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