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 * per-zone information in memory controller.
117 struct mem_cgroup_per_node {
118 struct lruvec lruvec;
120 struct lruvec_stat __percpu *lruvec_stat_cpu;
121 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
123 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
125 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
127 struct rb_node tree_node; /* RB tree node */
128 unsigned long usage_in_excess;/* Set to the value by which */
129 /* the soft limit is exceeded*/
131 bool congested; /* memcg has many dirty pages */
132 /* backed by a congested BDI */
134 struct mem_cgroup *memcg; /* Back pointer, we cannot */
135 /* use container_of */
138 struct mem_cgroup_threshold {
139 struct eventfd_ctx *eventfd;
140 unsigned long threshold;
144 struct mem_cgroup_threshold_ary {
145 /* An array index points to threshold just below or equal to usage. */
146 int current_threshold;
147 /* Size of entries[] */
149 /* Array of thresholds */
150 struct mem_cgroup_threshold entries[0];
153 struct mem_cgroup_thresholds {
154 /* Primary thresholds array */
155 struct mem_cgroup_threshold_ary *primary;
157 * Spare threshold array.
158 * This is needed to make mem_cgroup_unregister_event() "never fail".
159 * It must be able to store at least primary->size - 1 entries.
161 struct mem_cgroup_threshold_ary *spare;
164 enum memcg_kmem_state {
170 #if defined(CONFIG_SMP)
171 struct memcg_padding {
173 } ____cacheline_internodealigned_in_smp;
174 #define MEMCG_PADDING(name) struct memcg_padding name;
176 #define MEMCG_PADDING(name)
180 * The memory controller data structure. The memory controller controls both
181 * page cache and RSS per cgroup. We would eventually like to provide
182 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
183 * to help the administrator determine what knobs to tune.
186 struct cgroup_subsys_state css;
188 /* Private memcg ID. Used to ID objects that outlive the cgroup */
189 struct mem_cgroup_id id;
191 /* Accounted resources */
192 struct page_counter memory;
193 struct page_counter swap;
195 /* Legacy consumer-oriented counters */
196 struct page_counter memsw;
197 struct page_counter kmem;
198 struct page_counter tcpmem;
200 /* Upper bound of normal memory consumption range */
203 /* Range enforcement for interrupt charges */
204 struct work_struct high_work;
206 unsigned long soft_limit;
208 /* vmpressure notifications */
209 struct vmpressure vmpressure;
212 * Should the accounting and control be hierarchical, per subtree?
216 /* protected by memcg_oom_lock */
221 /* OOM-Killer disable */
222 int oom_kill_disable;
225 struct cgroup_file events_file;
227 /* handle for "memory.swap.events" */
228 struct cgroup_file swap_events_file;
230 /* protect arrays of thresholds */
231 struct mutex thresholds_lock;
233 /* thresholds for memory usage. RCU-protected */
234 struct mem_cgroup_thresholds thresholds;
236 /* thresholds for mem+swap usage. RCU-protected */
237 struct mem_cgroup_thresholds memsw_thresholds;
239 /* For oom notifier event fd */
240 struct list_head oom_notify;
243 * Should we move charges of a task when a task is moved into this
244 * mem_cgroup ? And what type of charges should we move ?
246 unsigned long move_charge_at_immigrate;
247 /* taken only while moving_account > 0 */
248 spinlock_t move_lock;
249 unsigned long move_lock_flags;
251 MEMCG_PADDING(_pad1_);
254 * set > 0 if pages under this cgroup are moving to other cgroup.
256 atomic_t moving_account;
257 struct task_struct *move_lock_task;
260 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
262 MEMCG_PADDING(_pad2_);
264 atomic_long_t stat[MEMCG_NR_STAT];
265 atomic_long_t events[NR_VM_EVENT_ITEMS];
266 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
268 unsigned long socket_pressure;
270 /* Legacy tcp memory accounting */
274 #ifdef CONFIG_MEMCG_KMEM
275 /* Index in the kmem_cache->memcg_params.memcg_caches array */
277 enum memcg_kmem_state kmem_state;
278 struct list_head kmem_caches;
281 int last_scanned_node;
283 nodemask_t scan_nodes;
284 atomic_t numainfo_events;
285 atomic_t numainfo_updating;
288 #ifdef CONFIG_CGROUP_WRITEBACK
289 struct list_head cgwb_list;
290 struct wb_domain cgwb_domain;
293 /* List of events which userspace want to receive */
294 struct list_head event_list;
295 spinlock_t event_list_lock;
297 struct mem_cgroup_per_node *nodeinfo[0];
298 /* WARNING: nodeinfo must be the last member here */
302 * size of first charge trial. "32" comes from vmscan.c's magic value.
303 * TODO: maybe necessary to use big numbers in big irons.
305 #define MEMCG_CHARGE_BATCH 32U
307 extern struct mem_cgroup *root_mem_cgroup;
309 static inline bool mem_cgroup_disabled(void)
311 return !cgroup_subsys_enabled(memory_cgrp_subsys);
314 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
315 struct mem_cgroup *memcg);
317 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
318 gfp_t gfp_mask, struct mem_cgroup **memcgp,
320 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
321 gfp_t gfp_mask, struct mem_cgroup **memcgp,
323 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
324 bool lrucare, bool compound);
325 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
327 void mem_cgroup_uncharge(struct page *page);
328 void mem_cgroup_uncharge_list(struct list_head *page_list);
330 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
332 static struct mem_cgroup_per_node *
333 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
335 return memcg->nodeinfo[nid];
339 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
340 * @node: node of the wanted lruvec
341 * @memcg: memcg of the wanted lruvec
343 * Returns the lru list vector holding pages for a given @node or a given
344 * @memcg and @zone. This can be the node lruvec, if the memory controller
347 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
348 struct mem_cgroup *memcg)
350 struct mem_cgroup_per_node *mz;
351 struct lruvec *lruvec;
353 if (mem_cgroup_disabled()) {
354 lruvec = node_lruvec(pgdat);
358 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
359 lruvec = &mz->lruvec;
362 * Since a node can be onlined after the mem_cgroup was created,
363 * we have to be prepared to initialize lruvec->pgdat here;
364 * and if offlined then reonlined, we need to reinitialize it.
366 if (unlikely(lruvec->pgdat != pgdat))
367 lruvec->pgdat = pgdat;
371 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
373 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
374 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
376 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
378 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
381 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
382 return css ? container_of(css, struct mem_cgroup, css) : NULL;
385 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
388 css_put(&memcg->css);
391 #define mem_cgroup_from_counter(counter, member) \
392 container_of(counter, struct mem_cgroup, member)
394 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
396 struct mem_cgroup_reclaim_cookie *);
397 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
398 int mem_cgroup_scan_tasks(struct mem_cgroup *,
399 int (*)(struct task_struct *, void *), void *);
401 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
403 if (mem_cgroup_disabled())
408 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
410 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
412 struct mem_cgroup_per_node *mz;
414 if (mem_cgroup_disabled())
417 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
422 * parent_mem_cgroup - find the accounting parent of a memcg
423 * @memcg: memcg whose parent to find
425 * Returns the parent memcg, or NULL if this is the root or the memory
426 * controller is in legacy no-hierarchy mode.
428 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
430 if (!memcg->memory.parent)
432 return mem_cgroup_from_counter(memcg->memory.parent, memory);
435 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
436 struct mem_cgroup *root)
440 if (!root->use_hierarchy)
442 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
445 static inline bool mm_match_cgroup(struct mm_struct *mm,
446 struct mem_cgroup *memcg)
448 struct mem_cgroup *task_memcg;
452 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
454 match = mem_cgroup_is_descendant(task_memcg, memcg);
459 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
460 ino_t page_cgroup_ino(struct page *page);
462 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
464 if (mem_cgroup_disabled())
466 return !!(memcg->css.flags & CSS_ONLINE);
470 * For memory reclaim.
472 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
474 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
475 int zid, int nr_pages);
477 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
478 int nid, unsigned int lru_mask);
481 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
483 struct mem_cgroup_per_node *mz;
484 unsigned long nr_pages = 0;
487 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
488 for (zid = 0; zid < MAX_NR_ZONES; zid++)
489 nr_pages += mz->lru_zone_size[zid][lru];
494 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
495 enum lru_list lru, int zone_idx)
497 struct mem_cgroup_per_node *mz;
499 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
500 return mz->lru_zone_size[zone_idx][lru];
503 void mem_cgroup_handle_over_high(void);
505 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
507 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
508 struct task_struct *p);
510 static inline void mem_cgroup_enter_user_fault(void)
512 WARN_ON(current->in_user_fault);
513 current->in_user_fault = 1;
516 static inline void mem_cgroup_exit_user_fault(void)
518 WARN_ON(!current->in_user_fault);
519 current->in_user_fault = 0;
522 static inline bool task_in_memcg_oom(struct task_struct *p)
524 return p->memcg_in_oom;
527 bool mem_cgroup_oom_synchronize(bool wait);
529 #ifdef CONFIG_MEMCG_SWAP
530 extern int do_swap_account;
533 struct mem_cgroup *lock_page_memcg(struct page *page);
534 void __unlock_page_memcg(struct mem_cgroup *memcg);
535 void unlock_page_memcg(struct page *page);
537 /* idx can be of type enum memcg_stat_item or node_stat_item */
538 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
541 long x = atomic_long_read(&memcg->stat[idx]);
549 /* idx can be of type enum memcg_stat_item or node_stat_item */
550 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
555 if (mem_cgroup_disabled())
558 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
559 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
560 atomic_long_add(x, &memcg->stat[idx]);
563 __this_cpu_write(memcg->stat_cpu->count[idx], x);
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 local_irq_save(flags);
573 __mod_memcg_state(memcg, idx, val);
574 local_irq_restore(flags);
578 * mod_memcg_page_state - update page state statistics
580 * @idx: page state item to account
581 * @val: number of pages (positive or negative)
583 * The @page must be locked or the caller must use lock_page_memcg()
584 * to prevent double accounting when the page is concurrently being
585 * moved to another memcg:
587 * lock_page(page) or lock_page_memcg(page)
588 * if (TestClearPageState(page))
589 * mod_memcg_page_state(page, state, -1);
590 * unlock_page(page) or unlock_page_memcg(page)
592 * Kernel pages are an exception to this, since they'll never move.
594 static inline void __mod_memcg_page_state(struct page *page,
597 if (page->mem_cgroup)
598 __mod_memcg_state(page->mem_cgroup, idx, val);
601 static inline void mod_memcg_page_state(struct page *page,
604 if (page->mem_cgroup)
605 mod_memcg_state(page->mem_cgroup, idx, val);
608 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
609 enum node_stat_item idx)
611 struct mem_cgroup_per_node *pn;
614 if (mem_cgroup_disabled())
615 return node_page_state(lruvec_pgdat(lruvec), idx);
617 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
618 x = atomic_long_read(&pn->lruvec_stat[idx]);
626 static inline void __mod_lruvec_state(struct lruvec *lruvec,
627 enum node_stat_item idx, int val)
629 struct mem_cgroup_per_node *pn;
633 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
635 if (mem_cgroup_disabled())
638 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
641 __mod_memcg_state(pn->memcg, idx, val);
644 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
645 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
646 atomic_long_add(x, &pn->lruvec_stat[idx]);
649 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
652 static inline void mod_lruvec_state(struct lruvec *lruvec,
653 enum node_stat_item idx, int val)
657 local_irq_save(flags);
658 __mod_lruvec_state(lruvec, idx, val);
659 local_irq_restore(flags);
662 static inline void __mod_lruvec_page_state(struct page *page,
663 enum node_stat_item idx, int val)
665 pg_data_t *pgdat = page_pgdat(page);
666 struct lruvec *lruvec;
668 /* Untracked pages have no memcg, no lruvec. Update only the node */
669 if (!page->mem_cgroup) {
670 __mod_node_page_state(pgdat, idx, val);
674 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
675 __mod_lruvec_state(lruvec, idx, val);
678 static inline void mod_lruvec_page_state(struct page *page,
679 enum node_stat_item idx, int val)
683 local_irq_save(flags);
684 __mod_lruvec_page_state(page, idx, val);
685 local_irq_restore(flags);
688 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
690 unsigned long *total_scanned);
692 static inline void __count_memcg_events(struct mem_cgroup *memcg,
693 enum vm_event_item idx,
698 if (mem_cgroup_disabled())
701 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
702 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
703 atomic_long_add(x, &memcg->events[idx]);
706 __this_cpu_write(memcg->stat_cpu->events[idx], x);
709 static inline void count_memcg_events(struct mem_cgroup *memcg,
710 enum vm_event_item idx,
715 local_irq_save(flags);
716 __count_memcg_events(memcg, idx, count);
717 local_irq_restore(flags);
720 static inline void count_memcg_page_event(struct page *page,
721 enum vm_event_item idx)
723 if (page->mem_cgroup)
724 count_memcg_events(page->mem_cgroup, idx, 1);
727 static inline void count_memcg_event_mm(struct mm_struct *mm,
728 enum vm_event_item idx)
730 struct mem_cgroup *memcg;
732 if (mem_cgroup_disabled())
736 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
738 count_memcg_events(memcg, idx, 1);
742 static inline void memcg_memory_event(struct mem_cgroup *memcg,
743 enum memcg_memory_event event)
745 atomic_long_inc(&memcg->memory_events[event]);
746 cgroup_file_notify(&memcg->events_file);
749 static inline void memcg_memory_event_mm(struct mm_struct *mm,
750 enum memcg_memory_event event)
752 struct mem_cgroup *memcg;
754 if (mem_cgroup_disabled())
758 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
760 memcg_memory_event(memcg, event);
764 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
765 void mem_cgroup_split_huge_fixup(struct page *head);
768 #else /* CONFIG_MEMCG */
770 #define MEM_CGROUP_ID_SHIFT 0
771 #define MEM_CGROUP_ID_MAX 0
775 static inline bool mem_cgroup_disabled(void)
780 static inline void memcg_memory_event(struct mem_cgroup *memcg,
781 enum memcg_memory_event event)
785 static inline void memcg_memory_event_mm(struct mm_struct *mm,
786 enum memcg_memory_event event)
790 static inline enum mem_cgroup_protection mem_cgroup_protected(
791 struct mem_cgroup *root, struct mem_cgroup *memcg)
793 return MEMCG_PROT_NONE;
796 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
798 struct mem_cgroup **memcgp,
805 static inline int mem_cgroup_try_charge_delay(struct page *page,
806 struct mm_struct *mm,
808 struct mem_cgroup **memcgp,
815 static inline void mem_cgroup_commit_charge(struct page *page,
816 struct mem_cgroup *memcg,
817 bool lrucare, bool compound)
821 static inline void mem_cgroup_cancel_charge(struct page *page,
822 struct mem_cgroup *memcg,
827 static inline void mem_cgroup_uncharge(struct page *page)
831 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
835 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
839 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
840 struct mem_cgroup *memcg)
842 return node_lruvec(pgdat);
845 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
846 struct pglist_data *pgdat)
848 return &pgdat->lruvec;
851 static inline bool mm_match_cgroup(struct mm_struct *mm,
852 struct mem_cgroup *memcg)
857 static inline bool task_in_mem_cgroup(struct task_struct *task,
858 const struct mem_cgroup *memcg)
863 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
868 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
873 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
877 static inline struct mem_cgroup *
878 mem_cgroup_iter(struct mem_cgroup *root,
879 struct mem_cgroup *prev,
880 struct mem_cgroup_reclaim_cookie *reclaim)
885 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
886 struct mem_cgroup *prev)
890 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
891 int (*fn)(struct task_struct *, void *), void *arg)
896 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
901 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
904 /* XXX: This should always return root_mem_cgroup */
908 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
913 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
918 static inline unsigned long
919 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
924 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
925 enum lru_list lru, int zone_idx)
930 static inline unsigned long
931 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
932 int nid, unsigned int lru_mask)
937 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
943 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
947 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
952 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
956 static inline void unlock_page_memcg(struct page *page)
960 static inline void mem_cgroup_handle_over_high(void)
964 static inline void mem_cgroup_enter_user_fault(void)
968 static inline void mem_cgroup_exit_user_fault(void)
972 static inline bool task_in_memcg_oom(struct task_struct *p)
977 static inline bool mem_cgroup_oom_synchronize(bool wait)
982 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
988 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
994 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1000 static inline void __mod_memcg_page_state(struct page *page,
1006 static inline void mod_memcg_page_state(struct page *page,
1012 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1013 enum node_stat_item idx)
1015 return node_page_state(lruvec_pgdat(lruvec), idx);
1018 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1019 enum node_stat_item idx, int val)
1021 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1024 static inline void mod_lruvec_state(struct lruvec *lruvec,
1025 enum node_stat_item idx, int val)
1027 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1030 static inline void __mod_lruvec_page_state(struct page *page,
1031 enum node_stat_item idx, int val)
1033 __mod_node_page_state(page_pgdat(page), idx, val);
1036 static inline void mod_lruvec_page_state(struct page *page,
1037 enum node_stat_item idx, int val)
1039 mod_node_page_state(page_pgdat(page), idx, val);
1043 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1045 unsigned long *total_scanned)
1050 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1054 static inline void count_memcg_events(struct mem_cgroup *memcg,
1055 enum vm_event_item idx,
1056 unsigned long count)
1060 static inline void count_memcg_page_event(struct page *page,
1066 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1069 #endif /* CONFIG_MEMCG */
1071 /* idx can be of type enum memcg_stat_item or node_stat_item */
1072 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1075 __mod_memcg_state(memcg, idx, 1);
1078 /* idx can be of type enum memcg_stat_item or node_stat_item */
1079 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1082 __mod_memcg_state(memcg, idx, -1);
1085 /* idx can be of type enum memcg_stat_item or node_stat_item */
1086 static inline void __inc_memcg_page_state(struct page *page,
1089 __mod_memcg_page_state(page, idx, 1);
1092 /* idx can be of type enum memcg_stat_item or node_stat_item */
1093 static inline void __dec_memcg_page_state(struct page *page,
1096 __mod_memcg_page_state(page, idx, -1);
1099 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1100 enum node_stat_item idx)
1102 __mod_lruvec_state(lruvec, idx, 1);
1105 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1106 enum node_stat_item idx)
1108 __mod_lruvec_state(lruvec, idx, -1);
1111 static inline void __inc_lruvec_page_state(struct page *page,
1112 enum node_stat_item idx)
1114 __mod_lruvec_page_state(page, idx, 1);
1117 static inline void __dec_lruvec_page_state(struct page *page,
1118 enum node_stat_item idx)
1120 __mod_lruvec_page_state(page, idx, -1);
1123 /* idx can be of type enum memcg_stat_item or node_stat_item */
1124 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1127 mod_memcg_state(memcg, idx, 1);
1130 /* idx can be of type enum memcg_stat_item or node_stat_item */
1131 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1134 mod_memcg_state(memcg, idx, -1);
1137 /* idx can be of type enum memcg_stat_item or node_stat_item */
1138 static inline void inc_memcg_page_state(struct page *page,
1141 mod_memcg_page_state(page, idx, 1);
1144 /* idx can be of type enum memcg_stat_item or node_stat_item */
1145 static inline void dec_memcg_page_state(struct page *page,
1148 mod_memcg_page_state(page, idx, -1);
1151 static inline void inc_lruvec_state(struct lruvec *lruvec,
1152 enum node_stat_item idx)
1154 mod_lruvec_state(lruvec, idx, 1);
1157 static inline void dec_lruvec_state(struct lruvec *lruvec,
1158 enum node_stat_item idx)
1160 mod_lruvec_state(lruvec, idx, -1);
1163 static inline void inc_lruvec_page_state(struct page *page,
1164 enum node_stat_item idx)
1166 mod_lruvec_page_state(page, idx, 1);
1169 static inline void dec_lruvec_page_state(struct page *page,
1170 enum node_stat_item idx)
1172 mod_lruvec_page_state(page, idx, -1);
1175 #ifdef CONFIG_CGROUP_WRITEBACK
1177 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1178 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1179 unsigned long *pheadroom, unsigned long *pdirty,
1180 unsigned long *pwriteback);
1182 #else /* CONFIG_CGROUP_WRITEBACK */
1184 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1189 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1190 unsigned long *pfilepages,
1191 unsigned long *pheadroom,
1192 unsigned long *pdirty,
1193 unsigned long *pwriteback)
1197 #endif /* CONFIG_CGROUP_WRITEBACK */
1200 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1201 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1203 extern struct static_key_false memcg_sockets_enabled_key;
1204 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1205 void mem_cgroup_sk_alloc(struct sock *sk);
1206 void mem_cgroup_sk_free(struct sock *sk);
1207 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1209 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1212 if (time_before(jiffies, memcg->socket_pressure))
1214 } while ((memcg = parent_mem_cgroup(memcg)));
1218 #define mem_cgroup_sockets_enabled 0
1219 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1220 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1221 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1227 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1228 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1229 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1230 struct mem_cgroup *memcg);
1231 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1232 void memcg_kmem_uncharge(struct page *page, int order);
1234 #ifdef CONFIG_MEMCG_KMEM
1235 extern struct static_key_false memcg_kmem_enabled_key;
1236 extern struct workqueue_struct *memcg_kmem_cache_wq;
1238 extern int memcg_nr_cache_ids;
1239 void memcg_get_cache_ids(void);
1240 void memcg_put_cache_ids(void);
1243 * Helper macro to loop through all memcg-specific caches. Callers must still
1244 * check if the cache is valid (it is either valid or NULL).
1245 * the slab_mutex must be held when looping through those caches
1247 #define for_each_memcg_cache_index(_idx) \
1248 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1250 static inline bool memcg_kmem_enabled(void)
1252 return static_branch_unlikely(&memcg_kmem_enabled_key);
1256 * helper for accessing a memcg's index. It will be used as an index in the
1257 * child cache array in kmem_cache, and also to derive its name. This function
1258 * will return -1 when this is not a kmem-limited memcg.
1260 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1262 return memcg ? memcg->kmemcg_id : -1;
1266 #define for_each_memcg_cache_index(_idx) \
1269 static inline bool memcg_kmem_enabled(void)
1274 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1279 static inline void memcg_get_cache_ids(void)
1283 static inline void memcg_put_cache_ids(void)
1287 #endif /* CONFIG_MEMCG_KMEM */
1289 #endif /* _LINUX_MEMCONTROL_H */