mm: memcontrol: generalize locking for the page->mem_cgroup binding
[deliverable/linux.git] / include / linux / memcontrol.h
CommitLineData
8cdea7c0
BS
1/* memcontrol.h - Memory Controller
2 *
3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
5 *
78fb7466
PE
6 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org>
8 *
8cdea7c0
BS
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.
13 *
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.
18 */
19
20#ifndef _LINUX_MEMCONTROL_H
21#define _LINUX_MEMCONTROL_H
f8d66542 22#include <linux/cgroup.h>
456f998e 23#include <linux/vm_event_item.h>
7ae1e1d0 24#include <linux/hardirq.h>
a8964b9b 25#include <linux/jump_label.h>
33398cf2
MH
26#include <linux/page_counter.h>
27#include <linux/vmpressure.h>
28#include <linux/eventfd.h>
29#include <linux/mmzone.h>
30#include <linux/writeback.h>
456f998e 31
78fb7466 32struct mem_cgroup;
8697d331
BS
33struct page;
34struct mm_struct;
2633d7a0 35struct kmem_cache;
78fb7466 36
68b4876d
SZ
37/*
38 * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
39 * These two lists should keep in accord with each other.
40 */
41enum mem_cgroup_stat_index {
42 /*
43 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
44 */
45 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
46 MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */
47 MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */
48 MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */
c4843a75 49 MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */
3ea67d06 50 MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
68b4876d
SZ
51 MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
52 MEM_CGROUP_STAT_NSTATS,
b2807f07 53 /* default hierarchy stats */
c792e824 54 MEMCG_SOCK = MEM_CGROUP_STAT_NSTATS,
b2807f07 55 MEMCG_NR_STAT,
2a7106f2
GT
56};
57
5660048c
JW
58struct mem_cgroup_reclaim_cookie {
59 struct zone *zone;
60 int priority;
61 unsigned int generation;
62};
63
241994ed
JW
64enum mem_cgroup_events_index {
65 MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */
66 MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */
67 MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */
68 MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */
69 MEM_CGROUP_EVENTS_NSTATS,
70 /* default hierarchy events */
71 MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
72 MEMCG_HIGH,
73 MEMCG_MAX,
74 MEMCG_OOM,
75 MEMCG_NR_EVENTS,
76};
77
33398cf2
MH
78/*
79 * Per memcg event counter is incremented at every pagein/pageout. With THP,
80 * it will be incremated by the number of pages. This counter is used for
81 * for trigger some periodic events. This is straightforward and better
82 * than using jiffies etc. to handle periodic memcg event.
83 */
84enum mem_cgroup_events_target {
85 MEM_CGROUP_TARGET_THRESH,
86 MEM_CGROUP_TARGET_SOFTLIMIT,
87 MEM_CGROUP_TARGET_NUMAINFO,
88 MEM_CGROUP_NTARGETS,
89};
90
c255a458 91#ifdef CONFIG_MEMCG
33398cf2 92struct mem_cgroup_stat_cpu {
b2807f07 93 long count[MEMCG_NR_STAT];
33398cf2
MH
94 unsigned long events[MEMCG_NR_EVENTS];
95 unsigned long nr_page_events;
96 unsigned long targets[MEM_CGROUP_NTARGETS];
97};
98
99struct mem_cgroup_reclaim_iter {
100 struct mem_cgroup *position;
101 /* scan generation, increased every round-trip */
102 unsigned int generation;
103};
104
105/*
106 * per-zone information in memory controller.
107 */
108struct mem_cgroup_per_zone {
109 struct lruvec lruvec;
110 unsigned long lru_size[NR_LRU_LISTS];
111
112 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
113
114 struct rb_node tree_node; /* RB tree node */
115 unsigned long usage_in_excess;/* Set to the value by which */
116 /* the soft limit is exceeded*/
117 bool on_tree;
118 struct mem_cgroup *memcg; /* Back pointer, we cannot */
119 /* use container_of */
120};
121
122struct mem_cgroup_per_node {
123 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
124};
125
126struct mem_cgroup_threshold {
127 struct eventfd_ctx *eventfd;
128 unsigned long threshold;
129};
130
131/* For threshold */
132struct mem_cgroup_threshold_ary {
133 /* An array index points to threshold just below or equal to usage. */
134 int current_threshold;
135 /* Size of entries[] */
136 unsigned int size;
137 /* Array of thresholds */
138 struct mem_cgroup_threshold entries[0];
139};
140
141struct mem_cgroup_thresholds {
142 /* Primary thresholds array */
143 struct mem_cgroup_threshold_ary *primary;
144 /*
145 * Spare threshold array.
146 * This is needed to make mem_cgroup_unregister_event() "never fail".
147 * It must be able to store at least primary->size - 1 entries.
148 */
149 struct mem_cgroup_threshold_ary *spare;
150};
151
567e9ab2
JW
152enum memcg_kmem_state {
153 KMEM_NONE,
154 KMEM_ALLOCATED,
155 KMEM_ONLINE,
156};
157
33398cf2
MH
158/*
159 * The memory controller data structure. The memory controller controls both
160 * page cache and RSS per cgroup. We would eventually like to provide
161 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
162 * to help the administrator determine what knobs to tune.
163 */
164struct mem_cgroup {
165 struct cgroup_subsys_state css;
166
167 /* Accounted resources */
168 struct page_counter memory;
37e84351 169 struct page_counter swap;
0db15298
JW
170
171 /* Legacy consumer-oriented counters */
33398cf2
MH
172 struct page_counter memsw;
173 struct page_counter kmem;
0db15298 174 struct page_counter tcpmem;
33398cf2
MH
175
176 /* Normal memory consumption range */
177 unsigned long low;
178 unsigned long high;
179
f7e1cb6e
JW
180 /* Range enforcement for interrupt charges */
181 struct work_struct high_work;
182
33398cf2
MH
183 unsigned long soft_limit;
184
185 /* vmpressure notifications */
186 struct vmpressure vmpressure;
187
33398cf2
MH
188 /*
189 * Should the accounting and control be hierarchical, per subtree?
190 */
191 bool use_hierarchy;
192
193 /* protected by memcg_oom_lock */
194 bool oom_lock;
195 int under_oom;
196
197 int swappiness;
198 /* OOM-Killer disable */
199 int oom_kill_disable;
200
472912a2
TH
201 /* handle for "memory.events" */
202 struct cgroup_file events_file;
203
33398cf2
MH
204 /* protect arrays of thresholds */
205 struct mutex thresholds_lock;
206
207 /* thresholds for memory usage. RCU-protected */
208 struct mem_cgroup_thresholds thresholds;
209
210 /* thresholds for mem+swap usage. RCU-protected */
211 struct mem_cgroup_thresholds memsw_thresholds;
212
213 /* For oom notifier event fd */
214 struct list_head oom_notify;
215
216 /*
217 * Should we move charges of a task when a task is moved into this
218 * mem_cgroup ? And what type of charges should we move ?
219 */
220 unsigned long move_charge_at_immigrate;
221 /*
222 * set > 0 if pages under this cgroup are moving to other cgroup.
223 */
224 atomic_t moving_account;
225 /* taken only while moving_account > 0 */
226 spinlock_t move_lock;
227 struct task_struct *move_lock_task;
228 unsigned long move_lock_flags;
229 /*
230 * percpu counter.
231 */
232 struct mem_cgroup_stat_cpu __percpu *stat;
33398cf2 233
d886f4e4
JW
234 unsigned long socket_pressure;
235
236 /* Legacy tcp memory accounting */
0db15298
JW
237 bool tcpmem_active;
238 int tcpmem_pressure;
d886f4e4 239
127424c8 240#ifndef CONFIG_SLOB
33398cf2
MH
241 /* Index in the kmem_cache->memcg_params.memcg_caches array */
242 int kmemcg_id;
567e9ab2 243 enum memcg_kmem_state kmem_state;
33398cf2
MH
244#endif
245
246 int last_scanned_node;
247#if MAX_NUMNODES > 1
248 nodemask_t scan_nodes;
249 atomic_t numainfo_events;
250 atomic_t numainfo_updating;
251#endif
252
253#ifdef CONFIG_CGROUP_WRITEBACK
254 struct list_head cgwb_list;
255 struct wb_domain cgwb_domain;
256#endif
257
258 /* List of events which userspace want to receive */
259 struct list_head event_list;
260 spinlock_t event_list_lock;
261
262 struct mem_cgroup_per_node *nodeinfo[0];
263 /* WARNING: nodeinfo must be the last member here */
264};
7d828602
JW
265
266extern struct mem_cgroup *root_mem_cgroup;
56161634 267
33398cf2
MH
268/**
269 * mem_cgroup_events - count memory events against a cgroup
270 * @memcg: the memory cgroup
271 * @idx: the event index
272 * @nr: the number of events to account for
273 */
274static inline void mem_cgroup_events(struct mem_cgroup *memcg,
241994ed 275 enum mem_cgroup_events_index idx,
33398cf2
MH
276 unsigned int nr)
277{
278 this_cpu_add(memcg->stat->events[idx], nr);
472912a2 279 cgroup_file_notify(&memcg->events_file);
33398cf2 280}
241994ed
JW
281
282bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
283
00501b53 284int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
f627c2f5
KS
285 gfp_t gfp_mask, struct mem_cgroup **memcgp,
286 bool compound);
00501b53 287void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
f627c2f5
KS
288 bool lrucare, bool compound);
289void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
290 bool compound);
0a31bc97 291void mem_cgroup_uncharge(struct page *page);
747db954 292void mem_cgroup_uncharge_list(struct list_head *page_list);
569b846d 293
45637bab 294void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
569b846d 295
0a31bc97
JW
296struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
297struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
c9b0ed51 298
2314b42d 299bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
64219994 300struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
e993d905 301
33398cf2
MH
302static inline
303struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
304 return css ? container_of(css, struct mem_cgroup, css) : NULL;
305}
306
8e8ae645
JW
307#define mem_cgroup_from_counter(counter, member) \
308 container_of(counter, struct mem_cgroup, member)
309
33398cf2
MH
310struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
311 struct mem_cgroup *,
312 struct mem_cgroup_reclaim_cookie *);
313void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
314
8e8ae645
JW
315/**
316 * parent_mem_cgroup - find the accounting parent of a memcg
317 * @memcg: memcg whose parent to find
318 *
319 * Returns the parent memcg, or NULL if this is the root or the memory
320 * controller is in legacy no-hierarchy mode.
321 */
322static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
323{
324 if (!memcg->memory.parent)
325 return NULL;
326 return mem_cgroup_from_counter(memcg->memory.parent, memory);
327}
328
33398cf2
MH
329static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
330 struct mem_cgroup *root)
331{
332 if (root == memcg)
333 return true;
334 if (!root->use_hierarchy)
335 return false;
336 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
337}
e1aab161 338
2314b42d
JW
339static inline bool mm_match_cgroup(struct mm_struct *mm,
340 struct mem_cgroup *memcg)
2e4d4091 341{
587af308 342 struct mem_cgroup *task_memcg;
413918bb 343 bool match = false;
c3ac9a8a 344
2e4d4091 345 rcu_read_lock();
587af308 346 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
413918bb 347 if (task_memcg)
2314b42d 348 match = mem_cgroup_is_descendant(task_memcg, memcg);
2e4d4091 349 rcu_read_unlock();
c3ac9a8a 350 return match;
2e4d4091 351}
8a9f3ccd 352
64219994 353struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
2fc04524 354ino_t page_cgroup_ino(struct page *page);
d324236b 355
33398cf2
MH
356static inline bool mem_cgroup_disabled(void)
357{
fc5ed1e9 358 return !cgroup_subsys_enabled(memory_cgrp_subsys);
33398cf2 359}
5660048c 360
eb01aaab
VD
361static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
362{
363 if (mem_cgroup_disabled())
364 return true;
365 return !!(memcg->css.flags & CSS_ONLINE);
366}
367
58ae83db
KH
368/*
369 * For memory reclaim.
370 */
889976db 371int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
33398cf2
MH
372
373void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
374 int nr_pages);
375
33398cf2
MH
376static inline
377unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
378{
379 struct mem_cgroup_per_zone *mz;
380
381 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
382 return mz->lru_size[lru];
383}
384
13308ca9 385static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
33398cf2
MH
386{
387 unsigned long inactive_ratio;
388 unsigned long inactive;
389 unsigned long active;
390 unsigned long gb;
391
392 inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
393 active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
394
395 gb = (inactive + active) >> (30 - PAGE_SHIFT);
396 if (gb)
397 inactive_ratio = int_sqrt(10 * gb);
398 else
399 inactive_ratio = 1;
400
401 return inactive * inactive_ratio < active;
402}
403
b23afb93
TH
404void mem_cgroup_handle_over_high(void);
405
64219994
MH
406void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
407 struct task_struct *p);
58ae83db 408
49426420 409static inline void mem_cgroup_oom_enable(void)
519e5247 410{
626ebc41
TH
411 WARN_ON(current->memcg_may_oom);
412 current->memcg_may_oom = 1;
519e5247
JW
413}
414
49426420 415static inline void mem_cgroup_oom_disable(void)
519e5247 416{
626ebc41
TH
417 WARN_ON(!current->memcg_may_oom);
418 current->memcg_may_oom = 0;
519e5247
JW
419}
420
3812c8c8
JW
421static inline bool task_in_memcg_oom(struct task_struct *p)
422{
626ebc41 423 return p->memcg_in_oom;
3812c8c8
JW
424}
425
49426420 426bool mem_cgroup_oom_synchronize(bool wait);
3812c8c8 427
c255a458 428#ifdef CONFIG_MEMCG_SWAP
c077719b
KH
429extern int do_swap_account;
430#endif
f8d66542 431
81f8c3a4
JW
432struct mem_cgroup *lock_page_memcg(struct page *page);
433void unlock_page_memcg(struct mem_cgroup *memcg);
d7365e78 434
33398cf2
MH
435/**
436 * mem_cgroup_update_page_stat - update page state statistics
437 * @memcg: memcg to account against
438 * @idx: page state item to account
439 * @val: number of pages (positive or negative)
440 *
81f8c3a4
JW
441 * Callers must use lock_page_memcg() to prevent double accounting
442 * when the page is concurrently being moved to another memcg:
443 *
444 * memcg = lock_page_memcg(page);
445 * if (TestClearPageState(page))
446 * mem_cgroup_update_page_stat(memcg, state, -1);
447 * unlock_page_memcg(memcg);
33398cf2
MH
448 */
449static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
450 enum mem_cgroup_stat_index idx, int val)
451{
452 VM_BUG_ON(!rcu_read_lock_held());
453
454 if (memcg)
455 this_cpu_add(memcg->stat->count[idx], val);
456}
457
d7365e78 458static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
68b4876d 459 enum mem_cgroup_stat_index idx)
2a7106f2 460{
d7365e78 461 mem_cgroup_update_page_stat(memcg, idx, 1);
2a7106f2
GT
462}
463
d7365e78 464static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
68b4876d 465 enum mem_cgroup_stat_index idx)
2a7106f2 466{
d7365e78 467 mem_cgroup_update_page_stat(memcg, idx, -1);
2a7106f2
GT
468}
469
0608f43d
AM
470unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
471 gfp_t gfp_mask,
472 unsigned long *total_scanned);
a63d83f4 473
68ae564b
DR
474static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
475 enum vm_event_item idx)
476{
33398cf2
MH
477 struct mem_cgroup *memcg;
478
68ae564b
DR
479 if (mem_cgroup_disabled())
480 return;
33398cf2
MH
481
482 rcu_read_lock();
483 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
484 if (unlikely(!memcg))
485 goto out;
486
487 switch (idx) {
488 case PGFAULT:
489 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
490 break;
491 case PGMAJFAULT:
492 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
493 break;
494 default:
495 BUG();
496 }
497out:
498 rcu_read_unlock();
68ae564b 499}
ca3e0214 500#ifdef CONFIG_TRANSPARENT_HUGEPAGE
e94c8a9c 501void mem_cgroup_split_huge_fixup(struct page *head);
ca3e0214
KH
502#endif
503
c255a458 504#else /* CONFIG_MEMCG */
7a81b88c
KH
505struct mem_cgroup;
506
241994ed
JW
507static inline void mem_cgroup_events(struct mem_cgroup *memcg,
508 enum mem_cgroup_events_index idx,
509 unsigned int nr)
510{
511}
512
513static inline bool mem_cgroup_low(struct mem_cgroup *root,
514 struct mem_cgroup *memcg)
515{
516 return false;
517}
518
00501b53
JW
519static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
520 gfp_t gfp_mask,
f627c2f5
KS
521 struct mem_cgroup **memcgp,
522 bool compound)
7a81b88c 523{
00501b53 524 *memcgp = NULL;
7a81b88c
KH
525 return 0;
526}
527
00501b53
JW
528static inline void mem_cgroup_commit_charge(struct page *page,
529 struct mem_cgroup *memcg,
f627c2f5 530 bool lrucare, bool compound)
7a81b88c
KH
531{
532}
533
00501b53 534static inline void mem_cgroup_cancel_charge(struct page *page,
f627c2f5
KS
535 struct mem_cgroup *memcg,
536 bool compound)
7a81b88c
KH
537{
538}
539
0a31bc97 540static inline void mem_cgroup_uncharge(struct page *page)
569b846d
KH
541{
542}
543
747db954 544static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
8a9f3ccd
BS
545{
546}
547
45637bab 548static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
69029cd5
KH
549{
550}
551
925b7673
JW
552static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
553 struct mem_cgroup *memcg)
08e552c6 554{
925b7673 555 return &zone->lruvec;
08e552c6
KH
556}
557
fa9add64
HD
558static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
559 struct zone *zone)
66e1707b 560{
925b7673 561 return &zone->lruvec;
66e1707b
BS
562}
563
587af308 564static inline bool mm_match_cgroup(struct mm_struct *mm,
c0ff4b85 565 struct mem_cgroup *memcg)
bed7161a 566{
587af308 567 return true;
bed7161a
BS
568}
569
ffbdccf5
DR
570static inline bool task_in_mem_cgroup(struct task_struct *task,
571 const struct mem_cgroup *memcg)
4c4a2214 572{
ffbdccf5 573 return true;
4c4a2214
DR
574}
575
5660048c
JW
576static inline struct mem_cgroup *
577mem_cgroup_iter(struct mem_cgroup *root,
578 struct mem_cgroup *prev,
579 struct mem_cgroup_reclaim_cookie *reclaim)
580{
581 return NULL;
582}
583
584static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
585 struct mem_cgroup *prev)
586{
587}
588
f8d66542
HT
589static inline bool mem_cgroup_disabled(void)
590{
591 return true;
592}
a636b327 593
eb01aaab 594static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
14797e23 595{
13308ca9 596 return true;
14797e23
KM
597}
598
eb01aaab
VD
599static inline bool
600mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
90cbc250
VD
601{
602 return true;
603}
604
a3d8e054 605static inline unsigned long
4d7dcca2 606mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
a3d8e054
KM
607{
608 return 0;
609}
610
fa9add64
HD
611static inline void
612mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
613 int increment)
3e2f41f1 614{
3e2f41f1
KM
615}
616
e222432b
BS
617static inline void
618mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
619{
620}
621
81f8c3a4 622static inline struct mem_cgroup *lock_page_memcg(struct page *page)
89c06bd5 623{
d7365e78 624 return NULL;
89c06bd5
KH
625}
626
81f8c3a4 627static inline void unlock_page_memcg(struct mem_cgroup *memcg)
89c06bd5
KH
628{
629}
630
b23afb93
TH
631static inline void mem_cgroup_handle_over_high(void)
632{
633}
634
49426420 635static inline void mem_cgroup_oom_enable(void)
519e5247
JW
636{
637}
638
49426420 639static inline void mem_cgroup_oom_disable(void)
519e5247
JW
640{
641}
642
3812c8c8
JW
643static inline bool task_in_memcg_oom(struct task_struct *p)
644{
645 return false;
646}
647
49426420 648static inline bool mem_cgroup_oom_synchronize(bool wait)
3812c8c8
JW
649{
650 return false;
651}
652
d7365e78 653static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
68b4876d 654 enum mem_cgroup_stat_index idx)
2a7106f2
GT
655{
656}
657
d7365e78 658static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
68b4876d 659 enum mem_cgroup_stat_index idx)
d69b042f
BS
660{
661}
662
4e416953 663static inline
0608f43d
AM
664unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
665 gfp_t gfp_mask,
666 unsigned long *total_scanned)
4e416953 667{
0608f43d 668 return 0;
4e416953
BS
669}
670
e94c8a9c 671static inline void mem_cgroup_split_huge_fixup(struct page *head)
ca3e0214
KH
672{
673}
674
456f998e
YH
675static inline
676void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
677{
678}
c255a458 679#endif /* CONFIG_MEMCG */
78fb7466 680
52ebea74 681#ifdef CONFIG_CGROUP_WRITEBACK
841710aa 682
52ebea74 683struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
841710aa 684struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
c5edf9cd
TH
685void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
686 unsigned long *pheadroom, unsigned long *pdirty,
687 unsigned long *pwriteback);
841710aa
TH
688
689#else /* CONFIG_CGROUP_WRITEBACK */
690
691static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
692{
693 return NULL;
694}
695
c2aa723a 696static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
c5edf9cd
TH
697 unsigned long *pfilepages,
698 unsigned long *pheadroom,
c2aa723a
TH
699 unsigned long *pdirty,
700 unsigned long *pwriteback)
701{
702}
703
841710aa 704#endif /* CONFIG_CGROUP_WRITEBACK */
52ebea74 705
e1aab161 706struct sock;
e1aab161
GC
707void sock_update_memcg(struct sock *sk);
708void sock_release_memcg(struct sock *sk);
baac50bb
JW
709bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
710void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
d886f4e4 711#ifdef CONFIG_MEMCG
ef12947c
JW
712extern struct static_key_false memcg_sockets_enabled_key;
713#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
baac50bb 714static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
e805605c 715{
0db15298 716 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
8e8ae645 717 return true;
8e8ae645
JW
718 do {
719 if (time_before(jiffies, memcg->socket_pressure))
720 return true;
721 } while ((memcg = parent_mem_cgroup(memcg)));
722 return false;
e805605c
JW
723}
724#else
80e95fe0 725#define mem_cgroup_sockets_enabled 0
baac50bb 726static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
e805605c
JW
727{
728 return false;
729}
730#endif
7ae1e1d0 731
127424c8 732#if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
ef12947c 733extern struct static_key_false memcg_kmem_enabled_key;
749c5415 734
dbcf73e2 735extern int memcg_nr_cache_ids;
64219994
MH
736void memcg_get_cache_ids(void);
737void memcg_put_cache_ids(void);
ebe945c2
GC
738
739/*
740 * Helper macro to loop through all memcg-specific caches. Callers must still
741 * check if the cache is valid (it is either valid or NULL).
742 * the slab_mutex must be held when looping through those caches
743 */
749c5415 744#define for_each_memcg_cache_index(_idx) \
dbcf73e2 745 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
749c5415 746
7ae1e1d0
GC
747static inline bool memcg_kmem_enabled(void)
748{
ef12947c 749 return static_branch_unlikely(&memcg_kmem_enabled_key);
7ae1e1d0
GC
750}
751
567e9ab2 752static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
33398cf2 753{
567e9ab2 754 return memcg->kmem_state == KMEM_ONLINE;
33398cf2 755}
cb731d6c 756
7ae1e1d0
GC
757/*
758 * In general, we'll do everything in our power to not incur in any overhead
759 * for non-memcg users for the kmem functions. Not even a function call, if we
760 * can avoid it.
761 *
762 * Therefore, we'll inline all those functions so that in the best case, we'll
763 * see that kmemcg is off for everybody and proceed quickly. If it is on,
764 * we'll still do most of the flag checking inline. We check a lot of
765 * conditions, but because they are pretty simple, they are expected to be
766 * fast.
767 */
f3ccb2c4
VD
768int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
769 struct mem_cgroup *memcg);
d05e83a6
VD
770int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
771void __memcg_kmem_uncharge(struct page *page, int order);
7ae1e1d0 772
33398cf2 773/*
9f706d68 774 * helper for accessing a memcg's index. It will be used as an index in the
33398cf2
MH
775 * child cache array in kmem_cache, and also to derive its name. This function
776 * will return -1 when this is not a kmem-limited memcg.
777 */
778static inline int memcg_cache_id(struct mem_cgroup *memcg)
779{
780 return memcg ? memcg->kmemcg_id : -1;
781}
5722d094 782
230e9fc2 783struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
8135be5a 784void __memcg_kmem_put_cache(struct kmem_cache *cachep);
d7f25f8a 785
230e9fc2 786static inline bool __memcg_kmem_bypass(void)
7ae1e1d0
GC
787{
788 if (!memcg_kmem_enabled())
789 return true;
7ae1e1d0
GC
790 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
791 return true;
cbfb4798 792 return false;
7ae1e1d0
GC
793}
794
795/**
d05e83a6
VD
796 * memcg_kmem_charge: charge a kmem page
797 * @page: page to charge
798 * @gfp: reclaim mode
799 * @order: allocation order
cbfb4798 800 *
d05e83a6 801 * Returns 0 on success, an error code on failure.
7ae1e1d0 802 */
d05e83a6
VD
803static __always_inline int memcg_kmem_charge(struct page *page,
804 gfp_t gfp, int order)
7ae1e1d0 805{
230e9fc2
VD
806 if (__memcg_kmem_bypass())
807 return 0;
808 if (!(gfp & __GFP_ACCOUNT))
d05e83a6
VD
809 return 0;
810 return __memcg_kmem_charge(page, gfp, order);
7ae1e1d0
GC
811}
812
813/**
d05e83a6
VD
814 * memcg_kmem_uncharge: uncharge a kmem page
815 * @page: page to uncharge
816 * @order: allocation order
7ae1e1d0 817 */
d05e83a6 818static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
7ae1e1d0 819{
7ae1e1d0 820 if (memcg_kmem_enabled())
d05e83a6 821 __memcg_kmem_uncharge(page, order);
7ae1e1d0
GC
822}
823
d7f25f8a
GC
824/**
825 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
826 * @cachep: the original global kmem cache
d7f25f8a 827 *
5dfb4175 828 * All memory allocated from a per-memcg cache is charged to the owner memcg.
d7f25f8a
GC
829 */
830static __always_inline struct kmem_cache *
831memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
832{
230e9fc2 833 if (__memcg_kmem_bypass())
d7f25f8a 834 return cachep;
230e9fc2 835 return __memcg_kmem_get_cache(cachep, gfp);
d7f25f8a 836}
8135be5a
VD
837
838static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
839{
840 if (memcg_kmem_enabled())
841 __memcg_kmem_put_cache(cachep);
842}
7ae1e1d0 843#else
749c5415
GC
844#define for_each_memcg_cache_index(_idx) \
845 for (; NULL; )
846
b9ce5ef4
GC
847static inline bool memcg_kmem_enabled(void)
848{
849 return false;
850}
851
567e9ab2 852static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
cb731d6c
VD
853{
854 return false;
855}
856
d05e83a6 857static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
7ae1e1d0 858{
d05e83a6 859 return 0;
7ae1e1d0
GC
860}
861
d05e83a6 862static inline void memcg_kmem_uncharge(struct page *page, int order)
7ae1e1d0
GC
863{
864}
2633d7a0
GC
865
866static inline int memcg_cache_id(struct mem_cgroup *memcg)
867{
868 return -1;
869}
870
05257a1a
VD
871static inline void memcg_get_cache_ids(void)
872{
873}
874
875static inline void memcg_put_cache_ids(void)
876{
877}
878
d7f25f8a
GC
879static inline struct kmem_cache *
880memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
881{
882 return cachep;
883}
8135be5a
VD
884
885static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
886{
887}
127424c8
JW
888#endif /* CONFIG_MEMCG && !CONFIG_SLOB */
889
8cdea7c0 890#endif /* _LINUX_MEMCONTROL_H */
This page took 0.960153 seconds and 5 git commands to generate.