160 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1162 unsigned long *total_scanned)
1167 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1171 static inline void count_memcg_events(struct mem_cgroup *memcg,
1172 enum vm_event_item idx,
1173 unsigned long count)
1177 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1178 enum vm_event_item idx,
1179 unsigned long count)
1183 static inline void count_memcg_page_event(struct page *page,
1189 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1192 #endif /* CONFIG_MEMCG */
1194 /* idx can be of type enum memcg_stat_item or node_stat_item */
1195 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1198 __mod_memcg_state(memcg, idx, 1);
1201 /* idx can be of type enum memcg_stat_item or node_stat_item */
1202 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1205 __mod_memcg_state(memcg, idx, -1);
1208 /* idx can be of type enum memcg_stat_item or node_stat_item */
1209 static inline void __inc_memcg_page_state(struct page *page,
1212 __mod_memcg_page_state(page, idx, 1);
1215 /* idx can be of type enum memcg_stat_item or node_stat_item */
1216 static inline void __dec_memcg_page_state(struct page *page,
1219 __mod_memcg_page_state(page, idx, -1);
1222 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1223 enum node_stat_item idx)
1225 __mod_lruvec_state(lruvec, idx, 1);
1228 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1229 enum node_stat_item idx)
1231 __mod_lruvec_state(lruvec, idx, -1);
1234 static inline void __inc_lruvec_page_state(struct page *page,
1235 enum node_stat_item idx)
1237 __mod_lruvec_page_state(page, idx, 1);
1240 static inline void __dec_lruvec_page_state(struct page *page,
1241 enum node_stat_item idx)
1243 __mod_lruvec_page_state(page, idx, -1);
1246 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1248 __mod_lruvec_slab_state(p, idx, 1);
1251 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1253 __mod_lruvec_slab_state(p, idx, -1);
1256 /* idx can be of type enum memcg_stat_item or node_stat_item */
1257 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1260 mod_memcg_state(memcg, idx, 1);
1263 /* idx can be of type enum memcg_stat_item or node_stat_item */
1264 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1267 mod_memcg_state(memcg, idx, -1);
1270 /* idx can be of type enum memcg_stat_item or node_stat_item */
1271 static inline void inc_memcg_page_state(struct page *page,
1274 mod_memcg_page_state(page, idx, 1);
1277 /* idx can be of type enum memcg_stat_item or node_stat_item */
1278 static inline void dec_memcg_page_state(struct page *page,
1281 mod_memcg_page_state(page, idx, -1);
1284 static inline void inc_lruvec_state(struct lruvec *lruvec,
1285 enum node_stat_item idx)
1287 mod_lruvec_state(lruvec, idx, 1);
1290 static inline void dec_lruvec_state(struct lruvec *lruvec,
1291 enum node_stat_item idx)
1293 mod_lruvec_state(lruvec, idx, -1);
1296 static inline void inc_lruvec_page_state(struct page *page,
1297 enum node_stat_item idx)
1299 mod_lruvec_page_state(page, idx, 1);
1302 static inline void dec_lruvec_page_state(struct page *page,
1303 enum node_stat_item idx)
1305 mod_lruvec_page_state(page, idx, -1);
1308 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1310 struct mem_cgroup *memcg;
1312 memcg = lruvec_memcg(lruvec);
1315 memcg = parent_mem_cgroup(memcg);
1318 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1321 #ifdef CONFIG_CGROUP_WRITEBACK
1323 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1324 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1325 unsigned long *pheadroom, unsigned long *pdirty,
1326 unsigned long *pwriteback);
1328 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1329 struct bdi_writeback *wb);
1331 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1332 struct bdi_writeback *wb)
1334 if (mem_cgroup_disabled())
1337 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1338 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1341 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1343 #else /* CONFIG_CGROUP_WRITEBACK */
1345 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1350 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1351 unsigned long *pfilepages,
1352 unsigned long *pheadroom,
1353 unsigned long *pdirty,
1354 unsigned long *pwriteback)
1358 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1359 struct bdi_writeback *wb)
1363 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1367 #endif /* CONFIG_CGROUP_WRITEBACK */
1370 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1371 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1373 extern struct static_key_false memcg_sockets_enabled_key;
1374 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1375 void mem_cgroup_sk_alloc(struct sock *sk);
1376 void mem_cgroup_sk_free(struct sock *sk);
1377 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1379 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1382 if (time_before(jiffies, memcg->socket_pressure))
1384 } while ((memcg = parent_mem_cgroup(memcg)));
1388 extern int memcg_expand_shrinker_maps(int new_id);
1390 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1391 int nid, int shrinker_id);
1393 #define mem_cgroup_sockets_enabled 0
1394 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1395 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1396 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1401 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1402 int nid, int shrinker_id)
1407 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1408 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1410 #ifdef CONFIG_MEMCG_KMEM
1411 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1412 unsigned int nr_pages);
1413 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1414 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1415 void __memcg_kmem_uncharge_page(struct page *page, int order);
1417 struct obj_cgroup *get_obj_cgroup_from_current(void);
1419 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1420 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1422 extern struct static_key_false memcg_kmem_enabled_key;
1423 extern struct workqueue_struct *memcg_kmem_cache_wq;
1425 extern int memcg_nr_cache_ids;
1426 void memcg_get_cache_ids(void);
1427 void memcg_put_cache_ids(void);
1430 * Helper macro to loop through all memcg-specific caches. Callers must still
1431 * check if the cache is valid (it is either valid or NULL).
1432 * the slab_mutex must be held when looping through those caches
1434 #define for_each_memcg_cache_index(_idx) \
1435 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1437 static inline bool memcg_kmem_enabled(void)
1439 return static_branch_unlikely(&memcg_kmem_enabled_key);
1442 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1445 if (memcg_kmem_enabled())
1446 return __memcg_kmem_charge_page(page, gfp, order);
1450 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1452 if (memcg_kmem_enabled())
1453 __memcg_kmem_uncharge_page(page, order);
1456 static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1457 unsigned int nr_pages)
1459 if (memcg_kmem_enabled())
1460 return __memcg_kmem_charge(memcg, gfp, nr_pages);
1464 static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg,
1465 unsigned int nr_pages)
1467 if (memcg_kmem_enabled())
1468 __memcg_kmem_uncharge(memcg, nr_pages);
1472 * helper for accessing a memcg's index. It will be used as an index in the
1473 * child cache array in kmem_cache, and also to derive its name. This function
1474 * will return -1 when this is not a kmem-limited memcg.
1476 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1478 return memcg ? memcg->kmemcg_id : -1;
1481 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1485 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1491 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1495 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1501 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1505 #define for_each_memcg_cache_index(_idx) \
1508 static inline bool memcg_kmem_enabled(void)
1513 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1518 static inline void memcg_get_cache_ids(void)
1522 static inline void memcg_put_cache_ids(void)
1526 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1531 #endif /* CONFIG_MEMCG_KMEM */
1533 #endif /* _LINUX_MEMCONTROL_H */