include/linux/memcontrol.h

   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_SOCK = MEM_CGROUP_STAT_NSTATS,
  56         MEMCG_SLAB_RECLAIMABLE,
  57         MEMCG_SLAB_UNRECLAIMABLE,
  58         MEMCG_NR_STAT,
  59 };
  60
  61 struct mem_cgroup_reclaim_cookie {
  62         struct zone *zone;
  63         int priority;
  64         unsigned int generation;
  65 };
  66
  67 enum mem_cgroup_events_index {
  68         MEM_CGROUP_EVENTS_PGPGIN,       /* # of pages paged in */
  69         MEM_CGROUP_EVENTS_PGPGOUT,      /* # of pages paged out */
  70         MEM_CGROUP_EVENTS_PGFAULT,      /* # of page-faults */
  71         MEM_CGROUP_EVENTS_PGMAJFAULT,   /* # of major page-faults */
  72         MEM_CGROUP_EVENTS_NSTATS,
  73         /* default hierarchy events */
  74         MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
  75         MEMCG_HIGH,
  76         MEMCG_MAX,
  77         MEMCG_OOM,
  78         MEMCG_NR_EVENTS,
  79 };
  80
  81 /*
  82  * Per memcg event counter is incremented at every pagein/pageout. With THP,
  83  * it will be incremated by the number of pages. This counter is used for
  84  * for trigger some periodic events. This is straightforward and better
  85  * than using jiffies etc. to handle periodic memcg event.
  86  */
  87 enum mem_cgroup_events_target {
  88         MEM_CGROUP_TARGET_THRESH,
  89         MEM_CGROUP_TARGET_SOFTLIMIT,
  90         MEM_CGROUP_TARGET_NUMAINFO,
  91         MEM_CGROUP_NTARGETS,
  92 };
  93
  94 #ifdef CONFIG_MEMCG
  95
  96 #define MEM_CGROUP_ID_SHIFT     16
  97 #define MEM_CGROUP_ID_MAX       USHRT_MAX
  98
  99 struct mem_cgroup_stat_cpu {
 100         long count[MEMCG_NR_STAT];
 101         unsigned long events[MEMCG_NR_EVENTS];
 102         unsigned long nr_page_events;
 103         unsigned long targets[MEM_CGROUP_NTARGETS];
 104 };
 105
 106 struct mem_cgroup_reclaim_iter {
 107         struct mem_cgroup *position;
 108         /* scan generation, increased every round-trip */
 109         unsigned int generation;
 110 };
 111
 112 /*
 113  * per-zone information in memory controller.
 114  */
 115 struct mem_cgroup_per_zone {
 116         struct lruvec           lruvec;
 117         unsigned long           lru_size[NR_LRU_LISTS];
 118
 119         struct mem_cgroup_reclaim_iter  iter[DEF_PRIORITY + 1];
 120
 121         struct rb_node          tree_node;      /* RB tree node */
 122         unsigned long           usage_in_excess;/* Set to the value by which */
 123                                                 /* the soft limit is exceeded*/
 124         bool                    on_tree;
 125         struct mem_cgroup       *memcg;         /* Back pointer, we cannot */
 126                                                 /* use container_of        */
 127 };
 128
 129 struct mem_cgroup_per_node {
 130         struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
 131 };
 132
 133 struct mem_cgroup_threshold {
 134         struct eventfd_ctx *eventfd;
 135         unsigned long threshold;
 136 };
 137
 138 /* For threshold */
 139 struct mem_cgroup_threshold_ary {
 140         /* An array index points to threshold just below or equal to usage. */
 141         int current_threshold;
 142         /* Size of entries[] */
 143         unsigned int size;
 144         /* Array of thresholds */
 145         struct mem_cgroup_threshold entries[0];
 146 };
 147
 148 struct mem_cgroup_thresholds {
 149         /* Primary thresholds array */
 150         struct mem_cgroup_threshold_ary *primary;
 151         /*
 152          * Spare threshold array.
 153          * This is needed to make mem_cgroup_unregister_event() "never fail".
 154          * It must be able to store at least primary->size - 1 entries.
 155          */
 156         struct mem_cgroup_threshold_ary *spare;
 157 };
 158
 159 enum memcg_kmem_state {
 160         KMEM_NONE,
 161         KMEM_ALLOCATED,
 162         KMEM_ONLINE,
 163 };
 164
 165 /*
 166  * The memory controller data structure. The memory controller controls both
 167  * page cache and RSS per cgroup. We would eventually like to provide
 168  * statistics based on the statistics developed by Rik Van Riel for clock-pro,
 169  * to help the administrator determine what knobs to tune.
 170  */
 171 struct mem_cgroup {
 172         struct cgroup_subsys_state css;
 173
 174         /* Accounted resources */
 175         struct page_counter memory;
 176         struct page_counter swap;
 177
 178         /* Legacy consumer-oriented counters */
 179         struct page_counter memsw;
 180         struct page_counter kmem;
 181         struct page_counter tcpmem;
 182
 183         /* Normal memory consumption range */
 184         unsigned long low;
 185         unsigned long high;
 186
 187         /* Range enforcement for interrupt charges */
 188         struct work_struct high_work;
 189
 190         unsigned long soft_limit;
 191
 192         /* vmpressure notifications */
 193         struct vmpressure vmpressure;
 194
 195         /*
 196          * Should the accounting and control be hierarchical, per subtree?
 197          */
 198         bool use_hierarchy;
 199
 200         /* protected by memcg_oom_lock */
 201         bool            oom_lock;
 202         int             under_oom;
 203
 204         int     swappiness;
 205         /* OOM-Killer disable */
 206         int             oom_kill_disable;
 207
 208         /* handle for "memory.events" */
 209         struct cgroup_file events_file;
 210
 211         /* protect arrays of thresholds */
 212         struct mutex thresholds_lock;
 213
 214         /* thresholds for memory usage. RCU-protected */
 215         struct mem_cgroup_thresholds thresholds;
 216
 217         /* thresholds for mem+swap usage. RCU-protected */
 218         struct mem_cgroup_thresholds memsw_thresholds;
 219
 220         /* For oom notifier event fd */
 221         struct list_head oom_notify;
 222
 223         /*
 224          * Should we move charges of a task when a task is moved into this
 225          * mem_cgroup ? And what type of charges should we move ?
 226          */
 227         unsigned long move_charge_at_immigrate;
 228         /*
 229          * set > 0 if pages under this cgroup are moving to other cgroup.
 230          */
 231         atomic_t                moving_account;
 232         /* taken only while moving_account > 0 */
 233         spinlock_t              move_lock;
 234         struct task_struct      *move_lock_task;
 235         unsigned long           move_lock_flags;
 236         /*
 237          * percpu counter.
 238          */
 239         struct mem_cgroup_stat_cpu __percpu *stat;
 240
 241         unsigned long           socket_pressure;
 242
 243         /* Legacy tcp memory accounting */
 244         bool                    tcpmem_active;
 245         int                     tcpmem_pressure;
 246
 247 #ifndef CONFIG_SLOB
 248         /* Index in the kmem_cache->memcg_params.memcg_caches array */
 249         int kmemcg_id;
 250         enum memcg_kmem_state kmem_state;
 251 #endif
 252
 253         int last_scanned_node;
 254 #if MAX_NUMNODES > 1
 255         nodemask_t      scan_nodes;
 256         atomic_t        numainfo_events;
 257         atomic_t        numainfo_updating;
 258 #endif
 259
 260 #ifdef CONFIG_CGROUP_WRITEBACK
 261         struct list_head cgwb_list;
 262         struct wb_domain cgwb_domain;
 263 #endif
 264
 265         /* List of events which userspace want to receive */
 266         struct list_head event_list;
 267         spinlock_t event_list_lock;
 268
 269         struct mem_cgroup_per_node *nodeinfo[0];
 270         /* WARNING: nodeinfo must be the last member here */
 271 };
 272
 273 extern struct mem_cgroup *root_mem_cgroup;
 274
 275 static inline bool mem_cgroup_disabled(void)
 276 {
 277         return !cgroup_subsys_enabled(memory_cgrp_subsys);
 278 }
 279
 280 /**
 281  * mem_cgroup_events - count memory events against a cgroup
 282  * @memcg: the memory cgroup
 283  * @idx: the event index
 284  * @nr: the number of events to account for
 285  */
 286 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 287                        enum mem_cgroup_events_index idx,
 288                        unsigned int nr)
 289 {
 290         this_cpu_add(memcg->stat->events[idx], nr);
 291         cgroup_file_notify(&memcg->events_file);
 292 }
 293
 294 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
 295
 296 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 297                           gfp_t gfp_mask, struct mem_cgroup **memcgp,
 298                           bool compound);
 299 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
 300                               bool lrucare, bool compound);
 301 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
 302                 bool compound);
 303 void mem_cgroup_uncharge(struct page *page);
 304 void mem_cgroup_uncharge_list(struct list_head *page_list);
 305
 306 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
 307
 308 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
 309 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
 310
 311 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
 312 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 313
 314 static inline
 315 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 316         return css ? container_of(css, struct mem_cgroup, css) : NULL;
 317 }
 318
 319 #define mem_cgroup_from_counter(counter, member)        \
 320         container_of(counter, struct mem_cgroup, member)
 321
 322 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 323                                    struct mem_cgroup *,
 324                                    struct mem_cgroup_reclaim_cookie *);
 325 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 326
 327 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 328 {
 329         if (mem_cgroup_disabled())
 330                 return 0;
 331
 332         return memcg->css.id;
 333 }
 334
 335 /**
 336  * mem_cgroup_from_id - look up a memcg from an id
 337  * @id: the id to look up
 338  *
 339  * Caller must hold rcu_read_lock() and use css_tryget() as necessary.
 340  */
 341 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
 342 {
 343         struct cgroup_subsys_state *css;
 344
 345         css = css_from_id(id, &memory_cgrp_subsys);
 346         return mem_cgroup_from_css(css);
 347 }
 348
 349 /**
 350  * parent_mem_cgroup - find the accounting parent of a memcg
 351  * @memcg: memcg whose parent to find
 352  *
 353  * Returns the parent memcg, or NULL if this is the root or the memory
 354  * controller is in legacy no-hierarchy mode.
 355  */
 356 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
 357 {
 358         if (!memcg->memory.parent)
 359                 return NULL;
 360         return mem_cgroup_from_counter(memcg->memory.parent, memory);
 361 }
 362
 363 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 364                               struct mem_cgroup *root)
 365 {
 366         if (root == memcg)
 367                 return true;
 368         if (!root->use_hierarchy)
 369                 return false;
 370         return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 371 }
 372
 373 static inline bool mm_match_cgroup(struct mm_struct *mm,
 374                                    struct mem_cgroup *memcg)
 375 {
 376         struct mem_cgroup *task_memcg;
 377         bool match = false;
 378
 379         rcu_read_lock();
 380         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 381         if (task_memcg)
 382                 match = mem_cgroup_is_descendant(task_memcg, memcg);
 383         rcu_read_unlock();
 384         return match;
 385 }
 386
 387 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 388 ino_t page_cgroup_ino(struct page *page);
 389
 390 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 391 {
 392         if (mem_cgroup_disabled())
 393                 return true;
 394         return !!(memcg->css.flags & CSS_ONLINE);
 395 }
 396
 397 /*
 398  * For memory reclaim.
 399  */
 400 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 401
 402 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 403                 int nr_pages);
 404
 405 static inline
 406 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 407 {
 408         struct mem_cgroup_per_zone *mz;
 409
 410         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 411         return mz->lru_size[lru];
 412 }
 413
 414 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 415 {
 416         unsigned long inactive_ratio;
 417         unsigned long inactive;
 418         unsigned long active;
 419         unsigned long gb;
 420
 421         inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
 422         active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
 423
 424         gb = (inactive + active) >> (30 - PAGE_SHIFT);
 425         if (gb)
 426                 inactive_ratio = int_sqrt(10 * gb);
 427         else
 428                 inactive_ratio = 1;
 429
 430         return inactive * inactive_ratio < active;
 431 }
 432
 433 void mem_cgroup_handle_over_high(void);
 434
 435 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 436                                 struct task_struct *p);
 437
 438 static inline void mem_cgroup_oom_enable(void)
 439 {
 440         WARN_ON(current->memcg_may_oom);
 441         current->memcg_may_oom = 1;
 442 }
 443
 444 static inline void mem_cgroup_oom_disable(void)
 445 {
 446         WARN_ON(!current->memcg_may_oom);
 447         current->memcg_may_oom = 0;
 448 }
 449
 450 static inline bool task_in_memcg_oom(struct task_struct *p)
 451 {
 452         return p->memcg_in_oom;
 453 }
 454
 455 bool mem_cgroup_oom_synchronize(bool wait);
 456
 457 #ifdef CONFIG_MEMCG_SWAP
 458 extern int do_swap_account;
 459 #endif
 460
 461 void lock_page_memcg(struct page *page);
 462 void unlock_page_memcg(struct page *page);
 463
 464 /**
 465  * mem_cgroup_update_page_stat - update page state statistics
 466  * @page: the page
 467  * @idx: page state item to account
 468  * @val: number of pages (positive or negative)
 469  *
 470  * The @page must be locked or the caller must use lock_page_memcg()
 471  * to prevent double accounting when the page is concurrently being
 472  * moved to another memcg:
 473  *
 474  *   lock_page(page) or lock_page_memcg(page)
 475  *   if (TestClearPageState(page))
 476  *     mem_cgroup_update_page_stat(page, state, -1);
 477  *   unlock_page(page) or unlock_page_memcg(page)
 478  */
 479 static inline void mem_cgroup_update_page_stat(struct page *page,
 480                                  enum mem_cgroup_stat_index idx, int val)
 481 {
 482         VM_BUG_ON(!(rcu_read_lock_held() || PageLocked(page)));
 483
 484         if (page->mem_cgroup)
 485                 this_cpu_add(page->mem_cgroup->stat->count[idx], val);
 486 }
 487
 488 static inline void mem_cgroup_inc_page_stat(struct page *page,
 489                                             enum mem_cgroup_stat_index idx)
 490 {
 491         mem_cgroup_update_page_stat(page, idx, 1);
 492 }
 493
 494 static inline void mem_cgroup_dec_page_stat(struct page *page,
 495                                             enum mem_cgroup_stat_index idx)
 496 {
 497         mem_cgroup_update_page_stat(page, idx, -1);
 498 }
 499
 500 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 501                                                 gfp_t gfp_mask,
 502                                                 unsigned long *total_scanned);
 503
 504 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 505                                              enum vm_event_item idx)
 506 {
 507         struct mem_cgroup *memcg;
 508
 509         if (mem_cgroup_disabled())
 510                 return;
 511
 512         rcu_read_lock();
 513         memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 514         if (unlikely(!memcg))
 515                 goto out;
 516
 517         switch (idx) {
 518         case PGFAULT:
 519                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
 520                 break;
 521         case PGMAJFAULT:
 522                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
 523                 break;
 524         default:
 525                 BUG();
 526         }
 527 out:
 528         rcu_read_unlock();
 529 }
 530 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 531 void mem_cgroup_split_huge_fixup(struct page *head);
 532 #endif
 533
 534 #else /* CONFIG_MEMCG */
 535
 536 #define MEM_CGROUP_ID_SHIFT     0
 537 #define MEM_CGROUP_ID_MAX       0
 538
 539 struct mem_cgroup;
 540
 541 static inline bool mem_cgroup_disabled(void)
 542 {
 543         return true;
 544 }
 545
 546 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 547                                      enum mem_cgroup_events_index idx,
 548                                      unsigned int nr)
 549 {
 550 }
 551
 552 static inline bool mem_cgroup_low(struct mem_cgroup *root,
 553                                   struct mem_cgroup *memcg)
 554 {
 555         return false;
 556 }
 557
 558 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 559                                         gfp_t gfp_mask,
 560                                         struct mem_cgroup **memcgp,
 561                                         bool compound)
 562 {
 563         *memcgp = NULL;
 564         return 0;
 565 }
 566
 567 static inline void mem_cgroup_commit_charge(struct page *page,
 568                                             struct mem_cgroup *memcg,
 569                                             bool lrucare, bool compound)
 570 {
 571 }
 572
 573 static inline void mem_cgroup_cancel_charge(struct page *page,
 574                                             struct mem_cgroup *memcg,
 575                                             bool compound)
 576 {
 577 }
 578
 579 static inline void mem_cgroup_uncharge(struct page *page)
 580 {
 581 }
 582
 583 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 584 {
 585 }
 586
 587 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
 588 {
 589 }
 590
 591 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 592                                                     struct mem_cgroup *memcg)
 593 {
 594         return &zone->lruvec;
 595 }
 596
 597 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 598                                                     struct zone *zone)
 599 {
 600         return &zone->lruvec;
 601 }
 602
 603 static inline bool mm_match_cgroup(struct mm_struct *mm,
 604                 struct mem_cgroup *memcg)
 605 {
 606         return true;
 607 }
 608
 609 static inline bool task_in_mem_cgroup(struct task_struct *task,
 610                                       const struct mem_cgroup *memcg)
 611 {
 612         return true;
 613 }
 614
 615 static inline struct mem_cgroup *
 616 mem_cgroup_iter(struct mem_cgroup *root,
 617                 struct mem_cgroup *prev,
 618                 struct mem_cgroup_reclaim_cookie *reclaim)
 619 {
 620         return NULL;
 621 }
 622
 623 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 624                                          struct mem_cgroup *prev)
 625 {
 626 }
 627
 628 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 629 {
 630         return 0;
 631 }
 632
 633 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
 634 {
 635         WARN_ON_ONCE(id);
 636         /* XXX: This should always return root_mem_cgroup */
 637         return NULL;
 638 }
 639
 640 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 641 {
 642         return true;
 643 }
 644
 645 static inline bool
 646 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 647 {
 648         return true;
 649 }
 650
 651 static inline unsigned long
 652 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 653 {
 654         return 0;
 655 }
 656
 657 static inline void
 658 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 659                               int increment)
 660 {
 661 }
 662
 663 static inline void
 664 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 665 {
 666 }
 667
 668 static inline void lock_page_memcg(struct page *page)
 669 {
 670 }
 671
 672 static inline void unlock_page_memcg(struct page *page)
 673 {
 674 }
 675
 676 static inline void mem_cgroup_handle_over_high(void)
 677 {
 678 }
 679
 680 static inline void mem_cgroup_oom_enable(void)
 681 {
 682 }
 683
 684 static inline void mem_cgroup_oom_disable(void)
 685 {
 686 }
 687
 688 static inline bool task_in_memcg_oom(struct task_struct *p)
 689 {
 690         return false;
 691 }
 692
 693 static inline bool mem_cgroup_oom_synchronize(bool wait)
 694 {
 695         return false;
 696 }
 697
 698 static inline void mem_cgroup_inc_page_stat(struct page *page,
 699                                             enum mem_cgroup_stat_index idx)
 700 {
 701 }
 702
 703 static inline void mem_cgroup_dec_page_stat(struct page *page,
 704                                             enum mem_cgroup_stat_index idx)
 705 {
 706 }
 707
 708 static inline
 709 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 710                                             gfp_t gfp_mask,
 711                                             unsigned long *total_scanned)
 712 {
 713         return 0;
 714 }
 715
 716 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 717 {
 718 }
 719
 720 static inline
 721 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 722 {
 723 }
 724 #endif /* CONFIG_MEMCG */
 725
 726 #ifdef CONFIG_CGROUP_WRITEBACK
 727
 728 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
 729 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
 730 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
 731                          unsigned long *pheadroom, unsigned long *pdirty,
 732                          unsigned long *pwriteback);
 733
 734 #else   /* CONFIG_CGROUP_WRITEBACK */
 735
 736 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
 737 {
 738         return NULL;
 739 }
 740
 741 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
 742                                        unsigned long *pfilepages,
 743                                        unsigned long *pheadroom,
 744                                        unsigned long *pdirty,
 745                                        unsigned long *pwriteback)
 746 {
 747 }
 748
 749 #endif  /* CONFIG_CGROUP_WRITEBACK */
 750
 751 struct sock;
 752 void sock_update_memcg(struct sock *sk);
 753 void sock_release_memcg(struct sock *sk);
 754 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 755 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 756 #ifdef CONFIG_MEMCG
 757 extern struct static_key_false memcg_sockets_enabled_key;
 758 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
 759 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 760 {
 761         if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
 762                 return true;
 763         do {
 764                 if (time_before(jiffies, memcg->socket_pressure))
 765                         return true;
 766         } while ((memcg = parent_mem_cgroup(memcg)));
 767         return false;
 768 }
 769 #else
 770 #define mem_cgroup_sockets_enabled 0
 771 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 772 {
 773         return false;
 774 }
 775 #endif
 776
 777 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
 778 extern struct static_key_false memcg_kmem_enabled_key;
 779
 780 extern int memcg_nr_cache_ids;
 781 void memcg_get_cache_ids(void);
 782 void memcg_put_cache_ids(void);
 783
 784 /*
 785  * Helper macro to loop through all memcg-specific caches. Callers must still
 786  * check if the cache is valid (it is either valid or NULL).
 787  * the slab_mutex must be held when looping through those caches
 788  */
 789 #define for_each_memcg_cache_index(_idx)        \
 790         for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
 791
 792 static inline bool memcg_kmem_enabled(void)
 793 {
 794         return static_branch_unlikely(&memcg_kmem_enabled_key);
 795 }
 796
 797 static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
 798 {
 799         return memcg->kmem_state == KMEM_ONLINE;
 800 }
 801
 802 /*
 803  * In general, we'll do everything in our power to not incur in any overhead
 804  * for non-memcg users for the kmem functions. Not even a function call, if we
 805  * can avoid it.
 806  *
 807  * Therefore, we'll inline all those functions so that in the best case, we'll
 808  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 809  * we'll still do most of the flag checking inline. We check a lot of
 810  * conditions, but because they are pretty simple, they are expected to be
 811  * fast.
 812  */
 813 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
 814                               struct mem_cgroup *memcg);
 815 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
 816 void __memcg_kmem_uncharge(struct page *page, int order);
 817
 818 /*
 819  * helper for accessing a memcg's index. It will be used as an index in the
 820  * child cache array in kmem_cache, and also to derive its name. This function
 821  * will return -1 when this is not a kmem-limited memcg.
 822  */
 823 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 824 {
 825         return memcg ? memcg->kmemcg_id : -1;
 826 }
 827
 828 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
 829 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
 830
 831 static inline bool __memcg_kmem_bypass(void)
 832 {
 833         if (!memcg_kmem_enabled())
 834                 return true;
 835         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 836                 return true;
 837         return false;
 838 }
 839
 840 /**
 841  * memcg_kmem_charge: charge a kmem page
 842  * @page: page to charge
 843  * @gfp: reclaim mode
 844  * @order: allocation order
 845  *
 846  * Returns 0 on success, an error code on failure.
 847  */
 848 static __always_inline int memcg_kmem_charge(struct page *page,
 849                                              gfp_t gfp, int order)
 850 {
 851         if (__memcg_kmem_bypass())
 852                 return 0;
 853         if (!(gfp & __GFP_ACCOUNT))
 854                 return 0;
 855         return __memcg_kmem_charge(page, gfp, order);
 856 }
 857
 858 /**
 859  * memcg_kmem_uncharge: uncharge a kmem page
 860  * @page: page to uncharge
 861  * @order: allocation order
 862  */
 863 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
 864 {
 865         if (memcg_kmem_enabled())
 866                 __memcg_kmem_uncharge(page, order);
 867 }
 868
 869 /**
 870  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 871  * @cachep: the original global kmem cache
 872  *
 873  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 874  */
 875 static __always_inline struct kmem_cache *
 876 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 877 {
 878         if (__memcg_kmem_bypass())
 879                 return cachep;
 880         return __memcg_kmem_get_cache(cachep, gfp);
 881 }
 882
 883 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 884 {
 885         if (memcg_kmem_enabled())
 886                 __memcg_kmem_put_cache(cachep);
 887 }
 888
 889 /**
 890  * memcg_kmem_update_page_stat - update kmem page state statistics
 891  * @page: the page
 892  * @idx: page state item to account
 893  * @val: number of pages (positive or negative)
 894  */
 895 static inline void memcg_kmem_update_page_stat(struct page *page,
 896                                 enum mem_cgroup_stat_index idx, int val)
 897 {
 898         if (memcg_kmem_enabled() && page->mem_cgroup)
 899                 this_cpu_add(page->mem_cgroup->stat->count[idx], val);
 900 }
 901
 902 #else
 903 #define for_each_memcg_cache_index(_idx)        \
 904         for (; NULL; )
 905
 906 static inline bool memcg_kmem_enabled(void)
 907 {
 908         return false;
 909 }
 910
 911 static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
 912 {
 913         return false;
 914 }
 915
 916 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
 917 {
 918         return 0;
 919 }
 920
 921 static inline void memcg_kmem_uncharge(struct page *page, int order)
 922 {
 923 }
 924
 925 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 926 {
 927         return -1;
 928 }
 929
 930 static inline void memcg_get_cache_ids(void)
 931 {
 932 }
 933
 934 static inline void memcg_put_cache_ids(void)
 935 {
 936 }
 937
 938 static inline struct kmem_cache *
 939 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 940 {
 941         return cachep;
 942 }
 943
 944 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 945 {
 946 }
 947
 948 static inline void memcg_kmem_update_page_stat(struct page *page,
 949                                 enum mem_cgroup_stat_index idx, int val)
 950 {
 951 }
 952 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
 953
 954 #endif /* _LINUX_MEMCONTROL_H */