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 enum mem_cgroup_protection {
67 struct mem_cgroup_reclaim_cookie {
70 unsigned int generation;
75 #define MEM_CGROUP_ID_SHIFT 16
76 #define MEM_CGROUP_ID_MAX USHRT_MAX
78 struct mem_cgroup_id {
84 * Per memcg event counter is incremented at every pagein/pageout. With THP,
85 * it will be incremated by the number of pages. This counter is used for
86 * for trigger some periodic events. This is straightforward and better
87 * than using jiffies etc. to handle periodic memcg event.
89 enum mem_cgroup_events_target {
90 MEM_CGROUP_TARGET_THRESH,
91 MEM_CGROUP_TARGET_SOFTLIMIT,
92 MEM_CGROUP_TARGET_NUMAINFO,
96 struct mem_cgroup_stat_cpu {
97 long count[MEMCG_NR_STAT];
98 unsigned long events[NR_VM_EVENT_ITEMS];
99 unsigned long nr_page_events;
100 unsigned long targets[MEM_CGROUP_NTARGETS];
103 struct mem_cgroup_reclaim_iter {
104 struct mem_cgroup *position;
105 /* scan generation, increased every round-trip */
106 unsigned int generation;
110 long count[NR_VM_NODE_STAT_ITEMS];
114 * per-zone information in memory controller.
116 struct mem_cgroup_per_node {
117 struct lruvec lruvec;
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[DEF_PRIORITY + 1];
126 struct rb_node tree_node; /* RB tree node */
127 unsigned long usage_in_excess;/* Set to the value by which */
128 /* the soft limit is exceeded*/
130 bool congested; /* memcg has many dirty pages */
131 /* backed by a congested BDI */
133 struct mem_cgroup *memcg; /* Back pointer, we cannot */
134 /* use container_of */
137 struct mem_cgroup_threshold {
138 struct eventfd_ctx *eventfd;
139 unsigned long threshold;
143 struct mem_cgroup_threshold_ary {
144 /* An array index points to threshold just below or equal to usage. */
145 int current_threshold;
146 /* Size of entries[] */
148 /* Array of thresholds */
149 struct mem_cgroup_threshold entries[0];
152 struct mem_cgroup_thresholds {
153 /* Primary thresholds array */
154 struct mem_cgroup_threshold_ary *primary;
156 * Spare threshold array.
157 * This is needed to make mem_cgroup_unregister_event() "never fail".
158 * It must be able to store at least primary->size - 1 entries.
160 struct mem_cgroup_threshold_ary *spare;
163 enum memcg_kmem_state {
169 #if defined(CONFIG_SMP)
170 struct memcg_padding {
172 } ____cacheline_internodealigned_in_smp;
173 #define MEMCG_PADDING(name) struct memcg_padding name;
175 #define MEMCG_PADDING(name)
179 * The memory controller data structure. The memory controller controls both
180 * page cache and RSS per cgroup. We would eventually like to provide
181 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
182 * to help the administrator determine what knobs to tune.
185 struct cgroup_subsys_state css;
187 /* Private memcg ID. Used to ID objects that outlive the cgroup */
188 struct mem_cgroup_id id;
190 /* Accounted resources */
191 struct page_counter memory;
192 struct page_counter swap;
194 /* Legacy consumer-oriented counters */
195 struct page_counter memsw;
196 struct page_counter kmem;
197 struct page_counter tcpmem;
199 /* Upper bound of normal memory consumption range */
202 /* Range enforcement for interrupt charges */
203 struct work_struct high_work;
205 unsigned long soft_limit;
207 /* vmpressure notifications */
208 struct vmpressure vmpressure;
211 * Should the accounting and control be hierarchical, per subtree?
215 /* protected by memcg_oom_lock */
220 /* OOM-Killer disable */
221 int oom_kill_disable;
224 struct cgroup_file events_file;
226 /* handle for "memory.swap.events" */
227 struct cgroup_file swap_events_file;
229 /* protect arrays of thresholds */
230 struct mutex thresholds_lock;
232 /* thresholds for memory usage. RCU-protected */
233 struct mem_cgroup_thresholds thresholds;
235 /* thresholds for mem+swap usage. RCU-protected */
236 struct mem_cgroup_thresholds memsw_thresholds;
238 /* For oom notifier event fd */
239 struct list_head oom_notify;
242 * Should we move charges of a task when a task is moved into this
243 * mem_cgroup ? And what type of charges should we move ?
245 unsigned long move_charge_at_immigrate;
246 /* taken only while moving_account > 0 */
247 spinlock_t move_lock;
248 unsigned long move_lock_flags;
250 MEMCG_PADDING(_pad1_);
253 * set > 0 if pages under this cgroup are moving to other cgroup.
255 atomic_t moving_account;
256 struct task_struct *move_lock_task;
259 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
261 MEMCG_PADDING(_pad2_);
263 atomic_long_t stat[MEMCG_NR_STAT];
264 atomic_long_t events[NR_VM_EVENT_ITEMS];
265 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
267 unsigned long socket_pressure;
269 /* Legacy tcp memory accounting */
274 /* Index in the kmem_cache->memcg_params.memcg_caches array */
276 enum memcg_kmem_state kmem_state;
277 struct list_head kmem_caches;
280 int last_scanned_node;
282 nodemask_t scan_nodes;
283 atomic_t numainfo_events;
284 atomic_t numainfo_updating;
287 #ifdef CONFIG_CGROUP_WRITEBACK
288 struct list_head cgwb_list;
289 struct wb_domain cgwb_domain;
292 /* List of events which userspace want to receive */
293 struct list_head event_list;
294 spinlock_t event_list_lock;
296 struct mem_cgroup_per_node *nodeinfo[0];
297 /* WARNING: nodeinfo must be the last member here */
301 * size of first charge trial. "32" comes from vmscan.c's magic value.
302 * TODO: maybe necessary to use big numbers in big irons.
304 #define MEMCG_CHARGE_BATCH 32U
306 extern struct mem_cgroup *root_mem_cgroup;
308 static inline bool mem_cgroup_disabled(void)
310 return !cgroup_subsys_enabled(memory_cgrp_subsys);
313 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
314 struct mem_cgroup *memcg);
316 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
317 gfp_t gfp_mask, struct mem_cgroup **memcgp,
319 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
320 bool lrucare, bool compound);
321 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
323 void mem_cgroup_uncharge(struct page *page);
324 void mem_cgroup_uncharge_list(struct list_head *page_list);
326 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
328 static struct mem_cgroup_per_node *
329 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
331 return memcg->nodeinfo[nid];
335 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
336 * @node: node of the wanted lruvec
337 * @memcg: memcg of the wanted lruvec
339 * Returns the lru list vector holding pages for a given @node or a given
340 * @memcg and @zone. This can be the node lruvec, if the memory controller
343 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
344 struct mem_cgroup *memcg)
346 struct mem_cgroup_per_node *mz;
347 struct lruvec *lruvec;
349 if (mem_cgroup_disabled()) {
350 lruvec = node_lruvec(pgdat);
354 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
355 lruvec = &mz->lruvec;
358 * Since a node can be onlined after the mem_cgroup was created,
359 * we have to be prepared to initialize lruvec->pgdat here;
360 * and if offlined then reonlined, we need to reinitialize it.
362 if (unlikely(lruvec->pgdat != pgdat))
363 lruvec->pgdat = pgdat;
367 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
369 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
370 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
373 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
374 return css ? container_of(css, struct mem_cgroup, css) : NULL;
377 #define mem_cgroup_from_counter(counter, member) \
378 container_of(counter, struct mem_cgroup, member)
380 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
382 struct mem_cgroup_reclaim_cookie *);
383 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
384 int mem_cgroup_scan_tasks(struct mem_cgroup *,
385 int (*)(struct task_struct *, void *), void *);
387 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
389 if (mem_cgroup_disabled())
394 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
396 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
398 struct mem_cgroup_per_node *mz;
400 if (mem_cgroup_disabled())
403 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
408 * parent_mem_cgroup - find the accounting parent of a memcg
409 * @memcg: memcg whose parent to find
411 * Returns the parent memcg, or NULL if this is the root or the memory
412 * controller is in legacy no-hierarchy mode.
414 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
416 if (!memcg->memory.parent)
418 return mem_cgroup_from_counter(memcg->memory.parent, memory);
421 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
422 struct mem_cgroup *root)
426 if (!root->use_hierarchy)
428 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
431 static inline bool mm_match_cgroup(struct mm_struct *mm,
432 struct mem_cgroup *memcg)
434 struct mem_cgroup *task_memcg;
438 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
440 match = mem_cgroup_is_descendant(task_memcg, memcg);
445 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
446 ino_t page_cgroup_ino(struct page *page);
448 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
450 if (mem_cgroup_disabled())
452 return !!(memcg->css.flags & CSS_ONLINE);
456 * For memory reclaim.
458 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
460 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
461 int zid, int nr_pages);
463 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
464 int nid, unsigned int lru_mask);
467 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
469 struct mem_cgroup_per_node *mz;
470 unsigned long nr_pages = 0;
473 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
474 for (zid = 0; zid < MAX_NR_ZONES; zid++)
475 nr_pages += mz->lru_zone_size[zid][lru];
480 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
481 enum lru_list lru, int zone_idx)
483 struct mem_cgroup_per_node *mz;
485 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
486 return mz->lru_zone_size[zone_idx][lru];
489 void mem_cgroup_handle_over_high(void);
491 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
493 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
494 struct task_struct *p);
496 static inline void mem_cgroup_oom_enable(void)
498 WARN_ON(current->memcg_may_oom);
499 current->memcg_may_oom = 1;
502 static inline void mem_cgroup_oom_disable(void)
504 WARN_ON(!current->memcg_may_oom);
505 current->memcg_may_oom = 0;
508 static inline bool task_in_memcg_oom(struct task_struct *p)
510 return p->memcg_in_oom;
513 bool mem_cgroup_oom_synchronize(bool wait);
515 #ifdef CONFIG_MEMCG_SWAP
516 extern int do_swap_account;
519 struct mem_cgroup *lock_page_memcg(struct page *page);
520 void __unlock_page_memcg(struct mem_cgroup *memcg);
521 void unlock_page_memcg(struct page *page);
523 /* idx can be of type enum memcg_stat_item or node_stat_item */
524 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
527 long x = atomic_long_read(&memcg->stat[idx]);
535 /* idx can be of type enum memcg_stat_item or node_stat_item */
536 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
541 if (mem_cgroup_disabled())
544 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
545 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
546 atomic_long_add(x, &memcg->stat[idx]);
549 __this_cpu_write(memcg->stat_cpu->count[idx], x);
552 /* idx can be of type enum memcg_stat_item or node_stat_item */
553 static inline void mod_memcg_state(struct mem_cgroup *memcg,
558 local_irq_save(flags);
559 __mod_memcg_state(memcg, idx, val);
560 local_irq_restore(flags);
564 * mod_memcg_page_state - update page state statistics
566 * @idx: page state item to account
567 * @val: number of pages (positive or negative)
569 * The @page must be locked or the caller must use lock_page_memcg()
570 * to prevent double accounting when the page is concurrently being
571 * moved to another memcg:
573 * lock_page(page) or lock_page_memcg(page)
574 * if (TestClearPageState(page))
575 * mod_memcg_page_state(page, state, -1);
576 * unlock_page(page) or unlock_page_memcg(page)
578 * Kernel pages are an exception to this, since they'll never move.
580 static inline void __mod_memcg_page_state(struct page *page,
583 if (page->mem_cgroup)
584 __mod_memcg_state(page->mem_cgroup, idx, val);
587 static inline void mod_memcg_page_state(struct page *page,
590 if (page->mem_cgroup)
591 mod_memcg_state(page->mem_cgroup, idx, val);
594 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
595 enum node_stat_item idx)
597 struct mem_cgroup_per_node *pn;
600 if (mem_cgroup_disabled())
601 return node_page_state(lruvec_pgdat(lruvec), idx);
603 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
604 x = atomic_long_read(&pn->lruvec_stat[idx]);
612 static inline void __mod_lruvec_state(struct lruvec *lruvec,
613 enum node_stat_item idx, int val)
615 struct mem_cgroup_per_node *pn;
619 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
621 if (mem_cgroup_disabled())
624 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
627 __mod_memcg_state(pn->memcg, idx, val);
630 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
631 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
632 atomic_long_add(x, &pn->lruvec_stat[idx]);
635 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
638 static inline void mod_lruvec_state(struct lruvec *lruvec,
639 enum node_stat_item idx, int val)
643 local_irq_save(flags);
644 __mod_lruvec_state(lruvec, idx, val);
645 local_irq_restore(flags);
648 static inline void __mod_lruvec_page_state(struct page *page,
649 enum node_stat_item idx, int val)
651 pg_data_t *pgdat = page_pgdat(page);
652 struct lruvec *lruvec;
654 /* Untracked pages have no memcg, no lruvec. Update only the node */
655 if (!page->mem_cgroup) {
656 __mod_node_page_state(pgdat, idx, val);
660 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
661 __mod_lruvec_state(lruvec, idx, val);
664 static inline void mod_lruvec_page_state(struct page *page,
665 enum node_stat_item idx, int val)
669 local_irq_save(flags);
670 __mod_lruvec_page_state(page, idx, val);
671 local_irq_restore(flags);
674 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
676 unsigned long *total_scanned);
678 static inline void __count_memcg_events(struct mem_cgroup *memcg,
679 enum vm_event_item idx,
684 if (mem_cgroup_disabled())
687 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
688 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
689 atomic_long_add(x, &memcg->events[idx]);
692 __this_cpu_write(memcg->stat_cpu->events[idx], x);
695 static inline void count_memcg_events(struct mem_cgroup *memcg,
696 enum vm_event_item idx,
701 local_irq_save(flags);
702 __count_memcg_events(memcg, idx, count);
703 local_irq_restore(flags);
706 static inline void count_memcg_page_event(struct page *page,
707 enum vm_event_item idx)
709 if (page->mem_cgroup)
710 count_memcg_events(page->mem_cgroup, idx, 1);
713 static inline void count_memcg_event_mm(struct mm_struct *mm,
714 enum vm_event_item idx)
716 struct mem_cgroup *memcg;
718 if (mem_cgroup_disabled())
722 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
724 count_memcg_events(memcg, idx, 1);
726 cgroup_file_notify(&memcg->events_file);
731 static inline void memcg_memory_event(struct mem_cgroup *memcg,
732 enum memcg_memory_event event)
734 atomic_long_inc(&memcg->memory_events[event]);
735 cgroup_file_notify(&memcg->events_file);
738 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
739 void mem_cgroup_split_huge_fixup(struct page *head);
742 #else /* CONFIG_MEMCG */
744 #define MEM_CGROUP_ID_SHIFT 0
745 #define MEM_CGROUP_ID_MAX 0
749 static inline bool mem_cgroup_disabled(void)
754 static inline void memcg_memory_event(struct mem_cgroup *memcg,
755 enum memcg_memory_event event)
759 static inline enum mem_cgroup_protection mem_cgroup_protected(
760 struct mem_cgroup *root, struct mem_cgroup *memcg)
762 return MEMCG_PROT_NONE;
765 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
767 struct mem_cgroup **memcgp,
774 static inline void mem_cgroup_commit_charge(struct page *page,
775 struct mem_cgroup *memcg,
776 bool lrucare, bool compound)
780 static inline void mem_cgroup_cancel_charge(struct page *page,
781 struct mem_cgroup *memcg,
786 static inline void mem_cgroup_uncharge(struct page *page)
790 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
794 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
798 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
799 struct mem_cgroup *memcg)
801 return node_lruvec(pgdat);
804 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
805 struct pglist_data *pgdat)
807 return &pgdat->lruvec;
810 static inline bool mm_match_cgroup(struct mm_struct *mm,
811 struct mem_cgroup *memcg)
816 static inline bool task_in_mem_cgroup(struct task_struct *task,
817 const struct mem_cgroup *memcg)
822 static inline struct mem_cgroup *
823 mem_cgroup_iter(struct mem_cgroup *root,
824 struct mem_cgroup *prev,
825 struct mem_cgroup_reclaim_cookie *reclaim)
830 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
831 struct mem_cgroup *prev)
835 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
836 int (*fn)(struct task_struct *, void *), void *arg)
841 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
846 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
849 /* XXX: This should always return root_mem_cgroup */
853 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
858 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
863 static inline unsigned long
864 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
869 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
870 enum lru_list lru, int zone_idx)
875 static inline unsigned long
876 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
877 int nid, unsigned int lru_mask)
882 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
888 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
892 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
897 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
901 static inline void unlock_page_memcg(struct page *page)
905 static inline void mem_cgroup_handle_over_high(void)
909 static inline void mem_cgroup_oom_enable(void)
913 static inline void mem_cgroup_oom_disable(void)
917 static inline bool task_in_memcg_oom(struct task_struct *p)
922 static inline bool mem_cgroup_oom_synchronize(bool wait)
927 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
933 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
939 static inline void mod_memcg_state(struct mem_cgroup *memcg,
945 static inline void __mod_memcg_page_state(struct page *page,
951 static inline void mod_memcg_page_state(struct page *page,
957 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
958 enum node_stat_item idx)
960 return node_page_state(lruvec_pgdat(lruvec), idx);
963 static inline void __mod_lruvec_state(struct lruvec *lruvec,
964 enum node_stat_item idx, int val)
966 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
969 static inline void mod_lruvec_state(struct lruvec *lruvec,
970 enum node_stat_item idx, int val)
972 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
975 static inline void __mod_lruvec_page_state(struct page *page,
976 enum node_stat_item idx, int val)
978 __mod_node_page_state(page_pgdat(page), idx, val);
981 static inline void mod_lruvec_page_state(struct page *page,
982 enum node_stat_item idx, int val)
984 mod_node_page_state(page_pgdat(page), idx, val);
988 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
990 unsigned long *total_scanned)
995 static inline void mem_cgroup_split_huge_fixup(struct page *head)
999 static inline void count_memcg_events(struct mem_cgroup *memcg,
1000 enum vm_event_item idx,
1001 unsigned long count)
1005 static inline void count_memcg_page_event(struct page *page,
1011 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1014 #endif /* CONFIG_MEMCG */
1016 /* idx can be of type enum memcg_stat_item or node_stat_item */
1017 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1020 __mod_memcg_state(memcg, idx, 1);
1023 /* idx can be of type enum memcg_stat_item or node_stat_item */
1024 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1027 __mod_memcg_state(memcg, idx, -1);
1030 /* idx can be of type enum memcg_stat_item or node_stat_item */
1031 static inline void __inc_memcg_page_state(struct page *page,
1034 __mod_memcg_page_state(page, idx, 1);
1037 /* idx can be of type enum memcg_stat_item or node_stat_item */
1038 static inline void __dec_memcg_page_state(struct page *page,
1041 __mod_memcg_page_state(page, idx, -1);
1044 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1045 enum node_stat_item idx)
1047 __mod_lruvec_state(lruvec, idx, 1);
1050 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1051 enum node_stat_item idx)
1053 __mod_lruvec_state(lruvec, idx, -1);
1056 static inline void __inc_lruvec_page_state(struct page *page,
1057 enum node_stat_item idx)
1059 __mod_lruvec_page_state(page, idx, 1);
1062 static inline void __dec_lruvec_page_state(struct page *page,
1063 enum node_stat_item idx)
1065 __mod_lruvec_page_state(page, idx, -1);
1068 /* idx can be of type enum memcg_stat_item or node_stat_item */
1069 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1072 mod_memcg_state(memcg, idx, 1);
1075 /* idx can be of type enum memcg_stat_item or node_stat_item */
1076 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1079 mod_memcg_state(memcg, idx, -1);
1082 /* idx can be of type enum memcg_stat_item or node_stat_item */
1083 static inline void inc_memcg_page_state(struct page *page,
1086 mod_memcg_page_state(page, idx, 1);
1089 /* idx can be of type enum memcg_stat_item or node_stat_item */
1090 static inline void dec_memcg_page_state(struct page *page,
1093 mod_memcg_page_state(page, idx, -1);
1096 static inline void inc_lruvec_state(struct lruvec *lruvec,
1097 enum node_stat_item idx)
1099 mod_lruvec_state(lruvec, idx, 1);
1102 static inline void dec_lruvec_state(struct lruvec *lruvec,
1103 enum node_stat_item idx)
1105 mod_lruvec_state(lruvec, idx, -1);
1108 static inline void inc_lruvec_page_state(struct page *page,
1109 enum node_stat_item idx)
1111 mod_lruvec_page_state(page, idx, 1);
1114 static inline void dec_lruvec_page_state(struct page *page,
1115 enum node_stat_item idx)
1117 mod_lruvec_page_state(page, idx, -1);
1120 #ifdef CONFIG_CGROUP_WRITEBACK
1122 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1123 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1124 unsigned long *pheadroom, unsigned long *pdirty,
1125 unsigned long *pwriteback);
1127 #else /* CONFIG_CGROUP_WRITEBACK */
1129 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1134 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1135 unsigned long *pfilepages,
1136 unsigned long *pheadroom,
1137 unsigned long *pdirty,
1138 unsigned long *pwriteback)
1142 #endif /* CONFIG_CGROUP_WRITEBACK */
1145 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1146 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1148 extern struct static_key_false memcg_sockets_enabled_key;
1149 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1150 void mem_cgroup_sk_alloc(struct sock *sk);
1151 void mem_cgroup_sk_free(struct sock *sk);
1152 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1154 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1157 if (time_before(jiffies, memcg->socket_pressure))
1159 } while ((memcg = parent_mem_cgroup(memcg)));
1163 #define mem_cgroup_sockets_enabled 0
1164 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1165 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1166 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1172 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1173 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1174 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1175 struct mem_cgroup *memcg);
1176 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1177 void memcg_kmem_uncharge(struct page *page, int order);
1179 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1180 extern struct static_key_false memcg_kmem_enabled_key;
1181 extern struct workqueue_struct *memcg_kmem_cache_wq;
1183 extern int memcg_nr_cache_ids;
1184 void memcg_get_cache_ids(void);
1185 void memcg_put_cache_ids(void);
1188 * Helper macro to loop through all memcg-specific caches. Callers must still
1189 * check if the cache is valid (it is either valid or NULL).
1190 * the slab_mutex must be held when looping through those caches
1192 #define for_each_memcg_cache_index(_idx) \
1193 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1195 static inline bool memcg_kmem_enabled(void)
1197 return static_branch_unlikely(&memcg_kmem_enabled_key);
1201 * helper for accessing a memcg's index. It will be used as an index in the
1202 * child cache array in kmem_cache, and also to derive its name. This function
1203 * will return -1 when this is not a kmem-limited memcg.
1205 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1207 return memcg ? memcg->kmemcg_id : -1;
1211 #define for_each_memcg_cache_index(_idx) \
1214 static inline bool memcg_kmem_enabled(void)
1219 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1224 static inline void memcg_get_cache_ids(void)
1228 static inline void memcg_put_cache_ids(void)
1232 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1234 #endif /* _LINUX_MEMCONTROL_H */