| 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 | |
| 20 | #ifndef _LINUX_MEMCONTROL_H |
| 21 | #define _LINUX_MEMCONTROL_H |
| 22 | #include <linux/cgroup.h> |
| 23 | #include <linux/vm_event_item.h> |
| 24 | #include <linux/hardirq.h> |
| 25 | #include <linux/jump_label.h> |
| 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> |
| 31 | #include <linux/page-flags.h> |
| 32 | |
| 33 | struct mem_cgroup; |
| 34 | struct page; |
| 35 | struct mm_struct; |
| 36 | struct kmem_cache; |
| 37 | |
| 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 */ |
| 50 | MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */ |
| 51 | MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */ |
| 52 | MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */ |
| 53 | MEM_CGROUP_STAT_NSTATS, |
| 54 | /* default hierarchy stats */ |
| 55 | MEMCG_KERNEL_STACK = MEM_CGROUP_STAT_NSTATS, |
| 56 | MEMCG_SLAB_RECLAIMABLE, |
| 57 | MEMCG_SLAB_UNRECLAIMABLE, |
| 58 | MEMCG_SOCK, |
| 59 | MEMCG_NR_STAT, |
| 60 | }; |
| 61 | |
| 62 | struct mem_cgroup_reclaim_cookie { |
| 63 | struct zone *zone; |
| 64 | int priority; |
| 65 | unsigned int generation; |
| 66 | }; |
| 67 | |
| 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 | |
| 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 | |
| 95 | #ifdef CONFIG_MEMCG |
| 96 | |
| 97 | #define MEM_CGROUP_ID_SHIFT 16 |
| 98 | #define MEM_CGROUP_ID_MAX USHRT_MAX |
| 99 | |
| 100 | struct mem_cgroup_stat_cpu { |
| 101 | long count[MEMCG_NR_STAT]; |
| 102 | unsigned long events[MEMCG_NR_EVENTS]; |
| 103 | unsigned long nr_page_events; |
| 104 | unsigned long targets[MEM_CGROUP_NTARGETS]; |
| 105 | }; |
| 106 | |
| 107 | struct mem_cgroup_reclaim_iter { |
| 108 | struct mem_cgroup *position; |
| 109 | /* scan generation, increased every round-trip */ |
| 110 | unsigned int generation; |
| 111 | }; |
| 112 | |
| 113 | /* |
| 114 | * per-zone information in memory controller. |
| 115 | */ |
| 116 | struct mem_cgroup_per_zone { |
| 117 | struct lruvec lruvec; |
| 118 | unsigned long lru_size[NR_LRU_LISTS]; |
| 119 | |
| 120 | struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1]; |
| 121 | |
| 122 | struct rb_node tree_node; /* RB tree node */ |
| 123 | unsigned long usage_in_excess;/* Set to the value by which */ |
| 124 | /* the soft limit is exceeded*/ |
| 125 | bool on_tree; |
| 126 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ |
| 127 | /* use container_of */ |
| 128 | }; |
| 129 | |
| 130 | struct mem_cgroup_per_node { |
| 131 | struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES]; |
| 132 | }; |
| 133 | |
| 134 | struct mem_cgroup_threshold { |
| 135 | struct eventfd_ctx *eventfd; |
| 136 | unsigned long threshold; |
| 137 | }; |
| 138 | |
| 139 | /* For threshold */ |
| 140 | struct mem_cgroup_threshold_ary { |
| 141 | /* An array index points to threshold just below or equal to usage. */ |
| 142 | int current_threshold; |
| 143 | /* Size of entries[] */ |
| 144 | unsigned int size; |
| 145 | /* Array of thresholds */ |
| 146 | struct mem_cgroup_threshold entries[0]; |
| 147 | }; |
| 148 | |
| 149 | struct mem_cgroup_thresholds { |
| 150 | /* Primary thresholds array */ |
| 151 | struct mem_cgroup_threshold_ary *primary; |
| 152 | /* |
| 153 | * Spare threshold array. |
| 154 | * This is needed to make mem_cgroup_unregister_event() "never fail". |
| 155 | * It must be able to store at least primary->size - 1 entries. |
| 156 | */ |
| 157 | struct mem_cgroup_threshold_ary *spare; |
| 158 | }; |
| 159 | |
| 160 | enum memcg_kmem_state { |
| 161 | KMEM_NONE, |
| 162 | KMEM_ALLOCATED, |
| 163 | KMEM_ONLINE, |
| 164 | }; |
| 165 | |
| 166 | /* |
| 167 | * The memory controller data structure. The memory controller controls both |
| 168 | * page cache and RSS per cgroup. We would eventually like to provide |
| 169 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, |
| 170 | * to help the administrator determine what knobs to tune. |
| 171 | */ |
| 172 | struct mem_cgroup { |
| 173 | struct cgroup_subsys_state css; |
| 174 | |
| 175 | /* Accounted resources */ |
| 176 | struct page_counter memory; |
| 177 | struct page_counter swap; |
| 178 | |
| 179 | /* Legacy consumer-oriented counters */ |
| 180 | struct page_counter memsw; |
| 181 | struct page_counter kmem; |
| 182 | struct page_counter tcpmem; |
| 183 | |
| 184 | /* Normal memory consumption range */ |
| 185 | unsigned long low; |
| 186 | unsigned long high; |
| 187 | |
| 188 | /* Range enforcement for interrupt charges */ |
| 189 | struct work_struct high_work; |
| 190 | |
| 191 | unsigned long soft_limit; |
| 192 | |
| 193 | /* vmpressure notifications */ |
| 194 | struct vmpressure vmpressure; |
| 195 | |
| 196 | /* |
| 197 | * Should the accounting and control be hierarchical, per subtree? |
| 198 | */ |
| 199 | bool use_hierarchy; |
| 200 | |
| 201 | /* protected by memcg_oom_lock */ |
| 202 | bool oom_lock; |
| 203 | int under_oom; |
| 204 | |
| 205 | int swappiness; |
| 206 | /* OOM-Killer disable */ |
| 207 | int oom_kill_disable; |
| 208 | |
| 209 | /* handle for "memory.events" */ |
| 210 | struct cgroup_file events_file; |
| 211 | |
| 212 | /* protect arrays of thresholds */ |
| 213 | struct mutex thresholds_lock; |
| 214 | |
| 215 | /* thresholds for memory usage. RCU-protected */ |
| 216 | struct mem_cgroup_thresholds thresholds; |
| 217 | |
| 218 | /* thresholds for mem+swap usage. RCU-protected */ |
| 219 | struct mem_cgroup_thresholds memsw_thresholds; |
| 220 | |
| 221 | /* For oom notifier event fd */ |
| 222 | struct list_head oom_notify; |
| 223 | |
| 224 | /* |
| 225 | * Should we move charges of a task when a task is moved into this |
| 226 | * mem_cgroup ? And what type of charges should we move ? |
| 227 | */ |
| 228 | unsigned long move_charge_at_immigrate; |
| 229 | /* |
| 230 | * set > 0 if pages under this cgroup are moving to other cgroup. |
| 231 | */ |
| 232 | atomic_t moving_account; |
| 233 | /* taken only while moving_account > 0 */ |
| 234 | spinlock_t move_lock; |
| 235 | struct task_struct *move_lock_task; |
| 236 | unsigned long move_lock_flags; |
| 237 | /* |
| 238 | * percpu counter. |
| 239 | */ |
| 240 | struct mem_cgroup_stat_cpu __percpu *stat; |
| 241 | |
| 242 | unsigned long socket_pressure; |
| 243 | |
| 244 | /* Legacy tcp memory accounting */ |
| 245 | bool tcpmem_active; |
| 246 | int tcpmem_pressure; |
| 247 | |
| 248 | #ifndef CONFIG_SLOB |
| 249 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ |
| 250 | int kmemcg_id; |
| 251 | enum memcg_kmem_state kmem_state; |
| 252 | #endif |
| 253 | |
| 254 | int last_scanned_node; |
| 255 | #if MAX_NUMNODES > 1 |
| 256 | nodemask_t scan_nodes; |
| 257 | atomic_t numainfo_events; |
| 258 | atomic_t numainfo_updating; |
| 259 | #endif |
| 260 | |
| 261 | #ifdef CONFIG_CGROUP_WRITEBACK |
| 262 | struct list_head cgwb_list; |
| 263 | struct wb_domain cgwb_domain; |
| 264 | #endif |
| 265 | |
| 266 | /* List of events which userspace want to receive */ |
| 267 | struct list_head event_list; |
| 268 | spinlock_t event_list_lock; |
| 269 | |
| 270 | struct mem_cgroup_per_node *nodeinfo[0]; |
| 271 | /* WARNING: nodeinfo must be the last member here */ |
| 272 | }; |
| 273 | |
| 274 | extern struct mem_cgroup *root_mem_cgroup; |
| 275 | |
| 276 | static inline bool mem_cgroup_disabled(void) |
| 277 | { |
| 278 | return !cgroup_subsys_enabled(memory_cgrp_subsys); |
| 279 | } |
| 280 | |
| 281 | /** |
| 282 | * mem_cgroup_events - count memory events against a cgroup |
| 283 | * @memcg: the memory cgroup |
| 284 | * @idx: the event index |
| 285 | * @nr: the number of events to account for |
| 286 | */ |
| 287 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, |
| 288 | enum mem_cgroup_events_index idx, |
| 289 | unsigned int nr) |
| 290 | { |
| 291 | this_cpu_add(memcg->stat->events[idx], nr); |
| 292 | cgroup_file_notify(&memcg->events_file); |
| 293 | } |
| 294 | |
| 295 | bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg); |
| 296 | |
| 297 | int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
| 298 | gfp_t gfp_mask, struct mem_cgroup **memcgp, |
| 299 | bool compound); |
| 300 | void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg, |
| 301 | bool lrucare, bool compound); |
| 302 | void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg, |
| 303 | bool compound); |
| 304 | void mem_cgroup_uncharge(struct page *page); |
| 305 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
| 306 | |
| 307 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage); |
| 308 | |
| 309 | struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *); |
| 310 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *); |
| 311 | |
| 312 | bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg); |
| 313 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
| 314 | |
| 315 | static inline |
| 316 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ |
| 317 | return css ? container_of(css, struct mem_cgroup, css) : NULL; |
| 318 | } |
| 319 | |
| 320 | #define mem_cgroup_from_counter(counter, member) \ |
| 321 | container_of(counter, struct mem_cgroup, member) |
| 322 | |
| 323 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
| 324 | struct mem_cgroup *, |
| 325 | struct mem_cgroup_reclaim_cookie *); |
| 326 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); |
| 327 | |
| 328 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
| 329 | { |
| 330 | if (mem_cgroup_disabled()) |
| 331 | return 0; |
| 332 | |
| 333 | return memcg->css.id; |
| 334 | } |
| 335 | |
| 336 | /** |
| 337 | * mem_cgroup_from_id - look up a memcg from an id |
| 338 | * @id: the id to look up |
| 339 | * |
| 340 | * Caller must hold rcu_read_lock() and use css_tryget() as necessary. |
| 341 | */ |
| 342 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) |
| 343 | { |
| 344 | struct cgroup_subsys_state *css; |
| 345 | |
| 346 | css = css_from_id(id, &memory_cgrp_subsys); |
| 347 | return mem_cgroup_from_css(css); |
| 348 | } |
| 349 | |
| 350 | /** |
| 351 | * parent_mem_cgroup - find the accounting parent of a memcg |
| 352 | * @memcg: memcg whose parent to find |
| 353 | * |
| 354 | * Returns the parent memcg, or NULL if this is the root or the memory |
| 355 | * controller is in legacy no-hierarchy mode. |
| 356 | */ |
| 357 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) |
| 358 | { |
| 359 | if (!memcg->memory.parent) |
| 360 | return NULL; |
| 361 | return mem_cgroup_from_counter(memcg->memory.parent, memory); |
| 362 | } |
| 363 | |
| 364 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
| 365 | struct mem_cgroup *root) |
| 366 | { |
| 367 | if (root == memcg) |
| 368 | return true; |
| 369 | if (!root->use_hierarchy) |
| 370 | return false; |
| 371 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); |
| 372 | } |
| 373 | |
| 374 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
| 375 | struct mem_cgroup *memcg) |
| 376 | { |
| 377 | struct mem_cgroup *task_memcg; |
| 378 | bool match = false; |
| 379 | |
| 380 | rcu_read_lock(); |
| 381 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
| 382 | if (task_memcg) |
| 383 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
| 384 | rcu_read_unlock(); |
| 385 | return match; |
| 386 | } |
| 387 | |
| 388 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
| 389 | ino_t page_cgroup_ino(struct page *page); |
| 390 | |
| 391 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
| 392 | { |
| 393 | if (mem_cgroup_disabled()) |
| 394 | return true; |
| 395 | return !!(memcg->css.flags & CSS_ONLINE); |
| 396 | } |
| 397 | |
| 398 | /* |
| 399 | * For memory reclaim. |
| 400 | */ |
| 401 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
| 402 | |
| 403 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, |
| 404 | int nr_pages); |
| 405 | |
| 406 | unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg, |
| 407 | int nid, unsigned int lru_mask); |
| 408 | |
| 409 | static inline |
| 410 | unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
| 411 | { |
| 412 | struct mem_cgroup_per_zone *mz; |
| 413 | |
| 414 | mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec); |
| 415 | return mz->lru_size[lru]; |
| 416 | } |
| 417 | |
| 418 | static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) |
| 419 | { |
| 420 | unsigned long inactive_ratio; |
| 421 | unsigned long inactive; |
| 422 | unsigned long active; |
| 423 | unsigned long gb; |
| 424 | |
| 425 | inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON); |
| 426 | active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON); |
| 427 | |
| 428 | gb = (inactive + active) >> (30 - PAGE_SHIFT); |
| 429 | if (gb) |
| 430 | inactive_ratio = int_sqrt(10 * gb); |
| 431 | else |
| 432 | inactive_ratio = 1; |
| 433 | |
| 434 | return inactive * inactive_ratio < active; |
| 435 | } |
| 436 | |
| 437 | void mem_cgroup_handle_over_high(void); |
| 438 | |
| 439 | void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, |
| 440 | struct task_struct *p); |
| 441 | |
| 442 | static inline void mem_cgroup_oom_enable(void) |
| 443 | { |
| 444 | WARN_ON(current->memcg_may_oom); |
| 445 | current->memcg_may_oom = 1; |
| 446 | } |
| 447 | |
| 448 | static inline void mem_cgroup_oom_disable(void) |
| 449 | { |
| 450 | WARN_ON(!current->memcg_may_oom); |
| 451 | current->memcg_may_oom = 0; |
| 452 | } |
| 453 | |
| 454 | static inline bool task_in_memcg_oom(struct task_struct *p) |
| 455 | { |
| 456 | return p->memcg_in_oom; |
| 457 | } |
| 458 | |
| 459 | bool mem_cgroup_oom_synchronize(bool wait); |
| 460 | |
| 461 | #ifdef CONFIG_MEMCG_SWAP |
| 462 | extern int do_swap_account; |
| 463 | #endif |
| 464 | |
| 465 | void lock_page_memcg(struct page *page); |
| 466 | void unlock_page_memcg(struct page *page); |
| 467 | |
| 468 | /** |
| 469 | * mem_cgroup_update_page_stat - update page state statistics |
| 470 | * @page: the page |
| 471 | * @idx: page state item to account |
| 472 | * @val: number of pages (positive or negative) |
| 473 | * |
| 474 | * The @page must be locked or the caller must use lock_page_memcg() |
| 475 | * to prevent double accounting when the page is concurrently being |
| 476 | * moved to another memcg: |
| 477 | * |
| 478 | * lock_page(page) or lock_page_memcg(page) |
| 479 | * if (TestClearPageState(page)) |
| 480 | * mem_cgroup_update_page_stat(page, state, -1); |
| 481 | * unlock_page(page) or unlock_page_memcg(page) |
| 482 | */ |
| 483 | static inline void mem_cgroup_update_page_stat(struct page *page, |
| 484 | enum mem_cgroup_stat_index idx, int val) |
| 485 | { |
| 486 | VM_BUG_ON(!(rcu_read_lock_held() || PageLocked(page))); |
| 487 | |
| 488 | if (page->mem_cgroup) |
| 489 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); |
| 490 | } |
| 491 | |
| 492 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
| 493 | enum mem_cgroup_stat_index idx) |
| 494 | { |
| 495 | mem_cgroup_update_page_stat(page, idx, 1); |
| 496 | } |
| 497 | |
| 498 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
| 499 | enum mem_cgroup_stat_index idx) |
| 500 | { |
| 501 | mem_cgroup_update_page_stat(page, idx, -1); |
| 502 | } |
| 503 | |
| 504 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
| 505 | gfp_t gfp_mask, |
| 506 | unsigned long *total_scanned); |
| 507 | |
| 508 | static inline void mem_cgroup_count_vm_event(struct mm_struct *mm, |
| 509 | enum vm_event_item idx) |
| 510 | { |
| 511 | struct mem_cgroup *memcg; |
| 512 | |
| 513 | if (mem_cgroup_disabled()) |
| 514 | return; |
| 515 | |
| 516 | rcu_read_lock(); |
| 517 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
| 518 | if (unlikely(!memcg)) |
| 519 | goto out; |
| 520 | |
| 521 | switch (idx) { |
| 522 | case PGFAULT: |
| 523 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]); |
| 524 | break; |
| 525 | case PGMAJFAULT: |
| 526 | this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]); |
| 527 | break; |
| 528 | default: |
| 529 | BUG(); |
| 530 | } |
| 531 | out: |
| 532 | rcu_read_unlock(); |
| 533 | } |
| 534 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 535 | void mem_cgroup_split_huge_fixup(struct page *head); |
| 536 | #endif |
| 537 | |
| 538 | #else /* CONFIG_MEMCG */ |
| 539 | |
| 540 | #define MEM_CGROUP_ID_SHIFT 0 |
| 541 | #define MEM_CGROUP_ID_MAX 0 |
| 542 | |
| 543 | struct mem_cgroup; |
| 544 | |
| 545 | static inline bool mem_cgroup_disabled(void) |
| 546 | { |
| 547 | return true; |
| 548 | } |
| 549 | |
| 550 | static inline void mem_cgroup_events(struct mem_cgroup *memcg, |
| 551 | enum mem_cgroup_events_index idx, |
| 552 | unsigned int nr) |
| 553 | { |
| 554 | } |
| 555 | |
| 556 | static inline bool mem_cgroup_low(struct mem_cgroup *root, |
| 557 | struct mem_cgroup *memcg) |
| 558 | { |
| 559 | return false; |
| 560 | } |
| 561 | |
| 562 | static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm, |
| 563 | gfp_t gfp_mask, |
| 564 | struct mem_cgroup **memcgp, |
| 565 | bool compound) |
| 566 | { |
| 567 | *memcgp = NULL; |
| 568 | return 0; |
| 569 | } |
| 570 | |
| 571 | static inline void mem_cgroup_commit_charge(struct page *page, |
| 572 | struct mem_cgroup *memcg, |
| 573 | bool lrucare, bool compound) |
| 574 | { |
| 575 | } |
| 576 | |
| 577 | static inline void mem_cgroup_cancel_charge(struct page *page, |
| 578 | struct mem_cgroup *memcg, |
| 579 | bool compound) |
| 580 | { |
| 581 | } |
| 582 | |
| 583 | static inline void mem_cgroup_uncharge(struct page *page) |
| 584 | { |
| 585 | } |
| 586 | |
| 587 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
| 588 | { |
| 589 | } |
| 590 | |
| 591 | static inline void mem_cgroup_migrate(struct page *old, struct page *new) |
| 592 | { |
| 593 | } |
| 594 | |
| 595 | static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone, |
| 596 | struct mem_cgroup *memcg) |
| 597 | { |
| 598 | return &zone->lruvec; |
| 599 | } |
| 600 | |
| 601 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
| 602 | struct zone *zone) |
| 603 | { |
| 604 | return &zone->lruvec; |
| 605 | } |
| 606 | |
| 607 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
| 608 | struct mem_cgroup *memcg) |
| 609 | { |
| 610 | return true; |
| 611 | } |
| 612 | |
| 613 | static inline bool task_in_mem_cgroup(struct task_struct *task, |
| 614 | const struct mem_cgroup *memcg) |
| 615 | { |
| 616 | return true; |
| 617 | } |
| 618 | |
| 619 | static inline struct mem_cgroup * |
| 620 | mem_cgroup_iter(struct mem_cgroup *root, |
| 621 | struct mem_cgroup *prev, |
| 622 | struct mem_cgroup_reclaim_cookie *reclaim) |
| 623 | { |
| 624 | return NULL; |
| 625 | } |
| 626 | |
| 627 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, |
| 628 | struct mem_cgroup *prev) |
| 629 | { |
| 630 | } |
| 631 | |
| 632 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
| 633 | { |
| 634 | return 0; |
| 635 | } |
| 636 | |
| 637 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) |
| 638 | { |
| 639 | WARN_ON_ONCE(id); |
| 640 | /* XXX: This should always return root_mem_cgroup */ |
| 641 | return NULL; |
| 642 | } |
| 643 | |
| 644 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
| 645 | { |
| 646 | return true; |
| 647 | } |
| 648 | |
| 649 | static inline bool |
| 650 | mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec) |
| 651 | { |
| 652 | return true; |
| 653 | } |
| 654 | |
| 655 | static inline unsigned long |
| 656 | mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru) |
| 657 | { |
| 658 | return 0; |
| 659 | } |
| 660 | |
| 661 | static inline void |
| 662 | mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, |
| 663 | int increment) |
| 664 | { |
| 665 | } |
| 666 | |
| 667 | static inline unsigned long |
| 668 | mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg, |
| 669 | int nid, unsigned int lru_mask) |
| 670 | { |
| 671 | return 0; |
| 672 | } |
| 673 | |
| 674 | static inline void |
| 675 | mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p) |
| 676 | { |
| 677 | } |
| 678 | |
| 679 | static inline void lock_page_memcg(struct page *page) |
| 680 | { |
| 681 | } |
| 682 | |
| 683 | static inline void unlock_page_memcg(struct page *page) |
| 684 | { |
| 685 | } |
| 686 | |
| 687 | static inline void mem_cgroup_handle_over_high(void) |
| 688 | { |
| 689 | } |
| 690 | |
| 691 | static inline void mem_cgroup_oom_enable(void) |
| 692 | { |
| 693 | } |
| 694 | |
| 695 | static inline void mem_cgroup_oom_disable(void) |
| 696 | { |
| 697 | } |
| 698 | |
| 699 | static inline bool task_in_memcg_oom(struct task_struct *p) |
| 700 | { |
| 701 | return false; |
| 702 | } |
| 703 | |
| 704 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
| 705 | { |
| 706 | return false; |
| 707 | } |
| 708 | |
| 709 | static inline void mem_cgroup_inc_page_stat(struct page *page, |
| 710 | enum mem_cgroup_stat_index idx) |
| 711 | { |
| 712 | } |
| 713 | |
| 714 | static inline void mem_cgroup_dec_page_stat(struct page *page, |
| 715 | enum mem_cgroup_stat_index idx) |
| 716 | { |
| 717 | } |
| 718 | |
| 719 | static inline |
| 720 | unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order, |
| 721 | gfp_t gfp_mask, |
| 722 | unsigned long *total_scanned) |
| 723 | { |
| 724 | return 0; |
| 725 | } |
| 726 | |
| 727 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
| 728 | { |
| 729 | } |
| 730 | |
| 731 | static inline |
| 732 | void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx) |
| 733 | { |
| 734 | } |
| 735 | #endif /* CONFIG_MEMCG */ |
| 736 | |
| 737 | #ifdef CONFIG_CGROUP_WRITEBACK |
| 738 | |
| 739 | struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); |
| 740 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
| 741 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
| 742 | unsigned long *pheadroom, unsigned long *pdirty, |
| 743 | unsigned long *pwriteback); |
| 744 | |
| 745 | #else /* CONFIG_CGROUP_WRITEBACK */ |
| 746 | |
| 747 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) |
| 748 | { |
| 749 | return NULL; |
| 750 | } |
| 751 | |
| 752 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
| 753 | unsigned long *pfilepages, |
| 754 | unsigned long *pheadroom, |
| 755 | unsigned long *pdirty, |
| 756 | unsigned long *pwriteback) |
| 757 | { |
| 758 | } |
| 759 | |
| 760 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
| 761 | |
| 762 | struct sock; |
| 763 | void sock_update_memcg(struct sock *sk); |
| 764 | void sock_release_memcg(struct sock *sk); |
| 765 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
| 766 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
| 767 | #ifdef CONFIG_MEMCG |
| 768 | extern struct static_key_false memcg_sockets_enabled_key; |
| 769 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) |
| 770 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
| 771 | { |
| 772 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
| 773 | return true; |
| 774 | do { |
| 775 | if (time_before(jiffies, memcg->socket_pressure)) |
| 776 | return true; |
| 777 | } while ((memcg = parent_mem_cgroup(memcg))); |
| 778 | return false; |
| 779 | } |
| 780 | #else |
| 781 | #define mem_cgroup_sockets_enabled 0 |
| 782 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
| 783 | { |
| 784 | return false; |
| 785 | } |
| 786 | #endif |
| 787 | |
| 788 | #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB) |
| 789 | extern struct static_key_false memcg_kmem_enabled_key; |
| 790 | |
| 791 | extern int memcg_nr_cache_ids; |
| 792 | void memcg_get_cache_ids(void); |
| 793 | void memcg_put_cache_ids(void); |
| 794 | |
| 795 | /* |
| 796 | * Helper macro to loop through all memcg-specific caches. Callers must still |
| 797 | * check if the cache is valid (it is either valid or NULL). |
| 798 | * the slab_mutex must be held when looping through those caches |
| 799 | */ |
| 800 | #define for_each_memcg_cache_index(_idx) \ |
| 801 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
| 802 | |
| 803 | static inline bool memcg_kmem_enabled(void) |
| 804 | { |
| 805 | return static_branch_unlikely(&memcg_kmem_enabled_key); |
| 806 | } |
| 807 | |
| 808 | /* |
| 809 | * In general, we'll do everything in our power to not incur in any overhead |
| 810 | * for non-memcg users for the kmem functions. Not even a function call, if we |
| 811 | * can avoid it. |
| 812 | * |
| 813 | * Therefore, we'll inline all those functions so that in the best case, we'll |
| 814 | * see that kmemcg is off for everybody and proceed quickly. If it is on, |
| 815 | * we'll still do most of the flag checking inline. We check a lot of |
| 816 | * conditions, but because they are pretty simple, they are expected to be |
| 817 | * fast. |
| 818 | */ |
| 819 | int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order, |
| 820 | struct mem_cgroup *memcg); |
| 821 | int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order); |
| 822 | void __memcg_kmem_uncharge(struct page *page, int order); |
| 823 | |
| 824 | /* |
| 825 | * helper for accessing a memcg's index. It will be used as an index in the |
| 826 | * child cache array in kmem_cache, and also to derive its name. This function |
| 827 | * will return -1 when this is not a kmem-limited memcg. |
| 828 | */ |
| 829 | static inline int memcg_cache_id(struct mem_cgroup *memcg) |
| 830 | { |
| 831 | return memcg ? memcg->kmemcg_id : -1; |
| 832 | } |
| 833 | |
| 834 | struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp); |
| 835 | void __memcg_kmem_put_cache(struct kmem_cache *cachep); |
| 836 | |
| 837 | static inline bool __memcg_kmem_bypass(void) |
| 838 | { |
| 839 | if (!memcg_kmem_enabled()) |
| 840 | return true; |
| 841 | if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD)) |
| 842 | return true; |
| 843 | return false; |
| 844 | } |
| 845 | |
| 846 | /** |
| 847 | * memcg_kmem_charge: charge a kmem page |
| 848 | * @page: page to charge |
| 849 | * @gfp: reclaim mode |
| 850 | * @order: allocation order |
| 851 | * |
| 852 | * Returns 0 on success, an error code on failure. |
| 853 | */ |
| 854 | static __always_inline int memcg_kmem_charge(struct page *page, |
| 855 | gfp_t gfp, int order) |
| 856 | { |
| 857 | if (__memcg_kmem_bypass()) |
| 858 | return 0; |
| 859 | if (!(gfp & __GFP_ACCOUNT)) |
| 860 | return 0; |
| 861 | return __memcg_kmem_charge(page, gfp, order); |
| 862 | } |
| 863 | |
| 864 | /** |
| 865 | * memcg_kmem_uncharge: uncharge a kmem page |
| 866 | * @page: page to uncharge |
| 867 | * @order: allocation order |
| 868 | */ |
| 869 | static __always_inline void memcg_kmem_uncharge(struct page *page, int order) |
| 870 | { |
| 871 | if (memcg_kmem_enabled()) |
| 872 | __memcg_kmem_uncharge(page, order); |
| 873 | } |
| 874 | |
| 875 | /** |
| 876 | * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation |
| 877 | * @cachep: the original global kmem cache |
| 878 | * |
| 879 | * All memory allocated from a per-memcg cache is charged to the owner memcg. |
| 880 | */ |
| 881 | static __always_inline struct kmem_cache * |
| 882 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) |
| 883 | { |
| 884 | if (__memcg_kmem_bypass()) |
| 885 | return cachep; |
| 886 | return __memcg_kmem_get_cache(cachep, gfp); |
| 887 | } |
| 888 | |
| 889 | static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep) |
| 890 | { |
| 891 | if (memcg_kmem_enabled()) |
| 892 | __memcg_kmem_put_cache(cachep); |
| 893 | } |
| 894 | |
| 895 | /** |
| 896 | * memcg_kmem_update_page_stat - update kmem page state statistics |
| 897 | * @page: the page |
| 898 | * @idx: page state item to account |
| 899 | * @val: number of pages (positive or negative) |
| 900 | */ |
| 901 | static inline void memcg_kmem_update_page_stat(struct page *page, |
| 902 | enum mem_cgroup_stat_index idx, int val) |
| 903 | { |
| 904 | if (memcg_kmem_enabled() && page->mem_cgroup) |
| 905 | this_cpu_add(page->mem_cgroup->stat->count[idx], val); |
| 906 | } |
| 907 | |
| 908 | #else |
| 909 | #define for_each_memcg_cache_index(_idx) \ |
| 910 | for (; NULL; ) |
| 911 | |
| 912 | static inline bool memcg_kmem_enabled(void) |
| 913 | { |
| 914 | return false; |
| 915 | } |
| 916 | |
| 917 | static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order) |
| 918 | { |
| 919 | return 0; |
| 920 | } |
| 921 | |
| 922 | static inline void memcg_kmem_uncharge(struct page *page, int order) |
| 923 | { |
| 924 | } |
| 925 | |
| 926 | static inline int memcg_cache_id(struct mem_cgroup *memcg) |
| 927 | { |
| 928 | return -1; |
| 929 | } |
| 930 | |
| 931 | static inline void memcg_get_cache_ids(void) |
| 932 | { |
| 933 | } |
| 934 | |
| 935 | static inline void memcg_put_cache_ids(void) |
| 936 | { |
| 937 | } |
| 938 | |
| 939 | static inline struct kmem_cache * |
| 940 | memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp) |
| 941 | { |
| 942 | return cachep; |
| 943 | } |
| 944 | |
| 945 | static inline void memcg_kmem_put_cache(struct kmem_cache *cachep) |
| 946 | { |
| 947 | } |
| 948 | |
| 949 | static inline void memcg_kmem_update_page_stat(struct page *page, |
| 950 | enum mem_cgroup_stat_index idx, int val) |
| 951 | { |
| 952 | } |
| 953 | #endif /* CONFIG_MEMCG && !CONFIG_SLOB */ |
| 954 | |
| 955 | #endif /* _LINUX_MEMCONTROL_H */ |