]>
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 */ | |
49 | MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */ | |
c4843a75 | 50 | MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */ |
3ea67d06 | 51 | MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */ |
68b4876d SZ |
52 | MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */ |
53 | MEM_CGROUP_STAT_NSTATS, | |
b2807f07 | 54 | /* default hierarchy stats */ |
c792e824 | 55 | MEMCG_SOCK = MEM_CGROUP_STAT_NSTATS, |
b2807f07 | 56 | MEMCG_NR_STAT, |
2a7106f2 GT |
57 | }; |
58 | ||
5660048c JW |
59 | struct mem_cgroup_reclaim_cookie { |
60 | struct zone *zone; | |
61 | int priority; | |
62 | unsigned int generation; | |
63 | }; | |
64 | ||
241994ed JW |
65 | enum mem_cgroup_events_index { |
66 | MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */ | |
67 | MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */ | |
68 | MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */ | |
69 | MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */ | |
70 | MEM_CGROUP_EVENTS_NSTATS, | |
71 | /* default hierarchy events */ | |
72 | MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS, | |
73 | MEMCG_HIGH, | |
74 | MEMCG_MAX, | |
75 | MEMCG_OOM, | |
76 | MEMCG_NR_EVENTS, | |
77 | }; | |
78 | ||
33398cf2 MH |
79 | /* |
80 | * Per memcg event counter is incremented at every pagein/pageout. With THP, | |
81 | * it will be incremated by the number of pages. This counter is used for | |
82 | * for trigger some periodic events. This is straightforward and better | |
83 | * than using jiffies etc. to handle periodic memcg event. | |
84 | */ | |
85 | enum mem_cgroup_events_target { | |
86 | MEM_CGROUP_TARGET_THRESH, | |
87 | MEM_CGROUP_TARGET_SOFTLIMIT, | |
88 | MEM_CGROUP_TARGET_NUMAINFO, | |
89 | MEM_CGROUP_NTARGETS, | |
90 | }; | |
91 | ||
c255a458 | 92 | #ifdef CONFIG_MEMCG |
23047a96 JW |
93 | |
94 | #define MEM_CGROUP_ID_SHIFT 16 | |
95 | #define MEM_CGROUP_ID_MAX USHRT_MAX | |
96 | ||
33398cf2 | 97 | struct mem_cgroup_stat_cpu { |
b2807f07 | 98 | long count[MEMCG_NR_STAT]; |
33398cf2 MH |
99 | unsigned long events[MEMCG_NR_EVENTS]; |
100 | unsigned long nr_page_events; | |
101 | unsigned long targets[MEM_CGROUP_NTARGETS]; | |
102 | }; | |
103 | ||
104 | struct mem_cgroup_reclaim_iter { | |
105 | struct mem_cgroup *position; | |
106 | /* scan generation, increased every round-trip */ | |
107 | unsigned int generation; | |
108 | }; | |
109 | ||
110 | /* | |
111 | * per-zone information in memory controller. | |
112 | */ | |
113 | struct mem_cgroup_per_zone { | |
114 | struct lruvec lruvec; | |
115 | unsigned long lru_size[NR_LRU_LISTS]; | |
116 | ||
117 | struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1]; | |
118 | ||
119 | struct rb_node tree_node; /* RB tree node */ | |
120 | unsigned long usage_in_excess;/* Set to the value by which */ | |
121 | /* the soft limit is exceeded*/ | |
122 | bool on_tree; | |
123 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ | |
124 | /* use container_of */ | |
125 | }; | |
126 | ||
127 | struct mem_cgroup_per_node { | |
128 | struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; | |
129 | }; | |
130 | ||
131 | struct mem_cgroup_threshold { | |
132 | struct eventfd_ctx *eventfd; | |
133 | unsigned long threshold; | |
134 | }; | |
135 | ||
136 | /* For threshold */ | |
137 | struct mem_cgroup_threshold_ary { | |
138 | /* An array index points to threshold just below or equal to usage. */ | |
139 | int current_threshold; | |
140 | /* Size of entries[] */ | |
141 | unsigned int size; | |
142 | /* Array of thresholds */ | |
143 | struct mem_cgroup_threshold entries[0]; | |
144 | }; | |
145 | ||
146 | struct mem_cgroup_thresholds { | |
147 | /* Primary thresholds array */ | |
148 | struct mem_cgroup_threshold_ary *primary; | |
149 | /* | |
150 | * Spare threshold array. | |
151 | * This is needed to make mem_cgroup_unregister_event() "never fail". | |
152 | * It must be able to store at least primary->size - 1 entries. | |
153 | */ | |
154 | struct mem_cgroup_threshold_ary *spare; | |
155 | }; | |
156 | ||
567e9ab2 JW |
157 | enum memcg_kmem_state { |
158 | KMEM_NONE, | |
159 | KMEM_ALLOCATED, | |
160 | KMEM_ONLINE, | |
161 | }; | |
162 | ||
33398cf2 MH |
163 | /* |
164 | * The memory controller data structure. The memory controller controls both | |
165 | * page cache and RSS per cgroup. We would eventually like to provide | |
166 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, | |
167 | * to help the administrator determine what knobs to tune. | |
168 | */ | |
169 | struct mem_cgroup { | |
170 | struct cgroup_subsys_state css; | |
171 | ||
172 | /* Accounted resources */ | |
173 | struct page_counter memory; | |
37e84351 | 174 | struct page_counter swap; |
0db15298 JW |
175 | |
176 | /* Legacy consumer-oriented counters */ | |
33398cf2 MH |
177 | struct page_counter memsw; |
178 | struct page_counter kmem; | |
0db15298 | 179 | struct page_counter tcpmem; |
33398cf2 MH |
180 | |
181 | /* Normal memory consumption range */ | |
182 | unsigned long low; | |
183 | unsigned long high; | |
184 | ||
f7e1cb6e JW |
185 | /* Range enforcement for interrupt charges */ |
186 | struct work_struct high_work; | |
187 | ||
33398cf2 MH |
188 | unsigned long soft_limit; |
189 | ||
190 | /* vmpressure notifications */ | |
191 | struct vmpressure vmpressure; | |
192 | ||
33398cf2 MH |
193 | /* |
194 | * Should the accounting and control be hierarchical, per subtree? | |
195 | */ | |
196 | bool use_hierarchy; | |
197 | ||
198 | /* protected by memcg_oom_lock */ | |
199 | bool oom_lock; | |
200 | int under_oom; | |
201 | ||
202 | int swappiness; | |
203 | /* OOM-Killer disable */ | |
204 | int oom_kill_disable; | |
205 | ||
472912a2 TH |
206 | /* handle for "memory.events" */ |
207 | struct cgroup_file events_file; | |
208 | ||
33398cf2 MH |
209 | /* protect arrays of thresholds */ |
210 | struct mutex thresholds_lock; | |
211 | ||
212 | /* thresholds for memory usage. RCU-protected */ | |
213 | struct mem_cgroup_thresholds thresholds; | |
214 | ||
215 | /* thresholds for mem+swap usage. RCU-protected */ | |
216 | struct mem_cgroup_thresholds memsw_thresholds; | |
217 | ||
218 | /* For oom notifier event fd */ | |
219 | struct list_head oom_notify; | |
220 | ||
221 | /* | |
222 | * Should we move charges of a task when a task is moved into this | |
223 | * mem_cgroup ? And what type of charges should we move ? | |
224 | */ | |
225 | unsigned long move_charge_at_immigrate; | |
226 | /* | |
227 | * set > 0 if pages under this cgroup are moving to other cgroup. | |
228 | */ | |
229 | atomic_t moving_account; | |
230 | /* taken only while moving_account > 0 */ | |
231 | spinlock_t move_lock; | |
232 | struct task_struct *move_lock_task; | |
233 | unsigned long move_lock_flags; | |
234 | /* | |
235 | * percpu counter. | |
236 | */ | |
237 | struct mem_cgroup_stat_cpu __percpu *stat; | |
33398cf2 | 238 | |
d886f4e4 JW |
239 | unsigned long socket_pressure; |
240 | ||
241 | /* Legacy tcp memory accounting */ | |
0db15298 JW |
242 | bool tcpmem_active; |
243 | int tcpmem_pressure; | |
d886f4e4 | 244 | |
127424c8 | 245 | #ifndef CONFIG_SLOB |
33398cf2 MH |
246 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ |
247 | int kmemcg_id; | |
567e9ab2 | 248 | enum memcg_kmem_state kmem_state; |
33398cf2 MH |
249 | #endif |
250 | ||
251 | int last_scanned_node; | |
252 | #if MAX_NUMNODES > 1 | |
253 | nodemask_t scan_nodes; | |
254 | atomic_t numainfo_events; | |
255 | atomic_t numainfo_updating; | |
256 | #endif | |
257 | ||
258 | #ifdef CONFIG_CGROUP_WRITEBACK | |
259 | struct list_head cgwb_list; | |
260 | struct wb_domain cgwb_domain; | |
261 | #endif | |
262 | ||
263 | /* List of events which userspace want to receive */ | |
264 | struct list_head event_list; | |
265 | spinlock_t event_list_lock; | |
266 | ||
267 | struct mem_cgroup_per_node *nodeinfo[0]; | |
268 | /* WARNING: nodeinfo must be the last member here */ | |
269 | }; | |
7d828602 JW |
270 | |
271 | extern struct mem_cgroup *root_mem_cgroup; | |
56161634 | 272 | |
23047a96 JW |
273 | static inline bool mem_cgroup_disabled(void) |
274 | { | |
275 | return !cgroup_subsys_enabled(memory_cgrp_subsys); | |
276 | } | |
277 | ||
33398cf2 MH |
278 | /** |
279 | * mem_cgroup_events - count memory events against a cgroup | |
280 | * @memcg: the memory cgroup | |
281 | * @idx: the event index | |
282 | * @nr: the number of events to account for | |
283 | */ | |
284 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, | |
241994ed | 285 | enum mem_cgroup_events_index idx, |
33398cf2 MH |
286 | unsigned int nr) |
287 | { | |
288 | this_cpu_add(memcg->stat->events[idx], nr); | |
472912a2 | 289 | cgroup_file_notify(&memcg->events_file); |
33398cf2 | 290 | } |
241994ed JW |
291 | |
292 | bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg); | |
293 | ||
00501b53 | 294 | int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
f627c2f5 KS |
295 | gfp_t gfp_mask, struct mem_cgroup **memcgp, |
296 | bool compound); | |
00501b53 | 297 | void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg, |
f627c2f5 KS |
298 | bool lrucare, bool compound); |
299 | void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg, | |
300 | bool compound); | |
0a31bc97 | 301 | void mem_cgroup_uncharge(struct page *page); |
747db954 | 302 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
569b846d | 303 | |
6a93ca8f | 304 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage); |
569b846d | 305 | |
0a31bc97 JW |
306 | struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *); |
307 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *); | |
c9b0ed51 | 308 | |
2314b42d | 309 | bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg); |
64219994 | 310 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
e993d905 | 311 | |
33398cf2 MH |
312 | static inline |
313 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ | |
314 | return css ? container_of(css, struct mem_cgroup, css) : NULL; | |
315 | } | |
316 | ||
8e8ae645 JW |
317 | #define mem_cgroup_from_counter(counter, member) \ |
318 | container_of(counter, struct mem_cgroup, member) | |
319 | ||
33398cf2 MH |
320 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
321 | struct mem_cgroup *, | |
322 | struct mem_cgroup_reclaim_cookie *); | |
323 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); | |
324 | ||
23047a96 JW |
325 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
326 | { | |
327 | if (mem_cgroup_disabled()) | |
328 | return 0; | |
329 | ||
330 | return memcg->css.id; | |
331 | } | |
332 | ||
333 | /** | |
334 | * mem_cgroup_from_id - look up a memcg from an id | |
335 | * @id: the id to look up | |
336 | * | |
337 | * Caller must hold rcu_read_lock() and use css_tryget() as necessary. | |
338 | */ | |
339 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
340 | { | |
341 | struct cgroup_subsys_state *css; | |
342 | ||
343 | css = css_from_id(id, &memory_cgrp_subsys); | |
344 | return mem_cgroup_from_css(css); | |
345 | } | |
346 | ||
8e8ae645 JW |
347 | /** |
348 | * parent_mem_cgroup - find the accounting parent of a memcg | |
349 | * @memcg: memcg whose parent to find | |
350 | * | |
351 | * Returns the parent memcg, or NULL if this is the root or the memory | |
352 | * controller is in legacy no-hierarchy mode. | |
353 | */ | |
354 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) | |
355 | { | |
356 | if (!memcg->memory.parent) | |
357 | return NULL; | |
358 | return mem_cgroup_from_counter(memcg->memory.parent, memory); | |
359 | } | |
360 | ||
33398cf2 MH |
361 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
362 | struct mem_cgroup *root) | |
363 | { | |
364 | if (root == memcg) | |
365 | return true; | |
366 | if (!root->use_hierarchy) | |
367 | return false; | |
368 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); | |
369 | } | |
e1aab161 | 370 | |
2314b42d JW |
371 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
372 | struct mem_cgroup *memcg) | |
2e4d4091 | 373 | { |
587af308 | 374 | struct mem_cgroup *task_memcg; |
413918bb | 375 | bool match = false; |
c3ac9a8a | 376 | |
2e4d4091 | 377 | rcu_read_lock(); |
587af308 | 378 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
413918bb | 379 | if (task_memcg) |
2314b42d | 380 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
2e4d4091 | 381 | rcu_read_unlock(); |
c3ac9a8a | 382 | return match; |
2e4d4091 | 383 | } |
8a9f3ccd | 384 | |
64219994 | 385 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
2fc04524 | 386 | ino_t page_cgroup_ino(struct page *page); |
d324236b | 387 | |
eb01aaab VD |
388 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
389 | { | |
390 | if (mem_cgroup_disabled()) | |
391 | return true; | |
392 | return !!(memcg->css.flags & CSS_ONLINE); | |
393 | } | |
394 | ||
58ae83db KH |
395 | /* |
396 | * For memory reclaim. | |
397 | */ | |
889976db | 398 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
33398cf2 MH |
399 | |
400 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
401 | int nr_pages); | |
402 | ||
33398cf2 MH |
403 | static inline |
404 | unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) | |
405 | { | |
406 | struct mem_cgroup_per_zone *mz; | |
407 | ||
408 | mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec); | |
409 | return mz->lru_size[lru]; | |
410 | } | |
411 | ||
13308ca9 | 412 | static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) |
33398cf2 MH |
413 | { |
414 | unsigned long inactive_ratio; | |
415 | unsigned long inactive; | |
416 | unsigned long active; | |
417 | unsigned long gb; | |
418 | ||
419 | inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON); | |
420 | active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON); | |
421 | ||
422 | gb = (inactive + active) >> (30 - PAGE_SHIFT); | |
423 | if (gb) | |
424 | inactive_ratio = int_sqrt(10 * gb); | |
425 | else | |
426 | inactive_ratio = 1; | |
427 | ||
428 | return inactive * inactive_ratio < active; | |
429 | } | |
430 | ||
b23afb93 TH |
431 | void mem_cgroup_handle_over_high(void); |
432 | ||
64219994 MH |
433 | void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, |
434 | struct task_struct *p); | |
58ae83db | 435 | |
49426420 | 436 | static inline void mem_cgroup_oom_enable(void) |
519e5247 | 437 | { |
626ebc41 TH |
438 | WARN_ON(current->memcg_may_oom); |
439 | current->memcg_may_oom = 1; | |
519e5247 JW |
440 | } |
441 | ||
49426420 | 442 | static inline void mem_cgroup_oom_disable(void) |
519e5247 | 443 | { |
626ebc41 TH |
444 | WARN_ON(!current->memcg_may_oom); |
445 | current->memcg_may_oom = 0; | |
519e5247 JW |
446 | } |
447 | ||
3812c8c8 JW |
448 | static inline bool task_in_memcg_oom(struct task_struct *p) |
449 | { | |
626ebc41 | 450 | return p->memcg_in_oom; |
3812c8c8 JW |
451 | } |
452 | ||
49426420 | 453 | bool mem_cgroup_oom_synchronize(bool wait); |
3812c8c8 | 454 | |
c255a458 | 455 | #ifdef CONFIG_MEMCG_SWAP |
c077719b KH |
456 | extern int do_swap_account; |
457 | #endif | |
f8d66542 | 458 | |
62cccb8c JW |
459 | void lock_page_memcg(struct page *page); |
460 | void unlock_page_memcg(struct page *page); | |
d7365e78 | 461 | |
33398cf2 MH |
462 | /** |
463 | * mem_cgroup_update_page_stat - update page state statistics | |
62cccb8c | 464 | * @page: the page |
33398cf2 MH |
465 | * @idx: page state item to account |
466 | * @val: number of pages (positive or negative) | |
467 | * | |
fdf1cdb9 JW |
468 | * The @page must be locked or the caller must use lock_page_memcg() |
469 | * to prevent double accounting when the page is concurrently being | |
470 | * moved to another memcg: | |
81f8c3a4 | 471 | * |
fdf1cdb9 | 472 | * lock_page(page) or lock_page_memcg(page) |
81f8c3a4 | 473 | * if (TestClearPageState(page)) |
62cccb8c | 474 | * mem_cgroup_update_page_stat(page, state, -1); |
fdf1cdb9 | 475 | * unlock_page(page) or unlock_page_memcg(page) |
33398cf2 | 476 | */ |
62cccb8c | 477 | static inline void mem_cgroup_update_page_stat(struct page *page, |
33398cf2 MH |
478 | enum mem_cgroup_stat_index idx, int val) |
479 | { | |
fdf1cdb9 | 480 | VM_BUG_ON(!(rcu_read_lock_held() || PageLocked(page))); |
33398cf2 | 481 | |
62cccb8c JW |
482 | if (page->mem_cgroup) |
483 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); | |
33398cf2 MH |
484 | } |
485 | ||
62cccb8c | 486 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 487 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 488 | { |
62cccb8c | 489 | mem_cgroup_update_page_stat(page, idx, 1); |
2a7106f2 GT |
490 | } |
491 | ||
62cccb8c | 492 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 493 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 494 | { |
62cccb8c | 495 | mem_cgroup_update_page_stat(page, idx, -1); |
2a7106f2 GT |
496 | } |
497 | ||
0608f43d AM |
498 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
499 | gfp_t gfp_mask, | |
500 | unsigned long *total_scanned); | |
a63d83f4 | 501 | |
68ae564b DR |
502 | static inline void mem_cgroup_count_vm_event(struct mm_struct *mm, |
503 | enum vm_event_item idx) | |
504 | { | |
33398cf2 MH |
505 | struct mem_cgroup *memcg; |
506 | ||
68ae564b DR |
507 | if (mem_cgroup_disabled()) |
508 | return; | |
33398cf2 MH |
509 | |
510 | rcu_read_lock(); | |
511 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
512 | if (unlikely(!memcg)) | |
513 | goto out; | |
514 | ||
515 | switch (idx) { | |
516 | case PGFAULT: | |
517 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]); | |
518 | break; | |
519 | case PGMAJFAULT: | |
520 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]); | |
521 | break; | |
522 | default: | |
523 | BUG(); | |
524 | } | |
525 | out: | |
526 | rcu_read_unlock(); | |
68ae564b | 527 | } |
ca3e0214 | 528 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
e94c8a9c | 529 | void mem_cgroup_split_huge_fixup(struct page *head); |
ca3e0214 KH |
530 | #endif |
531 | ||
c255a458 | 532 | #else /* CONFIG_MEMCG */ |
23047a96 JW |
533 | |
534 | #define MEM_CGROUP_ID_SHIFT 0 | |
535 | #define MEM_CGROUP_ID_MAX 0 | |
536 | ||
7a81b88c KH |
537 | struct mem_cgroup; |
538 | ||
23047a96 JW |
539 | static inline bool mem_cgroup_disabled(void) |
540 | { | |
541 | return true; | |
542 | } | |
543 | ||
241994ed JW |
544 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, |
545 | enum mem_cgroup_events_index idx, | |
546 | unsigned int nr) | |
547 | { | |
548 | } | |
549 | ||
550 | static inline bool mem_cgroup_low(struct mem_cgroup *root, | |
551 | struct mem_cgroup *memcg) | |
552 | { | |
553 | return false; | |
554 | } | |
555 | ||
00501b53 JW |
556 | static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
557 | gfp_t gfp_mask, | |
f627c2f5 KS |
558 | struct mem_cgroup **memcgp, |
559 | bool compound) | |
7a81b88c | 560 | { |
00501b53 | 561 | *memcgp = NULL; |
7a81b88c KH |
562 | return 0; |
563 | } | |
564 | ||
00501b53 JW |
565 | static inline void mem_cgroup_commit_charge(struct page *page, |
566 | struct mem_cgroup *memcg, | |
f627c2f5 | 567 | bool lrucare, bool compound) |
7a81b88c KH |
568 | { |
569 | } | |
570 | ||
00501b53 | 571 | static inline void mem_cgroup_cancel_charge(struct page *page, |
f627c2f5 KS |
572 | struct mem_cgroup *memcg, |
573 | bool compound) | |
7a81b88c KH |
574 | { |
575 | } | |
576 | ||
0a31bc97 | 577 | static inline void mem_cgroup_uncharge(struct page *page) |
569b846d KH |
578 | { |
579 | } | |
580 | ||
747db954 | 581 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
8a9f3ccd BS |
582 | { |
583 | } | |
584 | ||
6a93ca8f | 585 | static inline void mem_cgroup_migrate(struct page *old, struct page *new) |
69029cd5 KH |
586 | { |
587 | } | |
588 | ||
925b7673 JW |
589 | static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone, |
590 | struct mem_cgroup *memcg) | |
08e552c6 | 591 | { |
925b7673 | 592 | return &zone->lruvec; |
08e552c6 KH |
593 | } |
594 | ||
fa9add64 HD |
595 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
596 | struct zone *zone) | |
66e1707b | 597 | { |
925b7673 | 598 | return &zone->lruvec; |
66e1707b BS |
599 | } |
600 | ||
587af308 | 601 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
c0ff4b85 | 602 | struct mem_cgroup *memcg) |
bed7161a | 603 | { |
587af308 | 604 | return true; |
bed7161a BS |
605 | } |
606 | ||
ffbdccf5 DR |
607 | static inline bool task_in_mem_cgroup(struct task_struct *task, |
608 | const struct mem_cgroup *memcg) | |
4c4a2214 | 609 | { |
ffbdccf5 | 610 | return true; |
4c4a2214 DR |
611 | } |
612 | ||
5660048c JW |
613 | static inline struct mem_cgroup * |
614 | mem_cgroup_iter(struct mem_cgroup *root, | |
615 | struct mem_cgroup *prev, | |
616 | struct mem_cgroup_reclaim_cookie *reclaim) | |
617 | { | |
618 | return NULL; | |
619 | } | |
620 | ||
621 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, | |
622 | struct mem_cgroup *prev) | |
623 | { | |
624 | } | |
625 | ||
23047a96 | 626 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
f8d66542 | 627 | { |
23047a96 JW |
628 | return 0; |
629 | } | |
630 | ||
631 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
632 | { | |
633 | WARN_ON_ONCE(id); | |
634 | /* XXX: This should always return root_mem_cgroup */ | |
635 | return NULL; | |
f8d66542 | 636 | } |
a636b327 | 637 | |
eb01aaab | 638 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
14797e23 | 639 | { |
13308ca9 | 640 | return true; |
14797e23 KM |
641 | } |
642 | ||
eb01aaab VD |
643 | static inline bool |
644 | mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) | |
90cbc250 VD |
645 | { |
646 | return true; | |
647 | } | |
648 | ||
a3d8e054 | 649 | static inline unsigned long |
4d7dcca2 | 650 | mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
a3d8e054 KM |
651 | { |
652 | return 0; | |
653 | } | |
654 | ||
fa9add64 HD |
655 | static inline void |
656 | mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
657 | int increment) | |
3e2f41f1 | 658 | { |
3e2f41f1 KM |
659 | } |
660 | ||
e222432b BS |
661 | static inline void |
662 | mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p) | |
663 | { | |
664 | } | |
665 | ||
62cccb8c | 666 | static inline void lock_page_memcg(struct page *page) |
89c06bd5 KH |
667 | { |
668 | } | |
669 | ||
62cccb8c | 670 | static inline void unlock_page_memcg(struct page *page) |
89c06bd5 KH |
671 | { |
672 | } | |
673 | ||
b23afb93 TH |
674 | static inline void mem_cgroup_handle_over_high(void) |
675 | { | |
676 | } | |
677 | ||
49426420 | 678 | static inline void mem_cgroup_oom_enable(void) |
519e5247 JW |
679 | { |
680 | } | |
681 | ||
49426420 | 682 | static inline void mem_cgroup_oom_disable(void) |
519e5247 JW |
683 | { |
684 | } | |
685 | ||
3812c8c8 JW |
686 | static inline bool task_in_memcg_oom(struct task_struct *p) |
687 | { | |
688 | return false; | |
689 | } | |
690 | ||
49426420 | 691 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
3812c8c8 JW |
692 | { |
693 | return false; | |
694 | } | |
695 | ||
62cccb8c | 696 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 697 | enum mem_cgroup_stat_index idx) |
2a7106f2 GT |
698 | { |
699 | } | |
700 | ||
62cccb8c | 701 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 702 | enum mem_cgroup_stat_index idx) |
d69b042f BS |
703 | { |
704 | } | |
705 | ||
4e416953 | 706 | static inline |
0608f43d AM |
707 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
708 | gfp_t gfp_mask, | |
709 | unsigned long *total_scanned) | |
4e416953 | 710 | { |
0608f43d | 711 | return 0; |
4e416953 BS |
712 | } |
713 | ||
e94c8a9c | 714 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
ca3e0214 KH |
715 | { |
716 | } | |
717 | ||
456f998e YH |
718 | static inline |
719 | void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx) | |
720 | { | |
721 | } | |
c255a458 | 722 | #endif /* CONFIG_MEMCG */ |
78fb7466 | 723 | |
52ebea74 | 724 | #ifdef CONFIG_CGROUP_WRITEBACK |
841710aa | 725 | |
52ebea74 | 726 | struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); |
841710aa | 727 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
c5edf9cd TH |
728 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
729 | unsigned long *pheadroom, unsigned long *pdirty, | |
730 | unsigned long *pwriteback); | |
841710aa TH |
731 | |
732 | #else /* CONFIG_CGROUP_WRITEBACK */ | |
733 | ||
734 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) | |
735 | { | |
736 | return NULL; | |
737 | } | |
738 | ||
c2aa723a | 739 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
c5edf9cd TH |
740 | unsigned long *pfilepages, |
741 | unsigned long *pheadroom, | |
c2aa723a TH |
742 | unsigned long *pdirty, |
743 | unsigned long *pwriteback) | |
744 | { | |
745 | } | |
746 | ||
841710aa | 747 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
52ebea74 | 748 | |
e1aab161 | 749 | struct sock; |
e1aab161 GC |
750 | void sock_update_memcg(struct sock *sk); |
751 | void sock_release_memcg(struct sock *sk); | |
baac50bb JW |
752 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
753 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); | |
d886f4e4 | 754 | #ifdef CONFIG_MEMCG |
ef12947c JW |
755 | extern struct static_key_false memcg_sockets_enabled_key; |
756 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) | |
baac50bb | 757 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c | 758 | { |
0db15298 | 759 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
8e8ae645 | 760 | return true; |
8e8ae645 JW |
761 | do { |
762 | if (time_before(jiffies, memcg->socket_pressure)) | |
763 | return true; | |
764 | } while ((memcg = parent_mem_cgroup(memcg))); | |
765 | return false; | |
e805605c JW |
766 | } |
767 | #else | |
80e95fe0 | 768 | #define mem_cgroup_sockets_enabled 0 |
baac50bb | 769 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c JW |
770 | { |
771 | return false; | |
772 | } | |
773 | #endif | |
7ae1e1d0 | 774 | |
127424c8 | 775 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
ef12947c | 776 | extern struct static_key_false memcg_kmem_enabled_key; |
749c5415 | 777 | |
dbcf73e2 | 778 | extern int memcg_nr_cache_ids; |
64219994 MH |
779 | void memcg_get_cache_ids(void); |
780 | void memcg_put_cache_ids(void); | |
ebe945c2 GC |
781 | |
782 | /* | |
783 | * Helper macro to loop through all memcg-specific caches. Callers must still | |
784 | * check if the cache is valid (it is either valid or NULL). | |
785 | * the slab_mutex must be held when looping through those caches | |
786 | */ | |
749c5415 | 787 | #define for_each_memcg_cache_index(_idx) \ |
dbcf73e2 | 788 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
749c5415 | 789 | |
7ae1e1d0 GC |
790 | static inline bool memcg_kmem_enabled(void) |
791 | { | |
ef12947c | 792 | return static_branch_unlikely(&memcg_kmem_enabled_key); |
7ae1e1d0 GC |
793 | } |
794 | ||
567e9ab2 | 795 | static inline bool memcg_kmem_online(struct mem_cgroup *memcg) |
33398cf2 | 796 | { |
567e9ab2 | 797 | return memcg->kmem_state == KMEM_ONLINE; |
33398cf2 | 798 | } |
cb731d6c | 799 | |
7ae1e1d0 GC |
800 | /* |
801 | * In general, we'll do everything in our power to not incur in any overhead | |
802 | * for non-memcg users for the kmem functions. Not even a function call, if we | |
803 | * can avoid it. | |
804 | * | |
805 | * Therefore, we'll inline all those functions so that in the best case, we'll | |
806 | * see that kmemcg is off for everybody and proceed quickly. If it is on, | |
807 | * we'll still do most of the flag checking inline. We check a lot of | |
808 | * conditions, but because they are pretty simple, they are expected to be | |
809 | * fast. | |
810 | */ | |
f3ccb2c4 VD |
811 | int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order, |
812 | struct mem_cgroup *memcg); | |
d05e83a6 VD |
813 | int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order); |
814 | void __memcg_kmem_uncharge(struct page *page, int order); | |
7ae1e1d0 | 815 | |
33398cf2 | 816 | /* |
9f706d68 | 817 | * helper for accessing a memcg's index. It will be used as an index in the |
33398cf2 MH |
818 | * child cache array in kmem_cache, and also to derive its name. This function |
819 | * will return -1 when this is not a kmem-limited memcg. | |
820 | */ | |
821 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
822 | { | |
823 | return memcg ? memcg->kmemcg_id : -1; | |
824 | } | |
5722d094 | 825 | |
230e9fc2 | 826 | struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp); |
8135be5a | 827 | void __memcg_kmem_put_cache(struct kmem_cache *cachep); |
d7f25f8a | 828 | |
230e9fc2 | 829 | static inline bool __memcg_kmem_bypass(void) |
7ae1e1d0 GC |
830 | { |
831 | if (!memcg_kmem_enabled()) | |
832 | return true; | |
7ae1e1d0 GC |
833 | if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD)) |
834 | return true; | |
cbfb4798 | 835 | return false; |
7ae1e1d0 GC |
836 | } |
837 | ||
838 | /** | |
d05e83a6 VD |
839 | * memcg_kmem_charge: charge a kmem page |
840 | * @page: page to charge | |
841 | * @gfp: reclaim mode | |
842 | * @order: allocation order | |
cbfb4798 | 843 | * |
d05e83a6 | 844 | * Returns 0 on success, an error code on failure. |
7ae1e1d0 | 845 | */ |
d05e83a6 VD |
846 | static __always_inline int memcg_kmem_charge(struct page *page, |
847 | gfp_t gfp, int order) | |
7ae1e1d0 | 848 | { |
230e9fc2 VD |
849 | if (__memcg_kmem_bypass()) |
850 | return 0; | |
851 | if (!(gfp & __GFP_ACCOUNT)) | |
d05e83a6 VD |
852 | return 0; |
853 | return __memcg_kmem_charge(page, gfp, order); | |
7ae1e1d0 GC |
854 | } |
855 | ||
856 | /** | |
d05e83a6 VD |
857 | * memcg_kmem_uncharge: uncharge a kmem page |
858 | * @page: page to uncharge | |
859 | * @order: allocation order | |
7ae1e1d0 | 860 | */ |
d05e83a6 | 861 | static __always_inline void memcg_kmem_uncharge(struct page *page, int order) |
7ae1e1d0 | 862 | { |
7ae1e1d0 | 863 | if (memcg_kmem_enabled()) |
d05e83a6 | 864 | __memcg_kmem_uncharge(page, order); |
7ae1e1d0 GC |
865 | } |
866 | ||
d7f25f8a GC |
867 | /** |
868 | * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation | |
869 | * @cachep: the original global kmem cache | |
d7f25f8a | 870 | * |
5dfb4175 | 871 | * All memory allocated from a per-memcg cache is charged to the owner memcg. |
d7f25f8a GC |
872 | */ |
873 | static __always_inline struct kmem_cache * | |
874 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
875 | { | |
230e9fc2 | 876 | if (__memcg_kmem_bypass()) |
d7f25f8a | 877 | return cachep; |
230e9fc2 | 878 | return __memcg_kmem_get_cache(cachep, gfp); |
d7f25f8a | 879 | } |
8135be5a VD |
880 | |
881 | static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
882 | { | |
883 | if (memcg_kmem_enabled()) | |
884 | __memcg_kmem_put_cache(cachep); | |
885 | } | |
7ae1e1d0 | 886 | #else |
749c5415 GC |
887 | #define for_each_memcg_cache_index(_idx) \ |
888 | for (; NULL; ) | |
889 | ||
b9ce5ef4 GC |
890 | static inline bool memcg_kmem_enabled(void) |
891 | { | |
892 | return false; | |
893 | } | |
894 | ||
567e9ab2 | 895 | static inline bool memcg_kmem_online(struct mem_cgroup *memcg) |
cb731d6c VD |
896 | { |
897 | return false; | |
898 | } | |
899 | ||
d05e83a6 | 900 | static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order) |
7ae1e1d0 | 901 | { |
d05e83a6 | 902 | return 0; |
7ae1e1d0 GC |
903 | } |
904 | ||
d05e83a6 | 905 | static inline void memcg_kmem_uncharge(struct page *page, int order) |
7ae1e1d0 GC |
906 | { |
907 | } | |
2633d7a0 GC |
908 | |
909 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
910 | { | |
911 | return -1; | |
912 | } | |
913 | ||
05257a1a VD |
914 | static inline void memcg_get_cache_ids(void) |
915 | { | |
916 | } | |
917 | ||
918 | static inline void memcg_put_cache_ids(void) | |
919 | { | |
920 | } | |
921 | ||
d7f25f8a GC |
922 | static inline struct kmem_cache * |
923 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
924 | { | |
925 | return cachep; | |
926 | } | |
8135be5a VD |
927 | |
928 | static inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
929 | { | |
930 | } | |
127424c8 JW |
931 | #endif /* CONFIG_MEMCG && !CONFIG_SLOB */ |
932 | ||
8cdea7c0 | 933 | #endif /* _LINUX_MEMCONTROL_H */ |