blob: d0c724f53691e9de36a8959331271e7989764f88 [file] [log] [blame]
Balbir Singh8cdea7c2008-02-07 00:13:50 -08001/* memcontrol.h - Memory Controller
2 *
3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
5 *
Pavel Emelianov78fb7462008-02-07 00:13:51 -08006 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org>
8 *
Balbir Singh8cdea7c2008-02-07 00:13:50 -08009 * 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
Hirokazu Takahashif8d665422009-01-07 18:08:02 -080022#include <linux/cgroup.h>
Ying Han456f9982011-05-26 16:25:38 -070023#include <linux/vm_event_item.h>
Glauber Costa7ae1e1d2012-12-18 14:21:56 -080024#include <linux/hardirq.h>
Glauber Costaa8964b92012-12-18 14:22:09 -080025#include <linux/jump_label.h>
Michal Hocko33398cf2015-09-08 15:01:02 -070026#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>
Ying Han456f9982011-05-26 16:25:38 -070031
Pavel Emelianov78fb7462008-02-07 00:13:51 -080032struct mem_cgroup;
Balbir Singh8697d332008-02-07 00:13:59 -080033struct page;
34struct mm_struct;
Glauber Costa2633d7a2012-12-18 14:22:34 -080035struct kmem_cache;
Pavel Emelianov78fb7462008-02-07 00:13:51 -080036
Sha Zhengju68b48762013-09-12 15:13:50 -070037/*
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 */
Greg Thelenc4843a72015-05-22 17:13:16 -040049 MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */
Sha Zhengju3ea67d02013-09-12 15:13:53 -070050 MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
Sha Zhengju68b48762013-09-12 15:13:50 -070051 MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
52 MEM_CGROUP_STAT_NSTATS,
Greg Thelen2a7106f2011-01-13 15:47:37 -080053};
54
Johannes Weiner56600482012-01-12 17:17:59 -080055struct mem_cgroup_reclaim_cookie {
56 struct zone *zone;
57 int priority;
58 unsigned int generation;
59};
60
Johannes Weiner241994ed2015-02-11 15:26:06 -080061enum mem_cgroup_events_index {
62 MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */
63 MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */
64 MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */
65 MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */
66 MEM_CGROUP_EVENTS_NSTATS,
67 /* default hierarchy events */
68 MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
69 MEMCG_HIGH,
70 MEMCG_MAX,
71 MEMCG_OOM,
72 MEMCG_NR_EVENTS,
73};
74
Michal Hocko33398cf2015-09-08 15:01:02 -070075/*
76 * Per memcg event counter is incremented at every pagein/pageout. With THP,
77 * it will be incremated by the number of pages. This counter is used for
78 * for trigger some periodic events. This is straightforward and better
79 * than using jiffies etc. to handle periodic memcg event.
80 */
81enum mem_cgroup_events_target {
82 MEM_CGROUP_TARGET_THRESH,
83 MEM_CGROUP_TARGET_SOFTLIMIT,
84 MEM_CGROUP_TARGET_NUMAINFO,
85 MEM_CGROUP_NTARGETS,
86};
87
Michal Hocko33398cf2015-09-08 15:01:02 -070088struct cg_proto {
89 struct page_counter memory_allocated; /* Current allocated memory. */
90 struct percpu_counter sockets_allocated; /* Current number of sockets. */
91 int memory_pressure;
Vladimir Davydov9ee11ba2016-01-14 15:19:41 -080092 bool active;
Michal Hocko33398cf2015-09-08 15:01:02 -070093 long sysctl_mem[3];
Michal Hocko33398cf2015-09-08 15:01:02 -070094 /*
95 * memcg field is used to find which memcg we belong directly
96 * Each memcg struct can hold more than one cg_proto, so container_of
97 * won't really cut.
98 *
99 * The elegant solution would be having an inverse function to
100 * proto_cgroup in struct proto, but that means polluting the structure
101 * for everybody, instead of just for memcg users.
102 */
103 struct mem_cgroup *memcg;
104};
105
Andrew Mortonc255a452012-07-31 16:43:02 -0700106#ifdef CONFIG_MEMCG
Michal Hocko33398cf2015-09-08 15:01:02 -0700107struct mem_cgroup_stat_cpu {
108 long count[MEM_CGROUP_STAT_NSTATS];
109 unsigned long events[MEMCG_NR_EVENTS];
110 unsigned long nr_page_events;
111 unsigned long targets[MEM_CGROUP_NTARGETS];
112};
113
114struct mem_cgroup_reclaim_iter {
115 struct mem_cgroup *position;
116 /* scan generation, increased every round-trip */
117 unsigned int generation;
118};
119
120/*
121 * per-zone information in memory controller.
122 */
123struct mem_cgroup_per_zone {
124 struct lruvec lruvec;
125 unsigned long lru_size[NR_LRU_LISTS];
126
127 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
128
129 struct rb_node tree_node; /* RB tree node */
130 unsigned long usage_in_excess;/* Set to the value by which */
131 /* the soft limit is exceeded*/
132 bool on_tree;
133 struct mem_cgroup *memcg; /* Back pointer, we cannot */
134 /* use container_of */
135};
136
137struct mem_cgroup_per_node {
138 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
139};
140
141struct mem_cgroup_threshold {
142 struct eventfd_ctx *eventfd;
143 unsigned long threshold;
144};
145
146/* For threshold */
147struct mem_cgroup_threshold_ary {
148 /* An array index points to threshold just below or equal to usage. */
149 int current_threshold;
150 /* Size of entries[] */
151 unsigned int size;
152 /* Array of thresholds */
153 struct mem_cgroup_threshold entries[0];
154};
155
156struct mem_cgroup_thresholds {
157 /* Primary thresholds array */
158 struct mem_cgroup_threshold_ary *primary;
159 /*
160 * Spare threshold array.
161 * This is needed to make mem_cgroup_unregister_event() "never fail".
162 * It must be able to store at least primary->size - 1 entries.
163 */
164 struct mem_cgroup_threshold_ary *spare;
165};
166
167/*
168 * The memory controller data structure. The memory controller controls both
169 * page cache and RSS per cgroup. We would eventually like to provide
170 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
171 * to help the administrator determine what knobs to tune.
172 */
173struct mem_cgroup {
174 struct cgroup_subsys_state css;
175
176 /* Accounted resources */
177 struct page_counter memory;
178 struct page_counter memsw;
179 struct page_counter kmem;
180
181 /* Normal memory consumption range */
182 unsigned long low;
183 unsigned long high;
184
185 unsigned long soft_limit;
186
187 /* vmpressure notifications */
188 struct vmpressure vmpressure;
189
190 /* css_online() has been completed */
191 int initialized;
192
193 /*
194 * Should the accounting and control be hierarchical, per subtree?
195 */
196 bool use_hierarchy;
197
198 /* protected by memcg_oom_lock */
199 bool oom_lock;
200 int under_oom;
201
202 int swappiness;
203 /* OOM-Killer disable */
204 int oom_kill_disable;
205
Tejun Heo472912a2015-09-18 18:01:59 -0400206 /* handle for "memory.events" */
207 struct cgroup_file events_file;
208
Michal Hocko33398cf2015-09-08 15:01:02 -0700209 /* protect arrays of thresholds */
210 struct mutex thresholds_lock;
211
212 /* thresholds for memory usage. RCU-protected */
213 struct mem_cgroup_thresholds thresholds;
214
215 /* thresholds for mem+swap usage. RCU-protected */
216 struct mem_cgroup_thresholds memsw_thresholds;
217
218 /* For oom notifier event fd */
219 struct list_head oom_notify;
220
221 /*
222 * Should we move charges of a task when a task is moved into this
223 * mem_cgroup ? And what type of charges should we move ?
224 */
225 unsigned long move_charge_at_immigrate;
226 /*
227 * set > 0 if pages under this cgroup are moving to other cgroup.
228 */
229 atomic_t moving_account;
230 /* taken only while moving_account > 0 */
231 spinlock_t move_lock;
232 struct task_struct *move_lock_task;
233 unsigned long move_lock_flags;
234 /*
235 * percpu counter.
236 */
237 struct mem_cgroup_stat_cpu __percpu *stat;
Michal Hocko33398cf2015-09-08 15:01:02 -0700238
239#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
240 struct cg_proto tcp_mem;
241#endif
242#if defined(CONFIG_MEMCG_KMEM)
243 /* Index in the kmem_cache->memcg_params.memcg_caches array */
244 int kmemcg_id;
245 bool kmem_acct_activated;
246 bool kmem_acct_active;
247#endif
248
249 int last_scanned_node;
250#if MAX_NUMNODES > 1
251 nodemask_t scan_nodes;
252 atomic_t numainfo_events;
253 atomic_t numainfo_updating;
254#endif
255
256#ifdef CONFIG_CGROUP_WRITEBACK
257 struct list_head cgwb_list;
258 struct wb_domain cgwb_domain;
259#endif
260
261 /* List of events which userspace want to receive */
262 struct list_head event_list;
263 spinlock_t event_list_lock;
264
265 struct mem_cgroup_per_node *nodeinfo[0];
266 /* WARNING: nodeinfo must be the last member here */
267};
Johannes Weiner7d828602016-01-14 15:20:56 -0800268
269extern struct mem_cgroup *root_mem_cgroup;
Tejun Heo56161632015-05-22 17:13:20 -0400270
Michal Hocko33398cf2015-09-08 15:01:02 -0700271/**
272 * mem_cgroup_events - count memory events against a cgroup
273 * @memcg: the memory cgroup
274 * @idx: the event index
275 * @nr: the number of events to account for
276 */
277static inline void mem_cgroup_events(struct mem_cgroup *memcg,
Johannes Weiner241994ed2015-02-11 15:26:06 -0800278 enum mem_cgroup_events_index idx,
Michal Hocko33398cf2015-09-08 15:01:02 -0700279 unsigned int nr)
280{
281 this_cpu_add(memcg->stat->events[idx], nr);
Tejun Heo472912a2015-09-18 18:01:59 -0400282 cgroup_file_notify(&memcg->events_file);
Michal Hocko33398cf2015-09-08 15:01:02 -0700283}
Johannes Weiner241994ed2015-02-11 15:26:06 -0800284
285bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
286
Johannes Weiner00501b52014-08-08 14:19:20 -0700287int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
288 gfp_t gfp_mask, struct mem_cgroup **memcgp);
289void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
290 bool lrucare);
291void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700292void mem_cgroup_uncharge(struct page *page);
Johannes Weiner747db952014-08-08 14:19:24 -0700293void mem_cgroup_uncharge_list(struct list_head *page_list);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700294
Hugh Dickins45637ba2015-11-05 18:49:40 -0800295void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700296
Johannes Weiner925b7672012-01-12 17:18:15 -0800297struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700298struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -0800299
Johannes Weiner2314b422014-12-10 15:44:33 -0800300bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
Michal Hocko64219992015-09-08 15:01:07 -0700301struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
Michal Hocko64219992015-09-08 15:01:07 -0700302struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
Vladimir Davydove993d902015-09-09 15:35:35 -0700303
Michal Hocko33398cf2015-09-08 15:01:02 -0700304static inline
305struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
306 return css ? container_of(css, struct mem_cgroup, css) : NULL;
307}
308
309struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
310 struct mem_cgroup *,
311 struct mem_cgroup_reclaim_cookie *);
312void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
313
314static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
315 struct mem_cgroup *root)
316{
317 if (root == memcg)
318 return true;
319 if (!root->use_hierarchy)
320 return false;
321 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
322}
Glauber Costae1aab162011-12-11 21:47:03 +0000323
Johannes Weiner2314b422014-12-10 15:44:33 -0800324static inline bool mm_match_cgroup(struct mm_struct *mm,
325 struct mem_cgroup *memcg)
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800326{
Johannes Weiner587af302012-10-08 16:34:12 -0700327 struct mem_cgroup *task_memcg;
Johannes Weiner413918b2014-12-10 15:44:30 -0800328 bool match = false;
Johannes Weinerc3ac9a82012-05-29 15:06:25 -0700329
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800330 rcu_read_lock();
Johannes Weiner587af302012-10-08 16:34:12 -0700331 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
Johannes Weiner413918b2014-12-10 15:44:30 -0800332 if (task_memcg)
Johannes Weiner2314b422014-12-10 15:44:33 -0800333 match = mem_cgroup_is_descendant(task_memcg, memcg);
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800334 rcu_read_unlock();
Johannes Weinerc3ac9a82012-05-29 15:06:25 -0700335 return match;
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800336}
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800337
Michal Hocko64219992015-09-08 15:01:07 -0700338struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
Vladimir Davydov2fc04522015-09-09 15:35:28 -0700339ino_t page_cgroup_ino(struct page *page);
Wu Fengguangd3242362009-12-16 12:19:59 +0100340
Michal Hocko33398cf2015-09-08 15:01:02 -0700341static inline bool mem_cgroup_disabled(void)
342{
Tejun Heofc5ed1e2015-09-18 11:56:28 -0400343 return !cgroup_subsys_enabled(memory_cgrp_subsys);
Michal Hocko33398cf2015-09-08 15:01:02 -0700344}
Johannes Weiner56600482012-01-12 17:17:59 -0800345
KAMEZAWA Hiroyuki58ae83d2008-02-07 00:14:32 -0800346/*
347 * For memory reclaim.
348 */
Ying Han889976d2011-05-26 16:25:33 -0700349int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
Michal Hocko33398cf2015-09-08 15:01:02 -0700350
351void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
352 int nr_pages);
353
354static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
355{
356 struct mem_cgroup_per_zone *mz;
357 struct mem_cgroup *memcg;
358
359 if (mem_cgroup_disabled())
360 return true;
361
362 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
363 memcg = mz->memcg;
364
365 return !!(memcg->css.flags & CSS_ONLINE);
366}
367
368static inline
369unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
370{
371 struct mem_cgroup_per_zone *mz;
372
373 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
374 return mz->lru_size[lru];
375}
376
Yaowei Bai13308ca2015-11-05 18:47:40 -0800377static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
Michal Hocko33398cf2015-09-08 15:01:02 -0700378{
379 unsigned long inactive_ratio;
380 unsigned long inactive;
381 unsigned long active;
382 unsigned long gb;
383
384 inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
385 active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
386
387 gb = (inactive + active) >> (30 - PAGE_SHIFT);
388 if (gb)
389 inactive_ratio = int_sqrt(10 * gb);
390 else
391 inactive_ratio = 1;
392
393 return inactive * inactive_ratio < active;
394}
395
Tejun Heob23afb92015-11-05 18:46:11 -0800396void mem_cgroup_handle_over_high(void);
397
Michal Hocko64219992015-09-08 15:01:07 -0700398void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
399 struct task_struct *p);
KAMEZAWA Hiroyuki58ae83d2008-02-07 00:14:32 -0800400
Johannes Weiner49426422013-10-16 13:46:59 -0700401static inline void mem_cgroup_oom_enable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700402{
Tejun Heo626ebc42015-11-05 18:46:09 -0800403 WARN_ON(current->memcg_may_oom);
404 current->memcg_may_oom = 1;
Johannes Weiner519e5242013-09-12 15:13:42 -0700405}
406
Johannes Weiner49426422013-10-16 13:46:59 -0700407static inline void mem_cgroup_oom_disable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700408{
Tejun Heo626ebc42015-11-05 18:46:09 -0800409 WARN_ON(!current->memcg_may_oom);
410 current->memcg_may_oom = 0;
Johannes Weiner519e5242013-09-12 15:13:42 -0700411}
412
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700413static inline bool task_in_memcg_oom(struct task_struct *p)
414{
Tejun Heo626ebc42015-11-05 18:46:09 -0800415 return p->memcg_in_oom;
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700416}
417
Johannes Weiner49426422013-10-16 13:46:59 -0700418bool mem_cgroup_oom_synchronize(bool wait);
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700419
Andrew Mortonc255a452012-07-31 16:43:02 -0700420#ifdef CONFIG_MEMCG_SWAP
KAMEZAWA Hiroyukic0777192009-01-07 18:07:57 -0800421extern int do_swap_account;
422#endif
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800423
Johannes Weiner6de22612015-02-11 15:25:01 -0800424struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
Johannes Weiner6de22612015-02-11 15:25:01 -0800425void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700426
Michal Hocko33398cf2015-09-08 15:01:02 -0700427/**
428 * mem_cgroup_update_page_stat - update page state statistics
429 * @memcg: memcg to account against
430 * @idx: page state item to account
431 * @val: number of pages (positive or negative)
432 *
433 * See mem_cgroup_begin_page_stat() for locking requirements.
434 */
435static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
436 enum mem_cgroup_stat_index idx, int val)
437{
438 VM_BUG_ON(!rcu_read_lock_held());
439
440 if (memcg)
441 this_cpu_add(memcg->stat->count[idx], val);
442}
443
Johannes Weinerd7365e72014-10-29 14:50:48 -0700444static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700445 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800446{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700447 mem_cgroup_update_page_stat(memcg, idx, 1);
Greg Thelen2a7106f2011-01-13 15:47:37 -0800448}
449
Johannes Weinerd7365e72014-10-29 14:50:48 -0700450static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700451 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800452{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700453 mem_cgroup_update_page_stat(memcg, idx, -1);
Greg Thelen2a7106f2011-01-13 15:47:37 -0800454}
455
Andrew Morton0608f432013-09-24 15:27:41 -0700456unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
457 gfp_t gfp_mask,
458 unsigned long *total_scanned);
David Rientjesa63d83f2010-08-09 17:19:46 -0700459
David Rientjes68ae5642012-12-12 13:51:57 -0800460static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
461 enum vm_event_item idx)
462{
Michal Hocko33398cf2015-09-08 15:01:02 -0700463 struct mem_cgroup *memcg;
464
David Rientjes68ae5642012-12-12 13:51:57 -0800465 if (mem_cgroup_disabled())
466 return;
Michal Hocko33398cf2015-09-08 15:01:02 -0700467
468 rcu_read_lock();
469 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
470 if (unlikely(!memcg))
471 goto out;
472
473 switch (idx) {
474 case PGFAULT:
475 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
476 break;
477 case PGMAJFAULT:
478 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
479 break;
480 default:
481 BUG();
482 }
483out:
484 rcu_read_unlock();
David Rientjes68ae5642012-12-12 13:51:57 -0800485}
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800486#ifdef CONFIG_TRANSPARENT_HUGEPAGE
KAMEZAWA Hiroyukie94c8a92012-01-12 17:18:20 -0800487void mem_cgroup_split_huge_fixup(struct page *head);
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800488#endif
489
Andrew Mortonc255a452012-07-31 16:43:02 -0700490#else /* CONFIG_MEMCG */
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800491struct mem_cgroup;
492
Johannes Weiner241994ed2015-02-11 15:26:06 -0800493static inline void mem_cgroup_events(struct mem_cgroup *memcg,
494 enum mem_cgroup_events_index idx,
495 unsigned int nr)
496{
497}
498
499static inline bool mem_cgroup_low(struct mem_cgroup *root,
500 struct mem_cgroup *memcg)
501{
502 return false;
503}
504
Johannes Weiner00501b52014-08-08 14:19:20 -0700505static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
506 gfp_t gfp_mask,
507 struct mem_cgroup **memcgp)
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800508{
Johannes Weiner00501b52014-08-08 14:19:20 -0700509 *memcgp = NULL;
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800510 return 0;
511}
512
Johannes Weiner00501b52014-08-08 14:19:20 -0700513static inline void mem_cgroup_commit_charge(struct page *page,
514 struct mem_cgroup *memcg,
515 bool lrucare)
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800516{
517}
518
Johannes Weiner00501b52014-08-08 14:19:20 -0700519static inline void mem_cgroup_cancel_charge(struct page *page,
520 struct mem_cgroup *memcg)
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800521{
522}
523
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700524static inline void mem_cgroup_uncharge(struct page *page)
525{
526}
527
Johannes Weiner747db952014-08-08 14:19:24 -0700528static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -0800529{
530}
531
Hugh Dickins45637ba2015-11-05 18:49:40 -0800532static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
KAMEZAWA Hiroyuki69029cd2008-07-25 01:47:14 -0700533{
534}
535
Johannes Weiner925b7672012-01-12 17:18:15 -0800536static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
537 struct mem_cgroup *memcg)
538{
539 return &zone->lruvec;
540}
541
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700542static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
543 struct zone *zone)
Minchan Kim3f58a822011-03-22 16:32:53 -0700544{
Johannes Weiner925b7672012-01-12 17:18:15 -0800545 return &zone->lruvec;
Balbir Singh66e17072008-02-07 00:13:56 -0800546}
547
Johannes Weiner587af302012-10-08 16:34:12 -0700548static inline bool mm_match_cgroup(struct mm_struct *mm,
Raghavendra K Tc0ff4b82011-11-02 13:38:15 -0700549 struct mem_cgroup *memcg)
Balbir Singhbed71612008-02-07 00:14:01 -0800550{
Johannes Weiner587af302012-10-08 16:34:12 -0700551 return true;
Balbir Singhbed71612008-02-07 00:14:01 -0800552}
553
David Rientjesffbdccf2013-07-03 15:01:23 -0700554static inline bool task_in_mem_cgroup(struct task_struct *task,
555 const struct mem_cgroup *memcg)
David Rientjes4c4a2212008-02-07 00:14:06 -0800556{
David Rientjesffbdccf2013-07-03 15:01:23 -0700557 return true;
David Rientjes4c4a2212008-02-07 00:14:06 -0800558}
559
Johannes Weiner56600482012-01-12 17:17:59 -0800560static inline struct mem_cgroup *
561mem_cgroup_iter(struct mem_cgroup *root,
562 struct mem_cgroup *prev,
563 struct mem_cgroup_reclaim_cookie *reclaim)
564{
565 return NULL;
566}
567
568static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
569 struct mem_cgroup *prev)
570{
571}
572
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800573static inline bool mem_cgroup_disabled(void)
574{
575 return true;
576}
KAMEZAWA Hiroyukia636b322009-01-07 18:08:08 -0800577
Yaowei Bai13308ca2015-11-05 18:47:40 -0800578static inline bool
Konstantin Khlebnikovc56d5c72012-05-29 15:07:00 -0700579mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
KOSAKI Motohiro14797e22009-01-07 18:08:18 -0800580{
Yaowei Bai13308ca2015-11-05 18:47:40 -0800581 return true;
KOSAKI Motohiro14797e22009-01-07 18:08:18 -0800582}
583
Vladimir Davydov90cbc252015-02-11 15:25:55 -0800584static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
585{
586 return true;
587}
588
KOSAKI Motohiroa3d8e052009-01-07 18:08:19 -0800589static inline unsigned long
Hugh Dickins4d7dcca2012-05-29 15:07:08 -0700590mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
KOSAKI Motohiroa3d8e052009-01-07 18:08:19 -0800591{
592 return 0;
593}
594
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700595static inline void
596mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
597 int increment)
KOSAKI Motohiro3e2f41f2009-01-07 18:08:20 -0800598{
KOSAKI Motohiro3e2f41f2009-01-07 18:08:20 -0800599}
600
Balbir Singhe2224322009-04-02 16:57:39 -0700601static inline void
602mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
603{
604}
605
Johannes Weiner6de22612015-02-11 15:25:01 -0800606static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700607{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700608 return NULL;
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700609}
610
Johannes Weiner6de22612015-02-11 15:25:01 -0800611static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700612{
613}
614
Tejun Heob23afb92015-11-05 18:46:11 -0800615static inline void mem_cgroup_handle_over_high(void)
616{
617}
618
Johannes Weiner49426422013-10-16 13:46:59 -0700619static inline void mem_cgroup_oom_enable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700620{
621}
622
Johannes Weiner49426422013-10-16 13:46:59 -0700623static inline void mem_cgroup_oom_disable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700624{
625}
626
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700627static inline bool task_in_memcg_oom(struct task_struct *p)
628{
629 return false;
630}
631
Johannes Weiner49426422013-10-16 13:46:59 -0700632static inline bool mem_cgroup_oom_synchronize(bool wait)
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700633{
634 return false;
635}
636
Johannes Weinerd7365e72014-10-29 14:50:48 -0700637static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700638 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800639{
640}
641
Johannes Weinerd7365e72014-10-29 14:50:48 -0700642static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700643 enum mem_cgroup_stat_index idx)
Balbir Singhd69b0422009-06-17 16:26:34 -0700644{
645}
646
Balbir Singh4e416952009-09-23 15:56:39 -0700647static inline
Andrew Morton0608f432013-09-24 15:27:41 -0700648unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
649 gfp_t gfp_mask,
650 unsigned long *total_scanned)
Balbir Singh4e416952009-09-23 15:56:39 -0700651{
Andrew Morton0608f432013-09-24 15:27:41 -0700652 return 0;
Balbir Singh4e416952009-09-23 15:56:39 -0700653}
654
KAMEZAWA Hiroyukie94c8a92012-01-12 17:18:20 -0800655static inline void mem_cgroup_split_huge_fixup(struct page *head)
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800656{
657}
658
Ying Han456f9982011-05-26 16:25:38 -0700659static inline
660void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
661{
662}
Andrew Mortonc255a452012-07-31 16:43:02 -0700663#endif /* CONFIG_MEMCG */
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800664
Glauber Costae1aab162011-12-11 21:47:03 +0000665enum {
666 UNDER_LIMIT,
667 SOFT_LIMIT,
668 OVER_LIMIT,
669};
670
Tejun Heo52ebea72015-05-22 17:13:37 -0400671#ifdef CONFIG_CGROUP_WRITEBACK
Tejun Heo841710a2015-05-22 18:23:33 -0400672
Tejun Heo52ebea72015-05-22 17:13:37 -0400673struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
Tejun Heo841710a2015-05-22 18:23:33 -0400674struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
Tejun Heoc5edf9c2015-09-29 13:04:26 -0400675void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
676 unsigned long *pheadroom, unsigned long *pdirty,
677 unsigned long *pwriteback);
Tejun Heo841710a2015-05-22 18:23:33 -0400678
679#else /* CONFIG_CGROUP_WRITEBACK */
680
681static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
682{
683 return NULL;
684}
685
Tejun Heoc2aa7232015-05-22 18:23:35 -0400686static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
Tejun Heoc5edf9c2015-09-29 13:04:26 -0400687 unsigned long *pfilepages,
688 unsigned long *pheadroom,
Tejun Heoc2aa7232015-05-22 18:23:35 -0400689 unsigned long *pdirty,
690 unsigned long *pwriteback)
691{
692}
693
Tejun Heo841710a2015-05-22 18:23:33 -0400694#endif /* CONFIG_CGROUP_WRITEBACK */
Tejun Heo52ebea72015-05-22 17:13:37 -0400695
Glauber Costae1aab162011-12-11 21:47:03 +0000696struct sock;
David Rientjescd590852012-10-10 15:54:08 -0700697#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
Glauber Costae1aab162011-12-11 21:47:03 +0000698void sock_update_memcg(struct sock *sk);
699void sock_release_memcg(struct sock *sk);
700#else
701static inline void sock_update_memcg(struct sock *sk)
702{
703}
704static inline void sock_release_memcg(struct sock *sk)
705{
706}
David Rientjescd590852012-10-10 15:54:08 -0700707#endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800708
709#ifdef CONFIG_MEMCG_KMEM
Glauber Costaa8964b92012-12-18 14:22:09 -0800710extern struct static_key memcg_kmem_enabled_key;
Glauber Costa749c5412012-12-18 14:23:01 -0800711
Vladimir Davydovdbcf73e2015-02-12 14:58:57 -0800712extern int memcg_nr_cache_ids;
Michal Hocko64219992015-09-08 15:01:07 -0700713void memcg_get_cache_ids(void);
714void memcg_put_cache_ids(void);
Glauber Costaebe945c2012-12-18 14:23:10 -0800715
716/*
717 * Helper macro to loop through all memcg-specific caches. Callers must still
718 * check if the cache is valid (it is either valid or NULL).
719 * the slab_mutex must be held when looping through those caches
720 */
Glauber Costa749c5412012-12-18 14:23:01 -0800721#define for_each_memcg_cache_index(_idx) \
Vladimir Davydovdbcf73e2015-02-12 14:58:57 -0800722 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
Glauber Costa749c5412012-12-18 14:23:01 -0800723
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800724static inline bool memcg_kmem_enabled(void)
725{
Glauber Costaa8964b92012-12-18 14:22:09 -0800726 return static_key_false(&memcg_kmem_enabled_key);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800727}
728
Michal Hocko33398cf2015-09-08 15:01:02 -0700729static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
730{
731 return memcg->kmem_acct_active;
732}
Vladimir Davydovcb731d62015-02-12 14:58:54 -0800733
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800734/*
735 * In general, we'll do everything in our power to not incur in any overhead
736 * for non-memcg users for the kmem functions. Not even a function call, if we
737 * can avoid it.
738 *
739 * Therefore, we'll inline all those functions so that in the best case, we'll
740 * see that kmemcg is off for everybody and proceed quickly. If it is on,
741 * we'll still do most of the flag checking inline. We check a lot of
742 * conditions, but because they are pretty simple, they are expected to be
743 * fast.
744 */
Vladimir Davydovf3ccb2c42015-11-05 18:49:01 -0800745int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
746 struct mem_cgroup *memcg);
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800747int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
748void __memcg_kmem_uncharge(struct page *page, int order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800749
Michal Hocko33398cf2015-09-08 15:01:02 -0700750/*
751 * helper for acessing a memcg's index. It will be used as an index in the
752 * child cache array in kmem_cache, and also to derive its name. This function
753 * will return -1 when this is not a kmem-limited memcg.
754 */
755static inline int memcg_cache_id(struct mem_cgroup *memcg)
756{
757 return memcg ? memcg->kmemcg_id : -1;
758}
Vladimir Davydov5722d092014-04-07 15:39:24 -0700759
Vladimir Davydov230e9fc2016-01-14 15:18:15 -0800760struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
Vladimir Davydov8135be52014-12-12 16:56:38 -0800761void __memcg_kmem_put_cache(struct kmem_cache *cachep);
Glauber Costad7f25f82012-12-18 14:22:40 -0800762
Vladimir Davydov230e9fc2016-01-14 15:18:15 -0800763static inline bool __memcg_kmem_bypass(void)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800764{
765 if (!memcg_kmem_enabled())
766 return true;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800767 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
768 return true;
Tejun Heocbfb4792015-11-05 18:46:14 -0800769 return false;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800770}
771
772/**
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800773 * memcg_kmem_charge: charge a kmem page
774 * @page: page to charge
775 * @gfp: reclaim mode
776 * @order: allocation order
Tejun Heocbfb4792015-11-05 18:46:14 -0800777 *
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800778 * Returns 0 on success, an error code on failure.
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800779 */
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800780static __always_inline int memcg_kmem_charge(struct page *page,
781 gfp_t gfp, int order)
Tejun Heocbfb4792015-11-05 18:46:14 -0800782{
Vladimir Davydov230e9fc2016-01-14 15:18:15 -0800783 if (__memcg_kmem_bypass())
784 return 0;
785 if (!(gfp & __GFP_ACCOUNT))
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800786 return 0;
787 return __memcg_kmem_charge(page, gfp, order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800788}
789
790/**
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800791 * memcg_kmem_uncharge: uncharge a kmem page
792 * @page: page to uncharge
793 * @order: allocation order
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800794 */
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800795static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800796{
797 if (memcg_kmem_enabled())
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800798 __memcg_kmem_uncharge(page, order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800799}
800
Glauber Costad7f25f82012-12-18 14:22:40 -0800801/**
802 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
803 * @cachep: the original global kmem cache
Glauber Costad7f25f82012-12-18 14:22:40 -0800804 *
Vladimir Davydov5dfb4172014-06-04 16:06:38 -0700805 * All memory allocated from a per-memcg cache is charged to the owner memcg.
Glauber Costad7f25f82012-12-18 14:22:40 -0800806 */
807static __always_inline struct kmem_cache *
808memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
809{
Vladimir Davydov230e9fc2016-01-14 15:18:15 -0800810 if (__memcg_kmem_bypass())
Glauber Costad7f25f82012-12-18 14:22:40 -0800811 return cachep;
Vladimir Davydov230e9fc2016-01-14 15:18:15 -0800812 return __memcg_kmem_get_cache(cachep, gfp);
Glauber Costad7f25f82012-12-18 14:22:40 -0800813}
Vladimir Davydov8135be52014-12-12 16:56:38 -0800814
815static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
816{
817 if (memcg_kmem_enabled())
818 __memcg_kmem_put_cache(cachep);
819}
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800820#else
Glauber Costa749c5412012-12-18 14:23:01 -0800821#define for_each_memcg_cache_index(_idx) \
822 for (; NULL; )
823
Glauber Costab9ce5ef2012-12-18 14:22:46 -0800824static inline bool memcg_kmem_enabled(void)
825{
826 return false;
827}
828
Vladimir Davydovcb731d62015-02-12 14:58:54 -0800829static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
830{
831 return false;
832}
833
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800834static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800835{
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800836 return 0;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800837}
838
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800839static inline void memcg_kmem_uncharge(struct page *page, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800840{
841}
Glauber Costa2633d7a2012-12-18 14:22:34 -0800842
843static inline int memcg_cache_id(struct mem_cgroup *memcg)
844{
845 return -1;
846}
847
Vladimir Davydov05257a12015-02-12 14:59:01 -0800848static inline void memcg_get_cache_ids(void)
849{
850}
851
852static inline void memcg_put_cache_ids(void)
853{
854}
855
Glauber Costad7f25f82012-12-18 14:22:40 -0800856static inline struct kmem_cache *
857memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
858{
859 return cachep;
860}
Vladimir Davydov8135be52014-12-12 16:56:38 -0800861
862static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
863{
864}
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800865#endif /* CONFIG_MEMCG_KMEM */
Balbir Singh8cdea7c2008-02-07 00:13:50 -0800866#endif /* _LINUX_MEMCONTROL_H */