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 {
58 MEMCG_NR_MEMORY_EVENTS,
61 struct mem_cgroup_reclaim_cookie {
64 unsigned int generation;
69 #define MEM_CGROUP_ID_SHIFT 16
70 #define MEM_CGROUP_ID_MAX USHRT_MAX
72 struct mem_cgroup_id {
78 * Per memcg event counter is incremented at every pagein/pageout. With THP,
79 * it will be incremated by the number of pages. This counter is used for
80 * for trigger some periodic events. This is straightforward and better
81 * than using jiffies etc. to handle periodic memcg event.
83 enum mem_cgroup_events_target {
84 MEM_CGROUP_TARGET_THRESH,
85 MEM_CGROUP_TARGET_SOFTLIMIT,
86 MEM_CGROUP_TARGET_NUMAINFO,
90 struct mem_cgroup_stat_cpu {
91 long count[MEMCG_NR_STAT];
92 unsigned long events[NR_VM_EVENT_ITEMS];
93 unsigned long nr_page_events;
94 unsigned long targets[MEM_CGROUP_NTARGETS];
97 struct mem_cgroup_reclaim_iter {
98 struct mem_cgroup *position;
99 /* scan generation, increased every round-trip */
100 unsigned int generation;
104 long count[NR_VM_NODE_STAT_ITEMS];
108 * per-zone information in memory controller.
110 struct mem_cgroup_per_node {
111 struct lruvec lruvec;
113 struct lruvec_stat __percpu *lruvec_stat_cpu;
114 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
116 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
118 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
120 struct rb_node tree_node; /* RB tree node */
121 unsigned long usage_in_excess;/* Set to the value by which */
122 /* the soft limit is exceeded*/
124 bool congested; /* memcg has many dirty pages */
125 /* backed by a congested BDI */
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[0];
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 {
164 * The memory controller data structure. The memory controller controls both
165 * page cache and RSS per cgroup. We would eventually like to provide
166 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
167 * to help the administrator determine what knobs to tune.
170 struct cgroup_subsys_state css;
172 /* Private memcg ID. Used to ID objects that outlive the cgroup */
173 struct mem_cgroup_id id;
175 /* Accounted resources */
176 struct page_counter memory;
177 struct page_counter swap;
179 /* Legacy consumer-oriented counters */
180 struct page_counter memsw;
181 struct page_counter kmem;
182 struct page_counter tcpmem;
184 /* Normal memory consumption range */
188 /* Range enforcement for interrupt charges */
189 struct work_struct high_work;
191 unsigned long soft_limit;
193 /* vmpressure notifications */
194 struct vmpressure vmpressure;
197 * Should the accounting and control be hierarchical, per subtree?
201 /* protected by memcg_oom_lock */
206 /* OOM-Killer disable */
207 int oom_kill_disable;
210 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
211 struct cgroup_file events_file;
213 /* handle for "memory.swap.events" */
214 struct cgroup_file swap_events_file;
216 /* protect arrays of thresholds */
217 struct mutex thresholds_lock;
219 /* thresholds for memory usage. RCU-protected */
220 struct mem_cgroup_thresholds thresholds;
222 /* thresholds for mem+swap usage. RCU-protected */
223 struct mem_cgroup_thresholds memsw_thresholds;
225 /* For oom notifier event fd */
226 struct list_head oom_notify;
229 * Should we move charges of a task when a task is moved into this
230 * mem_cgroup ? And what type of charges should we move ?
232 unsigned long move_charge_at_immigrate;
234 * set > 0 if pages under this cgroup are moving to other cgroup.
236 atomic_t moving_account;
237 /* taken only while moving_account > 0 */
238 spinlock_t move_lock;
239 struct task_struct *move_lock_task;
240 unsigned long move_lock_flags;
243 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
244 atomic_long_t stat[MEMCG_NR_STAT];
245 atomic_long_t events[NR_VM_EVENT_ITEMS];
247 unsigned long socket_pressure;
249 /* Legacy tcp memory accounting */
254 /* Index in the kmem_cache->memcg_params.memcg_caches array */
256 enum memcg_kmem_state kmem_state;
257 struct list_head kmem_caches;
260 int last_scanned_node;
262 nodemask_t scan_nodes;
263 atomic_t numainfo_events;
264 atomic_t numainfo_updating;
267 #ifdef CONFIG_CGROUP_WRITEBACK
268 struct list_head cgwb_list;
269 struct wb_domain cgwb_domain;
272 /* List of events which userspace want to receive */
273 struct list_head event_list;
274 spinlock_t event_list_lock;
276 struct mem_cgroup_per_node *nodeinfo[0];
277 /* WARNING: nodeinfo must be the last member here */
281 * size of first charge trial. "32" comes from vmscan.c's magic value.
282 * TODO: maybe necessary to use big numbers in big irons.
284 #define MEMCG_CHARGE_BATCH 32U
286 extern struct mem_cgroup *root_mem_cgroup;
288 static inline bool mem_cgroup_disabled(void)
290 return !cgroup_subsys_enabled(memory_cgrp_subsys);
293 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
295 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
296 gfp_t gfp_mask, struct mem_cgroup **memcgp,
298 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
299 bool lrucare, bool compound);
300 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
302 void mem_cgroup_uncharge(struct page *page);
303 void mem_cgroup_uncharge_list(struct list_head *page_list);
305 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
307 static struct mem_cgroup_per_node *
308 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
310 return memcg->nodeinfo[nid];
314 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
315 * @node: node of the wanted lruvec
316 * @memcg: memcg of the wanted lruvec
318 * Returns the lru list vector holding pages for a given @node or a given
319 * @memcg and @zone. This can be the node lruvec, if the memory controller
322 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
323 struct mem_cgroup *memcg)
325 struct mem_cgroup_per_node *mz;
326 struct lruvec *lruvec;
328 if (mem_cgroup_disabled()) {
329 lruvec = node_lruvec(pgdat);
333 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
334 lruvec = &mz->lruvec;
337 * Since a node can be onlined after the mem_cgroup was created,
338 * we have to be prepared to initialize lruvec->pgdat here;
339 * and if offlined then reonlined, we need to reinitialize it.
341 if (unlikely(lruvec->pgdat != pgdat))
342 lruvec->pgdat = pgdat;
346 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
348 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
349 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
352 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
353 return css ? container_of(css, struct mem_cgroup, css) : NULL;
356 #define mem_cgroup_from_counter(counter, member) \
357 container_of(counter, struct mem_cgroup, member)
359 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
361 struct mem_cgroup_reclaim_cookie *);
362 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
363 int mem_cgroup_scan_tasks(struct mem_cgroup *,
364 int (*)(struct task_struct *, void *), void *);
366 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
368 if (mem_cgroup_disabled())
373 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
375 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
377 struct mem_cgroup_per_node *mz;
379 if (mem_cgroup_disabled())
382 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
387 * parent_mem_cgroup - find the accounting parent of a memcg
388 * @memcg: memcg whose parent to find
390 * Returns the parent memcg, or NULL if this is the root or the memory
391 * controller is in legacy no-hierarchy mode.
393 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
395 if (!memcg->memory.parent)
397 return mem_cgroup_from_counter(memcg->memory.parent, memory);
400 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
401 struct mem_cgroup *root)
405 if (!root->use_hierarchy)
407 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
410 static inline bool mm_match_cgroup(struct mm_struct *mm,
411 struct mem_cgroup *memcg)
413 struct mem_cgroup *task_memcg;
417 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
419 match = mem_cgroup_is_descendant(task_memcg, memcg);
424 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
425 ino_t page_cgroup_ino(struct page *page);
427 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
429 if (mem_cgroup_disabled())
431 return !!(memcg->css.flags & CSS_ONLINE);
435 * For memory reclaim.
437 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
439 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
440 int zid, int nr_pages);
442 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
443 int nid, unsigned int lru_mask);
446 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
448 struct mem_cgroup_per_node *mz;
449 unsigned long nr_pages = 0;
452 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
453 for (zid = 0; zid < MAX_NR_ZONES; zid++)
454 nr_pages += mz->lru_zone_size[zid][lru];
459 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
460 enum lru_list lru, int zone_idx)
462 struct mem_cgroup_per_node *mz;
464 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
465 return mz->lru_zone_size[zone_idx][lru];
468 void mem_cgroup_handle_over_high(void);
470 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
472 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
473 struct task_struct *p);
475 static inline void mem_cgroup_oom_enable(void)
477 WARN_ON(current->memcg_may_oom);
478 current->memcg_may_oom = 1;
481 static inline void mem_cgroup_oom_disable(void)
483 WARN_ON(!current->memcg_may_oom);
484 current->memcg_may_oom = 0;
487 static inline bool task_in_memcg_oom(struct task_struct *p)
489 return p->memcg_in_oom;
492 bool mem_cgroup_oom_synchronize(bool wait);
494 #ifdef CONFIG_MEMCG_SWAP
495 extern int do_swap_account;
498 struct mem_cgroup *lock_page_memcg(struct page *page);
499 void __unlock_page_memcg(struct mem_cgroup *memcg);
500 void unlock_page_memcg(struct page *page);
502 /* idx can be of type enum memcg_stat_item or node_stat_item */
503 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
506 long x = atomic_long_read(&memcg->stat[idx]);
514 /* idx can be of type enum memcg_stat_item or node_stat_item */
515 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
520 if (mem_cgroup_disabled())
523 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
524 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
525 atomic_long_add(x, &memcg->stat[idx]);
528 __this_cpu_write(memcg->stat_cpu->count[idx], x);
531 /* idx can be of type enum memcg_stat_item or node_stat_item */
532 static inline void mod_memcg_state(struct mem_cgroup *memcg,
537 local_irq_save(flags);
538 __mod_memcg_state(memcg, idx, val);
539 local_irq_restore(flags);
543 * mod_memcg_page_state - update page state statistics
545 * @idx: page state item to account
546 * @val: number of pages (positive or negative)
548 * The @page must be locked or the caller must use lock_page_memcg()
549 * to prevent double accounting when the page is concurrently being
550 * moved to another memcg:
552 * lock_page(page) or lock_page_memcg(page)
553 * if (TestClearPageState(page))
554 * mod_memcg_page_state(page, state, -1);
555 * unlock_page(page) or unlock_page_memcg(page)
557 * Kernel pages are an exception to this, since they'll never move.
559 static inline void __mod_memcg_page_state(struct page *page,
562 if (page->mem_cgroup)
563 __mod_memcg_state(page->mem_cgroup, idx, val);
566 static inline void mod_memcg_page_state(struct page *page,
569 if (page->mem_cgroup)
570 mod_memcg_state(page->mem_cgroup, idx, val);
573 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
574 enum node_stat_item idx)
576 struct mem_cgroup_per_node *pn;
579 if (mem_cgroup_disabled())
580 return node_page_state(lruvec_pgdat(lruvec), idx);
582 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
583 x = atomic_long_read(&pn->lruvec_stat[idx]);
591 static inline void __mod_lruvec_state(struct lruvec *lruvec,
592 enum node_stat_item idx, int val)
594 struct mem_cgroup_per_node *pn;
598 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
600 if (mem_cgroup_disabled())
603 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
606 __mod_memcg_state(pn->memcg, idx, val);
609 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
610 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
611 atomic_long_add(x, &pn->lruvec_stat[idx]);
614 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
617 static inline void mod_lruvec_state(struct lruvec *lruvec,
618 enum node_stat_item idx, int val)
622 local_irq_save(flags);
623 __mod_lruvec_state(lruvec, idx, val);
624 local_irq_restore(flags);
627 static inline void __mod_lruvec_page_state(struct page *page,
628 enum node_stat_item idx, int val)
630 pg_data_t *pgdat = page_pgdat(page);
631 struct lruvec *lruvec;
633 /* Untracked pages have no memcg, no lruvec. Update only the node */
634 if (!page->mem_cgroup) {
635 __mod_node_page_state(pgdat, idx, val);
639 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
640 __mod_lruvec_state(lruvec, idx, val);
643 static inline void mod_lruvec_page_state(struct page *page,
644 enum node_stat_item idx, int val)
648 local_irq_save(flags);
649 __mod_lruvec_page_state(page, idx, val);
650 local_irq_restore(flags);
653 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
655 unsigned long *total_scanned);
657 static inline void __count_memcg_events(struct mem_cgroup *memcg,
658 enum vm_event_item idx,
663 if (mem_cgroup_disabled())
666 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
667 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
668 atomic_long_add(x, &memcg->events[idx]);
671 __this_cpu_write(memcg->stat_cpu->events[idx], x);
674 static inline void count_memcg_events(struct mem_cgroup *memcg,
675 enum vm_event_item idx,
680 local_irq_save(flags);
681 __count_memcg_events(memcg, idx, count);
682 local_irq_restore(flags);
685 static inline void count_memcg_page_event(struct page *page,
686 enum vm_event_item idx)
688 if (page->mem_cgroup)
689 count_memcg_events(page->mem_cgroup, idx, 1);
692 static inline void count_memcg_event_mm(struct mm_struct *mm,
693 enum vm_event_item idx)
695 struct mem_cgroup *memcg;
697 if (mem_cgroup_disabled())
701 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
703 count_memcg_events(memcg, idx, 1);
705 cgroup_file_notify(&memcg->events_file);
710 static inline void memcg_memory_event(struct mem_cgroup *memcg,
711 enum memcg_memory_event event)
713 atomic_long_inc(&memcg->memory_events[event]);
714 cgroup_file_notify(&memcg->events_file);
717 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
718 void mem_cgroup_split_huge_fixup(struct page *head);
721 #else /* CONFIG_MEMCG */
723 #define MEM_CGROUP_ID_SHIFT 0
724 #define MEM_CGROUP_ID_MAX 0
728 static inline bool mem_cgroup_disabled(void)
733 static inline void memcg_memory_event(struct mem_cgroup *memcg,
734 enum memcg_memory_event event)
738 static inline bool mem_cgroup_low(struct mem_cgroup *root,
739 struct mem_cgroup *memcg)
744 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
746 struct mem_cgroup **memcgp,
753 static inline void mem_cgroup_commit_charge(struct page *page,
754 struct mem_cgroup *memcg,
755 bool lrucare, bool compound)
759 static inline void mem_cgroup_cancel_charge(struct page *page,
760 struct mem_cgroup *memcg,
765 static inline void mem_cgroup_uncharge(struct page *page)
769 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
773 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
777 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
778 struct mem_cgroup *memcg)
780 return node_lruvec(pgdat);
783 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
784 struct pglist_data *pgdat)
786 return &pgdat->lruvec;
789 static inline bool mm_match_cgroup(struct mm_struct *mm,
790 struct mem_cgroup *memcg)
795 static inline bool task_in_mem_cgroup(struct task_struct *task,
796 const struct mem_cgroup *memcg)
801 static inline struct mem_cgroup *
802 mem_cgroup_iter(struct mem_cgroup *root,
803 struct mem_cgroup *prev,
804 struct mem_cgroup_reclaim_cookie *reclaim)
809 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
810 struct mem_cgroup *prev)
814 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
815 int (*fn)(struct task_struct *, void *), void *arg)
820 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
825 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
828 /* XXX: This should always return root_mem_cgroup */
832 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
837 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
842 static inline unsigned long
843 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
848 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
849 enum lru_list lru, int zone_idx)
854 static inline unsigned long
855 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
856 int nid, unsigned int lru_mask)
861 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
867 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
871 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
876 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
880 static inline void unlock_page_memcg(struct page *page)
884 static inline void mem_cgroup_handle_over_high(void)
888 static inline void mem_cgroup_oom_enable(void)
892 static inline void mem_cgroup_oom_disable(void)
896 static inline bool task_in_memcg_oom(struct task_struct *p)
901 static inline bool mem_cgroup_oom_synchronize(bool wait)
906 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
912 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
918 static inline void mod_memcg_state(struct mem_cgroup *memcg,
924 static inline void __mod_memcg_page_state(struct page *page,
930 static inline void mod_memcg_page_state(struct page *page,
936 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
937 enum node_stat_item idx)
939 return node_page_state(lruvec_pgdat(lruvec), idx);
942 static inline void __mod_lruvec_state(struct lruvec *lruvec,
943 enum node_stat_item idx, int val)
945 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
948 static inline void mod_lruvec_state(struct lruvec *lruvec,
949 enum node_stat_item idx, int val)
951 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
954 static inline void __mod_lruvec_page_state(struct page *page,
955 enum node_stat_item idx, int val)
957 __mod_node_page_state(page_pgdat(page), idx, val);
960 static inline void mod_lruvec_page_state(struct page *page,
961 enum node_stat_item idx, int val)
963 mod_node_page_state(page_pgdat(page), idx, val);
967 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
969 unsigned long *total_scanned)
974 static inline void mem_cgroup_split_huge_fixup(struct page *head)
978 static inline void count_memcg_events(struct mem_cgroup *memcg,
979 enum vm_event_item idx,
984 static inline void count_memcg_page_event(struct page *page,
990 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
993 #endif /* CONFIG_MEMCG */
995 /* idx can be of type enum memcg_stat_item or node_stat_item */
996 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
999 __mod_memcg_state(memcg, idx, 1);
1002 /* idx can be of type enum memcg_stat_item or node_stat_item */
1003 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1006 __mod_memcg_state(memcg, idx, -1);
1009 /* idx can be of type enum memcg_stat_item or node_stat_item */
1010 static inline void __inc_memcg_page_state(struct page *page,
1013 __mod_memcg_page_state(page, idx, 1);
1016 /* idx can be of type enum memcg_stat_item or node_stat_item */
1017 static inline void __dec_memcg_page_state(struct page *page,
1020 __mod_memcg_page_state(page, idx, -1);
1023 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1024 enum node_stat_item idx)
1026 __mod_lruvec_state(lruvec, idx, 1);
1029 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1030 enum node_stat_item idx)
1032 __mod_lruvec_state(lruvec, idx, -1);
1035 static inline void __inc_lruvec_page_state(struct page *page,
1036 enum node_stat_item idx)
1038 __mod_lruvec_page_state(page, idx, 1);
1041 static inline void __dec_lruvec_page_state(struct page *page,
1042 enum node_stat_item idx)
1044 __mod_lruvec_page_state(page, idx, -1);
1047 /* idx can be of type enum memcg_stat_item or node_stat_item */
1048 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1051 mod_memcg_state(memcg, idx, 1);
1054 /* idx can be of type enum memcg_stat_item or node_stat_item */
1055 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1058 mod_memcg_state(memcg, idx, -1);
1061 /* idx can be of type enum memcg_stat_item or node_stat_item */
1062 static inline void inc_memcg_page_state(struct page *page,
1065 mod_memcg_page_state(page, idx, 1);
1068 /* idx can be of type enum memcg_stat_item or node_stat_item */
1069 static inline void dec_memcg_page_state(struct page *page,
1072 mod_memcg_page_state(page, idx, -1);
1075 static inline void inc_lruvec_state(struct lruvec *lruvec,
1076 enum node_stat_item idx)
1078 mod_lruvec_state(lruvec, idx, 1);
1081 static inline void dec_lruvec_state(struct lruvec *lruvec,
1082 enum node_stat_item idx)
1084 mod_lruvec_state(lruvec, idx, -1);
1087 static inline void inc_lruvec_page_state(struct page *page,
1088 enum node_stat_item idx)
1090 mod_lruvec_page_state(page, idx, 1);
1093 static inline void dec_lruvec_page_state(struct page *page,
1094 enum node_stat_item idx)
1096 mod_lruvec_page_state(page, idx, -1);
1099 #ifdef CONFIG_CGROUP_WRITEBACK
1101 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
1102 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1103 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1104 unsigned long *pheadroom, unsigned long *pdirty,
1105 unsigned long *pwriteback);
1107 #else /* CONFIG_CGROUP_WRITEBACK */
1109 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1114 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1115 unsigned long *pfilepages,
1116 unsigned long *pheadroom,
1117 unsigned long *pdirty,
1118 unsigned long *pwriteback)
1122 #endif /* CONFIG_CGROUP_WRITEBACK */
1125 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1126 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1128 extern struct static_key_false memcg_sockets_enabled_key;
1129 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1130 void mem_cgroup_sk_alloc(struct sock *sk);
1131 void mem_cgroup_sk_free(struct sock *sk);
1132 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1134 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1137 if (time_before(jiffies, memcg->socket_pressure))
1139 } while ((memcg = parent_mem_cgroup(memcg)));
1143 #define mem_cgroup_sockets_enabled 0
1144 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1145 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1146 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1152 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1153 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1154 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1155 struct mem_cgroup *memcg);
1156 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1157 void memcg_kmem_uncharge(struct page *page, int order);
1159 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1160 extern struct static_key_false memcg_kmem_enabled_key;
1161 extern struct workqueue_struct *memcg_kmem_cache_wq;
1163 extern int memcg_nr_cache_ids;
1164 void memcg_get_cache_ids(void);
1165 void memcg_put_cache_ids(void);
1168 * Helper macro to loop through all memcg-specific caches. Callers must still
1169 * check if the cache is valid (it is either valid or NULL).
1170 * the slab_mutex must be held when looping through those caches
1172 #define for_each_memcg_cache_index(_idx) \
1173 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1175 static inline bool memcg_kmem_enabled(void)
1177 return static_branch_unlikely(&memcg_kmem_enabled_key);
1181 * helper for accessing a memcg's index. It will be used as an index in the
1182 * child cache array in kmem_cache, and also to derive its name. This function
1183 * will return -1 when this is not a kmem-limited memcg.
1185 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1187 return memcg ? memcg->kmemcg_id : -1;
1191 #define for_each_memcg_cache_index(_idx) \
1194 static inline bool memcg_kmem_enabled(void)
1199 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1204 static inline void memcg_get_cache_ids(void)
1208 static inline void memcg_put_cache_ids(void)
1212 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1214 #endif /* _LINUX_MEMCONTROL_H */