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 {
57 MEMCG_NR_MEMORY_EVENTS,
60 struct mem_cgroup_reclaim_cookie {
63 unsigned int generation;
68 #define MEM_CGROUP_ID_SHIFT 16
69 #define MEM_CGROUP_ID_MAX USHRT_MAX
71 struct mem_cgroup_id {
77 * Per memcg event counter is incremented at every pagein/pageout. With THP,
78 * it will be incremated by the number of pages. This counter is used for
79 * for trigger some periodic events. This is straightforward and better
80 * than using jiffies etc. to handle periodic memcg event.
82 enum mem_cgroup_events_target {
83 MEM_CGROUP_TARGET_THRESH,
84 MEM_CGROUP_TARGET_SOFTLIMIT,
85 MEM_CGROUP_TARGET_NUMAINFO,
89 struct mem_cgroup_stat_cpu {
90 long count[MEMCG_NR_STAT];
91 unsigned long events[NR_VM_EVENT_ITEMS];
92 unsigned long nr_page_events;
93 unsigned long targets[MEM_CGROUP_NTARGETS];
96 struct mem_cgroup_reclaim_iter {
97 struct mem_cgroup *position;
98 /* scan generation, increased every round-trip */
99 unsigned int generation;
103 long count[NR_VM_NODE_STAT_ITEMS];
107 * per-zone information in memory controller.
109 struct mem_cgroup_per_node {
110 struct lruvec lruvec;
112 struct lruvec_stat __percpu *lruvec_stat_cpu;
113 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
115 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
117 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
119 struct rb_node tree_node; /* RB tree node */
120 unsigned long usage_in_excess;/* Set to the value by which */
121 /* the soft limit is exceeded*/
123 struct mem_cgroup *memcg; /* Back pointer, we cannot */
124 /* use container_of */
127 struct mem_cgroup_threshold {
128 struct eventfd_ctx *eventfd;
129 unsigned long threshold;
133 struct mem_cgroup_threshold_ary {
134 /* An array index points to threshold just below or equal to usage. */
135 int current_threshold;
136 /* Size of entries[] */
138 /* Array of thresholds */
139 struct mem_cgroup_threshold entries[0];
142 struct mem_cgroup_thresholds {
143 /* Primary thresholds array */
144 struct mem_cgroup_threshold_ary *primary;
146 * Spare threshold array.
147 * This is needed to make mem_cgroup_unregister_event() "never fail".
148 * It must be able to store at least primary->size - 1 entries.
150 struct mem_cgroup_threshold_ary *spare;
153 enum memcg_kmem_state {
159 #if defined(CONFIG_SMP)
160 struct memcg_padding {
162 } ____cacheline_internodealigned_in_smp;
163 #define MEMCG_PADDING(name) struct memcg_padding name;
165 #define MEMCG_PADDING(name)
169 * The memory controller data structure. The memory controller controls both
170 * page cache and RSS per cgroup. We would eventually like to provide
171 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
172 * to help the administrator determine what knobs to tune.
175 struct cgroup_subsys_state css;
177 /* Private memcg ID. Used to ID objects that outlive the cgroup */
178 struct mem_cgroup_id id;
180 /* Accounted resources */
181 struct page_counter memory;
182 struct page_counter swap;
184 /* Legacy consumer-oriented counters */
185 struct page_counter memsw;
186 struct page_counter kmem;
187 struct page_counter tcpmem;
189 /* Normal memory consumption range */
193 /* Range enforcement for interrupt charges */
194 struct work_struct high_work;
196 unsigned long soft_limit;
198 /* vmpressure notifications */
199 struct vmpressure vmpressure;
202 * Should the accounting and control be hierarchical, per subtree?
206 /* protected by memcg_oom_lock */
211 /* OOM-Killer disable */
212 int oom_kill_disable;
215 struct cgroup_file events_file;
217 /* protect arrays of thresholds */
218 struct mutex thresholds_lock;
220 /* thresholds for memory usage. RCU-protected */
221 struct mem_cgroup_thresholds thresholds;
223 /* thresholds for mem+swap usage. RCU-protected */
224 struct mem_cgroup_thresholds memsw_thresholds;
226 /* For oom notifier event fd */
227 struct list_head oom_notify;
230 * Should we move charges of a task when a task is moved into this
231 * mem_cgroup ? And what type of charges should we move ?
233 unsigned long move_charge_at_immigrate;
234 /* taken only while moving_account > 0 */
235 spinlock_t move_lock;
236 unsigned long move_lock_flags;
238 MEMCG_PADDING(_pad1_);
241 * set > 0 if pages under this cgroup are moving to other cgroup.
243 atomic_t moving_account;
244 struct task_struct *move_lock_task;
247 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
249 MEMCG_PADDING(_pad2_);
251 atomic_long_t stat[MEMCG_NR_STAT];
252 atomic_long_t events[NR_VM_EVENT_ITEMS];
253 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
255 unsigned long socket_pressure;
257 /* Legacy tcp memory accounting */
262 /* Index in the kmem_cache->memcg_params.memcg_caches array */
264 enum memcg_kmem_state kmem_state;
265 struct list_head kmem_caches;
268 int last_scanned_node;
270 nodemask_t scan_nodes;
271 atomic_t numainfo_events;
272 atomic_t numainfo_updating;
275 #ifdef CONFIG_CGROUP_WRITEBACK
276 struct list_head cgwb_list;
277 struct wb_domain cgwb_domain;
280 /* List of events which userspace want to receive */
281 struct list_head event_list;
282 spinlock_t event_list_lock;
284 struct mem_cgroup_per_node *nodeinfo[0];
285 /* WARNING: nodeinfo must be the last member here */
289 * size of first charge trial. "32" comes from vmscan.c's magic value.
290 * TODO: maybe necessary to use big numbers in big irons.
292 #define MEMCG_CHARGE_BATCH 32U
294 extern struct mem_cgroup *root_mem_cgroup;
296 static inline bool mem_cgroup_disabled(void)
298 return !cgroup_subsys_enabled(memory_cgrp_subsys);
301 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
303 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
304 gfp_t gfp_mask, struct mem_cgroup **memcgp,
306 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
307 bool lrucare, bool compound);
308 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
310 void mem_cgroup_uncharge(struct page *page);
311 void mem_cgroup_uncharge_list(struct list_head *page_list);
313 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
315 static struct mem_cgroup_per_node *
316 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
318 return memcg->nodeinfo[nid];
322 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
323 * @node: node of the wanted lruvec
324 * @memcg: memcg of the wanted lruvec
326 * Returns the lru list vector holding pages for a given @node or a given
327 * @memcg and @zone. This can be the node lruvec, if the memory controller
330 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
331 struct mem_cgroup *memcg)
333 struct mem_cgroup_per_node *mz;
334 struct lruvec *lruvec;
336 if (mem_cgroup_disabled()) {
337 lruvec = node_lruvec(pgdat);
341 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
342 lruvec = &mz->lruvec;
345 * Since a node can be onlined after the mem_cgroup was created,
346 * we have to be prepared to initialize lruvec->pgdat here;
347 * and if offlined then reonlined, we need to reinitialize it.
349 if (unlikely(lruvec->pgdat != pgdat))
350 lruvec->pgdat = pgdat;
354 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
356 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
357 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
360 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
361 return css ? container_of(css, struct mem_cgroup, css) : NULL;
364 #define mem_cgroup_from_counter(counter, member) \
365 container_of(counter, struct mem_cgroup, member)
367 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
369 struct mem_cgroup_reclaim_cookie *);
370 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
371 int mem_cgroup_scan_tasks(struct mem_cgroup *,
372 int (*)(struct task_struct *, void *), void *);
374 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
376 if (mem_cgroup_disabled())
381 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
383 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
385 struct mem_cgroup_per_node *mz;
387 if (mem_cgroup_disabled())
390 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
395 * parent_mem_cgroup - find the accounting parent of a memcg
396 * @memcg: memcg whose parent to find
398 * Returns the parent memcg, or NULL if this is the root or the memory
399 * controller is in legacy no-hierarchy mode.
401 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
403 if (!memcg->memory.parent)
405 return mem_cgroup_from_counter(memcg->memory.parent, memory);
408 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
409 struct mem_cgroup *root)
413 if (!root->use_hierarchy)
415 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
418 static inline bool mm_match_cgroup(struct mm_struct *mm,
419 struct mem_cgroup *memcg)
421 struct mem_cgroup *task_memcg;
425 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
427 match = mem_cgroup_is_descendant(task_memcg, memcg);
432 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
433 ino_t page_cgroup_ino(struct page *page);
435 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
437 if (mem_cgroup_disabled())
439 return !!(memcg->css.flags & CSS_ONLINE);
443 * For memory reclaim.
445 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
447 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
448 int zid, int nr_pages);
450 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
451 int nid, unsigned int lru_mask);
454 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
456 struct mem_cgroup_per_node *mz;
457 unsigned long nr_pages = 0;
460 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
461 for (zid = 0; zid < MAX_NR_ZONES; zid++)
462 nr_pages += mz->lru_zone_size[zid][lru];
467 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
468 enum lru_list lru, int zone_idx)
470 struct mem_cgroup_per_node *mz;
472 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
473 return mz->lru_zone_size[zone_idx][lru];
476 void mem_cgroup_handle_over_high(void);
478 unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg);
480 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
481 struct task_struct *p);
483 static inline void mem_cgroup_oom_enable(void)
485 WARN_ON(current->memcg_may_oom);
486 current->memcg_may_oom = 1;
489 static inline void mem_cgroup_oom_disable(void)
491 WARN_ON(!current->memcg_may_oom);
492 current->memcg_may_oom = 0;
495 static inline bool task_in_memcg_oom(struct task_struct *p)
497 return p->memcg_in_oom;
500 bool mem_cgroup_oom_synchronize(bool wait);
502 #ifdef CONFIG_MEMCG_SWAP
503 extern int do_swap_account;
506 struct mem_cgroup *lock_page_memcg(struct page *page);
507 void __unlock_page_memcg(struct mem_cgroup *memcg);
508 void unlock_page_memcg(struct page *page);
511 * idx can be of type enum memcg_stat_item or node_stat_item.
512 * Keep in sync with memcg_exact_page_state().
514 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
517 long x = atomic_long_read(&memcg->stat[idx]);
525 /* idx can be of type enum memcg_stat_item or node_stat_item */
526 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
531 if (mem_cgroup_disabled())
534 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
535 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
536 atomic_long_add(x, &memcg->stat[idx]);
539 __this_cpu_write(memcg->stat_cpu->count[idx], x);
542 /* idx can be of type enum memcg_stat_item or node_stat_item */
543 static inline void mod_memcg_state(struct mem_cgroup *memcg,
548 local_irq_save(flags);
549 __mod_memcg_state(memcg, idx, val);
550 local_irq_restore(flags);
554 * mod_memcg_page_state - update page state statistics
556 * @idx: page state item to account
557 * @val: number of pages (positive or negative)
559 * The @page must be locked or the caller must use lock_page_memcg()
560 * to prevent double accounting when the page is concurrently being
561 * moved to another memcg:
563 * lock_page(page) or lock_page_memcg(page)
564 * if (TestClearPageState(page))
565 * mod_memcg_page_state(page, state, -1);
566 * unlock_page(page) or unlock_page_memcg(page)
568 * Kernel pages are an exception to this, since they'll never move.
570 static inline void __mod_memcg_page_state(struct page *page,
573 if (page->mem_cgroup)
574 __mod_memcg_state(page->mem_cgroup, idx, val);
577 static inline void mod_memcg_page_state(struct page *page,
580 if (page->mem_cgroup)
581 mod_memcg_state(page->mem_cgroup, idx, val);
584 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
585 enum node_stat_item idx)
587 struct mem_cgroup_per_node *pn;
590 if (mem_cgroup_disabled())
591 return node_page_state(lruvec_pgdat(lruvec), idx);
593 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
594 x = atomic_long_read(&pn->lruvec_stat[idx]);
602 static inline void __mod_lruvec_state(struct lruvec *lruvec,
603 enum node_stat_item idx, int val)
605 struct mem_cgroup_per_node *pn;
609 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
611 if (mem_cgroup_disabled())
614 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
617 __mod_memcg_state(pn->memcg, idx, val);
620 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
621 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
622 atomic_long_add(x, &pn->lruvec_stat[idx]);
625 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
628 static inline void mod_lruvec_state(struct lruvec *lruvec,
629 enum node_stat_item idx, int val)
633 local_irq_save(flags);
634 __mod_lruvec_state(lruvec, idx, val);
635 local_irq_restore(flags);
638 static inline void __mod_lruvec_page_state(struct page *page,
639 enum node_stat_item idx, int val)
641 pg_data_t *pgdat = page_pgdat(page);
642 struct lruvec *lruvec;
644 /* Untracked pages have no memcg, no lruvec. Update only the node */
645 if (!page->mem_cgroup) {
646 __mod_node_page_state(pgdat, idx, val);
650 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
651 __mod_lruvec_state(lruvec, idx, val);
654 static inline void mod_lruvec_page_state(struct page *page,
655 enum node_stat_item idx, int val)
659 local_irq_save(flags);
660 __mod_lruvec_page_state(page, idx, val);
661 local_irq_restore(flags);
664 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
666 unsigned long *total_scanned);
668 static inline void __count_memcg_events(struct mem_cgroup *memcg,
669 enum vm_event_item idx,
674 if (mem_cgroup_disabled())
677 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
678 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
679 atomic_long_add(x, &memcg->events[idx]);
682 __this_cpu_write(memcg->stat_cpu->events[idx], x);
685 static inline void count_memcg_events(struct mem_cgroup *memcg,
686 enum vm_event_item idx,
691 local_irq_save(flags);
692 __count_memcg_events(memcg, idx, count);
693 local_irq_restore(flags);
696 static inline void count_memcg_page_event(struct page *page,
697 enum vm_event_item idx)
699 if (page->mem_cgroup)
700 count_memcg_events(page->mem_cgroup, idx, 1);
703 static inline void count_memcg_event_mm(struct mm_struct *mm,
704 enum vm_event_item idx)
706 struct mem_cgroup *memcg;
708 if (mem_cgroup_disabled())
712 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
714 count_memcg_events(memcg, idx, 1);
718 static inline void memcg_memory_event(struct mem_cgroup *memcg,
719 enum memcg_memory_event event)
721 atomic_long_inc(&memcg->memory_events[event]);
722 cgroup_file_notify(&memcg->events_file);
725 static inline void memcg_memory_event_mm(struct mm_struct *mm,
726 enum memcg_memory_event event)
728 struct mem_cgroup *memcg;
730 if (mem_cgroup_disabled())
734 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
736 memcg_memory_event(memcg, event);
740 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
741 void mem_cgroup_split_huge_fixup(struct page *head);
744 #else /* CONFIG_MEMCG */
746 #define MEM_CGROUP_ID_SHIFT 0
747 #define MEM_CGROUP_ID_MAX 0
751 static inline bool mem_cgroup_disabled(void)
756 static inline void memcg_memory_event(struct mem_cgroup *memcg,
757 enum memcg_memory_event event)
761 static inline void memcg_memory_event_mm(struct mm_struct *mm,
762 enum memcg_memory_event event)
766 static inline bool mem_cgroup_low(struct mem_cgroup *root,
767 struct mem_cgroup *memcg)
772 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
774 struct mem_cgroup **memcgp,
781 static inline void mem_cgroup_commit_charge(struct page *page,
782 struct mem_cgroup *memcg,
783 bool lrucare, bool compound)
787 static inline void mem_cgroup_cancel_charge(struct page *page,
788 struct mem_cgroup *memcg,
793 static inline void mem_cgroup_uncharge(struct page *page)
797 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
801 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
805 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
806 struct mem_cgroup *memcg)
808 return node_lruvec(pgdat);
811 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
812 struct pglist_data *pgdat)
814 return &pgdat->lruvec;
817 static inline bool mm_match_cgroup(struct mm_struct *mm,
818 struct mem_cgroup *memcg)
823 static inline bool task_in_mem_cgroup(struct task_struct *task,
824 const struct mem_cgroup *memcg)
829 static inline struct mem_cgroup *
830 mem_cgroup_iter(struct mem_cgroup *root,
831 struct mem_cgroup *prev,
832 struct mem_cgroup_reclaim_cookie *reclaim)
837 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
838 struct mem_cgroup *prev)
842 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
843 int (*fn)(struct task_struct *, void *), void *arg)
848 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
853 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
856 /* XXX: This should always return root_mem_cgroup */
860 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
865 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
870 static inline unsigned long
871 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
876 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
877 enum lru_list lru, int zone_idx)
882 static inline unsigned long
883 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
884 int nid, unsigned int lru_mask)
889 static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
895 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
899 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
904 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
908 static inline void unlock_page_memcg(struct page *page)
912 static inline void mem_cgroup_handle_over_high(void)
916 static inline void mem_cgroup_oom_enable(void)
920 static inline void mem_cgroup_oom_disable(void)
924 static inline bool task_in_memcg_oom(struct task_struct *p)
929 static inline bool mem_cgroup_oom_synchronize(bool wait)
934 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
940 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
946 static inline void mod_memcg_state(struct mem_cgroup *memcg,
952 static inline void __mod_memcg_page_state(struct page *page,
958 static inline void mod_memcg_page_state(struct page *page,
964 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
965 enum node_stat_item idx)
967 return node_page_state(lruvec_pgdat(lruvec), idx);
970 static inline void __mod_lruvec_state(struct lruvec *lruvec,
971 enum node_stat_item idx, int val)
973 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
976 static inline void mod_lruvec_state(struct lruvec *lruvec,
977 enum node_stat_item idx, int val)
979 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
982 static inline void __mod_lruvec_page_state(struct page *page,
983 enum node_stat_item idx, int val)
985 __mod_node_page_state(page_pgdat(page), idx, val);
988 static inline void mod_lruvec_page_state(struct page *page,
989 enum node_stat_item idx, int val)
991 mod_node_page_state(page_pgdat(page), idx, val);
995 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
997 unsigned long *total_scanned)
1002 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1006 static inline void count_memcg_events(struct mem_cgroup *memcg,
1007 enum vm_event_item idx,
1008 unsigned long count)
1012 static inline void count_memcg_page_event(struct page *page,
1018 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1021 #endif /* CONFIG_MEMCG */
1023 /* idx can be of type enum memcg_stat_item or node_stat_item */
1024 static inline void __inc_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 __dec_memcg_state(struct mem_cgroup *memcg,
1034 __mod_memcg_state(memcg, idx, -1);
1037 /* idx can be of type enum memcg_stat_item or node_stat_item */
1038 static inline void __inc_memcg_page_state(struct page *page,
1041 __mod_memcg_page_state(page, idx, 1);
1044 /* idx can be of type enum memcg_stat_item or node_stat_item */
1045 static inline void __dec_memcg_page_state(struct page *page,
1048 __mod_memcg_page_state(page, idx, -1);
1051 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1052 enum node_stat_item idx)
1054 __mod_lruvec_state(lruvec, idx, 1);
1057 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1058 enum node_stat_item idx)
1060 __mod_lruvec_state(lruvec, idx, -1);
1063 static inline void __inc_lruvec_page_state(struct page *page,
1064 enum node_stat_item idx)
1066 __mod_lruvec_page_state(page, idx, 1);
1069 static inline void __dec_lruvec_page_state(struct page *page,
1070 enum node_stat_item idx)
1072 __mod_lruvec_page_state(page, idx, -1);
1075 /* idx can be of type enum memcg_stat_item or node_stat_item */
1076 static inline void inc_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 dec_memcg_state(struct mem_cgroup *memcg,
1086 mod_memcg_state(memcg, idx, -1);
1089 /* idx can be of type enum memcg_stat_item or node_stat_item */
1090 static inline void inc_memcg_page_state(struct page *page,
1093 mod_memcg_page_state(page, idx, 1);
1096 /* idx can be of type enum memcg_stat_item or node_stat_item */
1097 static inline void dec_memcg_page_state(struct page *page,
1100 mod_memcg_page_state(page, idx, -1);
1103 static inline void inc_lruvec_state(struct lruvec *lruvec,
1104 enum node_stat_item idx)
1106 mod_lruvec_state(lruvec, idx, 1);
1109 static inline void dec_lruvec_state(struct lruvec *lruvec,
1110 enum node_stat_item idx)
1112 mod_lruvec_state(lruvec, idx, -1);
1115 static inline void inc_lruvec_page_state(struct page *page,
1116 enum node_stat_item idx)
1118 mod_lruvec_page_state(page, idx, 1);
1121 static inline void dec_lruvec_page_state(struct page *page,
1122 enum node_stat_item idx)
1124 mod_lruvec_page_state(page, idx, -1);
1127 #ifdef CONFIG_CGROUP_WRITEBACK
1129 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
1130 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1131 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1132 unsigned long *pheadroom, unsigned long *pdirty,
1133 unsigned long *pwriteback);
1135 #else /* CONFIG_CGROUP_WRITEBACK */
1137 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1142 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1143 unsigned long *pfilepages,
1144 unsigned long *pheadroom,
1145 unsigned long *pdirty,
1146 unsigned long *pwriteback)
1150 #endif /* CONFIG_CGROUP_WRITEBACK */
1153 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1154 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1156 extern struct static_key_false memcg_sockets_enabled_key;
1157 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1158 void mem_cgroup_sk_alloc(struct sock *sk);
1159 void mem_cgroup_sk_free(struct sock *sk);
1160 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1162 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1165 if (time_before(jiffies, memcg->socket_pressure))
1167 } while ((memcg = parent_mem_cgroup(memcg)));
1171 #define mem_cgroup_sockets_enabled 0
1172 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1173 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1174 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1180 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1181 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1182 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1183 struct mem_cgroup *memcg);
1184 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1185 void memcg_kmem_uncharge(struct page *page, int order);
1187 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1188 extern struct static_key_false memcg_kmem_enabled_key;
1189 extern struct workqueue_struct *memcg_kmem_cache_wq;
1191 extern int memcg_nr_cache_ids;
1192 void memcg_get_cache_ids(void);
1193 void memcg_put_cache_ids(void);
1196 * Helper macro to loop through all memcg-specific caches. Callers must still
1197 * check if the cache is valid (it is either valid or NULL).
1198 * the slab_mutex must be held when looping through those caches
1200 #define for_each_memcg_cache_index(_idx) \
1201 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1203 static inline bool memcg_kmem_enabled(void)
1205 return static_branch_unlikely(&memcg_kmem_enabled_key);
1209 * helper for accessing a memcg's index. It will be used as an index in the
1210 * child cache array in kmem_cache, and also to derive its name. This function
1211 * will return -1 when this is not a kmem-limited memcg.
1213 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1215 return memcg ? memcg->kmemcg_id : -1;
1219 #define for_each_memcg_cache_index(_idx) \
1222 static inline bool memcg_kmem_enabled(void)
1227 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1232 static inline void memcg_get_cache_ids(void)
1236 static inline void memcg_put_cache_ids(void)
1240 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1242 #endif /* _LINUX_MEMCONTROL_H */