blob: c06c70b4404ea5d66503e9fb5f5fb9662c577de1 [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
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
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;
241 /*
242 * percpu counter.
243 */
244 struct mem_cgroup_stat_cpu __percpu *stat;
Michal Hocko33398cf2015-09-08 15:01:02 -0700245
246#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
247 struct cg_proto tcp_mem;
248#endif
249#if defined(CONFIG_MEMCG_KMEM)
250 /* Index in the kmem_cache->memcg_params.memcg_caches array */
251 int kmemcg_id;
252 bool kmem_acct_activated;
253 bool kmem_acct_active;
254#endif
255
256 int last_scanned_node;
257#if MAX_NUMNODES > 1
258 nodemask_t scan_nodes;
259 atomic_t numainfo_events;
260 atomic_t numainfo_updating;
261#endif
262
263#ifdef CONFIG_CGROUP_WRITEBACK
264 struct list_head cgwb_list;
265 struct wb_domain cgwb_domain;
266#endif
267
268 /* List of events which userspace want to receive */
269 struct list_head event_list;
270 spinlock_t event_list_lock;
271
272 struct mem_cgroup_per_node *nodeinfo[0];
273 /* WARNING: nodeinfo must be the last member here */
274};
Tejun Heo56161632015-05-22 17:13:20 -0400275extern struct cgroup_subsys_state *mem_cgroup_root_css;
276
Michal Hocko33398cf2015-09-08 15:01:02 -0700277/**
278 * mem_cgroup_events - count memory events against a cgroup
279 * @memcg: the memory cgroup
280 * @idx: the event index
281 * @nr: the number of events to account for
282 */
283static inline void mem_cgroup_events(struct mem_cgroup *memcg,
Johannes Weiner241994ed2015-02-11 15:26:06 -0800284 enum mem_cgroup_events_index idx,
Michal Hocko33398cf2015-09-08 15:01:02 -0700285 unsigned int nr)
286{
287 this_cpu_add(memcg->stat->events[idx], nr);
288}
Johannes Weiner241994ed2015-02-11 15:26:06 -0800289
290bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
291
Johannes Weiner00501b52014-08-08 14:19:20 -0700292int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
293 gfp_t gfp_mask, struct mem_cgroup **memcgp);
294void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
295 bool lrucare);
296void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700297void mem_cgroup_uncharge(struct page *page);
Johannes Weiner747db952014-08-08 14:19:24 -0700298void mem_cgroup_uncharge_list(struct list_head *page_list);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700299
Hugh Dickins45637ba2015-11-05 18:49:40 -0800300void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700301
Johannes Weiner925b7672012-01-12 17:18:15 -0800302struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700303struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -0800304
Johannes Weiner2314b422014-12-10 15:44:33 -0800305bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
Michal Hocko64219992015-09-08 15:01:07 -0700306struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
Michal Hocko64219992015-09-08 15:01:07 -0700307struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
Vladimir Davydove993d902015-09-09 15:35:35 -0700308
Michal Hocko33398cf2015-09-08 15:01:02 -0700309static inline
310struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
311 return css ? container_of(css, struct mem_cgroup, css) : NULL;
312}
313
314struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
315 struct mem_cgroup *,
316 struct mem_cgroup_reclaim_cookie *);
317void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
318
319static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
320 struct mem_cgroup *root)
321{
322 if (root == memcg)
323 return true;
324 if (!root->use_hierarchy)
325 return false;
326 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
327}
Glauber Costae1aab162011-12-11 21:47:03 +0000328
Johannes Weiner2314b422014-12-10 15:44:33 -0800329static inline bool mm_match_cgroup(struct mm_struct *mm,
330 struct mem_cgroup *memcg)
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800331{
Johannes Weiner587af302012-10-08 16:34:12 -0700332 struct mem_cgroup *task_memcg;
Johannes Weiner413918b2014-12-10 15:44:30 -0800333 bool match = false;
Johannes Weinerc3ac9a82012-05-29 15:06:25 -0700334
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800335 rcu_read_lock();
Johannes Weiner587af302012-10-08 16:34:12 -0700336 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
Johannes Weiner413918b2014-12-10 15:44:30 -0800337 if (task_memcg)
Johannes Weiner2314b422014-12-10 15:44:33 -0800338 match = mem_cgroup_is_descendant(task_memcg, memcg);
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800339 rcu_read_unlock();
Johannes Weinerc3ac9a82012-05-29 15:06:25 -0700340 return match;
Lai Jiangshan2e4d4092009-01-07 18:08:07 -0800341}
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800342
Michal Hocko64219992015-09-08 15:01:07 -0700343struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
Vladimir Davydov2fc04522015-09-09 15:35:28 -0700344ino_t page_cgroup_ino(struct page *page);
Wu Fengguangd3242362009-12-16 12:19:59 +0100345
Michal Hocko33398cf2015-09-08 15:01:02 -0700346static inline bool mem_cgroup_disabled(void)
347{
348 if (memory_cgrp_subsys.disabled)
349 return true;
350 return false;
351}
Johannes Weiner56600482012-01-12 17:17:59 -0800352
KAMEZAWA Hiroyuki58ae83d2008-02-07 00:14:32 -0800353/*
354 * For memory reclaim.
355 */
Ying Han889976d2011-05-26 16:25:33 -0700356int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
Michal Hocko33398cf2015-09-08 15:01:02 -0700357
358void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
359 int nr_pages);
360
361static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
362{
363 struct mem_cgroup_per_zone *mz;
364 struct mem_cgroup *memcg;
365
366 if (mem_cgroup_disabled())
367 return true;
368
369 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
370 memcg = mz->memcg;
371
372 return !!(memcg->css.flags & CSS_ONLINE);
373}
374
375static inline
376unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
377{
378 struct mem_cgroup_per_zone *mz;
379
380 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
381 return mz->lru_size[lru];
382}
383
Yaowei Bai13308ca2015-11-05 18:47:40 -0800384static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
Michal Hocko33398cf2015-09-08 15:01:02 -0700385{
386 unsigned long inactive_ratio;
387 unsigned long inactive;
388 unsigned long active;
389 unsigned long gb;
390
391 inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
392 active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
393
394 gb = (inactive + active) >> (30 - PAGE_SHIFT);
395 if (gb)
396 inactive_ratio = int_sqrt(10 * gb);
397 else
398 inactive_ratio = 1;
399
400 return inactive * inactive_ratio < active;
401}
402
Tejun Heob23afb92015-11-05 18:46:11 -0800403void mem_cgroup_handle_over_high(void);
404
Michal Hocko64219992015-09-08 15:01:07 -0700405void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
406 struct task_struct *p);
KAMEZAWA Hiroyuki58ae83d2008-02-07 00:14:32 -0800407
Johannes Weiner49426422013-10-16 13:46:59 -0700408static inline void mem_cgroup_oom_enable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700409{
Tejun Heo626ebc42015-11-05 18:46:09 -0800410 WARN_ON(current->memcg_may_oom);
411 current->memcg_may_oom = 1;
Johannes Weiner519e5242013-09-12 15:13:42 -0700412}
413
Johannes Weiner49426422013-10-16 13:46:59 -0700414static inline void mem_cgroup_oom_disable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700415{
Tejun Heo626ebc42015-11-05 18:46:09 -0800416 WARN_ON(!current->memcg_may_oom);
417 current->memcg_may_oom = 0;
Johannes Weiner519e5242013-09-12 15:13:42 -0700418}
419
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700420static inline bool task_in_memcg_oom(struct task_struct *p)
421{
Tejun Heo626ebc42015-11-05 18:46:09 -0800422 return p->memcg_in_oom;
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700423}
424
Johannes Weiner49426422013-10-16 13:46:59 -0700425bool mem_cgroup_oom_synchronize(bool wait);
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700426
Andrew Mortonc255a452012-07-31 16:43:02 -0700427#ifdef CONFIG_MEMCG_SWAP
KAMEZAWA Hiroyukic0777192009-01-07 18:07:57 -0800428extern int do_swap_account;
429#endif
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800430
Johannes Weiner6de22612015-02-11 15:25:01 -0800431struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
Johannes Weiner6de22612015-02-11 15:25:01 -0800432void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700433
Michal Hocko33398cf2015-09-08 15:01:02 -0700434/**
435 * mem_cgroup_update_page_stat - update page state statistics
436 * @memcg: memcg to account against
437 * @idx: page state item to account
438 * @val: number of pages (positive or negative)
439 *
440 * See mem_cgroup_begin_page_stat() for locking requirements.
441 */
442static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
443 enum mem_cgroup_stat_index idx, int val)
444{
445 VM_BUG_ON(!rcu_read_lock_held());
446
447 if (memcg)
448 this_cpu_add(memcg->stat->count[idx], val);
449}
450
Johannes Weinerd7365e72014-10-29 14:50:48 -0700451static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700452 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800453{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700454 mem_cgroup_update_page_stat(memcg, idx, 1);
Greg Thelen2a7106f2011-01-13 15:47:37 -0800455}
456
Johannes Weinerd7365e72014-10-29 14:50:48 -0700457static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700458 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800459{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700460 mem_cgroup_update_page_stat(memcg, idx, -1);
Greg Thelen2a7106f2011-01-13 15:47:37 -0800461}
462
Andrew Morton0608f432013-09-24 15:27:41 -0700463unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
464 gfp_t gfp_mask,
465 unsigned long *total_scanned);
David Rientjesa63d83f2010-08-09 17:19:46 -0700466
David Rientjes68ae5642012-12-12 13:51:57 -0800467static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
468 enum vm_event_item idx)
469{
Michal Hocko33398cf2015-09-08 15:01:02 -0700470 struct mem_cgroup *memcg;
471
David Rientjes68ae5642012-12-12 13:51:57 -0800472 if (mem_cgroup_disabled())
473 return;
Michal Hocko33398cf2015-09-08 15:01:02 -0700474
475 rcu_read_lock();
476 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
477 if (unlikely(!memcg))
478 goto out;
479
480 switch (idx) {
481 case PGFAULT:
482 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
483 break;
484 case PGMAJFAULT:
485 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
486 break;
487 default:
488 BUG();
489 }
490out:
491 rcu_read_unlock();
David Rientjes68ae5642012-12-12 13:51:57 -0800492}
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800493#ifdef CONFIG_TRANSPARENT_HUGEPAGE
KAMEZAWA Hiroyukie94c8a92012-01-12 17:18:20 -0800494void mem_cgroup_split_huge_fixup(struct page *head);
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800495#endif
496
Andrew Mortonc255a452012-07-31 16:43:02 -0700497#else /* CONFIG_MEMCG */
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800498struct mem_cgroup;
499
Johannes Weiner241994ed2015-02-11 15:26:06 -0800500static inline void mem_cgroup_events(struct mem_cgroup *memcg,
501 enum mem_cgroup_events_index idx,
502 unsigned int nr)
503{
504}
505
506static inline bool mem_cgroup_low(struct mem_cgroup *root,
507 struct mem_cgroup *memcg)
508{
509 return false;
510}
511
Johannes Weiner00501b52014-08-08 14:19:20 -0700512static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
513 gfp_t gfp_mask,
514 struct mem_cgroup **memcgp)
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800515{
Johannes Weiner00501b52014-08-08 14:19:20 -0700516 *memcgp = NULL;
Balbir Singh8a9f3cc2008-02-07 00:13:53 -0800517 return 0;
518}
519
Johannes Weiner00501b52014-08-08 14:19:20 -0700520static inline void mem_cgroup_commit_charge(struct page *page,
521 struct mem_cgroup *memcg,
522 bool lrucare)
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800523{
524}
525
Johannes Weiner00501b52014-08-08 14:19:20 -0700526static inline void mem_cgroup_cancel_charge(struct page *page,
527 struct mem_cgroup *memcg)
KAMEZAWA Hiroyuki7a81b882009-01-07 18:07:48 -0800528{
529}
530
Johannes Weiner0a31bc92014-08-08 14:19:22 -0700531static inline void mem_cgroup_uncharge(struct page *page)
532{
533}
534
Johannes Weiner747db952014-08-08 14:19:24 -0700535static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -0800536{
537}
538
Hugh Dickins45637ba2015-11-05 18:49:40 -0800539static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
KAMEZAWA Hiroyuki69029cd2008-07-25 01:47:14 -0700540{
541}
542
Johannes Weiner925b7672012-01-12 17:18:15 -0800543static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
544 struct mem_cgroup *memcg)
545{
546 return &zone->lruvec;
547}
548
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700549static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
550 struct zone *zone)
Minchan Kim3f58a822011-03-22 16:32:53 -0700551{
Johannes Weiner925b7672012-01-12 17:18:15 -0800552 return &zone->lruvec;
Balbir Singh66e17072008-02-07 00:13:56 -0800553}
554
Johannes Weiner587af302012-10-08 16:34:12 -0700555static inline bool mm_match_cgroup(struct mm_struct *mm,
Raghavendra K Tc0ff4b82011-11-02 13:38:15 -0700556 struct mem_cgroup *memcg)
Balbir Singhbed71612008-02-07 00:14:01 -0800557{
Johannes Weiner587af302012-10-08 16:34:12 -0700558 return true;
Balbir Singhbed71612008-02-07 00:14:01 -0800559}
560
David Rientjesffbdccf2013-07-03 15:01:23 -0700561static inline bool task_in_mem_cgroup(struct task_struct *task,
562 const struct mem_cgroup *memcg)
David Rientjes4c4a2212008-02-07 00:14:06 -0800563{
David Rientjesffbdccf2013-07-03 15:01:23 -0700564 return true;
David Rientjes4c4a2212008-02-07 00:14:06 -0800565}
566
Johannes Weiner56600482012-01-12 17:17:59 -0800567static inline struct mem_cgroup *
568mem_cgroup_iter(struct mem_cgroup *root,
569 struct mem_cgroup *prev,
570 struct mem_cgroup_reclaim_cookie *reclaim)
571{
572 return NULL;
573}
574
575static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
576 struct mem_cgroup *prev)
577{
578}
579
Hirokazu Takahashif8d665422009-01-07 18:08:02 -0800580static inline bool mem_cgroup_disabled(void)
581{
582 return true;
583}
KAMEZAWA Hiroyukia636b322009-01-07 18:08:08 -0800584
Yaowei Bai13308ca2015-11-05 18:47:40 -0800585static inline bool
Konstantin Khlebnikovc56d5c72012-05-29 15:07:00 -0700586mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
KOSAKI Motohiro14797e22009-01-07 18:08:18 -0800587{
Yaowei Bai13308ca2015-11-05 18:47:40 -0800588 return true;
KOSAKI Motohiro14797e22009-01-07 18:08:18 -0800589}
590
Vladimir Davydov90cbc252015-02-11 15:25:55 -0800591static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
592{
593 return true;
594}
595
KOSAKI Motohiroa3d8e052009-01-07 18:08:19 -0800596static inline unsigned long
Hugh Dickins4d7dcca2012-05-29 15:07:08 -0700597mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
KOSAKI Motohiroa3d8e052009-01-07 18:08:19 -0800598{
599 return 0;
600}
601
Hugh Dickinsfa9add62012-05-29 15:07:09 -0700602static inline void
603mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
604 int increment)
KOSAKI Motohiro3e2f41f2009-01-07 18:08:20 -0800605{
KOSAKI Motohiro3e2f41f2009-01-07 18:08:20 -0800606}
607
Balbir Singhe2224322009-04-02 16:57:39 -0700608static inline void
609mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
610{
611}
612
Johannes Weiner6de22612015-02-11 15:25:01 -0800613static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700614{
Johannes Weinerd7365e72014-10-29 14:50:48 -0700615 return NULL;
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700616}
617
Johannes Weiner6de22612015-02-11 15:25:01 -0800618static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
KAMEZAWA Hiroyuki89c06bd2012-03-21 16:34:25 -0700619{
620}
621
Tejun Heob23afb92015-11-05 18:46:11 -0800622static inline void mem_cgroup_handle_over_high(void)
623{
624}
625
Johannes Weiner49426422013-10-16 13:46:59 -0700626static inline void mem_cgroup_oom_enable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700627{
628}
629
Johannes Weiner49426422013-10-16 13:46:59 -0700630static inline void mem_cgroup_oom_disable(void)
Johannes Weiner519e5242013-09-12 15:13:42 -0700631{
632}
633
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700634static inline bool task_in_memcg_oom(struct task_struct *p)
635{
636 return false;
637}
638
Johannes Weiner49426422013-10-16 13:46:59 -0700639static inline bool mem_cgroup_oom_synchronize(bool wait)
Johannes Weiner3812c8c2013-09-12 15:13:44 -0700640{
641 return false;
642}
643
Johannes Weinerd7365e72014-10-29 14:50:48 -0700644static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700645 enum mem_cgroup_stat_index idx)
Greg Thelen2a7106f2011-01-13 15:47:37 -0800646{
647}
648
Johannes Weinerd7365e72014-10-29 14:50:48 -0700649static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
Sha Zhengju68b48762013-09-12 15:13:50 -0700650 enum mem_cgroup_stat_index idx)
Balbir Singhd69b0422009-06-17 16:26:34 -0700651{
652}
653
Balbir Singh4e416952009-09-23 15:56:39 -0700654static inline
Andrew Morton0608f432013-09-24 15:27:41 -0700655unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
656 gfp_t gfp_mask,
657 unsigned long *total_scanned)
Balbir Singh4e416952009-09-23 15:56:39 -0700658{
Andrew Morton0608f432013-09-24 15:27:41 -0700659 return 0;
Balbir Singh4e416952009-09-23 15:56:39 -0700660}
661
KAMEZAWA Hiroyukie94c8a92012-01-12 17:18:20 -0800662static inline void mem_cgroup_split_huge_fixup(struct page *head)
KAMEZAWA Hiroyukica3e0212011-01-20 14:44:24 -0800663{
664}
665
Ying Han456f9982011-05-26 16:25:38 -0700666static inline
667void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
668{
669}
Andrew Mortonc255a452012-07-31 16:43:02 -0700670#endif /* CONFIG_MEMCG */
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800671
Glauber Costae1aab162011-12-11 21:47:03 +0000672enum {
673 UNDER_LIMIT,
674 SOFT_LIMIT,
675 OVER_LIMIT,
676};
677
Tejun Heo52ebea72015-05-22 17:13:37 -0400678#ifdef CONFIG_CGROUP_WRITEBACK
Tejun Heo841710a2015-05-22 18:23:33 -0400679
Tejun Heo52ebea72015-05-22 17:13:37 -0400680struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
Tejun Heo841710a2015-05-22 18:23:33 -0400681struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
Tejun Heoc5edf9c2015-09-29 13:04:26 -0400682void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
683 unsigned long *pheadroom, unsigned long *pdirty,
684 unsigned long *pwriteback);
Tejun Heo841710a2015-05-22 18:23:33 -0400685
686#else /* CONFIG_CGROUP_WRITEBACK */
687
688static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
689{
690 return NULL;
691}
692
Tejun Heoc2aa7232015-05-22 18:23:35 -0400693static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
Tejun Heoc5edf9c2015-09-29 13:04:26 -0400694 unsigned long *pfilepages,
695 unsigned long *pheadroom,
Tejun Heoc2aa7232015-05-22 18:23:35 -0400696 unsigned long *pdirty,
697 unsigned long *pwriteback)
698{
699}
700
Tejun Heo841710a2015-05-22 18:23:33 -0400701#endif /* CONFIG_CGROUP_WRITEBACK */
Tejun Heo52ebea72015-05-22 17:13:37 -0400702
Glauber Costae1aab162011-12-11 21:47:03 +0000703struct sock;
David Rientjescd590852012-10-10 15:54:08 -0700704#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
Glauber Costae1aab162011-12-11 21:47:03 +0000705void sock_update_memcg(struct sock *sk);
706void sock_release_memcg(struct sock *sk);
707#else
708static inline void sock_update_memcg(struct sock *sk)
709{
710}
711static inline void sock_release_memcg(struct sock *sk)
712{
713}
David Rientjescd590852012-10-10 15:54:08 -0700714#endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800715
716#ifdef CONFIG_MEMCG_KMEM
Glauber Costaa8964b92012-12-18 14:22:09 -0800717extern struct static_key memcg_kmem_enabled_key;
Glauber Costa749c5412012-12-18 14:23:01 -0800718
Vladimir Davydovdbcf73e2015-02-12 14:58:57 -0800719extern int memcg_nr_cache_ids;
Michal Hocko64219992015-09-08 15:01:07 -0700720void memcg_get_cache_ids(void);
721void memcg_put_cache_ids(void);
Glauber Costaebe945c2012-12-18 14:23:10 -0800722
723/*
724 * Helper macro to loop through all memcg-specific caches. Callers must still
725 * check if the cache is valid (it is either valid or NULL).
726 * the slab_mutex must be held when looping through those caches
727 */
Glauber Costa749c5412012-12-18 14:23:01 -0800728#define for_each_memcg_cache_index(_idx) \
Vladimir Davydovdbcf73e2015-02-12 14:58:57 -0800729 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
Glauber Costa749c5412012-12-18 14:23:01 -0800730
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800731static inline bool memcg_kmem_enabled(void)
732{
Glauber Costaa8964b92012-12-18 14:22:09 -0800733 return static_key_false(&memcg_kmem_enabled_key);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800734}
735
Michal Hocko33398cf2015-09-08 15:01:02 -0700736static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
737{
738 return memcg->kmem_acct_active;
739}
Vladimir Davydovcb731d62015-02-12 14:58:54 -0800740
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800741/*
742 * In general, we'll do everything in our power to not incur in any overhead
743 * for non-memcg users for the kmem functions. Not even a function call, if we
744 * can avoid it.
745 *
746 * Therefore, we'll inline all those functions so that in the best case, we'll
747 * see that kmemcg is off for everybody and proceed quickly. If it is on,
748 * we'll still do most of the flag checking inline. We check a lot of
749 * conditions, but because they are pretty simple, they are expected to be
750 * fast.
751 */
Vladimir Davydovf3ccb2c42015-11-05 18:49:01 -0800752int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
753 struct mem_cgroup *memcg);
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800754int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
755void __memcg_kmem_uncharge(struct page *page, int order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800756
Michal Hocko33398cf2015-09-08 15:01:02 -0700757/*
758 * helper for acessing a memcg's index. It will be used as an index in the
759 * child cache array in kmem_cache, and also to derive its name. This function
760 * will return -1 when this is not a kmem-limited memcg.
761 */
762static inline int memcg_cache_id(struct mem_cgroup *memcg)
763{
764 return memcg ? memcg->kmemcg_id : -1;
765}
Vladimir Davydov5722d092014-04-07 15:39:24 -0700766
Vladimir Davydov8135be52014-12-12 16:56:38 -0800767struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep);
768void __memcg_kmem_put_cache(struct kmem_cache *cachep);
Glauber Costad7f25f82012-12-18 14:22:40 -0800769
Tejun Heocbfb4792015-11-05 18:46:14 -0800770static inline bool __memcg_kmem_bypass(gfp_t gfp)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800771{
772 if (!memcg_kmem_enabled())
773 return true;
Vladimir Davydov8f4fc072015-05-14 15:16:55 -0700774 if (gfp & __GFP_NOACCOUNT)
775 return true;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800776 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
777 return true;
Tejun Heocbfb4792015-11-05 18:46:14 -0800778 return false;
779}
780
781/**
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800782 * memcg_kmem_charge: charge a kmem page
783 * @page: page to charge
784 * @gfp: reclaim mode
785 * @order: allocation order
Tejun Heocbfb4792015-11-05 18:46:14 -0800786 *
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800787 * Returns 0 on success, an error code on failure.
Tejun Heocbfb4792015-11-05 18:46:14 -0800788 */
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800789static __always_inline int memcg_kmem_charge(struct page *page,
790 gfp_t gfp, int order)
Tejun Heocbfb4792015-11-05 18:46:14 -0800791{
792 if (__memcg_kmem_bypass(gfp))
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800793 return 0;
794 return __memcg_kmem_charge(page, gfp, order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800795}
796
797/**
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800798 * memcg_kmem_uncharge: uncharge a kmem page
799 * @page: page to uncharge
800 * @order: allocation order
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800801 */
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800802static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800803{
804 if (memcg_kmem_enabled())
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800805 __memcg_kmem_uncharge(page, order);
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800806}
807
Glauber Costad7f25f82012-12-18 14:22:40 -0800808/**
809 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
810 * @cachep: the original global kmem cache
811 * @gfp: allocation flags.
812 *
Vladimir Davydov5dfb4172014-06-04 16:06:38 -0700813 * All memory allocated from a per-memcg cache is charged to the owner memcg.
Glauber Costad7f25f82012-12-18 14:22:40 -0800814 */
815static __always_inline struct kmem_cache *
816memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
817{
Tejun Heocbfb4792015-11-05 18:46:14 -0800818 if (__memcg_kmem_bypass(gfp))
Glauber Costad7f25f82012-12-18 14:22:40 -0800819 return cachep;
Zhang Zhen056b7cc2014-12-12 16:55:38 -0800820 return __memcg_kmem_get_cache(cachep);
Glauber Costad7f25f82012-12-18 14:22:40 -0800821}
Vladimir Davydov8135be52014-12-12 16:56:38 -0800822
823static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
824{
825 if (memcg_kmem_enabled())
826 __memcg_kmem_put_cache(cachep);
827}
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800828#else
Glauber Costa749c5412012-12-18 14:23:01 -0800829#define for_each_memcg_cache_index(_idx) \
830 for (; NULL; )
831
Glauber Costab9ce5ef2012-12-18 14:22:46 -0800832static inline bool memcg_kmem_enabled(void)
833{
834 return false;
835}
836
Vladimir Davydovcb731d62015-02-12 14:58:54 -0800837static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
838{
839 return false;
840}
841
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800842static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800843{
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800844 return 0;
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800845}
846
Vladimir Davydovd05e83a2015-11-05 18:48:59 -0800847static inline void memcg_kmem_uncharge(struct page *page, int order)
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800848{
849}
Glauber Costa2633d7a2012-12-18 14:22:34 -0800850
851static inline int memcg_cache_id(struct mem_cgroup *memcg)
852{
853 return -1;
854}
855
Vladimir Davydov05257a12015-02-12 14:59:01 -0800856static inline void memcg_get_cache_ids(void)
857{
858}
859
860static inline void memcg_put_cache_ids(void)
861{
862}
863
Glauber Costad7f25f82012-12-18 14:22:40 -0800864static inline struct kmem_cache *
865memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
866{
867 return cachep;
868}
Vladimir Davydov8135be52014-12-12 16:56:38 -0800869
870static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
871{
872}
Glauber Costa7ae1e1d2012-12-18 14:21:56 -0800873#endif /* CONFIG_MEMCG_KMEM */
Balbir Singh8cdea7c2008-02-07 00:13:50 -0800874#endif /* _LINUX_MEMCONTROL_H */