]>
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, | |
2a7106f2 GT |
53 | }; |
54 | ||
5660048c JW |
55 | struct mem_cgroup_reclaim_cookie { |
56 | struct zone *zone; | |
57 | int priority; | |
58 | unsigned int generation; | |
59 | }; | |
60 | ||
241994ed JW |
61 | enum mem_cgroup_events_index { |
62 | MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */ | |
63 | MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */ | |
64 | MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */ | |
65 | MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */ | |
66 | MEM_CGROUP_EVENTS_NSTATS, | |
67 | /* default hierarchy events */ | |
68 | MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS, | |
69 | MEMCG_HIGH, | |
70 | MEMCG_MAX, | |
71 | MEMCG_OOM, | |
72 | MEMCG_NR_EVENTS, | |
73 | }; | |
74 | ||
33398cf2 MH |
75 | /* |
76 | * Per memcg event counter is incremented at every pagein/pageout. With THP, | |
77 | * it will be incremated by the number of pages. This counter is used for | |
78 | * for trigger some periodic events. This is straightforward and better | |
79 | * than using jiffies etc. to handle periodic memcg event. | |
80 | */ | |
81 | enum mem_cgroup_events_target { | |
82 | MEM_CGROUP_TARGET_THRESH, | |
83 | MEM_CGROUP_TARGET_SOFTLIMIT, | |
84 | MEM_CGROUP_TARGET_NUMAINFO, | |
85 | MEM_CGROUP_NTARGETS, | |
86 | }; | |
87 | ||
88 | /* | |
89 | * Bits in struct cg_proto.flags | |
90 | */ | |
91 | enum cg_proto_flags { | |
92 | /* Currently active and new sockets should be assigned to cgroups */ | |
93 | MEMCG_SOCK_ACTIVE, | |
94 | /* It was ever activated; we must disarm static keys on destruction */ | |
95 | MEMCG_SOCK_ACTIVATED, | |
96 | }; | |
97 | ||
98 | struct cg_proto { | |
99 | struct page_counter memory_allocated; /* Current allocated memory. */ | |
100 | struct percpu_counter sockets_allocated; /* Current number of sockets. */ | |
101 | int memory_pressure; | |
102 | long sysctl_mem[3]; | |
103 | unsigned long flags; | |
104 | /* | |
105 | * memcg field is used to find which memcg we belong directly | |
106 | * Each memcg struct can hold more than one cg_proto, so container_of | |
107 | * won't really cut. | |
108 | * | |
109 | * The elegant solution would be having an inverse function to | |
110 | * proto_cgroup in struct proto, but that means polluting the structure | |
111 | * for everybody, instead of just for memcg users. | |
112 | */ | |
113 | struct mem_cgroup *memcg; | |
114 | }; | |
115 | ||
c255a458 | 116 | #ifdef CONFIG_MEMCG |
33398cf2 MH |
117 | struct mem_cgroup_stat_cpu { |
118 | long count[MEM_CGROUP_STAT_NSTATS]; | |
119 | unsigned long events[MEMCG_NR_EVENTS]; | |
120 | unsigned long nr_page_events; | |
121 | unsigned long targets[MEM_CGROUP_NTARGETS]; | |
122 | }; | |
123 | ||
124 | struct mem_cgroup_reclaim_iter { | |
125 | struct mem_cgroup *position; | |
126 | /* scan generation, increased every round-trip */ | |
127 | unsigned int generation; | |
128 | }; | |
129 | ||
130 | /* | |
131 | * per-zone information in memory controller. | |
132 | */ | |
133 | struct mem_cgroup_per_zone { | |
134 | struct lruvec lruvec; | |
135 | unsigned long lru_size[NR_LRU_LISTS]; | |
136 | ||
137 | struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1]; | |
138 | ||
139 | struct rb_node tree_node; /* RB tree node */ | |
140 | unsigned long usage_in_excess;/* Set to the value by which */ | |
141 | /* the soft limit is exceeded*/ | |
142 | bool on_tree; | |
143 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ | |
144 | /* use container_of */ | |
145 | }; | |
146 | ||
147 | struct mem_cgroup_per_node { | |
148 | struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; | |
149 | }; | |
150 | ||
151 | struct mem_cgroup_threshold { | |
152 | struct eventfd_ctx *eventfd; | |
153 | unsigned long threshold; | |
154 | }; | |
155 | ||
156 | /* For threshold */ | |
157 | struct mem_cgroup_threshold_ary { | |
158 | /* An array index points to threshold just below or equal to usage. */ | |
159 | int current_threshold; | |
160 | /* Size of entries[] */ | |
161 | unsigned int size; | |
162 | /* Array of thresholds */ | |
163 | struct mem_cgroup_threshold entries[0]; | |
164 | }; | |
165 | ||
166 | struct mem_cgroup_thresholds { | |
167 | /* Primary thresholds array */ | |
168 | struct mem_cgroup_threshold_ary *primary; | |
169 | /* | |
170 | * Spare threshold array. | |
171 | * This is needed to make mem_cgroup_unregister_event() "never fail". | |
172 | * It must be able to store at least primary->size - 1 entries. | |
173 | */ | |
174 | struct mem_cgroup_threshold_ary *spare; | |
175 | }; | |
176 | ||
177 | /* | |
178 | * The memory controller data structure. The memory controller controls both | |
179 | * page cache and RSS per cgroup. We would eventually like to provide | |
180 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, | |
181 | * to help the administrator determine what knobs to tune. | |
182 | */ | |
183 | struct mem_cgroup { | |
184 | struct cgroup_subsys_state css; | |
185 | ||
186 | /* Accounted resources */ | |
187 | struct page_counter memory; | |
188 | struct page_counter memsw; | |
189 | struct page_counter kmem; | |
190 | ||
191 | /* Normal memory consumption range */ | |
192 | unsigned long low; | |
193 | unsigned long high; | |
194 | ||
195 | unsigned long soft_limit; | |
196 | ||
197 | /* vmpressure notifications */ | |
198 | struct vmpressure vmpressure; | |
199 | ||
200 | /* css_online() has been completed */ | |
201 | int initialized; | |
202 | ||
203 | /* | |
204 | * Should the accounting and control be hierarchical, per subtree? | |
205 | */ | |
206 | bool use_hierarchy; | |
207 | ||
208 | /* protected by memcg_oom_lock */ | |
209 | bool oom_lock; | |
210 | int under_oom; | |
211 | ||
212 | int swappiness; | |
213 | /* OOM-Killer disable */ | |
214 | int oom_kill_disable; | |
215 | ||
216 | /* protect arrays of thresholds */ | |
217 | struct mutex thresholds_lock; | |
218 | ||
219 | /* thresholds for memory usage. RCU-protected */ | |
220 | struct mem_cgroup_thresholds thresholds; | |
221 | ||
222 | /* thresholds for mem+swap usage. RCU-protected */ | |
223 | struct mem_cgroup_thresholds memsw_thresholds; | |
224 | ||
225 | /* For oom notifier event fd */ | |
226 | struct list_head oom_notify; | |
227 | ||
228 | /* | |
229 | * Should we move charges of a task when a task is moved into this | |
230 | * mem_cgroup ? And what type of charges should we move ? | |
231 | */ | |
232 | unsigned long move_charge_at_immigrate; | |
233 | /* | |
234 | * set > 0 if pages under this cgroup are moving to other cgroup. | |
235 | */ | |
236 | atomic_t moving_account; | |
237 | /* taken only while moving_account > 0 */ | |
238 | spinlock_t move_lock; | |
239 | struct task_struct *move_lock_task; | |
240 | unsigned long move_lock_flags; | |
241 | /* | |
242 | * percpu counter. | |
243 | */ | |
244 | struct mem_cgroup_stat_cpu __percpu *stat; | |
245 | spinlock_t pcp_counter_lock; | |
246 | ||
247 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET) | |
248 | struct cg_proto tcp_mem; | |
249 | #endif | |
250 | #if defined(CONFIG_MEMCG_KMEM) | |
251 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ | |
252 | int kmemcg_id; | |
253 | bool kmem_acct_activated; | |
254 | bool kmem_acct_active; | |
255 | #endif | |
256 | ||
257 | int last_scanned_node; | |
258 | #if MAX_NUMNODES > 1 | |
259 | nodemask_t scan_nodes; | |
260 | atomic_t numainfo_events; | |
261 | atomic_t numainfo_updating; | |
262 | #endif | |
263 | ||
264 | #ifdef CONFIG_CGROUP_WRITEBACK | |
265 | struct list_head cgwb_list; | |
266 | struct wb_domain cgwb_domain; | |
267 | #endif | |
268 | ||
269 | /* List of events which userspace want to receive */ | |
270 | struct list_head event_list; | |
271 | spinlock_t event_list_lock; | |
272 | ||
273 | struct mem_cgroup_per_node *nodeinfo[0]; | |
274 | /* WARNING: nodeinfo must be the last member here */ | |
275 | }; | |
56161634 TH |
276 | extern struct cgroup_subsys_state *mem_cgroup_root_css; |
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); | |
289 | } | |
241994ed JW |
290 | |
291 | bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg); | |
292 | ||
00501b53 JW |
293 | int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
294 | gfp_t gfp_mask, struct mem_cgroup **memcgp); | |
295 | void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg, | |
296 | bool lrucare); | |
297 | void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg); | |
0a31bc97 | 298 | void mem_cgroup_uncharge(struct page *page); |
747db954 | 299 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
569b846d | 300 | |
0a31bc97 JW |
301 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage, |
302 | bool lrucare); | |
569b846d | 303 | |
0a31bc97 JW |
304 | struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *); |
305 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *); | |
c9b0ed51 | 306 | |
2314b42d | 307 | bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg); |
64219994 | 308 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
64219994 | 309 | struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg); |
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 | ||
316 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, | |
317 | struct mem_cgroup *, | |
318 | struct mem_cgroup_reclaim_cookie *); | |
319 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); | |
320 | ||
321 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, | |
322 | struct mem_cgroup *root) | |
323 | { | |
324 | if (root == memcg) | |
325 | return true; | |
326 | if (!root->use_hierarchy) | |
327 | return false; | |
328 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); | |
329 | } | |
e1aab161 | 330 | |
2314b42d JW |
331 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
332 | struct mem_cgroup *memcg) | |
2e4d4091 | 333 | { |
587af308 | 334 | struct mem_cgroup *task_memcg; |
413918bb | 335 | bool match = false; |
c3ac9a8a | 336 | |
2e4d4091 | 337 | rcu_read_lock(); |
587af308 | 338 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
413918bb | 339 | if (task_memcg) |
2314b42d | 340 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
2e4d4091 | 341 | rcu_read_unlock(); |
c3ac9a8a | 342 | return match; |
2e4d4091 | 343 | } |
8a9f3ccd | 344 | |
64219994 | 345 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
2fc04524 | 346 | ino_t page_cgroup_ino(struct page *page); |
d324236b | 347 | |
33398cf2 MH |
348 | static inline bool mem_cgroup_disabled(void) |
349 | { | |
fc5ed1e9 | 350 | return !cgroup_subsys_enabled(memory_cgrp_subsys); |
33398cf2 | 351 | } |
5660048c | 352 | |
58ae83db KH |
353 | /* |
354 | * For memory reclaim. | |
355 | */ | |
889976db | 356 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
33398cf2 MH |
357 | |
358 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
359 | int nr_pages); | |
360 | ||
361 | static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec) | |
362 | { | |
363 | struct mem_cgroup_per_zone *mz; | |
364 | struct mem_cgroup *memcg; | |
365 | ||
366 | if (mem_cgroup_disabled()) | |
367 | return true; | |
368 | ||
369 | mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec); | |
370 | memcg = mz->memcg; | |
371 | ||
372 | return !!(memcg->css.flags & CSS_ONLINE); | |
373 | } | |
374 | ||
375 | static inline | |
376 | unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) | |
377 | { | |
378 | struct mem_cgroup_per_zone *mz; | |
379 | ||
380 | mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec); | |
381 | return mz->lru_size[lru]; | |
382 | } | |
383 | ||
384 | static inline int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) | |
385 | { | |
386 | unsigned long inactive_ratio; | |
387 | unsigned long inactive; | |
388 | unsigned long active; | |
389 | unsigned long gb; | |
390 | ||
391 | inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON); | |
392 | active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON); | |
393 | ||
394 | gb = (inactive + active) >> (30 - PAGE_SHIFT); | |
395 | if (gb) | |
396 | inactive_ratio = int_sqrt(10 * gb); | |
397 | else | |
398 | inactive_ratio = 1; | |
399 | ||
400 | return inactive * inactive_ratio < active; | |
401 | } | |
402 | ||
64219994 MH |
403 | void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, |
404 | struct task_struct *p); | |
58ae83db | 405 | |
49426420 | 406 | static inline void mem_cgroup_oom_enable(void) |
519e5247 | 407 | { |
49426420 JW |
408 | WARN_ON(current->memcg_oom.may_oom); |
409 | current->memcg_oom.may_oom = 1; | |
519e5247 JW |
410 | } |
411 | ||
49426420 | 412 | static inline void mem_cgroup_oom_disable(void) |
519e5247 | 413 | { |
49426420 JW |
414 | WARN_ON(!current->memcg_oom.may_oom); |
415 | current->memcg_oom.may_oom = 0; | |
519e5247 JW |
416 | } |
417 | ||
3812c8c8 JW |
418 | static inline bool task_in_memcg_oom(struct task_struct *p) |
419 | { | |
49426420 | 420 | return p->memcg_oom.memcg; |
3812c8c8 JW |
421 | } |
422 | ||
49426420 | 423 | bool mem_cgroup_oom_synchronize(bool wait); |
3812c8c8 | 424 | |
c255a458 | 425 | #ifdef CONFIG_MEMCG_SWAP |
c077719b KH |
426 | extern int do_swap_account; |
427 | #endif | |
f8d66542 | 428 | |
6de22619 | 429 | struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page); |
6de22619 | 430 | void mem_cgroup_end_page_stat(struct mem_cgroup *memcg); |
d7365e78 | 431 | |
33398cf2 MH |
432 | /** |
433 | * mem_cgroup_update_page_stat - update page state statistics | |
434 | * @memcg: memcg to account against | |
435 | * @idx: page state item to account | |
436 | * @val: number of pages (positive or negative) | |
437 | * | |
438 | * See mem_cgroup_begin_page_stat() for locking requirements. | |
439 | */ | |
440 | static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg, | |
441 | enum mem_cgroup_stat_index idx, int val) | |
442 | { | |
443 | VM_BUG_ON(!rcu_read_lock_held()); | |
444 | ||
445 | if (memcg) | |
446 | this_cpu_add(memcg->stat->count[idx], val); | |
447 | } | |
448 | ||
d7365e78 | 449 | static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg, |
68b4876d | 450 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 451 | { |
d7365e78 | 452 | mem_cgroup_update_page_stat(memcg, idx, 1); |
2a7106f2 GT |
453 | } |
454 | ||
d7365e78 | 455 | static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg, |
68b4876d | 456 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 457 | { |
d7365e78 | 458 | mem_cgroup_update_page_stat(memcg, idx, -1); |
2a7106f2 GT |
459 | } |
460 | ||
0608f43d AM |
461 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
462 | gfp_t gfp_mask, | |
463 | unsigned long *total_scanned); | |
a63d83f4 | 464 | |
68ae564b DR |
465 | static inline void mem_cgroup_count_vm_event(struct mm_struct *mm, |
466 | enum vm_event_item idx) | |
467 | { | |
33398cf2 MH |
468 | struct mem_cgroup *memcg; |
469 | ||
68ae564b DR |
470 | if (mem_cgroup_disabled()) |
471 | return; | |
33398cf2 MH |
472 | |
473 | rcu_read_lock(); | |
474 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
475 | if (unlikely(!memcg)) | |
476 | goto out; | |
477 | ||
478 | switch (idx) { | |
479 | case PGFAULT: | |
480 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]); | |
481 | break; | |
482 | case PGMAJFAULT: | |
483 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]); | |
484 | break; | |
485 | default: | |
486 | BUG(); | |
487 | } | |
488 | out: | |
489 | rcu_read_unlock(); | |
68ae564b | 490 | } |
ca3e0214 | 491 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
e94c8a9c | 492 | void mem_cgroup_split_huge_fixup(struct page *head); |
ca3e0214 KH |
493 | #endif |
494 | ||
c255a458 | 495 | #else /* CONFIG_MEMCG */ |
7a81b88c KH |
496 | struct mem_cgroup; |
497 | ||
241994ed JW |
498 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, |
499 | enum mem_cgroup_events_index idx, | |
500 | unsigned int nr) | |
501 | { | |
502 | } | |
503 | ||
504 | static inline bool mem_cgroup_low(struct mem_cgroup *root, | |
505 | struct mem_cgroup *memcg) | |
506 | { | |
507 | return false; | |
508 | } | |
509 | ||
00501b53 JW |
510 | static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
511 | gfp_t gfp_mask, | |
512 | struct mem_cgroup **memcgp) | |
7a81b88c | 513 | { |
00501b53 | 514 | *memcgp = NULL; |
7a81b88c KH |
515 | return 0; |
516 | } | |
517 | ||
00501b53 JW |
518 | static inline void mem_cgroup_commit_charge(struct page *page, |
519 | struct mem_cgroup *memcg, | |
520 | bool lrucare) | |
7a81b88c KH |
521 | { |
522 | } | |
523 | ||
00501b53 JW |
524 | static inline void mem_cgroup_cancel_charge(struct page *page, |
525 | struct mem_cgroup *memcg) | |
7a81b88c KH |
526 | { |
527 | } | |
528 | ||
0a31bc97 | 529 | static inline void mem_cgroup_uncharge(struct page *page) |
569b846d KH |
530 | { |
531 | } | |
532 | ||
747db954 | 533 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
8a9f3ccd BS |
534 | { |
535 | } | |
536 | ||
0a31bc97 JW |
537 | static inline void mem_cgroup_migrate(struct page *oldpage, |
538 | struct page *newpage, | |
539 | bool lrucare) | |
69029cd5 KH |
540 | { |
541 | } | |
542 | ||
925b7673 JW |
543 | static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone, |
544 | struct mem_cgroup *memcg) | |
08e552c6 | 545 | { |
925b7673 | 546 | return &zone->lruvec; |
08e552c6 KH |
547 | } |
548 | ||
fa9add64 HD |
549 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
550 | struct zone *zone) | |
66e1707b | 551 | { |
925b7673 | 552 | return &zone->lruvec; |
66e1707b BS |
553 | } |
554 | ||
587af308 | 555 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
c0ff4b85 | 556 | struct mem_cgroup *memcg) |
bed7161a | 557 | { |
587af308 | 558 | return true; |
bed7161a BS |
559 | } |
560 | ||
ffbdccf5 DR |
561 | static inline bool task_in_mem_cgroup(struct task_struct *task, |
562 | const struct mem_cgroup *memcg) | |
4c4a2214 | 563 | { |
ffbdccf5 | 564 | return true; |
4c4a2214 DR |
565 | } |
566 | ||
5660048c JW |
567 | static inline struct mem_cgroup * |
568 | mem_cgroup_iter(struct mem_cgroup *root, | |
569 | struct mem_cgroup *prev, | |
570 | struct mem_cgroup_reclaim_cookie *reclaim) | |
571 | { | |
572 | return NULL; | |
573 | } | |
574 | ||
575 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, | |
576 | struct mem_cgroup *prev) | |
577 | { | |
578 | } | |
579 | ||
f8d66542 HT |
580 | static inline bool mem_cgroup_disabled(void) |
581 | { | |
582 | return true; | |
583 | } | |
a636b327 | 584 | |
14797e23 | 585 | static inline int |
c56d5c7d | 586 | mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) |
14797e23 KM |
587 | { |
588 | return 1; | |
589 | } | |
590 | ||
90cbc250 VD |
591 | static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec) |
592 | { | |
593 | return true; | |
594 | } | |
595 | ||
a3d8e054 | 596 | static inline unsigned long |
4d7dcca2 | 597 | mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
a3d8e054 KM |
598 | { |
599 | return 0; | |
600 | } | |
601 | ||
fa9add64 HD |
602 | static inline void |
603 | mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
604 | int increment) | |
3e2f41f1 | 605 | { |
3e2f41f1 KM |
606 | } |
607 | ||
e222432b BS |
608 | static inline void |
609 | mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p) | |
610 | { | |
611 | } | |
612 | ||
6de22619 | 613 | static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page) |
89c06bd5 | 614 | { |
d7365e78 | 615 | return NULL; |
89c06bd5 KH |
616 | } |
617 | ||
6de22619 | 618 | static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg) |
89c06bd5 KH |
619 | { |
620 | } | |
621 | ||
49426420 | 622 | static inline void mem_cgroup_oom_enable(void) |
519e5247 JW |
623 | { |
624 | } | |
625 | ||
49426420 | 626 | static inline void mem_cgroup_oom_disable(void) |
519e5247 JW |
627 | { |
628 | } | |
629 | ||
3812c8c8 JW |
630 | static inline bool task_in_memcg_oom(struct task_struct *p) |
631 | { | |
632 | return false; | |
633 | } | |
634 | ||
49426420 | 635 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
3812c8c8 JW |
636 | { |
637 | return false; | |
638 | } | |
639 | ||
d7365e78 | 640 | static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg, |
68b4876d | 641 | enum mem_cgroup_stat_index idx) |
2a7106f2 GT |
642 | { |
643 | } | |
644 | ||
d7365e78 | 645 | static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg, |
68b4876d | 646 | enum mem_cgroup_stat_index idx) |
d69b042f BS |
647 | { |
648 | } | |
649 | ||
4e416953 | 650 | static inline |
0608f43d AM |
651 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
652 | gfp_t gfp_mask, | |
653 | unsigned long *total_scanned) | |
4e416953 | 654 | { |
0608f43d | 655 | return 0; |
4e416953 BS |
656 | } |
657 | ||
e94c8a9c | 658 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
ca3e0214 KH |
659 | { |
660 | } | |
661 | ||
456f998e YH |
662 | static inline |
663 | void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx) | |
664 | { | |
665 | } | |
c255a458 | 666 | #endif /* CONFIG_MEMCG */ |
78fb7466 | 667 | |
e1aab161 GC |
668 | enum { |
669 | UNDER_LIMIT, | |
670 | SOFT_LIMIT, | |
671 | OVER_LIMIT, | |
672 | }; | |
673 | ||
52ebea74 | 674 | #ifdef CONFIG_CGROUP_WRITEBACK |
841710aa | 675 | |
52ebea74 | 676 | struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); |
841710aa | 677 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
c2aa723a TH |
678 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pavail, |
679 | unsigned long *pdirty, unsigned long *pwriteback); | |
841710aa TH |
680 | |
681 | #else /* CONFIG_CGROUP_WRITEBACK */ | |
682 | ||
683 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) | |
684 | { | |
685 | return NULL; | |
686 | } | |
687 | ||
c2aa723a TH |
688 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
689 | unsigned long *pavail, | |
690 | unsigned long *pdirty, | |
691 | unsigned long *pwriteback) | |
692 | { | |
693 | } | |
694 | ||
841710aa | 695 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
52ebea74 | 696 | |
e1aab161 | 697 | struct sock; |
cd59085a | 698 | #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM) |
e1aab161 GC |
699 | void sock_update_memcg(struct sock *sk); |
700 | void sock_release_memcg(struct sock *sk); | |
701 | #else | |
702 | static inline void sock_update_memcg(struct sock *sk) | |
703 | { | |
704 | } | |
705 | static inline void sock_release_memcg(struct sock *sk) | |
706 | { | |
707 | } | |
cd59085a | 708 | #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */ |
7ae1e1d0 GC |
709 | |
710 | #ifdef CONFIG_MEMCG_KMEM | |
a8964b9b | 711 | extern struct static_key memcg_kmem_enabled_key; |
749c5415 | 712 | |
dbcf73e2 | 713 | extern int memcg_nr_cache_ids; |
64219994 MH |
714 | void memcg_get_cache_ids(void); |
715 | void memcg_put_cache_ids(void); | |
ebe945c2 GC |
716 | |
717 | /* | |
718 | * Helper macro to loop through all memcg-specific caches. Callers must still | |
719 | * check if the cache is valid (it is either valid or NULL). | |
720 | * the slab_mutex must be held when looping through those caches | |
721 | */ | |
749c5415 | 722 | #define for_each_memcg_cache_index(_idx) \ |
dbcf73e2 | 723 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
749c5415 | 724 | |
7ae1e1d0 GC |
725 | static inline bool memcg_kmem_enabled(void) |
726 | { | |
a8964b9b | 727 | return static_key_false(&memcg_kmem_enabled_key); |
7ae1e1d0 GC |
728 | } |
729 | ||
33398cf2 MH |
730 | static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg) |
731 | { | |
732 | return memcg->kmem_acct_active; | |
733 | } | |
cb731d6c | 734 | |
7ae1e1d0 GC |
735 | /* |
736 | * In general, we'll do everything in our power to not incur in any overhead | |
737 | * for non-memcg users for the kmem functions. Not even a function call, if we | |
738 | * can avoid it. | |
739 | * | |
740 | * Therefore, we'll inline all those functions so that in the best case, we'll | |
741 | * see that kmemcg is off for everybody and proceed quickly. If it is on, | |
742 | * we'll still do most of the flag checking inline. We check a lot of | |
743 | * conditions, but because they are pretty simple, they are expected to be | |
744 | * fast. | |
745 | */ | |
746 | bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, | |
747 | int order); | |
748 | void __memcg_kmem_commit_charge(struct page *page, | |
749 | struct mem_cgroup *memcg, int order); | |
750 | void __memcg_kmem_uncharge_pages(struct page *page, int order); | |
751 | ||
33398cf2 MH |
752 | /* |
753 | * helper for acessing a memcg's index. It will be used as an index in the | |
754 | * child cache array in kmem_cache, and also to derive its name. This function | |
755 | * will return -1 when this is not a kmem-limited memcg. | |
756 | */ | |
757 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
758 | { | |
759 | return memcg ? memcg->kmemcg_id : -1; | |
760 | } | |
5722d094 | 761 | |
8135be5a VD |
762 | struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep); |
763 | void __memcg_kmem_put_cache(struct kmem_cache *cachep); | |
d7f25f8a | 764 | |
60d3fd32 VD |
765 | struct mem_cgroup *__mem_cgroup_from_kmem(void *ptr); |
766 | ||
dbf22eb6 VD |
767 | int memcg_charge_kmem(struct mem_cgroup *memcg, gfp_t gfp, |
768 | unsigned long nr_pages); | |
769 | void memcg_uncharge_kmem(struct mem_cgroup *memcg, unsigned long nr_pages); | |
5dfb4175 | 770 | |
7ae1e1d0 GC |
771 | /** |
772 | * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed. | |
773 | * @gfp: the gfp allocation flags. | |
774 | * @memcg: a pointer to the memcg this was charged against. | |
775 | * @order: allocation order. | |
776 | * | |
777 | * returns true if the memcg where the current task belongs can hold this | |
778 | * allocation. | |
779 | * | |
780 | * We return true automatically if this allocation is not to be accounted to | |
781 | * any memcg. | |
782 | */ | |
783 | static inline bool | |
784 | memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order) | |
785 | { | |
786 | if (!memcg_kmem_enabled()) | |
787 | return true; | |
788 | ||
8f4fc071 VD |
789 | if (gfp & __GFP_NOACCOUNT) |
790 | return true; | |
7ae1e1d0 GC |
791 | /* |
792 | * __GFP_NOFAIL allocations will move on even if charging is not | |
793 | * possible. Therefore we don't even try, and have this allocation | |
3e32cb2e JW |
794 | * unaccounted. We could in theory charge it forcibly, but we hope |
795 | * those allocations are rare, and won't be worth the trouble. | |
7ae1e1d0 | 796 | */ |
52383431 | 797 | if (gfp & __GFP_NOFAIL) |
7ae1e1d0 GC |
798 | return true; |
799 | if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD)) | |
800 | return true; | |
801 | ||
802 | /* If the test is dying, just let it go. */ | |
803 | if (unlikely(fatal_signal_pending(current))) | |
804 | return true; | |
805 | ||
806 | return __memcg_kmem_newpage_charge(gfp, memcg, order); | |
807 | } | |
808 | ||
809 | /** | |
810 | * memcg_kmem_uncharge_pages: uncharge pages from memcg | |
811 | * @page: pointer to struct page being freed | |
812 | * @order: allocation order. | |
7ae1e1d0 GC |
813 | */ |
814 | static inline void | |
815 | memcg_kmem_uncharge_pages(struct page *page, int order) | |
816 | { | |
817 | if (memcg_kmem_enabled()) | |
818 | __memcg_kmem_uncharge_pages(page, order); | |
819 | } | |
820 | ||
821 | /** | |
822 | * memcg_kmem_commit_charge: embeds correct memcg in a page | |
823 | * @page: pointer to struct page recently allocated | |
824 | * @memcg: the memcg structure we charged against | |
825 | * @order: allocation order. | |
826 | * | |
827 | * Needs to be called after memcg_kmem_newpage_charge, regardless of success or | |
828 | * failure of the allocation. if @page is NULL, this function will revert the | |
1306a85a | 829 | * charges. Otherwise, it will commit @page to @memcg. |
7ae1e1d0 GC |
830 | */ |
831 | static inline void | |
832 | memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order) | |
833 | { | |
834 | if (memcg_kmem_enabled() && memcg) | |
835 | __memcg_kmem_commit_charge(page, memcg, order); | |
836 | } | |
837 | ||
d7f25f8a GC |
838 | /** |
839 | * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation | |
840 | * @cachep: the original global kmem cache | |
841 | * @gfp: allocation flags. | |
842 | * | |
5dfb4175 | 843 | * All memory allocated from a per-memcg cache is charged to the owner memcg. |
d7f25f8a GC |
844 | */ |
845 | static __always_inline struct kmem_cache * | |
846 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
847 | { | |
848 | if (!memcg_kmem_enabled()) | |
849 | return cachep; | |
8f4fc071 VD |
850 | if (gfp & __GFP_NOACCOUNT) |
851 | return cachep; | |
d7f25f8a GC |
852 | if (gfp & __GFP_NOFAIL) |
853 | return cachep; | |
854 | if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD)) | |
855 | return cachep; | |
856 | if (unlikely(fatal_signal_pending(current))) | |
857 | return cachep; | |
858 | ||
056b7cce | 859 | return __memcg_kmem_get_cache(cachep); |
d7f25f8a | 860 | } |
8135be5a VD |
861 | |
862 | static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
863 | { | |
864 | if (memcg_kmem_enabled()) | |
865 | __memcg_kmem_put_cache(cachep); | |
866 | } | |
60d3fd32 VD |
867 | |
868 | static __always_inline struct mem_cgroup *mem_cgroup_from_kmem(void *ptr) | |
869 | { | |
870 | if (!memcg_kmem_enabled()) | |
871 | return NULL; | |
872 | return __mem_cgroup_from_kmem(ptr); | |
873 | } | |
7ae1e1d0 | 874 | #else |
749c5415 GC |
875 | #define for_each_memcg_cache_index(_idx) \ |
876 | for (; NULL; ) | |
877 | ||
b9ce5ef4 GC |
878 | static inline bool memcg_kmem_enabled(void) |
879 | { | |
880 | return false; | |
881 | } | |
882 | ||
cb731d6c VD |
883 | static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg) |
884 | { | |
885 | return false; | |
886 | } | |
887 | ||
7ae1e1d0 GC |
888 | static inline bool |
889 | memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order) | |
890 | { | |
891 | return true; | |
892 | } | |
893 | ||
894 | static inline void memcg_kmem_uncharge_pages(struct page *page, int order) | |
895 | { | |
896 | } | |
897 | ||
898 | static inline void | |
899 | memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order) | |
900 | { | |
901 | } | |
2633d7a0 GC |
902 | |
903 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
904 | { | |
905 | return -1; | |
906 | } | |
907 | ||
05257a1a VD |
908 | static inline void memcg_get_cache_ids(void) |
909 | { | |
910 | } | |
911 | ||
912 | static inline void memcg_put_cache_ids(void) | |
913 | { | |
914 | } | |
915 | ||
d7f25f8a GC |
916 | static inline struct kmem_cache * |
917 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
918 | { | |
919 | return cachep; | |
920 | } | |
8135be5a VD |
921 | |
922 | static inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
923 | { | |
924 | } | |
60d3fd32 VD |
925 | |
926 | static inline struct mem_cgroup *mem_cgroup_from_kmem(void *ptr) | |
927 | { | |
928 | return NULL; | |
929 | } | |
7ae1e1d0 | 930 | #endif /* CONFIG_MEMCG_KMEM */ |
8cdea7c0 BS |
931 | #endif /* _LINUX_MEMCONTROL_H */ |
932 |