1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* memcontrol.h - Memory Controller
4 * Copyright IBM Corporation, 2007
5 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
7 * Copyright 2007 OpenVZ SWsoft Inc
8 * Author: Pavel Emelianov <xemul@openvz.org>
11 #ifndef _LINUX_MEMCONTROL_H
12 #define _LINUX_MEMCONTROL_H
13 #include <linux/cgroup.h>
14 #include <linux/vm_event_item.h>
15 #include <linux/hardirq.h>
16 #include <linux/jump_label.h>
17 #include <linux/page_counter.h>
18 #include <linux/vmpressure.h>
19 #include <linux/eventfd.h>
21 #include <linux/vmstat.h>
22 #include <linux/writeback.h>
23 #include <linux/page-flags.h>
30 /* Cgroup-specific page state, on top of universal node page state */
31 enum memcg_stat_item {
32 MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
37 /* XXX: why are these zone and not node counters? */
38 MEMCG_KERNEL_STACK_KB,
42 enum memcg_memory_event {
50 MEMCG_NR_MEMORY_EVENTS,
53 enum mem_cgroup_protection {
59 struct mem_cgroup_reclaim_cookie {
62 unsigned int generation;
67 #define MEM_CGROUP_ID_SHIFT 16
68 #define MEM_CGROUP_ID_MAX USHRT_MAX
70 struct mem_cgroup_id {
76 * Per memcg event counter is incremented at every pagein/pageout. With THP,
77 * it will be incremated by the number of pages. This counter is used for
78 * for trigger some periodic events. This is straightforward and better
79 * than using jiffies etc. to handle periodic memcg event.
81 enum mem_cgroup_events_target {
82 MEM_CGROUP_TARGET_THRESH,
83 MEM_CGROUP_TARGET_SOFTLIMIT,
84 MEM_CGROUP_TARGET_NUMAINFO,
88 struct memcg_vmstats_percpu {
89 long stat[MEMCG_NR_STAT];
90 unsigned long events[NR_VM_EVENT_ITEMS];
91 unsigned long nr_page_events;
92 unsigned long targets[MEM_CGROUP_NTARGETS];
95 struct mem_cgroup_reclaim_iter {
96 struct mem_cgroup *position;
97 /* scan generation, increased every round-trip */
98 unsigned int generation;
102 long count[NR_VM_NODE_STAT_ITEMS];
106 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
107 * which have elements charged to this memcg.
109 struct memcg_shrinker_map {
111 unsigned long map[0];
115 * per-zone information in memory controller.
117 struct mem_cgroup_per_node {
118 struct lruvec lruvec;
120 /* Legacy local VM stats */
121 struct lruvec_stat __percpu *lruvec_stat_local;
123 /* Subtree VM stats (batched updates) */
124 struct lruvec_stat __percpu *lruvec_stat_cpu;
125 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
127 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
129 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
131 struct memcg_shrinker_map __rcu *shrinker_map;
133 struct rb_node tree_node; /* RB tree node */
134 unsigned long usage_in_excess;/* Set to the value by which */
135 /* the soft limit is exceeded*/
137 bool congested; /* memcg has many dirty pages */
138 /* backed by a congested BDI */
140 struct mem_cgroup *memcg; /* Back pointer, we cannot */
141 /* use container_of */
144 struct mem_cgroup_threshold {
145 struct eventfd_ctx *eventfd;
146 unsigned long threshold;
150 struct mem_cgroup_threshold_ary {
151 /* An array index points to threshold just below or equal to usage. */
152 int current_threshold;
153 /* Size of entries[] */
155 /* Array of thresholds */
156 struct mem_cgroup_threshold entries[0];
159 struct mem_cgroup_thresholds {
160 /* Primary thresholds array */
161 struct mem_cgroup_threshold_ary *primary;
163 * Spare threshold array.
164 * This is needed to make mem_cgroup_unregister_event() "never fail".
165 * It must be able to store at least primary->size - 1 entries.
167 struct mem_cgroup_threshold_ary *spare;
170 enum memcg_kmem_state {
176 #if defined(CONFIG_SMP)
177 struct memcg_padding {
179 } ____cacheline_internodealigned_in_smp;
180 #define MEMCG_PADDING(name) struct memcg_padding name;
182 #define MEMCG_PADDING(name)
186 * Remember four most recent foreign writebacks with dirty pages in this
187 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
188 * one in a given round, we're likely to catch it later if it keeps
189 * foreign-dirtying, so a fairly low count should be enough.
191 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
193 #define MEMCG_CGWB_FRN_CNT 4
195 struct memcg_cgwb_frn {
196 u64 bdi_id; /* bdi->id of the foreign inode */
197 int memcg_id; /* memcg->css.id of foreign inode */
198 u64 at; /* jiffies_64 at the time of dirtying */
199 struct wb_completion done; /* tracks in-flight foreign writebacks */
203 * The memory controller data structure. The memory controller controls both
204 * page cache and RSS per cgroup. We would eventually like to provide
205 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
206 * to help the administrator determine what knobs to tune.
209 struct cgroup_subsys_state css;
211 /* Private memcg ID. Used to ID objects that outlive the cgroup */
212 struct mem_cgroup_id id;
214 /* Accounted resources */
215 struct page_counter memory;
216 struct page_counter swap;
218 /* Legacy consumer-oriented counters */
219 struct page_counter memsw;
220 struct page_counter kmem;
221 struct page_counter tcpmem;
223 /* Upper bound of normal memory consumption range */
226 /* Range enforcement for interrupt charges */
227 struct work_struct high_work;
229 unsigned long soft_limit;
231 /* vmpressure notifications */
232 struct vmpressure vmpressure;
235 * Should the accounting and control be hierarchical, per subtree?
240 * Should the OOM killer kill all belonging tasks, had it kill one?
244 /* protected by memcg_oom_lock */
249 /* OOM-Killer disable */
250 int oom_kill_disable;
252 /* memory.events and memory.events.local */
253 struct cgroup_file events_file;
254 struct cgroup_file events_local_file;
256 /* handle for "memory.swap.events" */
257 struct cgroup_file swap_events_file;
259 /* protect arrays of thresholds */
260 struct mutex thresholds_lock;
262 /* thresholds for memory usage. RCU-protected */
263 struct mem_cgroup_thresholds thresholds;
265 /* thresholds for mem+swap usage. RCU-protected */
266 struct mem_cgroup_thresholds memsw_thresholds;
268 /* For oom notifier event fd */
269 struct list_head oom_notify;
272 * Should we move charges of a task when a task is moved into this
273 * mem_cgroup ? And what type of charges should we move ?
275 unsigned long move_charge_at_immigrate;
276 /* taken only while moving_account > 0 */
277 spinlock_t move_lock;
278 unsigned long move_lock_flags;
280 MEMCG_PADDING(_pad1_);
283 * set > 0 if pages under this cgroup are moving to other cgroup.
285 atomic_t moving_account;
286 struct task_struct *move_lock_task;
288 /* Legacy local VM stats and events */
289 struct memcg_vmstats_percpu __percpu *vmstats_local;
291 /* Subtree VM stats and events (batched updates) */
292 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
294 MEMCG_PADDING(_pad2_);
296 atomic_long_t vmstats[MEMCG_NR_STAT];
297 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
300 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
301 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
303 unsigned long socket_pressure;
305 /* Legacy tcp memory accounting */
309 #ifdef CONFIG_MEMCG_KMEM
310 /* Index in the kmem_cache->memcg_params.memcg_caches array */
312 enum memcg_kmem_state kmem_state;
313 struct list_head kmem_caches;
316 int last_scanned_node;
318 nodemask_t scan_nodes;
319 atomic_t numainfo_events;
320 atomic_t numainfo_updating;
323 #ifdef CONFIG_CGROUP_WRITEBACK
324 struct list_head cgwb_list;
325 struct wb_domain cgwb_domain;
326 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
329 /* List of events which userspace want to receive */
330 struct list_head event_list;
331 spinlock_t event_list_lock;
333 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
334 struct deferred_split deferred_split_queue;
337 struct mem_cgroup_per_node *nodeinfo[0];
338 /* WARNING: nodeinfo must be the last member here */
342 * size of first charge trial. "32" comes from vmscan.c's magic value.
343 * TODO: maybe necessary to use big numbers in big irons.
345 #define MEMCG_CHARGE_BATCH 32U
347 extern struct mem_cgroup *root_mem_cgroup;
349 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
351 return (memcg == root_mem_cgroup);
354 static inline bool mem_cgroup_disabled(void)
356 return !cgroup_subsys_enabled(memory_cgrp_subsys);
359 static inline void mem_cgroup_protection(struct mem_cgroup *root,
360 struct mem_cgroup *memcg,
366 if (mem_cgroup_disabled())
370 * There is no reclaim protection applied to a targeted reclaim.
371 * We are special casing this specific case here because
372 * mem_cgroup_protected calculation is not robust enough to keep
373 * the protection invariant for calculated effective values for
374 * parallel reclaimers with different reclaim target. This is
375 * especially a problem for tail memcgs (as they have pages on LRU)
376 * which would want to have effective values 0 for targeted reclaim
377 * but a different value for external reclaim.
380 * Let's have global and A's reclaim in parallel:
382 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
384 * | C (low = 1G, usage = 2.5G)
385 * B (low = 1G, usage = 0.5G)
387 * For the global reclaim
389 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
390 * C.elow = min(C.usage, C.low)
392 * With the effective values resetting we have A reclaim
397 * If the global reclaim races with A's reclaim then
398 * B.elow = C.elow = 0 because children_low_usage > A.elow)
399 * is possible and reclaiming B would be violating the protection.
405 *min = READ_ONCE(memcg->memory.emin);
406 *low = READ_ONCE(memcg->memory.elow);
409 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
410 struct mem_cgroup *memcg);
412 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
413 gfp_t gfp_mask, struct mem_cgroup **memcgp,
415 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
416 gfp_t gfp_mask, struct mem_cgroup **memcgp,
418 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
419 bool lrucare, bool compound);
420 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
422 void mem_cgroup_uncharge(struct page *page);
423 void mem_cgroup_uncharge_list(struct list_head *page_list);
425 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
427 static struct mem_cgroup_per_node *
428 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
430 return memcg->nodeinfo[nid];
434 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
435 * @node: node of the wanted lruvec
436 * @memcg: memcg of the wanted lruvec
438 * Returns the lru list vector holding pages for a given @node or a given
439 * @memcg and @zone. This can be the node lruvec, if the memory controller
442 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
443 struct mem_cgroup *memcg)
445 struct mem_cgroup_per_node *mz;
446 struct lruvec *lruvec;
448 if (mem_cgroup_disabled()) {
449 lruvec = node_lruvec(pgdat);
453 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
454 lruvec = &mz->lruvec;
457 * Since a node can be onlined after the mem_cgroup was created,
458 * we have to be prepared to initialize lruvec->pgdat here;
459 * and if offlined then reonlined, we need to reinitialize it.
461 if (unlikely(lruvec->pgdat != pgdat))
462 lruvec->pgdat = pgdat;
466 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
468 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
470 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
472 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
475 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
476 return css ? container_of(css, struct mem_cgroup, css) : NULL;
479 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
482 css_put(&memcg->css);
485 #define mem_cgroup_from_counter(counter, member) \
486 container_of(counter, struct mem_cgroup, member)
488 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
490 struct mem_cgroup_reclaim_cookie *);
491 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
492 int mem_cgroup_scan_tasks(struct mem_cgroup *,
493 int (*)(struct task_struct *, void *), void *);
495 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
497 if (mem_cgroup_disabled())
502 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
504 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
506 return mem_cgroup_from_css(seq_css(m));
509 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
511 struct mem_cgroup_per_node *mz;
513 if (mem_cgroup_disabled())
516 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
521 * parent_mem_cgroup - find the accounting parent of a memcg
522 * @memcg: memcg whose parent to find
524 * Returns the parent memcg, or NULL if this is the root or the memory
525 * controller is in legacy no-hierarchy mode.
527 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
529 if (!memcg->memory.parent)
531 return mem_cgroup_from_counter(memcg->memory.parent, memory);
534 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
535 struct mem_cgroup *root)
539 if (!root->use_hierarchy)
541 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
544 static inline bool mm_match_cgroup(struct mm_struct *mm,
545 struct mem_cgroup *memcg)
547 struct mem_cgroup *task_memcg;
551 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
553 match = mem_cgroup_is_descendant(task_memcg, memcg);
558 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
559 ino_t page_cgroup_ino(struct page *page);
561 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
563 if (mem_cgroup_disabled())
565 return !!(memcg->css.flags & CSS_ONLINE);
569 * For memory reclaim.
571 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
573 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
574 int zid, int nr_pages);
577 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
578 enum lru_list lru, int zone_idx)
580 struct mem_cgroup_per_node *mz;
582 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
583 return mz->lru_zone_size[zone_idx][lru];
586 void mem_cgroup_handle_over_high(void);
588 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
590 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
592 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
593 struct task_struct *p);
595 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
597 static inline void mem_cgroup_enter_user_fault(void)
599 WARN_ON(current->in_user_fault);
600 current->in_user_fault = 1;
603 static inline void mem_cgroup_exit_user_fault(void)
605 WARN_ON(!current->in_user_fault);
606 current->in_user_fault = 0;
609 static inline bool task_in_memcg_oom(struct task_struct *p)
611 return p->memcg_in_oom;
614 bool mem_cgroup_oom_synchronize(bool wait);
615 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
616 struct mem_cgroup *oom_domain);
617 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
619 #ifdef CONFIG_MEMCG_SWAP
620 extern int do_swap_account;
623 struct mem_cgroup *lock_page_memcg(struct page *page);
624 void __unlock_page_memcg(struct mem_cgroup *memcg);
625 void unlock_page_memcg(struct page *page);
628 * idx can be of type enum memcg_stat_item or node_stat_item.
629 * Keep in sync with memcg_exact_page_state().
631 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
633 long x = atomic_long_read(&memcg->vmstats[idx]);
642 * idx can be of type enum memcg_stat_item or node_stat_item.
643 * Keep in sync with memcg_exact_page_state().
645 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
651 for_each_possible_cpu(cpu)
652 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
660 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
662 /* idx can be of type enum memcg_stat_item or node_stat_item */
663 static inline void mod_memcg_state(struct mem_cgroup *memcg,
668 local_irq_save(flags);
669 __mod_memcg_state(memcg, idx, val);
670 local_irq_restore(flags);
674 * mod_memcg_page_state - update page state statistics
676 * @idx: page state item to account
677 * @val: number of pages (positive or negative)
679 * The @page must be locked or the caller must use lock_page_memcg()
680 * to prevent double accounting when the page is concurrently being
681 * moved to another memcg:
683 * lock_page(page) or lock_page_memcg(page)
684 * if (TestClearPageState(page))
685 * mod_memcg_page_state(page, state, -1);
686 * unlock_page(page) or unlock_page_memcg(page)
688 * Kernel pages are an exception to this, since they'll never move.
690 static inline void __mod_memcg_page_state(struct page *page,
693 if (page->mem_cgroup)
694 __mod_memcg_state(page->mem_cgroup, idx, val);
697 static inline void mod_memcg_page_state(struct page *page,
700 if (page->mem_cgroup)
701 mod_memcg_state(page->mem_cgroup, idx, val);
704 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
705 enum node_stat_item idx)
707 struct mem_cgroup_per_node *pn;
710 if (mem_cgroup_disabled())
711 return node_page_state(lruvec_pgdat(lruvec), idx);
713 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
714 x = atomic_long_read(&pn->lruvec_stat[idx]);
722 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
723 enum node_stat_item idx)
725 struct mem_cgroup_per_node *pn;
729 if (mem_cgroup_disabled())
730 return node_page_state(lruvec_pgdat(lruvec), idx);
732 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
733 for_each_possible_cpu(cpu)
734 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
742 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
744 void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val);
745 void mod_memcg_obj_state(void *p, int idx, int val);
747 static inline void mod_lruvec_state(struct lruvec *lruvec,
748 enum node_stat_item idx, int val)
752 local_irq_save(flags);
753 __mod_lruvec_state(lruvec, idx, val);
754 local_irq_restore(flags);
757 static inline void __mod_lruvec_page_state(struct page *page,
758 enum node_stat_item idx, int val)
760 pg_data_t *pgdat = page_pgdat(page);
761 struct lruvec *lruvec;
763 /* Untracked pages have no memcg, no lruvec. Update only the node */
764 if (!page->mem_cgroup) {
765 __mod_node_page_state(pgdat, idx, val);
769 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
770 __mod_lruvec_state(lruvec, idx, val);
773 static inline void mod_lruvec_page_state(struct page *page,
774 enum node_stat_item idx, int val)
778 local_irq_save(flags);
779 __mod_lruvec_page_state(page, idx, val);
780 local_irq_restore(flags);
783 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
785 unsigned long *total_scanned);
787 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
788 unsigned long count);
790 static inline void count_memcg_events(struct mem_cgroup *memcg,
791 enum vm_event_item idx,
796 local_irq_save(flags);
797 __count_memcg_events(memcg, idx, count);
798 local_irq_restore(flags);
801 static inline void count_memcg_page_event(struct page *page,
802 enum vm_event_item idx)
804 if (page->mem_cgroup)
805 count_memcg_events(page->mem_cgroup, idx, 1);
808 static inline void count_memcg_event_mm(struct mm_struct *mm,
809 enum vm_event_item idx)
811 struct mem_cgroup *memcg;
813 if (mem_cgroup_disabled())
817 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
819 count_memcg_events(memcg, idx, 1);
823 static inline void memcg_memory_event(struct mem_cgroup *memcg,
824 enum memcg_memory_event event)
826 atomic_long_inc(&memcg->memory_events_local[event]);
827 cgroup_file_notify(&memcg->events_local_file);
830 atomic_long_inc(&memcg->memory_events[event]);
831 cgroup_file_notify(&memcg->events_file);
833 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
835 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
837 } while ((memcg = parent_mem_cgroup(memcg)) &&
838 !mem_cgroup_is_root(memcg));
841 static inline void memcg_memory_event_mm(struct mm_struct *mm,
842 enum memcg_memory_event event)
844 struct mem_cgroup *memcg;
846 if (mem_cgroup_disabled())
850 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
852 memcg_memory_event(memcg, event);
856 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
857 void mem_cgroup_split_huge_fixup(struct page *head);
860 #else /* CONFIG_MEMCG */
862 #define MEM_CGROUP_ID_SHIFT 0
863 #define MEM_CGROUP_ID_MAX 0
867 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
872 static inline bool mem_cgroup_disabled(void)
877 static inline void memcg_memory_event(struct mem_cgroup *memcg,
878 enum memcg_memory_event event)
882 static inline void memcg_memory_event_mm(struct mm_struct *mm,
883 enum memcg_memory_event event)
887 static inline void mem_cgroup_protection(struct mem_cgroup *root,
888 struct mem_cgroup *memcg,
895 static inline enum mem_cgroup_protection mem_cgroup_protected(
896 struct mem_cgroup *root, struct mem_cgroup *memcg)
898 return MEMCG_PROT_NONE;
901 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
903 struct mem_cgroup **memcgp,
910 static inline int mem_cgroup_try_charge_delay(struct page *page,
911 struct mm_struct *mm,
913 struct mem_cgroup **memcgp,
920 static inline void mem_cgroup_commit_charge(struct page *page,
921 struct mem_cgroup *memcg,
922 bool lrucare, bool compound)
926 static inline void mem_cgroup_cancel_charge(struct page *page,
927 struct mem_cgroup *memcg,
932 static inline void mem_cgroup_uncharge(struct page *page)
936 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
940 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
944 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
945 struct mem_cgroup *memcg)
947 return node_lruvec(pgdat);
950 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
951 struct pglist_data *pgdat)
953 return &pgdat->lruvec;
956 static inline bool mm_match_cgroup(struct mm_struct *mm,
957 struct mem_cgroup *memcg)
962 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
967 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
972 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
976 static inline struct mem_cgroup *
977 mem_cgroup_iter(struct mem_cgroup *root,
978 struct mem_cgroup *prev,
979 struct mem_cgroup_reclaim_cookie *reclaim)
984 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
985 struct mem_cgroup *prev)
989 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
990 int (*fn)(struct task_struct *, void *), void *arg)
995 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1000 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1003 /* XXX: This should always return root_mem_cgroup */
1007 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1012 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1017 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1023 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1024 enum lru_list lru, int zone_idx)
1029 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1034 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1040 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1045 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1049 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1054 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1058 static inline void unlock_page_memcg(struct page *page)
1062 static inline void mem_cgroup_handle_over_high(void)
1066 static inline void mem_cgroup_enter_user_fault(void)
1070 static inline void mem_cgroup_exit_user_fault(void)
1074 static inline bool task_in_memcg_oom(struct task_struct *p)
1079 static inline bool mem_cgroup_oom_synchronize(bool wait)
1084 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1085 struct task_struct *victim, struct mem_cgroup *oom_domain)
1090 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1094 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1099 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1105 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1111 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1117 static inline void __mod_memcg_page_state(struct page *page,
1123 static inline void mod_memcg_page_state(struct page *page,
1129 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1130 enum node_stat_item idx)
1132 return node_page_state(lruvec_pgdat(lruvec), idx);
1135 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1136 enum node_stat_item idx)
1138 return node_page_state(lruvec_pgdat(lruvec), idx);
1141 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1142 enum node_stat_item idx, int val)
1144 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1147 static inline void mod_lruvec_state(struct lruvec *lruvec,
1148 enum node_stat_item idx, int val)
1150 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1153 static inline void __mod_lruvec_page_state(struct page *page,
1154 enum node_stat_item idx, int val)
1156 __mod_node_page_state(page_pgdat(page), idx, val);
1159 static inline void mod_lruvec_page_state(struct page *page,
1160 enum node_stat_item idx, int val)
1162 mod_node_page_state(page_pgdat(page), idx, val);
1165 static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx,
1168 struct page *page = virt_to_head_page(p);
1170 __mod_node_page_state(page_pgdat(page), idx, val);
1173 static inline void mod_memcg_obj_state(void *p, int idx, int val)
1178 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1180 unsigned long *total_scanned)
1185 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1189 static inline void count_memcg_events(struct mem_cgroup *memcg,
1190 enum vm_event_item idx,
1191 unsigned long count)
1195 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1196 enum vm_event_item idx,
1197 unsigned long count)
1201 static inline void count_memcg_page_event(struct page *page,
1207 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1210 #endif /* CONFIG_MEMCG */
1212 /* idx can be of type enum memcg_stat_item or node_stat_item */
1213 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1216 __mod_memcg_state(memcg, idx, 1);
1219 /* idx can be of type enum memcg_stat_item or node_stat_item */
1220 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1223 __mod_memcg_state(memcg, idx, -1);
1226 /* idx can be of type enum memcg_stat_item or node_stat_item */
1227 static inline void __inc_memcg_page_state(struct page *page,
1230 __mod_memcg_page_state(page, idx, 1);
1233 /* idx can be of type enum memcg_stat_item or node_stat_item */
1234 static inline void __dec_memcg_page_state(struct page *page,
1237 __mod_memcg_page_state(page, idx, -1);
1240 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1241 enum node_stat_item idx)
1243 __mod_lruvec_state(lruvec, idx, 1);
1246 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1247 enum node_stat_item idx)
1249 __mod_lruvec_state(lruvec, idx, -1);
1252 static inline void __inc_lruvec_page_state(struct page *page,
1253 enum node_stat_item idx)
1255 __mod_lruvec_page_state(page, idx, 1);
1258 static inline void __dec_lruvec_page_state(struct page *page,
1259 enum node_stat_item idx)
1261 __mod_lruvec_page_state(page, idx, -1);
1264 static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx)
1266 __mod_lruvec_slab_state(p, idx, 1);
1269 static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx)
1271 __mod_lruvec_slab_state(p, idx, -1);
1274 /* idx can be of type enum memcg_stat_item or node_stat_item */
1275 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1278 mod_memcg_state(memcg, idx, 1);
1281 /* idx can be of type enum memcg_stat_item or node_stat_item */
1282 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1285 mod_memcg_state(memcg, idx, -1);
1288 /* idx can be of type enum memcg_stat_item or node_stat_item */
1289 static inline void inc_memcg_page_state(struct page *page,
1292 mod_memcg_page_state(page, idx, 1);
1295 /* idx can be of type enum memcg_stat_item or node_stat_item */
1296 static inline void dec_memcg_page_state(struct page *page,
1299 mod_memcg_page_state(page, idx, -1);
1302 static inline void inc_lruvec_state(struct lruvec *lruvec,
1303 enum node_stat_item idx)
1305 mod_lruvec_state(lruvec, idx, 1);
1308 static inline void dec_lruvec_state(struct lruvec *lruvec,
1309 enum node_stat_item idx)
1311 mod_lruvec_state(lruvec, idx, -1);
1314 static inline void inc_lruvec_page_state(struct page *page,
1315 enum node_stat_item idx)
1317 mod_lruvec_page_state(page, idx, 1);
1320 static inline void dec_lruvec_page_state(struct page *page,
1321 enum node_stat_item idx)
1323 mod_lruvec_page_state(page, idx, -1);
1326 #ifdef CONFIG_CGROUP_WRITEBACK
1328 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1329 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1330 unsigned long *pheadroom, unsigned long *pdirty,
1331 unsigned long *pwriteback);
1333 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1334 struct bdi_writeback *wb);
1336 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1337 struct bdi_writeback *wb)
1339 if (mem_cgroup_disabled())
1342 if (unlikely(&page->mem_cgroup->css != wb->memcg_css))
1343 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1346 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1348 #else /* CONFIG_CGROUP_WRITEBACK */
1350 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1355 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1356 unsigned long *pfilepages,
1357 unsigned long *pheadroom,
1358 unsigned long *pdirty,
1359 unsigned long *pwriteback)
1363 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1364 struct bdi_writeback *wb)
1368 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1372 #endif /* CONFIG_CGROUP_WRITEBACK */
1375 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1376 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1378 extern struct static_key_false memcg_sockets_enabled_key;
1379 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1380 void mem_cgroup_sk_alloc(struct sock *sk);
1381 void mem_cgroup_sk_free(struct sock *sk);
1382 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1384 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1387 if (time_before(jiffies, memcg->socket_pressure))
1389 } while ((memcg = parent_mem_cgroup(memcg)));
1393 extern int memcg_expand_shrinker_maps(int new_id);
1395 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1396 int nid, int shrinker_id);
1398 #define mem_cgroup_sockets_enabled 0
1399 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1400 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1401 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1406 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1407 int nid, int shrinker_id)
1412 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1413 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1415 #ifdef CONFIG_MEMCG_KMEM
1416 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1417 void __memcg_kmem_uncharge(struct page *page, int order);
1418 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1419 struct mem_cgroup *memcg);
1420 void __memcg_kmem_uncharge_memcg(struct mem_cgroup *memcg,
1421 unsigned int nr_pages);
1423 extern struct static_key_false memcg_kmem_enabled_key;
1424 extern struct workqueue_struct *memcg_kmem_cache_wq;
1426 extern int memcg_nr_cache_ids;
1427 void memcg_get_cache_ids(void);
1428 void memcg_put_cache_ids(void);
1431 * Helper macro to loop through all memcg-specific caches. Callers must still
1432 * check if the cache is valid (it is either valid or NULL).
1433 * the slab_mutex must be held when looping through those caches
1435 #define for_each_memcg_cache_index(_idx) \
1436 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1438 static inline bool memcg_kmem_enabled(void)
1440 return static_branch_unlikely(&memcg_kmem_enabled_key);
1443 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1445 if (memcg_kmem_enabled())
1446 return __memcg_kmem_charge(page, gfp, order);
1450 static inline void memcg_kmem_uncharge(struct page *page, int order)
1452 if (memcg_kmem_enabled())
1453 __memcg_kmem_uncharge(page, order);
1456 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1457 int order, struct mem_cgroup *memcg)
1459 if (memcg_kmem_enabled())
1460 return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1464 static inline void memcg_kmem_uncharge_memcg(struct page *page, int order,
1465 struct mem_cgroup *memcg)
1467 if (memcg_kmem_enabled())
1468 __memcg_kmem_uncharge_memcg(memcg, 1 << order);
1472 * helper for accessing a memcg's index. It will be used as an index in the
1473 * child cache array in kmem_cache, and also to derive its name. This function
1474 * will return -1 when this is not a kmem-limited memcg.
1476 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1478 return memcg ? memcg->kmemcg_id : -1;
1481 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1485 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1490 static inline void memcg_kmem_uncharge(struct page *page, int order)
1494 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1499 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1503 #define for_each_memcg_cache_index(_idx) \
1506 static inline bool memcg_kmem_enabled(void)
1511 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1516 static inline void memcg_get_cache_ids(void)
1520 static inline void memcg_put_cache_ids(void)
1524 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1529 #endif /* CONFIG_MEMCG_KMEM */
1531 #endif /* _LINUX_MEMCONTROL_H */