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 /* Cgroup-specific events, on top of universal VM events */
52 enum memcg_event_item
{
53 MEMCG_LOW
= NR_VM_EVENT_ITEMS
,
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
[MEMCG_NR_EVENTS
];
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
;
111 struct lruvec_stat __percpu
*lruvec_stat
;
112 unsigned long lru_zone_size
[MAX_NR_ZONES
][NR_LRU_LISTS
];
114 struct mem_cgroup_reclaim_iter iter
[DEF_PRIORITY
+ 1];
116 struct rb_node tree_node
; /* RB tree node */
117 unsigned long usage_in_excess
;/* Set to the value by which */
118 /* the soft limit is exceeded*/
120 struct mem_cgroup
*memcg
; /* Back pointer, we cannot */
121 /* use container_of */
124 struct mem_cgroup_threshold
{
125 struct eventfd_ctx
*eventfd
;
126 unsigned long threshold
;
130 struct mem_cgroup_threshold_ary
{
131 /* An array index points to threshold just below or equal to usage. */
132 int current_threshold
;
133 /* Size of entries[] */
135 /* Array of thresholds */
136 struct mem_cgroup_threshold entries
[0];
139 struct mem_cgroup_thresholds
{
140 /* Primary thresholds array */
141 struct mem_cgroup_threshold_ary
*primary
;
143 * Spare threshold array.
144 * This is needed to make mem_cgroup_unregister_event() "never fail".
145 * It must be able to store at least primary->size - 1 entries.
147 struct mem_cgroup_threshold_ary
*spare
;
150 enum memcg_kmem_state
{
157 * The memory controller data structure. The memory controller controls both
158 * page cache and RSS per cgroup. We would eventually like to provide
159 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
160 * to help the administrator determine what knobs to tune.
163 struct cgroup_subsys_state css
;
165 /* Private memcg ID. Used to ID objects that outlive the cgroup */
166 struct mem_cgroup_id id
;
168 /* Accounted resources */
169 struct page_counter memory
;
170 struct page_counter swap
;
172 /* Legacy consumer-oriented counters */
173 struct page_counter memsw
;
174 struct page_counter kmem
;
175 struct page_counter tcpmem
;
177 /* Normal memory consumption range */
181 /* Range enforcement for interrupt charges */
182 struct work_struct high_work
;
184 unsigned long soft_limit
;
186 /* vmpressure notifications */
187 struct vmpressure vmpressure
;
190 * Should the accounting and control be hierarchical, per subtree?
194 /* protected by memcg_oom_lock */
199 /* OOM-Killer disable */
200 int oom_kill_disable
;
202 /* handle for "memory.events" */
203 struct cgroup_file events_file
;
205 /* protect arrays of thresholds */
206 struct mutex thresholds_lock
;
208 /* thresholds for memory usage. RCU-protected */
209 struct mem_cgroup_thresholds thresholds
;
211 /* thresholds for mem+swap usage. RCU-protected */
212 struct mem_cgroup_thresholds memsw_thresholds
;
214 /* For oom notifier event fd */
215 struct list_head oom_notify
;
218 * Should we move charges of a task when a task is moved into this
219 * mem_cgroup ? And what type of charges should we move ?
221 unsigned long move_charge_at_immigrate
;
223 * set > 0 if pages under this cgroup are moving to other cgroup.
225 atomic_t moving_account
;
226 /* taken only while moving_account > 0 */
227 spinlock_t move_lock
;
228 struct task_struct
*move_lock_task
;
229 unsigned long move_lock_flags
;
233 struct mem_cgroup_stat_cpu __percpu
*stat
;
235 unsigned long socket_pressure
;
237 /* Legacy tcp memory accounting */
242 /* Index in the kmem_cache->memcg_params.memcg_caches array */
244 enum memcg_kmem_state kmem_state
;
245 struct list_head kmem_caches
;
248 int last_scanned_node
;
250 nodemask_t scan_nodes
;
251 atomic_t numainfo_events
;
252 atomic_t numainfo_updating
;
255 #ifdef CONFIG_CGROUP_WRITEBACK
256 struct list_head cgwb_list
;
257 struct wb_domain cgwb_domain
;
260 /* List of events which userspace want to receive */
261 struct list_head event_list
;
262 spinlock_t event_list_lock
;
264 struct mem_cgroup_per_node
*nodeinfo
[0];
265 /* WARNING: nodeinfo must be the last member here */
268 extern struct mem_cgroup
*root_mem_cgroup
;
270 static inline bool mem_cgroup_disabled(void)
272 return !cgroup_subsys_enabled(memory_cgrp_subsys
);
275 static inline void mem_cgroup_event(struct mem_cgroup
*memcg
,
276 enum memcg_event_item event
)
278 this_cpu_inc(memcg
->stat
->events
[event
]);
279 cgroup_file_notify(&memcg
->events_file
);
282 bool mem_cgroup_low(struct mem_cgroup
*root
, struct mem_cgroup
*memcg
);
284 int mem_cgroup_try_charge(struct page
*page
, struct mm_struct
*mm
,
285 gfp_t gfp_mask
, struct mem_cgroup
**memcgp
,
287 void mem_cgroup_commit_charge(struct page
*page
, struct mem_cgroup
*memcg
,
288 bool lrucare
, bool compound
);
289 void mem_cgroup_cancel_charge(struct page
*page
, struct mem_cgroup
*memcg
,
291 void mem_cgroup_uncharge(struct page
*page
);
292 void mem_cgroup_uncharge_list(struct list_head
*page_list
);
294 void mem_cgroup_migrate(struct page
*oldpage
, struct page
*newpage
);
296 static struct mem_cgroup_per_node
*
297 mem_cgroup_nodeinfo(struct mem_cgroup
*memcg
, int nid
)
299 return memcg
->nodeinfo
[nid
];
303 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
304 * @node: node of the wanted lruvec
305 * @memcg: memcg of the wanted lruvec
307 * Returns the lru list vector holding pages for a given @node or a given
308 * @memcg and @zone. This can be the node lruvec, if the memory controller
311 static inline struct lruvec
*mem_cgroup_lruvec(struct pglist_data
*pgdat
,
312 struct mem_cgroup
*memcg
)
314 struct mem_cgroup_per_node
*mz
;
315 struct lruvec
*lruvec
;
317 if (mem_cgroup_disabled()) {
318 lruvec
= node_lruvec(pgdat
);
322 mz
= mem_cgroup_nodeinfo(memcg
, pgdat
->node_id
);
323 lruvec
= &mz
->lruvec
;
326 * Since a node can be onlined after the mem_cgroup was created,
327 * we have to be prepared to initialize lruvec->pgdat here;
328 * and if offlined then reonlined, we need to reinitialize it.
330 if (unlikely(lruvec
->pgdat
!= pgdat
))
331 lruvec
->pgdat
= pgdat
;
335 struct lruvec
*mem_cgroup_page_lruvec(struct page
*, struct pglist_data
*);
337 bool task_in_mem_cgroup(struct task_struct
*task
, struct mem_cgroup
*memcg
);
338 struct mem_cgroup
*mem_cgroup_from_task(struct task_struct
*p
);
341 struct mem_cgroup
*mem_cgroup_from_css(struct cgroup_subsys_state
*css
){
342 return css
? container_of(css
, struct mem_cgroup
, css
) : NULL
;
345 #define mem_cgroup_from_counter(counter, member) \
346 container_of(counter, struct mem_cgroup, member)
348 struct mem_cgroup
*mem_cgroup_iter(struct mem_cgroup
*,
350 struct mem_cgroup_reclaim_cookie
*);
351 void mem_cgroup_iter_break(struct mem_cgroup
*, struct mem_cgroup
*);
352 int mem_cgroup_scan_tasks(struct mem_cgroup
*,
353 int (*)(struct task_struct
*, void *), void *);
355 static inline unsigned short mem_cgroup_id(struct mem_cgroup
*memcg
)
357 if (mem_cgroup_disabled())
362 struct mem_cgroup
*mem_cgroup_from_id(unsigned short id
);
364 static inline struct mem_cgroup
*lruvec_memcg(struct lruvec
*lruvec
)
366 struct mem_cgroup_per_node
*mz
;
368 if (mem_cgroup_disabled())
371 mz
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
376 * parent_mem_cgroup - find the accounting parent of a memcg
377 * @memcg: memcg whose parent to find
379 * Returns the parent memcg, or NULL if this is the root or the memory
380 * controller is in legacy no-hierarchy mode.
382 static inline struct mem_cgroup
*parent_mem_cgroup(struct mem_cgroup
*memcg
)
384 if (!memcg
->memory
.parent
)
386 return mem_cgroup_from_counter(memcg
->memory
.parent
, memory
);
389 static inline bool mem_cgroup_is_descendant(struct mem_cgroup
*memcg
,
390 struct mem_cgroup
*root
)
394 if (!root
->use_hierarchy
)
396 return cgroup_is_descendant(memcg
->css
.cgroup
, root
->css
.cgroup
);
399 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
400 struct mem_cgroup
*memcg
)
402 struct mem_cgroup
*task_memcg
;
406 task_memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
408 match
= mem_cgroup_is_descendant(task_memcg
, memcg
);
413 struct cgroup_subsys_state
*mem_cgroup_css_from_page(struct page
*page
);
414 ino_t
page_cgroup_ino(struct page
*page
);
416 static inline bool mem_cgroup_online(struct mem_cgroup
*memcg
)
418 if (mem_cgroup_disabled())
420 return !!(memcg
->css
.flags
& CSS_ONLINE
);
424 * For memory reclaim.
426 int mem_cgroup_select_victim_node(struct mem_cgroup
*memcg
);
428 void mem_cgroup_update_lru_size(struct lruvec
*lruvec
, enum lru_list lru
,
429 int zid
, int nr_pages
);
431 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup
*memcg
,
432 int nid
, unsigned int lru_mask
);
435 unsigned long mem_cgroup_get_lru_size(struct lruvec
*lruvec
, enum lru_list lru
)
437 struct mem_cgroup_per_node
*mz
;
438 unsigned long nr_pages
= 0;
441 mz
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
442 for (zid
= 0; zid
< MAX_NR_ZONES
; zid
++)
443 nr_pages
+= mz
->lru_zone_size
[zid
][lru
];
448 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec
*lruvec
,
449 enum lru_list lru
, int zone_idx
)
451 struct mem_cgroup_per_node
*mz
;
453 mz
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
454 return mz
->lru_zone_size
[zone_idx
][lru
];
457 void mem_cgroup_handle_over_high(void);
459 unsigned long mem_cgroup_get_limit(struct mem_cgroup
*memcg
);
461 void mem_cgroup_print_oom_info(struct mem_cgroup
*memcg
,
462 struct task_struct
*p
);
464 static inline void mem_cgroup_oom_enable(void)
466 WARN_ON(current
->memcg_may_oom
);
467 current
->memcg_may_oom
= 1;
470 static inline void mem_cgroup_oom_disable(void)
472 WARN_ON(!current
->memcg_may_oom
);
473 current
->memcg_may_oom
= 0;
476 static inline bool task_in_memcg_oom(struct task_struct
*p
)
478 return p
->memcg_in_oom
;
481 bool mem_cgroup_oom_synchronize(bool wait
);
483 #ifdef CONFIG_MEMCG_SWAP
484 extern int do_swap_account
;
487 struct mem_cgroup
*lock_page_memcg(struct page
*page
);
488 void __unlock_page_memcg(struct mem_cgroup
*memcg
);
489 void unlock_page_memcg(struct page
*page
);
491 /* idx can be of type enum memcg_stat_item or node_stat_item */
492 static inline unsigned long memcg_page_state(struct mem_cgroup
*memcg
,
498 for_each_possible_cpu(cpu
)
499 val
+= per_cpu(memcg
->stat
->count
[idx
], cpu
);
507 /* idx can be of type enum memcg_stat_item or node_stat_item */
508 static inline void __mod_memcg_state(struct mem_cgroup
*memcg
,
511 if (!mem_cgroup_disabled())
512 __this_cpu_add(memcg
->stat
->count
[idx
], val
);
515 /* idx can be of type enum memcg_stat_item or node_stat_item */
516 static inline void mod_memcg_state(struct mem_cgroup
*memcg
,
519 if (!mem_cgroup_disabled())
520 this_cpu_add(memcg
->stat
->count
[idx
], val
);
524 * mod_memcg_page_state - update page state statistics
526 * @idx: page state item to account
527 * @val: number of pages (positive or negative)
529 * The @page must be locked or the caller must use lock_page_memcg()
530 * to prevent double accounting when the page is concurrently being
531 * moved to another memcg:
533 * lock_page(page) or lock_page_memcg(page)
534 * if (TestClearPageState(page))
535 * mod_memcg_page_state(page, state, -1);
536 * unlock_page(page) or unlock_page_memcg(page)
538 * Kernel pages are an exception to this, since they'll never move.
540 static inline void __mod_memcg_page_state(struct page
*page
,
543 if (page
->mem_cgroup
)
544 __mod_memcg_state(page
->mem_cgroup
, idx
, val
);
547 static inline void mod_memcg_page_state(struct page
*page
,
550 if (page
->mem_cgroup
)
551 mod_memcg_state(page
->mem_cgroup
, idx
, val
);
554 static inline unsigned long lruvec_page_state(struct lruvec
*lruvec
,
555 enum node_stat_item idx
)
557 struct mem_cgroup_per_node
*pn
;
561 if (mem_cgroup_disabled())
562 return node_page_state(lruvec_pgdat(lruvec
), idx
);
564 pn
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
565 for_each_possible_cpu(cpu
)
566 val
+= per_cpu(pn
->lruvec_stat
->count
[idx
], cpu
);
574 static inline void __mod_lruvec_state(struct lruvec
*lruvec
,
575 enum node_stat_item idx
, int val
)
577 struct mem_cgroup_per_node
*pn
;
579 __mod_node_page_state(lruvec_pgdat(lruvec
), idx
, val
);
580 if (mem_cgroup_disabled())
582 pn
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
583 __mod_memcg_state(pn
->memcg
, idx
, val
);
584 __this_cpu_add(pn
->lruvec_stat
->count
[idx
], val
);
587 static inline void mod_lruvec_state(struct lruvec
*lruvec
,
588 enum node_stat_item idx
, int val
)
590 struct mem_cgroup_per_node
*pn
;
592 mod_node_page_state(lruvec_pgdat(lruvec
), idx
, val
);
593 if (mem_cgroup_disabled())
595 pn
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
596 mod_memcg_state(pn
->memcg
, idx
, val
);
597 this_cpu_add(pn
->lruvec_stat
->count
[idx
], val
);
600 static inline void __mod_lruvec_page_state(struct page
*page
,
601 enum node_stat_item idx
, int val
)
603 struct mem_cgroup_per_node
*pn
;
605 __mod_node_page_state(page_pgdat(page
), idx
, val
);
606 if (mem_cgroup_disabled() || !page
->mem_cgroup
)
608 __mod_memcg_state(page
->mem_cgroup
, idx
, val
);
609 pn
= page
->mem_cgroup
->nodeinfo
[page_to_nid(page
)];
610 __this_cpu_add(pn
->lruvec_stat
->count
[idx
], val
);
613 static inline void mod_lruvec_page_state(struct page
*page
,
614 enum node_stat_item idx
, int val
)
616 struct mem_cgroup_per_node
*pn
;
618 mod_node_page_state(page_pgdat(page
), idx
, val
);
619 if (mem_cgroup_disabled() || !page
->mem_cgroup
)
621 mod_memcg_state(page
->mem_cgroup
, idx
, val
);
622 pn
= page
->mem_cgroup
->nodeinfo
[page_to_nid(page
)];
623 this_cpu_add(pn
->lruvec_stat
->count
[idx
], val
);
626 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t
*pgdat
, int order
,
628 unsigned long *total_scanned
);
630 static inline void count_memcg_events(struct mem_cgroup
*memcg
,
631 enum vm_event_item idx
,
634 if (!mem_cgroup_disabled())
635 this_cpu_add(memcg
->stat
->events
[idx
], count
);
638 /* idx can be of type enum memcg_stat_item or node_stat_item */
639 static inline void count_memcg_page_event(struct page
*page
,
642 if (page
->mem_cgroup
)
643 count_memcg_events(page
->mem_cgroup
, idx
, 1);
646 static inline void count_memcg_event_mm(struct mm_struct
*mm
,
647 enum vm_event_item idx
)
649 struct mem_cgroup
*memcg
;
651 if (mem_cgroup_disabled())
655 memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
657 this_cpu_inc(memcg
->stat
->events
[idx
]);
659 cgroup_file_notify(&memcg
->events_file
);
663 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
664 void mem_cgroup_split_huge_fixup(struct page
*head
);
667 #else /* CONFIG_MEMCG */
669 #define MEM_CGROUP_ID_SHIFT 0
670 #define MEM_CGROUP_ID_MAX 0
674 static inline bool mem_cgroup_disabled(void)
679 static inline void mem_cgroup_event(struct mem_cgroup
*memcg
,
680 enum memcg_event_item event
)
684 static inline bool mem_cgroup_low(struct mem_cgroup
*root
,
685 struct mem_cgroup
*memcg
)
690 static inline int mem_cgroup_try_charge(struct page
*page
, struct mm_struct
*mm
,
692 struct mem_cgroup
**memcgp
,
699 static inline void mem_cgroup_commit_charge(struct page
*page
,
700 struct mem_cgroup
*memcg
,
701 bool lrucare
, bool compound
)
705 static inline void mem_cgroup_cancel_charge(struct page
*page
,
706 struct mem_cgroup
*memcg
,
711 static inline void mem_cgroup_uncharge(struct page
*page
)
715 static inline void mem_cgroup_uncharge_list(struct list_head
*page_list
)
719 static inline void mem_cgroup_migrate(struct page
*old
, struct page
*new)
723 static inline struct lruvec
*mem_cgroup_lruvec(struct pglist_data
*pgdat
,
724 struct mem_cgroup
*memcg
)
726 return node_lruvec(pgdat
);
729 static inline struct lruvec
*mem_cgroup_page_lruvec(struct page
*page
,
730 struct pglist_data
*pgdat
)
732 return &pgdat
->lruvec
;
735 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
736 struct mem_cgroup
*memcg
)
741 static inline bool task_in_mem_cgroup(struct task_struct
*task
,
742 const struct mem_cgroup
*memcg
)
747 static inline struct mem_cgroup
*
748 mem_cgroup_iter(struct mem_cgroup
*root
,
749 struct mem_cgroup
*prev
,
750 struct mem_cgroup_reclaim_cookie
*reclaim
)
755 static inline void mem_cgroup_iter_break(struct mem_cgroup
*root
,
756 struct mem_cgroup
*prev
)
760 static inline int mem_cgroup_scan_tasks(struct mem_cgroup
*memcg
,
761 int (*fn
)(struct task_struct
*, void *), void *arg
)
766 static inline unsigned short mem_cgroup_id(struct mem_cgroup
*memcg
)
771 static inline struct mem_cgroup
*mem_cgroup_from_id(unsigned short id
)
774 /* XXX: This should always return root_mem_cgroup */
778 static inline struct mem_cgroup
*lruvec_memcg(struct lruvec
*lruvec
)
783 static inline bool mem_cgroup_online(struct mem_cgroup
*memcg
)
788 static inline unsigned long
789 mem_cgroup_get_lru_size(struct lruvec
*lruvec
, enum lru_list lru
)
794 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec
*lruvec
,
795 enum lru_list lru
, int zone_idx
)
800 static inline unsigned long
801 mem_cgroup_node_nr_lru_pages(struct mem_cgroup
*memcg
,
802 int nid
, unsigned int lru_mask
)
807 static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup
*memcg
)
813 mem_cgroup_print_oom_info(struct mem_cgroup
*memcg
, struct task_struct
*p
)
817 static inline struct mem_cgroup
*lock_page_memcg(struct page
*page
)
822 static inline void __unlock_page_memcg(struct mem_cgroup
*memcg
)
826 static inline void unlock_page_memcg(struct page
*page
)
830 static inline void mem_cgroup_handle_over_high(void)
834 static inline void mem_cgroup_oom_enable(void)
838 static inline void mem_cgroup_oom_disable(void)
842 static inline bool task_in_memcg_oom(struct task_struct
*p
)
847 static inline bool mem_cgroup_oom_synchronize(bool wait
)
852 static inline unsigned long memcg_page_state(struct mem_cgroup
*memcg
,
858 static inline void __mod_memcg_state(struct mem_cgroup
*memcg
,
864 static inline void mod_memcg_state(struct mem_cgroup
*memcg
,
870 static inline void __mod_memcg_page_state(struct page
*page
,
876 static inline void mod_memcg_page_state(struct page
*page
,
882 static inline unsigned long lruvec_page_state(struct lruvec
*lruvec
,
883 enum node_stat_item idx
)
885 return node_page_state(lruvec_pgdat(lruvec
), idx
);
888 static inline void __mod_lruvec_state(struct lruvec
*lruvec
,
889 enum node_stat_item idx
, int val
)
891 __mod_node_page_state(lruvec_pgdat(lruvec
), idx
, val
);
894 static inline void mod_lruvec_state(struct lruvec
*lruvec
,
895 enum node_stat_item idx
, int val
)
897 mod_node_page_state(lruvec_pgdat(lruvec
), idx
, val
);
900 static inline void __mod_lruvec_page_state(struct page
*page
,
901 enum node_stat_item idx
, int val
)
903 __mod_node_page_state(page_pgdat(page
), idx
, val
);
906 static inline void mod_lruvec_page_state(struct page
*page
,
907 enum node_stat_item idx
, int val
)
909 mod_node_page_state(page_pgdat(page
), idx
, val
);
913 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t
*pgdat
, int order
,
915 unsigned long *total_scanned
)
920 static inline void mem_cgroup_split_huge_fixup(struct page
*head
)
924 static inline void count_memcg_events(struct mem_cgroup
*memcg
,
925 enum vm_event_item idx
,
930 static inline void count_memcg_page_event(struct page
*page
,
936 void count_memcg_event_mm(struct mm_struct
*mm
, enum vm_event_item idx
)
939 #endif /* CONFIG_MEMCG */
941 /* idx can be of type enum memcg_stat_item or node_stat_item */
942 static inline void __inc_memcg_state(struct mem_cgroup
*memcg
,
945 __mod_memcg_state(memcg
, idx
, 1);
948 /* idx can be of type enum memcg_stat_item or node_stat_item */
949 static inline void __dec_memcg_state(struct mem_cgroup
*memcg
,
952 __mod_memcg_state(memcg
, idx
, -1);
955 /* idx can be of type enum memcg_stat_item or node_stat_item */
956 static inline void __inc_memcg_page_state(struct page
*page
,
959 __mod_memcg_page_state(page
, idx
, 1);
962 /* idx can be of type enum memcg_stat_item or node_stat_item */
963 static inline void __dec_memcg_page_state(struct page
*page
,
966 __mod_memcg_page_state(page
, idx
, -1);
969 static inline void __inc_lruvec_state(struct lruvec
*lruvec
,
970 enum node_stat_item idx
)
972 __mod_lruvec_state(lruvec
, idx
, 1);
975 static inline void __dec_lruvec_state(struct lruvec
*lruvec
,
976 enum node_stat_item idx
)
978 __mod_lruvec_state(lruvec
, idx
, -1);
981 static inline void __inc_lruvec_page_state(struct page
*page
,
982 enum node_stat_item idx
)
984 __mod_lruvec_page_state(page
, idx
, 1);
987 static inline void __dec_lruvec_page_state(struct page
*page
,
988 enum node_stat_item idx
)
990 __mod_lruvec_page_state(page
, idx
, -1);
993 /* idx can be of type enum memcg_stat_item or node_stat_item */
994 static inline void inc_memcg_state(struct mem_cgroup
*memcg
,
997 mod_memcg_state(memcg
, idx
, 1);
1000 /* idx can be of type enum memcg_stat_item or node_stat_item */
1001 static inline void dec_memcg_state(struct mem_cgroup
*memcg
,
1004 mod_memcg_state(memcg
, idx
, -1);
1007 /* idx can be of type enum memcg_stat_item or node_stat_item */
1008 static inline void inc_memcg_page_state(struct page
*page
,
1011 mod_memcg_page_state(page
, idx
, 1);
1014 /* idx can be of type enum memcg_stat_item or node_stat_item */
1015 static inline void dec_memcg_page_state(struct page
*page
,
1018 mod_memcg_page_state(page
, idx
, -1);
1021 static inline void inc_lruvec_state(struct lruvec
*lruvec
,
1022 enum node_stat_item idx
)
1024 mod_lruvec_state(lruvec
, idx
, 1);
1027 static inline void dec_lruvec_state(struct lruvec
*lruvec
,
1028 enum node_stat_item idx
)
1030 mod_lruvec_state(lruvec
, idx
, -1);
1033 static inline void inc_lruvec_page_state(struct page
*page
,
1034 enum node_stat_item idx
)
1036 mod_lruvec_page_state(page
, idx
, 1);
1039 static inline void dec_lruvec_page_state(struct page
*page
,
1040 enum node_stat_item idx
)
1042 mod_lruvec_page_state(page
, idx
, -1);
1045 #ifdef CONFIG_CGROUP_WRITEBACK
1047 struct list_head
*mem_cgroup_cgwb_list(struct mem_cgroup
*memcg
);
1048 struct wb_domain
*mem_cgroup_wb_domain(struct bdi_writeback
*wb
);
1049 void mem_cgroup_wb_stats(struct bdi_writeback
*wb
, unsigned long *pfilepages
,
1050 unsigned long *pheadroom
, unsigned long *pdirty
,
1051 unsigned long *pwriteback
);
1053 #else /* CONFIG_CGROUP_WRITEBACK */
1055 static inline struct wb_domain
*mem_cgroup_wb_domain(struct bdi_writeback
*wb
)
1060 static inline void mem_cgroup_wb_stats(struct bdi_writeback
*wb
,
1061 unsigned long *pfilepages
,
1062 unsigned long *pheadroom
,
1063 unsigned long *pdirty
,
1064 unsigned long *pwriteback
)
1068 #endif /* CONFIG_CGROUP_WRITEBACK */
1071 bool mem_cgroup_charge_skmem(struct mem_cgroup
*memcg
, unsigned int nr_pages
);
1072 void mem_cgroup_uncharge_skmem(struct mem_cgroup
*memcg
, unsigned int nr_pages
);
1074 extern struct static_key_false memcg_sockets_enabled_key
;
1075 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1076 void mem_cgroup_sk_alloc(struct sock
*sk
);
1077 void mem_cgroup_sk_free(struct sock
*sk
);
1078 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup
*memcg
)
1080 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys
) && memcg
->tcpmem_pressure
)
1083 if (time_before(jiffies
, memcg
->socket_pressure
))
1085 } while ((memcg
= parent_mem_cgroup(memcg
)));
1089 #define mem_cgroup_sockets_enabled 0
1090 static inline void mem_cgroup_sk_alloc(struct sock
*sk
) { };
1091 static inline void mem_cgroup_sk_free(struct sock
*sk
) { };
1092 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup
*memcg
)
1098 struct kmem_cache
*memcg_kmem_get_cache(struct kmem_cache
*cachep
);
1099 void memcg_kmem_put_cache(struct kmem_cache
*cachep
);
1100 int memcg_kmem_charge_memcg(struct page
*page
, gfp_t gfp
, int order
,
1101 struct mem_cgroup
*memcg
);
1102 int memcg_kmem_charge(struct page
*page
, gfp_t gfp
, int order
);
1103 void memcg_kmem_uncharge(struct page
*page
, int order
);
1105 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1106 extern struct static_key_false memcg_kmem_enabled_key
;
1107 extern struct workqueue_struct
*memcg_kmem_cache_wq
;
1109 extern int memcg_nr_cache_ids
;
1110 void memcg_get_cache_ids(void);
1111 void memcg_put_cache_ids(void);
1114 * Helper macro to loop through all memcg-specific caches. Callers must still
1115 * check if the cache is valid (it is either valid or NULL).
1116 * the slab_mutex must be held when looping through those caches
1118 #define for_each_memcg_cache_index(_idx) \
1119 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1121 static inline bool memcg_kmem_enabled(void)
1123 return static_branch_unlikely(&memcg_kmem_enabled_key
);
1127 * helper for accessing a memcg's index. It will be used as an index in the
1128 * child cache array in kmem_cache, and also to derive its name. This function
1129 * will return -1 when this is not a kmem-limited memcg.
1131 static inline int memcg_cache_id(struct mem_cgroup
*memcg
)
1133 return memcg
? memcg
->kmemcg_id
: -1;
1137 #define for_each_memcg_cache_index(_idx) \
1140 static inline bool memcg_kmem_enabled(void)
1145 static inline int memcg_cache_id(struct mem_cgroup
*memcg
)
1150 static inline void memcg_get_cache_ids(void)
1154 static inline void memcg_put_cache_ids(void)
1158 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1160 #endif /* _LINUX_MEMCONTROL_H */