memcontrol.h 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924
  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. #include <linux/page-flags.h>
  31. struct mem_cgroup;
  32. struct page;
  33. struct mm_struct;
  34. struct kmem_cache;
  35. /*
  36. * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
  37. * These two lists should keep in accord with each other.
  38. */
  39. enum mem_cgroup_stat_index {
  40. /*
  41. * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
  42. */
  43. MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
  44. MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */
  45. MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */
  46. MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */
  47. MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */
  48. MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
  49. MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
  50. MEM_CGROUP_STAT_NSTATS,
  51. /* default hierarchy stats */
  52. MEMCG_KERNEL_STACK = MEM_CGROUP_STAT_NSTATS,
  53. MEMCG_SLAB_RECLAIMABLE,
  54. MEMCG_SLAB_UNRECLAIMABLE,
  55. MEMCG_SOCK,
  56. MEMCG_NR_STAT,
  57. };
  58. struct mem_cgroup_reclaim_cookie {
  59. struct zone *zone;
  60. int priority;
  61. unsigned int generation;
  62. };
  63. enum mem_cgroup_events_index {
  64. MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */
  65. MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */
  66. MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */
  67. MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */
  68. MEM_CGROUP_EVENTS_NSTATS,
  69. /* default hierarchy events */
  70. MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
  71. MEMCG_HIGH,
  72. MEMCG_MAX,
  73. MEMCG_OOM,
  74. MEMCG_NR_EVENTS,
  75. };
  76. /*
  77. * Per memcg event counter is incremented at every pagein/pageout. With THP,
  78. * it will be incremated by the number of pages. This counter is used for
  79. * for trigger some periodic events. This is straightforward and better
  80. * than using jiffies etc. to handle periodic memcg event.
  81. */
  82. enum mem_cgroup_events_target {
  83. MEM_CGROUP_TARGET_THRESH,
  84. MEM_CGROUP_TARGET_SOFTLIMIT,
  85. MEM_CGROUP_TARGET_NUMAINFO,
  86. MEM_CGROUP_NTARGETS,
  87. };
  88. #ifdef CONFIG_MEMCG
  89. #define MEM_CGROUP_ID_SHIFT 16
  90. #define MEM_CGROUP_ID_MAX USHRT_MAX
  91. struct mem_cgroup_stat_cpu {
  92. long count[MEMCG_NR_STAT];
  93. unsigned long events[MEMCG_NR_EVENTS];
  94. unsigned long nr_page_events;
  95. unsigned long targets[MEM_CGROUP_NTARGETS];
  96. };
  97. struct mem_cgroup_reclaim_iter {
  98. struct mem_cgroup *position;
  99. /* scan generation, increased every round-trip */
  100. unsigned int generation;
  101. };
  102. /*
  103. * per-zone information in memory controller.
  104. */
  105. struct mem_cgroup_per_zone {
  106. struct lruvec lruvec;
  107. unsigned long lru_size[NR_LRU_LISTS];
  108. struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
  109. struct rb_node tree_node; /* RB tree node */
  110. unsigned long usage_in_excess;/* Set to the value by which */
  111. /* the soft limit is exceeded*/
  112. bool on_tree;
  113. struct mem_cgroup *memcg; /* Back pointer, we cannot */
  114. /* use container_of */
  115. };
  116. struct mem_cgroup_per_node {
  117. struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
  118. };
  119. struct mem_cgroup_threshold {
  120. struct eventfd_ctx *eventfd;
  121. unsigned long threshold;
  122. };
  123. /* For threshold */
  124. struct mem_cgroup_threshold_ary {
  125. /* An array index points to threshold just below or equal to usage. */
  126. int current_threshold;
  127. /* Size of entries[] */
  128. unsigned int size;
  129. /* Array of thresholds */
  130. struct mem_cgroup_threshold entries[0];
  131. };
  132. struct mem_cgroup_thresholds {
  133. /* Primary thresholds array */
  134. struct mem_cgroup_threshold_ary *primary;
  135. /*
  136. * Spare threshold array.
  137. * This is needed to make mem_cgroup_unregister_event() "never fail".
  138. * It must be able to store at least primary->size - 1 entries.
  139. */
  140. struct mem_cgroup_threshold_ary *spare;
  141. };
  142. enum memcg_kmem_state {
  143. KMEM_NONE,
  144. KMEM_ALLOCATED,
  145. KMEM_ONLINE,
  146. };
  147. /*
  148. * The memory controller data structure. The memory controller controls both
  149. * page cache and RSS per cgroup. We would eventually like to provide
  150. * statistics based on the statistics developed by Rik Van Riel for clock-pro,
  151. * to help the administrator determine what knobs to tune.
  152. */
  153. struct mem_cgroup {
  154. struct cgroup_subsys_state css;
  155. /* Accounted resources */
  156. struct page_counter memory;
  157. struct page_counter swap;
  158. /* Legacy consumer-oriented counters */
  159. struct page_counter memsw;
  160. struct page_counter kmem;
  161. struct page_counter tcpmem;
  162. /* Normal memory consumption range */
  163. unsigned long low;
  164. unsigned long high;
  165. /* Range enforcement for interrupt charges */
  166. struct work_struct high_work;
  167. unsigned long soft_limit;
  168. /* vmpressure notifications */
  169. struct vmpressure vmpressure;
  170. /*
  171. * Should the accounting and control be hierarchical, per subtree?
  172. */
  173. bool use_hierarchy;
  174. /* protected by memcg_oom_lock */
  175. bool oom_lock;
  176. int under_oom;
  177. int swappiness;
  178. /* OOM-Killer disable */
  179. int oom_kill_disable;
  180. /* handle for "memory.events" */
  181. struct cgroup_file events_file;
  182. /* protect arrays of thresholds */
  183. struct mutex thresholds_lock;
  184. /* thresholds for memory usage. RCU-protected */
  185. struct mem_cgroup_thresholds thresholds;
  186. /* thresholds for mem+swap usage. RCU-protected */
  187. struct mem_cgroup_thresholds memsw_thresholds;
  188. /* For oom notifier event fd */
  189. struct list_head oom_notify;
  190. /*
  191. * Should we move charges of a task when a task is moved into this
  192. * mem_cgroup ? And what type of charges should we move ?
  193. */
  194. unsigned long move_charge_at_immigrate;
  195. /*
  196. * set > 0 if pages under this cgroup are moving to other cgroup.
  197. */
  198. atomic_t moving_account;
  199. /* taken only while moving_account > 0 */
  200. spinlock_t move_lock;
  201. struct task_struct *move_lock_task;
  202. unsigned long move_lock_flags;
  203. /*
  204. * percpu counter.
  205. */
  206. struct mem_cgroup_stat_cpu __percpu *stat;
  207. unsigned long socket_pressure;
  208. /* Legacy tcp memory accounting */
  209. bool tcpmem_active;
  210. int tcpmem_pressure;
  211. #ifndef CONFIG_SLOB
  212. /* Index in the kmem_cache->memcg_params.memcg_caches array */
  213. int kmemcg_id;
  214. enum memcg_kmem_state kmem_state;
  215. #endif
  216. int last_scanned_node;
  217. #if MAX_NUMNODES > 1
  218. nodemask_t scan_nodes;
  219. atomic_t numainfo_events;
  220. atomic_t numainfo_updating;
  221. #endif
  222. #ifdef CONFIG_CGROUP_WRITEBACK
  223. struct list_head cgwb_list;
  224. struct wb_domain cgwb_domain;
  225. #endif
  226. /* List of events which userspace want to receive */
  227. struct list_head event_list;
  228. spinlock_t event_list_lock;
  229. struct mem_cgroup_per_node *nodeinfo[0];
  230. /* WARNING: nodeinfo must be the last member here */
  231. };
  232. extern struct mem_cgroup *root_mem_cgroup;
  233. static inline bool mem_cgroup_disabled(void)
  234. {
  235. return !cgroup_subsys_enabled(memory_cgrp_subsys);
  236. }
  237. /**
  238. * mem_cgroup_events - count memory events against a cgroup
  239. * @memcg: the memory cgroup
  240. * @idx: the event index
  241. * @nr: the number of events to account for
  242. */
  243. static inline void mem_cgroup_events(struct mem_cgroup *memcg,
  244. enum mem_cgroup_events_index idx,
  245. unsigned int nr)
  246. {
  247. this_cpu_add(memcg->stat->events[idx], nr);
  248. cgroup_file_notify(&memcg->events_file);
  249. }
  250. bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
  251. int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
  252. gfp_t gfp_mask, struct mem_cgroup **memcgp,
  253. bool compound);
  254. void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
  255. bool lrucare, bool compound);
  256. void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
  257. bool compound);
  258. void mem_cgroup_uncharge(struct page *page);
  259. void mem_cgroup_uncharge_list(struct list_head *page_list);
  260. void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
  261. struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
  262. struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
  263. bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
  264. struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
  265. static inline
  266. struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
  267. return css ? container_of(css, struct mem_cgroup, css) : NULL;
  268. }
  269. #define mem_cgroup_from_counter(counter, member) \
  270. container_of(counter, struct mem_cgroup, member)
  271. struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
  272. struct mem_cgroup *,
  273. struct mem_cgroup_reclaim_cookie *);
  274. void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
  275. static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
  276. {
  277. if (mem_cgroup_disabled())
  278. return 0;
  279. return memcg->css.id;
  280. }
  281. /**
  282. * mem_cgroup_from_id - look up a memcg from an id
  283. * @id: the id to look up
  284. *
  285. * Caller must hold rcu_read_lock() and use css_tryget() as necessary.
  286. */
  287. static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
  288. {
  289. struct cgroup_subsys_state *css;
  290. css = css_from_id(id, &memory_cgrp_subsys);
  291. return mem_cgroup_from_css(css);
  292. }
  293. /**
  294. * parent_mem_cgroup - find the accounting parent of a memcg
  295. * @memcg: memcg whose parent to find
  296. *
  297. * Returns the parent memcg, or NULL if this is the root or the memory
  298. * controller is in legacy no-hierarchy mode.
  299. */
  300. static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
  301. {
  302. if (!memcg->memory.parent)
  303. return NULL;
  304. return mem_cgroup_from_counter(memcg->memory.parent, memory);
  305. }
  306. static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
  307. struct mem_cgroup *root)
  308. {
  309. if (root == memcg)
  310. return true;
  311. if (!root->use_hierarchy)
  312. return false;
  313. return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
  314. }
  315. static inline bool mm_match_cgroup(struct mm_struct *mm,
  316. struct mem_cgroup *memcg)
  317. {
  318. struct mem_cgroup *task_memcg;
  319. bool match = false;
  320. rcu_read_lock();
  321. task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
  322. if (task_memcg)
  323. match = mem_cgroup_is_descendant(task_memcg, memcg);
  324. rcu_read_unlock();
  325. return match;
  326. }
  327. struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
  328. ino_t page_cgroup_ino(struct page *page);
  329. static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
  330. {
  331. if (mem_cgroup_disabled())
  332. return true;
  333. return !!(memcg->css.flags & CSS_ONLINE);
  334. }
  335. /*
  336. * For memory reclaim.
  337. */
  338. int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
  339. void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
  340. int nr_pages);
  341. unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
  342. int nid, unsigned int lru_mask);
  343. static inline
  344. unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
  345. {
  346. struct mem_cgroup_per_zone *mz;
  347. mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
  348. return mz->lru_size[lru];
  349. }
  350. void mem_cgroup_handle_over_high(void);
  351. void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
  352. struct task_struct *p);
  353. static inline void mem_cgroup_oom_enable(void)
  354. {
  355. WARN_ON(current->memcg_may_oom);
  356. current->memcg_may_oom = 1;
  357. }
  358. static inline void mem_cgroup_oom_disable(void)
  359. {
  360. WARN_ON(!current->memcg_may_oom);
  361. current->memcg_may_oom = 0;
  362. }
  363. static inline bool task_in_memcg_oom(struct task_struct *p)
  364. {
  365. return p->memcg_in_oom;
  366. }
  367. bool mem_cgroup_oom_synchronize(bool wait);
  368. #ifdef CONFIG_MEMCG_SWAP
  369. extern int do_swap_account;
  370. #endif
  371. void lock_page_memcg(struct page *page);
  372. void unlock_page_memcg(struct page *page);
  373. /**
  374. * mem_cgroup_update_page_stat - update page state statistics
  375. * @page: the page
  376. * @idx: page state item to account
  377. * @val: number of pages (positive or negative)
  378. *
  379. * The @page must be locked or the caller must use lock_page_memcg()
  380. * to prevent double accounting when the page is concurrently being
  381. * moved to another memcg:
  382. *
  383. * lock_page(page) or lock_page_memcg(page)
  384. * if (TestClearPageState(page))
  385. * mem_cgroup_update_page_stat(page, state, -1);
  386. * unlock_page(page) or unlock_page_memcg(page)
  387. */
  388. static inline void mem_cgroup_update_page_stat(struct page *page,
  389. enum mem_cgroup_stat_index idx, int val)
  390. {
  391. VM_BUG_ON(!(rcu_read_lock_held() || PageLocked(page)));
  392. if (page->mem_cgroup)
  393. this_cpu_add(page->mem_cgroup->stat->count[idx], val);
  394. }
  395. static inline void mem_cgroup_inc_page_stat(struct page *page,
  396. enum mem_cgroup_stat_index idx)
  397. {
  398. mem_cgroup_update_page_stat(page, idx, 1);
  399. }
  400. static inline void mem_cgroup_dec_page_stat(struct page *page,
  401. enum mem_cgroup_stat_index idx)
  402. {
  403. mem_cgroup_update_page_stat(page, idx, -1);
  404. }
  405. unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
  406. gfp_t gfp_mask,
  407. unsigned long *total_scanned);
  408. static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
  409. enum vm_event_item idx)
  410. {
  411. struct mem_cgroup *memcg;
  412. if (mem_cgroup_disabled())
  413. return;
  414. rcu_read_lock();
  415. memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
  416. if (unlikely(!memcg))
  417. goto out;
  418. switch (idx) {
  419. case PGFAULT:
  420. this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
  421. break;
  422. case PGMAJFAULT:
  423. this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
  424. break;
  425. default:
  426. BUG();
  427. }
  428. out:
  429. rcu_read_unlock();
  430. }
  431. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  432. void mem_cgroup_split_huge_fixup(struct page *head);
  433. #endif
  434. #else /* CONFIG_MEMCG */
  435. #define MEM_CGROUP_ID_SHIFT 0
  436. #define MEM_CGROUP_ID_MAX 0
  437. struct mem_cgroup;
  438. static inline bool mem_cgroup_disabled(void)
  439. {
  440. return true;
  441. }
  442. static inline void mem_cgroup_events(struct mem_cgroup *memcg,
  443. enum mem_cgroup_events_index idx,
  444. unsigned int nr)
  445. {
  446. }
  447. static inline bool mem_cgroup_low(struct mem_cgroup *root,
  448. struct mem_cgroup *memcg)
  449. {
  450. return false;
  451. }
  452. static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
  453. gfp_t gfp_mask,
  454. struct mem_cgroup **memcgp,
  455. bool compound)
  456. {
  457. *memcgp = NULL;
  458. return 0;
  459. }
  460. static inline void mem_cgroup_commit_charge(struct page *page,
  461. struct mem_cgroup *memcg,
  462. bool lrucare, bool compound)
  463. {
  464. }
  465. static inline void mem_cgroup_cancel_charge(struct page *page,
  466. struct mem_cgroup *memcg,
  467. bool compound)
  468. {
  469. }
  470. static inline void mem_cgroup_uncharge(struct page *page)
  471. {
  472. }
  473. static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
  474. {
  475. }
  476. static inline void mem_cgroup_migrate(struct page *old, struct page *new)
  477. {
  478. }
  479. static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
  480. struct mem_cgroup *memcg)
  481. {
  482. return &zone->lruvec;
  483. }
  484. static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
  485. struct zone *zone)
  486. {
  487. return &zone->lruvec;
  488. }
  489. static inline bool mm_match_cgroup(struct mm_struct *mm,
  490. struct mem_cgroup *memcg)
  491. {
  492. return true;
  493. }
  494. static inline bool task_in_mem_cgroup(struct task_struct *task,
  495. const struct mem_cgroup *memcg)
  496. {
  497. return true;
  498. }
  499. static inline struct mem_cgroup *
  500. mem_cgroup_iter(struct mem_cgroup *root,
  501. struct mem_cgroup *prev,
  502. struct mem_cgroup_reclaim_cookie *reclaim)
  503. {
  504. return NULL;
  505. }
  506. static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
  507. struct mem_cgroup *prev)
  508. {
  509. }
  510. static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
  511. {
  512. return 0;
  513. }
  514. static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
  515. {
  516. WARN_ON_ONCE(id);
  517. /* XXX: This should always return root_mem_cgroup */
  518. return NULL;
  519. }
  520. static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
  521. {
  522. return true;
  523. }
  524. static inline unsigned long
  525. mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
  526. {
  527. return 0;
  528. }
  529. static inline unsigned long
  530. mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
  531. int nid, unsigned int lru_mask)
  532. {
  533. return 0;
  534. }
  535. static inline void
  536. mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
  537. {
  538. }
  539. static inline void lock_page_memcg(struct page *page)
  540. {
  541. }
  542. static inline void unlock_page_memcg(struct page *page)
  543. {
  544. }
  545. static inline void mem_cgroup_handle_over_high(void)
  546. {
  547. }
  548. static inline void mem_cgroup_oom_enable(void)
  549. {
  550. }
  551. static inline void mem_cgroup_oom_disable(void)
  552. {
  553. }
  554. static inline bool task_in_memcg_oom(struct task_struct *p)
  555. {
  556. return false;
  557. }
  558. static inline bool mem_cgroup_oom_synchronize(bool wait)
  559. {
  560. return false;
  561. }
  562. static inline void mem_cgroup_inc_page_stat(struct page *page,
  563. enum mem_cgroup_stat_index idx)
  564. {
  565. }
  566. static inline void mem_cgroup_dec_page_stat(struct page *page,
  567. enum mem_cgroup_stat_index idx)
  568. {
  569. }
  570. static inline
  571. unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
  572. gfp_t gfp_mask,
  573. unsigned long *total_scanned)
  574. {
  575. return 0;
  576. }
  577. static inline void mem_cgroup_split_huge_fixup(struct page *head)
  578. {
  579. }
  580. static inline
  581. void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
  582. {
  583. }
  584. #endif /* CONFIG_MEMCG */
  585. #ifdef CONFIG_CGROUP_WRITEBACK
  586. struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
  587. struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
  588. void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
  589. unsigned long *pheadroom, unsigned long *pdirty,
  590. unsigned long *pwriteback);
  591. #else /* CONFIG_CGROUP_WRITEBACK */
  592. static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
  593. {
  594. return NULL;
  595. }
  596. static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
  597. unsigned long *pfilepages,
  598. unsigned long *pheadroom,
  599. unsigned long *pdirty,
  600. unsigned long *pwriteback)
  601. {
  602. }
  603. #endif /* CONFIG_CGROUP_WRITEBACK */
  604. struct sock;
  605. void sock_update_memcg(struct sock *sk);
  606. void sock_release_memcg(struct sock *sk);
  607. bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
  608. void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
  609. #ifdef CONFIG_MEMCG
  610. extern struct static_key_false memcg_sockets_enabled_key;
  611. #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
  612. static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
  613. {
  614. if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
  615. return true;
  616. do {
  617. if (time_before(jiffies, memcg->socket_pressure))
  618. return true;
  619. } while ((memcg = parent_mem_cgroup(memcg)));
  620. return false;
  621. }
  622. #else
  623. #define mem_cgroup_sockets_enabled 0
  624. static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
  625. {
  626. return false;
  627. }
  628. #endif
  629. #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
  630. extern struct static_key_false memcg_kmem_enabled_key;
  631. extern int memcg_nr_cache_ids;
  632. void memcg_get_cache_ids(void);
  633. void memcg_put_cache_ids(void);
  634. /*
  635. * Helper macro to loop through all memcg-specific caches. Callers must still
  636. * check if the cache is valid (it is either valid or NULL).
  637. * the slab_mutex must be held when looping through those caches
  638. */
  639. #define for_each_memcg_cache_index(_idx) \
  640. for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
  641. static inline bool memcg_kmem_enabled(void)
  642. {
  643. return static_branch_unlikely(&memcg_kmem_enabled_key);
  644. }
  645. /*
  646. * In general, we'll do everything in our power to not incur in any overhead
  647. * for non-memcg users for the kmem functions. Not even a function call, if we
  648. * can avoid it.
  649. *
  650. * Therefore, we'll inline all those functions so that in the best case, we'll
  651. * see that kmemcg is off for everybody and proceed quickly. If it is on,
  652. * we'll still do most of the flag checking inline. We check a lot of
  653. * conditions, but because they are pretty simple, they are expected to be
  654. * fast.
  655. */
  656. int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
  657. struct mem_cgroup *memcg);
  658. int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
  659. void __memcg_kmem_uncharge(struct page *page, int order);
  660. /*
  661. * helper for accessing a memcg's index. It will be used as an index in the
  662. * child cache array in kmem_cache, and also to derive its name. This function
  663. * will return -1 when this is not a kmem-limited memcg.
  664. */
  665. static inline int memcg_cache_id(struct mem_cgroup *memcg)
  666. {
  667. return memcg ? memcg->kmemcg_id : -1;
  668. }
  669. struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
  670. void __memcg_kmem_put_cache(struct kmem_cache *cachep);
  671. static inline bool __memcg_kmem_bypass(void)
  672. {
  673. if (!memcg_kmem_enabled())
  674. return true;
  675. if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
  676. return true;
  677. return false;
  678. }
  679. /**
  680. * memcg_kmem_charge: charge a kmem page
  681. * @page: page to charge
  682. * @gfp: reclaim mode
  683. * @order: allocation order
  684. *
  685. * Returns 0 on success, an error code on failure.
  686. */
  687. static __always_inline int memcg_kmem_charge(struct page *page,
  688. gfp_t gfp, int order)
  689. {
  690. if (__memcg_kmem_bypass())
  691. return 0;
  692. if (!(gfp & __GFP_ACCOUNT))
  693. return 0;
  694. return __memcg_kmem_charge(page, gfp, order);
  695. }
  696. /**
  697. * memcg_kmem_uncharge: uncharge a kmem page
  698. * @page: page to uncharge
  699. * @order: allocation order
  700. */
  701. static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
  702. {
  703. if (memcg_kmem_enabled())
  704. __memcg_kmem_uncharge(page, order);
  705. }
  706. /**
  707. * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
  708. * @cachep: the original global kmem cache
  709. *
  710. * All memory allocated from a per-memcg cache is charged to the owner memcg.
  711. */
  712. static __always_inline struct kmem_cache *
  713. memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
  714. {
  715. if (__memcg_kmem_bypass())
  716. return cachep;
  717. return __memcg_kmem_get_cache(cachep, gfp);
  718. }
  719. static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
  720. {
  721. if (memcg_kmem_enabled())
  722. __memcg_kmem_put_cache(cachep);
  723. }
  724. /**
  725. * memcg_kmem_update_page_stat - update kmem page state statistics
  726. * @page: the page
  727. * @idx: page state item to account
  728. * @val: number of pages (positive or negative)
  729. */
  730. static inline void memcg_kmem_update_page_stat(struct page *page,
  731. enum mem_cgroup_stat_index idx, int val)
  732. {
  733. if (memcg_kmem_enabled() && page->mem_cgroup)
  734. this_cpu_add(page->mem_cgroup->stat->count[idx], val);
  735. }
  736. #else
  737. #define for_each_memcg_cache_index(_idx) \
  738. for (; NULL; )
  739. static inline bool memcg_kmem_enabled(void)
  740. {
  741. return false;
  742. }
  743. static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
  744. {
  745. return 0;
  746. }
  747. static inline void memcg_kmem_uncharge(struct page *page, int order)
  748. {
  749. }
  750. static inline int memcg_cache_id(struct mem_cgroup *memcg)
  751. {
  752. return -1;
  753. }
  754. static inline void memcg_get_cache_ids(void)
  755. {
  756. }
  757. static inline void memcg_put_cache_ids(void)
  758. {
  759. }
  760. static inline struct kmem_cache *
  761. memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
  762. {
  763. return cachep;
  764. }
  765. static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
  766. {
  767. }
  768. static inline void memcg_kmem_update_page_stat(struct page *page,
  769. enum mem_cgroup_stat_index idx, int val)
  770. {
  771. }
  772. #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
  773. #endif /* _LINUX_MEMCONTROL_H */