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,
38 enum memcg_memory_event {
47 MEMCG_NR_MEMORY_EVENTS,
50 enum mem_cgroup_protection {
56 struct mem_cgroup_reclaim_cookie {
58 unsigned int generation;
63 #define MEM_CGROUP_ID_SHIFT 16
64 #define MEM_CGROUP_ID_MAX USHRT_MAX
66 struct mem_cgroup_id {
72 * Per memcg event counter is incremented at every pagein/pageout. With THP,
73 * it will be incremated by the number of pages. This counter is used for
74 * for trigger some periodic events. This is straightforward and better
75 * than using jiffies etc. to handle periodic memcg event.
77 enum mem_cgroup_events_target {
78 MEM_CGROUP_TARGET_THRESH,
79 MEM_CGROUP_TARGET_SOFTLIMIT,
83 struct memcg_vmstats_percpu {
84 long stat[MEMCG_NR_STAT];
85 unsigned long events[NR_VM_EVENT_ITEMS];
86 unsigned long nr_page_events;
87 unsigned long targets[MEM_CGROUP_NTARGETS];
90 struct mem_cgroup_reclaim_iter {
91 struct mem_cgroup *position;
92 /* scan generation, increased every round-trip */
93 unsigned int generation;
97 long count[NR_VM_NODE_STAT_ITEMS];
101 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
102 * which have elements charged to this memcg.
104 struct memcg_shrinker_map {
110 * per-node information in memory controller.
112 struct mem_cgroup_per_node {
113 struct lruvec lruvec;
115 /* Legacy local VM stats */
116 struct lruvec_stat __percpu *lruvec_stat_local;
118 /* Subtree VM stats (batched updates) */
119 struct lruvec_stat __percpu *lruvec_stat_cpu;
120 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
122 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
124 struct mem_cgroup_reclaim_iter iter;
126 struct memcg_shrinker_map __rcu *shrinker_map;
128 struct rb_node tree_node; /* RB tree node */
129 unsigned long usage_in_excess;/* Set to the value by which */
130 /* the soft limit is exceeded*/
132 struct mem_cgroup *memcg; /* Back pointer, we cannot */
133 /* use container_of */
136 struct mem_cgroup_threshold {
137 struct eventfd_ctx *eventfd;
138 unsigned long threshold;
142 struct mem_cgroup_threshold_ary {
143 /* An array index points to threshold just below or equal to usage. */
144 int current_threshold;
145 /* Size of entries[] */
147 /* Array of thresholds */
148 struct mem_cgroup_threshold entries[];
151 struct mem_cgroup_thresholds {
152 /* Primary thresholds array */
153 struct mem_cgroup_threshold_ary *primary;
155 * Spare threshold array.
156 * This is needed to make mem_cgroup_unregister_event() "never fail".
157 * It must be able to store at least primary->size - 1 entries.
159 struct mem_cgroup_threshold_ary *spare;
162 enum memcg_kmem_state {
168 #if defined(CONFIG_SMP)
169 struct memcg_padding {
171 } ____cacheline_internodealigned_in_smp;
172 #define MEMCG_PADDING(name) struct memcg_padding name;
174 #define MEMCG_PADDING(name)
178 * Remember four most recent foreign writebacks with dirty pages in this
179 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
180 * one in a given round, we're likely to catch it later if it keeps
181 * foreign-dirtying, so a fairly low count should be enough.
183 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
185 #define MEMCG_CGWB_FRN_CNT 4
187 struct memcg_cgwb_frn {
188 u64 bdi_id; /* bdi->id of the foreign inode */
189 int memcg_id; /* memcg->css.id of foreign inode */
190 u64 at; /* jiffies_64 at the time of dirtying */
191 struct wb_completion done; /* tracks in-flight foreign writebacks */
195 * Bucket for arbitrarily byte-sized objects charged to a memory
196 * cgroup. The bucket can be reparented in one piece when the cgroup
197 * is destroyed, without having to round up the individual references
198 * of all live memory objects in the wild.
201 struct percpu_ref refcnt;
202 struct mem_cgroup *memcg;
203 atomic_t nr_charged_bytes;
205 struct list_head list;
211 * The memory controller data structure. The memory controller controls both
212 * page cache and RSS per cgroup. We would eventually like to provide
213 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
214 * to help the administrator determine what knobs to tune.
217 struct cgroup_subsys_state css;
219 /* Private memcg ID. Used to ID objects that outlive the cgroup */
220 struct mem_cgroup_id id;
222 /* Accounted resources */
223 struct page_counter memory;
224 struct page_counter swap;
226 /* Legacy consumer-oriented counters */
227 struct page_counter memsw;
228 struct page_counter kmem;
229 struct page_counter tcpmem;
231 /* Range enforcement for interrupt charges */
232 struct work_struct high_work;
234 unsigned long soft_limit;
236 /* vmpressure notifications */
237 struct vmpressure vmpressure;
240 * Should the accounting and control be hierarchical, per subtree?
245 * Should the OOM killer kill all belonging tasks, had it kill one?
249 /* protected by memcg_oom_lock */
254 /* OOM-Killer disable */
255 int oom_kill_disable;
257 /* memory.events and memory.events.local */
258 struct cgroup_file events_file;
259 struct cgroup_file events_local_file;
261 /* handle for "memory.swap.events" */
262 struct cgroup_file swap_events_file;
264 /* protect arrays of thresholds */
265 struct mutex thresholds_lock;
267 /* thresholds for memory usage. RCU-protected */
268 struct mem_cgroup_thresholds thresholds;
270 /* thresholds for mem+swap usage. RCU-protected */
271 struct mem_cgroup_thresholds memsw_thresholds;
273 /* For oom notifier event fd */
274 struct list_head oom_notify;
277 * Should we move charges of a task when a task is moved into this
278 * mem_cgroup ? And what type of charges should we move ?
280 unsigned long move_charge_at_immigrate;
281 /* taken only while moving_account > 0 */
282 spinlock_t move_lock;
283 unsigned long move_lock_flags;
285 MEMCG_PADDING(_pad1_);
288 * set > 0 if pages under this cgroup are moving to other cgroup.
290 atomic_t moving_account;
291 struct task_struct *move_lock_task;
293 /* Legacy local VM stats and events */
294 struct memcg_vmstats_percpu __percpu *vmstats_local;
296 /* Subtree VM stats and events (batched updates) */
297 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
299 MEMCG_PADDING(_pad2_);
301 atomic_long_t vmstats[MEMCG_NR_STAT];
302 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
305 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
306 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
308 unsigned long socket_pressure;
310 /* Legacy tcp memory accounting */
314 #ifdef CONFIG_MEMCG_KMEM
315 /* Index in the kmem_cache->memcg_params.memcg_caches array */
317 enum memcg_kmem_state kmem_state;
318 struct obj_cgroup __rcu *objcg;
319 struct list_head objcg_list; /* list of inherited objcgs */
322 #ifdef CONFIG_CGROUP_WRITEBACK
323 struct list_head cgwb_list;
324 struct wb_domain cgwb_domain;
325 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
328 /* List of events which userspace want to receive */
329 struct list_head event_list;
330 spinlock_t event_list_lock;
332 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
333 struct deferred_split deferred_split_queue;
336 struct mem_cgroup_per_node *nodeinfo[0];
337 /* WARNING: nodeinfo must be the last member here */
341 * size of first charge trial. "32" comes from vmscan.c's magic value.
342 * TODO: maybe necessary to use big numbers in big irons.
344 #define MEMCG_CHARGE_BATCH 32U
346 extern struct mem_cgroup *root_mem_cgroup;
348 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
350 return (memcg == root_mem_cgroup);
353 static inline bool mem_cgroup_disabled(void)
355 return !cgroup_subsys_enabled(memory_cgrp_subsys);
358 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *root,
359 struct mem_cgroup *memcg,
362 if (mem_cgroup_disabled())
366 * There is no reclaim protection applied to a targeted reclaim.
367 * We are special casing this specific case here because
368 * mem_cgroup_protected calculation is not robust enough to keep
369 * the protection invariant for calculated effective values for
370 * parallel reclaimers with different reclaim target. This is
371 * especially a problem for tail memcgs (as they have pages on LRU)
372 * which would want to have effective values 0 for targeted reclaim
373 * but a different value for external reclaim.
376 * Let's have global and A's reclaim in parallel:
378 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
380 * | C (low = 1G, usage = 2.5G)
381 * B (low = 1G, usage = 0.5G)
383 * For the global reclaim
385 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
386 * C.elow = min(C.usage, C.low)
388 * With the effective values resetting we have A reclaim
393 * If the global reclaim races with A's reclaim then
394 * B.elow = C.elow = 0 because children_low_usage > A.elow)
395 * is possible and reclaiming B would be violating the protection.
402 return READ_ONCE(memcg->memory.emin);
404 return max(READ_ONCE(memcg->memory.emin),
405 READ_ONCE(memcg->memory.elow));
408 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
409 struct mem_cgroup *memcg);
411 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask);
413 void mem_cgroup_uncharge(struct page *page);
414 void mem_cgroup_uncharge_list(struct list_head *page_list);
416 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
418 static struct mem_cgroup_per_node *
419 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
421 return memcg->nodeinfo[nid];
425 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
426 * @memcg: memcg of the wanted lruvec
428 * Returns the lru list vector holding pages for a given @memcg &
429 * @node combination. This can be the node lruvec, if the memory
430 * controller is disabled.
432 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
433 struct pglist_data *pgdat)
435 struct mem_cgroup_per_node *mz;
436 struct lruvec *lruvec;
438 if (mem_cgroup_disabled()) {
439 lruvec = &pgdat->__lruvec;
444 memcg = root_mem_cgroup;
446 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
447 lruvec = &mz->lruvec;
450 * Since a node can be onlined after the mem_cgroup was created,
451 * we have to be prepared to initialize lruvec->pgdat here;
452 * and if offlined then reonlined, we need to reinitialize it.
454 if (unlikely(lruvec->pgdat != pgdat))
455 lruvec->pgdat = pgdat;
459 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
461 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
463 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
465 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
468 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
469 return css ? container_of(css, struct mem_cgroup, css) : NULL;
472 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
474 return percpu_ref_tryget(&objcg->refcnt);
477 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
479 percpu_ref_get(&objcg->refcnt);
482 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
484 percpu_ref_put(&objcg->refcnt);
488 * After the initialization objcg->memcg is always pointing at
489 * a valid memcg, but can be atomically swapped to the parent memcg.
491 * The caller must ensure that the returned memcg won't be released:
492 * e.g. acquire the rcu_read_lock or css_set_lock.
494 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
496 return READ_ONCE(objcg->memcg);
499 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
502 css_put(&memcg->css);
505 #define mem_cgroup_from_counter(counter, member) \
506 container_of(counter, struct mem_cgroup, member)
508 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
510 struct mem_cgroup_reclaim_cookie *);
511 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
512 int mem_cgroup_scan_tasks(struct mem_cgroup *,
513 int (*)(struct task_struct *, void *), void *);
515 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
517 if (mem_cgroup_disabled())
522 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
524 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
526 return mem_cgroup_from_css(seq_css(m));
529 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
531 struct mem_cgroup_per_node *mz;
533 if (mem_cgroup_disabled())
536 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
541 * parent_mem_cgroup - find the accounting parent of a memcg
542 * @memcg: memcg whose parent to find
544 * Returns the parent memcg, or NULL if this is the root or the memory
545 * controller is in legacy no-hierarchy mode.
547 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
549 if (!memcg->memory.parent)
551 return mem_cgroup_from_counter(memcg->memory.parent, memory);
554 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
555 struct mem_cgroup *root)
559 if (!root->use_hierarchy)
561 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
564 static inline bool mm_match_cgroup(struct mm_struct *mm,
565 struct mem_cgroup *memcg)
567 struct mem_cgroup *task_memcg;
571 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
573 match = mem_cgroup_is_descendant(task_memcg, memcg);
578 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
579 ino_t page_cgroup_ino(struct page *page);
581 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
583 if (mem_cgroup_disabled())
585 return !!(memcg->css.flags & CSS_ONLINE);
589 * For memory reclaim.
591 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
593 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
594 int zid, int nr_pages);
597 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
598 enum lru_list lru, int zone_idx)
600 struct mem_cgroup_per_node *mz;
602 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
603 return mz->lru_zone_size[zone_idx][lru];
606 void mem_cgroup_handle_over_high(void);
608 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
610 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
612 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
613 struct task_struct *p);
615 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
617 static inline void mem_cgroup_enter_user_fault(void)
619 WARN_ON(current->in_user_fault);
620 current->in_user_fault = 1;
623 static inline void mem_cgroup_exit_user_fault(void)
625 WARN_ON(!current->in_user_fault);
626 current->in_user_fault = 0;
629 static inline bool task_in_memcg_oom(struct task_struct *p)
631 return p->memcg_in_oom;
634 bool mem_cgroup_oom_synchronize(bool wait);
635 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
636 struct mem_cgroup *oom_domain);
637 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
639 #ifdef CONFIG_MEMCG_SWAP
640 extern bool cgroup_memory_noswap;
643 struct mem_cgroup *lock_page_memcg(struct page *page);
644 void __unlock_page_memcg(struct mem_cgroup *memcg);
645 void unlock_page_memcg(struct page *page);
648 * idx can be of type enum memcg_stat_item or node_stat_item.
649 * Keep in sync with memcg_exact_page_state().
651 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
653 long x = atomic_long_read(&memcg->vmstats[idx]);
662 * idx can be of type enum memcg_stat_item or node_stat_item.
663 * Keep in sync with memcg_exact_page_state().
665 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
671 for_each_possible_cpu(cpu)
672 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
680 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
682 /* idx can be of type enum memcg_stat_item or node_stat_item */
683 static inline void mod_memcg_state(struct mem_cgroup *memcg,
688 local_irq_save(flags);
689 __mod_memcg_state(memcg, idx, val);
690 local_irq_restore(flags);
694 * mod_memcg_page_state - update page state statistics
696 * @idx: page state item to account
697 * @val: number of pages (positive or negative)
699 * The @page must be locked or the caller must use lock_page_memcg()
700 * to prevent double accounting when the page is concurrently being
701 * moved to another memcg:
703 * lock_page(page) or lock_page_memcg(page)
704 * if (TestClearPageState(page))
705 * mod_memcg_page_state(page, state, -1);
706 * unlock_page(page) or unlock_page_memcg(page)
708 * Kernel pages are an exception to this, since they'll never move.
710 static inline void __mod_memcg_page_state(struct page *page,
713 if (page->mem_cgroup)
714 __mod_memcg_state(page->mem_cgroup, idx, val);
717 static inline void mod_memcg_page_state(struct page *page,
720 if (page->mem_cgroup)
721 mod_memcg_state(page->mem_cgroup, idx, val);
724 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
725 enum node_stat_item idx)
727 struct mem_cgroup_per_node *pn;
730 if (mem_cgroup_disabled())
731 return node_page_state(lruvec_pgdat(lruvec), idx);
733 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
734 x = atomic_long_read(&pn->lruvec_stat[idx]);
742 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
743 enum node_stat_item idx)
745 struct mem_cgroup_per_node *pn;
749 if (mem_cgroup_disabled())
750 return node_page_state(lruvec_pgdat(lruvec), idx);
752 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
753 for_each_possible_cpu(cpu)
754 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
762 void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
764 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
766 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
768 void mod_memcg_obj_state(void *p, int idx, int val);
770 static inline void mod_lruvec_slab_state(void *p, enum node_stat_item idx,
775 local_irq_save(flags);
776 __mod_lruvec_slab_state(p, idx, val);
777 local_irq_restore(flags);
780 static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
781 enum node_stat_item idx, int val)
785 local_irq_save(flags);
786 __mod_memcg_lruvec_state(lruvec, idx, val);
787 local_irq_restore(flags);
790 static inline void mod_lruvec_state(struct lruvec *lruvec,
791 enum node_stat_item idx, int val)
795 local_irq_save(flags);
796 __mod_lruvec_state(lruvec, idx, val);
797 local_irq_restore(flags);
800 static inline void __mod_lruvec_page_state(struct page *page,
801 enum node_stat_item idx, int val)
803 struct page *head = compound_head(page); /* rmap on tail pages */
804 pg_data_t *pgdat = page_pgdat(page);
805 struct lruvec *lruvec;
807 /* Untracked pages have no memcg, no lruvec. Update only the node */
808 if (!head->mem_cgroup) {
809 __mod_node_page_state(pgdat, idx, val);
813 lruvec = mem_cgroup_lruvec(head->mem_cgroup, pgdat);
814 __mod_lruvec_state(lruvec, idx, val);
817 static inline void mod_lruvec_page_state(struct page *page,
818 enum node_stat_item idx, int val)
822 local_irq_save(flags);
823 __mod_lruvec_page_state(page, idx, val);
824 local_irq_restore(flags);
827 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
829 unsigned long *total_scanned);
831 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
832 unsigned long count);
834 static inline void count_memcg_events(struct mem_cgroup *memcg,
835 enum vm_event_item idx,
840 local_irq_save(flags);
841 __count_memcg_events(memcg, idx, count);
842 local_irq_restore(flags);
845 static inline void count_memcg_page_event(struct page *page,
846 enum vm_event_item idx)
848 if (page->mem_cgroup)
849 count_memcg_events(page->mem_cgroup, idx, 1);
852 static inline void count_memcg_event_mm(struct mm_struct *mm,
853 enum vm_event_item idx)
855 struct mem_cgroup *memcg;
857 if (mem_cgroup_disabled())
861 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
863 count_memcg_events(memcg, idx, 1);
867 static inline void memcg_memory_event(struct mem_cgroup *memcg,
868 enum memcg_memory_event event)
870 atomic_long_inc(&memcg->memory_events_local[event]);
871 cgroup_file_notify(&memcg->events_local_file);
874 atomic_long_inc(&memcg->memory_events[event]);
875 cgroup_file_notify(&memcg->events_file);
877 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
879 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
881 } while ((memcg = parent_mem_cgroup(memcg)) &&
882 !mem_cgroup_is_root(memcg));
885 static inline void memcg_memory_event_mm(struct mm_struct *mm,
886 enum memcg_memory_event event)
888 struct mem_cgroup *memcg;
890 if (mem_cgroup_disabled())
894 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
896 memcg_memory_event(memcg, event);
900 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
901 void mem_cgroup_split_huge_fixup(struct page *head);
904 #else /* CONFIG_MEMCG */
906 #define MEM_CGROUP_ID_SHIFT 0
907 #define MEM_CGROUP_ID_MAX 0
911 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
916 static inline bool mem_cgroup_disabled(void)
921 static inline void memcg_memory_event(struct mem_cgroup *memcg,
922 enum memcg_memory_event event)
926 static inline void memcg_memory_event_mm(struct mm_struct *mm,
927 enum memcg_memory_event event)
931 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *root,
932 struct mem_cgroup *memcg,
938 static inline enum mem_cgroup_protection mem_cgroup_protected(
939 struct mem_cgroup *root, struct mem_cgroup *memcg)
941 return MEMCG_PROT_NONE;
944 static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
950 static inline void mem_cgroup_uncharge(struct page *page)
954 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
958 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
962 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
963 struct pglist_data *pgdat)
965 return &pgdat->__lruvec;
968 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
969 struct pglist_data *pgdat)
971 return &pgdat->__lruvec;
974 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
979 static inline bool mm_match_cgroup(struct mm_struct *mm,
980 struct mem_cgroup *memcg)
985 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
990 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
995 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
999 static inline struct mem_cgroup *
1000 mem_cgroup_iter(struct mem_cgroup *root,
1001 struct mem_cgroup *prev,
1002 struct mem_cgroup_reclaim_cookie *reclaim)
1007 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1008 struct mem_cgroup *prev)
1012 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1013 int (*fn)(struct task_struct *, void *), void *arg)
1018 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1023 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1026 /* XXX: This should always return root_mem_cgroup */
1030 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1035 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1040 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1046 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1047 enum lru_list lru, int zone_idx)
1052 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1057 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1063 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1068 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1072 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1077 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1081 static inline void unlock_page_memcg(struct page *page)
1085 static inline void mem_cgroup_handle_over_high(void)
1089 static inline void mem_cgroup_enter_user_fault(void)
1093 static inline void mem_cgroup_exit_user_fault(void)
1097 static inline bool task_in_memcg_oom(struct task_struct *p)
1102 static inline bool mem_cgroup_oom_synchronize(bool wait)
1107 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1108 struct task_struct *victim, struct mem_cgroup *oom_domain)
1113 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1117 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1122 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1128 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1134 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1140 static inline void __mod_memcg_page_state(struct page *page,
1146 static inline void mod_memcg_page_state(struct page *page,
1152 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1153 enum node_stat_item idx)
1155 return node_page_state(lruvec_pgdat(lruvec), idx);
1158 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1159 enum node_stat_item idx)
1161 return node_page_state(lruvec_pgdat(lruvec), idx);
1164 static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1165 enum node_stat_item idx, int val)
1169 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1170 enum node_stat_item idx, int val)
1172 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1175 static inline void mod_lruvec_state(struct lruvec *lruvec,
1176 enum node_stat_item idx, int val)
1178 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1181 static inline void __mod_lruvec_page_state(struct page *page,
1182 enum node_stat_item idx, int val)
1184 __mod_node_page_state(page_pgdat(page), idx, val);
1187 static inline void mod_lruvec_page_state(struct page *page,
1188 enum node_stat_item idx, int val)
1190 mod_node_page_state(page_pgdat(page), idx, val);
1193 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1196 struct page *page = virt_to_head_page(p);
1198 __mod_node_page_state(page_pgdat(page), idx, val);
1201 static inline void mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1204 struct page *page = virt_to_head_page(p);
1206 mod_node_page_state(page_pgdat(page), idx, val);
1209 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1214 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1216 unsigned long *total_scanned)
1221 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1225 static inline void count_memcg_events(struct mem_cgroup *memcg,
1226 enum vm_event_item idx,
1227 unsigned long count)
1231 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1232 enum vm_event_item idx,
1233 unsigned long count)
1237 static inline void count_memcg_page_event(struct page *page,
1243 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1246 #endif /* CONFIG_MEMCG */
1248 /* idx can be of type enum memcg_stat_item or node_stat_item */
1249 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1252 __mod_memcg_state(memcg, idx, 1);
1255 /* idx can be of type enum memcg_stat_item or node_stat_item */
1256 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1259 __mod_memcg_state(memcg, idx, -1);
1262 /* idx can be of type enum memcg_stat_item or node_stat_item */
1263 static inline void __inc_memcg_page_state(struct page *page,
1266 __mod_memcg_page_state(page, idx, 1);
1269 /* idx can be of type enum memcg_stat_item or node_stat_item */
1270 static inline void __dec_memcg_page_state(struct page *page,
1273 __mod_memcg_page_state(page, idx, -1);
1276 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1277 enum node_stat_item idx)
1279 __mod_lruvec_state(lruvec, idx, 1);
1282 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1283 enum node_stat_item idx)
1285 __mod_lruvec_state(lruvec, idx, -1);
1288 static inline void __inc_lruvec_page_state(struct page *page,
1289 enum node_stat_item idx)
1291 __mod_lruvec_page_state(page, idx, 1);
1294 static inline void __dec_lruvec_page_state(struct page *page,
1295 enum node_stat_item idx)
1297 __mod_lruvec_page_state(page, idx, -1);
1300 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1302 __mod_lruvec_slab_state(p, idx, 1);
1305 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1307 __mod_lruvec_slab_state(p, idx, -1);
1310 /* idx can be of type enum memcg_stat_item or node_stat_item */
1311 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1314 mod_memcg_state(memcg, idx, 1);
1317 /* idx can be of type enum memcg_stat_item or node_stat_item */
1318 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1321 mod_memcg_state(memcg, idx, -1);
1324 /* idx can be of type enum memcg_stat_item or node_stat_item */
1325 static inline void inc_memcg_page_state(struct page *page,
1328 mod_memcg_page_state(page, idx, 1);
1331 /* idx can be of type enum memcg_stat_item or node_stat_item */
1332 static inline void dec_memcg_page_state(struct page *page,
1335 mod_memcg_page_state(page, idx, -1);
1338 static inline void inc_lruvec_state(struct lruvec *lruvec,
1339 enum node_stat_item idx)
1341 mod_lruvec_state(lruvec, idx, 1);
1344 static inline void dec_lruvec_state(struct lruvec *lruvec,
1345 enum node_stat_item idx)
1347 mod_lruvec_state(lruvec, idx, -1);
1350 static inline void inc_lruvec_page_state(struct page *page,
1351 enum node_stat_item idx)
1353 mod_lruvec_page_state(page, idx, 1);
1356 static inline void dec_lruvec_page_state(struct page *page,
1357 enum node_stat_item idx)
1359 mod_lruvec_page_state(page, idx, -1);
1362 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1364 struct mem_cgroup *memcg;
1366 memcg = lruvec_memcg(lruvec);
1369 memcg = parent_mem_cgroup(memcg);
1372 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1375 #ifdef CONFIG_CGROUP_WRITEBACK
1377 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1378 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1379 unsigned long *pheadroom, unsigned long *pdirty,
1380 unsigned long *pwriteback);
1382 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1383 struct bdi_writeback *wb);
1385 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1386 struct bdi_writeback *wb)
1388 if (mem_cgroup_disabled())
1391 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1392 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1395 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1397 #else /* CONFIG_CGROUP_WRITEBACK */
1399 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1404 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1405 unsigned long *pfilepages,
1406 unsigned long *pheadroom,
1407 unsigned long *pdirty,
1408 unsigned long *pwriteback)
1412 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1413 struct bdi_writeback *wb)
1417 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1421 #endif /* CONFIG_CGROUP_WRITEBACK */
1424 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1425 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1427 extern struct static_key_false memcg_sockets_enabled_key;
1428 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1429 void mem_cgroup_sk_alloc(struct sock *sk);
1430 void mem_cgroup_sk_free(struct sock *sk);
1431 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1433 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1436 if (time_before(jiffies, memcg->socket_pressure))
1438 } while ((memcg = parent_mem_cgroup(memcg)));
1442 extern int memcg_expand_shrinker_maps(int new_id);
1444 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1445 int nid, int shrinker_id);
1447 #define mem_cgroup_sockets_enabled 0
1448 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1449 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1450 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1455 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1456 int nid, int shrinker_id)
1461 #ifdef CONFIG_MEMCG_KMEM
1462 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1463 unsigned int nr_pages);
1464 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1465 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1466 void __memcg_kmem_uncharge_page(struct page *page, int order);
1468 struct obj_cgroup *get_obj_cgroup_from_current(void);
1470 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1471 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1473 extern struct static_key_false memcg_kmem_enabled_key;
1475 extern int memcg_nr_cache_ids;
1476 void memcg_get_cache_ids(void);
1477 void memcg_put_cache_ids(void);
1480 * Helper macro to loop through all memcg-specific caches. Callers must still
1481 * check if the cache is valid (it is either valid or NULL).
1482 * the slab_mutex must be held when looping through those caches
1484 #define for_each_memcg_cache_index(_idx) \
1485 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1487 static inline bool memcg_kmem_enabled(void)
1489 return static_branch_likely(&memcg_kmem_enabled_key);
1492 static inline bool memcg_kmem_bypass(void)
1497 /* Allow remote memcg charging in kthread contexts. */
1498 if ((!current->mm || (current->flags & PF_KTHREAD)) &&
1499 !current->active_memcg)
1504 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1507 if (memcg_kmem_enabled())
1508 return __memcg_kmem_charge_page(page, gfp, order);
1512 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1514 if (memcg_kmem_enabled())
1515 __memcg_kmem_uncharge_page(page, order);
1518 static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1519 unsigned int nr_pages)
1521 if (memcg_kmem_enabled())
1522 return __memcg_kmem_charge(memcg, gfp, nr_pages);
1526 static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg,
1527 unsigned int nr_pages)
1529 if (memcg_kmem_enabled())
1530 __memcg_kmem_uncharge(memcg, nr_pages);
1534 * helper for accessing a memcg's index. It will be used as an index in the
1535 * child cache array in kmem_cache, and also to derive its name. This function
1536 * will return -1 when this is not a kmem-limited memcg.
1538 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1540 return memcg ? memcg->kmemcg_id : -1;
1543 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1547 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1553 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1557 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1563 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1567 #define for_each_memcg_cache_index(_idx) \
1570 static inline bool memcg_kmem_enabled(void)
1575 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1580 static inline void memcg_get_cache_ids(void)
1584 static inline void memcg_put_cache_ids(void)
1588 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1593 #endif /* CONFIG_MEMCG_KMEM */
1595 #endif /* _LINUX_MEMCONTROL_H */