]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - include/linux/memcontrol.h
Kbuild: always define endianess in kconfig.h
[thirdparty/kernel/stable.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>
00f3ca2c
JW
29#include <linux/mm.h>
30#include <linux/vmstat.h>
33398cf2 31#include <linux/writeback.h>
fdf1cdb9 32#include <linux/page-flags.h>
456f998e 33
78fb7466 34struct mem_cgroup;
8697d331
BS
35struct page;
36struct mm_struct;
2633d7a0 37struct kmem_cache;
78fb7466 38
71cd3113
JW
39/* Cgroup-specific page state, on top of universal node page state */
40enum memcg_stat_item {
41 MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
42 MEMCG_RSS,
43 MEMCG_RSS_HUGE,
44 MEMCG_SWAP,
45 MEMCG_SOCK,
46 /* XXX: why are these zone and not node counters? */
47 MEMCG_KERNEL_STACK_KB,
b2807f07 48 MEMCG_NR_STAT,
2a7106f2
GT
49};
50
71cd3113
JW
51/* Cgroup-specific events, on top of universal VM events */
52enum memcg_event_item {
53 MEMCG_LOW = NR_VM_EVENT_ITEMS,
54 MEMCG_HIGH,
55 MEMCG_MAX,
56 MEMCG_OOM,
57 MEMCG_NR_EVENTS,
58};
59
5660048c 60struct mem_cgroup_reclaim_cookie {
ef8f2327 61 pg_data_t *pgdat;
5660048c
JW
62 int priority;
63 unsigned int generation;
64};
65
71cd3113
JW
66#ifdef CONFIG_MEMCG
67
68#define MEM_CGROUP_ID_SHIFT 16
69#define MEM_CGROUP_ID_MAX USHRT_MAX
70
71struct mem_cgroup_id {
72 int id;
73 atomic_t ref;
74};
75
33398cf2
MH
76/*
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.
81 */
82enum mem_cgroup_events_target {
83 MEM_CGROUP_TARGET_THRESH,
84 MEM_CGROUP_TARGET_SOFTLIMIT,
85 MEM_CGROUP_TARGET_NUMAINFO,
86 MEM_CGROUP_NTARGETS,
87};
88
33398cf2 89struct mem_cgroup_stat_cpu {
b2807f07 90 long count[MEMCG_NR_STAT];
33398cf2
MH
91 unsigned long events[MEMCG_NR_EVENTS];
92 unsigned long nr_page_events;
93 unsigned long targets[MEM_CGROUP_NTARGETS];
94};
95
96struct mem_cgroup_reclaim_iter {
97 struct mem_cgroup *position;
98 /* scan generation, increased every round-trip */
99 unsigned int generation;
100};
101
00f3ca2c
JW
102struct lruvec_stat {
103 long count[NR_VM_NODE_STAT_ITEMS];
104};
105
33398cf2
MH
106/*
107 * per-zone information in memory controller.
108 */
ef8f2327 109struct mem_cgroup_per_node {
33398cf2 110 struct lruvec lruvec;
a983b5eb
JW
111
112 struct lruvec_stat __percpu *lruvec_stat_cpu;
113 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
114
b4536f0c 115 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
33398cf2
MH
116
117 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
118
119 struct rb_node tree_node; /* RB tree node */
120 unsigned long usage_in_excess;/* Set to the value by which */
121 /* the soft limit is exceeded*/
122 bool on_tree;
123 struct mem_cgroup *memcg; /* Back pointer, we cannot */
124 /* use container_of */
125};
126
33398cf2
MH
127struct mem_cgroup_threshold {
128 struct eventfd_ctx *eventfd;
129 unsigned long threshold;
130};
131
132/* For threshold */
133struct mem_cgroup_threshold_ary {
134 /* An array index points to threshold just below or equal to usage. */
135 int current_threshold;
136 /* Size of entries[] */
137 unsigned int size;
138 /* Array of thresholds */
139 struct mem_cgroup_threshold entries[0];
140};
141
142struct mem_cgroup_thresholds {
143 /* Primary thresholds array */
144 struct mem_cgroup_threshold_ary *primary;
145 /*
146 * Spare threshold array.
147 * This is needed to make mem_cgroup_unregister_event() "never fail".
148 * It must be able to store at least primary->size - 1 entries.
149 */
150 struct mem_cgroup_threshold_ary *spare;
151};
152
567e9ab2
JW
153enum memcg_kmem_state {
154 KMEM_NONE,
155 KMEM_ALLOCATED,
156 KMEM_ONLINE,
157};
158
33398cf2
MH
159/*
160 * The memory controller data structure. The memory controller controls both
161 * page cache and RSS per cgroup. We would eventually like to provide
162 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
163 * to help the administrator determine what knobs to tune.
164 */
165struct mem_cgroup {
166 struct cgroup_subsys_state css;
167
73f576c0
JW
168 /* Private memcg ID. Used to ID objects that outlive the cgroup */
169 struct mem_cgroup_id id;
170
33398cf2
MH
171 /* Accounted resources */
172 struct page_counter memory;
37e84351 173 struct page_counter swap;
0db15298
JW
174
175 /* Legacy consumer-oriented counters */
33398cf2
MH
176 struct page_counter memsw;
177 struct page_counter kmem;
0db15298 178 struct page_counter tcpmem;
33398cf2
MH
179
180 /* Normal memory consumption range */
181 unsigned long low;
182 unsigned long high;
183
f7e1cb6e
JW
184 /* Range enforcement for interrupt charges */
185 struct work_struct high_work;
186
33398cf2
MH
187 unsigned long soft_limit;
188
189 /* vmpressure notifications */
190 struct vmpressure vmpressure;
191
33398cf2
MH
192 /*
193 * Should the accounting and control be hierarchical, per subtree?
194 */
195 bool use_hierarchy;
196
197 /* protected by memcg_oom_lock */
198 bool oom_lock;
199 int under_oom;
200
201 int swappiness;
202 /* OOM-Killer disable */
203 int oom_kill_disable;
204
472912a2
TH
205 /* handle for "memory.events" */
206 struct cgroup_file events_file;
207
33398cf2
MH
208 /* protect arrays of thresholds */
209 struct mutex thresholds_lock;
210
211 /* thresholds for memory usage. RCU-protected */
212 struct mem_cgroup_thresholds thresholds;
213
214 /* thresholds for mem+swap usage. RCU-protected */
215 struct mem_cgroup_thresholds memsw_thresholds;
216
217 /* For oom notifier event fd */
218 struct list_head oom_notify;
219
220 /*
221 * Should we move charges of a task when a task is moved into this
222 * mem_cgroup ? And what type of charges should we move ?
223 */
224 unsigned long move_charge_at_immigrate;
225 /*
226 * set > 0 if pages under this cgroup are moving to other cgroup.
227 */
228 atomic_t moving_account;
229 /* taken only while moving_account > 0 */
230 spinlock_t move_lock;
231 struct task_struct *move_lock_task;
232 unsigned long move_lock_flags;
a983b5eb
JW
233
234 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
235 atomic_long_t stat[MEMCG_NR_STAT];
236 atomic_long_t events[MEMCG_NR_EVENTS];
33398cf2 237
d886f4e4
JW
238 unsigned long socket_pressure;
239
240 /* Legacy tcp memory accounting */
0db15298
JW
241 bool tcpmem_active;
242 int tcpmem_pressure;
d886f4e4 243
127424c8 244#ifndef CONFIG_SLOB
33398cf2
MH
245 /* Index in the kmem_cache->memcg_params.memcg_caches array */
246 int kmemcg_id;
567e9ab2 247 enum memcg_kmem_state kmem_state;
bc2791f8 248 struct list_head kmem_caches;
33398cf2
MH
249#endif
250
251 int last_scanned_node;
252#if MAX_NUMNODES > 1
253 nodemask_t scan_nodes;
254 atomic_t numainfo_events;
255 atomic_t numainfo_updating;
256#endif
257
258#ifdef CONFIG_CGROUP_WRITEBACK
259 struct list_head cgwb_list;
260 struct wb_domain cgwb_domain;
261#endif
262
263 /* List of events which userspace want to receive */
264 struct list_head event_list;
265 spinlock_t event_list_lock;
266
267 struct mem_cgroup_per_node *nodeinfo[0];
268 /* WARNING: nodeinfo must be the last member here */
269};
7d828602 270
a983b5eb
JW
271/*
272 * size of first charge trial. "32" comes from vmscan.c's magic value.
273 * TODO: maybe necessary to use big numbers in big irons.
274 */
275#define MEMCG_CHARGE_BATCH 32U
276
7d828602 277extern struct mem_cgroup *root_mem_cgroup;
56161634 278
23047a96
JW
279static inline bool mem_cgroup_disabled(void)
280{
281 return !cgroup_subsys_enabled(memory_cgrp_subsys);
282}
283
241994ed
JW
284bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
285
00501b53 286int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
f627c2f5
KS
287 gfp_t gfp_mask, struct mem_cgroup **memcgp,
288 bool compound);
00501b53 289void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
f627c2f5
KS
290 bool lrucare, bool compound);
291void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
292 bool compound);
0a31bc97 293void mem_cgroup_uncharge(struct page *page);
747db954 294void mem_cgroup_uncharge_list(struct list_head *page_list);
569b846d 295
6a93ca8f 296void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
569b846d 297
ef8f2327
MG
298static struct mem_cgroup_per_node *
299mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
55779ec7 300{
ef8f2327 301 return memcg->nodeinfo[nid];
55779ec7
JW
302}
303
304/**
a9dd0a83
MG
305 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
306 * @node: node of the wanted lruvec
55779ec7
JW
307 * @memcg: memcg of the wanted lruvec
308 *
a9dd0a83
MG
309 * Returns the lru list vector holding pages for a given @node or a given
310 * @memcg and @zone. This can be the node lruvec, if the memory controller
55779ec7
JW
311 * is disabled.
312 */
a9dd0a83 313static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
ef8f2327 314 struct mem_cgroup *memcg)
55779ec7 315{
ef8f2327 316 struct mem_cgroup_per_node *mz;
55779ec7
JW
317 struct lruvec *lruvec;
318
319 if (mem_cgroup_disabled()) {
a9dd0a83 320 lruvec = node_lruvec(pgdat);
55779ec7
JW
321 goto out;
322 }
323
ef8f2327 324 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
55779ec7
JW
325 lruvec = &mz->lruvec;
326out:
327 /*
328 * Since a node can be onlined after the mem_cgroup was created,
599d0c95 329 * we have to be prepared to initialize lruvec->pgdat here;
55779ec7
JW
330 * and if offlined then reonlined, we need to reinitialize it.
331 */
ef8f2327
MG
332 if (unlikely(lruvec->pgdat != pgdat))
333 lruvec->pgdat = pgdat;
55779ec7
JW
334 return lruvec;
335}
336
599d0c95 337struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
c9b0ed51 338
2314b42d 339bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
64219994 340struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
e993d905 341
33398cf2
MH
342static inline
343struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
344 return css ? container_of(css, struct mem_cgroup, css) : NULL;
345}
346
8e8ae645
JW
347#define mem_cgroup_from_counter(counter, member) \
348 container_of(counter, struct mem_cgroup, member)
349
33398cf2
MH
350struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
351 struct mem_cgroup *,
352 struct mem_cgroup_reclaim_cookie *);
353void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
7c5f64f8
VD
354int mem_cgroup_scan_tasks(struct mem_cgroup *,
355 int (*)(struct task_struct *, void *), void *);
33398cf2 356
23047a96
JW
357static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
358{
359 if (mem_cgroup_disabled())
360 return 0;
361
73f576c0 362 return memcg->id.id;
23047a96 363}
73f576c0 364struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
23047a96 365
2262185c
RG
366static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
367{
368 struct mem_cgroup_per_node *mz;
369
370 if (mem_cgroup_disabled())
371 return NULL;
372
373 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
374 return mz->memcg;
375}
376
8e8ae645
JW
377/**
378 * parent_mem_cgroup - find the accounting parent of a memcg
379 * @memcg: memcg whose parent to find
380 *
381 * Returns the parent memcg, or NULL if this is the root or the memory
382 * controller is in legacy no-hierarchy mode.
383 */
384static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
385{
386 if (!memcg->memory.parent)
387 return NULL;
388 return mem_cgroup_from_counter(memcg->memory.parent, memory);
389}
390
33398cf2
MH
391static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
392 struct mem_cgroup *root)
393{
394 if (root == memcg)
395 return true;
396 if (!root->use_hierarchy)
397 return false;
398 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
399}
e1aab161 400
2314b42d
JW
401static inline bool mm_match_cgroup(struct mm_struct *mm,
402 struct mem_cgroup *memcg)
2e4d4091 403{
587af308 404 struct mem_cgroup *task_memcg;
413918bb 405 bool match = false;
c3ac9a8a 406
2e4d4091 407 rcu_read_lock();
587af308 408 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
413918bb 409 if (task_memcg)
2314b42d 410 match = mem_cgroup_is_descendant(task_memcg, memcg);
2e4d4091 411 rcu_read_unlock();
c3ac9a8a 412 return match;
2e4d4091 413}
8a9f3ccd 414
64219994 415struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
2fc04524 416ino_t page_cgroup_ino(struct page *page);
d324236b 417
eb01aaab
VD
418static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
419{
420 if (mem_cgroup_disabled())
421 return true;
422 return !!(memcg->css.flags & CSS_ONLINE);
423}
424
58ae83db
KH
425/*
426 * For memory reclaim.
427 */
889976db 428int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
33398cf2
MH
429
430void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
b4536f0c 431 int zid, int nr_pages);
33398cf2 432
0a6b76dd
VD
433unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
434 int nid, unsigned int lru_mask);
435
33398cf2
MH
436static inline
437unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
438{
ef8f2327 439 struct mem_cgroup_per_node *mz;
b4536f0c
MH
440 unsigned long nr_pages = 0;
441 int zid;
33398cf2 442
ef8f2327 443 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
b4536f0c
MH
444 for (zid = 0; zid < MAX_NR_ZONES; zid++)
445 nr_pages += mz->lru_zone_size[zid][lru];
446 return nr_pages;
447}
448
449static inline
450unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
451 enum lru_list lru, int zone_idx)
452{
453 struct mem_cgroup_per_node *mz;
454
455 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
456 return mz->lru_zone_size[zone_idx][lru];
33398cf2
MH
457}
458
b23afb93
TH
459void mem_cgroup_handle_over_high(void);
460
7c5f64f8
VD
461unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg);
462
64219994
MH
463void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
464 struct task_struct *p);
58ae83db 465
49426420 466static inline void mem_cgroup_oom_enable(void)
519e5247 467{
626ebc41
TH
468 WARN_ON(current->memcg_may_oom);
469 current->memcg_may_oom = 1;
519e5247
JW
470}
471
49426420 472static inline void mem_cgroup_oom_disable(void)
519e5247 473{
626ebc41
TH
474 WARN_ON(!current->memcg_may_oom);
475 current->memcg_may_oom = 0;
519e5247
JW
476}
477
3812c8c8
JW
478static inline bool task_in_memcg_oom(struct task_struct *p)
479{
626ebc41 480 return p->memcg_in_oom;
3812c8c8
JW
481}
482
49426420 483bool mem_cgroup_oom_synchronize(bool wait);
3812c8c8 484
c255a458 485#ifdef CONFIG_MEMCG_SWAP
c077719b
KH
486extern int do_swap_account;
487#endif
f8d66542 488
739f79fc
JW
489struct mem_cgroup *lock_page_memcg(struct page *page);
490void __unlock_page_memcg(struct mem_cgroup *memcg);
62cccb8c 491void unlock_page_memcg(struct page *page);
d7365e78 492
04fecbf5 493/* idx can be of type enum memcg_stat_item or node_stat_item */
ccda7f43 494static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
04fecbf5 495 int idx)
2a2e4885 496{
a983b5eb
JW
497 long x = atomic_long_read(&memcg->stat[idx]);
498#ifdef CONFIG_SMP
499 if (x < 0)
500 x = 0;
501#endif
502 return x;
2a2e4885
JW
503}
504
04fecbf5 505/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 506static inline void __mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 507 int idx, int val)
2a2e4885 508{
a983b5eb
JW
509 long x;
510
511 if (mem_cgroup_disabled())
512 return;
513
514 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
515 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
516 atomic_long_add(x, &memcg->stat[idx]);
517 x = 0;
518 }
519 __this_cpu_write(memcg->stat_cpu->count[idx], x);
2a2e4885
JW
520}
521
04fecbf5 522/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 523static inline void mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 524 int idx, int val)
2a2e4885 525{
a983b5eb
JW
526 preempt_disable();
527 __mod_memcg_state(memcg, idx, val);
528 preempt_enable();
2a2e4885
JW
529}
530
33398cf2 531/**
ccda7f43 532 * mod_memcg_page_state - update page state statistics
62cccb8c 533 * @page: the page
33398cf2
MH
534 * @idx: page state item to account
535 * @val: number of pages (positive or negative)
536 *
fdf1cdb9
JW
537 * The @page must be locked or the caller must use lock_page_memcg()
538 * to prevent double accounting when the page is concurrently being
539 * moved to another memcg:
81f8c3a4 540 *
fdf1cdb9 541 * lock_page(page) or lock_page_memcg(page)
81f8c3a4 542 * if (TestClearPageState(page))
ccda7f43 543 * mod_memcg_page_state(page, state, -1);
fdf1cdb9 544 * unlock_page(page) or unlock_page_memcg(page)
2a2e4885
JW
545 *
546 * Kernel pages are an exception to this, since they'll never move.
33398cf2 547 */
00f3ca2c 548static inline void __mod_memcg_page_state(struct page *page,
04fecbf5 549 int idx, int val)
00f3ca2c
JW
550{
551 if (page->mem_cgroup)
552 __mod_memcg_state(page->mem_cgroup, idx, val);
553}
554
ccda7f43 555static inline void mod_memcg_page_state(struct page *page,
04fecbf5 556 int idx, int val)
33398cf2 557{
62cccb8c 558 if (page->mem_cgroup)
ccda7f43 559 mod_memcg_state(page->mem_cgroup, idx, val);
33398cf2
MH
560}
561
00f3ca2c
JW
562static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
563 enum node_stat_item idx)
2a7106f2 564{
00f3ca2c 565 struct mem_cgroup_per_node *pn;
a983b5eb 566 long x;
00f3ca2c
JW
567
568 if (mem_cgroup_disabled())
569 return node_page_state(lruvec_pgdat(lruvec), idx);
570
571 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
a983b5eb
JW
572 x = atomic_long_read(&pn->lruvec_stat[idx]);
573#ifdef CONFIG_SMP
574 if (x < 0)
575 x = 0;
576#endif
577 return x;
2a7106f2
GT
578}
579
00f3ca2c
JW
580static inline void __mod_lruvec_state(struct lruvec *lruvec,
581 enum node_stat_item idx, int val)
2a7106f2 582{
00f3ca2c 583 struct mem_cgroup_per_node *pn;
a983b5eb 584 long x;
00f3ca2c 585
28454265 586 /* Update node */
00f3ca2c 587 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
28454265 588
00f3ca2c
JW
589 if (mem_cgroup_disabled())
590 return;
28454265 591
00f3ca2c 592 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
28454265
JW
593
594 /* Update memcg */
00f3ca2c 595 __mod_memcg_state(pn->memcg, idx, val);
28454265
JW
596
597 /* Update lruvec */
a983b5eb
JW
598 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
599 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
600 atomic_long_add(x, &pn->lruvec_stat[idx]);
601 x = 0;
602 }
603 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
00f3ca2c
JW
604}
605
606static inline void mod_lruvec_state(struct lruvec *lruvec,
607 enum node_stat_item idx, int val)
608{
28454265
JW
609 preempt_disable();
610 __mod_lruvec_state(lruvec, idx, val);
611 preempt_enable();
00f3ca2c
JW
612}
613
614static inline void __mod_lruvec_page_state(struct page *page,
615 enum node_stat_item idx, int val)
616{
28454265
JW
617 pg_data_t *pgdat = page_pgdat(page);
618 struct lruvec *lruvec;
00f3ca2c 619
28454265
JW
620 /* Untracked pages have no memcg, no lruvec. Update only the node */
621 if (!page->mem_cgroup) {
622 __mod_node_page_state(pgdat, idx, val);
00f3ca2c 623 return;
28454265
JW
624 }
625
626 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
627 __mod_lruvec_state(lruvec, idx, val);
00f3ca2c
JW
628}
629
630static inline void mod_lruvec_page_state(struct page *page,
631 enum node_stat_item idx, int val)
632{
28454265
JW
633 preempt_disable();
634 __mod_lruvec_page_state(page, idx, val);
635 preempt_enable();
2a7106f2
GT
636}
637
ef8f2327 638unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
0608f43d
AM
639 gfp_t gfp_mask,
640 unsigned long *total_scanned);
a63d83f4 641
c9019e9b
JW
642/* idx can be of type enum memcg_event_item or vm_event_item */
643static inline void __count_memcg_events(struct mem_cgroup *memcg,
644 int idx, unsigned long count)
645{
a983b5eb
JW
646 unsigned long x;
647
648 if (mem_cgroup_disabled())
649 return;
650
651 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
652 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
653 atomic_long_add(x, &memcg->events[idx]);
654 x = 0;
655 }
656 __this_cpu_write(memcg->stat_cpu->events[idx], x);
c9019e9b
JW
657}
658
2262185c 659static inline void count_memcg_events(struct mem_cgroup *memcg,
c9019e9b 660 int idx, unsigned long count)
2262185c 661{
a983b5eb
JW
662 preempt_disable();
663 __count_memcg_events(memcg, idx, count);
664 preempt_enable();
2262185c
RG
665}
666
c9019e9b 667/* idx can be of type enum memcg_event_item or vm_event_item */
2262185c 668static inline void count_memcg_page_event(struct page *page,
04fecbf5 669 int idx)
2262185c
RG
670{
671 if (page->mem_cgroup)
672 count_memcg_events(page->mem_cgroup, idx, 1);
673}
674
675static inline void count_memcg_event_mm(struct mm_struct *mm,
676 enum vm_event_item idx)
68ae564b 677{
33398cf2
MH
678 struct mem_cgroup *memcg;
679
68ae564b
DR
680 if (mem_cgroup_disabled())
681 return;
33398cf2
MH
682
683 rcu_read_lock();
684 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
8e675f7a 685 if (likely(memcg)) {
c9019e9b 686 count_memcg_events(memcg, idx, 1);
8e675f7a
KK
687 if (idx == OOM_KILL)
688 cgroup_file_notify(&memcg->events_file);
689 }
33398cf2 690 rcu_read_unlock();
68ae564b 691}
c9019e9b
JW
692
693static inline void mem_cgroup_event(struct mem_cgroup *memcg,
694 enum memcg_event_item event)
695{
696 count_memcg_events(memcg, event, 1);
697 cgroup_file_notify(&memcg->events_file);
698}
699
ca3e0214 700#ifdef CONFIG_TRANSPARENT_HUGEPAGE
e94c8a9c 701void mem_cgroup_split_huge_fixup(struct page *head);
ca3e0214
KH
702#endif
703
c255a458 704#else /* CONFIG_MEMCG */
23047a96
JW
705
706#define MEM_CGROUP_ID_SHIFT 0
707#define MEM_CGROUP_ID_MAX 0
708
7a81b88c
KH
709struct mem_cgroup;
710
23047a96
JW
711static inline bool mem_cgroup_disabled(void)
712{
713 return true;
714}
715
31176c78 716static inline void mem_cgroup_event(struct mem_cgroup *memcg,
df0e53d0 717 enum memcg_event_item event)
241994ed
JW
718{
719}
720
721static inline bool mem_cgroup_low(struct mem_cgroup *root,
722 struct mem_cgroup *memcg)
723{
724 return false;
725}
726
00501b53
JW
727static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
728 gfp_t gfp_mask,
f627c2f5
KS
729 struct mem_cgroup **memcgp,
730 bool compound)
7a81b88c 731{
00501b53 732 *memcgp = NULL;
7a81b88c
KH
733 return 0;
734}
735
00501b53
JW
736static inline void mem_cgroup_commit_charge(struct page *page,
737 struct mem_cgroup *memcg,
f627c2f5 738 bool lrucare, bool compound)
7a81b88c
KH
739{
740}
741
00501b53 742static inline void mem_cgroup_cancel_charge(struct page *page,
f627c2f5
KS
743 struct mem_cgroup *memcg,
744 bool compound)
7a81b88c
KH
745{
746}
747
0a31bc97 748static inline void mem_cgroup_uncharge(struct page *page)
569b846d
KH
749{
750}
751
747db954 752static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
8a9f3ccd
BS
753{
754}
755
6a93ca8f 756static inline void mem_cgroup_migrate(struct page *old, struct page *new)
69029cd5
KH
757{
758}
759
a9dd0a83 760static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
ef8f2327 761 struct mem_cgroup *memcg)
08e552c6 762{
a9dd0a83 763 return node_lruvec(pgdat);
08e552c6
KH
764}
765
fa9add64 766static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
599d0c95 767 struct pglist_data *pgdat)
66e1707b 768{
599d0c95 769 return &pgdat->lruvec;
66e1707b
BS
770}
771
587af308 772static inline bool mm_match_cgroup(struct mm_struct *mm,
c0ff4b85 773 struct mem_cgroup *memcg)
bed7161a 774{
587af308 775 return true;
bed7161a
BS
776}
777
ffbdccf5
DR
778static inline bool task_in_mem_cgroup(struct task_struct *task,
779 const struct mem_cgroup *memcg)
4c4a2214 780{
ffbdccf5 781 return true;
4c4a2214
DR
782}
783
5660048c
JW
784static inline struct mem_cgroup *
785mem_cgroup_iter(struct mem_cgroup *root,
786 struct mem_cgroup *prev,
787 struct mem_cgroup_reclaim_cookie *reclaim)
788{
789 return NULL;
790}
791
792static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
793 struct mem_cgroup *prev)
794{
795}
796
7c5f64f8
VD
797static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
798 int (*fn)(struct task_struct *, void *), void *arg)
799{
800 return 0;
801}
802
23047a96 803static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
f8d66542 804{
23047a96
JW
805 return 0;
806}
807
808static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
809{
810 WARN_ON_ONCE(id);
811 /* XXX: This should always return root_mem_cgroup */
812 return NULL;
f8d66542 813}
a636b327 814
2262185c
RG
815static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
816{
817 return NULL;
818}
819
eb01aaab 820static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
14797e23 821{
13308ca9 822 return true;
14797e23
KM
823}
824
a3d8e054 825static inline unsigned long
4d7dcca2 826mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
a3d8e054
KM
827{
828 return 0;
829}
b4536f0c
MH
830static inline
831unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
832 enum lru_list lru, int zone_idx)
833{
834 return 0;
835}
a3d8e054 836
0a6b76dd
VD
837static inline unsigned long
838mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
839 int nid, unsigned int lru_mask)
840{
841 return 0;
842}
843
7c5f64f8
VD
844static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
845{
846 return 0;
847}
848
e222432b
BS
849static inline void
850mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
851{
852}
853
739f79fc
JW
854static inline struct mem_cgroup *lock_page_memcg(struct page *page)
855{
856 return NULL;
857}
858
859static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
89c06bd5
KH
860{
861}
862
62cccb8c 863static inline void unlock_page_memcg(struct page *page)
89c06bd5
KH
864{
865}
866
b23afb93
TH
867static inline void mem_cgroup_handle_over_high(void)
868{
869}
870
49426420 871static inline void mem_cgroup_oom_enable(void)
519e5247
JW
872{
873}
874
49426420 875static inline void mem_cgroup_oom_disable(void)
519e5247
JW
876{
877}
878
3812c8c8
JW
879static inline bool task_in_memcg_oom(struct task_struct *p)
880{
881 return false;
882}
883
49426420 884static inline bool mem_cgroup_oom_synchronize(bool wait)
3812c8c8
JW
885{
886 return false;
887}
888
ccda7f43 889static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
04fecbf5 890 int idx)
2a2e4885
JW
891{
892 return 0;
893}
894
00f3ca2c 895static inline void __mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 896 int idx,
00f3ca2c 897 int nr)
2a2e4885
JW
898{
899}
900
00f3ca2c 901static inline void mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 902 int idx,
00f3ca2c 903 int nr)
2a2e4885
JW
904{
905}
906
00f3ca2c 907static inline void __mod_memcg_page_state(struct page *page,
04fecbf5 908 int idx,
00f3ca2c 909 int nr)
2a2e4885
JW
910{
911}
912
ccda7f43 913static inline void mod_memcg_page_state(struct page *page,
04fecbf5 914 int idx,
ccda7f43 915 int nr)
553af430
JW
916{
917}
918
00f3ca2c
JW
919static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
920 enum node_stat_item idx)
2a7106f2 921{
00f3ca2c 922 return node_page_state(lruvec_pgdat(lruvec), idx);
2a7106f2
GT
923}
924
00f3ca2c
JW
925static inline void __mod_lruvec_state(struct lruvec *lruvec,
926 enum node_stat_item idx, int val)
d69b042f 927{
00f3ca2c
JW
928 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
929}
930
931static inline void mod_lruvec_state(struct lruvec *lruvec,
932 enum node_stat_item idx, int val)
933{
934 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
935}
936
937static inline void __mod_lruvec_page_state(struct page *page,
938 enum node_stat_item idx, int val)
939{
940 __mod_node_page_state(page_pgdat(page), idx, val);
941}
942
943static inline void mod_lruvec_page_state(struct page *page,
944 enum node_stat_item idx, int val)
945{
946 mod_node_page_state(page_pgdat(page), idx, val);
d69b042f
BS
947}
948
4e416953 949static inline
ef8f2327 950unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
0608f43d
AM
951 gfp_t gfp_mask,
952 unsigned long *total_scanned)
4e416953 953{
0608f43d 954 return 0;
4e416953
BS
955}
956
e94c8a9c 957static inline void mem_cgroup_split_huge_fixup(struct page *head)
ca3e0214
KH
958{
959}
960
2262185c
RG
961static inline void count_memcg_events(struct mem_cgroup *memcg,
962 enum vm_event_item idx,
963 unsigned long count)
964{
965}
966
967static inline void count_memcg_page_event(struct page *page,
04fecbf5 968 int idx)
2262185c
RG
969{
970}
971
456f998e 972static inline
2262185c 973void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
456f998e
YH
974{
975}
c255a458 976#endif /* CONFIG_MEMCG */
78fb7466 977
04fecbf5 978/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 979static inline void __inc_memcg_state(struct mem_cgroup *memcg,
04fecbf5 980 int idx)
00f3ca2c
JW
981{
982 __mod_memcg_state(memcg, idx, 1);
983}
984
04fecbf5 985/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 986static inline void __dec_memcg_state(struct mem_cgroup *memcg,
04fecbf5 987 int idx)
00f3ca2c
JW
988{
989 __mod_memcg_state(memcg, idx, -1);
990}
991
04fecbf5 992/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 993static inline void __inc_memcg_page_state(struct page *page,
04fecbf5 994 int idx)
00f3ca2c
JW
995{
996 __mod_memcg_page_state(page, idx, 1);
997}
998
04fecbf5 999/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1000static inline void __dec_memcg_page_state(struct page *page,
04fecbf5 1001 int idx)
00f3ca2c
JW
1002{
1003 __mod_memcg_page_state(page, idx, -1);
1004}
1005
1006static inline void __inc_lruvec_state(struct lruvec *lruvec,
1007 enum node_stat_item idx)
1008{
1009 __mod_lruvec_state(lruvec, idx, 1);
1010}
1011
1012static inline void __dec_lruvec_state(struct lruvec *lruvec,
1013 enum node_stat_item idx)
1014{
1015 __mod_lruvec_state(lruvec, idx, -1);
1016}
1017
1018static inline void __inc_lruvec_page_state(struct page *page,
1019 enum node_stat_item idx)
1020{
1021 __mod_lruvec_page_state(page, idx, 1);
1022}
1023
1024static inline void __dec_lruvec_page_state(struct page *page,
1025 enum node_stat_item idx)
1026{
1027 __mod_lruvec_page_state(page, idx, -1);
1028}
1029
04fecbf5 1030/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1031static inline void inc_memcg_state(struct mem_cgroup *memcg,
04fecbf5 1032 int idx)
00f3ca2c
JW
1033{
1034 mod_memcg_state(memcg, idx, 1);
1035}
1036
04fecbf5 1037/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1038static inline void dec_memcg_state(struct mem_cgroup *memcg,
04fecbf5 1039 int idx)
00f3ca2c
JW
1040{
1041 mod_memcg_state(memcg, idx, -1);
1042}
1043
04fecbf5 1044/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1045static inline void inc_memcg_page_state(struct page *page,
04fecbf5 1046 int idx)
00f3ca2c
JW
1047{
1048 mod_memcg_page_state(page, idx, 1);
1049}
1050
04fecbf5 1051/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1052static inline void dec_memcg_page_state(struct page *page,
04fecbf5 1053 int idx)
00f3ca2c
JW
1054{
1055 mod_memcg_page_state(page, idx, -1);
1056}
1057
1058static inline void inc_lruvec_state(struct lruvec *lruvec,
1059 enum node_stat_item idx)
1060{
1061 mod_lruvec_state(lruvec, idx, 1);
1062}
1063
1064static inline void dec_lruvec_state(struct lruvec *lruvec,
1065 enum node_stat_item idx)
1066{
1067 mod_lruvec_state(lruvec, idx, -1);
1068}
1069
1070static inline void inc_lruvec_page_state(struct page *page,
1071 enum node_stat_item idx)
1072{
1073 mod_lruvec_page_state(page, idx, 1);
1074}
1075
1076static inline void dec_lruvec_page_state(struct page *page,
1077 enum node_stat_item idx)
1078{
1079 mod_lruvec_page_state(page, idx, -1);
1080}
1081
52ebea74 1082#ifdef CONFIG_CGROUP_WRITEBACK
841710aa 1083
52ebea74 1084struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
841710aa 1085struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
c5edf9cd
TH
1086void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1087 unsigned long *pheadroom, unsigned long *pdirty,
1088 unsigned long *pwriteback);
841710aa
TH
1089
1090#else /* CONFIG_CGROUP_WRITEBACK */
1091
1092static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1093{
1094 return NULL;
1095}
1096
c2aa723a 1097static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
c5edf9cd
TH
1098 unsigned long *pfilepages,
1099 unsigned long *pheadroom,
c2aa723a
TH
1100 unsigned long *pdirty,
1101 unsigned long *pwriteback)
1102{
1103}
1104
841710aa 1105#endif /* CONFIG_CGROUP_WRITEBACK */
52ebea74 1106
e1aab161 1107struct sock;
baac50bb
JW
1108bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1109void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
d886f4e4 1110#ifdef CONFIG_MEMCG
ef12947c
JW
1111extern struct static_key_false memcg_sockets_enabled_key;
1112#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
2d758073
JW
1113void mem_cgroup_sk_alloc(struct sock *sk);
1114void mem_cgroup_sk_free(struct sock *sk);
baac50bb 1115static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
e805605c 1116{
0db15298 1117 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
8e8ae645 1118 return true;
8e8ae645
JW
1119 do {
1120 if (time_before(jiffies, memcg->socket_pressure))
1121 return true;
1122 } while ((memcg = parent_mem_cgroup(memcg)));
1123 return false;
e805605c
JW
1124}
1125#else
80e95fe0 1126#define mem_cgroup_sockets_enabled 0
2d758073
JW
1127static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1128static inline void mem_cgroup_sk_free(struct sock *sk) { };
baac50bb 1129static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
e805605c
JW
1130{
1131 return false;
1132}
1133#endif
7ae1e1d0 1134
45264778
VD
1135struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1136void memcg_kmem_put_cache(struct kmem_cache *cachep);
1137int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1138 struct mem_cgroup *memcg);
1139int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1140void memcg_kmem_uncharge(struct page *page, int order);
1141
127424c8 1142#if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
ef12947c 1143extern struct static_key_false memcg_kmem_enabled_key;
17cc4dfe 1144extern struct workqueue_struct *memcg_kmem_cache_wq;
749c5415 1145
dbcf73e2 1146extern int memcg_nr_cache_ids;
64219994
MH
1147void memcg_get_cache_ids(void);
1148void memcg_put_cache_ids(void);
ebe945c2
GC
1149
1150/*
1151 * Helper macro to loop through all memcg-specific caches. Callers must still
1152 * check if the cache is valid (it is either valid or NULL).
1153 * the slab_mutex must be held when looping through those caches
1154 */
749c5415 1155#define for_each_memcg_cache_index(_idx) \
dbcf73e2 1156 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
749c5415 1157
7ae1e1d0
GC
1158static inline bool memcg_kmem_enabled(void)
1159{
ef12947c 1160 return static_branch_unlikely(&memcg_kmem_enabled_key);
7ae1e1d0
GC
1161}
1162
33398cf2 1163/*
9f706d68 1164 * helper for accessing a memcg's index. It will be used as an index in the
33398cf2
MH
1165 * child cache array in kmem_cache, and also to derive its name. This function
1166 * will return -1 when this is not a kmem-limited memcg.
1167 */
1168static inline int memcg_cache_id(struct mem_cgroup *memcg)
1169{
1170 return memcg ? memcg->kmemcg_id : -1;
1171}
5722d094 1172
7ae1e1d0 1173#else
749c5415
GC
1174#define for_each_memcg_cache_index(_idx) \
1175 for (; NULL; )
1176
b9ce5ef4
GC
1177static inline bool memcg_kmem_enabled(void)
1178{
1179 return false;
1180}
1181
2633d7a0
GC
1182static inline int memcg_cache_id(struct mem_cgroup *memcg)
1183{
1184 return -1;
1185}
1186
05257a1a
VD
1187static inline void memcg_get_cache_ids(void)
1188{
1189}
1190
1191static inline void memcg_put_cache_ids(void)
1192{
1193}
1194
127424c8
JW
1195#endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1196
8cdea7c0 1197#endif /* _LINUX_MEMCONTROL_H */