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