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,
39 enum memcg_memory_event {
48 MEMCG_NR_MEMORY_EVENTS,
51 struct mem_cgroup_reclaim_cookie {
53 unsigned int generation;
58 #define MEM_CGROUP_ID_SHIFT 16
59 #define MEM_CGROUP_ID_MAX USHRT_MAX
61 struct mem_cgroup_id {
67 * Per memcg event counter is incremented at every pagein/pageout. With THP,
68 * it will be incremented by the number of pages. This counter is used
69 * to trigger some periodic events. This is straightforward and better
70 * than using jiffies etc. to handle periodic memcg event.
72 enum mem_cgroup_events_target {
73 MEM_CGROUP_TARGET_THRESH,
74 MEM_CGROUP_TARGET_SOFTLIMIT,
78 struct memcg_vmstats_percpu {
79 long stat[MEMCG_NR_STAT];
80 unsigned long events[NR_VM_EVENT_ITEMS];
81 unsigned long nr_page_events;
82 unsigned long targets[MEM_CGROUP_NTARGETS];
85 struct mem_cgroup_reclaim_iter {
86 struct mem_cgroup *position;
87 /* scan generation, increased every round-trip */
88 unsigned int generation;
92 long count[NR_VM_NODE_STAT_ITEMS];
96 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
97 * which have elements charged to this memcg.
99 struct memcg_shrinker_map {
105 * per-node information in memory controller.
107 struct mem_cgroup_per_node {
108 struct lruvec lruvec;
110 /* Legacy local VM stats */
111 struct lruvec_stat __percpu *lruvec_stat_local;
113 /* Subtree VM stats (batched updates) */
114 struct lruvec_stat __percpu *lruvec_stat_cpu;
115 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
117 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
119 struct mem_cgroup_reclaim_iter iter;
121 struct memcg_shrinker_map __rcu *shrinker_map;
123 struct rb_node tree_node; /* RB tree node */
124 unsigned long usage_in_excess;/* Set to the value by which */
125 /* the soft limit is exceeded*/
127 struct mem_cgroup *memcg; /* Back pointer, we cannot */
128 /* use container_of */
131 struct mem_cgroup_threshold {
132 struct eventfd_ctx *eventfd;
133 unsigned long threshold;
137 struct mem_cgroup_threshold_ary {
138 /* An array index points to threshold just below or equal to usage. */
139 int current_threshold;
140 /* Size of entries[] */
142 /* Array of thresholds */
143 struct mem_cgroup_threshold entries[];
146 struct mem_cgroup_thresholds {
147 /* Primary thresholds array */
148 struct mem_cgroup_threshold_ary *primary;
150 * Spare threshold array.
151 * This is needed to make mem_cgroup_unregister_event() "never fail".
152 * It must be able to store at least primary->size - 1 entries.
154 struct mem_cgroup_threshold_ary *spare;
157 enum memcg_kmem_state {
163 #if defined(CONFIG_SMP)
164 struct memcg_padding {
166 } ____cacheline_internodealigned_in_smp;
167 #define MEMCG_PADDING(name) struct memcg_padding name;
169 #define MEMCG_PADDING(name)
173 * Remember four most recent foreign writebacks with dirty pages in this
174 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
175 * one in a given round, we're likely to catch it later if it keeps
176 * foreign-dirtying, so a fairly low count should be enough.
178 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
180 #define MEMCG_CGWB_FRN_CNT 4
182 struct memcg_cgwb_frn {
183 u64 bdi_id; /* bdi->id of the foreign inode */
184 int memcg_id; /* memcg->css.id of foreign inode */
185 u64 at; /* jiffies_64 at the time of dirtying */
186 struct wb_completion done; /* tracks in-flight foreign writebacks */
190 * Bucket for arbitrarily byte-sized objects charged to a memory
191 * cgroup. The bucket can be reparented in one piece when the cgroup
192 * is destroyed, without having to round up the individual references
193 * of all live memory objects in the wild.
196 struct percpu_ref refcnt;
197 struct mem_cgroup *memcg;
198 atomic_t nr_charged_bytes;
200 struct list_head list; /* protected by objcg_lock */
206 * The memory controller data structure. The memory controller controls both
207 * page cache and RSS per cgroup. We would eventually like to provide
208 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
209 * to help the administrator determine what knobs to tune.
212 struct cgroup_subsys_state css;
214 /* Private memcg ID. Used to ID objects that outlive the cgroup */
215 struct mem_cgroup_id id;
217 /* Accounted resources */
218 struct page_counter memory; /* Both v1 & v2 */
221 struct page_counter swap; /* v2 only */
222 struct page_counter memsw; /* v1 only */
225 /* Legacy consumer-oriented counters */
226 struct page_counter kmem; /* v1 only */
227 struct page_counter tcpmem; /* v1 only */
229 /* Range enforcement for interrupt charges */
230 struct work_struct high_work;
232 unsigned long soft_limit;
234 /* vmpressure notifications */
235 struct vmpressure vmpressure;
238 * Should the accounting and control be hierarchical, per subtree?
243 * Should the OOM killer kill all belonging tasks, had it kill one?
247 /* protected by memcg_oom_lock */
252 /* OOM-Killer disable */
253 int oom_kill_disable;
255 /* memory.events and memory.events.local */
256 struct cgroup_file events_file;
257 struct cgroup_file events_local_file;
259 /* handle for "memory.swap.events" */
260 struct cgroup_file swap_events_file;
262 /* protect arrays of thresholds */
263 struct mutex thresholds_lock;
265 /* thresholds for memory usage. RCU-protected */
266 struct mem_cgroup_thresholds thresholds;
268 /* thresholds for mem+swap usage. RCU-protected */
269 struct mem_cgroup_thresholds memsw_thresholds;
271 /* For oom notifier event fd */
272 struct list_head oom_notify;
275 * Should we move charges of a task when a task is moved into this
276 * mem_cgroup ? And what type of charges should we move ?
278 unsigned long move_charge_at_immigrate;
279 /* taken only while moving_account > 0 */
280 spinlock_t move_lock;
281 unsigned long move_lock_flags;
283 MEMCG_PADDING(_pad1_);
285 atomic_long_t vmstats[MEMCG_NR_STAT];
286 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
289 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
290 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
292 unsigned long socket_pressure;
294 /* Legacy tcp memory accounting */
298 #ifdef CONFIG_MEMCG_KMEM
299 /* Index in the kmem_cache->memcg_params.memcg_caches array */
301 enum memcg_kmem_state kmem_state;
302 struct obj_cgroup __rcu *objcg;
303 /* list of inherited objcgs, protected by objcg_lock */
304 struct list_head objcg_list;
307 MEMCG_PADDING(_pad2_);
310 * set > 0 if pages under this cgroup are moving to other cgroup.
312 atomic_t moving_account;
313 struct task_struct *move_lock_task;
315 /* Legacy local VM stats and events */
316 struct memcg_vmstats_percpu __percpu *vmstats_local;
318 /* Subtree VM stats and events (batched updates) */
319 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
321 #ifdef CONFIG_CGROUP_WRITEBACK
322 struct list_head cgwb_list;
323 struct wb_domain cgwb_domain;
324 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
327 /* List of events which userspace want to receive */
328 struct list_head event_list;
329 spinlock_t event_list_lock;
331 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
332 struct deferred_split deferred_split_queue;
335 struct mem_cgroup_per_node *nodeinfo[0];
336 /* WARNING: nodeinfo must be the last member here */
340 * size of first charge trial. "32" comes from vmscan.c's magic value.
341 * TODO: maybe necessary to use big numbers in big irons.
343 #define MEMCG_CHARGE_BATCH 32U
345 extern struct mem_cgroup *root_mem_cgroup;
347 static __always_inline bool memcg_stat_item_in_bytes(int idx)
349 if (idx == MEMCG_PERCPU_B)
351 return vmstat_item_in_bytes(idx);
354 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
356 return (memcg == root_mem_cgroup);
359 static inline bool mem_cgroup_disabled(void)
361 return !cgroup_subsys_enabled(memory_cgrp_subsys);
364 static inline void mem_cgroup_protection(struct mem_cgroup *root,
365 struct mem_cgroup *memcg,
371 if (mem_cgroup_disabled())
375 * There is no reclaim protection applied to a targeted reclaim.
376 * We are special casing this specific case here because
377 * mem_cgroup_protected calculation is not robust enough to keep
378 * the protection invariant for calculated effective values for
379 * parallel reclaimers with different reclaim target. This is
380 * especially a problem for tail memcgs (as they have pages on LRU)
381 * which would want to have effective values 0 for targeted reclaim
382 * but a different value for external reclaim.
385 * Let's have global and A's reclaim in parallel:
387 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
389 * | C (low = 1G, usage = 2.5G)
390 * B (low = 1G, usage = 0.5G)
392 * For the global reclaim
394 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
395 * C.elow = min(C.usage, C.low)
397 * With the effective values resetting we have A reclaim
402 * If the global reclaim races with A's reclaim then
403 * B.elow = C.elow = 0 because children_low_usage > A.elow)
404 * is possible and reclaiming B would be violating the protection.
410 *min = READ_ONCE(memcg->memory.emin);
411 *low = READ_ONCE(memcg->memory.elow);
414 void mem_cgroup_calculate_protection(struct mem_cgroup *root,
415 struct mem_cgroup *memcg);
417 static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg)
420 * The root memcg doesn't account charges, and doesn't support
423 return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg);
427 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
429 if (!mem_cgroup_supports_protection(memcg))
432 return READ_ONCE(memcg->memory.elow) >=
433 page_counter_read(&memcg->memory);
436 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
438 if (!mem_cgroup_supports_protection(memcg))
441 return READ_ONCE(memcg->memory.emin) >=
442 page_counter_read(&memcg->memory);
445 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask);
447 void mem_cgroup_uncharge(struct page *page);
448 void mem_cgroup_uncharge_list(struct list_head *page_list);
450 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
452 static struct mem_cgroup_per_node *
453 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
455 return memcg->nodeinfo[nid];
459 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
460 * @memcg: memcg of the wanted lruvec
462 * Returns the lru list vector holding pages for a given @memcg &
463 * @node combination. This can be the node lruvec, if the memory
464 * controller is disabled.
466 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
467 struct pglist_data *pgdat)
469 struct mem_cgroup_per_node *mz;
470 struct lruvec *lruvec;
472 if (mem_cgroup_disabled()) {
473 lruvec = &pgdat->__lruvec;
478 memcg = root_mem_cgroup;
480 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
481 lruvec = &mz->lruvec;
484 * Since a node can be onlined after the mem_cgroup was created,
485 * we have to be prepared to initialize lruvec->pgdat here;
486 * and if offlined then reonlined, we need to reinitialize it.
488 if (unlikely(lruvec->pgdat != pgdat))
489 lruvec->pgdat = pgdat;
493 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
495 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
497 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
499 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
502 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
503 return css ? container_of(css, struct mem_cgroup, css) : NULL;
506 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
508 return percpu_ref_tryget(&objcg->refcnt);
511 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
513 percpu_ref_get(&objcg->refcnt);
516 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
518 percpu_ref_put(&objcg->refcnt);
522 * After the initialization objcg->memcg is always pointing at
523 * a valid memcg, but can be atomically swapped to the parent memcg.
525 * The caller must ensure that the returned memcg won't be released:
526 * e.g. acquire the rcu_read_lock or css_set_lock.
528 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
530 return READ_ONCE(objcg->memcg);
533 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
536 css_put(&memcg->css);
539 #define mem_cgroup_from_counter(counter, member) \
540 container_of(counter, struct mem_cgroup, member)
542 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
544 struct mem_cgroup_reclaim_cookie *);
545 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
546 int mem_cgroup_scan_tasks(struct mem_cgroup *,
547 int (*)(struct task_struct *, void *), void *);
549 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
551 if (mem_cgroup_disabled())
556 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
558 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
560 return mem_cgroup_from_css(seq_css(m));
563 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
565 struct mem_cgroup_per_node *mz;
567 if (mem_cgroup_disabled())
570 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
575 * parent_mem_cgroup - find the accounting parent of a memcg
576 * @memcg: memcg whose parent to find
578 * Returns the parent memcg, or NULL if this is the root or the memory
579 * controller is in legacy no-hierarchy mode.
581 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
583 if (!memcg->memory.parent)
585 return mem_cgroup_from_counter(memcg->memory.parent, memory);
588 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
589 struct mem_cgroup *root)
593 if (!root->use_hierarchy)
595 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
598 static inline bool mm_match_cgroup(struct mm_struct *mm,
599 struct mem_cgroup *memcg)
601 struct mem_cgroup *task_memcg;
605 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
607 match = mem_cgroup_is_descendant(task_memcg, memcg);
612 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
613 ino_t page_cgroup_ino(struct page *page);
615 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
617 if (mem_cgroup_disabled())
619 return !!(memcg->css.flags & CSS_ONLINE);
623 * For memory reclaim.
625 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
627 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
628 int zid, int nr_pages);
631 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
632 enum lru_list lru, int zone_idx)
634 struct mem_cgroup_per_node *mz;
636 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
637 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
640 void mem_cgroup_handle_over_high(void);
642 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
644 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
646 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
647 struct task_struct *p);
649 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
651 static inline void mem_cgroup_enter_user_fault(void)
653 WARN_ON(current->in_user_fault);
654 current->in_user_fault = 1;
657 static inline void mem_cgroup_exit_user_fault(void)
659 WARN_ON(!current->in_user_fault);
660 current->in_user_fault = 0;
663 static inline bool task_in_memcg_oom(struct task_struct *p)
665 return p->memcg_in_oom;
668 bool mem_cgroup_oom_synchronize(bool wait);
669 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
670 struct mem_cgroup *oom_domain);
671 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
673 #ifdef CONFIG_MEMCG_SWAP
674 extern bool cgroup_memory_noswap;
677 struct mem_cgroup *lock_page_memcg(struct page *page);
678 void __unlock_page_memcg(struct mem_cgroup *memcg);
679 void unlock_page_memcg(struct page *page);
682 * idx can be of type enum memcg_stat_item or node_stat_item.
683 * Keep in sync with memcg_exact_page_state().
685 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
687 long x = atomic_long_read(&memcg->vmstats[idx]);
696 * idx can be of type enum memcg_stat_item or node_stat_item.
697 * Keep in sync with memcg_exact_page_state().
699 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
705 for_each_possible_cpu(cpu)
706 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
714 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
716 /* idx can be of type enum memcg_stat_item or node_stat_item */
717 static inline void mod_memcg_state(struct mem_cgroup *memcg,
722 local_irq_save(flags);
723 __mod_memcg_state(memcg, idx, val);
724 local_irq_restore(flags);
728 * mod_memcg_page_state - update page state statistics
730 * @idx: page state item to account
731 * @val: number of pages (positive or negative)
733 * The @page must be locked or the caller must use lock_page_memcg()
734 * to prevent double accounting when the page is concurrently being
735 * moved to another memcg:
737 * lock_page(page) or lock_page_memcg(page)
738 * if (TestClearPageState(page))
739 * mod_memcg_page_state(page, state, -1);
740 * unlock_page(page) or unlock_page_memcg(page)
742 * Kernel pages are an exception to this, since they'll never move.
744 static inline void __mod_memcg_page_state(struct page *page,
747 if (page->mem_cgroup)
748 __mod_memcg_state(page->mem_cgroup, idx, val);
751 static inline void mod_memcg_page_state(struct page *page,
754 if (page->mem_cgroup)
755 mod_memcg_state(page->mem_cgroup, idx, val);
758 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
759 enum node_stat_item idx)
761 struct mem_cgroup_per_node *pn;
764 if (mem_cgroup_disabled())
765 return node_page_state(lruvec_pgdat(lruvec), idx);
767 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
768 x = atomic_long_read(&pn->lruvec_stat[idx]);
776 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
777 enum node_stat_item idx)
779 struct mem_cgroup_per_node *pn;
783 if (mem_cgroup_disabled())
784 return node_page_state(lruvec_pgdat(lruvec), idx);
786 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
787 for_each_possible_cpu(cpu)
788 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
796 void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
798 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
800 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
802 void mod_memcg_obj_state(void *p, int idx, int val);
804 static inline void mod_lruvec_slab_state(void *p, enum node_stat_item idx,
809 local_irq_save(flags);
810 __mod_lruvec_slab_state(p, idx, val);
811 local_irq_restore(flags);
814 static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
815 enum node_stat_item idx, int val)
819 local_irq_save(flags);
820 __mod_memcg_lruvec_state(lruvec, idx, val);
821 local_irq_restore(flags);
824 static inline void mod_lruvec_state(struct lruvec *lruvec,
825 enum node_stat_item idx, int val)
829 local_irq_save(flags);
830 __mod_lruvec_state(lruvec, idx, val);
831 local_irq_restore(flags);
834 static inline void __mod_lruvec_page_state(struct page *page,
835 enum node_stat_item idx, int val)
837 struct page *head = compound_head(page); /* rmap on tail pages */
838 pg_data_t *pgdat = page_pgdat(page);
839 struct lruvec *lruvec;
841 /* Untracked pages have no memcg, no lruvec. Update only the node */
842 if (!head->mem_cgroup) {
843 __mod_node_page_state(pgdat, idx, val);
847 lruvec = mem_cgroup_lruvec(head->mem_cgroup, pgdat);
848 __mod_lruvec_state(lruvec, idx, val);
851 static inline void mod_lruvec_page_state(struct page *page,
852 enum node_stat_item idx, int val)
856 local_irq_save(flags);
857 __mod_lruvec_page_state(page, idx, val);
858 local_irq_restore(flags);
861 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
863 unsigned long *total_scanned);
865 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
866 unsigned long count);
868 static inline void count_memcg_events(struct mem_cgroup *memcg,
869 enum vm_event_item idx,
874 local_irq_save(flags);
875 __count_memcg_events(memcg, idx, count);
876 local_irq_restore(flags);
879 static inline void count_memcg_page_event(struct page *page,
880 enum vm_event_item idx)
882 if (page->mem_cgroup)
883 count_memcg_events(page->mem_cgroup, idx, 1);
886 static inline void count_memcg_event_mm(struct mm_struct *mm,
887 enum vm_event_item idx)
889 struct mem_cgroup *memcg;
891 if (mem_cgroup_disabled())
895 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
897 count_memcg_events(memcg, idx, 1);
901 static inline void memcg_memory_event(struct mem_cgroup *memcg,
902 enum memcg_memory_event event)
904 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
905 event == MEMCG_SWAP_FAIL;
907 atomic_long_inc(&memcg->memory_events_local[event]);
909 cgroup_file_notify(&memcg->events_local_file);
912 atomic_long_inc(&memcg->memory_events[event]);
914 cgroup_file_notify(&memcg->swap_events_file);
916 cgroup_file_notify(&memcg->events_file);
918 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
920 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
922 } while ((memcg = parent_mem_cgroup(memcg)) &&
923 !mem_cgroup_is_root(memcg));
926 static inline void memcg_memory_event_mm(struct mm_struct *mm,
927 enum memcg_memory_event event)
929 struct mem_cgroup *memcg;
931 if (mem_cgroup_disabled())
935 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
937 memcg_memory_event(memcg, event);
941 void split_page_memcg(struct page *head, unsigned int nr);
943 #else /* CONFIG_MEMCG */
945 #define MEM_CGROUP_ID_SHIFT 0
946 #define MEM_CGROUP_ID_MAX 0
950 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
955 static inline bool mem_cgroup_disabled(void)
960 static inline void memcg_memory_event(struct mem_cgroup *memcg,
961 enum memcg_memory_event event)
965 static inline void memcg_memory_event_mm(struct mm_struct *mm,
966 enum memcg_memory_event event)
970 static inline void mem_cgroup_protection(struct mem_cgroup *root,
971 struct mem_cgroup *memcg,
978 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
979 struct mem_cgroup *memcg)
983 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
988 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
993 static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
999 static inline void mem_cgroup_uncharge(struct page *page)
1003 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
1007 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
1011 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1012 struct pglist_data *pgdat)
1014 return &pgdat->__lruvec;
1017 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
1018 struct pglist_data *pgdat)
1020 return &pgdat->__lruvec;
1023 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1028 static inline bool mm_match_cgroup(struct mm_struct *mm,
1029 struct mem_cgroup *memcg)
1034 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1039 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
1044 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1048 static inline struct mem_cgroup *
1049 mem_cgroup_iter(struct mem_cgroup *root,
1050 struct mem_cgroup *prev,
1051 struct mem_cgroup_reclaim_cookie *reclaim)
1056 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1057 struct mem_cgroup *prev)
1061 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1062 int (*fn)(struct task_struct *, void *), void *arg)
1067 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1072 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1075 /* XXX: This should always return root_mem_cgroup */
1079 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1084 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1089 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1095 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1096 enum lru_list lru, int zone_idx)
1101 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1106 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1112 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1117 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1121 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1126 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1130 static inline void unlock_page_memcg(struct page *page)
1134 static inline void mem_cgroup_handle_over_high(void)
1138 static inline void mem_cgroup_enter_user_fault(void)
1142 static inline void mem_cgroup_exit_user_fault(void)
1146 static inline bool task_in_memcg_oom(struct task_struct *p)
1151 static inline bool mem_cgroup_oom_synchronize(bool wait)
1156 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1157 struct task_struct *victim, struct mem_cgroup *oom_domain)
1162 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1166 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1171 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1177 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1183 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1189 static inline void __mod_memcg_page_state(struct page *page,
1195 static inline void mod_memcg_page_state(struct page *page,
1201 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1202 enum node_stat_item idx)
1204 return node_page_state(lruvec_pgdat(lruvec), idx);
1207 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1208 enum node_stat_item idx)
1210 return node_page_state(lruvec_pgdat(lruvec), idx);
1213 static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1214 enum node_stat_item idx, int val)
1218 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1219 enum node_stat_item idx, int val)
1221 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1224 static inline void mod_lruvec_state(struct lruvec *lruvec,
1225 enum node_stat_item idx, int val)
1227 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1230 static inline void __mod_lruvec_page_state(struct page *page,
1231 enum node_stat_item idx, int val)
1233 __mod_node_page_state(page_pgdat(page), idx, val);
1236 static inline void mod_lruvec_page_state(struct page *page,
1237 enum node_stat_item idx, int val)
1239 mod_node_page_state(page_pgdat(page), idx, val);
1242 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1245 struct page *page = virt_to_head_page(p);
1247 __mod_node_page_state(page_pgdat(page), idx, val);
1250 static inline void mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1253 struct page *page = virt_to_head_page(p);
1255 mod_node_page_state(page_pgdat(page), idx, val);
1258 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1263 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1265 unsigned long *total_scanned)
1270 static inline void split_page_memcg(struct page *head, unsigned int nr)
1274 static inline void count_memcg_events(struct mem_cgroup *memcg,
1275 enum vm_event_item idx,
1276 unsigned long count)
1280 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1281 enum vm_event_item idx,
1282 unsigned long count)
1286 static inline void count_memcg_page_event(struct page *page,
1292 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1295 #endif /* CONFIG_MEMCG */
1297 /* idx can be of type enum memcg_stat_item or node_stat_item */
1298 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1301 __mod_memcg_state(memcg, idx, 1);
1304 /* idx can be of type enum memcg_stat_item or node_stat_item */
1305 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1308 __mod_memcg_state(memcg, idx, -1);
1311 /* idx can be of type enum memcg_stat_item or node_stat_item */
1312 static inline void __inc_memcg_page_state(struct page *page,
1315 __mod_memcg_page_state(page, idx, 1);
1318 /* idx can be of type enum memcg_stat_item or node_stat_item */
1319 static inline void __dec_memcg_page_state(struct page *page,
1322 __mod_memcg_page_state(page, idx, -1);
1325 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1326 enum node_stat_item idx)
1328 __mod_lruvec_state(lruvec, idx, 1);
1331 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1332 enum node_stat_item idx)
1334 __mod_lruvec_state(lruvec, idx, -1);
1337 static inline void __inc_lruvec_page_state(struct page *page,
1338 enum node_stat_item idx)
1340 __mod_lruvec_page_state(page, idx, 1);
1343 static inline void __dec_lruvec_page_state(struct page *page,
1344 enum node_stat_item idx)
1346 __mod_lruvec_page_state(page, idx, -1);
1349 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1351 __mod_lruvec_slab_state(p, idx, 1);
1354 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1356 __mod_lruvec_slab_state(p, idx, -1);
1359 /* idx can be of type enum memcg_stat_item or node_stat_item */
1360 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1363 mod_memcg_state(memcg, idx, 1);
1366 /* idx can be of type enum memcg_stat_item or node_stat_item */
1367 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1370 mod_memcg_state(memcg, idx, -1);
1373 /* idx can be of type enum memcg_stat_item or node_stat_item */
1374 static inline void inc_memcg_page_state(struct page *page,
1377 mod_memcg_page_state(page, idx, 1);
1380 /* idx can be of type enum memcg_stat_item or node_stat_item */
1381 static inline void dec_memcg_page_state(struct page *page,
1384 mod_memcg_page_state(page, idx, -1);
1387 static inline void inc_lruvec_state(struct lruvec *lruvec,
1388 enum node_stat_item idx)
1390 mod_lruvec_state(lruvec, idx, 1);
1393 static inline void dec_lruvec_state(struct lruvec *lruvec,
1394 enum node_stat_item idx)
1396 mod_lruvec_state(lruvec, idx, -1);
1399 static inline void inc_lruvec_page_state(struct page *page,
1400 enum node_stat_item idx)
1402 mod_lruvec_page_state(page, idx, 1);
1405 static inline void dec_lruvec_page_state(struct page *page,
1406 enum node_stat_item idx)
1408 mod_lruvec_page_state(page, idx, -1);
1411 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1413 struct mem_cgroup *memcg;
1415 memcg = lruvec_memcg(lruvec);
1418 memcg = parent_mem_cgroup(memcg);
1421 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1424 #ifdef CONFIG_CGROUP_WRITEBACK
1426 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1427 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1428 unsigned long *pheadroom, unsigned long *pdirty,
1429 unsigned long *pwriteback);
1431 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1432 struct bdi_writeback *wb);
1434 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1435 struct bdi_writeback *wb)
1437 if (mem_cgroup_disabled())
1440 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1441 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1444 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1446 #else /* CONFIG_CGROUP_WRITEBACK */
1448 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1453 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1454 unsigned long *pfilepages,
1455 unsigned long *pheadroom,
1456 unsigned long *pdirty,
1457 unsigned long *pwriteback)
1461 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1462 struct bdi_writeback *wb)
1466 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1470 #endif /* CONFIG_CGROUP_WRITEBACK */
1473 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1474 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1476 extern struct static_key_false memcg_sockets_enabled_key;
1477 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1478 void mem_cgroup_sk_alloc(struct sock *sk);
1479 void mem_cgroup_sk_free(struct sock *sk);
1480 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1482 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1485 if (time_before(jiffies, memcg->socket_pressure))
1487 } while ((memcg = parent_mem_cgroup(memcg)));
1491 extern int memcg_expand_shrinker_maps(int new_id);
1493 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1494 int nid, int shrinker_id);
1496 #define mem_cgroup_sockets_enabled 0
1497 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1498 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1499 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1504 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1505 int nid, int shrinker_id)
1510 #ifdef CONFIG_MEMCG_KMEM
1511 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1512 unsigned int nr_pages);
1513 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1514 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1515 void __memcg_kmem_uncharge_page(struct page *page, int order);
1517 struct obj_cgroup *get_obj_cgroup_from_current(void);
1519 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1520 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1522 extern struct static_key_false memcg_kmem_enabled_key;
1524 extern int memcg_nr_cache_ids;
1525 void memcg_get_cache_ids(void);
1526 void memcg_put_cache_ids(void);
1529 * Helper macro to loop through all memcg-specific caches. Callers must still
1530 * check if the cache is valid (it is either valid or NULL).
1531 * the slab_mutex must be held when looping through those caches
1533 #define for_each_memcg_cache_index(_idx) \
1534 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1536 static inline bool memcg_kmem_enabled(void)
1538 return static_branch_likely(&memcg_kmem_enabled_key);
1541 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1544 if (memcg_kmem_enabled())
1545 return __memcg_kmem_charge_page(page, gfp, order);
1549 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1551 if (memcg_kmem_enabled())
1552 __memcg_kmem_uncharge_page(page, order);
1555 static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1556 unsigned int nr_pages)
1558 if (memcg_kmem_enabled())
1559 return __memcg_kmem_charge(memcg, gfp, nr_pages);
1563 static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg,
1564 unsigned int nr_pages)
1566 if (memcg_kmem_enabled())
1567 __memcg_kmem_uncharge(memcg, nr_pages);
1571 * helper for accessing a memcg's index. It will be used as an index in the
1572 * child cache array in kmem_cache, and also to derive its name. This function
1573 * will return -1 when this is not a kmem-limited memcg.
1575 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1577 return memcg ? memcg->kmemcg_id : -1;
1580 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1584 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1590 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1594 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1600 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1604 #define for_each_memcg_cache_index(_idx) \
1607 static inline bool memcg_kmem_enabled(void)
1612 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1617 static inline void memcg_get_cache_ids(void)
1621 static inline void memcg_put_cache_ids(void)
1625 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1630 #endif /* CONFIG_MEMCG_KMEM */
1632 #endif /* _LINUX_MEMCONTROL_H */