]>
Commit | Line | Data |
---|---|---|
8cdea7c0 BS |
1 | /* memcontrol.h - Memory Controller |
2 | * | |
3 | * Copyright IBM Corporation, 2007 | |
4 | * Author Balbir Singh <balbir@linux.vnet.ibm.com> | |
5 | * | |
78fb7466 PE |
6 | * Copyright 2007 OpenVZ SWsoft Inc |
7 | * Author: Pavel Emelianov <xemul@openvz.org> | |
8 | * | |
8cdea7c0 BS |
9 | * This program is free software; you can redistribute it and/or modify |
10 | * it under the terms of the GNU General Public License as published by | |
11 | * the Free Software Foundation; either version 2 of the License, or | |
12 | * (at your option) any later version. | |
13 | * | |
14 | * This program is distributed in the hope that it will be useful, | |
15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | * GNU General Public License for more details. | |
18 | */ | |
19 | ||
20 | #ifndef _LINUX_MEMCONTROL_H | |
21 | #define _LINUX_MEMCONTROL_H | |
f8d66542 | 22 | #include <linux/cgroup.h> |
456f998e | 23 | #include <linux/vm_event_item.h> |
7ae1e1d0 | 24 | #include <linux/hardirq.h> |
a8964b9b | 25 | #include <linux/jump_label.h> |
33398cf2 MH |
26 | #include <linux/page_counter.h> |
27 | #include <linux/vmpressure.h> | |
28 | #include <linux/eventfd.h> | |
29 | #include <linux/mmzone.h> | |
30 | #include <linux/writeback.h> | |
fdf1cdb9 | 31 | #include <linux/page-flags.h> |
456f998e | 32 | |
78fb7466 | 33 | struct mem_cgroup; |
8697d331 BS |
34 | struct page; |
35 | struct mm_struct; | |
2633d7a0 | 36 | struct kmem_cache; |
78fb7466 | 37 | |
68b4876d SZ |
38 | /* |
39 | * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c, | |
40 | * These two lists should keep in accord with each other. | |
41 | */ | |
42 | enum mem_cgroup_stat_index { | |
43 | /* | |
44 | * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss. | |
45 | */ | |
46 | MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */ | |
47 | MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */ | |
48 | MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */ | |
9a4caf1e | 49 | MEM_CGROUP_STAT_SHMEM, /* # of pages charged as shmem */ |
68b4876d | 50 | MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */ |
c4843a75 | 51 | MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */ |
3ea67d06 | 52 | MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */ |
68b4876d SZ |
53 | MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */ |
54 | MEM_CGROUP_STAT_NSTATS, | |
b2807f07 | 55 | /* default hierarchy stats */ |
efdc9490 | 56 | MEMCG_KERNEL_STACK_KB = MEM_CGROUP_STAT_NSTATS, |
27ee57c9 VD |
57 | MEMCG_SLAB_RECLAIMABLE, |
58 | MEMCG_SLAB_UNRECLAIMABLE, | |
12580e4b | 59 | MEMCG_SOCK, |
2a2e4885 JW |
60 | MEMCG_WORKINGSET_REFAULT, |
61 | MEMCG_WORKINGSET_ACTIVATE, | |
62 | MEMCG_WORKINGSET_NODERECLAIM, | |
b2807f07 | 63 | MEMCG_NR_STAT, |
2a7106f2 GT |
64 | }; |
65 | ||
5660048c | 66 | struct mem_cgroup_reclaim_cookie { |
ef8f2327 | 67 | pg_data_t *pgdat; |
5660048c JW |
68 | int priority; |
69 | unsigned int generation; | |
70 | }; | |
71 | ||
241994ed JW |
72 | enum mem_cgroup_events_index { |
73 | MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */ | |
74 | MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */ | |
75 | MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */ | |
76 | MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */ | |
77 | MEM_CGROUP_EVENTS_NSTATS, | |
78 | /* default hierarchy events */ | |
79 | MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS, | |
80 | MEMCG_HIGH, | |
81 | MEMCG_MAX, | |
82 | MEMCG_OOM, | |
83 | MEMCG_NR_EVENTS, | |
84 | }; | |
85 | ||
33398cf2 MH |
86 | /* |
87 | * Per memcg event counter is incremented at every pagein/pageout. With THP, | |
88 | * it will be incremated by the number of pages. This counter is used for | |
89 | * for trigger some periodic events. This is straightforward and better | |
90 | * than using jiffies etc. to handle periodic memcg event. | |
91 | */ | |
92 | enum mem_cgroup_events_target { | |
93 | MEM_CGROUP_TARGET_THRESH, | |
94 | MEM_CGROUP_TARGET_SOFTLIMIT, | |
95 | MEM_CGROUP_TARGET_NUMAINFO, | |
96 | MEM_CGROUP_NTARGETS, | |
97 | }; | |
98 | ||
c255a458 | 99 | #ifdef CONFIG_MEMCG |
23047a96 JW |
100 | |
101 | #define MEM_CGROUP_ID_SHIFT 16 | |
102 | #define MEM_CGROUP_ID_MAX USHRT_MAX | |
103 | ||
73f576c0 JW |
104 | struct mem_cgroup_id { |
105 | int id; | |
106 | atomic_t ref; | |
107 | }; | |
108 | ||
33398cf2 | 109 | struct mem_cgroup_stat_cpu { |
b2807f07 | 110 | long count[MEMCG_NR_STAT]; |
33398cf2 MH |
111 | unsigned long events[MEMCG_NR_EVENTS]; |
112 | unsigned long nr_page_events; | |
113 | unsigned long targets[MEM_CGROUP_NTARGETS]; | |
114 | }; | |
115 | ||
116 | struct mem_cgroup_reclaim_iter { | |
117 | struct mem_cgroup *position; | |
118 | /* scan generation, increased every round-trip */ | |
119 | unsigned int generation; | |
120 | }; | |
121 | ||
122 | /* | |
123 | * per-zone information in memory controller. | |
124 | */ | |
ef8f2327 | 125 | struct mem_cgroup_per_node { |
33398cf2 | 126 | struct lruvec lruvec; |
b4536f0c | 127 | unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; |
33398cf2 MH |
128 | |
129 | struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1]; | |
130 | ||
131 | struct rb_node tree_node; /* RB tree node */ | |
132 | unsigned long usage_in_excess;/* Set to the value by which */ | |
133 | /* the soft limit is exceeded*/ | |
134 | bool on_tree; | |
135 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ | |
136 | /* use container_of */ | |
137 | }; | |
138 | ||
33398cf2 MH |
139 | struct mem_cgroup_threshold { |
140 | struct eventfd_ctx *eventfd; | |
141 | unsigned long threshold; | |
142 | }; | |
143 | ||
144 | /* For threshold */ | |
145 | struct mem_cgroup_threshold_ary { | |
146 | /* An array index points to threshold just below or equal to usage. */ | |
147 | int current_threshold; | |
148 | /* Size of entries[] */ | |
149 | unsigned int size; | |
150 | /* Array of thresholds */ | |
151 | struct mem_cgroup_threshold entries[0]; | |
152 | }; | |
153 | ||
154 | struct mem_cgroup_thresholds { | |
155 | /* Primary thresholds array */ | |
156 | struct mem_cgroup_threshold_ary *primary; | |
157 | /* | |
158 | * Spare threshold array. | |
159 | * This is needed to make mem_cgroup_unregister_event() "never fail". | |
160 | * It must be able to store at least primary->size - 1 entries. | |
161 | */ | |
162 | struct mem_cgroup_threshold_ary *spare; | |
163 | }; | |
164 | ||
567e9ab2 JW |
165 | enum memcg_kmem_state { |
166 | KMEM_NONE, | |
167 | KMEM_ALLOCATED, | |
168 | KMEM_ONLINE, | |
169 | }; | |
170 | ||
33398cf2 MH |
171 | /* |
172 | * The memory controller data structure. The memory controller controls both | |
173 | * page cache and RSS per cgroup. We would eventually like to provide | |
174 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, | |
175 | * to help the administrator determine what knobs to tune. | |
176 | */ | |
177 | struct mem_cgroup { | |
178 | struct cgroup_subsys_state css; | |
179 | ||
73f576c0 JW |
180 | /* Private memcg ID. Used to ID objects that outlive the cgroup */ |
181 | struct mem_cgroup_id id; | |
182 | ||
33398cf2 MH |
183 | /* Accounted resources */ |
184 | struct page_counter memory; | |
37e84351 | 185 | struct page_counter swap; |
0db15298 JW |
186 | |
187 | /* Legacy consumer-oriented counters */ | |
33398cf2 MH |
188 | struct page_counter memsw; |
189 | struct page_counter kmem; | |
0db15298 | 190 | struct page_counter tcpmem; |
33398cf2 MH |
191 | |
192 | /* Normal memory consumption range */ | |
193 | unsigned long low; | |
194 | unsigned long high; | |
195 | ||
f7e1cb6e JW |
196 | /* Range enforcement for interrupt charges */ |
197 | struct work_struct high_work; | |
198 | ||
33398cf2 MH |
199 | unsigned long soft_limit; |
200 | ||
201 | /* vmpressure notifications */ | |
202 | struct vmpressure vmpressure; | |
203 | ||
33398cf2 MH |
204 | /* |
205 | * Should the accounting and control be hierarchical, per subtree? | |
206 | */ | |
207 | bool use_hierarchy; | |
208 | ||
209 | /* protected by memcg_oom_lock */ | |
210 | bool oom_lock; | |
211 | int under_oom; | |
212 | ||
213 | int swappiness; | |
214 | /* OOM-Killer disable */ | |
215 | int oom_kill_disable; | |
216 | ||
472912a2 TH |
217 | /* handle for "memory.events" */ |
218 | struct cgroup_file events_file; | |
219 | ||
33398cf2 MH |
220 | /* protect arrays of thresholds */ |
221 | struct mutex thresholds_lock; | |
222 | ||
223 | /* thresholds for memory usage. RCU-protected */ | |
224 | struct mem_cgroup_thresholds thresholds; | |
225 | ||
226 | /* thresholds for mem+swap usage. RCU-protected */ | |
227 | struct mem_cgroup_thresholds memsw_thresholds; | |
228 | ||
229 | /* For oom notifier event fd */ | |
230 | struct list_head oom_notify; | |
231 | ||
232 | /* | |
233 | * Should we move charges of a task when a task is moved into this | |
234 | * mem_cgroup ? And what type of charges should we move ? | |
235 | */ | |
236 | unsigned long move_charge_at_immigrate; | |
237 | /* | |
238 | * set > 0 if pages under this cgroup are moving to other cgroup. | |
239 | */ | |
240 | atomic_t moving_account; | |
241 | /* taken only while moving_account > 0 */ | |
242 | spinlock_t move_lock; | |
243 | struct task_struct *move_lock_task; | |
244 | unsigned long move_lock_flags; | |
245 | /* | |
246 | * percpu counter. | |
247 | */ | |
248 | struct mem_cgroup_stat_cpu __percpu *stat; | |
33398cf2 | 249 | |
d886f4e4 JW |
250 | unsigned long socket_pressure; |
251 | ||
252 | /* Legacy tcp memory accounting */ | |
0db15298 JW |
253 | bool tcpmem_active; |
254 | int tcpmem_pressure; | |
d886f4e4 | 255 | |
127424c8 | 256 | #ifndef CONFIG_SLOB |
33398cf2 MH |
257 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ |
258 | int kmemcg_id; | |
567e9ab2 | 259 | enum memcg_kmem_state kmem_state; |
bc2791f8 | 260 | struct list_head kmem_caches; |
33398cf2 MH |
261 | #endif |
262 | ||
263 | int last_scanned_node; | |
264 | #if MAX_NUMNODES > 1 | |
265 | nodemask_t scan_nodes; | |
266 | atomic_t numainfo_events; | |
267 | atomic_t numainfo_updating; | |
268 | #endif | |
269 | ||
270 | #ifdef CONFIG_CGROUP_WRITEBACK | |
271 | struct list_head cgwb_list; | |
272 | struct wb_domain cgwb_domain; | |
273 | #endif | |
274 | ||
275 | /* List of events which userspace want to receive */ | |
276 | struct list_head event_list; | |
277 | spinlock_t event_list_lock; | |
278 | ||
279 | struct mem_cgroup_per_node *nodeinfo[0]; | |
280 | /* WARNING: nodeinfo must be the last member here */ | |
281 | }; | |
7d828602 JW |
282 | |
283 | extern struct mem_cgroup *root_mem_cgroup; | |
56161634 | 284 | |
23047a96 JW |
285 | static inline bool mem_cgroup_disabled(void) |
286 | { | |
287 | return !cgroup_subsys_enabled(memory_cgrp_subsys); | |
288 | } | |
289 | ||
31176c78 JW |
290 | static inline void mem_cgroup_event(struct mem_cgroup *memcg, |
291 | enum mem_cgroup_events_index idx) | |
33398cf2 | 292 | { |
31176c78 | 293 | this_cpu_inc(memcg->stat->events[idx]); |
472912a2 | 294 | cgroup_file_notify(&memcg->events_file); |
33398cf2 | 295 | } |
241994ed JW |
296 | |
297 | bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg); | |
298 | ||
00501b53 | 299 | int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
f627c2f5 KS |
300 | gfp_t gfp_mask, struct mem_cgroup **memcgp, |
301 | bool compound); | |
00501b53 | 302 | void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg, |
f627c2f5 KS |
303 | bool lrucare, bool compound); |
304 | void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg, | |
305 | bool compound); | |
0a31bc97 | 306 | void mem_cgroup_uncharge(struct page *page); |
747db954 | 307 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
569b846d | 308 | |
6a93ca8f | 309 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage); |
569b846d | 310 | |
ef8f2327 MG |
311 | static struct mem_cgroup_per_node * |
312 | mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid) | |
55779ec7 | 313 | { |
ef8f2327 | 314 | return memcg->nodeinfo[nid]; |
55779ec7 JW |
315 | } |
316 | ||
317 | /** | |
a9dd0a83 MG |
318 | * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone |
319 | * @node: node of the wanted lruvec | |
55779ec7 JW |
320 | * @memcg: memcg of the wanted lruvec |
321 | * | |
a9dd0a83 MG |
322 | * Returns the lru list vector holding pages for a given @node or a given |
323 | * @memcg and @zone. This can be the node lruvec, if the memory controller | |
55779ec7 JW |
324 | * is disabled. |
325 | */ | |
a9dd0a83 | 326 | static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat, |
ef8f2327 | 327 | struct mem_cgroup *memcg) |
55779ec7 | 328 | { |
ef8f2327 | 329 | struct mem_cgroup_per_node *mz; |
55779ec7 JW |
330 | struct lruvec *lruvec; |
331 | ||
332 | if (mem_cgroup_disabled()) { | |
a9dd0a83 | 333 | lruvec = node_lruvec(pgdat); |
55779ec7 JW |
334 | goto out; |
335 | } | |
336 | ||
ef8f2327 | 337 | mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id); |
55779ec7 JW |
338 | lruvec = &mz->lruvec; |
339 | out: | |
340 | /* | |
341 | * Since a node can be onlined after the mem_cgroup was created, | |
599d0c95 | 342 | * we have to be prepared to initialize lruvec->pgdat here; |
55779ec7 JW |
343 | * and if offlined then reonlined, we need to reinitialize it. |
344 | */ | |
ef8f2327 MG |
345 | if (unlikely(lruvec->pgdat != pgdat)) |
346 | lruvec->pgdat = pgdat; | |
55779ec7 JW |
347 | return lruvec; |
348 | } | |
349 | ||
599d0c95 | 350 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *); |
c9b0ed51 | 351 | |
2314b42d | 352 | bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg); |
64219994 | 353 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
e993d905 | 354 | |
33398cf2 MH |
355 | static inline |
356 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ | |
357 | return css ? container_of(css, struct mem_cgroup, css) : NULL; | |
358 | } | |
359 | ||
8e8ae645 JW |
360 | #define mem_cgroup_from_counter(counter, member) \ |
361 | container_of(counter, struct mem_cgroup, member) | |
362 | ||
33398cf2 MH |
363 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
364 | struct mem_cgroup *, | |
365 | struct mem_cgroup_reclaim_cookie *); | |
366 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); | |
7c5f64f8 VD |
367 | int mem_cgroup_scan_tasks(struct mem_cgroup *, |
368 | int (*)(struct task_struct *, void *), void *); | |
33398cf2 | 369 | |
23047a96 JW |
370 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
371 | { | |
372 | if (mem_cgroup_disabled()) | |
373 | return 0; | |
374 | ||
73f576c0 | 375 | return memcg->id.id; |
23047a96 | 376 | } |
73f576c0 | 377 | struct mem_cgroup *mem_cgroup_from_id(unsigned short id); |
23047a96 | 378 | |
8e8ae645 JW |
379 | /** |
380 | * parent_mem_cgroup - find the accounting parent of a memcg | |
381 | * @memcg: memcg whose parent to find | |
382 | * | |
383 | * Returns the parent memcg, or NULL if this is the root or the memory | |
384 | * controller is in legacy no-hierarchy mode. | |
385 | */ | |
386 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) | |
387 | { | |
388 | if (!memcg->memory.parent) | |
389 | return NULL; | |
390 | return mem_cgroup_from_counter(memcg->memory.parent, memory); | |
391 | } | |
392 | ||
33398cf2 MH |
393 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
394 | struct mem_cgroup *root) | |
395 | { | |
396 | if (root == memcg) | |
397 | return true; | |
398 | if (!root->use_hierarchy) | |
399 | return false; | |
400 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); | |
401 | } | |
e1aab161 | 402 | |
2314b42d JW |
403 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
404 | struct mem_cgroup *memcg) | |
2e4d4091 | 405 | { |
587af308 | 406 | struct mem_cgroup *task_memcg; |
413918bb | 407 | bool match = false; |
c3ac9a8a | 408 | |
2e4d4091 | 409 | rcu_read_lock(); |
587af308 | 410 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
413918bb | 411 | if (task_memcg) |
2314b42d | 412 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
2e4d4091 | 413 | rcu_read_unlock(); |
c3ac9a8a | 414 | return match; |
2e4d4091 | 415 | } |
8a9f3ccd | 416 | |
64219994 | 417 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
2fc04524 | 418 | ino_t page_cgroup_ino(struct page *page); |
d324236b | 419 | |
eb01aaab VD |
420 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
421 | { | |
422 | if (mem_cgroup_disabled()) | |
423 | return true; | |
424 | return !!(memcg->css.flags & CSS_ONLINE); | |
425 | } | |
426 | ||
58ae83db KH |
427 | /* |
428 | * For memory reclaim. | |
429 | */ | |
889976db | 430 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
33398cf2 MH |
431 | |
432 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
b4536f0c | 433 | int zid, int nr_pages); |
33398cf2 | 434 | |
0a6b76dd VD |
435 | unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg, |
436 | int nid, unsigned int lru_mask); | |
437 | ||
33398cf2 MH |
438 | static inline |
439 | unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) | |
440 | { | |
ef8f2327 | 441 | struct mem_cgroup_per_node *mz; |
b4536f0c MH |
442 | unsigned long nr_pages = 0; |
443 | int zid; | |
33398cf2 | 444 | |
ef8f2327 | 445 | mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); |
b4536f0c MH |
446 | for (zid = 0; zid < MAX_NR_ZONES; zid++) |
447 | nr_pages += mz->lru_zone_size[zid][lru]; | |
448 | return nr_pages; | |
449 | } | |
450 | ||
451 | static inline | |
452 | unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, | |
453 | enum lru_list lru, int zone_idx) | |
454 | { | |
455 | struct mem_cgroup_per_node *mz; | |
456 | ||
457 | mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); | |
458 | return mz->lru_zone_size[zone_idx][lru]; | |
33398cf2 MH |
459 | } |
460 | ||
b23afb93 TH |
461 | void mem_cgroup_handle_over_high(void); |
462 | ||
7c5f64f8 VD |
463 | unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg); |
464 | ||
64219994 MH |
465 | void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, |
466 | struct task_struct *p); | |
58ae83db | 467 | |
49426420 | 468 | static inline void mem_cgroup_oom_enable(void) |
519e5247 | 469 | { |
626ebc41 TH |
470 | WARN_ON(current->memcg_may_oom); |
471 | current->memcg_may_oom = 1; | |
519e5247 JW |
472 | } |
473 | ||
49426420 | 474 | static inline void mem_cgroup_oom_disable(void) |
519e5247 | 475 | { |
626ebc41 TH |
476 | WARN_ON(!current->memcg_may_oom); |
477 | current->memcg_may_oom = 0; | |
519e5247 JW |
478 | } |
479 | ||
3812c8c8 JW |
480 | static inline bool task_in_memcg_oom(struct task_struct *p) |
481 | { | |
626ebc41 | 482 | return p->memcg_in_oom; |
3812c8c8 JW |
483 | } |
484 | ||
49426420 | 485 | bool mem_cgroup_oom_synchronize(bool wait); |
3812c8c8 | 486 | |
c255a458 | 487 | #ifdef CONFIG_MEMCG_SWAP |
c077719b KH |
488 | extern int do_swap_account; |
489 | #endif | |
f8d66542 | 490 | |
62cccb8c JW |
491 | void lock_page_memcg(struct page *page); |
492 | void unlock_page_memcg(struct page *page); | |
d7365e78 | 493 | |
2a2e4885 JW |
494 | static inline unsigned long mem_cgroup_read_stat(struct mem_cgroup *memcg, |
495 | enum mem_cgroup_stat_index idx) | |
496 | { | |
497 | long val = 0; | |
498 | int cpu; | |
499 | ||
500 | for_each_possible_cpu(cpu) | |
501 | val += per_cpu(memcg->stat->count[idx], cpu); | |
502 | ||
503 | if (val < 0) | |
504 | val = 0; | |
505 | ||
506 | return val; | |
507 | } | |
508 | ||
509 | static inline void mem_cgroup_update_stat(struct mem_cgroup *memcg, | |
510 | enum mem_cgroup_stat_index idx, int val) | |
511 | { | |
512 | if (!mem_cgroup_disabled()) | |
513 | this_cpu_add(memcg->stat->count[idx], val); | |
514 | } | |
515 | ||
516 | static inline void mem_cgroup_inc_stat(struct mem_cgroup *memcg, | |
517 | enum mem_cgroup_stat_index idx) | |
518 | { | |
519 | mem_cgroup_update_stat(memcg, idx, 1); | |
520 | } | |
521 | ||
522 | static inline void mem_cgroup_dec_stat(struct mem_cgroup *memcg, | |
523 | enum mem_cgroup_stat_index idx) | |
524 | { | |
525 | mem_cgroup_update_stat(memcg, idx, -1); | |
526 | } | |
527 | ||
33398cf2 MH |
528 | /** |
529 | * mem_cgroup_update_page_stat - update page state statistics | |
62cccb8c | 530 | * @page: the page |
33398cf2 MH |
531 | * @idx: page state item to account |
532 | * @val: number of pages (positive or negative) | |
533 | * | |
fdf1cdb9 JW |
534 | * The @page must be locked or the caller must use lock_page_memcg() |
535 | * to prevent double accounting when the page is concurrently being | |
536 | * moved to another memcg: | |
81f8c3a4 | 537 | * |
fdf1cdb9 | 538 | * lock_page(page) or lock_page_memcg(page) |
81f8c3a4 | 539 | * if (TestClearPageState(page)) |
62cccb8c | 540 | * mem_cgroup_update_page_stat(page, state, -1); |
fdf1cdb9 | 541 | * unlock_page(page) or unlock_page_memcg(page) |
2a2e4885 JW |
542 | * |
543 | * Kernel pages are an exception to this, since they'll never move. | |
33398cf2 | 544 | */ |
62cccb8c | 545 | static inline void mem_cgroup_update_page_stat(struct page *page, |
33398cf2 MH |
546 | enum mem_cgroup_stat_index idx, int val) |
547 | { | |
62cccb8c | 548 | if (page->mem_cgroup) |
2a2e4885 | 549 | mem_cgroup_update_stat(page->mem_cgroup, idx, val); |
33398cf2 MH |
550 | } |
551 | ||
62cccb8c | 552 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 553 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 554 | { |
62cccb8c | 555 | mem_cgroup_update_page_stat(page, idx, 1); |
2a7106f2 GT |
556 | } |
557 | ||
62cccb8c | 558 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 559 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 560 | { |
62cccb8c | 561 | mem_cgroup_update_page_stat(page, idx, -1); |
2a7106f2 GT |
562 | } |
563 | ||
ef8f2327 | 564 | unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, |
0608f43d AM |
565 | gfp_t gfp_mask, |
566 | unsigned long *total_scanned); | |
a63d83f4 | 567 | |
68ae564b DR |
568 | static inline void mem_cgroup_count_vm_event(struct mm_struct *mm, |
569 | enum vm_event_item idx) | |
570 | { | |
33398cf2 MH |
571 | struct mem_cgroup *memcg; |
572 | ||
68ae564b DR |
573 | if (mem_cgroup_disabled()) |
574 | return; | |
33398cf2 MH |
575 | |
576 | rcu_read_lock(); | |
577 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
578 | if (unlikely(!memcg)) | |
579 | goto out; | |
580 | ||
581 | switch (idx) { | |
582 | case PGFAULT: | |
583 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]); | |
584 | break; | |
585 | case PGMAJFAULT: | |
586 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]); | |
587 | break; | |
588 | default: | |
589 | BUG(); | |
590 | } | |
591 | out: | |
592 | rcu_read_unlock(); | |
68ae564b | 593 | } |
ca3e0214 | 594 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
e94c8a9c | 595 | void mem_cgroup_split_huge_fixup(struct page *head); |
ca3e0214 KH |
596 | #endif |
597 | ||
c255a458 | 598 | #else /* CONFIG_MEMCG */ |
23047a96 JW |
599 | |
600 | #define MEM_CGROUP_ID_SHIFT 0 | |
601 | #define MEM_CGROUP_ID_MAX 0 | |
602 | ||
7a81b88c KH |
603 | struct mem_cgroup; |
604 | ||
23047a96 JW |
605 | static inline bool mem_cgroup_disabled(void) |
606 | { | |
607 | return true; | |
608 | } | |
609 | ||
31176c78 JW |
610 | static inline void mem_cgroup_event(struct mem_cgroup *memcg, |
611 | enum mem_cgroup_events_index idx) | |
241994ed JW |
612 | { |
613 | } | |
614 | ||
615 | static inline bool mem_cgroup_low(struct mem_cgroup *root, | |
616 | struct mem_cgroup *memcg) | |
617 | { | |
618 | return false; | |
619 | } | |
620 | ||
00501b53 JW |
621 | static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
622 | gfp_t gfp_mask, | |
f627c2f5 KS |
623 | struct mem_cgroup **memcgp, |
624 | bool compound) | |
7a81b88c | 625 | { |
00501b53 | 626 | *memcgp = NULL; |
7a81b88c KH |
627 | return 0; |
628 | } | |
629 | ||
00501b53 JW |
630 | static inline void mem_cgroup_commit_charge(struct page *page, |
631 | struct mem_cgroup *memcg, | |
f627c2f5 | 632 | bool lrucare, bool compound) |
7a81b88c KH |
633 | { |
634 | } | |
635 | ||
00501b53 | 636 | static inline void mem_cgroup_cancel_charge(struct page *page, |
f627c2f5 KS |
637 | struct mem_cgroup *memcg, |
638 | bool compound) | |
7a81b88c KH |
639 | { |
640 | } | |
641 | ||
0a31bc97 | 642 | static inline void mem_cgroup_uncharge(struct page *page) |
569b846d KH |
643 | { |
644 | } | |
645 | ||
747db954 | 646 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
8a9f3ccd BS |
647 | { |
648 | } | |
649 | ||
6a93ca8f | 650 | static inline void mem_cgroup_migrate(struct page *old, struct page *new) |
69029cd5 KH |
651 | { |
652 | } | |
653 | ||
a9dd0a83 | 654 | static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat, |
ef8f2327 | 655 | struct mem_cgroup *memcg) |
08e552c6 | 656 | { |
a9dd0a83 | 657 | return node_lruvec(pgdat); |
08e552c6 KH |
658 | } |
659 | ||
fa9add64 | 660 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
599d0c95 | 661 | struct pglist_data *pgdat) |
66e1707b | 662 | { |
599d0c95 | 663 | return &pgdat->lruvec; |
66e1707b BS |
664 | } |
665 | ||
587af308 | 666 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
c0ff4b85 | 667 | struct mem_cgroup *memcg) |
bed7161a | 668 | { |
587af308 | 669 | return true; |
bed7161a BS |
670 | } |
671 | ||
ffbdccf5 DR |
672 | static inline bool task_in_mem_cgroup(struct task_struct *task, |
673 | const struct mem_cgroup *memcg) | |
4c4a2214 | 674 | { |
ffbdccf5 | 675 | return true; |
4c4a2214 DR |
676 | } |
677 | ||
5660048c JW |
678 | static inline struct mem_cgroup * |
679 | mem_cgroup_iter(struct mem_cgroup *root, | |
680 | struct mem_cgroup *prev, | |
681 | struct mem_cgroup_reclaim_cookie *reclaim) | |
682 | { | |
683 | return NULL; | |
684 | } | |
685 | ||
686 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, | |
687 | struct mem_cgroup *prev) | |
688 | { | |
689 | } | |
690 | ||
7c5f64f8 VD |
691 | static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg, |
692 | int (*fn)(struct task_struct *, void *), void *arg) | |
693 | { | |
694 | return 0; | |
695 | } | |
696 | ||
23047a96 | 697 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
f8d66542 | 698 | { |
23047a96 JW |
699 | return 0; |
700 | } | |
701 | ||
702 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
703 | { | |
704 | WARN_ON_ONCE(id); | |
705 | /* XXX: This should always return root_mem_cgroup */ | |
706 | return NULL; | |
f8d66542 | 707 | } |
a636b327 | 708 | |
eb01aaab | 709 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
14797e23 | 710 | { |
13308ca9 | 711 | return true; |
14797e23 KM |
712 | } |
713 | ||
a3d8e054 | 714 | static inline unsigned long |
4d7dcca2 | 715 | mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
a3d8e054 KM |
716 | { |
717 | return 0; | |
718 | } | |
b4536f0c MH |
719 | static inline |
720 | unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, | |
721 | enum lru_list lru, int zone_idx) | |
722 | { | |
723 | return 0; | |
724 | } | |
a3d8e054 | 725 | |
0a6b76dd VD |
726 | static inline unsigned long |
727 | mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg, | |
728 | int nid, unsigned int lru_mask) | |
729 | { | |
730 | return 0; | |
731 | } | |
732 | ||
7c5f64f8 VD |
733 | static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg) |
734 | { | |
735 | return 0; | |
736 | } | |
737 | ||
e222432b BS |
738 | static inline void |
739 | mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p) | |
740 | { | |
741 | } | |
742 | ||
62cccb8c | 743 | static inline void lock_page_memcg(struct page *page) |
89c06bd5 KH |
744 | { |
745 | } | |
746 | ||
62cccb8c | 747 | static inline void unlock_page_memcg(struct page *page) |
89c06bd5 KH |
748 | { |
749 | } | |
750 | ||
b23afb93 TH |
751 | static inline void mem_cgroup_handle_over_high(void) |
752 | { | |
753 | } | |
754 | ||
49426420 | 755 | static inline void mem_cgroup_oom_enable(void) |
519e5247 JW |
756 | { |
757 | } | |
758 | ||
49426420 | 759 | static inline void mem_cgroup_oom_disable(void) |
519e5247 JW |
760 | { |
761 | } | |
762 | ||
3812c8c8 JW |
763 | static inline bool task_in_memcg_oom(struct task_struct *p) |
764 | { | |
765 | return false; | |
766 | } | |
767 | ||
49426420 | 768 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
3812c8c8 JW |
769 | { |
770 | return false; | |
771 | } | |
772 | ||
2a2e4885 JW |
773 | static inline unsigned long mem_cgroup_read_stat(struct mem_cgroup *memcg, |
774 | enum mem_cgroup_stat_index idx) | |
775 | { | |
776 | return 0; | |
777 | } | |
778 | ||
779 | static inline void mem_cgroup_update_stat(struct mem_cgroup *memcg, | |
780 | enum mem_cgroup_stat_index idx, int val) | |
781 | { | |
782 | } | |
783 | ||
784 | static inline void mem_cgroup_inc_stat(struct mem_cgroup *memcg, | |
785 | enum mem_cgroup_stat_index idx) | |
786 | { | |
787 | } | |
788 | ||
789 | static inline void mem_cgroup_dec_stat(struct mem_cgroup *memcg, | |
790 | enum mem_cgroup_stat_index idx) | |
791 | { | |
792 | } | |
793 | ||
553af430 JW |
794 | static inline void mem_cgroup_update_page_stat(struct page *page, |
795 | enum mem_cgroup_stat_index idx, | |
796 | int nr) | |
797 | { | |
798 | } | |
799 | ||
62cccb8c | 800 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 801 | enum mem_cgroup_stat_index idx) |
2a7106f2 GT |
802 | { |
803 | } | |
804 | ||
62cccb8c | 805 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 806 | enum mem_cgroup_stat_index idx) |
d69b042f BS |
807 | { |
808 | } | |
809 | ||
4e416953 | 810 | static inline |
ef8f2327 | 811 | unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, |
0608f43d AM |
812 | gfp_t gfp_mask, |
813 | unsigned long *total_scanned) | |
4e416953 | 814 | { |
0608f43d | 815 | return 0; |
4e416953 BS |
816 | } |
817 | ||
e94c8a9c | 818 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
ca3e0214 KH |
819 | { |
820 | } | |
821 | ||
456f998e YH |
822 | static inline |
823 | void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx) | |
824 | { | |
825 | } | |
c255a458 | 826 | #endif /* CONFIG_MEMCG */ |
78fb7466 | 827 | |
52ebea74 | 828 | #ifdef CONFIG_CGROUP_WRITEBACK |
841710aa | 829 | |
52ebea74 | 830 | struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); |
841710aa | 831 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
c5edf9cd TH |
832 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
833 | unsigned long *pheadroom, unsigned long *pdirty, | |
834 | unsigned long *pwriteback); | |
841710aa TH |
835 | |
836 | #else /* CONFIG_CGROUP_WRITEBACK */ | |
837 | ||
838 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) | |
839 | { | |
840 | return NULL; | |
841 | } | |
842 | ||
c2aa723a | 843 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
c5edf9cd TH |
844 | unsigned long *pfilepages, |
845 | unsigned long *pheadroom, | |
c2aa723a TH |
846 | unsigned long *pdirty, |
847 | unsigned long *pwriteback) | |
848 | { | |
849 | } | |
850 | ||
841710aa | 851 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
52ebea74 | 852 | |
e1aab161 | 853 | struct sock; |
baac50bb JW |
854 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
855 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); | |
d886f4e4 | 856 | #ifdef CONFIG_MEMCG |
ef12947c JW |
857 | extern struct static_key_false memcg_sockets_enabled_key; |
858 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) | |
2d758073 JW |
859 | void mem_cgroup_sk_alloc(struct sock *sk); |
860 | void mem_cgroup_sk_free(struct sock *sk); | |
baac50bb | 861 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c | 862 | { |
0db15298 | 863 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
8e8ae645 | 864 | return true; |
8e8ae645 JW |
865 | do { |
866 | if (time_before(jiffies, memcg->socket_pressure)) | |
867 | return true; | |
868 | } while ((memcg = parent_mem_cgroup(memcg))); | |
869 | return false; | |
e805605c JW |
870 | } |
871 | #else | |
80e95fe0 | 872 | #define mem_cgroup_sockets_enabled 0 |
2d758073 JW |
873 | static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; |
874 | static inline void mem_cgroup_sk_free(struct sock *sk) { }; | |
baac50bb | 875 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c JW |
876 | { |
877 | return false; | |
878 | } | |
879 | #endif | |
7ae1e1d0 | 880 | |
45264778 VD |
881 | struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep); |
882 | void memcg_kmem_put_cache(struct kmem_cache *cachep); | |
883 | int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order, | |
884 | struct mem_cgroup *memcg); | |
885 | int memcg_kmem_charge(struct page *page, gfp_t gfp, int order); | |
886 | void memcg_kmem_uncharge(struct page *page, int order); | |
887 | ||
127424c8 | 888 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
ef12947c | 889 | extern struct static_key_false memcg_kmem_enabled_key; |
17cc4dfe | 890 | extern struct workqueue_struct *memcg_kmem_cache_wq; |
749c5415 | 891 | |
dbcf73e2 | 892 | extern int memcg_nr_cache_ids; |
64219994 MH |
893 | void memcg_get_cache_ids(void); |
894 | void memcg_put_cache_ids(void); | |
ebe945c2 GC |
895 | |
896 | /* | |
897 | * Helper macro to loop through all memcg-specific caches. Callers must still | |
898 | * check if the cache is valid (it is either valid or NULL). | |
899 | * the slab_mutex must be held when looping through those caches | |
900 | */ | |
749c5415 | 901 | #define for_each_memcg_cache_index(_idx) \ |
dbcf73e2 | 902 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
749c5415 | 903 | |
7ae1e1d0 GC |
904 | static inline bool memcg_kmem_enabled(void) |
905 | { | |
ef12947c | 906 | return static_branch_unlikely(&memcg_kmem_enabled_key); |
7ae1e1d0 GC |
907 | } |
908 | ||
33398cf2 | 909 | /* |
9f706d68 | 910 | * helper for accessing a memcg's index. It will be used as an index in the |
33398cf2 MH |
911 | * child cache array in kmem_cache, and also to derive its name. This function |
912 | * will return -1 when this is not a kmem-limited memcg. | |
913 | */ | |
914 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
915 | { | |
916 | return memcg ? memcg->kmemcg_id : -1; | |
917 | } | |
5722d094 | 918 | |
27ee57c9 VD |
919 | /** |
920 | * memcg_kmem_update_page_stat - update kmem page state statistics | |
921 | * @page: the page | |
922 | * @idx: page state item to account | |
923 | * @val: number of pages (positive or negative) | |
924 | */ | |
925 | static inline void memcg_kmem_update_page_stat(struct page *page, | |
926 | enum mem_cgroup_stat_index idx, int val) | |
927 | { | |
928 | if (memcg_kmem_enabled() && page->mem_cgroup) | |
929 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); | |
930 | } | |
931 | ||
7ae1e1d0 | 932 | #else |
749c5415 GC |
933 | #define for_each_memcg_cache_index(_idx) \ |
934 | for (; NULL; ) | |
935 | ||
b9ce5ef4 GC |
936 | static inline bool memcg_kmem_enabled(void) |
937 | { | |
938 | return false; | |
939 | } | |
940 | ||
2633d7a0 GC |
941 | static inline int memcg_cache_id(struct mem_cgroup *memcg) |
942 | { | |
943 | return -1; | |
944 | } | |
945 | ||
05257a1a VD |
946 | static inline void memcg_get_cache_ids(void) |
947 | { | |
948 | } | |
949 | ||
950 | static inline void memcg_put_cache_ids(void) | |
951 | { | |
952 | } | |
953 | ||
27ee57c9 VD |
954 | static inline void memcg_kmem_update_page_stat(struct page *page, |
955 | enum mem_cgroup_stat_index idx, int val) | |
956 | { | |
957 | } | |
127424c8 JW |
958 | #endif /* CONFIG_MEMCG && !CONFIG_SLOB */ |
959 | ||
8cdea7c0 | 960 | #endif /* _LINUX_MEMCONTROL_H */ |