]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/linux/memcontrol.h
mm/memblock: introduce PHYS_ADDR_MAX
[mirror_ubuntu-hirsute-kernel.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
e27be240
JW
51enum memcg_memory_event {
52 MEMCG_LOW,
71cd3113
JW
53 MEMCG_HIGH,
54 MEMCG_MAX,
55 MEMCG_OOM,
f3a53a3a
TH
56 MEMCG_SWAP_MAX,
57 MEMCG_SWAP_FAIL,
e27be240 58 MEMCG_NR_MEMORY_EVENTS,
71cd3113
JW
59};
60
5660048c 61struct mem_cgroup_reclaim_cookie {
ef8f2327 62 pg_data_t *pgdat;
5660048c
JW
63 int priority;
64 unsigned int generation;
65};
66
71cd3113
JW
67#ifdef CONFIG_MEMCG
68
69#define MEM_CGROUP_ID_SHIFT 16
70#define MEM_CGROUP_ID_MAX USHRT_MAX
71
72struct mem_cgroup_id {
73 int id;
74 atomic_t ref;
75};
76
33398cf2
MH
77/*
78 * Per memcg event counter is incremented at every pagein/pageout. With THP,
79 * it will be incremated by the number of pages. This counter is used for
80 * for trigger some periodic events. This is straightforward and better
81 * than using jiffies etc. to handle periodic memcg event.
82 */
83enum mem_cgroup_events_target {
84 MEM_CGROUP_TARGET_THRESH,
85 MEM_CGROUP_TARGET_SOFTLIMIT,
86 MEM_CGROUP_TARGET_NUMAINFO,
87 MEM_CGROUP_NTARGETS,
88};
89
33398cf2 90struct mem_cgroup_stat_cpu {
b2807f07 91 long count[MEMCG_NR_STAT];
e27be240 92 unsigned long events[NR_VM_EVENT_ITEMS];
33398cf2
MH
93 unsigned long nr_page_events;
94 unsigned long targets[MEM_CGROUP_NTARGETS];
95};
96
97struct mem_cgroup_reclaim_iter {
98 struct mem_cgroup *position;
99 /* scan generation, increased every round-trip */
100 unsigned int generation;
101};
102
00f3ca2c
JW
103struct lruvec_stat {
104 long count[NR_VM_NODE_STAT_ITEMS];
105};
106
33398cf2
MH
107/*
108 * per-zone information in memory controller.
109 */
ef8f2327 110struct mem_cgroup_per_node {
33398cf2 111 struct lruvec lruvec;
a983b5eb
JW
112
113 struct lruvec_stat __percpu *lruvec_stat_cpu;
114 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
115
b4536f0c 116 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
33398cf2
MH
117
118 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
119
120 struct rb_node tree_node; /* RB tree node */
121 unsigned long usage_in_excess;/* Set to the value by which */
122 /* the soft limit is exceeded*/
123 bool on_tree;
e3c1ac58
AR
124 bool congested; /* memcg has many dirty pages */
125 /* backed by a congested BDI */
126
33398cf2
MH
127 struct mem_cgroup *memcg; /* Back pointer, we cannot */
128 /* use container_of */
129};
130
33398cf2
MH
131struct mem_cgroup_threshold {
132 struct eventfd_ctx *eventfd;
133 unsigned long threshold;
134};
135
136/* For threshold */
137struct mem_cgroup_threshold_ary {
138 /* An array index points to threshold just below or equal to usage. */
139 int current_threshold;
140 /* Size of entries[] */
141 unsigned int size;
142 /* Array of thresholds */
143 struct mem_cgroup_threshold entries[0];
144};
145
146struct mem_cgroup_thresholds {
147 /* Primary thresholds array */
148 struct mem_cgroup_threshold_ary *primary;
149 /*
150 * Spare threshold array.
151 * This is needed to make mem_cgroup_unregister_event() "never fail".
152 * It must be able to store at least primary->size - 1 entries.
153 */
154 struct mem_cgroup_threshold_ary *spare;
155};
156
567e9ab2
JW
157enum memcg_kmem_state {
158 KMEM_NONE,
159 KMEM_ALLOCATED,
160 KMEM_ONLINE,
161};
162
33398cf2
MH
163/*
164 * The memory controller data structure. The memory controller controls both
165 * page cache and RSS per cgroup. We would eventually like to provide
166 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
167 * to help the administrator determine what knobs to tune.
168 */
169struct mem_cgroup {
170 struct cgroup_subsys_state css;
171
73f576c0
JW
172 /* Private memcg ID. Used to ID objects that outlive the cgroup */
173 struct mem_cgroup_id id;
174
33398cf2
MH
175 /* Accounted resources */
176 struct page_counter memory;
37e84351 177 struct page_counter swap;
0db15298
JW
178
179 /* Legacy consumer-oriented counters */
33398cf2
MH
180 struct page_counter memsw;
181 struct page_counter kmem;
0db15298 182 struct page_counter tcpmem;
33398cf2
MH
183
184 /* Normal memory consumption range */
185 unsigned long low;
186 unsigned long high;
187
f7e1cb6e
JW
188 /* Range enforcement for interrupt charges */
189 struct work_struct high_work;
190
33398cf2
MH
191 unsigned long soft_limit;
192
193 /* vmpressure notifications */
194 struct vmpressure vmpressure;
195
33398cf2
MH
196 /*
197 * Should the accounting and control be hierarchical, per subtree?
198 */
199 bool use_hierarchy;
200
201 /* protected by memcg_oom_lock */
202 bool oom_lock;
203 int under_oom;
204
205 int swappiness;
206 /* OOM-Killer disable */
207 int oom_kill_disable;
208
e27be240
JW
209 /* memory.events */
210 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
472912a2
TH
211 struct cgroup_file events_file;
212
f3a53a3a
TH
213 /* handle for "memory.swap.events" */
214 struct cgroup_file swap_events_file;
215
33398cf2
MH
216 /* protect arrays of thresholds */
217 struct mutex thresholds_lock;
218
219 /* thresholds for memory usage. RCU-protected */
220 struct mem_cgroup_thresholds thresholds;
221
222 /* thresholds for mem+swap usage. RCU-protected */
223 struct mem_cgroup_thresholds memsw_thresholds;
224
225 /* For oom notifier event fd */
226 struct list_head oom_notify;
227
228 /*
229 * Should we move charges of a task when a task is moved into this
230 * mem_cgroup ? And what type of charges should we move ?
231 */
232 unsigned long move_charge_at_immigrate;
233 /*
234 * set > 0 if pages under this cgroup are moving to other cgroup.
235 */
236 atomic_t moving_account;
237 /* taken only while moving_account > 0 */
238 spinlock_t move_lock;
239 struct task_struct *move_lock_task;
240 unsigned long move_lock_flags;
a983b5eb 241
e27be240 242 /* memory.stat */
a983b5eb
JW
243 struct mem_cgroup_stat_cpu __percpu *stat_cpu;
244 atomic_long_t stat[MEMCG_NR_STAT];
e27be240 245 atomic_long_t events[NR_VM_EVENT_ITEMS];
33398cf2 246
d886f4e4
JW
247 unsigned long socket_pressure;
248
249 /* Legacy tcp memory accounting */
0db15298
JW
250 bool tcpmem_active;
251 int tcpmem_pressure;
d886f4e4 252
127424c8 253#ifndef CONFIG_SLOB
33398cf2
MH
254 /* Index in the kmem_cache->memcg_params.memcg_caches array */
255 int kmemcg_id;
567e9ab2 256 enum memcg_kmem_state kmem_state;
bc2791f8 257 struct list_head kmem_caches;
33398cf2
MH
258#endif
259
260 int last_scanned_node;
261#if MAX_NUMNODES > 1
262 nodemask_t scan_nodes;
263 atomic_t numainfo_events;
264 atomic_t numainfo_updating;
265#endif
266
267#ifdef CONFIG_CGROUP_WRITEBACK
268 struct list_head cgwb_list;
269 struct wb_domain cgwb_domain;
270#endif
271
272 /* List of events which userspace want to receive */
273 struct list_head event_list;
274 spinlock_t event_list_lock;
275
276 struct mem_cgroup_per_node *nodeinfo[0];
277 /* WARNING: nodeinfo must be the last member here */
278};
7d828602 279
a983b5eb
JW
280/*
281 * size of first charge trial. "32" comes from vmscan.c's magic value.
282 * TODO: maybe necessary to use big numbers in big irons.
283 */
284#define MEMCG_CHARGE_BATCH 32U
285
7d828602 286extern struct mem_cgroup *root_mem_cgroup;
56161634 287
23047a96
JW
288static inline bool mem_cgroup_disabled(void)
289{
290 return !cgroup_subsys_enabled(memory_cgrp_subsys);
291}
292
241994ed
JW
293bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
294
00501b53 295int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
f627c2f5
KS
296 gfp_t gfp_mask, struct mem_cgroup **memcgp,
297 bool compound);
00501b53 298void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
f627c2f5
KS
299 bool lrucare, bool compound);
300void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
301 bool compound);
0a31bc97 302void mem_cgroup_uncharge(struct page *page);
747db954 303void mem_cgroup_uncharge_list(struct list_head *page_list);
569b846d 304
6a93ca8f 305void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
569b846d 306
ef8f2327
MG
307static struct mem_cgroup_per_node *
308mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
55779ec7 309{
ef8f2327 310 return memcg->nodeinfo[nid];
55779ec7
JW
311}
312
313/**
a9dd0a83
MG
314 * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
315 * @node: node of the wanted lruvec
55779ec7
JW
316 * @memcg: memcg of the wanted lruvec
317 *
a9dd0a83
MG
318 * Returns the lru list vector holding pages for a given @node or a given
319 * @memcg and @zone. This can be the node lruvec, if the memory controller
55779ec7
JW
320 * is disabled.
321 */
a9dd0a83 322static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
ef8f2327 323 struct mem_cgroup *memcg)
55779ec7 324{
ef8f2327 325 struct mem_cgroup_per_node *mz;
55779ec7
JW
326 struct lruvec *lruvec;
327
328 if (mem_cgroup_disabled()) {
a9dd0a83 329 lruvec = node_lruvec(pgdat);
55779ec7
JW
330 goto out;
331 }
332
ef8f2327 333 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
55779ec7
JW
334 lruvec = &mz->lruvec;
335out:
336 /*
337 * Since a node can be onlined after the mem_cgroup was created,
599d0c95 338 * we have to be prepared to initialize lruvec->pgdat here;
55779ec7
JW
339 * and if offlined then reonlined, we need to reinitialize it.
340 */
ef8f2327
MG
341 if (unlikely(lruvec->pgdat != pgdat))
342 lruvec->pgdat = pgdat;
55779ec7
JW
343 return lruvec;
344}
345
599d0c95 346struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
c9b0ed51 347
2314b42d 348bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
64219994 349struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
e993d905 350
33398cf2
MH
351static inline
352struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
353 return css ? container_of(css, struct mem_cgroup, css) : NULL;
354}
355
8e8ae645
JW
356#define mem_cgroup_from_counter(counter, member) \
357 container_of(counter, struct mem_cgroup, member)
358
33398cf2
MH
359struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
360 struct mem_cgroup *,
361 struct mem_cgroup_reclaim_cookie *);
362void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
7c5f64f8
VD
363int mem_cgroup_scan_tasks(struct mem_cgroup *,
364 int (*)(struct task_struct *, void *), void *);
33398cf2 365
23047a96
JW
366static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
367{
368 if (mem_cgroup_disabled())
369 return 0;
370
73f576c0 371 return memcg->id.id;
23047a96 372}
73f576c0 373struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
23047a96 374
2262185c
RG
375static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
376{
377 struct mem_cgroup_per_node *mz;
378
379 if (mem_cgroup_disabled())
380 return NULL;
381
382 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
383 return mz->memcg;
384}
385
8e8ae645
JW
386/**
387 * parent_mem_cgroup - find the accounting parent of a memcg
388 * @memcg: memcg whose parent to find
389 *
390 * Returns the parent memcg, or NULL if this is the root or the memory
391 * controller is in legacy no-hierarchy mode.
392 */
393static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
394{
395 if (!memcg->memory.parent)
396 return NULL;
397 return mem_cgroup_from_counter(memcg->memory.parent, memory);
398}
399
33398cf2
MH
400static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
401 struct mem_cgroup *root)
402{
403 if (root == memcg)
404 return true;
405 if (!root->use_hierarchy)
406 return false;
407 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
408}
e1aab161 409
2314b42d
JW
410static inline bool mm_match_cgroup(struct mm_struct *mm,
411 struct mem_cgroup *memcg)
2e4d4091 412{
587af308 413 struct mem_cgroup *task_memcg;
413918bb 414 bool match = false;
c3ac9a8a 415
2e4d4091 416 rcu_read_lock();
587af308 417 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
413918bb 418 if (task_memcg)
2314b42d 419 match = mem_cgroup_is_descendant(task_memcg, memcg);
2e4d4091 420 rcu_read_unlock();
c3ac9a8a 421 return match;
2e4d4091 422}
8a9f3ccd 423
64219994 424struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
2fc04524 425ino_t page_cgroup_ino(struct page *page);
d324236b 426
eb01aaab
VD
427static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
428{
429 if (mem_cgroup_disabled())
430 return true;
431 return !!(memcg->css.flags & CSS_ONLINE);
432}
433
58ae83db
KH
434/*
435 * For memory reclaim.
436 */
889976db 437int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
33398cf2
MH
438
439void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
b4536f0c 440 int zid, int nr_pages);
33398cf2 441
0a6b76dd
VD
442unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
443 int nid, unsigned int lru_mask);
444
33398cf2
MH
445static inline
446unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
447{
ef8f2327 448 struct mem_cgroup_per_node *mz;
b4536f0c
MH
449 unsigned long nr_pages = 0;
450 int zid;
33398cf2 451
ef8f2327 452 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
b4536f0c
MH
453 for (zid = 0; zid < MAX_NR_ZONES; zid++)
454 nr_pages += mz->lru_zone_size[zid][lru];
455 return nr_pages;
456}
457
458static inline
459unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
460 enum lru_list lru, int zone_idx)
461{
462 struct mem_cgroup_per_node *mz;
463
464 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
465 return mz->lru_zone_size[zone_idx][lru];
33398cf2
MH
466}
467
b23afb93
TH
468void mem_cgroup_handle_over_high(void);
469
7c5f64f8
VD
470unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg);
471
64219994
MH
472void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
473 struct task_struct *p);
58ae83db 474
49426420 475static inline void mem_cgroup_oom_enable(void)
519e5247 476{
626ebc41
TH
477 WARN_ON(current->memcg_may_oom);
478 current->memcg_may_oom = 1;
519e5247
JW
479}
480
49426420 481static inline void mem_cgroup_oom_disable(void)
519e5247 482{
626ebc41
TH
483 WARN_ON(!current->memcg_may_oom);
484 current->memcg_may_oom = 0;
519e5247
JW
485}
486
3812c8c8
JW
487static inline bool task_in_memcg_oom(struct task_struct *p)
488{
626ebc41 489 return p->memcg_in_oom;
3812c8c8
JW
490}
491
49426420 492bool mem_cgroup_oom_synchronize(bool wait);
3812c8c8 493
c255a458 494#ifdef CONFIG_MEMCG_SWAP
c077719b
KH
495extern int do_swap_account;
496#endif
f8d66542 497
739f79fc
JW
498struct mem_cgroup *lock_page_memcg(struct page *page);
499void __unlock_page_memcg(struct mem_cgroup *memcg);
62cccb8c 500void unlock_page_memcg(struct page *page);
d7365e78 501
04fecbf5 502/* idx can be of type enum memcg_stat_item or node_stat_item */
ccda7f43 503static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
04fecbf5 504 int idx)
2a2e4885 505{
a983b5eb
JW
506 long x = atomic_long_read(&memcg->stat[idx]);
507#ifdef CONFIG_SMP
508 if (x < 0)
509 x = 0;
510#endif
511 return x;
2a2e4885
JW
512}
513
04fecbf5 514/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 515static inline void __mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 516 int idx, int val)
2a2e4885 517{
a983b5eb
JW
518 long x;
519
520 if (mem_cgroup_disabled())
521 return;
522
523 x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
524 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
525 atomic_long_add(x, &memcg->stat[idx]);
526 x = 0;
527 }
528 __this_cpu_write(memcg->stat_cpu->count[idx], x);
2a2e4885
JW
529}
530
04fecbf5 531/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 532static inline void mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 533 int idx, int val)
2a2e4885 534{
c3cc3911
JW
535 unsigned long flags;
536
537 local_irq_save(flags);
a983b5eb 538 __mod_memcg_state(memcg, idx, val);
c3cc3911 539 local_irq_restore(flags);
2a2e4885
JW
540}
541
33398cf2 542/**
ccda7f43 543 * mod_memcg_page_state - update page state statistics
62cccb8c 544 * @page: the page
33398cf2
MH
545 * @idx: page state item to account
546 * @val: number of pages (positive or negative)
547 *
fdf1cdb9
JW
548 * The @page must be locked or the caller must use lock_page_memcg()
549 * to prevent double accounting when the page is concurrently being
550 * moved to another memcg:
81f8c3a4 551 *
fdf1cdb9 552 * lock_page(page) or lock_page_memcg(page)
81f8c3a4 553 * if (TestClearPageState(page))
ccda7f43 554 * mod_memcg_page_state(page, state, -1);
fdf1cdb9 555 * unlock_page(page) or unlock_page_memcg(page)
2a2e4885
JW
556 *
557 * Kernel pages are an exception to this, since they'll never move.
33398cf2 558 */
00f3ca2c 559static inline void __mod_memcg_page_state(struct page *page,
04fecbf5 560 int idx, int val)
00f3ca2c
JW
561{
562 if (page->mem_cgroup)
563 __mod_memcg_state(page->mem_cgroup, idx, val);
564}
565
ccda7f43 566static inline void mod_memcg_page_state(struct page *page,
04fecbf5 567 int idx, int val)
33398cf2 568{
62cccb8c 569 if (page->mem_cgroup)
ccda7f43 570 mod_memcg_state(page->mem_cgroup, idx, val);
33398cf2
MH
571}
572
00f3ca2c
JW
573static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
574 enum node_stat_item idx)
2a7106f2 575{
00f3ca2c 576 struct mem_cgroup_per_node *pn;
a983b5eb 577 long x;
00f3ca2c
JW
578
579 if (mem_cgroup_disabled())
580 return node_page_state(lruvec_pgdat(lruvec), idx);
581
582 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
a983b5eb
JW
583 x = atomic_long_read(&pn->lruvec_stat[idx]);
584#ifdef CONFIG_SMP
585 if (x < 0)
586 x = 0;
587#endif
588 return x;
2a7106f2
GT
589}
590
00f3ca2c
JW
591static inline void __mod_lruvec_state(struct lruvec *lruvec,
592 enum node_stat_item idx, int val)
2a7106f2 593{
00f3ca2c 594 struct mem_cgroup_per_node *pn;
a983b5eb 595 long x;
00f3ca2c 596
28454265 597 /* Update node */
00f3ca2c 598 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
28454265 599
00f3ca2c
JW
600 if (mem_cgroup_disabled())
601 return;
28454265 602
00f3ca2c 603 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
28454265
JW
604
605 /* Update memcg */
00f3ca2c 606 __mod_memcg_state(pn->memcg, idx, val);
28454265
JW
607
608 /* Update lruvec */
a983b5eb
JW
609 x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
610 if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
611 atomic_long_add(x, &pn->lruvec_stat[idx]);
612 x = 0;
613 }
614 __this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
00f3ca2c
JW
615}
616
617static inline void mod_lruvec_state(struct lruvec *lruvec,
618 enum node_stat_item idx, int val)
619{
c3cc3911
JW
620 unsigned long flags;
621
622 local_irq_save(flags);
28454265 623 __mod_lruvec_state(lruvec, idx, val);
c3cc3911 624 local_irq_restore(flags);
00f3ca2c
JW
625}
626
627static inline void __mod_lruvec_page_state(struct page *page,
628 enum node_stat_item idx, int val)
629{
28454265
JW
630 pg_data_t *pgdat = page_pgdat(page);
631 struct lruvec *lruvec;
00f3ca2c 632
28454265
JW
633 /* Untracked pages have no memcg, no lruvec. Update only the node */
634 if (!page->mem_cgroup) {
635 __mod_node_page_state(pgdat, idx, val);
00f3ca2c 636 return;
28454265
JW
637 }
638
639 lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
640 __mod_lruvec_state(lruvec, idx, val);
00f3ca2c
JW
641}
642
643static inline void mod_lruvec_page_state(struct page *page,
644 enum node_stat_item idx, int val)
645{
c3cc3911
JW
646 unsigned long flags;
647
648 local_irq_save(flags);
28454265 649 __mod_lruvec_page_state(page, idx, val);
c3cc3911 650 local_irq_restore(flags);
2a7106f2
GT
651}
652
ef8f2327 653unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
0608f43d
AM
654 gfp_t gfp_mask,
655 unsigned long *total_scanned);
a63d83f4 656
c9019e9b 657static inline void __count_memcg_events(struct mem_cgroup *memcg,
e27be240
JW
658 enum vm_event_item idx,
659 unsigned long count)
c9019e9b 660{
a983b5eb
JW
661 unsigned long x;
662
663 if (mem_cgroup_disabled())
664 return;
665
666 x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
667 if (unlikely(x > MEMCG_CHARGE_BATCH)) {
668 atomic_long_add(x, &memcg->events[idx]);
669 x = 0;
670 }
671 __this_cpu_write(memcg->stat_cpu->events[idx], x);
c9019e9b
JW
672}
673
2262185c 674static inline void count_memcg_events(struct mem_cgroup *memcg,
e27be240
JW
675 enum vm_event_item idx,
676 unsigned long count)
2262185c 677{
c3cc3911
JW
678 unsigned long flags;
679
680 local_irq_save(flags);
a983b5eb 681 __count_memcg_events(memcg, idx, count);
c3cc3911 682 local_irq_restore(flags);
2262185c
RG
683}
684
685static inline void count_memcg_page_event(struct page *page,
e27be240 686 enum vm_event_item idx)
2262185c
RG
687{
688 if (page->mem_cgroup)
689 count_memcg_events(page->mem_cgroup, idx, 1);
690}
691
692static inline void count_memcg_event_mm(struct mm_struct *mm,
693 enum vm_event_item idx)
68ae564b 694{
33398cf2
MH
695 struct mem_cgroup *memcg;
696
68ae564b
DR
697 if (mem_cgroup_disabled())
698 return;
33398cf2
MH
699
700 rcu_read_lock();
701 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
8e675f7a 702 if (likely(memcg)) {
c9019e9b 703 count_memcg_events(memcg, idx, 1);
8e675f7a
KK
704 if (idx == OOM_KILL)
705 cgroup_file_notify(&memcg->events_file);
706 }
33398cf2 707 rcu_read_unlock();
68ae564b 708}
c9019e9b 709
e27be240
JW
710static inline void memcg_memory_event(struct mem_cgroup *memcg,
711 enum memcg_memory_event event)
c9019e9b 712{
e27be240 713 atomic_long_inc(&memcg->memory_events[event]);
c9019e9b
JW
714 cgroup_file_notify(&memcg->events_file);
715}
716
ca3e0214 717#ifdef CONFIG_TRANSPARENT_HUGEPAGE
e94c8a9c 718void mem_cgroup_split_huge_fixup(struct page *head);
ca3e0214
KH
719#endif
720
c255a458 721#else /* CONFIG_MEMCG */
23047a96
JW
722
723#define MEM_CGROUP_ID_SHIFT 0
724#define MEM_CGROUP_ID_MAX 0
725
7a81b88c
KH
726struct mem_cgroup;
727
23047a96
JW
728static inline bool mem_cgroup_disabled(void)
729{
730 return true;
731}
732
e27be240
JW
733static inline void memcg_memory_event(struct mem_cgroup *memcg,
734 enum memcg_memory_event event)
241994ed
JW
735{
736}
737
738static inline bool mem_cgroup_low(struct mem_cgroup *root,
739 struct mem_cgroup *memcg)
740{
741 return false;
742}
743
00501b53
JW
744static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
745 gfp_t gfp_mask,
f627c2f5
KS
746 struct mem_cgroup **memcgp,
747 bool compound)
7a81b88c 748{
00501b53 749 *memcgp = NULL;
7a81b88c
KH
750 return 0;
751}
752
00501b53
JW
753static inline void mem_cgroup_commit_charge(struct page *page,
754 struct mem_cgroup *memcg,
f627c2f5 755 bool lrucare, bool compound)
7a81b88c
KH
756{
757}
758
00501b53 759static inline void mem_cgroup_cancel_charge(struct page *page,
f627c2f5
KS
760 struct mem_cgroup *memcg,
761 bool compound)
7a81b88c
KH
762{
763}
764
0a31bc97 765static inline void mem_cgroup_uncharge(struct page *page)
569b846d
KH
766{
767}
768
747db954 769static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
8a9f3ccd
BS
770{
771}
772
6a93ca8f 773static inline void mem_cgroup_migrate(struct page *old, struct page *new)
69029cd5
KH
774{
775}
776
a9dd0a83 777static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
ef8f2327 778 struct mem_cgroup *memcg)
08e552c6 779{
a9dd0a83 780 return node_lruvec(pgdat);
08e552c6
KH
781}
782
fa9add64 783static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
599d0c95 784 struct pglist_data *pgdat)
66e1707b 785{
599d0c95 786 return &pgdat->lruvec;
66e1707b
BS
787}
788
587af308 789static inline bool mm_match_cgroup(struct mm_struct *mm,
c0ff4b85 790 struct mem_cgroup *memcg)
bed7161a 791{
587af308 792 return true;
bed7161a
BS
793}
794
ffbdccf5
DR
795static inline bool task_in_mem_cgroup(struct task_struct *task,
796 const struct mem_cgroup *memcg)
4c4a2214 797{
ffbdccf5 798 return true;
4c4a2214
DR
799}
800
5660048c
JW
801static inline struct mem_cgroup *
802mem_cgroup_iter(struct mem_cgroup *root,
803 struct mem_cgroup *prev,
804 struct mem_cgroup_reclaim_cookie *reclaim)
805{
806 return NULL;
807}
808
809static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
810 struct mem_cgroup *prev)
811{
812}
813
7c5f64f8
VD
814static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
815 int (*fn)(struct task_struct *, void *), void *arg)
816{
817 return 0;
818}
819
23047a96 820static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
f8d66542 821{
23047a96
JW
822 return 0;
823}
824
825static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
826{
827 WARN_ON_ONCE(id);
828 /* XXX: This should always return root_mem_cgroup */
829 return NULL;
f8d66542 830}
a636b327 831
2262185c
RG
832static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
833{
834 return NULL;
835}
836
eb01aaab 837static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
14797e23 838{
13308ca9 839 return true;
14797e23
KM
840}
841
a3d8e054 842static inline unsigned long
4d7dcca2 843mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
a3d8e054
KM
844{
845 return 0;
846}
b4536f0c
MH
847static inline
848unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
849 enum lru_list lru, int zone_idx)
850{
851 return 0;
852}
a3d8e054 853
0a6b76dd
VD
854static inline unsigned long
855mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
856 int nid, unsigned int lru_mask)
857{
858 return 0;
859}
860
7c5f64f8
VD
861static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
862{
863 return 0;
864}
865
e222432b
BS
866static inline void
867mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
868{
869}
870
739f79fc
JW
871static inline struct mem_cgroup *lock_page_memcg(struct page *page)
872{
873 return NULL;
874}
875
876static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
89c06bd5
KH
877{
878}
879
62cccb8c 880static inline void unlock_page_memcg(struct page *page)
89c06bd5
KH
881{
882}
883
b23afb93
TH
884static inline void mem_cgroup_handle_over_high(void)
885{
886}
887
49426420 888static inline void mem_cgroup_oom_enable(void)
519e5247
JW
889{
890}
891
49426420 892static inline void mem_cgroup_oom_disable(void)
519e5247
JW
893{
894}
895
3812c8c8
JW
896static inline bool task_in_memcg_oom(struct task_struct *p)
897{
898 return false;
899}
900
49426420 901static inline bool mem_cgroup_oom_synchronize(bool wait)
3812c8c8
JW
902{
903 return false;
904}
905
ccda7f43 906static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
04fecbf5 907 int idx)
2a2e4885
JW
908{
909 return 0;
910}
911
00f3ca2c 912static inline void __mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 913 int idx,
00f3ca2c 914 int nr)
2a2e4885
JW
915{
916}
917
00f3ca2c 918static inline void mod_memcg_state(struct mem_cgroup *memcg,
04fecbf5 919 int idx,
00f3ca2c 920 int nr)
2a2e4885
JW
921{
922}
923
00f3ca2c 924static inline void __mod_memcg_page_state(struct page *page,
04fecbf5 925 int idx,
00f3ca2c 926 int nr)
2a2e4885
JW
927{
928}
929
ccda7f43 930static inline void mod_memcg_page_state(struct page *page,
04fecbf5 931 int idx,
ccda7f43 932 int nr)
553af430
JW
933{
934}
935
00f3ca2c
JW
936static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
937 enum node_stat_item idx)
2a7106f2 938{
00f3ca2c 939 return node_page_state(lruvec_pgdat(lruvec), idx);
2a7106f2
GT
940}
941
00f3ca2c
JW
942static inline void __mod_lruvec_state(struct lruvec *lruvec,
943 enum node_stat_item idx, int val)
d69b042f 944{
00f3ca2c
JW
945 __mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
946}
947
948static inline void mod_lruvec_state(struct lruvec *lruvec,
949 enum node_stat_item idx, int val)
950{
951 mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
952}
953
954static inline void __mod_lruvec_page_state(struct page *page,
955 enum node_stat_item idx, int val)
956{
957 __mod_node_page_state(page_pgdat(page), idx, val);
958}
959
960static inline void mod_lruvec_page_state(struct page *page,
961 enum node_stat_item idx, int val)
962{
963 mod_node_page_state(page_pgdat(page), idx, val);
d69b042f
BS
964}
965
4e416953 966static inline
ef8f2327 967unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
0608f43d
AM
968 gfp_t gfp_mask,
969 unsigned long *total_scanned)
4e416953 970{
0608f43d 971 return 0;
4e416953
BS
972}
973
e94c8a9c 974static inline void mem_cgroup_split_huge_fixup(struct page *head)
ca3e0214
KH
975{
976}
977
2262185c
RG
978static inline void count_memcg_events(struct mem_cgroup *memcg,
979 enum vm_event_item idx,
980 unsigned long count)
981{
982}
983
984static inline void count_memcg_page_event(struct page *page,
04fecbf5 985 int idx)
2262185c
RG
986{
987}
988
456f998e 989static inline
2262185c 990void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
456f998e
YH
991{
992}
c255a458 993#endif /* CONFIG_MEMCG */
78fb7466 994
04fecbf5 995/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 996static inline void __inc_memcg_state(struct mem_cgroup *memcg,
04fecbf5 997 int idx)
00f3ca2c
JW
998{
999 __mod_memcg_state(memcg, idx, 1);
1000}
1001
04fecbf5 1002/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1003static inline void __dec_memcg_state(struct mem_cgroup *memcg,
04fecbf5 1004 int idx)
00f3ca2c
JW
1005{
1006 __mod_memcg_state(memcg, idx, -1);
1007}
1008
04fecbf5 1009/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1010static inline void __inc_memcg_page_state(struct page *page,
04fecbf5 1011 int idx)
00f3ca2c
JW
1012{
1013 __mod_memcg_page_state(page, idx, 1);
1014}
1015
04fecbf5 1016/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1017static inline void __dec_memcg_page_state(struct page *page,
04fecbf5 1018 int idx)
00f3ca2c
JW
1019{
1020 __mod_memcg_page_state(page, idx, -1);
1021}
1022
1023static inline void __inc_lruvec_state(struct lruvec *lruvec,
1024 enum node_stat_item idx)
1025{
1026 __mod_lruvec_state(lruvec, idx, 1);
1027}
1028
1029static inline void __dec_lruvec_state(struct lruvec *lruvec,
1030 enum node_stat_item idx)
1031{
1032 __mod_lruvec_state(lruvec, idx, -1);
1033}
1034
1035static inline void __inc_lruvec_page_state(struct page *page,
1036 enum node_stat_item idx)
1037{
1038 __mod_lruvec_page_state(page, idx, 1);
1039}
1040
1041static inline void __dec_lruvec_page_state(struct page *page,
1042 enum node_stat_item idx)
1043{
1044 __mod_lruvec_page_state(page, idx, -1);
1045}
1046
04fecbf5 1047/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1048static inline void inc_memcg_state(struct mem_cgroup *memcg,
04fecbf5 1049 int idx)
00f3ca2c
JW
1050{
1051 mod_memcg_state(memcg, idx, 1);
1052}
1053
04fecbf5 1054/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1055static inline void dec_memcg_state(struct mem_cgroup *memcg,
04fecbf5 1056 int idx)
00f3ca2c
JW
1057{
1058 mod_memcg_state(memcg, idx, -1);
1059}
1060
04fecbf5 1061/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1062static inline void inc_memcg_page_state(struct page *page,
04fecbf5 1063 int idx)
00f3ca2c
JW
1064{
1065 mod_memcg_page_state(page, idx, 1);
1066}
1067
04fecbf5 1068/* idx can be of type enum memcg_stat_item or node_stat_item */
00f3ca2c 1069static inline void dec_memcg_page_state(struct page *page,
04fecbf5 1070 int idx)
00f3ca2c
JW
1071{
1072 mod_memcg_page_state(page, idx, -1);
1073}
1074
1075static inline void inc_lruvec_state(struct lruvec *lruvec,
1076 enum node_stat_item idx)
1077{
1078 mod_lruvec_state(lruvec, idx, 1);
1079}
1080
1081static inline void dec_lruvec_state(struct lruvec *lruvec,
1082 enum node_stat_item idx)
1083{
1084 mod_lruvec_state(lruvec, idx, -1);
1085}
1086
1087static inline void inc_lruvec_page_state(struct page *page,
1088 enum node_stat_item idx)
1089{
1090 mod_lruvec_page_state(page, idx, 1);
1091}
1092
1093static inline void dec_lruvec_page_state(struct page *page,
1094 enum node_stat_item idx)
1095{
1096 mod_lruvec_page_state(page, idx, -1);
1097}
1098
52ebea74 1099#ifdef CONFIG_CGROUP_WRITEBACK
841710aa 1100
52ebea74 1101struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
841710aa 1102struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
c5edf9cd
TH
1103void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1104 unsigned long *pheadroom, unsigned long *pdirty,
1105 unsigned long *pwriteback);
841710aa
TH
1106
1107#else /* CONFIG_CGROUP_WRITEBACK */
1108
1109static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1110{
1111 return NULL;
1112}
1113
c2aa723a 1114static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
c5edf9cd
TH
1115 unsigned long *pfilepages,
1116 unsigned long *pheadroom,
c2aa723a
TH
1117 unsigned long *pdirty,
1118 unsigned long *pwriteback)
1119{
1120}
1121
841710aa 1122#endif /* CONFIG_CGROUP_WRITEBACK */
52ebea74 1123
e1aab161 1124struct sock;
baac50bb
JW
1125bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1126void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
d886f4e4 1127#ifdef CONFIG_MEMCG
ef12947c
JW
1128extern struct static_key_false memcg_sockets_enabled_key;
1129#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
2d758073
JW
1130void mem_cgroup_sk_alloc(struct sock *sk);
1131void mem_cgroup_sk_free(struct sock *sk);
baac50bb 1132static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
e805605c 1133{
0db15298 1134 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
8e8ae645 1135 return true;
8e8ae645
JW
1136 do {
1137 if (time_before(jiffies, memcg->socket_pressure))
1138 return true;
1139 } while ((memcg = parent_mem_cgroup(memcg)));
1140 return false;
e805605c
JW
1141}
1142#else
80e95fe0 1143#define mem_cgroup_sockets_enabled 0
2d758073
JW
1144static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1145static inline void mem_cgroup_sk_free(struct sock *sk) { };
baac50bb 1146static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
e805605c
JW
1147{
1148 return false;
1149}
1150#endif
7ae1e1d0 1151
45264778
VD
1152struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1153void memcg_kmem_put_cache(struct kmem_cache *cachep);
1154int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1155 struct mem_cgroup *memcg);
1156int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1157void memcg_kmem_uncharge(struct page *page, int order);
1158
127424c8 1159#if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
ef12947c 1160extern struct static_key_false memcg_kmem_enabled_key;
17cc4dfe 1161extern struct workqueue_struct *memcg_kmem_cache_wq;
749c5415 1162
dbcf73e2 1163extern int memcg_nr_cache_ids;
64219994
MH
1164void memcg_get_cache_ids(void);
1165void memcg_put_cache_ids(void);
ebe945c2
GC
1166
1167/*
1168 * Helper macro to loop through all memcg-specific caches. Callers must still
1169 * check if the cache is valid (it is either valid or NULL).
1170 * the slab_mutex must be held when looping through those caches
1171 */
749c5415 1172#define for_each_memcg_cache_index(_idx) \
dbcf73e2 1173 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
749c5415 1174
7ae1e1d0
GC
1175static inline bool memcg_kmem_enabled(void)
1176{
ef12947c 1177 return static_branch_unlikely(&memcg_kmem_enabled_key);
7ae1e1d0
GC
1178}
1179
33398cf2 1180/*
9f706d68 1181 * helper for accessing a memcg's index. It will be used as an index in the
33398cf2
MH
1182 * child cache array in kmem_cache, and also to derive its name. This function
1183 * will return -1 when this is not a kmem-limited memcg.
1184 */
1185static inline int memcg_cache_id(struct mem_cgroup *memcg)
1186{
1187 return memcg ? memcg->kmemcg_id : -1;
1188}
5722d094 1189
7ae1e1d0 1190#else
749c5415
GC
1191#define for_each_memcg_cache_index(_idx) \
1192 for (; NULL; )
1193
b9ce5ef4
GC
1194static inline bool memcg_kmem_enabled(void)
1195{
1196 return false;
1197}
1198
2633d7a0
GC
1199static inline int memcg_cache_id(struct mem_cgroup *memcg)
1200{
1201 return -1;
1202}
1203
05257a1a
VD
1204static inline void memcg_get_cache_ids(void)
1205{
1206}
1207
1208static inline void memcg_put_cache_ids(void)
1209{
1210}
1211
127424c8
JW
1212#endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1213
8cdea7c0 1214#endif /* _LINUX_MEMCONTROL_H */