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 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
 285                               bool lrucare);
 286 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
 287 void mem_cgroup_uncharge(struct page *page);
 288 void mem_cgroup_uncharge_list(struct list_head *page_list);
 289
 290 void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
 291
 292 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
 293 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
 294
 295 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
 296 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 297
 298 static inline
 299 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 300         return css ? container_of(css, struct mem_cgroup, css) : NULL;
 301 }
 302
 303 #define mem_cgroup_from_counter(counter, member)        \
 304         container_of(counter, struct mem_cgroup, member)
 305
 306 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 307                                    struct mem_cgroup *,
 308                                    struct mem_cgroup_reclaim_cookie *);
 309 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 310
 311 /**
 312  * parent_mem_cgroup - find the accounting parent of a memcg
 313  * @memcg: memcg whose parent to find
 314  *
 315  * Returns the parent memcg, or NULL if this is the root or the memory
 316  * controller is in legacy no-hierarchy mode.
 317  */
 318 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
 319 {
 320         if (!memcg->memory.parent)
 321                 return NULL;
 322         return mem_cgroup_from_counter(memcg->memory.parent, memory);
 323 }
 324
 325 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 326                               struct mem_cgroup *root)
 327 {
 328         if (root == memcg)
 329                 return true;
 330         if (!root->use_hierarchy)
 331                 return false;
 332         return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 333 }
 334
 335 static inline bool mm_match_cgroup(struct mm_struct *mm,
 336                                    struct mem_cgroup *memcg)
 337 {
 338         struct mem_cgroup *task_memcg;
 339         bool match = false;
 340
 341         rcu_read_lock();
 342         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 343         if (task_memcg)
 344                 match = mem_cgroup_is_descendant(task_memcg, memcg);
 345         rcu_read_unlock();
 346         return match;
 347 }
 348
 349 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 350 ino_t page_cgroup_ino(struct page *page);
 351
 352 static inline bool mem_cgroup_disabled(void)
 353 {
 354         return !cgroup_subsys_enabled(memory_cgrp_subsys);
 355 }
 356
 357 /*
 358  * For memory reclaim.
 359  */
 360 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 361
 362 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 363                 int nr_pages);
 364
 365 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
 366 {
 367         struct mem_cgroup_per_zone *mz;
 368         struct mem_cgroup *memcg;
 369
 370         if (mem_cgroup_disabled())
 371                 return true;
 372
 373         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 374         memcg = mz->memcg;
 375
 376         return !!(memcg->css.flags & CSS_ONLINE);
 377 }
 378
 379 static inline
 380 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 381 {
 382         struct mem_cgroup_per_zone *mz;
 383
 384         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 385         return mz->lru_size[lru];
 386 }
 387
 388 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 389 {
 390         unsigned long inactive_ratio;
 391         unsigned long inactive;
 392         unsigned long active;
 393         unsigned long gb;
 394
 395         inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
 396         active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
 397
 398         gb = (inactive + active) >> (30 - PAGE_SHIFT);
 399         if (gb)
 400                 inactive_ratio = int_sqrt(10 * gb);
 401         else
 402                 inactive_ratio = 1;
 403
 404         return inactive * inactive_ratio < active;
 405 }
 406
 407 void mem_cgroup_handle_over_high(void);
 408
 409 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 410                                 struct task_struct *p);
 411
 412 static inline void mem_cgroup_oom_enable(void)
 413 {
 414         WARN_ON(current->memcg_may_oom);
 415         current->memcg_may_oom = 1;
 416 }
 417
 418 static inline void mem_cgroup_oom_disable(void)
 419 {
 420         WARN_ON(!current->memcg_may_oom);
 421         current->memcg_may_oom = 0;
 422 }
 423
 424 static inline bool task_in_memcg_oom(struct task_struct *p)
 425 {
 426         return p->memcg_in_oom;
 427 }
 428
 429 bool mem_cgroup_oom_synchronize(bool wait);
 430
 431 #ifdef CONFIG_MEMCG_SWAP
 432 extern int do_swap_account;
 433 #endif
 434
 435 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
 436 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
 437
 438 /**
 439  * mem_cgroup_update_page_stat - update page state statistics
 440  * @memcg: memcg to account against
 441  * @idx: page state item to account
 442  * @val: number of pages (positive or negative)
 443  *
 444  * See mem_cgroup_begin_page_stat() for locking requirements.
 445  */
 446 static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
 447                                  enum mem_cgroup_stat_index idx, int val)
 448 {
 449         VM_BUG_ON(!rcu_read_lock_held());
 450
 451         if (memcg)
 452                 this_cpu_add(memcg->stat->count[idx], val);
 453 }
 454
 455 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 456                                             enum mem_cgroup_stat_index idx)
 457 {
 458         mem_cgroup_update_page_stat(memcg, idx, 1);
 459 }
 460
 461 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 462                                             enum mem_cgroup_stat_index idx)
 463 {
 464         mem_cgroup_update_page_stat(memcg, idx, -1);
 465 }
 466
 467 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 468                                                 gfp_t gfp_mask,
 469                                                 unsigned long *total_scanned);
 470
 471 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 472                                              enum vm_event_item idx)
 473 {
 474         struct mem_cgroup *memcg;
 475
 476         if (mem_cgroup_disabled())
 477                 return;
 478
 479         rcu_read_lock();
 480         memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 481         if (unlikely(!memcg))
 482                 goto out;
 483
 484         switch (idx) {
 485         case PGFAULT:
 486                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
 487                 break;
 488         case PGMAJFAULT:
 489                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
 490                 break;
 491         default:
 492                 BUG();
 493         }
 494 out:
 495         rcu_read_unlock();
 496 }
 497 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 498 void mem_cgroup_split_huge_fixup(struct page *head);
 499 #endif
 500
 501 #else /* CONFIG_MEMCG */
 502 struct mem_cgroup;
 503
 504 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 505                                      enum mem_cgroup_events_index idx,
 506                                      unsigned int nr)
 507 {
 508 }
 509
 510 static inline bool mem_cgroup_low(struct mem_cgroup *root,
 511                                   struct mem_cgroup *memcg)
 512 {
 513         return false;
 514 }
 515
 516 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 517                                         gfp_t gfp_mask,
 518                                         struct mem_cgroup **memcgp)
 519 {
 520         *memcgp = NULL;
 521         return 0;
 522 }
 523
 524 static inline void mem_cgroup_commit_charge(struct page *page,
 525                                             struct mem_cgroup *memcg,
 526                                             bool lrucare)
 527 {
 528 }
 529
 530 static inline void mem_cgroup_cancel_charge(struct page *page,
 531                                             struct mem_cgroup *memcg)
 532 {
 533 }
 534
 535 static inline void mem_cgroup_uncharge(struct page *page)
 536 {
 537 }
 538
 539 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 540 {
 541 }
 542
 543 static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
 544 {
 545 }
 546
 547 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 548                                                     struct mem_cgroup *memcg)
 549 {
 550         return &zone->lruvec;
 551 }
 552
 553 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 554                                                     struct zone *zone)
 555 {
 556         return &zone->lruvec;
 557 }
 558
 559 static inline bool mm_match_cgroup(struct mm_struct *mm,
 560                 struct mem_cgroup *memcg)
 561 {
 562         return true;
 563 }
 564
 565 static inline bool task_in_mem_cgroup(struct task_struct *task,
 566                                       const struct mem_cgroup *memcg)
 567 {
 568         return true;
 569 }
 570
 571 static inline struct mem_cgroup *
 572 mem_cgroup_iter(struct mem_cgroup *root,
 573                 struct mem_cgroup *prev,
 574                 struct mem_cgroup_reclaim_cookie *reclaim)
 575 {
 576         return NULL;
 577 }
 578
 579 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 580                                          struct mem_cgroup *prev)
 581 {
 582 }
 583
 584 static inline bool mem_cgroup_disabled(void)
 585 {
 586         return true;
 587 }
 588
 589 static inline bool
 590 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 591 {
 592         return true;
 593 }
 594
 595 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
 596 {
 597         return true;
 598 }
 599
 600 static inline unsigned long
 601 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 602 {
 603         return 0;
 604 }
 605
 606 static inline void
 607 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 608                               int increment)
 609 {
 610 }
 611
 612 static inline void
 613 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 614 {
 615 }
 616
 617 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
 618 {
 619         return NULL;
 620 }
 621
 622 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
 623 {
 624 }
 625
 626 static inline void mem_cgroup_handle_over_high(void)
 627 {
 628 }
 629
 630 static inline void mem_cgroup_oom_enable(void)
 631 {
 632 }
 633
 634 static inline void mem_cgroup_oom_disable(void)
 635 {
 636 }
 637
 638 static inline bool task_in_memcg_oom(struct task_struct *p)
 639 {
 640         return false;
 641 }
 642
 643 static inline bool mem_cgroup_oom_synchronize(bool wait)
 644 {
 645         return false;
 646 }
 647
 648 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 649                                             enum mem_cgroup_stat_index idx)
 650 {
 651 }
 652
 653 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 654                                             enum mem_cgroup_stat_index idx)
 655 {
 656 }
 657
 658 static inline
 659 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 660                                             gfp_t gfp_mask,
 661                                             unsigned long *total_scanned)
 662 {
 663         return 0;
 664 }
 665
 666 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 667 {
 668 }
 669
 670 static inline
 671 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 672 {
 673 }
 674 #endif /* CONFIG_MEMCG */
 675
 676 #ifdef CONFIG_CGROUP_WRITEBACK
 677
 678 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
 679 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
 680 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
 681                          unsigned long *pheadroom, unsigned long *pdirty,
 682                          unsigned long *pwriteback);
 683
 684 #else   /* CONFIG_CGROUP_WRITEBACK */
 685
 686 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
 687 {
 688         return NULL;
 689 }
 690
 691 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
 692                                        unsigned long *pfilepages,
 693                                        unsigned long *pheadroom,
 694                                        unsigned long *pdirty,
 695                                        unsigned long *pwriteback)
 696 {
 697 }
 698
 699 #endif  /* CONFIG_CGROUP_WRITEBACK */
 700
 701 struct sock;
 702 void sock_update_memcg(struct sock *sk);
 703 void sock_release_memcg(struct sock *sk);
 704 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 705 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 706 #if defined(CONFIG_MEMCG) && defined(CONFIG_INET)
 707 extern struct static_key_false memcg_sockets_enabled_key;
 708 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
 709 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 710 {
 711 #ifdef CONFIG_MEMCG_KMEM
 712         if (memcg->tcp_mem.memory_pressure)
 713                 return true;
 714 #endif
 715         do {
 716                 if (time_before(jiffies, memcg->socket_pressure))
 717                         return true;
 718         } while ((memcg = parent_mem_cgroup(memcg)));
 719         return false;
 720 }
 721 #else
 722 #define mem_cgroup_sockets_enabled 0
 723 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 724 {
 725         return false;
 726 }
 727 #endif
 728
 729 #ifdef CONFIG_MEMCG_KMEM
 730 extern struct static_key_false memcg_kmem_enabled_key;
 731
 732 extern int memcg_nr_cache_ids;
 733 void memcg_get_cache_ids(void);
 734 void memcg_put_cache_ids(void);
 735
 736 /*
 737  * Helper macro to loop through all memcg-specific caches. Callers must still
 738  * check if the cache is valid (it is either valid or NULL).
 739  * the slab_mutex must be held when looping through those caches
 740  */
 741 #define for_each_memcg_cache_index(_idx)        \
 742         for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
 743
 744 static inline bool memcg_kmem_enabled(void)
 745 {
 746         return static_branch_unlikely(&memcg_kmem_enabled_key);
 747 }
 748
 749 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
 750 {
 751         return memcg->kmem_acct_active;
 752 }
 753
 754 /*
 755  * In general, we'll do everything in our power to not incur in any overhead
 756  * for non-memcg users for the kmem functions. Not even a function call, if we
 757  * can avoid it.
 758  *
 759  * Therefore, we'll inline all those functions so that in the best case, we'll
 760  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 761  * we'll still do most of the flag checking inline. We check a lot of
 762  * conditions, but because they are pretty simple, they are expected to be
 763  * fast.
 764  */
 765 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
 766                               struct mem_cgroup *memcg);
 767 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
 768 void __memcg_kmem_uncharge(struct page *page, int order);
 769
 770 /*
 771  * helper for acessing a memcg's index. It will be used as an index in the
 772  * child cache array in kmem_cache, and also to derive its name. This function
 773  * will return -1 when this is not a kmem-limited memcg.
 774  */
 775 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 776 {
 777         return memcg ? memcg->kmemcg_id : -1;
 778 }
 779
 780 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
 781 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
 782
 783 static inline bool __memcg_kmem_bypass(void)
 784 {
 785         if (!memcg_kmem_enabled())
 786                 return true;
 787         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 788                 return true;
 789         return false;
 790 }
 791
 792 /**
 793  * memcg_kmem_charge: charge a kmem page
 794  * @page: page to charge
 795  * @gfp: reclaim mode
 796  * @order: allocation order
 797  *
 798  * Returns 0 on success, an error code on failure.
 799  */
 800 static __always_inline int memcg_kmem_charge(struct page *page,
 801                                              gfp_t gfp, int order)
 802 {
 803         if (__memcg_kmem_bypass())
 804                 return 0;
 805         if (!(gfp & __GFP_ACCOUNT))
 806                 return 0;
 807         return __memcg_kmem_charge(page, gfp, order);
 808 }
 809
 810 /**
 811  * memcg_kmem_uncharge: uncharge a kmem page
 812  * @page: page to uncharge
 813  * @order: allocation order
 814  */
 815 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
 816 {
 817         if (memcg_kmem_enabled())
 818                 __memcg_kmem_uncharge(page, order);
 819 }
 820
 821 /**
 822  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 823  * @cachep: the original global kmem cache
 824  *
 825  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 826  */
 827 static __always_inline struct kmem_cache *
 828 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 829 {
 830         if (__memcg_kmem_bypass())
 831                 return cachep;
 832         return __memcg_kmem_get_cache(cachep, gfp);
 833 }
 834
 835 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 836 {
 837         if (memcg_kmem_enabled())
 838                 __memcg_kmem_put_cache(cachep);
 839 }
 840 #else
 841 #define for_each_memcg_cache_index(_idx)        \
 842         for (; NULL; )
 843
 844 static inline bool memcg_kmem_enabled(void)
 845 {
 846         return false;
 847 }
 848
 849 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
 850 {
 851         return false;
 852 }
 853
 854 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
 855 {
 856         return 0;
 857 }
 858
 859 static inline void memcg_kmem_uncharge(struct page *page, int order)
 860 {
 861 }
 862
 863 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 864 {
 865         return -1;
 866 }
 867
 868 static inline void memcg_get_cache_ids(void)
 869 {
 870 }
 871
 872 static inline void memcg_put_cache_ids(void)
 873 {
 874 }
 875
 876 static inline struct kmem_cache *
 877 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 878 {
 879         return cachep;
 880 }
 881
 882 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 883 {
 884 }
 885 #endif /* CONFIG_MEMCG_KMEM */
 886 #endif /* _LINUX_MEMCONTROL_H */