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
  32 struct mem_cgroup;
  33 struct page;
  34 struct mm_struct;
  35 struct kmem_cache;
  36
  37 /*
  38  * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
  39  * These two lists should keep in accord with each other.
  40  */
  41 enum mem_cgroup_stat_index {
  42         /*
  43          * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
  44          */
  45         MEM_CGROUP_STAT_CACHE,          /* # of pages charged as cache */
  46         MEM_CGROUP_STAT_RSS,            /* # of pages charged as anon rss */
  47         MEM_CGROUP_STAT_RSS_HUGE,       /* # of pages charged as anon huge */
  48         MEM_CGROUP_STAT_FILE_MAPPED,    /* # of pages charged as file rss */
  49         MEM_CGROUP_STAT_DIRTY,          /* # of dirty pages in page cache */
  50         MEM_CGROUP_STAT_WRITEBACK,      /* # of pages under writeback */
  51         MEM_CGROUP_STAT_SWAP,           /* # of pages, swapped out */
  52         MEM_CGROUP_STAT_NSTATS,
  53         /* default hierarchy stats */
  54         MEMCG_SOCK = MEM_CGROUP_STAT_NSTATS,
  55         MEMCG_NR_STAT,
  56 };
  57
  58 struct mem_cgroup_reclaim_cookie {
  59         struct zone *zone;
  60         int priority;
  61         unsigned int generation;
  62 };
  63
  64 enum mem_cgroup_events_index {
  65         MEM_CGROUP_EVENTS_PGPGIN,       /* # of pages paged in */
  66         MEM_CGROUP_EVENTS_PGPGOUT,      /* # of pages paged out */
  67         MEM_CGROUP_EVENTS_PGFAULT,      /* # of page-faults */
  68         MEM_CGROUP_EVENTS_PGMAJFAULT,   /* # of major page-faults */
  69         MEM_CGROUP_EVENTS_NSTATS,
  70         /* default hierarchy events */
  71         MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
  72         MEMCG_HIGH,
  73         MEMCG_MAX,
  74         MEMCG_OOM,
  75         MEMCG_NR_EVENTS,
  76 };
  77
  78 /*
  79  * Per memcg event counter is incremented at every pagein/pageout. With THP,
  80  * it will be incremated by the number of pages. This counter is used for
  81  * for trigger some periodic events. This is straightforward and better
  82  * than using jiffies etc. to handle periodic memcg event.
  83  */
  84 enum mem_cgroup_events_target {
  85         MEM_CGROUP_TARGET_THRESH,
  86         MEM_CGROUP_TARGET_SOFTLIMIT,
  87         MEM_CGROUP_TARGET_NUMAINFO,
  88         MEM_CGROUP_NTARGETS,
  89 };
  90
  91 #ifdef CONFIG_MEMCG
  92
  93 #define MEM_CGROUP_ID_SHIFT     16
  94 #define MEM_CGROUP_ID_MAX       USHRT_MAX
  95
  96 struct mem_cgroup_stat_cpu {
  97         long count[MEMCG_NR_STAT];
  98         unsigned long events[MEMCG_NR_EVENTS];
  99         unsigned long nr_page_events;
 100         unsigned long targets[MEM_CGROUP_NTARGETS];
 101 };
 102
 103 struct mem_cgroup_reclaim_iter {
 104         struct mem_cgroup *position;
 105         /* scan generation, increased every round-trip */
 106         unsigned int generation;
 107 };
 108
 109 /*
 110  * per-zone information in memory controller.
 111  */
 112 struct mem_cgroup_per_zone {
 113         struct lruvec           lruvec;
 114         unsigned long           lru_size[NR_LRU_LISTS];
 115
 116         struct mem_cgroup_reclaim_iter  iter[DEF_PRIORITY + 1];
 117
 118         struct rb_node          tree_node;      /* RB tree node */
 119         unsigned long           usage_in_excess;/* Set to the value by which */
 120                                                 /* the soft limit is exceeded*/
 121         bool                    on_tree;
 122         struct mem_cgroup       *memcg;         /* Back pointer, we cannot */
 123                                                 /* use container_of        */
 124 };
 125
 126 struct mem_cgroup_per_node {
 127         struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
 128 };
 129
 130 struct mem_cgroup_threshold {
 131         struct eventfd_ctx *eventfd;
 132         unsigned long threshold;
 133 };
 134
 135 /* For threshold */
 136 struct mem_cgroup_threshold_ary {
 137         /* An array index points to threshold just below or equal to usage. */
 138         int current_threshold;
 139         /* Size of entries[] */
 140         unsigned int size;
 141         /* Array of thresholds */
 142         struct mem_cgroup_threshold entries[0];
 143 };
 144
 145 struct mem_cgroup_thresholds {
 146         /* Primary thresholds array */
 147         struct mem_cgroup_threshold_ary *primary;
 148         /*
 149          * Spare threshold array.
 150          * This is needed to make mem_cgroup_unregister_event() "never fail".
 151          * It must be able to store at least primary->size - 1 entries.
 152          */
 153         struct mem_cgroup_threshold_ary *spare;
 154 };
 155
 156 enum memcg_kmem_state {
 157         KMEM_NONE,
 158         KMEM_ALLOCATED,
 159         KMEM_ONLINE,
 160 };
 161
 162 /*
 163  * The memory controller data structure. The memory controller controls both
 164  * page cache and RSS per cgroup. We would eventually like to provide
 165  * statistics based on the statistics developed by Rik Van Riel for clock-pro,
 166  * to help the administrator determine what knobs to tune.
 167  */
 168 struct mem_cgroup {
 169         struct cgroup_subsys_state css;
 170
 171         /* Accounted resources */
 172         struct page_counter memory;
 173         struct page_counter swap;
 174
 175         /* Legacy consumer-oriented counters */
 176         struct page_counter memsw;
 177         struct page_counter kmem;
 178         struct page_counter tcpmem;
 179
 180         /* Normal memory consumption range */
 181         unsigned long low;
 182         unsigned long high;
 183
 184         /* Range enforcement for interrupt charges */
 185         struct work_struct high_work;
 186
 187         unsigned long soft_limit;
 188
 189         /* vmpressure notifications */
 190         struct vmpressure vmpressure;
 191
 192         /*
 193          * Should the accounting and control be hierarchical, per subtree?
 194          */
 195         bool use_hierarchy;
 196
 197         /* protected by memcg_oom_lock */
 198         bool            oom_lock;
 199         int             under_oom;
 200
 201         int     swappiness;
 202         /* OOM-Killer disable */
 203         int             oom_kill_disable;
 204
 205         /* handle for "memory.events" */
 206         struct cgroup_file events_file;
 207
 208         /* protect arrays of thresholds */
 209         struct mutex thresholds_lock;
 210
 211         /* thresholds for memory usage. RCU-protected */
 212         struct mem_cgroup_thresholds thresholds;
 213
 214         /* thresholds for mem+swap usage. RCU-protected */
 215         struct mem_cgroup_thresholds memsw_thresholds;
 216
 217         /* For oom notifier event fd */
 218         struct list_head oom_notify;
 219
 220         /*
 221          * Should we move charges of a task when a task is moved into this
 222          * mem_cgroup ? And what type of charges should we move ?
 223          */
 224         unsigned long move_charge_at_immigrate;
 225         /*
 226          * set > 0 if pages under this cgroup are moving to other cgroup.
 227          */
 228         atomic_t                moving_account;
 229         /* taken only while moving_account > 0 */
 230         spinlock_t              move_lock;
 231         struct task_struct      *move_lock_task;
 232         unsigned long           move_lock_flags;
 233         /*
 234          * percpu counter.
 235          */
 236         struct mem_cgroup_stat_cpu __percpu *stat;
 237
 238         unsigned long           socket_pressure;
 239
 240         /* Legacy tcp memory accounting */
 241         bool                    tcpmem_active;
 242         int                     tcpmem_pressure;
 243
 244 #ifndef CONFIG_SLOB
 245         /* Index in the kmem_cache->memcg_params.memcg_caches array */
 246         int kmemcg_id;
 247         enum memcg_kmem_state kmem_state;
 248 #endif
 249
 250         int last_scanned_node;
 251 #if MAX_NUMNODES > 1
 252         nodemask_t      scan_nodes;
 253         atomic_t        numainfo_events;
 254         atomic_t        numainfo_updating;
 255 #endif
 256
 257 #ifdef CONFIG_CGROUP_WRITEBACK
 258         struct list_head cgwb_list;
 259         struct wb_domain cgwb_domain;
 260 #endif
 261
 262         /* List of events which userspace want to receive */
 263         struct list_head event_list;
 264         spinlock_t event_list_lock;
 265
 266         struct mem_cgroup_per_node *nodeinfo[0];
 267         /* WARNING: nodeinfo must be the last member here */
 268 };
 269
 270 extern struct mem_cgroup *root_mem_cgroup;
 271
 272 static inline bool mem_cgroup_disabled(void)
 273 {
 274         return !cgroup_subsys_enabled(memory_cgrp_subsys);
 275 }
 276
 277 /**
 278  * mem_cgroup_events - count memory events against a cgroup
 279  * @memcg: the memory cgroup
 280  * @idx: the event index
 281  * @nr: the number of events to account for
 282  */
 283 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 284                        enum mem_cgroup_events_index idx,
 285                        unsigned int nr)
 286 {
 287         this_cpu_add(memcg->stat->events[idx], nr);
 288         cgroup_file_notify(&memcg->events_file);
 289 }
 290
 291 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
 292
 293 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 294                           gfp_t gfp_mask, struct mem_cgroup **memcgp,
 295                           bool compound);
 296 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
 297                               bool lrucare, bool compound);
 298 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
 299                 bool compound);
 300 void mem_cgroup_uncharge(struct page *page);
 301 void mem_cgroup_uncharge_list(struct list_head *page_list);
 302
 303 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
 304
 305 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
 306 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
 307
 308 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
 309 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 310
 311 static inline
 312 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 313         return css ? container_of(css, struct mem_cgroup, css) : NULL;
 314 }
 315
 316 #define mem_cgroup_from_counter(counter, member)        \
 317         container_of(counter, struct mem_cgroup, member)
 318
 319 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 320                                    struct mem_cgroup *,
 321                                    struct mem_cgroup_reclaim_cookie *);
 322 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 323
 324 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 325 {
 326         if (mem_cgroup_disabled())
 327                 return 0;
 328
 329         return memcg->css.id;
 330 }
 331
 332 /**
 333  * mem_cgroup_from_id - look up a memcg from an id
 334  * @id: the id to look up
 335  *
 336  * Caller must hold rcu_read_lock() and use css_tryget() as necessary.
 337  */
 338 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
 339 {
 340         struct cgroup_subsys_state *css;
 341
 342         css = css_from_id(id, &memory_cgrp_subsys);
 343         return mem_cgroup_from_css(css);
 344 }
 345
 346 /**
 347  * parent_mem_cgroup - find the accounting parent of a memcg
 348  * @memcg: memcg whose parent to find
 349  *
 350  * Returns the parent memcg, or NULL if this is the root or the memory
 351  * controller is in legacy no-hierarchy mode.
 352  */
 353 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
 354 {
 355         if (!memcg->memory.parent)
 356                 return NULL;
 357         return mem_cgroup_from_counter(memcg->memory.parent, memory);
 358 }
 359
 360 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 361                               struct mem_cgroup *root)
 362 {
 363         if (root == memcg)
 364                 return true;
 365         if (!root->use_hierarchy)
 366                 return false;
 367         return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 368 }
 369
 370 static inline bool mm_match_cgroup(struct mm_struct *mm,
 371                                    struct mem_cgroup *memcg)
 372 {
 373         struct mem_cgroup *task_memcg;
 374         bool match = false;
 375
 376         rcu_read_lock();
 377         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 378         if (task_memcg)
 379                 match = mem_cgroup_is_descendant(task_memcg, memcg);
 380         rcu_read_unlock();
 381         return match;
 382 }
 383
 384 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 385 ino_t page_cgroup_ino(struct page *page);
 386
 387 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 388 {
 389         if (mem_cgroup_disabled())
 390                 return true;
 391         return !!(memcg->css.flags & CSS_ONLINE);
 392 }
 393
 394 /*
 395  * For memory reclaim.
 396  */
 397 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 398
 399 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 400                 int nr_pages);
 401
 402 static inline
 403 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 404 {
 405         struct mem_cgroup_per_zone *mz;
 406
 407         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 408         return mz->lru_size[lru];
 409 }
 410
 411 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 412 {
 413         unsigned long inactive_ratio;
 414         unsigned long inactive;
 415         unsigned long active;
 416         unsigned long gb;
 417
 418         inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
 419         active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
 420
 421         gb = (inactive + active) >> (30 - PAGE_SHIFT);
 422         if (gb)
 423                 inactive_ratio = int_sqrt(10 * gb);
 424         else
 425                 inactive_ratio = 1;
 426
 427         return inactive * inactive_ratio < active;
 428 }
 429
 430 void mem_cgroup_handle_over_high(void);
 431
 432 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 433                                 struct task_struct *p);
 434
 435 static inline void mem_cgroup_oom_enable(void)
 436 {
 437         WARN_ON(current->memcg_may_oom);
 438         current->memcg_may_oom = 1;
 439 }
 440
 441 static inline void mem_cgroup_oom_disable(void)
 442 {
 443         WARN_ON(!current->memcg_may_oom);
 444         current->memcg_may_oom = 0;
 445 }
 446
 447 static inline bool task_in_memcg_oom(struct task_struct *p)
 448 {
 449         return p->memcg_in_oom;
 450 }
 451
 452 bool mem_cgroup_oom_synchronize(bool wait);
 453
 454 #ifdef CONFIG_MEMCG_SWAP
 455 extern int do_swap_account;
 456 #endif
 457
 458 struct mem_cgroup *lock_page_memcg(struct page *page);
 459 void unlock_page_memcg(struct mem_cgroup *memcg);
 460
 461 /**
 462  * mem_cgroup_update_page_stat - update page state statistics
 463  * @memcg: memcg to account against
 464  * @idx: page state item to account
 465  * @val: number of pages (positive or negative)
 466  *
 467  * Callers must use lock_page_memcg() to prevent double accounting
 468  * when the page is concurrently being moved to another memcg:
 469  *
 470  *   memcg = lock_page_memcg(page);
 471  *   if (TestClearPageState(page))
 472  *     mem_cgroup_update_page_stat(memcg, state, -1);
 473  *   unlock_page_memcg(memcg);
 474  */
 475 static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
 476                                  enum mem_cgroup_stat_index idx, int val)
 477 {
 478         VM_BUG_ON(!rcu_read_lock_held());
 479
 480         if (memcg)
 481                 this_cpu_add(memcg->stat->count[idx], val);
 482 }
 483
 484 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 485                                             enum mem_cgroup_stat_index idx)
 486 {
 487         mem_cgroup_update_page_stat(memcg, idx, 1);
 488 }
 489
 490 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 491                                             enum mem_cgroup_stat_index idx)
 492 {
 493         mem_cgroup_update_page_stat(memcg, idx, -1);
 494 }
 495
 496 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 497                                                 gfp_t gfp_mask,
 498                                                 unsigned long *total_scanned);
 499
 500 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 501                                              enum vm_event_item idx)
 502 {
 503         struct mem_cgroup *memcg;
 504
 505         if (mem_cgroup_disabled())
 506                 return;
 507
 508         rcu_read_lock();
 509         memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 510         if (unlikely(!memcg))
 511                 goto out;
 512
 513         switch (idx) {
 514         case PGFAULT:
 515                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
 516                 break;
 517         case PGMAJFAULT:
 518                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
 519                 break;
 520         default:
 521                 BUG();
 522         }
 523 out:
 524         rcu_read_unlock();
 525 }
 526 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 527 void mem_cgroup_split_huge_fixup(struct page *head);
 528 #endif
 529
 530 #else /* CONFIG_MEMCG */
 531
 532 #define MEM_CGROUP_ID_SHIFT     0
 533 #define MEM_CGROUP_ID_MAX       0
 534
 535 struct mem_cgroup;
 536
 537 static inline bool mem_cgroup_disabled(void)
 538 {
 539         return true;
 540 }
 541
 542 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 543                                      enum mem_cgroup_events_index idx,
 544                                      unsigned int nr)
 545 {
 546 }
 547
 548 static inline bool mem_cgroup_low(struct mem_cgroup *root,
 549                                   struct mem_cgroup *memcg)
 550 {
 551         return false;
 552 }
 553
 554 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 555                                         gfp_t gfp_mask,
 556                                         struct mem_cgroup **memcgp,
 557                                         bool compound)
 558 {
 559         *memcgp = NULL;
 560         return 0;
 561 }
 562
 563 static inline void mem_cgroup_commit_charge(struct page *page,
 564                                             struct mem_cgroup *memcg,
 565                                             bool lrucare, bool compound)
 566 {
 567 }
 568
 569 static inline void mem_cgroup_cancel_charge(struct page *page,
 570                                             struct mem_cgroup *memcg,
 571                                             bool compound)
 572 {
 573 }
 574
 575 static inline void mem_cgroup_uncharge(struct page *page)
 576 {
 577 }
 578
 579 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 580 {
 581 }
 582
 583 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
 584 {
 585 }
 586
 587 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 588                                                     struct mem_cgroup *memcg)
 589 {
 590         return &zone->lruvec;
 591 }
 592
 593 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 594                                                     struct zone *zone)
 595 {
 596         return &zone->lruvec;
 597 }
 598
 599 static inline bool mm_match_cgroup(struct mm_struct *mm,
 600                 struct mem_cgroup *memcg)
 601 {
 602         return true;
 603 }
 604
 605 static inline bool task_in_mem_cgroup(struct task_struct *task,
 606                                       const struct mem_cgroup *memcg)
 607 {
 608         return true;
 609 }
 610
 611 static inline struct mem_cgroup *
 612 mem_cgroup_iter(struct mem_cgroup *root,
 613                 struct mem_cgroup *prev,
 614                 struct mem_cgroup_reclaim_cookie *reclaim)
 615 {
 616         return NULL;
 617 }
 618
 619 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 620                                          struct mem_cgroup *prev)
 621 {
 622 }
 623
 624 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 625 {
 626         return 0;
 627 }
 628
 629 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
 630 {
 631         WARN_ON_ONCE(id);
 632         /* XXX: This should always return root_mem_cgroup */
 633         return NULL;
 634 }
 635
 636 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 637 {
 638         return true;
 639 }
 640
 641 static inline bool
 642 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 643 {
 644         return true;
 645 }
 646
 647 static inline unsigned long
 648 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 649 {
 650         return 0;
 651 }
 652
 653 static inline void
 654 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 655                               int increment)
 656 {
 657 }
 658
 659 static inline void
 660 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 661 {
 662 }
 663
 664 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
 665 {
 666         return NULL;
 667 }
 668
 669 static inline void unlock_page_memcg(struct mem_cgroup *memcg)
 670 {
 671 }
 672
 673 static inline void mem_cgroup_handle_over_high(void)
 674 {
 675 }
 676
 677 static inline void mem_cgroup_oom_enable(void)
 678 {
 679 }
 680
 681 static inline void mem_cgroup_oom_disable(void)
 682 {
 683 }
 684
 685 static inline bool task_in_memcg_oom(struct task_struct *p)
 686 {
 687         return false;
 688 }
 689
 690 static inline bool mem_cgroup_oom_synchronize(bool wait)
 691 {
 692         return false;
 693 }
 694
 695 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 696                                             enum mem_cgroup_stat_index idx)
 697 {
 698 }
 699
 700 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 701                                             enum mem_cgroup_stat_index idx)
 702 {
 703 }
 704
 705 static inline
 706 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 707                                             gfp_t gfp_mask,
 708                                             unsigned long *total_scanned)
 709 {
 710         return 0;
 711 }
 712
 713 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 714 {
 715 }
 716
 717 static inline
 718 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 719 {
 720 }
 721 #endif /* CONFIG_MEMCG */
 722
 723 #ifdef CONFIG_CGROUP_WRITEBACK
 724
 725 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
 726 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
 727 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
 728                          unsigned long *pheadroom, unsigned long *pdirty,
 729                          unsigned long *pwriteback);
 730
 731 #else   /* CONFIG_CGROUP_WRITEBACK */
 732
 733 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
 734 {
 735         return NULL;
 736 }
 737
 738 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
 739                                        unsigned long *pfilepages,
 740                                        unsigned long *pheadroom,
 741                                        unsigned long *pdirty,
 742                                        unsigned long *pwriteback)
 743 {
 744 }
 745
 746 #endif  /* CONFIG_CGROUP_WRITEBACK */
 747
 748 struct sock;
 749 void sock_update_memcg(struct sock *sk);
 750 void sock_release_memcg(struct sock *sk);
 751 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 752 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 753 #ifdef CONFIG_MEMCG
 754 extern struct static_key_false memcg_sockets_enabled_key;
 755 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
 756 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 757 {
 758         if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
 759                 return true;
 760         do {
 761                 if (time_before(jiffies, memcg->socket_pressure))
 762                         return true;
 763         } while ((memcg = parent_mem_cgroup(memcg)));
 764         return false;
 765 }
 766 #else
 767 #define mem_cgroup_sockets_enabled 0
 768 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 769 {
 770         return false;
 771 }
 772 #endif
 773
 774 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
 775 extern struct static_key_false memcg_kmem_enabled_key;
 776
 777 extern int memcg_nr_cache_ids;
 778 void memcg_get_cache_ids(void);
 779 void memcg_put_cache_ids(void);
 780
 781 /*
 782  * Helper macro to loop through all memcg-specific caches. Callers must still
 783  * check if the cache is valid (it is either valid or NULL).
 784  * the slab_mutex must be held when looping through those caches
 785  */
 786 #define for_each_memcg_cache_index(_idx)        \
 787         for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
 788
 789 static inline bool memcg_kmem_enabled(void)
 790 {
 791         return static_branch_unlikely(&memcg_kmem_enabled_key);
 792 }
 793
 794 static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
 795 {
 796         return memcg->kmem_state == KMEM_ONLINE;
 797 }
 798
 799 /*
 800  * In general, we'll do everything in our power to not incur in any overhead
 801  * for non-memcg users for the kmem functions. Not even a function call, if we
 802  * can avoid it.
 803  *
 804  * Therefore, we'll inline all those functions so that in the best case, we'll
 805  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 806  * we'll still do most of the flag checking inline. We check a lot of
 807  * conditions, but because they are pretty simple, they are expected to be
 808  * fast.
 809  */
 810 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
 811                               struct mem_cgroup *memcg);
 812 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
 813 void __memcg_kmem_uncharge(struct page *page, int order);
 814
 815 /*
 816  * helper for accessing a memcg's index. It will be used as an index in the
 817  * child cache array in kmem_cache, and also to derive its name. This function
 818  * will return -1 when this is not a kmem-limited memcg.
 819  */
 820 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 821 {
 822         return memcg ? memcg->kmemcg_id : -1;
 823 }
 824
 825 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
 826 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
 827
 828 static inline bool __memcg_kmem_bypass(void)
 829 {
 830         if (!memcg_kmem_enabled())
 831                 return true;
 832         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 833                 return true;
 834         return false;
 835 }
 836
 837 /**
 838  * memcg_kmem_charge: charge a kmem page
 839  * @page: page to charge
 840  * @gfp: reclaim mode
 841  * @order: allocation order
 842  *
 843  * Returns 0 on success, an error code on failure.
 844  */
 845 static __always_inline int memcg_kmem_charge(struct page *page,
 846                                              gfp_t gfp, int order)
 847 {
 848         if (__memcg_kmem_bypass())
 849                 return 0;
 850         if (!(gfp & __GFP_ACCOUNT))
 851                 return 0;
 852         return __memcg_kmem_charge(page, gfp, order);
 853 }
 854
 855 /**
 856  * memcg_kmem_uncharge: uncharge a kmem page
 857  * @page: page to uncharge
 858  * @order: allocation order
 859  */
 860 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
 861 {
 862         if (memcg_kmem_enabled())
 863                 __memcg_kmem_uncharge(page, order);
 864 }
 865
 866 /**
 867  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 868  * @cachep: the original global kmem cache
 869  *
 870  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 871  */
 872 static __always_inline struct kmem_cache *
 873 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 874 {
 875         if (__memcg_kmem_bypass())
 876                 return cachep;
 877         return __memcg_kmem_get_cache(cachep, gfp);
 878 }
 879
 880 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 881 {
 882         if (memcg_kmem_enabled())
 883                 __memcg_kmem_put_cache(cachep);
 884 }
 885 #else
 886 #define for_each_memcg_cache_index(_idx)        \
 887         for (; NULL; )
 888
 889 static inline bool memcg_kmem_enabled(void)
 890 {
 891         return false;
 892 }
 893
 894 static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
 895 {
 896         return false;
 897 }
 898
 899 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
 900 {
 901         return 0;
 902 }
 903
 904 static inline void memcg_kmem_uncharge(struct page *page, int order)
 905 {
 906 }
 907
 908 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 909 {
 910         return -1;
 911 }
 912
 913 static inline void memcg_get_cache_ids(void)
 914 {
 915 }
 916
 917 static inline void memcg_put_cache_ids(void)
 918 {
 919 }
 920
 921 static inline struct kmem_cache *
 922 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 923 {
 924         return cachep;
 925 }
 926
 927 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 928 {
 929 }
 930 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
 931
 932 #endif /* _LINUX_MEMCONTROL_H */