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