1 /* memcontrol.h - Memory Controller
3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
6 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
20 #ifndef _LINUX_MEMCONTROL_H
21 #define _LINUX_MEMCONTROL_H
22 #include <linux/cgroup.h>
23 #include <linux/vm_event_item.h>
24 #include <linux/hardirq.h>
25 #include <linux/jump_label.h>
26 #include <linux/page_counter.h>
27 #include <linux/vmpressure.h>
28 #include <linux/eventfd.h>
30 #include <linux/vmstat.h>
31 #include <linux/writeback.h>
32 #include <linux/page-flags.h>
39 /* Cgroup-specific page state, on top of universal node page state */
40 enum memcg_stat_item {
41 MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
46 /* XXX: why are these zone and not node counters? */
47 MEMCG_KERNEL_STACK_KB,
51 enum memcg_memory_event {
59 MEMCG_NR_MEMORY_EVENTS,
62 enum mem_cgroup_protection {
68 struct mem_cgroup_reclaim_cookie {
71 unsigned int generation;
76 #define MEM_CGROUP_ID_SHIFT 16
77 #define MEM_CGROUP_ID_MAX USHRT_MAX
79 struct mem_cgroup_id {
85 * Per memcg event counter is incremented at every pagein/pageout. With THP,
86 * it will be incremated by the number of pages. This counter is used for
87 * for trigger some periodic events. This is straightforward and better
88 * than using jiffies etc. to handle periodic memcg event.
90 enum mem_cgroup_events_target {
91 MEM_CGROUP_TARGET_THRESH,
92 MEM_CGROUP_TARGET_SOFTLIMIT,
93 MEM_CGROUP_TARGET_NUMAINFO,
97 struct mem_cgroup_stat_cpu {
98 long count[MEMCG_NR_STAT];
99 unsigned long events[NR_VM_EVENT_ITEMS];
100 unsigned long nr_page_events;
101 unsigned long targets[MEM_CGROUP_NTARGETS];
104 struct mem_cgroup_reclaim_iter {
105 struct mem_cgroup *position;
106 /* scan generation, increased every round-trip */
107 unsigned int generation;
111 long count[NR_VM_NODE_STAT_ITEMS];
115 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
116 * which have elements charged to this memcg.
118 struct memcg_shrinker_map {
120 unsigned long map[0];
124 * per-zone information in memory controller.
126 struct mem_cgroup_per_node {
127 struct lruvec lruvec;
129 struct lruvec_stat __percpu *lruvec_stat_cpu;
130 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
132 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
134 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
136 #ifdef CONFIG_MEMCG_KMEM
137 struct memcg_shrinker_map __rcu *shrinker_map;
139 struct rb_node tree_node; /* RB tree node */
140 unsigned long usage_in_excess;/* Set to the value by which */
141 /* the soft limit is exceeded*/
143 bool congested; /* memcg has many dirty pages */
144 /* backed by a congested BDI */
146 struct mem_cgroup *memcg; /* Back pointer, we cannot */
147 /* use container_of */
150 struct mem_cgroup_threshold {
151 struct eventfd_ctx *eventfd;
152 unsigned long threshold;
156 struct mem_cgroup_threshold_ary {
157 /* An array index points to threshold just below or equal to usage. */
158 int current_threshold;
159 /* Size of entries[] */
161 /* Array of thresholds */
162 struct mem_cgroup_threshold entries[0];
165 struct mem_cgroup_thresholds {
166 /* Primary thresholds array */
167 struct mem_cgroup_threshold_ary *primary;
169 * Spare threshold array.
170 * This is needed to make mem_cgroup_unregister_event() "never fail".
171 * It must be able to store at least primary->size - 1 entries.
173 struct mem_cgroup_threshold_ary *spare;
176 enum memcg_kmem_state {
182 #if defined(CONFIG_SMP)
183 struct memcg_padding {
185 } ____cacheline_internodealigned_in_smp;
186 #define MEMCG_PADDING(name) struct memcg_padding name;
188 #define MEMCG_PADDING(name)
192 * The memory controller data structure. The memory controller controls both
193 * page cache and RSS per cgroup. We would eventually like to provide
194 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
195 * to help the administrator determine what knobs to tune.
198 struct cgroup_subsys_state css;
200 /* Private memcg ID. Used to ID objects that outlive the cgroup */
201 struct mem_cgroup_id id;
203 /* Accounted resources */
204 struct page_counter memory;
205 struct page_counter swap;
207 /* Legacy consumer-oriented counters */
208 struct page_counter memsw;
209 struct page_counter kmem;
210 struct page_counter tcpmem;
212 /* Upper bound of normal memory consumption range */
215 /* Range enforcement for interrupt charges */
216 struct work_struct high_work;
218 unsigned long soft_limit;
220 /* vmpressure notifications */
221 struct vmpressure vmpressure;
224 * Should the accounting and control be hierarchical, per subtree?
228 /* protected by memcg_oom_lock */
233 /* OOM-Killer disable */
234 int oom_kill_disable;
237 struct cgroup_file events_file;
239 /* handle for "memory.swap.events" */
240 struct cgroup_file swap_events_file;
242 /* protect arrays of thresholds */
243 struct mutex thresholds_lock;
245 /* thresholds for memory usage. RCU-protected */
246 struct mem_cgroup_thresholds thresholds;
248 /* thresholds for mem+swap usage. RCU-protected */
249 struct mem_cgroup_thresholds memsw_thresholds;
251 /* For oom notifier event fd */
252 struct list_head oom_notify;
255 * Should we move charges of a task when a task is moved into this
256 * mem_cgroup ? And what type of charges should we move ?
258 unsigned long move_charge_at_immigrate;
259 /* taken only while moving_account > 0 */
260 spinlock_t move_lock;
261 unsigned long move_lock_flags;
263 MEMCG_PADDING(_pad1_);
266 * set > 0 if pages under this cgroup are moving to other cgroup.
268 atomic_t moving_account;
269 struct task_struct *move_lock_task;
272 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
274 MEMCG_PADDING(_pad2_);
276 atomic_long_t stat[MEMCG_NR_STAT];
277 atomic_long_t events[NR_VM_EVENT_ITEMS];
278 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
280 unsigned long socket_pressure;
282 /* Legacy tcp memory accounting */
286 #ifdef CONFIG_MEMCG_KMEM
287 /* Index in the kmem_cache->memcg_params.memcg_caches array */
289 enum memcg_kmem_state kmem_state;
290 struct list_head kmem_caches;
293 int last_scanned_node;
295 nodemask_t scan_nodes;
296 atomic_t numainfo_events;
297 atomic_t numainfo_updating;
300 #ifdef CONFIG_CGROUP_WRITEBACK
301 struct list_head cgwb_list;
302 struct wb_domain cgwb_domain;
305 /* List of events which userspace want to receive */
306 struct list_head event_list;
307 spinlock_t event_list_lock;
309 struct mem_cgroup_per_node *nodeinfo[0];
310 /* WARNING: nodeinfo must be the last member here */
314 * size of first charge trial. "32" comes from vmscan.c's magic value.
315 * TODO: maybe necessary to use big numbers in big irons.
317 #define MEMCG_CHARGE_BATCH 32U
319 extern struct mem_cgroup *root_mem_cgroup;
321 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
323 return (memcg == root_mem_cgroup);
326 static inline bool mem_cgroup_disabled(void)
328 return !cgroup_subsys_enabled(memory_cgrp_subsys);
331 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
332 struct mem_cgroup *memcg);
334 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
335 gfp_t gfp_mask, struct mem_cgroup **memcgp,
337 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
338 gfp_t gfp_mask, struct mem_cgroup **memcgp,
340 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
341 bool lrucare, bool compound);
342 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
344 void mem_cgroup_uncharge(struct page *page);
345 void mem_cgroup_uncharge_list(struct list_head *page_list);
347 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
349 static struct mem_cgroup_per_node *
350 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
352 return memcg->nodeinfo[nid];
356 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
357 * @node: node of the wanted lruvec
358 * @memcg: memcg of the wanted lruvec
360 * Returns the lru list vector holding pages for a given @node or a given
361 * @memcg and @zone. This can be the node lruvec, if the memory controller
364 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
365 struct mem_cgroup *memcg)
367 struct mem_cgroup_per_node *mz;
368 struct lruvec *lruvec;
370 if (mem_cgroup_disabled()) {
371 lruvec = node_lruvec(pgdat);
375 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
376 lruvec = &mz->lruvec;
379 * Since a node can be onlined after the mem_cgroup was created,
380 * we have to be prepared to initialize lruvec->pgdat here;
381 * and if offlined then reonlined, we need to reinitialize it.
383 if (unlikely(lruvec->pgdat != pgdat))
384 lruvec->pgdat = pgdat;
388 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
390 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
391 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
393 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
395 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
398 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
399 return css ? container_of(css, struct mem_cgroup, css) : NULL;
402 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
405 css_put(&memcg->css);
408 #define mem_cgroup_from_counter(counter, member) \
409 container_of(counter, struct mem_cgroup, member)
411 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
413 struct mem_cgroup_reclaim_cookie *);
414 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
415 int mem_cgroup_scan_tasks(struct mem_cgroup *,
416 int (*)(struct task_struct *, void *), void *);
418 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
420 if (mem_cgroup_disabled())
425 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
427 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
429 struct mem_cgroup_per_node *mz;
431 if (mem_cgroup_disabled())
434 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
439 * parent_mem_cgroup - find the accounting parent of a memcg
440 * @memcg: memcg whose parent to find
442 * Returns the parent memcg, or NULL if this is the root or the memory
443 * controller is in legacy no-hierarchy mode.
445 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
447 if (!memcg->memory.parent)
449 return mem_cgroup_from_counter(memcg->memory.parent, memory);
452 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
453 struct mem_cgroup *root)
457 if (!root->use_hierarchy)
459 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
462 static inline bool mm_match_cgroup(struct mm_struct *mm,
463 struct mem_cgroup *memcg)
465 struct mem_cgroup *task_memcg;
469 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
471 match = mem_cgroup_is_descendant(task_memcg, memcg);
476 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
477 ino_t page_cgroup_ino(struct page *page);
479 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
481 if (mem_cgroup_disabled())
483 return !!(memcg->css.flags & CSS_ONLINE);
487 * For memory reclaim.
489 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
491 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
492 int zid, int nr_pages);
494 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
495 int nid, unsigned int lru_mask);
498 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
500 struct mem_cgroup_per_node *mz;
501 unsigned long nr_pages = 0;
504 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
505 for (zid = 0; zid < MAX_NR_ZONES; zid++)
506 nr_pages += mz->lru_zone_size[zid][lru];
511 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
512 enum lru_list lru, int zone_idx)
514 struct mem_cgroup_per_node *mz;
516 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
517 return mz->lru_zone_size[zone_idx][lru];
520 void mem_cgroup_handle_over_high(void);
522 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
524 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
525 struct task_struct *p);
527 static inline void mem_cgroup_enter_user_fault(void)
529 WARN_ON(current->in_user_fault);
530 current->in_user_fault = 1;
533 static inline void mem_cgroup_exit_user_fault(void)
535 WARN_ON(!current->in_user_fault);
536 current->in_user_fault = 0;
539 static inline bool task_in_memcg_oom(struct task_struct *p)
541 return p->memcg_in_oom;
544 bool mem_cgroup_oom_synchronize(bool wait);
546 #ifdef CONFIG_MEMCG_SWAP
547 extern int do_swap_account;
550 struct mem_cgroup *lock_page_memcg(struct page *page);
551 void __unlock_page_memcg(struct mem_cgroup *memcg);
552 void unlock_page_memcg(struct page *page);
554 /* idx can be of type enum memcg_stat_item or node_stat_item */
555 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
558 long x = atomic_long_read(&memcg->stat[idx]);
566 /* idx can be of type enum memcg_stat_item or node_stat_item */
567 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
572 if (mem_cgroup_disabled())
575 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
576 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
577 atomic_long_add(x, &memcg->stat[idx]);
580 __this_cpu_write(memcg->stat_cpu->count[idx], x);
583 /* idx can be of type enum memcg_stat_item or node_stat_item */
584 static inline void mod_memcg_state(struct mem_cgroup *memcg,
589 local_irq_save(flags);
590 __mod_memcg_state(memcg, idx, val);
591 local_irq_restore(flags);
595 * mod_memcg_page_state - update page state statistics
597 * @idx: page state item to account
598 * @val: number of pages (positive or negative)
600 * The @page must be locked or the caller must use lock_page_memcg()
601 * to prevent double accounting when the page is concurrently being
602 * moved to another memcg:
604 * lock_page(page) or lock_page_memcg(page)
605 * if (TestClearPageState(page))
606 * mod_memcg_page_state(page, state, -1);
607 * unlock_page(page) or unlock_page_memcg(page)
609 * Kernel pages are an exception to this, since they'll never move.
611 static inline void __mod_memcg_page_state(struct page *page,
614 if (page->mem_cgroup)
615 __mod_memcg_state(page->mem_cgroup, idx, val);
618 static inline void mod_memcg_page_state(struct page *page,
621 if (page->mem_cgroup)
622 mod_memcg_state(page->mem_cgroup, idx, val);
625 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
626 enum node_stat_item idx)
628 struct mem_cgroup_per_node *pn;
631 if (mem_cgroup_disabled())
632 return node_page_state(lruvec_pgdat(lruvec), idx);
634 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
635 x = atomic_long_read(&pn->lruvec_stat[idx]);
643 static inline void __mod_lruvec_state(struct lruvec *lruvec,
644 enum node_stat_item idx, int val)
646 struct mem_cgroup_per_node *pn;
650 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
652 if (mem_cgroup_disabled())
655 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
658 __mod_memcg_state(pn->memcg, idx, val);
661 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
662 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
663 atomic_long_add(x, &pn->lruvec_stat[idx]);
666 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
669 static inline void mod_lruvec_state(struct lruvec *lruvec,
670 enum node_stat_item idx, int val)
674 local_irq_save(flags);
675 __mod_lruvec_state(lruvec, idx, val);
676 local_irq_restore(flags);
679 static inline void __mod_lruvec_page_state(struct page *page,
680 enum node_stat_item idx, int val)
682 pg_data_t *pgdat = page_pgdat(page);
683 struct lruvec *lruvec;
685 /* Untracked pages have no memcg, no lruvec. Update only the node */
686 if (!page->mem_cgroup) {
687 __mod_node_page_state(pgdat, idx, val);
691 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
692 __mod_lruvec_state(lruvec, idx, val);
695 static inline void mod_lruvec_page_state(struct page *page,
696 enum node_stat_item idx, int val)
700 local_irq_save(flags);
701 __mod_lruvec_page_state(page, idx, val);
702 local_irq_restore(flags);
705 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
707 unsigned long *total_scanned);
709 static inline void __count_memcg_events(struct mem_cgroup *memcg,
710 enum vm_event_item idx,
715 if (mem_cgroup_disabled())
718 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
719 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
720 atomic_long_add(x, &memcg->events[idx]);
723 __this_cpu_write(memcg->stat_cpu->events[idx], x);
726 static inline void count_memcg_events(struct mem_cgroup *memcg,
727 enum vm_event_item idx,
732 local_irq_save(flags);
733 __count_memcg_events(memcg, idx, count);
734 local_irq_restore(flags);
737 static inline void count_memcg_page_event(struct page *page,
738 enum vm_event_item idx)
740 if (page->mem_cgroup)
741 count_memcg_events(page->mem_cgroup, idx, 1);
744 static inline void count_memcg_event_mm(struct mm_struct *mm,
745 enum vm_event_item idx)
747 struct mem_cgroup *memcg;
749 if (mem_cgroup_disabled())
753 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
755 count_memcg_events(memcg, idx, 1);
759 static inline void memcg_memory_event(struct mem_cgroup *memcg,
760 enum memcg_memory_event event)
762 atomic_long_inc(&memcg->memory_events[event]);
763 cgroup_file_notify(&memcg->events_file);
766 static inline void memcg_memory_event_mm(struct mm_struct *mm,
767 enum memcg_memory_event event)
769 struct mem_cgroup *memcg;
771 if (mem_cgroup_disabled())
775 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
777 memcg_memory_event(memcg, event);
781 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
782 void mem_cgroup_split_huge_fixup(struct page *head);
785 #else /* CONFIG_MEMCG */
787 #define MEM_CGROUP_ID_SHIFT 0
788 #define MEM_CGROUP_ID_MAX 0
792 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
797 static inline bool mem_cgroup_disabled(void)
802 static inline void memcg_memory_event(struct mem_cgroup *memcg,
803 enum memcg_memory_event event)
807 static inline void memcg_memory_event_mm(struct mm_struct *mm,
808 enum memcg_memory_event event)
812 static inline enum mem_cgroup_protection mem_cgroup_protected(
813 struct mem_cgroup *root, struct mem_cgroup *memcg)
815 return MEMCG_PROT_NONE;
818 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
820 struct mem_cgroup **memcgp,
827 static inline int mem_cgroup_try_charge_delay(struct page *page,
828 struct mm_struct *mm,
830 struct mem_cgroup **memcgp,
837 static inline void mem_cgroup_commit_charge(struct page *page,
838 struct mem_cgroup *memcg,
839 bool lrucare, bool compound)
843 static inline void mem_cgroup_cancel_charge(struct page *page,
844 struct mem_cgroup *memcg,
849 static inline void mem_cgroup_uncharge(struct page *page)
853 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
857 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
861 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
862 struct mem_cgroup *memcg)
864 return node_lruvec(pgdat);
867 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
868 struct pglist_data *pgdat)
870 return &pgdat->lruvec;
873 static inline bool mm_match_cgroup(struct mm_struct *mm,
874 struct mem_cgroup *memcg)
879 static inline bool task_in_mem_cgroup(struct task_struct *task,
880 const struct mem_cgroup *memcg)
885 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
890 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
895 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
899 static inline struct mem_cgroup *
900 mem_cgroup_iter(struct mem_cgroup *root,
901 struct mem_cgroup *prev,
902 struct mem_cgroup_reclaim_cookie *reclaim)
907 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
908 struct mem_cgroup *prev)
912 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
913 int (*fn)(struct task_struct *, void *), void *arg)
918 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
923 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
926 /* XXX: This should always return root_mem_cgroup */
930 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
935 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
940 static inline unsigned long
941 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
946 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
947 enum lru_list lru, int zone_idx)
952 static inline unsigned long
953 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
954 int nid, unsigned int lru_mask)
959 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
965 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
969 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
974 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
978 static inline void unlock_page_memcg(struct page *page)
982 static inline void mem_cgroup_handle_over_high(void)
986 static inline void mem_cgroup_enter_user_fault(void)
990 static inline void mem_cgroup_exit_user_fault(void)
994 static inline bool task_in_memcg_oom(struct task_struct *p)
999 static inline bool mem_cgroup_oom_synchronize(bool wait)
1004 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
1010 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1016 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1022 static inline void __mod_memcg_page_state(struct page *page,
1028 static inline void mod_memcg_page_state(struct page *page,
1034 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1035 enum node_stat_item idx)
1037 return node_page_state(lruvec_pgdat(lruvec), idx);
1040 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1041 enum node_stat_item idx, int val)
1043 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1046 static inline void mod_lruvec_state(struct lruvec *lruvec,
1047 enum node_stat_item idx, int val)
1049 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1052 static inline void __mod_lruvec_page_state(struct page *page,
1053 enum node_stat_item idx, int val)
1055 __mod_node_page_state(page_pgdat(page), idx, val);
1058 static inline void mod_lruvec_page_state(struct page *page,
1059 enum node_stat_item idx, int val)
1061 mod_node_page_state(page_pgdat(page), idx, val);
1065 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1067 unsigned long *total_scanned)
1072 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1076 static inline void count_memcg_events(struct mem_cgroup *memcg,
1077 enum vm_event_item idx,
1078 unsigned long count)
1082 static inline void count_memcg_page_event(struct page *page,
1088 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1091 #endif /* CONFIG_MEMCG */
1093 /* idx can be of type enum memcg_stat_item or node_stat_item */
1094 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1097 __mod_memcg_state(memcg, idx, 1);
1100 /* idx can be of type enum memcg_stat_item or node_stat_item */
1101 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1104 __mod_memcg_state(memcg, idx, -1);
1107 /* idx can be of type enum memcg_stat_item or node_stat_item */
1108 static inline void __inc_memcg_page_state(struct page *page,
1111 __mod_memcg_page_state(page, idx, 1);
1114 /* idx can be of type enum memcg_stat_item or node_stat_item */
1115 static inline void __dec_memcg_page_state(struct page *page,
1118 __mod_memcg_page_state(page, idx, -1);
1121 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1122 enum node_stat_item idx)
1124 __mod_lruvec_state(lruvec, idx, 1);
1127 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1128 enum node_stat_item idx)
1130 __mod_lruvec_state(lruvec, idx, -1);
1133 static inline void __inc_lruvec_page_state(struct page *page,
1134 enum node_stat_item idx)
1136 __mod_lruvec_page_state(page, idx, 1);
1139 static inline void __dec_lruvec_page_state(struct page *page,
1140 enum node_stat_item idx)
1142 __mod_lruvec_page_state(page, idx, -1);
1145 /* idx can be of type enum memcg_stat_item or node_stat_item */
1146 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1149 mod_memcg_state(memcg, idx, 1);
1152 /* idx can be of type enum memcg_stat_item or node_stat_item */
1153 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1156 mod_memcg_state(memcg, idx, -1);
1159 /* idx can be of type enum memcg_stat_item or node_stat_item */
1160 static inline void inc_memcg_page_state(struct page *page,
1163 mod_memcg_page_state(page, idx, 1);
1166 /* idx can be of type enum memcg_stat_item or node_stat_item */
1167 static inline void dec_memcg_page_state(struct page *page,
1170 mod_memcg_page_state(page, idx, -1);
1173 static inline void inc_lruvec_state(struct lruvec *lruvec,
1174 enum node_stat_item idx)
1176 mod_lruvec_state(lruvec, idx, 1);
1179 static inline void dec_lruvec_state(struct lruvec *lruvec,
1180 enum node_stat_item idx)
1182 mod_lruvec_state(lruvec, idx, -1);
1185 static inline void inc_lruvec_page_state(struct page *page,
1186 enum node_stat_item idx)
1188 mod_lruvec_page_state(page, idx, 1);
1191 static inline void dec_lruvec_page_state(struct page *page,
1192 enum node_stat_item idx)
1194 mod_lruvec_page_state(page, idx, -1);
1197 #ifdef CONFIG_CGROUP_WRITEBACK
1199 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1200 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1201 unsigned long *pheadroom, unsigned long *pdirty,
1202 unsigned long *pwriteback);
1204 #else /* CONFIG_CGROUP_WRITEBACK */
1206 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1211 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1212 unsigned long *pfilepages,
1213 unsigned long *pheadroom,
1214 unsigned long *pdirty,
1215 unsigned long *pwriteback)
1219 #endif /* CONFIG_CGROUP_WRITEBACK */
1222 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1223 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1225 extern struct static_key_false memcg_sockets_enabled_key;
1226 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1227 void mem_cgroup_sk_alloc(struct sock *sk);
1228 void mem_cgroup_sk_free(struct sock *sk);
1229 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1231 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1234 if (time_before(jiffies, memcg->socket_pressure))
1236 } while ((memcg = parent_mem_cgroup(memcg)));
1240 #define mem_cgroup_sockets_enabled 0
1241 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1242 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1243 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1249 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1250 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1251 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1252 struct mem_cgroup *memcg);
1253 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1254 void memcg_kmem_uncharge(struct page *page, int order);
1256 #ifdef CONFIG_MEMCG_KMEM
1257 extern struct static_key_false memcg_kmem_enabled_key;
1258 extern struct workqueue_struct *memcg_kmem_cache_wq;
1260 extern int memcg_nr_cache_ids;
1261 void memcg_get_cache_ids(void);
1262 void memcg_put_cache_ids(void);
1265 * Helper macro to loop through all memcg-specific caches. Callers must still
1266 * check if the cache is valid (it is either valid or NULL).
1267 * the slab_mutex must be held when looping through those caches
1269 #define for_each_memcg_cache_index(_idx) \
1270 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1272 static inline bool memcg_kmem_enabled(void)
1274 return static_branch_unlikely(&memcg_kmem_enabled_key);
1278 * helper for accessing a memcg's index. It will be used as an index in the
1279 * child cache array in kmem_cache, and also to derive its name. This function
1280 * will return -1 when this is not a kmem-limited memcg.
1282 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1284 return memcg ? memcg->kmemcg_id : -1;
1287 extern int memcg_expand_shrinker_maps(int new_id);
1289 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1290 int nid, int shrinker_id);
1292 #define for_each_memcg_cache_index(_idx) \
1295 static inline bool memcg_kmem_enabled(void)
1300 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1305 static inline void memcg_get_cache_ids(void)
1309 static inline void memcg_put_cache_ids(void)
1313 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1314 int nid, int shrinker_id) { }
1315 #endif /* CONFIG_MEMCG_KMEM */
1317 #endif /* _LINUX_MEMCONTROL_H */