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_RSS_HUGE = 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 * The memory controller data structure. The memory controller controls both
198 * page cache and RSS per cgroup. We would eventually like to provide
199 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
200 * to help the administrator determine what knobs to tune.
203 struct cgroup_subsys_state css;
205 /* Private memcg ID. Used to ID objects that outlive the cgroup */
206 struct mem_cgroup_id id;
208 /* Accounted resources */
209 struct page_counter memory;
210 struct page_counter swap;
212 /* Legacy consumer-oriented counters */
213 struct page_counter memsw;
214 struct page_counter kmem;
215 struct page_counter tcpmem;
217 /* Range enforcement for interrupt charges */
218 struct work_struct high_work;
220 unsigned long soft_limit;
222 /* vmpressure notifications */
223 struct vmpressure vmpressure;
226 * Should the accounting and control be hierarchical, per subtree?
231 * Should the OOM killer kill all belonging tasks, had it kill one?
235 /* protected by memcg_oom_lock */
240 /* OOM-Killer disable */
241 int oom_kill_disable;
243 /* memory.events and memory.events.local */
244 struct cgroup_file events_file;
245 struct cgroup_file events_local_file;
247 /* handle for "memory.swap.events" */
248 struct cgroup_file swap_events_file;
250 /* protect arrays of thresholds */
251 struct mutex thresholds_lock;
253 /* thresholds for memory usage. RCU-protected */
254 struct mem_cgroup_thresholds thresholds;
256 /* thresholds for mem+swap usage. RCU-protected */
257 struct mem_cgroup_thresholds memsw_thresholds;
259 /* For oom notifier event fd */
260 struct list_head oom_notify;
263 * Should we move charges of a task when a task is moved into this
264 * mem_cgroup ? And what type of charges should we move ?
266 unsigned long move_charge_at_immigrate;
267 /* taken only while moving_account > 0 */
268 spinlock_t move_lock;
269 unsigned long move_lock_flags;
271 MEMCG_PADDING(_pad1_);
274 * set > 0 if pages under this cgroup are moving to other cgroup.
276 atomic_t moving_account;
277 struct task_struct *move_lock_task;
279 /* Legacy local VM stats and events */
280 struct memcg_vmstats_percpu __percpu *vmstats_local;
282 /* Subtree VM stats and events (batched updates) */
283 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
285 MEMCG_PADDING(_pad2_);
287 atomic_long_t vmstats[MEMCG_NR_STAT];
288 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
291 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
292 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
294 unsigned long socket_pressure;
296 /* Legacy tcp memory accounting */
300 #ifdef CONFIG_MEMCG_KMEM
301 /* Index in the kmem_cache->memcg_params.memcg_caches array */
303 enum memcg_kmem_state kmem_state;
304 struct list_head kmem_caches;
307 #ifdef CONFIG_CGROUP_WRITEBACK
308 struct list_head cgwb_list;
309 struct wb_domain cgwb_domain;
310 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
313 /* List of events which userspace want to receive */
314 struct list_head event_list;
315 spinlock_t event_list_lock;
317 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
318 struct deferred_split deferred_split_queue;
321 struct mem_cgroup_per_node *nodeinfo[0];
322 /* WARNING: nodeinfo must be the last member here */
326 * size of first charge trial. "32" comes from vmscan.c's magic value.
327 * TODO: maybe necessary to use big numbers in big irons.
329 #define MEMCG_CHARGE_BATCH 32U
331 extern struct mem_cgroup *root_mem_cgroup;
333 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
335 return (memcg == root_mem_cgroup);
338 static inline bool mem_cgroup_disabled(void)
340 return !cgroup_subsys_enabled(memory_cgrp_subsys);
343 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
346 if (mem_cgroup_disabled())
350 return READ_ONCE(memcg->memory.emin);
352 return max(READ_ONCE(memcg->memory.emin),
353 READ_ONCE(memcg->memory.elow));
356 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
357 struct mem_cgroup *memcg);
359 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
360 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);
363 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
365 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
367 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask,
370 void mem_cgroup_uncharge(struct page *page);
371 void mem_cgroup_uncharge_list(struct list_head *page_list);
373 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
375 static struct mem_cgroup_per_node *
376 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
378 return memcg->nodeinfo[nid];
382 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
383 * @memcg: memcg of the wanted lruvec
385 * Returns the lru list vector holding pages for a given @memcg &
386 * @node combination. This can be the node lruvec, if the memory
387 * controller is disabled.
389 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
390 struct pglist_data *pgdat)
392 struct mem_cgroup_per_node *mz;
393 struct lruvec *lruvec;
395 if (mem_cgroup_disabled()) {
396 lruvec = &pgdat->__lruvec;
401 memcg = root_mem_cgroup;
403 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
404 lruvec = &mz->lruvec;
407 * Since a node can be onlined after the mem_cgroup was created,
408 * we have to be prepared to initialize lruvec->pgdat here;
409 * and if offlined then reonlined, we need to reinitialize it.
411 if (unlikely(lruvec->pgdat != pgdat))
412 lruvec->pgdat = pgdat;
416 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
418 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
420 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
422 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
425 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
426 return css ? container_of(css, struct mem_cgroup, css) : NULL;
429 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
432 css_put(&memcg->css);
435 #define mem_cgroup_from_counter(counter, member) \
436 container_of(counter, struct mem_cgroup, member)
438 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
440 struct mem_cgroup_reclaim_cookie *);
441 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
442 int mem_cgroup_scan_tasks(struct mem_cgroup *,
443 int (*)(struct task_struct *, void *), void *);
445 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
447 if (mem_cgroup_disabled())
452 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
454 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
456 return mem_cgroup_from_css(seq_css(m));
459 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
461 struct mem_cgroup_per_node *mz;
463 if (mem_cgroup_disabled())
466 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
471 * parent_mem_cgroup - find the accounting parent of a memcg
472 * @memcg: memcg whose parent to find
474 * Returns the parent memcg, or NULL if this is the root or the memory
475 * controller is in legacy no-hierarchy mode.
477 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
479 if (!memcg->memory.parent)
481 return mem_cgroup_from_counter(memcg->memory.parent, memory);
484 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
485 struct mem_cgroup *root)
489 if (!root->use_hierarchy)
491 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
494 static inline bool mm_match_cgroup(struct mm_struct *mm,
495 struct mem_cgroup *memcg)
497 struct mem_cgroup *task_memcg;
501 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
503 match = mem_cgroup_is_descendant(task_memcg, memcg);
508 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
509 ino_t page_cgroup_ino(struct page *page);
511 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
513 if (mem_cgroup_disabled())
515 return !!(memcg->css.flags & CSS_ONLINE);
519 * For memory reclaim.
521 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
523 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
524 int zid, int nr_pages);
527 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
528 enum lru_list lru, int zone_idx)
530 struct mem_cgroup_per_node *mz;
532 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
533 return mz->lru_zone_size[zone_idx][lru];
536 void mem_cgroup_handle_over_high(void);
538 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
540 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
542 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
543 struct task_struct *p);
545 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
547 static inline void mem_cgroup_enter_user_fault(void)
549 WARN_ON(current->in_user_fault);
550 current->in_user_fault = 1;
553 static inline void mem_cgroup_exit_user_fault(void)
555 WARN_ON(!current->in_user_fault);
556 current->in_user_fault = 0;
559 static inline bool task_in_memcg_oom(struct task_struct *p)
561 return p->memcg_in_oom;
564 bool mem_cgroup_oom_synchronize(bool wait);
565 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
566 struct mem_cgroup *oom_domain);
567 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
569 #ifdef CONFIG_MEMCG_SWAP
570 extern int do_swap_account;
573 struct mem_cgroup *lock_page_memcg(struct page *page);
574 void __unlock_page_memcg(struct mem_cgroup *memcg);
575 void unlock_page_memcg(struct page *page);
578 * idx can be of type enum memcg_stat_item or node_stat_item.
579 * Keep in sync with memcg_exact_page_state().
581 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
583 long x = atomic_long_read(&memcg->vmstats[idx]);
592 * idx can be of type enum memcg_stat_item or node_stat_item.
593 * Keep in sync with memcg_exact_page_state().
595 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
601 for_each_possible_cpu(cpu)
602 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
610 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
612 /* idx can be of type enum memcg_stat_item or node_stat_item */
613 static inline void mod_memcg_state(struct mem_cgroup *memcg,
618 local_irq_save(flags);
619 __mod_memcg_state(memcg, idx, val);
620 local_irq_restore(flags);
624 * mod_memcg_page_state - update page state statistics
626 * @idx: page state item to account
627 * @val: number of pages (positive or negative)
629 * The @page must be locked or the caller must use lock_page_memcg()
630 * to prevent double accounting when the page is concurrently being
631 * moved to another memcg:
633 * lock_page(page) or lock_page_memcg(page)
634 * if (TestClearPageState(page))
635 * mod_memcg_page_state(page, state, -1);
636 * unlock_page(page) or unlock_page_memcg(page)
638 * Kernel pages are an exception to this, since they'll never move.
640 static inline void __mod_memcg_page_state(struct page *page,
643 if (page->mem_cgroup)
644 __mod_memcg_state(page->mem_cgroup, idx, val);
647 static inline void mod_memcg_page_state(struct page *page,
650 if (page->mem_cgroup)
651 mod_memcg_state(page->mem_cgroup, idx, val);
654 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
655 enum node_stat_item idx)
657 struct mem_cgroup_per_node *pn;
660 if (mem_cgroup_disabled())
661 return node_page_state(lruvec_pgdat(lruvec), idx);
663 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
664 x = atomic_long_read(&pn->lruvec_stat[idx]);
672 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
673 enum node_stat_item idx)
675 struct mem_cgroup_per_node *pn;
679 if (mem_cgroup_disabled())
680 return node_page_state(lruvec_pgdat(lruvec), idx);
682 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
683 for_each_possible_cpu(cpu)
684 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
692 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
694 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
695 void mod_memcg_obj_state(void *p, int idx, int val);
697 static inline void mod_lruvec_state(struct lruvec *lruvec,
698 enum node_stat_item idx, int val)
702 local_irq_save(flags);
703 __mod_lruvec_state(lruvec, idx, val);
704 local_irq_restore(flags);
707 static inline void __mod_lruvec_page_state(struct page *page,
708 enum node_stat_item idx, int val)
710 struct page *head = compound_head(page); /* rmap on tail pages */
711 pg_data_t *pgdat = page_pgdat(page);
712 struct lruvec *lruvec;
714 /* Untracked pages have no memcg, no lruvec. Update only the node */
715 if (!head->mem_cgroup) {
716 __mod_node_page_state(pgdat, idx, val);
720 lruvec = mem_cgroup_lruvec(head->mem_cgroup, pgdat);
721 __mod_lruvec_state(lruvec, idx, val);
724 static inline void mod_lruvec_page_state(struct page *page,
725 enum node_stat_item idx, int val)
729 local_irq_save(flags);
730 __mod_lruvec_page_state(page, idx, val);
731 local_irq_restore(flags);
734 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
736 unsigned long *total_scanned);
738 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
739 unsigned long count);
741 static inline void count_memcg_events(struct mem_cgroup *memcg,
742 enum vm_event_item idx,
747 local_irq_save(flags);
748 __count_memcg_events(memcg, idx, count);
749 local_irq_restore(flags);
752 static inline void count_memcg_page_event(struct page *page,
753 enum vm_event_item idx)
755 if (page->mem_cgroup)
756 count_memcg_events(page->mem_cgroup, idx, 1);
759 static inline void count_memcg_event_mm(struct mm_struct *mm,
760 enum vm_event_item idx)
762 struct mem_cgroup *memcg;
764 if (mem_cgroup_disabled())
768 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
770 count_memcg_events(memcg, idx, 1);
774 static inline void memcg_memory_event(struct mem_cgroup *memcg,
775 enum memcg_memory_event event)
777 atomic_long_inc(&memcg->memory_events_local[event]);
778 cgroup_file_notify(&memcg->events_local_file);
781 atomic_long_inc(&memcg->memory_events[event]);
782 cgroup_file_notify(&memcg->events_file);
784 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
786 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
788 } while ((memcg = parent_mem_cgroup(memcg)) &&
789 !mem_cgroup_is_root(memcg));
792 static inline void memcg_memory_event_mm(struct mm_struct *mm,
793 enum memcg_memory_event event)
795 struct mem_cgroup *memcg;
797 if (mem_cgroup_disabled())
801 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
803 memcg_memory_event(memcg, event);
807 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
808 void mem_cgroup_split_huge_fixup(struct page *head);
811 #else /* CONFIG_MEMCG */
813 #define MEM_CGROUP_ID_SHIFT 0
814 #define MEM_CGROUP_ID_MAX 0
818 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
823 static inline bool mem_cgroup_disabled(void)
828 static inline void memcg_memory_event(struct mem_cgroup *memcg,
829 enum memcg_memory_event event)
833 static inline void memcg_memory_event_mm(struct mm_struct *mm,
834 enum memcg_memory_event event)
838 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *memcg,
844 static inline enum mem_cgroup_protection mem_cgroup_protected(
845 struct mem_cgroup *root, struct mem_cgroup *memcg)
847 return MEMCG_PROT_NONE;
850 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
852 struct mem_cgroup **memcgp)
858 static inline int mem_cgroup_try_charge_delay(struct page *page,
859 struct mm_struct *mm,
861 struct mem_cgroup **memcgp)
867 static inline void mem_cgroup_commit_charge(struct page *page,
868 struct mem_cgroup *memcg,
873 static inline void mem_cgroup_cancel_charge(struct page *page,
874 struct mem_cgroup *memcg)
878 static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
879 gfp_t gfp_mask, bool lrucare)
884 static inline void mem_cgroup_uncharge(struct page *page)
888 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
892 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
896 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
897 struct pglist_data *pgdat)
899 return &pgdat->__lruvec;
902 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
903 struct pglist_data *pgdat)
905 return &pgdat->__lruvec;
908 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
913 static inline bool mm_match_cgroup(struct mm_struct *mm,
914 struct mem_cgroup *memcg)
919 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
924 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
929 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
933 static inline struct mem_cgroup *
934 mem_cgroup_iter(struct mem_cgroup *root,
935 struct mem_cgroup *prev,
936 struct mem_cgroup_reclaim_cookie *reclaim)
941 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
942 struct mem_cgroup *prev)
946 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
947 int (*fn)(struct task_struct *, void *), void *arg)
952 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
957 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
960 /* XXX: This should always return root_mem_cgroup */
964 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
969 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
974 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
980 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
981 enum lru_list lru, int zone_idx)
986 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
991 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
997 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1002 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1006 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1011 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1015 static inline void unlock_page_memcg(struct page *page)
1019 static inline void mem_cgroup_handle_over_high(void)
1023 static inline void mem_cgroup_enter_user_fault(void)
1027 static inline void mem_cgroup_exit_user_fault(void)
1031 static inline bool task_in_memcg_oom(struct task_struct *p)
1036 static inline bool mem_cgroup_oom_synchronize(bool wait)
1041 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1042 struct task_struct *victim, struct mem_cgroup *oom_domain)
1047 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1051 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1056 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1062 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1068 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1074 static inline void __mod_memcg_page_state(struct page *page,
1080 static inline void mod_memcg_page_state(struct page *page,
1086 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1087 enum node_stat_item idx)
1089 return node_page_state(lruvec_pgdat(lruvec), idx);
1092 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1093 enum node_stat_item idx)
1095 return node_page_state(lruvec_pgdat(lruvec), idx);
1098 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1099 enum node_stat_item idx, int val)
1101 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1104 static inline void mod_lruvec_state(struct lruvec *lruvec,
1105 enum node_stat_item idx, int val)
1107 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1110 static inline void __mod_lruvec_page_state(struct page *page,
1111 enum node_stat_item idx, int val)
1113 __mod_node_page_state(page_pgdat(page), idx, val);
1116 static inline void mod_lruvec_page_state(struct page *page,
1117 enum node_stat_item idx, int val)
1119 mod_node_page_state(page_pgdat(page), idx, val);
1122 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1125 struct page *page = virt_to_head_page(p);
1127 __mod_node_page_state(page_pgdat(page), idx, val);
1130 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1135 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1137 unsigned long *total_scanned)
1142 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1146 static inline void count_memcg_events(struct mem_cgroup *memcg,
1147 enum vm_event_item idx,
1148 unsigned long count)
1152 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1153 enum vm_event_item idx,
1154 unsigned long count)
1158 static inline void count_memcg_page_event(struct page *page,
1164 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1167 #endif /* CONFIG_MEMCG */
1169 /* idx can be of type enum memcg_stat_item or node_stat_item */
1170 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1173 __mod_memcg_state(memcg, idx, 1);
1176 /* idx can be of type enum memcg_stat_item or node_stat_item */
1177 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1180 __mod_memcg_state(memcg, idx, -1);
1183 /* idx can be of type enum memcg_stat_item or node_stat_item */
1184 static inline void __inc_memcg_page_state(struct page *page,
1187 __mod_memcg_page_state(page, idx, 1);
1190 /* idx can be of type enum memcg_stat_item or node_stat_item */
1191 static inline void __dec_memcg_page_state(struct page *page,
1194 __mod_memcg_page_state(page, idx, -1);
1197 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1198 enum node_stat_item idx)
1200 __mod_lruvec_state(lruvec, idx, 1);
1203 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1204 enum node_stat_item idx)
1206 __mod_lruvec_state(lruvec, idx, -1);
1209 static inline void __inc_lruvec_page_state(struct page *page,
1210 enum node_stat_item idx)
1212 __mod_lruvec_page_state(page, idx, 1);
1215 static inline void __dec_lruvec_page_state(struct page *page,
1216 enum node_stat_item idx)
1218 __mod_lruvec_page_state(page, idx, -1);
1221 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1223 __mod_lruvec_slab_state(p, idx, 1);
1226 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1228 __mod_lruvec_slab_state(p, idx, -1);
1231 /* idx can be of type enum memcg_stat_item or node_stat_item */
1232 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1235 mod_memcg_state(memcg, idx, 1);
1238 /* idx can be of type enum memcg_stat_item or node_stat_item */
1239 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1242 mod_memcg_state(memcg, idx, -1);
1245 /* idx can be of type enum memcg_stat_item or node_stat_item */
1246 static inline void inc_memcg_page_state(struct page *page,
1249 mod_memcg_page_state(page, idx, 1);
1252 /* idx can be of type enum memcg_stat_item or node_stat_item */
1253 static inline void dec_memcg_page_state(struct page *page,
1256 mod_memcg_page_state(page, idx, -1);
1259 static inline void inc_lruvec_state(struct lruvec *lruvec,
1260 enum node_stat_item idx)
1262 mod_lruvec_state(lruvec, idx, 1);
1265 static inline void dec_lruvec_state(struct lruvec *lruvec,
1266 enum node_stat_item idx)
1268 mod_lruvec_state(lruvec, idx, -1);
1271 static inline void inc_lruvec_page_state(struct page *page,
1272 enum node_stat_item idx)
1274 mod_lruvec_page_state(page, idx, 1);
1277 static inline void dec_lruvec_page_state(struct page *page,
1278 enum node_stat_item idx)
1280 mod_lruvec_page_state(page, idx, -1);
1283 #ifdef CONFIG_CGROUP_WRITEBACK
1285 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1286 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1287 unsigned long *pheadroom, unsigned long *pdirty,
1288 unsigned long *pwriteback);
1290 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1291 struct bdi_writeback *wb);
1293 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1294 struct bdi_writeback *wb)
1296 if (mem_cgroup_disabled())
1299 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1300 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1303 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1305 #else /* CONFIG_CGROUP_WRITEBACK */
1307 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1312 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1313 unsigned long *pfilepages,
1314 unsigned long *pheadroom,
1315 unsigned long *pdirty,
1316 unsigned long *pwriteback)
1320 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1321 struct bdi_writeback *wb)
1325 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1329 #endif /* CONFIG_CGROUP_WRITEBACK */
1332 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1333 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1335 extern struct static_key_false memcg_sockets_enabled_key;
1336 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1337 void mem_cgroup_sk_alloc(struct sock *sk);
1338 void mem_cgroup_sk_free(struct sock *sk);
1339 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1341 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1344 if (time_before(jiffies, memcg->socket_pressure))
1346 } while ((memcg = parent_mem_cgroup(memcg)));
1350 extern int memcg_expand_shrinker_maps(int new_id);
1352 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1353 int nid, int shrinker_id);
1355 #define mem_cgroup_sockets_enabled 0
1356 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1357 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1358 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1363 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1364 int nid, int shrinker_id)
1369 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1370 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1372 #ifdef CONFIG_MEMCG_KMEM
1373 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1374 unsigned int nr_pages);
1375 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1376 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1377 void __memcg_kmem_uncharge_page(struct page *page, int order);
1379 extern struct static_key_false memcg_kmem_enabled_key;
1380 extern struct workqueue_struct *memcg_kmem_cache_wq;
1382 extern int memcg_nr_cache_ids;
1383 void memcg_get_cache_ids(void);
1384 void memcg_put_cache_ids(void);
1387 * Helper macro to loop through all memcg-specific caches. Callers must still
1388 * check if the cache is valid (it is either valid or NULL).
1389 * the slab_mutex must be held when looping through those caches
1391 #define for_each_memcg_cache_index(_idx) \
1392 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1394 static inline bool memcg_kmem_enabled(void)
1396 return static_branch_unlikely(&memcg_kmem_enabled_key);
1399 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1402 if (memcg_kmem_enabled())
1403 return __memcg_kmem_charge_page(page, gfp, order);
1407 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1409 if (memcg_kmem_enabled())
1410 __memcg_kmem_uncharge_page(page, order);
1413 static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1414 unsigned int nr_pages)
1416 if (memcg_kmem_enabled())
1417 return __memcg_kmem_charge(memcg, gfp, nr_pages);
1421 static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg,
1422 unsigned int nr_pages)
1424 if (memcg_kmem_enabled())
1425 __memcg_kmem_uncharge(memcg, nr_pages);
1429 * helper for accessing a memcg's index. It will be used as an index in the
1430 * child cache array in kmem_cache, and also to derive its name. This function
1431 * will return -1 when this is not a kmem-limited memcg.
1433 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1435 return memcg ? memcg->kmemcg_id : -1;
1438 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1442 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1448 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1452 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1458 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1462 #define for_each_memcg_cache_index(_idx) \
1465 static inline bool memcg_kmem_enabled(void)
1470 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1475 static inline void memcg_get_cache_ids(void)
1479 static inline void memcg_put_cache_ids(void)
1483 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1488 #endif /* CONFIG_MEMCG_KMEM */
1490 #endif /* _LINUX_MEMCONTROL_H */