memcontrol.h 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862
  1. /* memcontrol.h - Memory Controller
  2. *
  3. * Copyright IBM Corporation, 2007
  4. * Author Balbir Singh <balbir@linux.vnet.ibm.com>
  5. *
  6. * Copyright 2007 OpenVZ SWsoft Inc
  7. * Author: Pavel Emelianov <xemul@openvz.org>
  8. *
  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. #ifndef _LINUX_MEMCONTROL_H
  20. #define _LINUX_MEMCONTROL_H
  21. #include <linux/cgroup.h>
  22. #include <linux/vm_event_item.h>
  23. #include <linux/hardirq.h>
  24. #include <linux/jump_label.h>
  25. #include <linux/page_counter.h>
  26. #include <linux/vmpressure.h>
  27. #include <linux/eventfd.h>
  28. #include <linux/mmzone.h>
  29. #include <linux/writeback.h>
  30. struct mem_cgroup;
  31. struct page;
  32. struct mm_struct;
  33. struct kmem_cache;
  34. /*
  35. * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
  36. * These two lists should keep in accord with each other.
  37. */
  38. enum mem_cgroup_stat_index {
  39. /*
  40. * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
  41. */
  42. MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
  43. MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */
  44. MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */
  45. MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */
  46. MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */
  47. MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
  48. MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
  49. MEM_CGROUP_STAT_NSTATS,
  50. };
  51. struct mem_cgroup_reclaim_cookie {
  52. struct zone *zone;
  53. int priority;
  54. unsigned int generation;
  55. };
  56. enum mem_cgroup_events_index {
  57. MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */
  58. MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */
  59. MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */
  60. MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */
  61. MEM_CGROUP_EVENTS_NSTATS,
  62. /* default hierarchy events */
  63. MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
  64. MEMCG_HIGH,
  65. MEMCG_MAX,
  66. MEMCG_OOM,
  67. MEMCG_NR_EVENTS,
  68. };
  69. /*
  70. * Per memcg event counter is incremented at every pagein/pageout. With THP,
  71. * it will be incremated by the number of pages. This counter is used for
  72. * for trigger some periodic events. This is straightforward and better
  73. * than using jiffies etc. to handle periodic memcg event.
  74. */
  75. enum mem_cgroup_events_target {
  76. MEM_CGROUP_TARGET_THRESH,
  77. MEM_CGROUP_TARGET_SOFTLIMIT,
  78. MEM_CGROUP_TARGET_NUMAINFO,
  79. MEM_CGROUP_NTARGETS,
  80. };
  81. struct cg_proto {
  82. struct page_counter memory_allocated; /* Current allocated memory. */
  83. int memory_pressure;
  84. bool active;
  85. };
  86. #ifdef CONFIG_MEMCG
  87. struct mem_cgroup_stat_cpu {
  88. long count[MEM_CGROUP_STAT_NSTATS];
  89. unsigned long events[MEMCG_NR_EVENTS];
  90. unsigned long nr_page_events;
  91. unsigned long targets[MEM_CGROUP_NTARGETS];
  92. };
  93. struct mem_cgroup_reclaim_iter {
  94. struct mem_cgroup *position;
  95. /* scan generation, increased every round-trip */
  96. unsigned int generation;
  97. };
  98. /*
  99. * per-zone information in memory controller.
  100. */
  101. struct mem_cgroup_per_zone {
  102. struct lruvec lruvec;
  103. unsigned long lru_size[NR_LRU_LISTS];
  104. struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
  105. struct rb_node tree_node; /* RB tree node */
  106. unsigned long usage_in_excess;/* Set to the value by which */
  107. /* the soft limit is exceeded*/
  108. bool on_tree;
  109. struct mem_cgroup *memcg; /* Back pointer, we cannot */
  110. /* use container_of */
  111. };
  112. struct mem_cgroup_per_node {
  113. struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
  114. };
  115. struct mem_cgroup_threshold {
  116. struct eventfd_ctx *eventfd;
  117. unsigned long threshold;
  118. };
  119. /* For threshold */
  120. struct mem_cgroup_threshold_ary {
  121. /* An array index points to threshold just below or equal to usage. */
  122. int current_threshold;
  123. /* Size of entries[] */
  124. unsigned int size;
  125. /* Array of thresholds */
  126. struct mem_cgroup_threshold entries[0];
  127. };
  128. struct mem_cgroup_thresholds {
  129. /* Primary thresholds array */
  130. struct mem_cgroup_threshold_ary *primary;
  131. /*
  132. * Spare threshold array.
  133. * This is needed to make mem_cgroup_unregister_event() "never fail".
  134. * It must be able to store at least primary->size - 1 entries.
  135. */
  136. struct mem_cgroup_threshold_ary *spare;
  137. };
  138. /*
  139. * The memory controller data structure. The memory controller controls both
  140. * page cache and RSS per cgroup. We would eventually like to provide
  141. * statistics based on the statistics developed by Rik Van Riel for clock-pro,
  142. * to help the administrator determine what knobs to tune.
  143. */
  144. struct mem_cgroup {
  145. struct cgroup_subsys_state css;
  146. /* Accounted resources */
  147. struct page_counter memory;
  148. struct page_counter memsw;
  149. struct page_counter kmem;
  150. /* Normal memory consumption range */
  151. unsigned long low;
  152. unsigned long high;
  153. /* Range enforcement for interrupt charges */
  154. struct work_struct high_work;
  155. unsigned long soft_limit;
  156. /* vmpressure notifications */
  157. struct vmpressure vmpressure;
  158. /* css_online() has been completed */
  159. int initialized;
  160. /*
  161. * Should the accounting and control be hierarchical, per subtree?
  162. */
  163. bool use_hierarchy;
  164. /* protected by memcg_oom_lock */
  165. bool oom_lock;
  166. int under_oom;
  167. int swappiness;
  168. /* OOM-Killer disable */
  169. int oom_kill_disable;
  170. /* handle for "memory.events" */
  171. struct cgroup_file events_file;
  172. /* protect arrays of thresholds */
  173. struct mutex thresholds_lock;
  174. /* thresholds for memory usage. RCU-protected */
  175. struct mem_cgroup_thresholds thresholds;
  176. /* thresholds for mem+swap usage. RCU-protected */
  177. struct mem_cgroup_thresholds memsw_thresholds;
  178. /* For oom notifier event fd */
  179. struct list_head oom_notify;
  180. /*
  181. * Should we move charges of a task when a task is moved into this
  182. * mem_cgroup ? And what type of charges should we move ?
  183. */
  184. unsigned long move_charge_at_immigrate;
  185. /*
  186. * set > 0 if pages under this cgroup are moving to other cgroup.
  187. */
  188. atomic_t moving_account;
  189. /* taken only while moving_account > 0 */
  190. spinlock_t move_lock;
  191. struct task_struct *move_lock_task;
  192. unsigned long move_lock_flags;
  193. /*
  194. * percpu counter.
  195. */
  196. struct mem_cgroup_stat_cpu __percpu *stat;
  197. #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
  198. struct cg_proto tcp_mem;
  199. #endif
  200. #if defined(CONFIG_MEMCG_KMEM)
  201. /* Index in the kmem_cache->memcg_params.memcg_caches array */
  202. int kmemcg_id;
  203. bool kmem_acct_activated;
  204. bool kmem_acct_active;
  205. #endif
  206. int last_scanned_node;
  207. #if MAX_NUMNODES > 1
  208. nodemask_t scan_nodes;
  209. atomic_t numainfo_events;
  210. atomic_t numainfo_updating;
  211. #endif
  212. #ifdef CONFIG_CGROUP_WRITEBACK
  213. struct list_head cgwb_list;
  214. struct wb_domain cgwb_domain;
  215. #endif
  216. /* List of events which userspace want to receive */
  217. struct list_head event_list;
  218. spinlock_t event_list_lock;
  219. struct mem_cgroup_per_node *nodeinfo[0];
  220. /* WARNING: nodeinfo must be the last member here */
  221. };
  222. extern struct mem_cgroup *root_mem_cgroup;
  223. /**
  224. * mem_cgroup_events - count memory events against a cgroup
  225. * @memcg: the memory cgroup
  226. * @idx: the event index
  227. * @nr: the number of events to account for
  228. */
  229. static inline void mem_cgroup_events(struct mem_cgroup *memcg,
  230. enum mem_cgroup_events_index idx,
  231. unsigned int nr)
  232. {
  233. this_cpu_add(memcg->stat->events[idx], nr);
  234. cgroup_file_notify(&memcg->events_file);
  235. }
  236. bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
  237. int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
  238. gfp_t gfp_mask, struct mem_cgroup **memcgp);
  239. void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
  240. bool lrucare);
  241. void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
  242. void mem_cgroup_uncharge(struct page *page);
  243. void mem_cgroup_uncharge_list(struct list_head *page_list);
  244. void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
  245. struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
  246. struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
  247. bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
  248. struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
  249. struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
  250. static inline
  251. struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
  252. return css ? container_of(css, struct mem_cgroup, css) : NULL;
  253. }
  254. struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
  255. struct mem_cgroup *,
  256. struct mem_cgroup_reclaim_cookie *);
  257. void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
  258. static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
  259. struct mem_cgroup *root)
  260. {
  261. if (root == memcg)
  262. return true;
  263. if (!root->use_hierarchy)
  264. return false;
  265. return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
  266. }
  267. static inline bool mm_match_cgroup(struct mm_struct *mm,
  268. struct mem_cgroup *memcg)
  269. {
  270. struct mem_cgroup *task_memcg;
  271. bool match = false;
  272. rcu_read_lock();
  273. task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
  274. if (task_memcg)
  275. match = mem_cgroup_is_descendant(task_memcg, memcg);
  276. rcu_read_unlock();
  277. return match;
  278. }
  279. struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
  280. ino_t page_cgroup_ino(struct page *page);
  281. static inline bool mem_cgroup_disabled(void)
  282. {
  283. return !cgroup_subsys_enabled(memory_cgrp_subsys);
  284. }
  285. /*
  286. * For memory reclaim.
  287. */
  288. int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
  289. void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
  290. int nr_pages);
  291. static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
  292. {
  293. struct mem_cgroup_per_zone *mz;
  294. struct mem_cgroup *memcg;
  295. if (mem_cgroup_disabled())
  296. return true;
  297. mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
  298. memcg = mz->memcg;
  299. return !!(memcg->css.flags & CSS_ONLINE);
  300. }
  301. static inline
  302. unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
  303. {
  304. struct mem_cgroup_per_zone *mz;
  305. mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
  306. return mz->lru_size[lru];
  307. }
  308. static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
  309. {
  310. unsigned long inactive_ratio;
  311. unsigned long inactive;
  312. unsigned long active;
  313. unsigned long gb;
  314. inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
  315. active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
  316. gb = (inactive + active) >> (30 - PAGE_SHIFT);
  317. if (gb)
  318. inactive_ratio = int_sqrt(10 * gb);
  319. else
  320. inactive_ratio = 1;
  321. return inactive * inactive_ratio < active;
  322. }
  323. void mem_cgroup_handle_over_high(void);
  324. void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
  325. struct task_struct *p);
  326. static inline void mem_cgroup_oom_enable(void)
  327. {
  328. WARN_ON(current->memcg_may_oom);
  329. current->memcg_may_oom = 1;
  330. }
  331. static inline void mem_cgroup_oom_disable(void)
  332. {
  333. WARN_ON(!current->memcg_may_oom);
  334. current->memcg_may_oom = 0;
  335. }
  336. static inline bool task_in_memcg_oom(struct task_struct *p)
  337. {
  338. return p->memcg_in_oom;
  339. }
  340. bool mem_cgroup_oom_synchronize(bool wait);
  341. #ifdef CONFIG_MEMCG_SWAP
  342. extern int do_swap_account;
  343. #endif
  344. struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
  345. void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
  346. /**
  347. * mem_cgroup_update_page_stat - update page state statistics
  348. * @memcg: memcg to account against
  349. * @idx: page state item to account
  350. * @val: number of pages (positive or negative)
  351. *
  352. * See mem_cgroup_begin_page_stat() for locking requirements.
  353. */
  354. static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
  355. enum mem_cgroup_stat_index idx, int val)
  356. {
  357. VM_BUG_ON(!rcu_read_lock_held());
  358. if (memcg)
  359. this_cpu_add(memcg->stat->count[idx], val);
  360. }
  361. static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
  362. enum mem_cgroup_stat_index idx)
  363. {
  364. mem_cgroup_update_page_stat(memcg, idx, 1);
  365. }
  366. static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
  367. enum mem_cgroup_stat_index idx)
  368. {
  369. mem_cgroup_update_page_stat(memcg, idx, -1);
  370. }
  371. unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
  372. gfp_t gfp_mask,
  373. unsigned long *total_scanned);
  374. static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
  375. enum vm_event_item idx)
  376. {
  377. struct mem_cgroup *memcg;
  378. if (mem_cgroup_disabled())
  379. return;
  380. rcu_read_lock();
  381. memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
  382. if (unlikely(!memcg))
  383. goto out;
  384. switch (idx) {
  385. case PGFAULT:
  386. this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
  387. break;
  388. case PGMAJFAULT:
  389. this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
  390. break;
  391. default:
  392. BUG();
  393. }
  394. out:
  395. rcu_read_unlock();
  396. }
  397. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  398. void mem_cgroup_split_huge_fixup(struct page *head);
  399. #endif
  400. #else /* CONFIG_MEMCG */
  401. struct mem_cgroup;
  402. static inline void mem_cgroup_events(struct mem_cgroup *memcg,
  403. enum mem_cgroup_events_index idx,
  404. unsigned int nr)
  405. {
  406. }
  407. static inline bool mem_cgroup_low(struct mem_cgroup *root,
  408. struct mem_cgroup *memcg)
  409. {
  410. return false;
  411. }
  412. static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
  413. gfp_t gfp_mask,
  414. struct mem_cgroup **memcgp)
  415. {
  416. *memcgp = NULL;
  417. return 0;
  418. }
  419. static inline void mem_cgroup_commit_charge(struct page *page,
  420. struct mem_cgroup *memcg,
  421. bool lrucare)
  422. {
  423. }
  424. static inline void mem_cgroup_cancel_charge(struct page *page,
  425. struct mem_cgroup *memcg)
  426. {
  427. }
  428. static inline void mem_cgroup_uncharge(struct page *page)
  429. {
  430. }
  431. static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
  432. {
  433. }
  434. static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
  435. {
  436. }
  437. static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
  438. struct mem_cgroup *memcg)
  439. {
  440. return &zone->lruvec;
  441. }
  442. static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
  443. struct zone *zone)
  444. {
  445. return &zone->lruvec;
  446. }
  447. static inline bool mm_match_cgroup(struct mm_struct *mm,
  448. struct mem_cgroup *memcg)
  449. {
  450. return true;
  451. }
  452. static inline bool task_in_mem_cgroup(struct task_struct *task,
  453. const struct mem_cgroup *memcg)
  454. {
  455. return true;
  456. }
  457. static inline struct mem_cgroup *
  458. mem_cgroup_iter(struct mem_cgroup *root,
  459. struct mem_cgroup *prev,
  460. struct mem_cgroup_reclaim_cookie *reclaim)
  461. {
  462. return NULL;
  463. }
  464. static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
  465. struct mem_cgroup *prev)
  466. {
  467. }
  468. static inline bool mem_cgroup_disabled(void)
  469. {
  470. return true;
  471. }
  472. static inline bool
  473. mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
  474. {
  475. return true;
  476. }
  477. static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
  478. {
  479. return true;
  480. }
  481. static inline unsigned long
  482. mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
  483. {
  484. return 0;
  485. }
  486. static inline void
  487. mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
  488. int increment)
  489. {
  490. }
  491. static inline void
  492. mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
  493. {
  494. }
  495. static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
  496. {
  497. return NULL;
  498. }
  499. static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
  500. {
  501. }
  502. static inline void mem_cgroup_handle_over_high(void)
  503. {
  504. }
  505. static inline void mem_cgroup_oom_enable(void)
  506. {
  507. }
  508. static inline void mem_cgroup_oom_disable(void)
  509. {
  510. }
  511. static inline bool task_in_memcg_oom(struct task_struct *p)
  512. {
  513. return false;
  514. }
  515. static inline bool mem_cgroup_oom_synchronize(bool wait)
  516. {
  517. return false;
  518. }
  519. static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
  520. enum mem_cgroup_stat_index idx)
  521. {
  522. }
  523. static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
  524. enum mem_cgroup_stat_index idx)
  525. {
  526. }
  527. static inline
  528. unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
  529. gfp_t gfp_mask,
  530. unsigned long *total_scanned)
  531. {
  532. return 0;
  533. }
  534. static inline void mem_cgroup_split_huge_fixup(struct page *head)
  535. {
  536. }
  537. static inline
  538. void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
  539. {
  540. }
  541. #endif /* CONFIG_MEMCG */
  542. #ifdef CONFIG_CGROUP_WRITEBACK
  543. struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
  544. struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
  545. void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
  546. unsigned long *pheadroom, unsigned long *pdirty,
  547. unsigned long *pwriteback);
  548. #else /* CONFIG_CGROUP_WRITEBACK */
  549. static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
  550. {
  551. return NULL;
  552. }
  553. static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
  554. unsigned long *pfilepages,
  555. unsigned long *pheadroom,
  556. unsigned long *pdirty,
  557. unsigned long *pwriteback)
  558. {
  559. }
  560. #endif /* CONFIG_CGROUP_WRITEBACK */
  561. struct sock;
  562. void sock_update_memcg(struct sock *sk);
  563. void sock_release_memcg(struct sock *sk);
  564. bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
  565. void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
  566. #if defined(CONFIG_MEMCG) && defined(CONFIG_INET)
  567. extern struct static_key memcg_sockets_enabled_key;
  568. #define mem_cgroup_sockets_enabled static_key_false(&memcg_sockets_enabled_key)
  569. static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
  570. {
  571. #ifdef CONFIG_MEMCG_KMEM
  572. return memcg->tcp_mem.memory_pressure;
  573. #else
  574. return false;
  575. #endif
  576. }
  577. #else
  578. #define mem_cgroup_sockets_enabled 0
  579. static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
  580. {
  581. return false;
  582. }
  583. #endif
  584. #ifdef CONFIG_MEMCG_KMEM
  585. extern struct static_key memcg_kmem_enabled_key;
  586. extern int memcg_nr_cache_ids;
  587. void memcg_get_cache_ids(void);
  588. void memcg_put_cache_ids(void);
  589. /*
  590. * Helper macro to loop through all memcg-specific caches. Callers must still
  591. * check if the cache is valid (it is either valid or NULL).
  592. * the slab_mutex must be held when looping through those caches
  593. */
  594. #define for_each_memcg_cache_index(_idx) \
  595. for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
  596. static inline bool memcg_kmem_enabled(void)
  597. {
  598. return static_key_false(&memcg_kmem_enabled_key);
  599. }
  600. static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
  601. {
  602. return memcg->kmem_acct_active;
  603. }
  604. /*
  605. * In general, we'll do everything in our power to not incur in any overhead
  606. * for non-memcg users for the kmem functions. Not even a function call, if we
  607. * can avoid it.
  608. *
  609. * Therefore, we'll inline all those functions so that in the best case, we'll
  610. * see that kmemcg is off for everybody and proceed quickly. If it is on,
  611. * we'll still do most of the flag checking inline. We check a lot of
  612. * conditions, but because they are pretty simple, they are expected to be
  613. * fast.
  614. */
  615. int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
  616. struct mem_cgroup *memcg);
  617. int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
  618. void __memcg_kmem_uncharge(struct page *page, int order);
  619. /*
  620. * helper for acessing a memcg's index. It will be used as an index in the
  621. * child cache array in kmem_cache, and also to derive its name. This function
  622. * will return -1 when this is not a kmem-limited memcg.
  623. */
  624. static inline int memcg_cache_id(struct mem_cgroup *memcg)
  625. {
  626. return memcg ? memcg->kmemcg_id : -1;
  627. }
  628. struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
  629. void __memcg_kmem_put_cache(struct kmem_cache *cachep);
  630. static inline bool __memcg_kmem_bypass(void)
  631. {
  632. if (!memcg_kmem_enabled())
  633. return true;
  634. if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
  635. return true;
  636. return false;
  637. }
  638. /**
  639. * memcg_kmem_charge: charge a kmem page
  640. * @page: page to charge
  641. * @gfp: reclaim mode
  642. * @order: allocation order
  643. *
  644. * Returns 0 on success, an error code on failure.
  645. */
  646. static __always_inline int memcg_kmem_charge(struct page *page,
  647. gfp_t gfp, int order)
  648. {
  649. if (__memcg_kmem_bypass())
  650. return 0;
  651. if (!(gfp & __GFP_ACCOUNT))
  652. return 0;
  653. return __memcg_kmem_charge(page, gfp, order);
  654. }
  655. /**
  656. * memcg_kmem_uncharge: uncharge a kmem page
  657. * @page: page to uncharge
  658. * @order: allocation order
  659. */
  660. static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
  661. {
  662. if (memcg_kmem_enabled())
  663. __memcg_kmem_uncharge(page, order);
  664. }
  665. /**
  666. * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
  667. * @cachep: the original global kmem cache
  668. *
  669. * All memory allocated from a per-memcg cache is charged to the owner memcg.
  670. */
  671. static __always_inline struct kmem_cache *
  672. memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
  673. {
  674. if (__memcg_kmem_bypass())
  675. return cachep;
  676. return __memcg_kmem_get_cache(cachep, gfp);
  677. }
  678. static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
  679. {
  680. if (memcg_kmem_enabled())
  681. __memcg_kmem_put_cache(cachep);
  682. }
  683. #else
  684. #define for_each_memcg_cache_index(_idx) \
  685. for (; NULL; )
  686. static inline bool memcg_kmem_enabled(void)
  687. {
  688. return false;
  689. }
  690. static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
  691. {
  692. return false;
  693. }
  694. static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
  695. {
  696. return 0;
  697. }
  698. static inline void memcg_kmem_uncharge(struct page *page, int order)
  699. {
  700. }
  701. static inline int memcg_cache_id(struct mem_cgroup *memcg)
  702. {
  703. return -1;
  704. }
  705. static inline void memcg_get_cache_ids(void)
  706. {
  707. }
  708. static inline void memcg_put_cache_ids(void)
  709. {
  710. }
  711. static inline struct kmem_cache *
  712. memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
  713. {
  714. return cachep;
  715. }
  716. static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
  717. {
  718. }
  719. #endif /* CONFIG_MEMCG_KMEM */
  720. #endif /* _LINUX_MEMCONTROL_H */