]>
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> | |
456f998e | 31 | |
78fb7466 | 32 | struct mem_cgroup; |
8697d331 BS |
33 | struct page; |
34 | struct mm_struct; | |
2633d7a0 | 35 | struct kmem_cache; |
78fb7466 | 36 | |
68b4876d SZ |
37 | /* |
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 | */ | |
41 | enum 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 */ | |
c4843a75 | 49 | MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */ |
3ea67d06 | 50 | MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */ |
68b4876d SZ |
51 | MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */ |
52 | MEM_CGROUP_STAT_NSTATS, | |
b2807f07 | 53 | /* default hierarchy stats */ |
c792e824 | 54 | MEMCG_SOCK = MEM_CGROUP_STAT_NSTATS, |
b2807f07 | 55 | MEMCG_NR_STAT, |
2a7106f2 GT |
56 | }; |
57 | ||
5660048c JW |
58 | struct mem_cgroup_reclaim_cookie { |
59 | struct zone *zone; | |
60 | int priority; | |
61 | unsigned int generation; | |
62 | }; | |
63 | ||
241994ed JW |
64 | enum mem_cgroup_events_index { |
65 | MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */ | |
66 | MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */ | |
67 | MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */ | |
68 | MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */ | |
69 | MEM_CGROUP_EVENTS_NSTATS, | |
70 | /* default hierarchy events */ | |
71 | MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS, | |
72 | MEMCG_HIGH, | |
73 | MEMCG_MAX, | |
74 | MEMCG_OOM, | |
75 | MEMCG_NR_EVENTS, | |
76 | }; | |
77 | ||
33398cf2 MH |
78 | /* |
79 | * Per memcg event counter is incremented at every pagein/pageout. With THP, | |
80 | * it will be incremated by the number of pages. This counter is used for | |
81 | * for trigger some periodic events. This is straightforward and better | |
82 | * than using jiffies etc. to handle periodic memcg event. | |
83 | */ | |
84 | enum mem_cgroup_events_target { | |
85 | MEM_CGROUP_TARGET_THRESH, | |
86 | MEM_CGROUP_TARGET_SOFTLIMIT, | |
87 | MEM_CGROUP_TARGET_NUMAINFO, | |
88 | MEM_CGROUP_NTARGETS, | |
89 | }; | |
90 | ||
c255a458 | 91 | #ifdef CONFIG_MEMCG |
23047a96 JW |
92 | |
93 | #define MEM_CGROUP_ID_SHIFT 16 | |
94 | #define MEM_CGROUP_ID_MAX USHRT_MAX | |
95 | ||
33398cf2 | 96 | struct mem_cgroup_stat_cpu { |
b2807f07 | 97 | long count[MEMCG_NR_STAT]; |
33398cf2 MH |
98 | unsigned long events[MEMCG_NR_EVENTS]; |
99 | unsigned long nr_page_events; | |
100 | unsigned long targets[MEM_CGROUP_NTARGETS]; | |
101 | }; | |
102 | ||
103 | struct mem_cgroup_reclaim_iter { | |
104 | struct mem_cgroup *position; | |
105 | /* scan generation, increased every round-trip */ | |
106 | unsigned int generation; | |
107 | }; | |
108 | ||
109 | /* | |
110 | * per-zone information in memory controller. | |
111 | */ | |
112 | struct mem_cgroup_per_zone { | |
113 | struct lruvec lruvec; | |
114 | unsigned long lru_size[NR_LRU_LISTS]; | |
115 | ||
116 | struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1]; | |
117 | ||
118 | struct rb_node tree_node; /* RB tree node */ | |
119 | unsigned long usage_in_excess;/* Set to the value by which */ | |
120 | /* the soft limit is exceeded*/ | |
121 | bool on_tree; | |
122 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ | |
123 | /* use container_of */ | |
124 | }; | |
125 | ||
126 | struct mem_cgroup_per_node { | |
127 | struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; | |
128 | }; | |
129 | ||
130 | struct mem_cgroup_threshold { | |
131 | struct eventfd_ctx *eventfd; | |
132 | unsigned long threshold; | |
133 | }; | |
134 | ||
135 | /* For threshold */ | |
136 | struct mem_cgroup_threshold_ary { | |
137 | /* An array index points to threshold just below or equal to usage. */ | |
138 | int current_threshold; | |
139 | /* Size of entries[] */ | |
140 | unsigned int size; | |
141 | /* Array of thresholds */ | |
142 | struct mem_cgroup_threshold entries[0]; | |
143 | }; | |
144 | ||
145 | struct mem_cgroup_thresholds { | |
146 | /* Primary thresholds array */ | |
147 | struct mem_cgroup_threshold_ary *primary; | |
148 | /* | |
149 | * Spare threshold array. | |
150 | * This is needed to make mem_cgroup_unregister_event() "never fail". | |
151 | * It must be able to store at least primary->size - 1 entries. | |
152 | */ | |
153 | struct mem_cgroup_threshold_ary *spare; | |
154 | }; | |
155 | ||
567e9ab2 JW |
156 | enum memcg_kmem_state { |
157 | KMEM_NONE, | |
158 | KMEM_ALLOCATED, | |
159 | KMEM_ONLINE, | |
160 | }; | |
161 | ||
33398cf2 MH |
162 | /* |
163 | * The memory controller data structure. The memory controller controls both | |
164 | * page cache and RSS per cgroup. We would eventually like to provide | |
165 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, | |
166 | * to help the administrator determine what knobs to tune. | |
167 | */ | |
168 | struct mem_cgroup { | |
169 | struct cgroup_subsys_state css; | |
170 | ||
171 | /* Accounted resources */ | |
172 | struct page_counter memory; | |
37e84351 | 173 | struct page_counter swap; |
0db15298 JW |
174 | |
175 | /* Legacy consumer-oriented counters */ | |
33398cf2 MH |
176 | struct page_counter memsw; |
177 | struct page_counter kmem; | |
0db15298 | 178 | struct page_counter tcpmem; |
33398cf2 MH |
179 | |
180 | /* Normal memory consumption range */ | |
181 | unsigned long low; | |
182 | unsigned long high; | |
183 | ||
f7e1cb6e JW |
184 | /* Range enforcement for interrupt charges */ |
185 | struct work_struct high_work; | |
186 | ||
33398cf2 MH |
187 | unsigned long soft_limit; |
188 | ||
189 | /* vmpressure notifications */ | |
190 | struct vmpressure vmpressure; | |
191 | ||
33398cf2 MH |
192 | /* |
193 | * Should the accounting and control be hierarchical, per subtree? | |
194 | */ | |
195 | bool use_hierarchy; | |
196 | ||
197 | /* protected by memcg_oom_lock */ | |
198 | bool oom_lock; | |
199 | int under_oom; | |
200 | ||
201 | int swappiness; | |
202 | /* OOM-Killer disable */ | |
203 | int oom_kill_disable; | |
204 | ||
472912a2 TH |
205 | /* handle for "memory.events" */ |
206 | struct cgroup_file events_file; | |
207 | ||
33398cf2 MH |
208 | /* protect arrays of thresholds */ |
209 | struct mutex thresholds_lock; | |
210 | ||
211 | /* thresholds for memory usage. RCU-protected */ | |
212 | struct mem_cgroup_thresholds thresholds; | |
213 | ||
214 | /* thresholds for mem+swap usage. RCU-protected */ | |
215 | struct mem_cgroup_thresholds memsw_thresholds; | |
216 | ||
217 | /* For oom notifier event fd */ | |
218 | struct list_head oom_notify; | |
219 | ||
220 | /* | |
221 | * Should we move charges of a task when a task is moved into this | |
222 | * mem_cgroup ? And what type of charges should we move ? | |
223 | */ | |
224 | unsigned long move_charge_at_immigrate; | |
225 | /* | |
226 | * set > 0 if pages under this cgroup are moving to other cgroup. | |
227 | */ | |
228 | atomic_t moving_account; | |
229 | /* taken only while moving_account > 0 */ | |
230 | spinlock_t move_lock; | |
231 | struct task_struct *move_lock_task; | |
232 | unsigned long move_lock_flags; | |
233 | /* | |
234 | * percpu counter. | |
235 | */ | |
236 | struct mem_cgroup_stat_cpu __percpu *stat; | |
33398cf2 | 237 | |
d886f4e4 JW |
238 | unsigned long socket_pressure; |
239 | ||
240 | /* Legacy tcp memory accounting */ | |
0db15298 JW |
241 | bool tcpmem_active; |
242 | int tcpmem_pressure; | |
d886f4e4 | 243 | |
127424c8 | 244 | #ifndef CONFIG_SLOB |
33398cf2 MH |
245 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ |
246 | int kmemcg_id; | |
567e9ab2 | 247 | enum memcg_kmem_state kmem_state; |
33398cf2 MH |
248 | #endif |
249 | ||
250 | int last_scanned_node; | |
251 | #if MAX_NUMNODES > 1 | |
252 | nodemask_t scan_nodes; | |
253 | atomic_t numainfo_events; | |
254 | atomic_t numainfo_updating; | |
255 | #endif | |
256 | ||
257 | #ifdef CONFIG_CGROUP_WRITEBACK | |
258 | struct list_head cgwb_list; | |
259 | struct wb_domain cgwb_domain; | |
260 | #endif | |
261 | ||
262 | /* List of events which userspace want to receive */ | |
263 | struct list_head event_list; | |
264 | spinlock_t event_list_lock; | |
265 | ||
266 | struct mem_cgroup_per_node *nodeinfo[0]; | |
267 | /* WARNING: nodeinfo must be the last member here */ | |
268 | }; | |
7d828602 JW |
269 | |
270 | extern struct mem_cgroup *root_mem_cgroup; | |
56161634 | 271 | |
23047a96 JW |
272 | static inline bool mem_cgroup_disabled(void) |
273 | { | |
274 | return !cgroup_subsys_enabled(memory_cgrp_subsys); | |
275 | } | |
276 | ||
33398cf2 MH |
277 | /** |
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 | */ | |
283 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, | |
241994ed | 284 | enum mem_cgroup_events_index idx, |
33398cf2 MH |
285 | unsigned int nr) |
286 | { | |
287 | this_cpu_add(memcg->stat->events[idx], nr); | |
472912a2 | 288 | cgroup_file_notify(&memcg->events_file); |
33398cf2 | 289 | } |
241994ed JW |
290 | |
291 | bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg); | |
292 | ||
00501b53 | 293 | int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
f627c2f5 KS |
294 | gfp_t gfp_mask, struct mem_cgroup **memcgp, |
295 | bool compound); | |
00501b53 | 296 | void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg, |
f627c2f5 KS |
297 | bool lrucare, bool compound); |
298 | void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg, | |
299 | bool compound); | |
0a31bc97 | 300 | void mem_cgroup_uncharge(struct page *page); |
747db954 | 301 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
569b846d | 302 | |
6a93ca8f | 303 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage); |
569b846d | 304 | |
0a31bc97 JW |
305 | struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *); |
306 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *); | |
c9b0ed51 | 307 | |
2314b42d | 308 | bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg); |
64219994 | 309 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
e993d905 | 310 | |
33398cf2 MH |
311 | static inline |
312 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ | |
313 | return css ? container_of(css, struct mem_cgroup, css) : NULL; | |
314 | } | |
315 | ||
8e8ae645 JW |
316 | #define mem_cgroup_from_counter(counter, member) \ |
317 | container_of(counter, struct mem_cgroup, member) | |
318 | ||
33398cf2 MH |
319 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
320 | struct mem_cgroup *, | |
321 | struct mem_cgroup_reclaim_cookie *); | |
322 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); | |
323 | ||
23047a96 JW |
324 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
325 | { | |
326 | if (mem_cgroup_disabled()) | |
327 | return 0; | |
328 | ||
329 | return memcg->css.id; | |
330 | } | |
331 | ||
332 | /** | |
333 | * mem_cgroup_from_id - look up a memcg from an id | |
334 | * @id: the id to look up | |
335 | * | |
336 | * Caller must hold rcu_read_lock() and use css_tryget() as necessary. | |
337 | */ | |
338 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
339 | { | |
340 | struct cgroup_subsys_state *css; | |
341 | ||
342 | css = css_from_id(id, &memory_cgrp_subsys); | |
343 | return mem_cgroup_from_css(css); | |
344 | } | |
345 | ||
8e8ae645 JW |
346 | /** |
347 | * parent_mem_cgroup - find the accounting parent of a memcg | |
348 | * @memcg: memcg whose parent to find | |
349 | * | |
350 | * Returns the parent memcg, or NULL if this is the root or the memory | |
351 | * controller is in legacy no-hierarchy mode. | |
352 | */ | |
353 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) | |
354 | { | |
355 | if (!memcg->memory.parent) | |
356 | return NULL; | |
357 | return mem_cgroup_from_counter(memcg->memory.parent, memory); | |
358 | } | |
359 | ||
33398cf2 MH |
360 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
361 | struct mem_cgroup *root) | |
362 | { | |
363 | if (root == memcg) | |
364 | return true; | |
365 | if (!root->use_hierarchy) | |
366 | return false; | |
367 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); | |
368 | } | |
e1aab161 | 369 | |
2314b42d JW |
370 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
371 | struct mem_cgroup *memcg) | |
2e4d4091 | 372 | { |
587af308 | 373 | struct mem_cgroup *task_memcg; |
413918bb | 374 | bool match = false; |
c3ac9a8a | 375 | |
2e4d4091 | 376 | rcu_read_lock(); |
587af308 | 377 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
413918bb | 378 | if (task_memcg) |
2314b42d | 379 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
2e4d4091 | 380 | rcu_read_unlock(); |
c3ac9a8a | 381 | return match; |
2e4d4091 | 382 | } |
8a9f3ccd | 383 | |
64219994 | 384 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
2fc04524 | 385 | ino_t page_cgroup_ino(struct page *page); |
d324236b | 386 | |
eb01aaab VD |
387 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
388 | { | |
389 | if (mem_cgroup_disabled()) | |
390 | return true; | |
391 | return !!(memcg->css.flags & CSS_ONLINE); | |
392 | } | |
393 | ||
58ae83db KH |
394 | /* |
395 | * For memory reclaim. | |
396 | */ | |
889976db | 397 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
33398cf2 MH |
398 | |
399 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
400 | int nr_pages); | |
401 | ||
33398cf2 MH |
402 | static inline |
403 | unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) | |
404 | { | |
405 | struct mem_cgroup_per_zone *mz; | |
406 | ||
407 | mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec); | |
408 | return mz->lru_size[lru]; | |
409 | } | |
410 | ||
13308ca9 | 411 | static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) |
33398cf2 MH |
412 | { |
413 | unsigned long inactive_ratio; | |
414 | unsigned long inactive; | |
415 | unsigned long active; | |
416 | unsigned long gb; | |
417 | ||
418 | inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON); | |
419 | active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON); | |
420 | ||
421 | gb = (inactive + active) >> (30 - PAGE_SHIFT); | |
422 | if (gb) | |
423 | inactive_ratio = int_sqrt(10 * gb); | |
424 | else | |
425 | inactive_ratio = 1; | |
426 | ||
427 | return inactive * inactive_ratio < active; | |
428 | } | |
429 | ||
b23afb93 TH |
430 | void mem_cgroup_handle_over_high(void); |
431 | ||
64219994 MH |
432 | void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, |
433 | struct task_struct *p); | |
58ae83db | 434 | |
49426420 | 435 | static inline void mem_cgroup_oom_enable(void) |
519e5247 | 436 | { |
626ebc41 TH |
437 | WARN_ON(current->memcg_may_oom); |
438 | current->memcg_may_oom = 1; | |
519e5247 JW |
439 | } |
440 | ||
49426420 | 441 | static inline void mem_cgroup_oom_disable(void) |
519e5247 | 442 | { |
626ebc41 TH |
443 | WARN_ON(!current->memcg_may_oom); |
444 | current->memcg_may_oom = 0; | |
519e5247 JW |
445 | } |
446 | ||
3812c8c8 JW |
447 | static inline bool task_in_memcg_oom(struct task_struct *p) |
448 | { | |
626ebc41 | 449 | return p->memcg_in_oom; |
3812c8c8 JW |
450 | } |
451 | ||
49426420 | 452 | bool mem_cgroup_oom_synchronize(bool wait); |
3812c8c8 | 453 | |
c255a458 | 454 | #ifdef CONFIG_MEMCG_SWAP |
c077719b KH |
455 | extern int do_swap_account; |
456 | #endif | |
f8d66542 | 457 | |
62cccb8c JW |
458 | void lock_page_memcg(struct page *page); |
459 | void unlock_page_memcg(struct page *page); | |
d7365e78 | 460 | |
33398cf2 MH |
461 | /** |
462 | * mem_cgroup_update_page_stat - update page state statistics | |
62cccb8c | 463 | * @page: the page |
33398cf2 MH |
464 | * @idx: page state item to account |
465 | * @val: number of pages (positive or negative) | |
466 | * | |
81f8c3a4 JW |
467 | * Callers must use lock_page_memcg() to prevent double accounting |
468 | * when the page is concurrently being moved to another memcg: | |
469 | * | |
62cccb8c | 470 | * lock_page_memcg(page); |
81f8c3a4 | 471 | * if (TestClearPageState(page)) |
62cccb8c JW |
472 | * mem_cgroup_update_page_stat(page, state, -1); |
473 | * unlock_page_memcg(page); | |
33398cf2 | 474 | */ |
62cccb8c | 475 | static inline void mem_cgroup_update_page_stat(struct page *page, |
33398cf2 MH |
476 | enum mem_cgroup_stat_index idx, int val) |
477 | { | |
478 | VM_BUG_ON(!rcu_read_lock_held()); | |
479 | ||
62cccb8c JW |
480 | if (page->mem_cgroup) |
481 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); | |
33398cf2 MH |
482 | } |
483 | ||
62cccb8c | 484 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 485 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 486 | { |
62cccb8c | 487 | mem_cgroup_update_page_stat(page, idx, 1); |
2a7106f2 GT |
488 | } |
489 | ||
62cccb8c | 490 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 491 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 492 | { |
62cccb8c | 493 | mem_cgroup_update_page_stat(page, idx, -1); |
2a7106f2 GT |
494 | } |
495 | ||
0608f43d AM |
496 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
497 | gfp_t gfp_mask, | |
498 | unsigned long *total_scanned); | |
a63d83f4 | 499 | |
68ae564b DR |
500 | static inline void mem_cgroup_count_vm_event(struct mm_struct *mm, |
501 | enum vm_event_item idx) | |
502 | { | |
33398cf2 MH |
503 | struct mem_cgroup *memcg; |
504 | ||
68ae564b DR |
505 | if (mem_cgroup_disabled()) |
506 | return; | |
33398cf2 MH |
507 | |
508 | rcu_read_lock(); | |
509 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
510 | if (unlikely(!memcg)) | |
511 | goto out; | |
512 | ||
513 | switch (idx) { | |
514 | case PGFAULT: | |
515 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]); | |
516 | break; | |
517 | case PGMAJFAULT: | |
518 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]); | |
519 | break; | |
520 | default: | |
521 | BUG(); | |
522 | } | |
523 | out: | |
524 | rcu_read_unlock(); | |
68ae564b | 525 | } |
ca3e0214 | 526 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
e94c8a9c | 527 | void mem_cgroup_split_huge_fixup(struct page *head); |
ca3e0214 KH |
528 | #endif |
529 | ||
c255a458 | 530 | #else /* CONFIG_MEMCG */ |
23047a96 JW |
531 | |
532 | #define MEM_CGROUP_ID_SHIFT 0 | |
533 | #define MEM_CGROUP_ID_MAX 0 | |
534 | ||
7a81b88c KH |
535 | struct mem_cgroup; |
536 | ||
23047a96 JW |
537 | static inline bool mem_cgroup_disabled(void) |
538 | { | |
539 | return true; | |
540 | } | |
541 | ||
241994ed JW |
542 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, |
543 | enum mem_cgroup_events_index idx, | |
544 | unsigned int nr) | |
545 | { | |
546 | } | |
547 | ||
548 | static inline bool mem_cgroup_low(struct mem_cgroup *root, | |
549 | struct mem_cgroup *memcg) | |
550 | { | |
551 | return false; | |
552 | } | |
553 | ||
00501b53 JW |
554 | static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
555 | gfp_t gfp_mask, | |
f627c2f5 KS |
556 | struct mem_cgroup **memcgp, |
557 | bool compound) | |
7a81b88c | 558 | { |
00501b53 | 559 | *memcgp = NULL; |
7a81b88c KH |
560 | return 0; |
561 | } | |
562 | ||
00501b53 JW |
563 | static inline void mem_cgroup_commit_charge(struct page *page, |
564 | struct mem_cgroup *memcg, | |
f627c2f5 | 565 | bool lrucare, bool compound) |
7a81b88c KH |
566 | { |
567 | } | |
568 | ||
00501b53 | 569 | static inline void mem_cgroup_cancel_charge(struct page *page, |
f627c2f5 KS |
570 | struct mem_cgroup *memcg, |
571 | bool compound) | |
7a81b88c KH |
572 | { |
573 | } | |
574 | ||
0a31bc97 | 575 | static inline void mem_cgroup_uncharge(struct page *page) |
569b846d KH |
576 | { |
577 | } | |
578 | ||
747db954 | 579 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
8a9f3ccd BS |
580 | { |
581 | } | |
582 | ||
6a93ca8f | 583 | static inline void mem_cgroup_migrate(struct page *old, struct page *new) |
69029cd5 KH |
584 | { |
585 | } | |
586 | ||
925b7673 JW |
587 | static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone, |
588 | struct mem_cgroup *memcg) | |
08e552c6 | 589 | { |
925b7673 | 590 | return &zone->lruvec; |
08e552c6 KH |
591 | } |
592 | ||
fa9add64 HD |
593 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
594 | struct zone *zone) | |
66e1707b | 595 | { |
925b7673 | 596 | return &zone->lruvec; |
66e1707b BS |
597 | } |
598 | ||
587af308 | 599 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
c0ff4b85 | 600 | struct mem_cgroup *memcg) |
bed7161a | 601 | { |
587af308 | 602 | return true; |
bed7161a BS |
603 | } |
604 | ||
ffbdccf5 DR |
605 | static inline bool task_in_mem_cgroup(struct task_struct *task, |
606 | const struct mem_cgroup *memcg) | |
4c4a2214 | 607 | { |
ffbdccf5 | 608 | return true; |
4c4a2214 DR |
609 | } |
610 | ||
5660048c JW |
611 | static inline struct mem_cgroup * |
612 | mem_cgroup_iter(struct mem_cgroup *root, | |
613 | struct mem_cgroup *prev, | |
614 | struct mem_cgroup_reclaim_cookie *reclaim) | |
615 | { | |
616 | return NULL; | |
617 | } | |
618 | ||
619 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, | |
620 | struct mem_cgroup *prev) | |
621 | { | |
622 | } | |
623 | ||
23047a96 | 624 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
f8d66542 | 625 | { |
23047a96 JW |
626 | return 0; |
627 | } | |
628 | ||
629 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
630 | { | |
631 | WARN_ON_ONCE(id); | |
632 | /* XXX: This should always return root_mem_cgroup */ | |
633 | return NULL; | |
f8d66542 | 634 | } |
a636b327 | 635 | |
eb01aaab | 636 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
14797e23 | 637 | { |
13308ca9 | 638 | return true; |
14797e23 KM |
639 | } |
640 | ||
eb01aaab VD |
641 | static inline bool |
642 | mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) | |
90cbc250 VD |
643 | { |
644 | return true; | |
645 | } | |
646 | ||
a3d8e054 | 647 | static inline unsigned long |
4d7dcca2 | 648 | mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
a3d8e054 KM |
649 | { |
650 | return 0; | |
651 | } | |
652 | ||
fa9add64 HD |
653 | static inline void |
654 | mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
655 | int increment) | |
3e2f41f1 | 656 | { |
3e2f41f1 KM |
657 | } |
658 | ||
e222432b BS |
659 | static inline void |
660 | mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p) | |
661 | { | |
662 | } | |
663 | ||
62cccb8c | 664 | static inline void lock_page_memcg(struct page *page) |
89c06bd5 KH |
665 | { |
666 | } | |
667 | ||
62cccb8c | 668 | static inline void unlock_page_memcg(struct page *page) |
89c06bd5 KH |
669 | { |
670 | } | |
671 | ||
b23afb93 TH |
672 | static inline void mem_cgroup_handle_over_high(void) |
673 | { | |
674 | } | |
675 | ||
49426420 | 676 | static inline void mem_cgroup_oom_enable(void) |
519e5247 JW |
677 | { |
678 | } | |
679 | ||
49426420 | 680 | static inline void mem_cgroup_oom_disable(void) |
519e5247 JW |
681 | { |
682 | } | |
683 | ||
3812c8c8 JW |
684 | static inline bool task_in_memcg_oom(struct task_struct *p) |
685 | { | |
686 | return false; | |
687 | } | |
688 | ||
49426420 | 689 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
3812c8c8 JW |
690 | { |
691 | return false; | |
692 | } | |
693 | ||
62cccb8c | 694 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 695 | enum mem_cgroup_stat_index idx) |
2a7106f2 GT |
696 | { |
697 | } | |
698 | ||
62cccb8c | 699 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 700 | enum mem_cgroup_stat_index idx) |
d69b042f BS |
701 | { |
702 | } | |
703 | ||
4e416953 | 704 | static inline |
0608f43d AM |
705 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
706 | gfp_t gfp_mask, | |
707 | unsigned long *total_scanned) | |
4e416953 | 708 | { |
0608f43d | 709 | return 0; |
4e416953 BS |
710 | } |
711 | ||
e94c8a9c | 712 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
ca3e0214 KH |
713 | { |
714 | } | |
715 | ||
456f998e YH |
716 | static inline |
717 | void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx) | |
718 | { | |
719 | } | |
c255a458 | 720 | #endif /* CONFIG_MEMCG */ |
78fb7466 | 721 | |
52ebea74 | 722 | #ifdef CONFIG_CGROUP_WRITEBACK |
841710aa | 723 | |
52ebea74 | 724 | struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); |
841710aa | 725 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
c5edf9cd TH |
726 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
727 | unsigned long *pheadroom, unsigned long *pdirty, | |
728 | unsigned long *pwriteback); | |
841710aa TH |
729 | |
730 | #else /* CONFIG_CGROUP_WRITEBACK */ | |
731 | ||
732 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) | |
733 | { | |
734 | return NULL; | |
735 | } | |
736 | ||
c2aa723a | 737 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
c5edf9cd TH |
738 | unsigned long *pfilepages, |
739 | unsigned long *pheadroom, | |
c2aa723a TH |
740 | unsigned long *pdirty, |
741 | unsigned long *pwriteback) | |
742 | { | |
743 | } | |
744 | ||
841710aa | 745 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
52ebea74 | 746 | |
e1aab161 | 747 | struct sock; |
e1aab161 GC |
748 | void sock_update_memcg(struct sock *sk); |
749 | void sock_release_memcg(struct sock *sk); | |
baac50bb JW |
750 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
751 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); | |
d886f4e4 | 752 | #ifdef CONFIG_MEMCG |
ef12947c JW |
753 | extern struct static_key_false memcg_sockets_enabled_key; |
754 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) | |
baac50bb | 755 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c | 756 | { |
0db15298 | 757 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
8e8ae645 | 758 | return true; |
8e8ae645 JW |
759 | do { |
760 | if (time_before(jiffies, memcg->socket_pressure)) | |
761 | return true; | |
762 | } while ((memcg = parent_mem_cgroup(memcg))); | |
763 | return false; | |
e805605c JW |
764 | } |
765 | #else | |
80e95fe0 | 766 | #define mem_cgroup_sockets_enabled 0 |
baac50bb | 767 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c JW |
768 | { |
769 | return false; | |
770 | } | |
771 | #endif | |
7ae1e1d0 | 772 | |
127424c8 | 773 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
ef12947c | 774 | extern struct static_key_false memcg_kmem_enabled_key; |
749c5415 | 775 | |
dbcf73e2 | 776 | extern int memcg_nr_cache_ids; |
64219994 MH |
777 | void memcg_get_cache_ids(void); |
778 | void memcg_put_cache_ids(void); | |
ebe945c2 GC |
779 | |
780 | /* | |
781 | * Helper macro to loop through all memcg-specific caches. Callers must still | |
782 | * check if the cache is valid (it is either valid or NULL). | |
783 | * the slab_mutex must be held when looping through those caches | |
784 | */ | |
749c5415 | 785 | #define for_each_memcg_cache_index(_idx) \ |
dbcf73e2 | 786 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
749c5415 | 787 | |
7ae1e1d0 GC |
788 | static inline bool memcg_kmem_enabled(void) |
789 | { | |
ef12947c | 790 | return static_branch_unlikely(&memcg_kmem_enabled_key); |
7ae1e1d0 GC |
791 | } |
792 | ||
567e9ab2 | 793 | static inline bool memcg_kmem_online(struct mem_cgroup *memcg) |
33398cf2 | 794 | { |
567e9ab2 | 795 | return memcg->kmem_state == KMEM_ONLINE; |
33398cf2 | 796 | } |
cb731d6c | 797 | |
7ae1e1d0 GC |
798 | /* |
799 | * In general, we'll do everything in our power to not incur in any overhead | |
800 | * for non-memcg users for the kmem functions. Not even a function call, if we | |
801 | * can avoid it. | |
802 | * | |
803 | * Therefore, we'll inline all those functions so that in the best case, we'll | |
804 | * see that kmemcg is off for everybody and proceed quickly. If it is on, | |
805 | * we'll still do most of the flag checking inline. We check a lot of | |
806 | * conditions, but because they are pretty simple, they are expected to be | |
807 | * fast. | |
808 | */ | |
f3ccb2c4 VD |
809 | int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order, |
810 | struct mem_cgroup *memcg); | |
d05e83a6 VD |
811 | int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order); |
812 | void __memcg_kmem_uncharge(struct page *page, int order); | |
7ae1e1d0 | 813 | |
33398cf2 | 814 | /* |
9f706d68 | 815 | * helper for accessing a memcg's index. It will be used as an index in the |
33398cf2 MH |
816 | * child cache array in kmem_cache, and also to derive its name. This function |
817 | * will return -1 when this is not a kmem-limited memcg. | |
818 | */ | |
819 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
820 | { | |
821 | return memcg ? memcg->kmemcg_id : -1; | |
822 | } | |
5722d094 | 823 | |
230e9fc2 | 824 | struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp); |
8135be5a | 825 | void __memcg_kmem_put_cache(struct kmem_cache *cachep); |
d7f25f8a | 826 | |
230e9fc2 | 827 | static inline bool __memcg_kmem_bypass(void) |
7ae1e1d0 GC |
828 | { |
829 | if (!memcg_kmem_enabled()) | |
830 | return true; | |
7ae1e1d0 GC |
831 | if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD)) |
832 | return true; | |
cbfb4798 | 833 | return false; |
7ae1e1d0 GC |
834 | } |
835 | ||
836 | /** | |
d05e83a6 VD |
837 | * memcg_kmem_charge: charge a kmem page |
838 | * @page: page to charge | |
839 | * @gfp: reclaim mode | |
840 | * @order: allocation order | |
cbfb4798 | 841 | * |
d05e83a6 | 842 | * Returns 0 on success, an error code on failure. |
7ae1e1d0 | 843 | */ |
d05e83a6 VD |
844 | static __always_inline int memcg_kmem_charge(struct page *page, |
845 | gfp_t gfp, int order) | |
7ae1e1d0 | 846 | { |
230e9fc2 VD |
847 | if (__memcg_kmem_bypass()) |
848 | return 0; | |
849 | if (!(gfp & __GFP_ACCOUNT)) | |
d05e83a6 VD |
850 | return 0; |
851 | return __memcg_kmem_charge(page, gfp, order); | |
7ae1e1d0 GC |
852 | } |
853 | ||
854 | /** | |
d05e83a6 VD |
855 | * memcg_kmem_uncharge: uncharge a kmem page |
856 | * @page: page to uncharge | |
857 | * @order: allocation order | |
7ae1e1d0 | 858 | */ |
d05e83a6 | 859 | static __always_inline void memcg_kmem_uncharge(struct page *page, int order) |
7ae1e1d0 | 860 | { |
7ae1e1d0 | 861 | if (memcg_kmem_enabled()) |
d05e83a6 | 862 | __memcg_kmem_uncharge(page, order); |
7ae1e1d0 GC |
863 | } |
864 | ||
d7f25f8a GC |
865 | /** |
866 | * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation | |
867 | * @cachep: the original global kmem cache | |
d7f25f8a | 868 | * |
5dfb4175 | 869 | * All memory allocated from a per-memcg cache is charged to the owner memcg. |
d7f25f8a GC |
870 | */ |
871 | static __always_inline struct kmem_cache * | |
872 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
873 | { | |
230e9fc2 | 874 | if (__memcg_kmem_bypass()) |
d7f25f8a | 875 | return cachep; |
230e9fc2 | 876 | return __memcg_kmem_get_cache(cachep, gfp); |
d7f25f8a | 877 | } |
8135be5a VD |
878 | |
879 | static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
880 | { | |
881 | if (memcg_kmem_enabled()) | |
882 | __memcg_kmem_put_cache(cachep); | |
883 | } | |
7ae1e1d0 | 884 | #else |
749c5415 GC |
885 | #define for_each_memcg_cache_index(_idx) \ |
886 | for (; NULL; ) | |
887 | ||
b9ce5ef4 GC |
888 | static inline bool memcg_kmem_enabled(void) |
889 | { | |
890 | return false; | |
891 | } | |
892 | ||
567e9ab2 | 893 | static inline bool memcg_kmem_online(struct mem_cgroup *memcg) |
cb731d6c VD |
894 | { |
895 | return false; | |
896 | } | |
897 | ||
d05e83a6 | 898 | static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order) |
7ae1e1d0 | 899 | { |
d05e83a6 | 900 | return 0; |
7ae1e1d0 GC |
901 | } |
902 | ||
d05e83a6 | 903 | static inline void memcg_kmem_uncharge(struct page *page, int order) |
7ae1e1d0 GC |
904 | { |
905 | } | |
2633d7a0 GC |
906 | |
907 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
908 | { | |
909 | return -1; | |
910 | } | |
911 | ||
05257a1a VD |
912 | static inline void memcg_get_cache_ids(void) |
913 | { | |
914 | } | |
915 | ||
916 | static inline void memcg_put_cache_ids(void) | |
917 | { | |
918 | } | |
919 | ||
d7f25f8a GC |
920 | static inline struct kmem_cache * |
921 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
922 | { | |
923 | return cachep; | |
924 | } | |
8135be5a VD |
925 | |
926 | static inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
927 | { | |
928 | } | |
127424c8 JW |
929 | #endif /* CONFIG_MEMCG && !CONFIG_SLOB */ |
930 | ||
8cdea7c0 | 931 | #endif /* _LINUX_MEMCONTROL_H */ |