]>
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 */ |
12580e4b | 55 | MEMCG_KERNEL_STACK = MEM_CGROUP_STAT_NSTATS, |
27ee57c9 VD |
56 | MEMCG_SLAB_RECLAIMABLE, |
57 | MEMCG_SLAB_UNRECLAIMABLE, | |
12580e4b | 58 | MEMCG_SOCK, |
b2807f07 | 59 | MEMCG_NR_STAT, |
2a7106f2 GT |
60 | }; |
61 | ||
5660048c JW |
62 | struct mem_cgroup_reclaim_cookie { |
63 | struct zone *zone; | |
64 | int priority; | |
65 | unsigned int generation; | |
66 | }; | |
67 | ||
241994ed JW |
68 | enum mem_cgroup_events_index { |
69 | MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */ | |
70 | MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */ | |
71 | MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */ | |
72 | MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */ | |
73 | MEM_CGROUP_EVENTS_NSTATS, | |
74 | /* default hierarchy events */ | |
75 | MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS, | |
76 | MEMCG_HIGH, | |
77 | MEMCG_MAX, | |
78 | MEMCG_OOM, | |
79 | MEMCG_NR_EVENTS, | |
80 | }; | |
81 | ||
33398cf2 MH |
82 | /* |
83 | * Per memcg event counter is incremented at every pagein/pageout. With THP, | |
84 | * it will be incremated by the number of pages. This counter is used for | |
85 | * for trigger some periodic events. This is straightforward and better | |
86 | * than using jiffies etc. to handle periodic memcg event. | |
87 | */ | |
88 | enum mem_cgroup_events_target { | |
89 | MEM_CGROUP_TARGET_THRESH, | |
90 | MEM_CGROUP_TARGET_SOFTLIMIT, | |
91 | MEM_CGROUP_TARGET_NUMAINFO, | |
92 | MEM_CGROUP_NTARGETS, | |
93 | }; | |
94 | ||
c255a458 | 95 | #ifdef CONFIG_MEMCG |
23047a96 JW |
96 | |
97 | #define MEM_CGROUP_ID_SHIFT 16 | |
98 | #define MEM_CGROUP_ID_MAX USHRT_MAX | |
99 | ||
73f576c0 JW |
100 | struct mem_cgroup_id { |
101 | int id; | |
102 | atomic_t ref; | |
103 | }; | |
104 | ||
33398cf2 | 105 | struct mem_cgroup_stat_cpu { |
b2807f07 | 106 | long count[MEMCG_NR_STAT]; |
33398cf2 MH |
107 | unsigned long events[MEMCG_NR_EVENTS]; |
108 | unsigned long nr_page_events; | |
109 | unsigned long targets[MEM_CGROUP_NTARGETS]; | |
110 | }; | |
111 | ||
112 | struct mem_cgroup_reclaim_iter { | |
113 | struct mem_cgroup *position; | |
114 | /* scan generation, increased every round-trip */ | |
115 | unsigned int generation; | |
116 | }; | |
117 | ||
118 | /* | |
119 | * per-zone information in memory controller. | |
120 | */ | |
121 | struct mem_cgroup_per_zone { | |
122 | struct lruvec lruvec; | |
123 | unsigned long lru_size[NR_LRU_LISTS]; | |
124 | ||
125 | struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1]; | |
126 | ||
127 | struct rb_node tree_node; /* RB tree node */ | |
128 | unsigned long usage_in_excess;/* Set to the value by which */ | |
129 | /* the soft limit is exceeded*/ | |
130 | bool on_tree; | |
131 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ | |
132 | /* use container_of */ | |
133 | }; | |
134 | ||
135 | struct mem_cgroup_per_node { | |
136 | struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; | |
137 | }; | |
138 | ||
139 | struct mem_cgroup_threshold { | |
140 | struct eventfd_ctx *eventfd; | |
141 | unsigned long threshold; | |
142 | }; | |
143 | ||
144 | /* For threshold */ | |
145 | struct mem_cgroup_threshold_ary { | |
146 | /* An array index points to threshold just below or equal to usage. */ | |
147 | int current_threshold; | |
148 | /* Size of entries[] */ | |
149 | unsigned int size; | |
150 | /* Array of thresholds */ | |
151 | struct mem_cgroup_threshold entries[0]; | |
152 | }; | |
153 | ||
154 | struct mem_cgroup_thresholds { | |
155 | /* Primary thresholds array */ | |
156 | struct mem_cgroup_threshold_ary *primary; | |
157 | /* | |
158 | * Spare threshold array. | |
159 | * This is needed to make mem_cgroup_unregister_event() "never fail". | |
160 | * It must be able to store at least primary->size - 1 entries. | |
161 | */ | |
162 | struct mem_cgroup_threshold_ary *spare; | |
163 | }; | |
164 | ||
567e9ab2 JW |
165 | enum memcg_kmem_state { |
166 | KMEM_NONE, | |
167 | KMEM_ALLOCATED, | |
168 | KMEM_ONLINE, | |
169 | }; | |
170 | ||
33398cf2 MH |
171 | /* |
172 | * The memory controller data structure. The memory controller controls both | |
173 | * page cache and RSS per cgroup. We would eventually like to provide | |
174 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, | |
175 | * to help the administrator determine what knobs to tune. | |
176 | */ | |
177 | struct mem_cgroup { | |
178 | struct cgroup_subsys_state css; | |
179 | ||
73f576c0 JW |
180 | /* Private memcg ID. Used to ID objects that outlive the cgroup */ |
181 | struct mem_cgroup_id id; | |
182 | ||
33398cf2 MH |
183 | /* Accounted resources */ |
184 | struct page_counter memory; | |
37e84351 | 185 | struct page_counter swap; |
0db15298 JW |
186 | |
187 | /* Legacy consumer-oriented counters */ | |
33398cf2 MH |
188 | struct page_counter memsw; |
189 | struct page_counter kmem; | |
0db15298 | 190 | struct page_counter tcpmem; |
33398cf2 MH |
191 | |
192 | /* Normal memory consumption range */ | |
193 | unsigned long low; | |
194 | unsigned long high; | |
195 | ||
f7e1cb6e JW |
196 | /* Range enforcement for interrupt charges */ |
197 | struct work_struct high_work; | |
198 | ||
33398cf2 MH |
199 | unsigned long soft_limit; |
200 | ||
201 | /* vmpressure notifications */ | |
202 | struct vmpressure vmpressure; | |
203 | ||
33398cf2 MH |
204 | /* |
205 | * Should the accounting and control be hierarchical, per subtree? | |
206 | */ | |
207 | bool use_hierarchy; | |
208 | ||
209 | /* protected by memcg_oom_lock */ | |
210 | bool oom_lock; | |
211 | int under_oom; | |
212 | ||
213 | int swappiness; | |
214 | /* OOM-Killer disable */ | |
215 | int oom_kill_disable; | |
216 | ||
472912a2 TH |
217 | /* handle for "memory.events" */ |
218 | struct cgroup_file events_file; | |
219 | ||
33398cf2 MH |
220 | /* protect arrays of thresholds */ |
221 | struct mutex thresholds_lock; | |
222 | ||
223 | /* thresholds for memory usage. RCU-protected */ | |
224 | struct mem_cgroup_thresholds thresholds; | |
225 | ||
226 | /* thresholds for mem+swap usage. RCU-protected */ | |
227 | struct mem_cgroup_thresholds memsw_thresholds; | |
228 | ||
229 | /* For oom notifier event fd */ | |
230 | struct list_head oom_notify; | |
231 | ||
232 | /* | |
233 | * Should we move charges of a task when a task is moved into this | |
234 | * mem_cgroup ? And what type of charges should we move ? | |
235 | */ | |
236 | unsigned long move_charge_at_immigrate; | |
237 | /* | |
238 | * set > 0 if pages under this cgroup are moving to other cgroup. | |
239 | */ | |
240 | atomic_t moving_account; | |
241 | /* taken only while moving_account > 0 */ | |
242 | spinlock_t move_lock; | |
243 | struct task_struct *move_lock_task; | |
244 | unsigned long move_lock_flags; | |
245 | /* | |
246 | * percpu counter. | |
247 | */ | |
248 | struct mem_cgroup_stat_cpu __percpu *stat; | |
33398cf2 | 249 | |
d886f4e4 JW |
250 | unsigned long socket_pressure; |
251 | ||
252 | /* Legacy tcp memory accounting */ | |
0db15298 JW |
253 | bool tcpmem_active; |
254 | int tcpmem_pressure; | |
d886f4e4 | 255 | |
127424c8 | 256 | #ifndef CONFIG_SLOB |
33398cf2 MH |
257 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ |
258 | int kmemcg_id; | |
567e9ab2 | 259 | enum memcg_kmem_state kmem_state; |
33398cf2 MH |
260 | #endif |
261 | ||
262 | int last_scanned_node; | |
263 | #if MAX_NUMNODES > 1 | |
264 | nodemask_t scan_nodes; | |
265 | atomic_t numainfo_events; | |
266 | atomic_t numainfo_updating; | |
267 | #endif | |
268 | ||
269 | #ifdef CONFIG_CGROUP_WRITEBACK | |
270 | struct list_head cgwb_list; | |
271 | struct wb_domain cgwb_domain; | |
272 | #endif | |
273 | ||
274 | /* List of events which userspace want to receive */ | |
275 | struct list_head event_list; | |
276 | spinlock_t event_list_lock; | |
277 | ||
278 | struct mem_cgroup_per_node *nodeinfo[0]; | |
279 | /* WARNING: nodeinfo must be the last member here */ | |
280 | }; | |
7d828602 JW |
281 | |
282 | extern struct mem_cgroup *root_mem_cgroup; | |
56161634 | 283 | |
23047a96 JW |
284 | static inline bool mem_cgroup_disabled(void) |
285 | { | |
286 | return !cgroup_subsys_enabled(memory_cgrp_subsys); | |
287 | } | |
288 | ||
33398cf2 MH |
289 | /** |
290 | * mem_cgroup_events - count memory events against a cgroup | |
291 | * @memcg: the memory cgroup | |
292 | * @idx: the event index | |
293 | * @nr: the number of events to account for | |
294 | */ | |
295 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, | |
241994ed | 296 | enum mem_cgroup_events_index idx, |
33398cf2 MH |
297 | unsigned int nr) |
298 | { | |
299 | this_cpu_add(memcg->stat->events[idx], nr); | |
472912a2 | 300 | cgroup_file_notify(&memcg->events_file); |
33398cf2 | 301 | } |
241994ed JW |
302 | |
303 | bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg); | |
304 | ||
00501b53 | 305 | int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
f627c2f5 KS |
306 | gfp_t gfp_mask, struct mem_cgroup **memcgp, |
307 | bool compound); | |
00501b53 | 308 | void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg, |
f627c2f5 KS |
309 | bool lrucare, bool compound); |
310 | void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg, | |
311 | bool compound); | |
0a31bc97 | 312 | void mem_cgroup_uncharge(struct page *page); |
747db954 | 313 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
569b846d | 314 | |
6a93ca8f | 315 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage); |
569b846d | 316 | |
0a31bc97 JW |
317 | struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *); |
318 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *); | |
c9b0ed51 | 319 | |
2314b42d | 320 | bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg); |
64219994 | 321 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
e993d905 | 322 | |
33398cf2 MH |
323 | static inline |
324 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ | |
325 | return css ? container_of(css, struct mem_cgroup, css) : NULL; | |
326 | } | |
327 | ||
8e8ae645 JW |
328 | #define mem_cgroup_from_counter(counter, member) \ |
329 | container_of(counter, struct mem_cgroup, member) | |
330 | ||
33398cf2 MH |
331 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
332 | struct mem_cgroup *, | |
333 | struct mem_cgroup_reclaim_cookie *); | |
334 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); | |
335 | ||
23047a96 JW |
336 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
337 | { | |
338 | if (mem_cgroup_disabled()) | |
339 | return 0; | |
340 | ||
73f576c0 | 341 | return memcg->id.id; |
23047a96 | 342 | } |
73f576c0 | 343 | struct mem_cgroup *mem_cgroup_from_id(unsigned short id); |
23047a96 | 344 | |
8e8ae645 JW |
345 | /** |
346 | * parent_mem_cgroup - find the accounting parent of a memcg | |
347 | * @memcg: memcg whose parent to find | |
348 | * | |
349 | * Returns the parent memcg, or NULL if this is the root or the memory | |
350 | * controller is in legacy no-hierarchy mode. | |
351 | */ | |
352 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) | |
353 | { | |
354 | if (!memcg->memory.parent) | |
355 | return NULL; | |
356 | return mem_cgroup_from_counter(memcg->memory.parent, memory); | |
357 | } | |
358 | ||
33398cf2 MH |
359 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
360 | struct mem_cgroup *root) | |
361 | { | |
362 | if (root == memcg) | |
363 | return true; | |
364 | if (!root->use_hierarchy) | |
365 | return false; | |
366 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); | |
367 | } | |
e1aab161 | 368 | |
2314b42d JW |
369 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
370 | struct mem_cgroup *memcg) | |
2e4d4091 | 371 | { |
587af308 | 372 | struct mem_cgroup *task_memcg; |
413918bb | 373 | bool match = false; |
c3ac9a8a | 374 | |
2e4d4091 | 375 | rcu_read_lock(); |
587af308 | 376 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
413918bb | 377 | if (task_memcg) |
2314b42d | 378 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
2e4d4091 | 379 | rcu_read_unlock(); |
c3ac9a8a | 380 | return match; |
2e4d4091 | 381 | } |
8a9f3ccd | 382 | |
64219994 | 383 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
2fc04524 | 384 | ino_t page_cgroup_ino(struct page *page); |
d324236b | 385 | |
eb01aaab VD |
386 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
387 | { | |
388 | if (mem_cgroup_disabled()) | |
389 | return true; | |
390 | return !!(memcg->css.flags & CSS_ONLINE); | |
391 | } | |
392 | ||
58ae83db KH |
393 | /* |
394 | * For memory reclaim. | |
395 | */ | |
889976db | 396 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
33398cf2 MH |
397 | |
398 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, | |
399 | int nr_pages); | |
400 | ||
0a6b76dd VD |
401 | unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg, |
402 | int nid, unsigned int lru_mask); | |
403 | ||
33398cf2 MH |
404 | static inline |
405 | unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) | |
406 | { | |
407 | struct mem_cgroup_per_zone *mz; | |
408 | ||
409 | mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec); | |
410 | return mz->lru_size[lru]; | |
411 | } | |
412 | ||
b23afb93 TH |
413 | void mem_cgroup_handle_over_high(void); |
414 | ||
64219994 MH |
415 | void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, |
416 | struct task_struct *p); | |
58ae83db | 417 | |
49426420 | 418 | static inline void mem_cgroup_oom_enable(void) |
519e5247 | 419 | { |
626ebc41 TH |
420 | WARN_ON(current->memcg_may_oom); |
421 | current->memcg_may_oom = 1; | |
519e5247 JW |
422 | } |
423 | ||
49426420 | 424 | static inline void mem_cgroup_oom_disable(void) |
519e5247 | 425 | { |
626ebc41 TH |
426 | WARN_ON(!current->memcg_may_oom); |
427 | current->memcg_may_oom = 0; | |
519e5247 JW |
428 | } |
429 | ||
3812c8c8 JW |
430 | static inline bool task_in_memcg_oom(struct task_struct *p) |
431 | { | |
626ebc41 | 432 | return p->memcg_in_oom; |
3812c8c8 JW |
433 | } |
434 | ||
49426420 | 435 | bool mem_cgroup_oom_synchronize(bool wait); |
3812c8c8 | 436 | |
c255a458 | 437 | #ifdef CONFIG_MEMCG_SWAP |
c077719b KH |
438 | extern int do_swap_account; |
439 | #endif | |
f8d66542 | 440 | |
62cccb8c JW |
441 | void lock_page_memcg(struct page *page); |
442 | void unlock_page_memcg(struct page *page); | |
d7365e78 | 443 | |
33398cf2 MH |
444 | /** |
445 | * mem_cgroup_update_page_stat - update page state statistics | |
62cccb8c | 446 | * @page: the page |
33398cf2 MH |
447 | * @idx: page state item to account |
448 | * @val: number of pages (positive or negative) | |
449 | * | |
fdf1cdb9 JW |
450 | * The @page must be locked or the caller must use lock_page_memcg() |
451 | * to prevent double accounting when the page is concurrently being | |
452 | * moved to another memcg: | |
81f8c3a4 | 453 | * |
fdf1cdb9 | 454 | * lock_page(page) or lock_page_memcg(page) |
81f8c3a4 | 455 | * if (TestClearPageState(page)) |
62cccb8c | 456 | * mem_cgroup_update_page_stat(page, state, -1); |
fdf1cdb9 | 457 | * unlock_page(page) or unlock_page_memcg(page) |
33398cf2 | 458 | */ |
62cccb8c | 459 | static inline void mem_cgroup_update_page_stat(struct page *page, |
33398cf2 MH |
460 | enum mem_cgroup_stat_index idx, int val) |
461 | { | |
fdf1cdb9 | 462 | VM_BUG_ON(!(rcu_read_lock_held() || PageLocked(page))); |
33398cf2 | 463 | |
62cccb8c JW |
464 | if (page->mem_cgroup) |
465 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); | |
33398cf2 MH |
466 | } |
467 | ||
62cccb8c | 468 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 469 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 470 | { |
62cccb8c | 471 | mem_cgroup_update_page_stat(page, idx, 1); |
2a7106f2 GT |
472 | } |
473 | ||
62cccb8c | 474 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 475 | enum mem_cgroup_stat_index idx) |
2a7106f2 | 476 | { |
62cccb8c | 477 | mem_cgroup_update_page_stat(page, idx, -1); |
2a7106f2 GT |
478 | } |
479 | ||
0608f43d AM |
480 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
481 | gfp_t gfp_mask, | |
482 | unsigned long *total_scanned); | |
a63d83f4 | 483 | |
68ae564b DR |
484 | static inline void mem_cgroup_count_vm_event(struct mm_struct *mm, |
485 | enum vm_event_item idx) | |
486 | { | |
33398cf2 MH |
487 | struct mem_cgroup *memcg; |
488 | ||
68ae564b DR |
489 | if (mem_cgroup_disabled()) |
490 | return; | |
33398cf2 MH |
491 | |
492 | rcu_read_lock(); | |
493 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
494 | if (unlikely(!memcg)) | |
495 | goto out; | |
496 | ||
497 | switch (idx) { | |
498 | case PGFAULT: | |
499 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]); | |
500 | break; | |
501 | case PGMAJFAULT: | |
502 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]); | |
503 | break; | |
504 | default: | |
505 | BUG(); | |
506 | } | |
507 | out: | |
508 | rcu_read_unlock(); | |
68ae564b | 509 | } |
ca3e0214 | 510 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
e94c8a9c | 511 | void mem_cgroup_split_huge_fixup(struct page *head); |
ca3e0214 KH |
512 | #endif |
513 | ||
c255a458 | 514 | #else /* CONFIG_MEMCG */ |
23047a96 JW |
515 | |
516 | #define MEM_CGROUP_ID_SHIFT 0 | |
517 | #define MEM_CGROUP_ID_MAX 0 | |
518 | ||
7a81b88c KH |
519 | struct mem_cgroup; |
520 | ||
23047a96 JW |
521 | static inline bool mem_cgroup_disabled(void) |
522 | { | |
523 | return true; | |
524 | } | |
525 | ||
241994ed JW |
526 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, |
527 | enum mem_cgroup_events_index idx, | |
528 | unsigned int nr) | |
529 | { | |
530 | } | |
531 | ||
532 | static inline bool mem_cgroup_low(struct mem_cgroup *root, | |
533 | struct mem_cgroup *memcg) | |
534 | { | |
535 | return false; | |
536 | } | |
537 | ||
00501b53 JW |
538 | static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
539 | gfp_t gfp_mask, | |
f627c2f5 KS |
540 | struct mem_cgroup **memcgp, |
541 | bool compound) | |
7a81b88c | 542 | { |
00501b53 | 543 | *memcgp = NULL; |
7a81b88c KH |
544 | return 0; |
545 | } | |
546 | ||
00501b53 JW |
547 | static inline void mem_cgroup_commit_charge(struct page *page, |
548 | struct mem_cgroup *memcg, | |
f627c2f5 | 549 | bool lrucare, bool compound) |
7a81b88c KH |
550 | { |
551 | } | |
552 | ||
00501b53 | 553 | static inline void mem_cgroup_cancel_charge(struct page *page, |
f627c2f5 KS |
554 | struct mem_cgroup *memcg, |
555 | bool compound) | |
7a81b88c KH |
556 | { |
557 | } | |
558 | ||
0a31bc97 | 559 | static inline void mem_cgroup_uncharge(struct page *page) |
569b846d KH |
560 | { |
561 | } | |
562 | ||
747db954 | 563 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
8a9f3ccd BS |
564 | { |
565 | } | |
566 | ||
6a93ca8f | 567 | static inline void mem_cgroup_migrate(struct page *old, struct page *new) |
69029cd5 KH |
568 | { |
569 | } | |
570 | ||
925b7673 JW |
571 | static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone, |
572 | struct mem_cgroup *memcg) | |
08e552c6 | 573 | { |
925b7673 | 574 | return &zone->lruvec; |
08e552c6 KH |
575 | } |
576 | ||
fa9add64 HD |
577 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
578 | struct zone *zone) | |
66e1707b | 579 | { |
925b7673 | 580 | return &zone->lruvec; |
66e1707b BS |
581 | } |
582 | ||
587af308 | 583 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
c0ff4b85 | 584 | struct mem_cgroup *memcg) |
bed7161a | 585 | { |
587af308 | 586 | return true; |
bed7161a BS |
587 | } |
588 | ||
ffbdccf5 DR |
589 | static inline bool task_in_mem_cgroup(struct task_struct *task, |
590 | const struct mem_cgroup *memcg) | |
4c4a2214 | 591 | { |
ffbdccf5 | 592 | return true; |
4c4a2214 DR |
593 | } |
594 | ||
5660048c JW |
595 | static inline struct mem_cgroup * |
596 | mem_cgroup_iter(struct mem_cgroup *root, | |
597 | struct mem_cgroup *prev, | |
598 | struct mem_cgroup_reclaim_cookie *reclaim) | |
599 | { | |
600 | return NULL; | |
601 | } | |
602 | ||
603 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, | |
604 | struct mem_cgroup *prev) | |
605 | { | |
606 | } | |
607 | ||
23047a96 | 608 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
f8d66542 | 609 | { |
23047a96 JW |
610 | return 0; |
611 | } | |
612 | ||
613 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
614 | { | |
615 | WARN_ON_ONCE(id); | |
616 | /* XXX: This should always return root_mem_cgroup */ | |
617 | return NULL; | |
f8d66542 | 618 | } |
a636b327 | 619 | |
eb01aaab | 620 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
14797e23 | 621 | { |
13308ca9 | 622 | return true; |
14797e23 KM |
623 | } |
624 | ||
a3d8e054 | 625 | static inline unsigned long |
4d7dcca2 | 626 | mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
a3d8e054 KM |
627 | { |
628 | return 0; | |
629 | } | |
630 | ||
0a6b76dd VD |
631 | static inline unsigned long |
632 | mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg, | |
633 | int nid, unsigned int lru_mask) | |
634 | { | |
635 | return 0; | |
636 | } | |
637 | ||
e222432b BS |
638 | static inline void |
639 | mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p) | |
640 | { | |
641 | } | |
642 | ||
62cccb8c | 643 | static inline void lock_page_memcg(struct page *page) |
89c06bd5 KH |
644 | { |
645 | } | |
646 | ||
62cccb8c | 647 | static inline void unlock_page_memcg(struct page *page) |
89c06bd5 KH |
648 | { |
649 | } | |
650 | ||
b23afb93 TH |
651 | static inline void mem_cgroup_handle_over_high(void) |
652 | { | |
653 | } | |
654 | ||
49426420 | 655 | static inline void mem_cgroup_oom_enable(void) |
519e5247 JW |
656 | { |
657 | } | |
658 | ||
49426420 | 659 | static inline void mem_cgroup_oom_disable(void) |
519e5247 JW |
660 | { |
661 | } | |
662 | ||
3812c8c8 JW |
663 | static inline bool task_in_memcg_oom(struct task_struct *p) |
664 | { | |
665 | return false; | |
666 | } | |
667 | ||
49426420 | 668 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
3812c8c8 JW |
669 | { |
670 | return false; | |
671 | } | |
672 | ||
62cccb8c | 673 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
68b4876d | 674 | enum mem_cgroup_stat_index idx) |
2a7106f2 GT |
675 | { |
676 | } | |
677 | ||
62cccb8c | 678 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
68b4876d | 679 | enum mem_cgroup_stat_index idx) |
d69b042f BS |
680 | { |
681 | } | |
682 | ||
4e416953 | 683 | static inline |
0608f43d AM |
684 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
685 | gfp_t gfp_mask, | |
686 | unsigned long *total_scanned) | |
4e416953 | 687 | { |
0608f43d | 688 | return 0; |
4e416953 BS |
689 | } |
690 | ||
e94c8a9c | 691 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
ca3e0214 KH |
692 | { |
693 | } | |
694 | ||
456f998e YH |
695 | static inline |
696 | void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx) | |
697 | { | |
698 | } | |
c255a458 | 699 | #endif /* CONFIG_MEMCG */ |
78fb7466 | 700 | |
52ebea74 | 701 | #ifdef CONFIG_CGROUP_WRITEBACK |
841710aa | 702 | |
52ebea74 | 703 | struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); |
841710aa | 704 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
c5edf9cd TH |
705 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
706 | unsigned long *pheadroom, unsigned long *pdirty, | |
707 | unsigned long *pwriteback); | |
841710aa TH |
708 | |
709 | #else /* CONFIG_CGROUP_WRITEBACK */ | |
710 | ||
711 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) | |
712 | { | |
713 | return NULL; | |
714 | } | |
715 | ||
c2aa723a | 716 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
c5edf9cd TH |
717 | unsigned long *pfilepages, |
718 | unsigned long *pheadroom, | |
c2aa723a TH |
719 | unsigned long *pdirty, |
720 | unsigned long *pwriteback) | |
721 | { | |
722 | } | |
723 | ||
841710aa | 724 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
52ebea74 | 725 | |
e1aab161 | 726 | struct sock; |
e1aab161 GC |
727 | void sock_update_memcg(struct sock *sk); |
728 | void sock_release_memcg(struct sock *sk); | |
baac50bb JW |
729 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
730 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); | |
d886f4e4 | 731 | #ifdef CONFIG_MEMCG |
ef12947c JW |
732 | extern struct static_key_false memcg_sockets_enabled_key; |
733 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) | |
baac50bb | 734 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c | 735 | { |
0db15298 | 736 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
8e8ae645 | 737 | return true; |
8e8ae645 JW |
738 | do { |
739 | if (time_before(jiffies, memcg->socket_pressure)) | |
740 | return true; | |
741 | } while ((memcg = parent_mem_cgroup(memcg))); | |
742 | return false; | |
e805605c JW |
743 | } |
744 | #else | |
80e95fe0 | 745 | #define mem_cgroup_sockets_enabled 0 |
baac50bb | 746 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c JW |
747 | { |
748 | return false; | |
749 | } | |
750 | #endif | |
7ae1e1d0 | 751 | |
127424c8 | 752 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
ef12947c | 753 | extern struct static_key_false memcg_kmem_enabled_key; |
749c5415 | 754 | |
dbcf73e2 | 755 | extern int memcg_nr_cache_ids; |
64219994 MH |
756 | void memcg_get_cache_ids(void); |
757 | void memcg_put_cache_ids(void); | |
ebe945c2 GC |
758 | |
759 | /* | |
760 | * Helper macro to loop through all memcg-specific caches. Callers must still | |
761 | * check if the cache is valid (it is either valid or NULL). | |
762 | * the slab_mutex must be held when looping through those caches | |
763 | */ | |
749c5415 | 764 | #define for_each_memcg_cache_index(_idx) \ |
dbcf73e2 | 765 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
749c5415 | 766 | |
7ae1e1d0 GC |
767 | static inline bool memcg_kmem_enabled(void) |
768 | { | |
ef12947c | 769 | return static_branch_unlikely(&memcg_kmem_enabled_key); |
7ae1e1d0 GC |
770 | } |
771 | ||
772 | /* | |
773 | * In general, we'll do everything in our power to not incur in any overhead | |
774 | * for non-memcg users for the kmem functions. Not even a function call, if we | |
775 | * can avoid it. | |
776 | * | |
777 | * Therefore, we'll inline all those functions so that in the best case, we'll | |
778 | * see that kmemcg is off for everybody and proceed quickly. If it is on, | |
779 | * we'll still do most of the flag checking inline. We check a lot of | |
780 | * conditions, but because they are pretty simple, they are expected to be | |
781 | * fast. | |
782 | */ | |
f3ccb2c4 VD |
783 | int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order, |
784 | struct mem_cgroup *memcg); | |
d05e83a6 VD |
785 | int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order); |
786 | void __memcg_kmem_uncharge(struct page *page, int order); | |
7ae1e1d0 | 787 | |
33398cf2 | 788 | /* |
9f706d68 | 789 | * helper for accessing a memcg's index. It will be used as an index in the |
33398cf2 MH |
790 | * child cache array in kmem_cache, and also to derive its name. This function |
791 | * will return -1 when this is not a kmem-limited memcg. | |
792 | */ | |
793 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
794 | { | |
795 | return memcg ? memcg->kmemcg_id : -1; | |
796 | } | |
5722d094 | 797 | |
230e9fc2 | 798 | struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp); |
8135be5a | 799 | void __memcg_kmem_put_cache(struct kmem_cache *cachep); |
d7f25f8a | 800 | |
230e9fc2 | 801 | static inline bool __memcg_kmem_bypass(void) |
7ae1e1d0 GC |
802 | { |
803 | if (!memcg_kmem_enabled()) | |
804 | return true; | |
7ae1e1d0 GC |
805 | if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD)) |
806 | return true; | |
cbfb4798 | 807 | return false; |
7ae1e1d0 GC |
808 | } |
809 | ||
810 | /** | |
d05e83a6 VD |
811 | * memcg_kmem_charge: charge a kmem page |
812 | * @page: page to charge | |
813 | * @gfp: reclaim mode | |
814 | * @order: allocation order | |
cbfb4798 | 815 | * |
d05e83a6 | 816 | * Returns 0 on success, an error code on failure. |
7ae1e1d0 | 817 | */ |
d05e83a6 VD |
818 | static __always_inline int memcg_kmem_charge(struct page *page, |
819 | gfp_t gfp, int order) | |
7ae1e1d0 | 820 | { |
230e9fc2 VD |
821 | if (__memcg_kmem_bypass()) |
822 | return 0; | |
823 | if (!(gfp & __GFP_ACCOUNT)) | |
d05e83a6 VD |
824 | return 0; |
825 | return __memcg_kmem_charge(page, gfp, order); | |
7ae1e1d0 GC |
826 | } |
827 | ||
828 | /** | |
d05e83a6 VD |
829 | * memcg_kmem_uncharge: uncharge a kmem page |
830 | * @page: page to uncharge | |
831 | * @order: allocation order | |
7ae1e1d0 | 832 | */ |
d05e83a6 | 833 | static __always_inline void memcg_kmem_uncharge(struct page *page, int order) |
7ae1e1d0 | 834 | { |
7ae1e1d0 | 835 | if (memcg_kmem_enabled()) |
d05e83a6 | 836 | __memcg_kmem_uncharge(page, order); |
7ae1e1d0 GC |
837 | } |
838 | ||
d7f25f8a GC |
839 | /** |
840 | * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation | |
841 | * @cachep: the original global kmem cache | |
d7f25f8a | 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 | { | |
230e9fc2 | 848 | if (__memcg_kmem_bypass()) |
d7f25f8a | 849 | return cachep; |
230e9fc2 | 850 | return __memcg_kmem_get_cache(cachep, gfp); |
d7f25f8a | 851 | } |
8135be5a VD |
852 | |
853 | static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
854 | { | |
855 | if (memcg_kmem_enabled()) | |
856 | __memcg_kmem_put_cache(cachep); | |
857 | } | |
27ee57c9 VD |
858 | |
859 | /** | |
860 | * memcg_kmem_update_page_stat - update kmem page state statistics | |
861 | * @page: the page | |
862 | * @idx: page state item to account | |
863 | * @val: number of pages (positive or negative) | |
864 | */ | |
865 | static inline void memcg_kmem_update_page_stat(struct page *page, | |
866 | enum mem_cgroup_stat_index idx, int val) | |
867 | { | |
868 | if (memcg_kmem_enabled() && page->mem_cgroup) | |
869 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); | |
870 | } | |
871 | ||
7ae1e1d0 | 872 | #else |
749c5415 GC |
873 | #define for_each_memcg_cache_index(_idx) \ |
874 | for (; NULL; ) | |
875 | ||
b9ce5ef4 GC |
876 | static inline bool memcg_kmem_enabled(void) |
877 | { | |
878 | return false; | |
879 | } | |
880 | ||
d05e83a6 | 881 | static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order) |
7ae1e1d0 | 882 | { |
d05e83a6 | 883 | return 0; |
7ae1e1d0 GC |
884 | } |
885 | ||
d05e83a6 | 886 | static inline void memcg_kmem_uncharge(struct page *page, int order) |
7ae1e1d0 GC |
887 | { |
888 | } | |
2633d7a0 GC |
889 | |
890 | static inline int memcg_cache_id(struct mem_cgroup *memcg) | |
891 | { | |
892 | return -1; | |
893 | } | |
894 | ||
05257a1a VD |
895 | static inline void memcg_get_cache_ids(void) |
896 | { | |
897 | } | |
898 | ||
899 | static inline void memcg_put_cache_ids(void) | |
900 | { | |
901 | } | |
902 | ||
d7f25f8a GC |
903 | static inline struct kmem_cache * |
904 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) | |
905 | { | |
906 | return cachep; | |
907 | } | |
8135be5a VD |
908 | |
909 | static inline void memcg_kmem_put_cache(struct kmem_cache *cachep) | |
910 | { | |
911 | } | |
27ee57c9 VD |
912 | |
913 | static inline void memcg_kmem_update_page_stat(struct page *page, | |
914 | enum mem_cgroup_stat_index idx, int val) | |
915 | { | |
916 | } | |
127424c8 JW |
917 | #endif /* CONFIG_MEMCG && !CONFIG_SLOB */ |
918 | ||
8cdea7c0 | 919 | #endif /* _LINUX_MEMCONTROL_H */ |