blob: cd0e2413c358d4cf58072ebbdc2cfea2684fd245 [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 Weiner241994e2015-02-11 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
88/*
89 * Bits in struct cg_proto.flags
90 */
91enum cg_proto_flags {
92 /* Currently active and new sockets should be assigned to cgroups */
93 MEMCG_SOCK_ACTIVE,
94 /* It was ever activated; we must disarm static keys on destruction */
95 MEMCG_SOCK_ACTIVATED,
96};
97
98struct cg_proto {
99 struct page_counter memory_allocated; /* Current allocated memory. */
100 struct percpu_counter sockets_allocated; /* Current number of sockets. */
101 int memory_pressure;
102 long sysctl_mem[3];
103 unsigned long flags;
104 /*
105 * memcg field is used to find which memcg we belong directly
106 * Each memcg struct can hold more than one cg_proto, so container_of
107 * won't really cut.
108 *
109 * The elegant solution would be having an inverse function to
110 * proto_cgroup in struct proto, but that means polluting the structure
111 * for everybody, instead of just for memcg users.
112 */
113 struct mem_cgroup *memcg;
114};
115
Andrew Mortonc255a452012-07-31 16:43:02 -0700116#ifdef CONFIG_MEMCG
Michal Hocko33398cf2015-09-08 15:01:02 -0700117struct mem_cgroup_stat_cpu {
118 long count[MEM_CGROUP_STAT_NSTATS];
119 unsigned long events[MEMCG_NR_EVENTS];
120 unsigned long nr_page_events;
121 unsigned long targets[MEM_CGROUP_NTARGETS];
122};
123
124struct mem_cgroup_reclaim_iter {
125 struct mem_cgroup *position;
126 /* scan generation, increased every round-trip */
127 unsigned int generation;
128};
129
130/*
131 * per-zone information in memory controller.
132 */
133struct mem_cgroup_per_zone {
134 struct lruvec lruvec;
135 unsigned long lru_size[NR_LRU_LISTS];
136
137 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
138
139 struct rb_node tree_node; /* RB tree node */
140 unsigned long usage_in_excess;/* Set to the value by which */
141 /* the soft limit is exceeded*/
142 bool on_tree;
143 struct mem_cgroup *memcg; /* Back pointer, we cannot */
144 /* use container_of */
145};
146
147struct mem_cgroup_per_node {
148 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
149};
150
151struct mem_cgroup_threshold {
152 struct eventfd_ctx *eventfd;
153 unsigned long threshold;
154};
155
156/* For threshold */
157struct mem_cgroup_threshold_ary {
158 /* An array index points to threshold just below or equal to usage. */
159 int current_threshold;
160 /* Size of entries[] */
161 unsigned int size;
162 /* Array of thresholds */
163 struct mem_cgroup_threshold entries[0];
164};
165
166struct mem_cgroup_thresholds {
167 /* Primary thresholds array */
168 struct mem_cgroup_threshold_ary *primary;
169 /*
170 * Spare threshold array.
171 * This is needed to make mem_cgroup_unregister_event() "never fail".
172 * It must be able to store at least primary->size - 1 entries.
173 */
174 struct mem_cgroup_threshold_ary *spare;
175};
176
177/*
178 * The memory controller data structure. The memory controller controls both
179 * page cache and RSS per cgroup. We would eventually like to provide
180 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
181 * to help the administrator determine what knobs to tune.
182 */
183struct mem_cgroup {
184 struct cgroup_subsys_state css;
185
186 /* Accounted resources */
187 struct page_counter memory;
188 struct page_counter memsw;
189 struct page_counter kmem;
190
191 /* Normal memory consumption range */
192 unsigned long low;
193 unsigned long high;
194
195 unsigned long soft_limit;
196
197 /* vmpressure notifications */
198 struct vmpressure vmpressure;
199
200 /* css_online() has been completed */
201 int initialized;
202
203 /*
204 * Should the accounting and control be hierarchical, per subtree?
205 */
206 bool use_hierarchy;
207
208 /* protected by memcg_oom_lock */
209 bool oom_lock;
210 int under_oom;
211
212 int swappiness;
213 /* OOM-Killer disable */
214 int oom_kill_disable;
215
Tejun Heo472912a2015-09-18 18:01:59 -0400216 /* handle for "memory.events" */
217 struct cgroup_file events_file;
218
Michal Hocko33398cf2015-09-08 15:01:02 -0700219 /* protect arrays of thresholds */
220 struct mutex thresholds_lock;
221
222 /* thresholds for memory usage. RCU-protected */
223 struct mem_cgroup_thresholds thresholds;
224
225 /* thresholds for mem+swap usage. RCU-protected */
226 struct mem_cgroup_thresholds memsw_thresholds;
227
228 /* For oom notifier event fd */
229 struct list_head oom_notify;
230
231 /*
232 * Should we move charges of a task when a task is moved into this
233 * mem_cgroup ? And what type of charges should we move ?
234 */
235 unsigned long move_charge_at_immigrate;
236 /*
237 * set > 0 if pages under this cgroup are moving to other cgroup.
238 */
239 atomic_t moving_account;
240 /* taken only while moving_account > 0 */
241 spinlock_t move_lock;
242 struct task_struct *move_lock_task;
243 unsigned long move_lock_flags;
244 /*
245 * percpu counter.
246 */
247 struct mem_cgroup_stat_cpu __percpu *stat;
Michal Hocko33398cf2015-09-08 15:01:02 -0700248
249#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
250 struct cg_proto tcp_mem;
251#endif
252#if defined(CONFIG_MEMCG_KMEM)
253 /* Index in the kmem_cache->memcg_params.memcg_caches array */
254 int kmemcg_id;
255 bool kmem_acct_activated;
256 bool kmem_acct_active;
257#endif
258
259 int last_scanned_node;
260#if MAX_NUMNODES > 1
261 nodemask_t scan_nodes;
262 atomic_t numainfo_events;
263 atomic_t numainfo_updating;
264#endif
265
266#ifdef CONFIG_CGROUP_WRITEBACK
267 struct list_head cgwb_list;
268 struct wb_domain cgwb_domain;
269#endif
270
271 /* List of events which userspace want to receive */
272 struct list_head event_list;
273 spinlock_t event_list_lock;
274
275 struct mem_cgroup_per_node *nodeinfo[0];
276 /* WARNING: nodeinfo must be the last member here */
277};
Tejun Heo56161632015-05-22 17:13:20 -0400278extern struct cgroup_subsys_state *mem_cgroup_root_css;
279
Michal Hocko33398cf2015-09-08 15:01:02 -0700280/**
281 * mem_cgroup_events - count memory events against a cgroup
282 * @memcg: the memory cgroup
283 * @idx: the event index
284 * @nr: the number of events to account for
285 */
286static inline void mem_cgroup_events(struct mem_cgroup *memcg,
Johannes Weiner241994e2015-02-11 15:26:06 -0800287 enum mem_cgroup_events_index idx,
Michal Hocko33398cf2015-09-08 15:01:02 -0700288 unsigned int nr)
289{
290 this_cpu_add(memcg->stat->events[idx], nr);
Tejun Heo472912a2015-09-18 18:01:59 -0400291 cgroup_file_notify(&memcg->events_file);
Michal Hocko33398cf2015-09-08 15:01:02 -0700292}
Johannes Weiner241994e2015-02-11 15:26:06 -0800293
294bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
295
Johannes Weiner00501b52014-08-08 14:19:20 -0700296int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
297 gfp_t gfp_mask, struct mem_cgroup **memcgp);
298void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
299 bool lrucare);
300void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700301void mem_cgroup_uncharge(struct page *page);
Johannes Weiner747db952014-08-08 14:19:24 -0700302void mem_cgroup_uncharge_list(struct list_head *page_list);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700303
Hugh Dickins45637ba2015-11-05 18:49:40 -0800304void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700305
Johannes Weiner925b7672012-01-12 17:18:15 -0800306struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700307struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -0800308
Johannes Weiner2314b422014-12-10 15:44:33 -0800309bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
Michal Hocko64219992015-09-08 15:01:07 -0700310struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
Michal Hocko64219992015-09-08 15:01:07 -0700311struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
Vladimir Davydove993d902015-09-09 15:35:35 -0700312
Michal Hocko33398cf2015-09-08 15:01:02 -0700313static inline
314struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
315 return css ? container_of(css, struct mem_cgroup, css) : NULL;
316}
317
318struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
319 struct mem_cgroup *,
320 struct mem_cgroup_reclaim_cookie *);
321void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
322
323static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
324 struct mem_cgroup *root)
325{
326 if (root == memcg)
327 return true;
328 if (!root->use_hierarchy)
329 return false;
330 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
331}
Glauber Costae1aab162011-12-11 21:47:03 +0000332
Johannes Weiner2314b422014-12-10 15:44:33 -0800333static inline bool mm_match_cgroup(struct mm_struct *mm,
334 struct mem_cgroup *memcg)
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800335{
Johannes Weiner587af302012-10-08 16:34:12 -0700336 struct mem_cgroup *task_memcg;
Johannes Weiner413918b2014-12-10 15:44:30 -0800337 bool match = false;
Johannes Weinerc3ac9a82012-05-29 15:06:25 -0700338
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800339 rcu_read_lock();
Johannes Weiner587af302012-10-08 16:34:12 -0700340 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
Johannes Weiner413918b2014-12-10 15:44:30 -0800341 if (task_memcg)
Johannes Weiner2314b422014-12-10 15:44:33 -0800342 match = mem_cgroup_is_descendant(task_memcg, memcg);
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800343 rcu_read_unlock();
Johannes Weinerc3ac9a82012-05-29 15:06:25 -0700344 return match;
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800345}
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800346
Michal Hocko64219992015-09-08 15:01:07 -0700347struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
Vladimir Davydov2fc04522015-09-09 15:35:28 -0700348ino_t page_cgroup_ino(struct page *page);
Wu Fengguangd3242362009-12-16 12:19:59 +0100349
Michal Hocko33398cf2015-09-08 15:01:02 -0700350static inline bool mem_cgroup_disabled(void)
351{
Tejun Heofc5ed1e2015-09-18 11:56:28 -0400352 return !cgroup_subsys_enabled(memory_cgrp_subsys);
Michal Hocko33398cf2015-09-08 15:01:02 -0700353}
Johannes Weiner56600482012-01-12 17:17:59 -0800354
KAMEZAWA Hiroyuki58ae83d2008-02-07 00:14:32 -0800355/*
356 * For memory reclaim.
357 */
Ying Han889976d2011-05-26 16:25:33 -0700358int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
Michal Hocko33398cf2015-09-08 15:01:02 -0700359
360void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
361 int nr_pages);
362
363static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
364{
365 struct mem_cgroup_per_zone *mz;
366 struct mem_cgroup *memcg;
367
368 if (mem_cgroup_disabled())
369 return true;
370
371 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
372 memcg = mz->memcg;
373
374 return !!(memcg->css.flags & CSS_ONLINE);
375}
376
377static inline
378unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
379{
380 struct mem_cgroup_per_zone *mz;
381
382 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
383 return mz->lru_size[lru];
384}
385
Yaowei Bai13308ca2015-11-05 18:47:40 -0800386static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
Michal Hocko33398cf2015-09-08 15:01:02 -0700387{
388 unsigned long inactive_ratio;
389 unsigned long inactive;
390 unsigned long active;
391 unsigned long gb;
392
393 inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
394 active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
395
396 gb = (inactive + active) >> (30 - PAGE_SHIFT);
397 if (gb)
398 inactive_ratio = int_sqrt(10 * gb);
399 else
400 inactive_ratio = 1;
401
402 return inactive * inactive_ratio < active;
403}
404
Tejun Heob23afb92015-11-05 18:46:11 -0800405void mem_cgroup_handle_over_high(void);
406
Michal Hocko64219992015-09-08 15:01:07 -0700407void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
408 struct task_struct *p);
KAMEZAWA Hiroyuki58ae83d2008-02-07 00:14:32 -0800409
Johannes Weiner49426422013-10-16 13:46:59 -0700410static inline void mem_cgroup_oom_enable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700411{
Tejun Heo626ebc42015-11-05 18:46:09 -0800412 WARN_ON(current->memcg_may_oom);
413 current->memcg_may_oom = 1;
Johannes Weiner519e5242013-09-12 15:13:42 -0700414}
415
Johannes Weiner49426422013-10-16 13:46:59 -0700416static inline void mem_cgroup_oom_disable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700417{
Tejun Heo626ebc42015-11-05 18:46:09 -0800418 WARN_ON(!current->memcg_may_oom);
419 current->memcg_may_oom = 0;
Johannes Weiner519e5242013-09-12 15:13:42 -0700420}
421
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700422static inline bool task_in_memcg_oom(struct task_struct *p)
423{
Tejun Heo626ebc42015-11-05 18:46:09 -0800424 return p->memcg_in_oom;
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700425}
426
Johannes Weiner49426422013-10-16 13:46:59 -0700427bool mem_cgroup_oom_synchronize(bool wait);
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700428
Andrew Mortonc255a452012-07-31 16:43:02 -0700429#ifdef CONFIG_MEMCG_SWAP
KAMEZAWA Hiroyukic0777192009-01-07 18:07:57 -0800430extern int do_swap_account;
431#endif
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800432
Johannes Weiner6de22612015-02-11 15:25:01 -0800433struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
Johannes Weiner6de22612015-02-11 15:25:01 -0800434void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700435
Michal Hocko33398cf2015-09-08 15:01:02 -0700436/**
437 * mem_cgroup_update_page_stat - update page state statistics
438 * @memcg: memcg to account against
439 * @idx: page state item to account
440 * @val: number of pages (positive or negative)
441 *
442 * See mem_cgroup_begin_page_stat() for locking requirements.
443 */
444static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
445 enum mem_cgroup_stat_index idx, int val)
446{
447 VM_BUG_ON(!rcu_read_lock_held());
448
449 if (memcg)
450 this_cpu_add(memcg->stat->count[idx], val);
451}
452
Johannes Weinerd7365e72014-10-29 14:50:48 -0700453static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700454 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800455{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700456 mem_cgroup_update_page_stat(memcg, idx, 1);
Greg Thelen2a7106f2011-01-13 15:47:37 -0800457}
458
Johannes Weinerd7365e72014-10-29 14:50:48 -0700459static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700460 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800461{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700462 mem_cgroup_update_page_stat(memcg, idx, -1);
Greg Thelen2a7106f2011-01-13 15:47:37 -0800463}
464
Andrew Morton0608f432013-09-24 15:27:41 -0700465unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
466 gfp_t gfp_mask,
467 unsigned long *total_scanned);
David Rientjesa63d83f2010-08-09 17:19:46 -0700468
David Rientjes68ae5642012-12-12 13:51:57 -0800469static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
470 enum vm_event_item idx)
471{
Michal Hocko33398cf2015-09-08 15:01:02 -0700472 struct mem_cgroup *memcg;
473
David Rientjes68ae5642012-12-12 13:51:57 -0800474 if (mem_cgroup_disabled())
475 return;
Michal Hocko33398cf2015-09-08 15:01:02 -0700476
477 rcu_read_lock();
478 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
479 if (unlikely(!memcg))
480 goto out;
481
482 switch (idx) {
483 case PGFAULT:
484 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
485 break;
486 case PGMAJFAULT:
487 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
488 break;
489 default:
490 BUG();
491 }
492out:
493 rcu_read_unlock();
David Rientjes68ae5642012-12-12 13:51:57 -0800494}
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800495#ifdef CONFIG_TRANSPARENT_HUGEPAGE
KAMEZAWA Hiroyukie94c8a92012-01-12 17:18:20 -0800496void mem_cgroup_split_huge_fixup(struct page *head);
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800497#endif
498
Andrew Mortonc255a452012-07-31 16:43:02 -0700499#else /* CONFIG_MEMCG */
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800500struct mem_cgroup;
501
Johannes Weiner241994e2015-02-11 15:26:06 -0800502static inline void mem_cgroup_events(struct mem_cgroup *memcg,
503 enum mem_cgroup_events_index idx,
504 unsigned int nr)
505{
506}
507
508static inline bool mem_cgroup_low(struct mem_cgroup *root,
509 struct mem_cgroup *memcg)
510{
511 return false;
512}
513
Johannes Weiner00501b52014-08-08 14:19:20 -0700514static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
515 gfp_t gfp_mask,
516 struct mem_cgroup **memcgp)
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800517{
Johannes Weiner00501b52014-08-08 14:19:20 -0700518 *memcgp = NULL;
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800519 return 0;
520}
521
Johannes Weiner00501b52014-08-08 14:19:20 -0700522static inline void mem_cgroup_commit_charge(struct page *page,
523 struct mem_cgroup *memcg,
524 bool lrucare)
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800525{
526}
527
Johannes Weiner00501b52014-08-08 14:19:20 -0700528static inline void mem_cgroup_cancel_charge(struct page *page,
529 struct mem_cgroup *memcg)
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800530{
531}
532
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700533static inline void mem_cgroup_uncharge(struct page *page)
534{
535}
536
Johannes Weiner747db952014-08-08 14:19:24 -0700537static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -0800538{
539}
540
Hugh Dickins45637ba2015-11-05 18:49:40 -0800541static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
KAMEZAWA Hiroyuki69029cd2008-07-25 01:47:14 -0700542{
543}
544
Johannes Weiner925b7672012-01-12 17:18:15 -0800545static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
546 struct mem_cgroup *memcg)
547{
548 return &zone->lruvec;
549}
550
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700551static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
552 struct zone *zone)
Minchan Kim3f58a822011-03-22 16:32:53 -0700553{
Johannes Weiner925b7672012-01-12 17:18:15 -0800554 return &zone->lruvec;
Balbir Singh66e17072008-02-07 00:13:56 -0800555}
556
Johannes Weiner587af302012-10-08 16:34:12 -0700557static inline bool mm_match_cgroup(struct mm_struct *mm,
Raghavendra K Tc0ff4b82011-11-02 13:38:15 -0700558 struct mem_cgroup *memcg)
Balbir Singhbed71612008-02-07 00:14:01 -0800559{
Johannes Weiner587af302012-10-08 16:34:12 -0700560 return true;
Balbir Singhbed71612008-02-07 00:14:01 -0800561}
562
David Rientjesffbdccf2013-07-03 15:01:23 -0700563static inline bool task_in_mem_cgroup(struct task_struct *task,
564 const struct mem_cgroup *memcg)
David Rientjes4c4a2212008-02-07 00:14:06 -0800565{
David Rientjesffbdccf2013-07-03 15:01:23 -0700566 return true;
David Rientjes4c4a2212008-02-07 00:14:06 -0800567}
568
Johannes Weiner56600482012-01-12 17:17:59 -0800569static inline struct mem_cgroup *
570mem_cgroup_iter(struct mem_cgroup *root,
571 struct mem_cgroup *prev,
572 struct mem_cgroup_reclaim_cookie *reclaim)
573{
574 return NULL;
575}
576
577static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
578 struct mem_cgroup *prev)
579{
580}
581
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800582static inline bool mem_cgroup_disabled(void)
583{
584 return true;
585}
KAMEZAWA Hiroyukia636b322009-01-07 18:08:08 -0800586
Yaowei Bai13308ca2015-11-05 18:47:40 -0800587static inline bool
Konstantin Khlebnikovc56d5c72012-05-29 15:07:00 -0700588mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
KOSAKI Motohiro14797e22009-01-07 18:08:18 -0800589{
Yaowei Bai13308ca2015-11-05 18:47:40 -0800590 return true;
KOSAKI Motohiro14797e22009-01-07 18:08:18 -0800591}
592
Vladimir Davydov90cbc252015-02-11 15:25:55 -0800593static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
594{
595 return true;
596}
597
KOSAKI Motohiroa3d8e052009-01-07 18:08:19 -0800598static inline unsigned long
Hugh Dickins4d7dcca2012-05-29 15:07:08 -0700599mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
KOSAKI Motohiroa3d8e052009-01-07 18:08:19 -0800600{
601 return 0;
602}
603
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700604static inline void
605mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
606 int increment)
KOSAKI Motohiro3e2f41f2009-01-07 18:08:20 -0800607{
KOSAKI Motohiro3e2f41f2009-01-07 18:08:20 -0800608}
609
Balbir Singhe2224322009-04-02 16:57:39 -0700610static inline void
611mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
612{
613}
614
Johannes Weiner6de22612015-02-11 15:25:01 -0800615static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700616{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700617 return NULL;
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700618}
619
Johannes Weiner6de22612015-02-11 15:25:01 -0800620static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700621{
622}
623
Tejun Heob23afb92015-11-05 18:46:11 -0800624static inline void mem_cgroup_handle_over_high(void)
625{
626}
627
Johannes Weiner49426422013-10-16 13:46:59 -0700628static inline void mem_cgroup_oom_enable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700629{
630}
631
Johannes Weiner49426422013-10-16 13:46:59 -0700632static inline void mem_cgroup_oom_disable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700633{
634}
635
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700636static inline bool task_in_memcg_oom(struct task_struct *p)
637{
638 return false;
639}
640
Johannes Weiner49426422013-10-16 13:46:59 -0700641static inline bool mem_cgroup_oom_synchronize(bool wait)
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700642{
643 return false;
644}
645
Johannes Weinerd7365e72014-10-29 14:50:48 -0700646static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700647 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800648{
649}
650
Johannes Weinerd7365e72014-10-29 14:50:48 -0700651static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700652 enum mem_cgroup_stat_index idx)
Balbir Singhd69b0422009-06-17 16:26:34 -0700653{
654}
655
Balbir Singh4e416952009-09-23 15:56:39 -0700656static inline
Andrew Morton0608f432013-09-24 15:27:41 -0700657unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
658 gfp_t gfp_mask,
659 unsigned long *total_scanned)
Balbir Singh4e416952009-09-23 15:56:39 -0700660{
Andrew Morton0608f432013-09-24 15:27:41 -0700661 return 0;
Balbir Singh4e416952009-09-23 15:56:39 -0700662}
663
KAMEZAWA Hiroyukie94c8a92012-01-12 17:18:20 -0800664static inline void mem_cgroup_split_huge_fixup(struct page *head)
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800665{
666}
667
Ying Han456f9982011-05-26 16:25:38 -0700668static inline
669void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
670{
671}
Andrew Mortonc255a452012-07-31 16:43:02 -0700672#endif /* CONFIG_MEMCG */
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800673
Glauber Costae1aab162011-12-11 21:47:03 +0000674enum {
675 UNDER_LIMIT,
676 SOFT_LIMIT,
677 OVER_LIMIT,
678};
679
Tejun Heo52ebea72015-05-22 17:13:37 -0400680#ifdef CONFIG_CGROUP_WRITEBACK
Tejun Heo841710a2015-05-22 18:23:33 -0400681
Tejun Heo52ebea72015-05-22 17:13:37 -0400682struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
Tejun Heo841710a2015-05-22 18:23:33 -0400683struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
Tejun Heoc5edf9c2015-09-29 13:04:26 -0400684void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
685 unsigned long *pheadroom, unsigned long *pdirty,
686 unsigned long *pwriteback);
Tejun Heo841710a2015-05-22 18:23:33 -0400687
688#else /* CONFIG_CGROUP_WRITEBACK */
689
690static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
691{
692 return NULL;
693}
694
Tejun Heoc2aa7232015-05-22 18:23:35 -0400695static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
Tejun Heoc5edf9c2015-09-29 13:04:26 -0400696 unsigned long *pfilepages,
697 unsigned long *pheadroom,
Tejun Heoc2aa7232015-05-22 18:23:35 -0400698 unsigned long *pdirty,
699 unsigned long *pwriteback)
700{
701}
702
Tejun Heo841710a2015-05-22 18:23:33 -0400703#endif /* CONFIG_CGROUP_WRITEBACK */
Tejun Heo52ebea72015-05-22 17:13:37 -0400704
Glauber Costae1aab162011-12-11 21:47:03 +0000705struct sock;
David Rientjescd590852012-10-10 15:54:08 -0700706#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
Glauber Costae1aab162011-12-11 21:47:03 +0000707void sock_update_memcg(struct sock *sk);
708void sock_release_memcg(struct sock *sk);
709#else
710static inline void sock_update_memcg(struct sock *sk)
711{
712}
713static inline void sock_release_memcg(struct sock *sk)
714{
715}
David Rientjescd590852012-10-10 15:54:08 -0700716#endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800717
718#ifdef CONFIG_MEMCG_KMEM
Glauber Costaa8964b92012-12-18 14:22:09 -0800719extern struct static_key memcg_kmem_enabled_key;
Glauber Costa749c5412012-12-18 14:23:01 -0800720
Vladimir Davydovdbcf73e2015-02-12 14:58:57 -0800721extern int memcg_nr_cache_ids;
Michal Hocko64219992015-09-08 15:01:07 -0700722void memcg_get_cache_ids(void);
723void memcg_put_cache_ids(void);
Glauber Costaebe945c2012-12-18 14:23:10 -0800724
725/*
726 * Helper macro to loop through all memcg-specific caches. Callers must still
727 * check if the cache is valid (it is either valid or NULL).
728 * the slab_mutex must be held when looping through those caches
729 */
Glauber Costa749c5412012-12-18 14:23:01 -0800730#define for_each_memcg_cache_index(_idx) \
Vladimir Davydovdbcf73e2015-02-12 14:58:57 -0800731 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
Glauber Costa749c5412012-12-18 14:23:01 -0800732
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800733static inline bool memcg_kmem_enabled(void)
734{
Glauber Costaa8964b92012-12-18 14:22:09 -0800735 return static_key_false(&memcg_kmem_enabled_key);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800736}
737
Michal Hocko33398cf2015-09-08 15:01:02 -0700738static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
739{
740 return memcg->kmem_acct_active;
741}
Vladimir Davydovcb731d62015-02-12 14:58:54 -0800742
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800743/*
744 * In general, we'll do everything in our power to not incur in any overhead
745 * for non-memcg users for the kmem functions. Not even a function call, if we
746 * can avoid it.
747 *
748 * Therefore, we'll inline all those functions so that in the best case, we'll
749 * see that kmemcg is off for everybody and proceed quickly. If it is on,
750 * we'll still do most of the flag checking inline. We check a lot of
751 * conditions, but because they are pretty simple, they are expected to be
752 * fast.
753 */
Vladimir Davydovf3ccb2c42015-11-05 18:49:01 -0800754int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
755 struct mem_cgroup *memcg);
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800756int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
757void __memcg_kmem_uncharge(struct page *page, int order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800758
Michal Hocko33398cf2015-09-08 15:01:02 -0700759/*
760 * helper for acessing a memcg's index. It will be used as an index in the
761 * child cache array in kmem_cache, and also to derive its name. This function
762 * will return -1 when this is not a kmem-limited memcg.
763 */
764static inline int memcg_cache_id(struct mem_cgroup *memcg)
765{
766 return memcg ? memcg->kmemcg_id : -1;
767}
Vladimir Davydov5722d092014-04-07 15:39:24 -0700768
Vladimir Davydov8135be52014-12-12 16:56:38 -0800769struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep);
770void __memcg_kmem_put_cache(struct kmem_cache *cachep);
Glauber Costad7f25f82012-12-18 14:22:40 -0800771
Tejun Heocbfb4792015-11-05 18:46:14 -0800772static inline bool __memcg_kmem_bypass(gfp_t gfp)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800773{
774 if (!memcg_kmem_enabled())
775 return true;
Vladimir Davydov8f4fc072015-05-14 15:16:55 -0700776 if (gfp & __GFP_NOACCOUNT)
777 return true;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800778 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
779 return true;
Tejun Heocbfb4792015-11-05 18:46:14 -0800780 return false;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800781}
782
783/**
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800784 * memcg_kmem_charge: charge a kmem page
785 * @page: page to charge
786 * @gfp: reclaim mode
787 * @order: allocation order
Tejun Heocbfb4792015-11-05 18:46:14 -0800788 *
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800789 * Returns 0 on success, an error code on failure.
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800790 */
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800791static __always_inline int memcg_kmem_charge(struct page *page,
792 gfp_t gfp, int order)
Tejun Heocbfb4792015-11-05 18:46:14 -0800793{
794 if (__memcg_kmem_bypass(gfp))
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800795 return 0;
796 return __memcg_kmem_charge(page, gfp, order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800797}
798
799/**
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800800 * memcg_kmem_uncharge: uncharge a kmem page
801 * @page: page to uncharge
802 * @order: allocation order
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800803 */
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800804static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800805{
806 if (memcg_kmem_enabled())
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800807 __memcg_kmem_uncharge(page, order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800808}
809
Glauber Costad7f25f82012-12-18 14:22:40 -0800810/**
811 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
812 * @cachep: the original global kmem cache
813 * @gfp: allocation flags.
814 *
Vladimir Davydov5dfb4172014-06-04 16:06:38 -0700815 * All memory allocated from a per-memcg cache is charged to the owner memcg.
Glauber Costad7f25f82012-12-18 14:22:40 -0800816 */
817static __always_inline struct kmem_cache *
818memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
819{
Tejun Heocbfb4792015-11-05 18:46:14 -0800820 if (__memcg_kmem_bypass(gfp))
Glauber Costad7f25f82012-12-18 14:22:40 -0800821 return cachep;
Zhang Zhen056b7cc2014-12-12 16:55:38 -0800822 return __memcg_kmem_get_cache(cachep);
Glauber Costad7f25f82012-12-18 14:22:40 -0800823}
Vladimir Davydov8135be52014-12-12 16:56:38 -0800824
825static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
826{
827 if (memcg_kmem_enabled())
828 __memcg_kmem_put_cache(cachep);
829}
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800830#else
Glauber Costa749c5412012-12-18 14:23:01 -0800831#define for_each_memcg_cache_index(_idx) \
832 for (; NULL; )
833
Glauber Costab9ce5ef2012-12-18 14:22:46 -0800834static inline bool memcg_kmem_enabled(void)
835{
836 return false;
837}
838
Vladimir Davydovcb731d62015-02-12 14:58:54 -0800839static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
840{
841 return false;
842}
843
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800844static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800845{
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800846 return 0;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800847}
848
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800849static inline void memcg_kmem_uncharge(struct page *page, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800850{
851}
Glauber Costa2633d7a2012-12-18 14:22:34 -0800852
853static inline int memcg_cache_id(struct mem_cgroup *memcg)
854{
855 return -1;
856}
857
Vladimir Davydov05257a12015-02-12 14:59:01 -0800858static inline void memcg_get_cache_ids(void)
859{
860}
861
862static inline void memcg_put_cache_ids(void)
863{
864}
865
Glauber Costad7f25f82012-12-18 14:22:40 -0800866static inline struct kmem_cache *
867memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
868{
869 return cachep;
870}
Vladimir Davydov8135be52014-12-12 16:56:38 -0800871
872static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
873{
874}
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800875#endif /* CONFIG_MEMCG_KMEM */
Balbir Singh8cdea7c2008-02-07 00:13:50 -0800876#endif /* _LINUX_MEMCONTROL_H */