include/linux/memcontrol.h

   1 /* memcontrol.h - Memory Controller
   2  *
   3  * Copyright IBM Corporation, 2007
   4  * Author Balbir Singh <balbir@linux.vnet.ibm.com>
   5  *
   6  * Copyright 2007 OpenVZ SWsoft Inc
   7  * Author: Pavel Emelianov <xemul@openvz.org>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  */
  19
  20 #ifndef _LINUX_MEMCONTROL_H
  21 #define _LINUX_MEMCONTROL_H
  22 #include <linux/cgroup.h>
  23 #include <linux/vm_event_item.h>
  24 #include <linux/hardirq.h>
  25 #include <linux/jump_label.h>
  26 #include <linux/page_counter.h>
  27 #include <linux/vmpressure.h>
  28 #include <linux/eventfd.h>
  29 #include <linux/mmzone.h>
  30 #include <linux/writeback.h>
  31 #include <linux/page-flags.h>
  32
  33 struct mem_cgroup;
  34 struct page;
  35 struct mm_struct;
  36 struct kmem_cache;
  37
  38 /*
  39  * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
  40  * These two lists should keep in accord with each other.
  41  */
  42 enum mem_cgroup_stat_index {
  43         /*
  44          * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
  45          */
  46         MEM_CGROUP_STAT_CACHE,          /* # of pages charged as cache */
  47         MEM_CGROUP_STAT_RSS,            /* # of pages charged as anon rss */
  48         MEM_CGROUP_STAT_RSS_HUGE,       /* # of pages charged as anon huge */
  49         MEM_CGROUP_STAT_FILE_MAPPED,    /* # of pages charged as file rss */
  50         MEM_CGROUP_STAT_DIRTY,          /* # of dirty pages in page cache */
  51         MEM_CGROUP_STAT_WRITEBACK,      /* # of pages under writeback */
  52         MEM_CGROUP_STAT_SWAP,           /* # of pages, swapped out */
  53         MEM_CGROUP_STAT_NSTATS,
  54         /* default hierarchy stats */
  55         MEMCG_SOCK = MEM_CGROUP_STAT_NSTATS,
  56         MEMCG_NR_STAT,
  57 };
  58
  59 struct mem_cgroup_reclaim_cookie {
  60         struct zone *zone;
  61         int priority;
  62         unsigned int generation;
  63 };
  64
  65 enum mem_cgroup_events_index {
  66         MEM_CGROUP_EVENTS_PGPGIN,       /* # of pages paged in */
  67         MEM_CGROUP_EVENTS_PGPGOUT,      /* # of pages paged out */
  68         MEM_CGROUP_EVENTS_PGFAULT,      /* # of page-faults */
  69         MEM_CGROUP_EVENTS_PGMAJFAULT,   /* # of major page-faults */
  70         MEM_CGROUP_EVENTS_NSTATS,
  71         /* default hierarchy events */
  72         MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
  73         MEMCG_HIGH,
  74         MEMCG_MAX,
  75         MEMCG_OOM,
  76         MEMCG_NR_EVENTS,
  77 };
  78
  79 /*
  80  * Per memcg event counter is incremented at every pagein/pageout. With THP,
  81  * it will be incremated by the number of pages. This counter is used for
  82  * for trigger some periodic events. This is straightforward and better
  83  * than using jiffies etc. to handle periodic memcg event.
  84  */
  85 enum mem_cgroup_events_target {
  86         MEM_CGROUP_TARGET_THRESH,
  87         MEM_CGROUP_TARGET_SOFTLIMIT,
  88         MEM_CGROUP_TARGET_NUMAINFO,
  89         MEM_CGROUP_NTARGETS,
  90 };
  91
  92 #ifdef CONFIG_MEMCG
  93
  94 #define MEM_CGROUP_ID_SHIFT     16
  95 #define MEM_CGROUP_ID_MAX       USHRT_MAX
  96
  97 struct mem_cgroup_stat_cpu {
  98         long count[MEMCG_NR_STAT];
  99         unsigned long events[MEMCG_NR_EVENTS];
 100         unsigned long nr_page_events;
 101         unsigned long targets[MEM_CGROUP_NTARGETS];
 102 };
 103
 104 struct mem_cgroup_reclaim_iter {
 105         struct mem_cgroup *position;
 106         /* scan generation, increased every round-trip */
 107         unsigned int generation;
 108 };
 109
 110 /*
 111  * per-zone information in memory controller.
 112  */
 113 struct mem_cgroup_per_zone {
 114         struct lruvec           lruvec;
 115         unsigned long           lru_size[NR_LRU_LISTS];
 116
 117         struct mem_cgroup_reclaim_iter  iter[DEF_PRIORITY + 1];
 118
 119         struct rb_node          tree_node;      /* RB tree node */
 120         unsigned long           usage_in_excess;/* Set to the value by which */
 121                                                 /* the soft limit is exceeded*/
 122         bool                    on_tree;
 123         struct mem_cgroup       *memcg;         /* Back pointer, we cannot */
 124                                                 /* use container_of        */
 125 };
 126
 127 struct mem_cgroup_per_node {
 128         struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
 129 };
 130
 131 struct mem_cgroup_threshold {
 132         struct eventfd_ctx *eventfd;
 133         unsigned long threshold;
 134 };
 135
 136 /* For threshold */
 137 struct mem_cgroup_threshold_ary {
 138         /* An array index points to threshold just below or equal to usage. */
 139         int current_threshold;
 140         /* Size of entries[] */
 141         unsigned int size;
 142         /* Array of thresholds */
 143         struct mem_cgroup_threshold entries[0];
 144 };
 145
 146 struct mem_cgroup_thresholds {
 147         /* Primary thresholds array */
 148         struct mem_cgroup_threshold_ary *primary;
 149         /*
 150          * Spare threshold array.
 151          * This is needed to make mem_cgroup_unregister_event() "never fail".
 152          * It must be able to store at least primary->size - 1 entries.
 153          */
 154         struct mem_cgroup_threshold_ary *spare;
 155 };
 156
 157 enum memcg_kmem_state {
 158         KMEM_NONE,
 159         KMEM_ALLOCATED,
 160         KMEM_ONLINE,
 161 };
 162
 163 /*
 164  * The memory controller data structure. The memory controller controls both
 165  * page cache and RSS per cgroup. We would eventually like to provide
 166  * statistics based on the statistics developed by Rik Van Riel for clock-pro,
 167  * to help the administrator determine what knobs to tune.
 168  */
 169 struct mem_cgroup {
 170         struct cgroup_subsys_state css;
 171
 172         /* Accounted resources */
 173         struct page_counter memory;
 174         struct page_counter swap;
 175
 176         /* Legacy consumer-oriented counters */
 177         struct page_counter memsw;
 178         struct page_counter kmem;
 179         struct page_counter tcpmem;
 180
 181         /* Normal memory consumption range */
 182         unsigned long low;
 183         unsigned long high;
 184
 185         /* Range enforcement for interrupt charges */
 186         struct work_struct high_work;
 187
 188         unsigned long soft_limit;
 189
 190         /* vmpressure notifications */
 191         struct vmpressure vmpressure;
 192
 193         /*
 194          * Should the accounting and control be hierarchical, per subtree?
 195          */
 196         bool use_hierarchy;
 197
 198         /* protected by memcg_oom_lock */
 199         bool            oom_lock;
 200         int             under_oom;
 201
 202         int     swappiness;
 203         /* OOM-Killer disable */
 204         int             oom_kill_disable;
 205
 206         /* handle for "memory.events" */
 207         struct cgroup_file events_file;
 208
 209         /* protect arrays of thresholds */
 210         struct mutex thresholds_lock;
 211
 212         /* thresholds for memory usage. RCU-protected */
 213         struct mem_cgroup_thresholds thresholds;
 214
 215         /* thresholds for mem+swap usage. RCU-protected */
 216         struct mem_cgroup_thresholds memsw_thresholds;
 217
 218         /* For oom notifier event fd */
 219         struct list_head oom_notify;
 220
 221         /*
 222          * Should we move charges of a task when a task is moved into this
 223          * mem_cgroup ? And what type of charges should we move ?
 224          */
 225         unsigned long move_charge_at_immigrate;
 226         /*
 227          * set > 0 if pages under this cgroup are moving to other cgroup.
 228          */
 229         atomic_t                moving_account;
 230         /* taken only while moving_account > 0 */
 231         spinlock_t              move_lock;
 232         struct task_struct      *move_lock_task;
 233         unsigned long           move_lock_flags;
 234         /*
 235          * percpu counter.
 236          */
 237         struct mem_cgroup_stat_cpu __percpu *stat;
 238
 239         unsigned long           socket_pressure;
 240
 241         /* Legacy tcp memory accounting */
 242         bool                    tcpmem_active;
 243         int                     tcpmem_pressure;
 244
 245 #ifndef CONFIG_SLOB
 246         /* Index in the kmem_cache->memcg_params.memcg_caches array */
 247         int kmemcg_id;
 248         enum memcg_kmem_state kmem_state;
 249 #endif
 250
 251         int last_scanned_node;
 252 #if MAX_NUMNODES > 1
 253         nodemask_t      scan_nodes;
 254         atomic_t        numainfo_events;
 255         atomic_t        numainfo_updating;
 256 #endif
 257
 258 #ifdef CONFIG_CGROUP_WRITEBACK
 259         struct list_head cgwb_list;
 260         struct wb_domain cgwb_domain;
 261 #endif
 262
 263         /* List of events which userspace want to receive */
 264         struct list_head event_list;
 265         spinlock_t event_list_lock;
 266
 267         struct mem_cgroup_per_node *nodeinfo[0];
 268         /* WARNING: nodeinfo must be the last member here */
 269 };
 270
 271 extern struct mem_cgroup *root_mem_cgroup;
 272
 273 static inline bool mem_cgroup_disabled(void)
 274 {
 275         return !cgroup_subsys_enabled(memory_cgrp_subsys);
 276 }
 277
 278 /**
 279  * mem_cgroup_events - count memory events against a cgroup
 280  * @memcg: the memory cgroup
 281  * @idx: the event index
 282  * @nr: the number of events to account for
 283  */
 284 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 285                        enum mem_cgroup_events_index idx,
 286                        unsigned int nr)
 287 {
 288         this_cpu_add(memcg->stat->events[idx], nr);
 289         cgroup_file_notify(&memcg->events_file);
 290 }
 291
 292 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
 293
 294 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 295                           gfp_t gfp_mask, struct mem_cgroup **memcgp,
 296                           bool compound);
 297 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
 298                               bool lrucare, bool compound);
 299 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
 300                 bool compound);
 301 void mem_cgroup_uncharge(struct page *page);
 302 void mem_cgroup_uncharge_list(struct list_head *page_list);
 303
 304 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
 305
 306 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
 307 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
 308
 309 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
 310 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 311
 312 static inline
 313 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 314         return css ? container_of(css, struct mem_cgroup, css) : NULL;
 315 }
 316
 317 #define mem_cgroup_from_counter(counter, member)        \
 318         container_of(counter, struct mem_cgroup, member)
 319
 320 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 321                                    struct mem_cgroup *,
 322                                    struct mem_cgroup_reclaim_cookie *);
 323 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 324
 325 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 326 {
 327         if (mem_cgroup_disabled())
 328                 return 0;
 329
 330         return memcg->css.id;
 331 }
 332
 333 /**
 334  * mem_cgroup_from_id - look up a memcg from an id
 335  * @id: the id to look up
 336  *
 337  * Caller must hold rcu_read_lock() and use css_tryget() as necessary.
 338  */
 339 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
 340 {
 341         struct cgroup_subsys_state *css;
 342
 343         css = css_from_id(id, &memory_cgrp_subsys);
 344         return mem_cgroup_from_css(css);
 345 }
 346
 347 /**
 348  * parent_mem_cgroup - find the accounting parent of a memcg
 349  * @memcg: memcg whose parent to find
 350  *
 351  * Returns the parent memcg, or NULL if this is the root or the memory
 352  * controller is in legacy no-hierarchy mode.
 353  */
 354 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
 355 {
 356         if (!memcg->memory.parent)
 357                 return NULL;
 358         return mem_cgroup_from_counter(memcg->memory.parent, memory);
 359 }
 360
 361 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 362                               struct mem_cgroup *root)
 363 {
 364         if (root == memcg)
 365                 return true;
 366         if (!root->use_hierarchy)
 367                 return false;
 368         return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 369 }
 370
 371 static inline bool mm_match_cgroup(struct mm_struct *mm,
 372                                    struct mem_cgroup *memcg)
 373 {
 374         struct mem_cgroup *task_memcg;
 375         bool match = false;
 376
 377         rcu_read_lock();
 378         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 379         if (task_memcg)
 380                 match = mem_cgroup_is_descendant(task_memcg, memcg);
 381         rcu_read_unlock();
 382         return match;
 383 }
 384
 385 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 386 ino_t page_cgroup_ino(struct page *page);
 387
 388 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 389 {
 390         if (mem_cgroup_disabled())
 391                 return true;
 392         return !!(memcg->css.flags & CSS_ONLINE);
 393 }
 394
 395 /*
 396  * For memory reclaim.
 397  */
 398 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 399
 400 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 401                 int nr_pages);
 402
 403 static inline
 404 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 405 {
 406         struct mem_cgroup_per_zone *mz;
 407
 408         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 409         return mz->lru_size[lru];
 410 }
 411
 412 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 413 {
 414         unsigned long inactive_ratio;
 415         unsigned long inactive;
 416         unsigned long active;
 417         unsigned long gb;
 418
 419         inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
 420         active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
 421
 422         gb = (inactive + active) >> (30 - PAGE_SHIFT);
 423         if (gb)
 424                 inactive_ratio = int_sqrt(10 * gb);
 425         else
 426                 inactive_ratio = 1;
 427
 428         return inactive * inactive_ratio < active;
 429 }
 430
 431 void mem_cgroup_handle_over_high(void);
 432
 433 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 434                                 struct task_struct *p);
 435
 436 static inline void mem_cgroup_oom_enable(void)
 437 {
 438         WARN_ON(current->memcg_may_oom);
 439         current->memcg_may_oom = 1;
 440 }
 441
 442 static inline void mem_cgroup_oom_disable(void)
 443 {
 444         WARN_ON(!current->memcg_may_oom);
 445         current->memcg_may_oom = 0;
 446 }
 447
 448 static inline bool task_in_memcg_oom(struct task_struct *p)
 449 {
 450         return p->memcg_in_oom;
 451 }
 452
 453 bool mem_cgroup_oom_synchronize(bool wait);
 454
 455 #ifdef CONFIG_MEMCG_SWAP
 456 extern int do_swap_account;
 457 #endif
 458
 459 void lock_page_memcg(struct page *page);
 460 void unlock_page_memcg(struct page *page);
 461
 462 /**
 463  * mem_cgroup_update_page_stat - update page state statistics
 464  * @page: the page
 465  * @idx: page state item to account
 466  * @val: number of pages (positive or negative)
 467  *
 468  * The @page must be locked or the caller must use lock_page_memcg()
 469  * to prevent double accounting when the page is concurrently being
 470  * moved to another memcg:
 471  *
 472  *   lock_page(page) or lock_page_memcg(page)
 473  *   if (TestClearPageState(page))
 474  *     mem_cgroup_update_page_stat(page, state, -1);
 475  *   unlock_page(page) or unlock_page_memcg(page)
 476  */
 477 static inline void mem_cgroup_update_page_stat(struct page *page,
 478                                  enum mem_cgroup_stat_index idx, int val)
 479 {
 480         VM_BUG_ON(!(rcu_read_lock_held() || PageLocked(page)));
 481
 482         if (page->mem_cgroup)
 483                 this_cpu_add(page->mem_cgroup->stat->count[idx], val);
 484 }
 485
 486 static inline void mem_cgroup_inc_page_stat(struct page *page,
 487                                             enum mem_cgroup_stat_index idx)
 488 {
 489         mem_cgroup_update_page_stat(page, idx, 1);
 490 }
 491
 492 static inline void mem_cgroup_dec_page_stat(struct page *page,
 493                                             enum mem_cgroup_stat_index idx)
 494 {
 495         mem_cgroup_update_page_stat(page, idx, -1);
 496 }
 497
 498 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 499                                                 gfp_t gfp_mask,
 500                                                 unsigned long *total_scanned);
 501
 502 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 503                                              enum vm_event_item idx)
 504 {
 505         struct mem_cgroup *memcg;
 506
 507         if (mem_cgroup_disabled())
 508                 return;
 509
 510         rcu_read_lock();
 511         memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 512         if (unlikely(!memcg))
 513                 goto out;
 514
 515         switch (idx) {
 516         case PGFAULT:
 517                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
 518                 break;
 519         case PGMAJFAULT:
 520                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
 521                 break;
 522         default:
 523                 BUG();
 524         }
 525 out:
 526         rcu_read_unlock();
 527 }
 528 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 529 void mem_cgroup_split_huge_fixup(struct page *head);
 530 #endif
 531
 532 #else /* CONFIG_MEMCG */
 533
 534 #define MEM_CGROUP_ID_SHIFT     0
 535 #define MEM_CGROUP_ID_MAX       0
 536
 537 struct mem_cgroup;
 538
 539 static inline bool mem_cgroup_disabled(void)
 540 {
 541         return true;
 542 }
 543
 544 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 545                                      enum mem_cgroup_events_index idx,
 546                                      unsigned int nr)
 547 {
 548 }
 549
 550 static inline bool mem_cgroup_low(struct mem_cgroup *root,
 551                                   struct mem_cgroup *memcg)
 552 {
 553         return false;
 554 }
 555
 556 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 557                                         gfp_t gfp_mask,
 558                                         struct mem_cgroup **memcgp,
 559                                         bool compound)
 560 {
 561         *memcgp = NULL;
 562         return 0;
 563 }
 564
 565 static inline void mem_cgroup_commit_charge(struct page *page,
 566                                             struct mem_cgroup *memcg,
 567                                             bool lrucare, bool compound)
 568 {
 569 }
 570
 571 static inline void mem_cgroup_cancel_charge(struct page *page,
 572                                             struct mem_cgroup *memcg,
 573                                             bool compound)
 574 {
 575 }
 576
 577 static inline void mem_cgroup_uncharge(struct page *page)
 578 {
 579 }
 580
 581 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 582 {
 583 }
 584
 585 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
 586 {
 587 }
 588
 589 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 590                                                     struct mem_cgroup *memcg)
 591 {
 592         return &zone->lruvec;
 593 }
 594
 595 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 596                                                     struct zone *zone)
 597 {
 598         return &zone->lruvec;
 599 }
 600
 601 static inline bool mm_match_cgroup(struct mm_struct *mm,
 602                 struct mem_cgroup *memcg)
 603 {
 604         return true;
 605 }
 606
 607 static inline bool task_in_mem_cgroup(struct task_struct *task,
 608                                       const struct mem_cgroup *memcg)
 609 {
 610         return true;
 611 }
 612
 613 static inline struct mem_cgroup *
 614 mem_cgroup_iter(struct mem_cgroup *root,
 615                 struct mem_cgroup *prev,
 616                 struct mem_cgroup_reclaim_cookie *reclaim)
 617 {
 618         return NULL;
 619 }
 620
 621 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 622                                          struct mem_cgroup *prev)
 623 {
 624 }
 625
 626 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 627 {
 628         return 0;
 629 }
 630
 631 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
 632 {
 633         WARN_ON_ONCE(id);
 634         /* XXX: This should always return root_mem_cgroup */
 635         return NULL;
 636 }
 637
 638 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 639 {
 640         return true;
 641 }
 642
 643 static inline bool
 644 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 645 {
 646         return true;
 647 }
 648
 649 static inline unsigned long
 650 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 651 {
 652         return 0;
 653 }
 654
 655 static inline void
 656 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 657                               int increment)
 658 {
 659 }
 660
 661 static inline void
 662 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 663 {
 664 }
 665
 666 static inline void lock_page_memcg(struct page *page)
 667 {
 668 }
 669
 670 static inline void unlock_page_memcg(struct page *page)
 671 {
 672 }
 673
 674 static inline void mem_cgroup_handle_over_high(void)
 675 {
 676 }
 677
 678 static inline void mem_cgroup_oom_enable(void)
 679 {
 680 }
 681
 682 static inline void mem_cgroup_oom_disable(void)
 683 {
 684 }
 685
 686 static inline bool task_in_memcg_oom(struct task_struct *p)
 687 {
 688         return false;
 689 }
 690
 691 static inline bool mem_cgroup_oom_synchronize(bool wait)
 692 {
 693         return false;
 694 }
 695
 696 static inline void mem_cgroup_inc_page_stat(struct page *page,
 697                                             enum mem_cgroup_stat_index idx)
 698 {
 699 }
 700
 701 static inline void mem_cgroup_dec_page_stat(struct page *page,
 702                                             enum mem_cgroup_stat_index idx)
 703 {
 704 }
 705
 706 static inline
 707 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 708                                             gfp_t gfp_mask,
 709                                             unsigned long *total_scanned)
 710 {
 711         return 0;
 712 }
 713
 714 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 715 {
 716 }
 717
 718 static inline
 719 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 720 {
 721 }
 722 #endif /* CONFIG_MEMCG */
 723
 724 #ifdef CONFIG_CGROUP_WRITEBACK
 725
 726 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
 727 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
 728 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
 729                          unsigned long *pheadroom, unsigned long *pdirty,
 730                          unsigned long *pwriteback);
 731
 732 #else   /* CONFIG_CGROUP_WRITEBACK */
 733
 734 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
 735 {
 736         return NULL;
 737 }
 738
 739 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
 740                                        unsigned long *pfilepages,
 741                                        unsigned long *pheadroom,
 742                                        unsigned long *pdirty,
 743                                        unsigned long *pwriteback)
 744 {
 745 }
 746
 747 #endif  /* CONFIG_CGROUP_WRITEBACK */
 748
 749 struct sock;
 750 void sock_update_memcg(struct sock *sk);
 751 void sock_release_memcg(struct sock *sk);
 752 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 753 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 754 #ifdef CONFIG_MEMCG
 755 extern struct static_key_false memcg_sockets_enabled_key;
 756 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
 757 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 758 {
 759         if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
 760                 return true;
 761         do {
 762                 if (time_before(jiffies, memcg->socket_pressure))
 763                         return true;
 764         } while ((memcg = parent_mem_cgroup(memcg)));
 765         return false;
 766 }
 767 #else
 768 #define mem_cgroup_sockets_enabled 0
 769 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 770 {
 771         return false;
 772 }
 773 #endif
 774
 775 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
 776 extern struct static_key_false memcg_kmem_enabled_key;
 777
 778 extern int memcg_nr_cache_ids;
 779 void memcg_get_cache_ids(void);
 780 void memcg_put_cache_ids(void);
 781
 782 /*
 783  * Helper macro to loop through all memcg-specific caches. Callers must still
 784  * check if the cache is valid (it is either valid or NULL).
 785  * the slab_mutex must be held when looping through those caches
 786  */
 787 #define for_each_memcg_cache_index(_idx)        \
 788         for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
 789
 790 static inline bool memcg_kmem_enabled(void)
 791 {
 792         return static_branch_unlikely(&memcg_kmem_enabled_key);
 793 }
 794
 795 static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
 796 {
 797         return memcg->kmem_state == KMEM_ONLINE;
 798 }
 799
 800 /*
 801  * In general, we'll do everything in our power to not incur in any overhead
 802  * for non-memcg users for the kmem functions. Not even a function call, if we
 803  * can avoid it.
 804  *
 805  * Therefore, we'll inline all those functions so that in the best case, we'll
 806  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 807  * we'll still do most of the flag checking inline. We check a lot of
 808  * conditions, but because they are pretty simple, they are expected to be
 809  * fast.
 810  */
 811 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
 812                               struct mem_cgroup *memcg);
 813 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
 814 void __memcg_kmem_uncharge(struct page *page, int order);
 815
 816 /*
 817  * helper for accessing a memcg's index. It will be used as an index in the
 818  * child cache array in kmem_cache, and also to derive its name. This function
 819  * will return -1 when this is not a kmem-limited memcg.
 820  */
 821 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 822 {
 823         return memcg ? memcg->kmemcg_id : -1;
 824 }
 825
 826 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
 827 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
 828
 829 static inline bool __memcg_kmem_bypass(void)
 830 {
 831         if (!memcg_kmem_enabled())
 832                 return true;
 833         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 834                 return true;
 835         return false;
 836 }
 837
 838 /**
 839  * memcg_kmem_charge: charge a kmem page
 840  * @page: page to charge
 841  * @gfp: reclaim mode
 842  * @order: allocation order
 843  *
 844  * Returns 0 on success, an error code on failure.
 845  */
 846 static __always_inline int memcg_kmem_charge(struct page *page,
 847                                              gfp_t gfp, int order)
 848 {
 849         if (__memcg_kmem_bypass())
 850                 return 0;
 851         if (!(gfp & __GFP_ACCOUNT))
 852                 return 0;
 853         return __memcg_kmem_charge(page, gfp, order);
 854 }
 855
 856 /**
 857  * memcg_kmem_uncharge: uncharge a kmem page
 858  * @page: page to uncharge
 859  * @order: allocation order
 860  */
 861 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
 862 {
 863         if (memcg_kmem_enabled())
 864                 __memcg_kmem_uncharge(page, order);
 865 }
 866
 867 /**
 868  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 869  * @cachep: the original global kmem cache
 870  *
 871  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 872  */
 873 static __always_inline struct kmem_cache *
 874 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 875 {
 876         if (__memcg_kmem_bypass())
 877                 return cachep;
 878         return __memcg_kmem_get_cache(cachep, gfp);
 879 }
 880
 881 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 882 {
 883         if (memcg_kmem_enabled())
 884                 __memcg_kmem_put_cache(cachep);
 885 }
 886 #else
 887 #define for_each_memcg_cache_index(_idx)        \
 888         for (; NULL; )
 889
 890 static inline bool memcg_kmem_enabled(void)
 891 {
 892         return false;
 893 }
 894
 895 static inline bool memcg_kmem_online(struct mem_cgroup *memcg)
 896 {
 897         return false;
 898 }
 899
 900 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
 901 {
 902         return 0;
 903 }
 904
 905 static inline void memcg_kmem_uncharge(struct page *page, int order)
 906 {
 907 }
 908
 909 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 910 {
 911         return -1;
 912 }
 913
 914 static inline void memcg_get_cache_ids(void)
 915 {
 916 }
 917
 918 static inline void memcg_put_cache_ids(void)
 919 {
 920 }
 921
 922 static inline struct kmem_cache *
 923 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 924 {
 925         return cachep;
 926 }
 927
 928 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 929 {
 930 }
 931 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
 932
 933 #endif /* _LINUX_MEMCONTROL_H */