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 /*
  89  * Bits in struct cg_proto.flags
  90  */
  91 enum cg_proto_flags {
  92         /* Currently active and new sockets should be assigned to cgroups */
  93         MEMCG_SOCK_ACTIVE,
  94         /* It was ever activated; we must disarm static keys on destruction */
  95         MEMCG_SOCK_ACTIVATED,
  96 };
  97
  98 struct cg_proto {
  99         struct page_counter     memory_allocated;       /* Current allocated memory. */
 100         struct percpu_counter   sockets_allocated;      /* Current number of sockets. */
 101         int                     memory_pressure;
 102         long                    sysctl_mem[3];
 103         unsigned long           flags;
 104         /*
 105          * memcg field is used to find which memcg we belong directly
 106          * Each memcg struct can hold more than one cg_proto, so container_of
 107          * won't really cut.
 108          *
 109          * The elegant solution would be having an inverse function to
 110          * proto_cgroup in struct proto, but that means polluting the structure
 111          * for everybody, instead of just for memcg users.
 112          */
 113         struct mem_cgroup       *memcg;
 114 };
 115
 116 #ifdef CONFIG_MEMCG
 117 struct mem_cgroup_stat_cpu {
 118         long count[MEM_CGROUP_STAT_NSTATS];
 119         unsigned long events[MEMCG_NR_EVENTS];
 120         unsigned long nr_page_events;
 121         unsigned long targets[MEM_CGROUP_NTARGETS];
 122 };
 123
 124 struct mem_cgroup_reclaim_iter {
 125         struct mem_cgroup *position;
 126         /* scan generation, increased every round-trip */
 127         unsigned int generation;
 128 };
 129
 130 /*
 131  * per-zone information in memory controller.
 132  */
 133 struct mem_cgroup_per_zone {
 134         struct lruvec           lruvec;
 135         unsigned long           lru_size[NR_LRU_LISTS];
 136
 137         struct mem_cgroup_reclaim_iter  iter[DEF_PRIORITY + 1];
 138
 139         struct rb_node          tree_node;      /* RB tree node */
 140         unsigned long           usage_in_excess;/* Set to the value by which */
 141                                                 /* the soft limit is exceeded*/
 142         bool                    on_tree;
 143         struct mem_cgroup       *memcg;         /* Back pointer, we cannot */
 144                                                 /* use container_of        */
 145 };
 146
 147 struct mem_cgroup_per_node {
 148         struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
 149 };
 150
 151 struct mem_cgroup_threshold {
 152         struct eventfd_ctx *eventfd;
 153         unsigned long threshold;
 154 };
 155
 156 /* For threshold */
 157 struct mem_cgroup_threshold_ary {
 158         /* An array index points to threshold just below or equal to usage. */
 159         int current_threshold;
 160         /* Size of entries[] */
 161         unsigned int size;
 162         /* Array of thresholds */
 163         struct mem_cgroup_threshold entries[0];
 164 };
 165
 166 struct mem_cgroup_thresholds {
 167         /* Primary thresholds array */
 168         struct mem_cgroup_threshold_ary *primary;
 169         /*
 170          * Spare threshold array.
 171          * This is needed to make mem_cgroup_unregister_event() "never fail".
 172          * It must be able to store at least primary->size - 1 entries.
 173          */
 174         struct mem_cgroup_threshold_ary *spare;
 175 };
 176
 177 /*
 178  * The memory controller data structure. The memory controller controls both
 179  * page cache and RSS per cgroup. We would eventually like to provide
 180  * statistics based on the statistics developed by Rik Van Riel for clock-pro,
 181  * to help the administrator determine what knobs to tune.
 182  */
 183 struct mem_cgroup {
 184         struct cgroup_subsys_state css;
 185
 186         /* Accounted resources */
 187         struct page_counter memory;
 188         struct page_counter memsw;
 189         struct page_counter kmem;
 190
 191         /* Normal memory consumption range */
 192         unsigned long low;
 193         unsigned long high;
 194
 195         unsigned long soft_limit;
 196
 197         /* vmpressure notifications */
 198         struct vmpressure vmpressure;
 199
 200         /* css_online() has been completed */
 201         int initialized;
 202
 203         /*
 204          * Should the accounting and control be hierarchical, per subtree?
 205          */
 206         bool use_hierarchy;
 207
 208         /* protected by memcg_oom_lock */
 209         bool            oom_lock;
 210         int             under_oom;
 211
 212         int     swappiness;
 213         /* OOM-Killer disable */
 214         int             oom_kill_disable;
 215
 216         /* protect arrays of thresholds */
 217         struct mutex thresholds_lock;
 218
 219         /* thresholds for memory usage. RCU-protected */
 220         struct mem_cgroup_thresholds thresholds;
 221
 222         /* thresholds for mem+swap usage. RCU-protected */
 223         struct mem_cgroup_thresholds memsw_thresholds;
 224
 225         /* For oom notifier event fd */
 226         struct list_head oom_notify;
 227
 228         /*
 229          * Should we move charges of a task when a task is moved into this
 230          * mem_cgroup ? And what type of charges should we move ?
 231          */
 232         unsigned long move_charge_at_immigrate;
 233         /*
 234          * set > 0 if pages under this cgroup are moving to other cgroup.
 235          */
 236         atomic_t                moving_account;
 237         /* taken only while moving_account > 0 */
 238         spinlock_t              move_lock;
 239         struct task_struct      *move_lock_task;
 240         unsigned long           move_lock_flags;
 241         /*
 242          * percpu counter.
 243          */
 244         struct mem_cgroup_stat_cpu __percpu *stat;
 245
 246 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
 247         struct cg_proto tcp_mem;
 248 #endif
 249 #if defined(CONFIG_MEMCG_KMEM)
 250         /* Index in the kmem_cache->memcg_params.memcg_caches array */
 251         int kmemcg_id;
 252         bool kmem_acct_activated;
 253         bool kmem_acct_active;
 254 #endif
 255
 256         int last_scanned_node;
 257 #if MAX_NUMNODES > 1
 258         nodemask_t      scan_nodes;
 259         atomic_t        numainfo_events;
 260         atomic_t        numainfo_updating;
 261 #endif
 262
 263 #ifdef CONFIG_CGROUP_WRITEBACK
 264         struct list_head cgwb_list;
 265         struct wb_domain cgwb_domain;
 266 #endif
 267
 268         /* List of events which userspace want to receive */
 269         struct list_head event_list;
 270         spinlock_t event_list_lock;
 271
 272         struct mem_cgroup_per_node *nodeinfo[0];
 273         /* WARNING: nodeinfo must be the last member here */
 274 };
 275 extern struct cgroup_subsys_state *mem_cgroup_root_css;
 276
 277 /**
 278  * mem_cgroup_events - count memory events against a cgroup
 279  * @memcg: the memory cgroup
 280  * @idx: the event index
 281  * @nr: the number of events to account for
 282  */
 283 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 284                        enum mem_cgroup_events_index idx,
 285                        unsigned int nr)
 286 {
 287         this_cpu_add(memcg->stat->events[idx], nr);
 288 }
 289
 290 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
 291
 292 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 293                           gfp_t gfp_mask, struct mem_cgroup **memcgp);
 294 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
 295                               bool lrucare);
 296 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
 297 void mem_cgroup_uncharge(struct page *page);
 298 void mem_cgroup_uncharge_list(struct list_head *page_list);
 299
 300 void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
 301
 302 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
 303 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
 304
 305 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
 306 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 307 struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
 308
 309 static inline
 310 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 311         return css ? container_of(css, struct mem_cgroup, css) : NULL;
 312 }
 313
 314 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 315                                    struct mem_cgroup *,
 316                                    struct mem_cgroup_reclaim_cookie *);
 317 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 318
 319 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 320                               struct mem_cgroup *root)
 321 {
 322         if (root == memcg)
 323                 return true;
 324         if (!root->use_hierarchy)
 325                 return false;
 326         return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 327 }
 328
 329 static inline bool mm_match_cgroup(struct mm_struct *mm,
 330                                    struct mem_cgroup *memcg)
 331 {
 332         struct mem_cgroup *task_memcg;
 333         bool match = false;
 334
 335         rcu_read_lock();
 336         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 337         if (task_memcg)
 338                 match = mem_cgroup_is_descendant(task_memcg, memcg);
 339         rcu_read_unlock();
 340         return match;
 341 }
 342
 343 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 344 ino_t page_cgroup_ino(struct page *page);
 345
 346 static inline bool mem_cgroup_disabled(void)
 347 {
 348         if (memory_cgrp_subsys.disabled)
 349                 return true;
 350         return false;
 351 }
 352
 353 /*
 354  * For memory reclaim.
 355  */
 356 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 357
 358 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 359                 int nr_pages);
 360
 361 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
 362 {
 363         struct mem_cgroup_per_zone *mz;
 364         struct mem_cgroup *memcg;
 365
 366         if (mem_cgroup_disabled())
 367                 return true;
 368
 369         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 370         memcg = mz->memcg;
 371
 372         return !!(memcg->css.flags & CSS_ONLINE);
 373 }
 374
 375 static inline
 376 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 377 {
 378         struct mem_cgroup_per_zone *mz;
 379
 380         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 381         return mz->lru_size[lru];
 382 }
 383
 384 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 385 {
 386         unsigned long inactive_ratio;
 387         unsigned long inactive;
 388         unsigned long active;
 389         unsigned long gb;
 390
 391         inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
 392         active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
 393
 394         gb = (inactive + active) >> (30 - PAGE_SHIFT);
 395         if (gb)
 396                 inactive_ratio = int_sqrt(10 * gb);
 397         else
 398                 inactive_ratio = 1;
 399
 400         return inactive * inactive_ratio < active;
 401 }
 402
 403 void mem_cgroup_handle_over_high(void);
 404
 405 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 406                                 struct task_struct *p);
 407
 408 static inline void mem_cgroup_oom_enable(void)
 409 {
 410         WARN_ON(current->memcg_may_oom);
 411         current->memcg_may_oom = 1;
 412 }
 413
 414 static inline void mem_cgroup_oom_disable(void)
 415 {
 416         WARN_ON(!current->memcg_may_oom);
 417         current->memcg_may_oom = 0;
 418 }
 419
 420 static inline bool task_in_memcg_oom(struct task_struct *p)
 421 {
 422         return p->memcg_in_oom;
 423 }
 424
 425 bool mem_cgroup_oom_synchronize(bool wait);
 426
 427 #ifdef CONFIG_MEMCG_SWAP
 428 extern int do_swap_account;
 429 #endif
 430
 431 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
 432 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
 433
 434 /**
 435  * mem_cgroup_update_page_stat - update page state statistics
 436  * @memcg: memcg to account against
 437  * @idx: page state item to account
 438  * @val: number of pages (positive or negative)
 439  *
 440  * See mem_cgroup_begin_page_stat() for locking requirements.
 441  */
 442 static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
 443                                  enum mem_cgroup_stat_index idx, int val)
 444 {
 445         VM_BUG_ON(!rcu_read_lock_held());
 446
 447         if (memcg)
 448                 this_cpu_add(memcg->stat->count[idx], val);
 449 }
 450
 451 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 452                                             enum mem_cgroup_stat_index idx)
 453 {
 454         mem_cgroup_update_page_stat(memcg, idx, 1);
 455 }
 456
 457 static inline void mem_cgroup_dec_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 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 464                                                 gfp_t gfp_mask,
 465                                                 unsigned long *total_scanned);
 466
 467 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 468                                              enum vm_event_item idx)
 469 {
 470         struct mem_cgroup *memcg;
 471
 472         if (mem_cgroup_disabled())
 473                 return;
 474
 475         rcu_read_lock();
 476         memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 477         if (unlikely(!memcg))
 478                 goto out;
 479
 480         switch (idx) {
 481         case PGFAULT:
 482                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
 483                 break;
 484         case PGMAJFAULT:
 485                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
 486                 break;
 487         default:
 488                 BUG();
 489         }
 490 out:
 491         rcu_read_unlock();
 492 }
 493 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 494 void mem_cgroup_split_huge_fixup(struct page *head);
 495 #endif
 496
 497 #else /* CONFIG_MEMCG */
 498 struct mem_cgroup;
 499
 500 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 501                                      enum mem_cgroup_events_index idx,
 502                                      unsigned int nr)
 503 {
 504 }
 505
 506 static inline bool mem_cgroup_low(struct mem_cgroup *root,
 507                                   struct mem_cgroup *memcg)
 508 {
 509         return false;
 510 }
 511
 512 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 513                                         gfp_t gfp_mask,
 514                                         struct mem_cgroup **memcgp)
 515 {
 516         *memcgp = NULL;
 517         return 0;
 518 }
 519
 520 static inline void mem_cgroup_commit_charge(struct page *page,
 521                                             struct mem_cgroup *memcg,
 522                                             bool lrucare)
 523 {
 524 }
 525
 526 static inline void mem_cgroup_cancel_charge(struct page *page,
 527                                             struct mem_cgroup *memcg)
 528 {
 529 }
 530
 531 static inline void mem_cgroup_uncharge(struct page *page)
 532 {
 533 }
 534
 535 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 536 {
 537 }
 538
 539 static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
 540 {
 541 }
 542
 543 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 544                                                     struct mem_cgroup *memcg)
 545 {
 546         return &zone->lruvec;
 547 }
 548
 549 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 550                                                     struct zone *zone)
 551 {
 552         return &zone->lruvec;
 553 }
 554
 555 static inline bool mm_match_cgroup(struct mm_struct *mm,
 556                 struct mem_cgroup *memcg)
 557 {
 558         return true;
 559 }
 560
 561 static inline bool task_in_mem_cgroup(struct task_struct *task,
 562                                       const struct mem_cgroup *memcg)
 563 {
 564         return true;
 565 }
 566
 567 static inline struct mem_cgroup *
 568 mem_cgroup_iter(struct mem_cgroup *root,
 569                 struct mem_cgroup *prev,
 570                 struct mem_cgroup_reclaim_cookie *reclaim)
 571 {
 572         return NULL;
 573 }
 574
 575 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 576                                          struct mem_cgroup *prev)
 577 {
 578 }
 579
 580 static inline bool mem_cgroup_disabled(void)
 581 {
 582         return true;
 583 }
 584
 585 static inline bool
 586 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 587 {
 588         return true;
 589 }
 590
 591 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
 592 {
 593         return true;
 594 }
 595
 596 static inline unsigned long
 597 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 598 {
 599         return 0;
 600 }
 601
 602 static inline void
 603 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 604                               int increment)
 605 {
 606 }
 607
 608 static inline void
 609 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 610 {
 611 }
 612
 613 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
 614 {
 615         return NULL;
 616 }
 617
 618 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
 619 {
 620 }
 621
 622 static inline void mem_cgroup_handle_over_high(void)
 623 {
 624 }
 625
 626 static inline void mem_cgroup_oom_enable(void)
 627 {
 628 }
 629
 630 static inline void mem_cgroup_oom_disable(void)
 631 {
 632 }
 633
 634 static inline bool task_in_memcg_oom(struct task_struct *p)
 635 {
 636         return false;
 637 }
 638
 639 static inline bool mem_cgroup_oom_synchronize(bool wait)
 640 {
 641         return false;
 642 }
 643
 644 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 645                                             enum mem_cgroup_stat_index idx)
 646 {
 647 }
 648
 649 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 650                                             enum mem_cgroup_stat_index idx)
 651 {
 652 }
 653
 654 static inline
 655 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 656                                             gfp_t gfp_mask,
 657                                             unsigned long *total_scanned)
 658 {
 659         return 0;
 660 }
 661
 662 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 663 {
 664 }
 665
 666 static inline
 667 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 668 {
 669 }
 670 #endif /* CONFIG_MEMCG */
 671
 672 enum {
 673         UNDER_LIMIT,
 674         SOFT_LIMIT,
 675         OVER_LIMIT,
 676 };
 677
 678 #ifdef CONFIG_CGROUP_WRITEBACK
 679
 680 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
 681 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
 682 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
 683                          unsigned long *pheadroom, unsigned long *pdirty,
 684                          unsigned long *pwriteback);
 685
 686 #else   /* CONFIG_CGROUP_WRITEBACK */
 687
 688 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
 689 {
 690         return NULL;
 691 }
 692
 693 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
 694                                        unsigned long *pfilepages,
 695                                        unsigned long *pheadroom,
 696                                        unsigned long *pdirty,
 697                                        unsigned long *pwriteback)
 698 {
 699 }
 700
 701 #endif  /* CONFIG_CGROUP_WRITEBACK */
 702
 703 struct sock;
 704 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
 705 void sock_update_memcg(struct sock *sk);
 706 void sock_release_memcg(struct sock *sk);
 707 #else
 708 static inline void sock_update_memcg(struct sock *sk)
 709 {
 710 }
 711 static inline void sock_release_memcg(struct sock *sk)
 712 {
 713 }
 714 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
 715
 716 #ifdef CONFIG_MEMCG_KMEM
 717 extern struct static_key memcg_kmem_enabled_key;
 718
 719 extern int memcg_nr_cache_ids;
 720 void memcg_get_cache_ids(void);
 721 void memcg_put_cache_ids(void);
 722
 723 /*
 724  * Helper macro to loop through all memcg-specific caches. Callers must still
 725  * check if the cache is valid (it is either valid or NULL).
 726  * the slab_mutex must be held when looping through those caches
 727  */
 728 #define for_each_memcg_cache_index(_idx)        \
 729         for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
 730
 731 static inline bool memcg_kmem_enabled(void)
 732 {
 733         return static_key_false(&memcg_kmem_enabled_key);
 734 }
 735
 736 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
 737 {
 738         return memcg->kmem_acct_active;
 739 }
 740
 741 /*
 742  * In general, we'll do everything in our power to not incur in any overhead
 743  * for non-memcg users for the kmem functions. Not even a function call, if we
 744  * can avoid it.
 745  *
 746  * Therefore, we'll inline all those functions so that in the best case, we'll
 747  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 748  * we'll still do most of the flag checking inline. We check a lot of
 749  * conditions, but because they are pretty simple, they are expected to be
 750  * fast.
 751  */
 752 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
 753                               struct mem_cgroup *memcg);
 754 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
 755 void __memcg_kmem_uncharge(struct page *page, int order);
 756
 757 /*
 758  * helper for acessing a memcg's index. It will be used as an index in the
 759  * child cache array in kmem_cache, and also to derive its name. This function
 760  * will return -1 when this is not a kmem-limited memcg.
 761  */
 762 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 763 {
 764         return memcg ? memcg->kmemcg_id : -1;
 765 }
 766
 767 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep);
 768 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
 769
 770 static inline bool __memcg_kmem_bypass(gfp_t gfp)
 771 {
 772         if (!memcg_kmem_enabled())
 773                 return true;
 774         if (gfp & __GFP_NOACCOUNT)
 775                 return true;
 776         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 777                 return true;
 778         return false;
 779 }
 780
 781 /**
 782  * memcg_kmem_charge: charge a kmem page
 783  * @page: page to charge
 784  * @gfp: reclaim mode
 785  * @order: allocation order
 786  *
 787  * Returns 0 on success, an error code on failure.
 788  */
 789 static __always_inline int memcg_kmem_charge(struct page *page,
 790                                              gfp_t gfp, int order)
 791 {
 792         if (__memcg_kmem_bypass(gfp))
 793                 return 0;
 794         return __memcg_kmem_charge(page, gfp, order);
 795 }
 796
 797 /**
 798  * memcg_kmem_uncharge: uncharge a kmem page
 799  * @page: page to uncharge
 800  * @order: allocation order
 801  */
 802 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
 803 {
 804         if (memcg_kmem_enabled())
 805                 __memcg_kmem_uncharge(page, order);
 806 }
 807
 808 /**
 809  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 810  * @cachep: the original global kmem cache
 811  * @gfp: allocation flags.
 812  *
 813  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 814  */
 815 static __always_inline struct kmem_cache *
 816 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 817 {
 818         if (__memcg_kmem_bypass(gfp))
 819                 return cachep;
 820         return __memcg_kmem_get_cache(cachep);
 821 }
 822
 823 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 824 {
 825         if (memcg_kmem_enabled())
 826                 __memcg_kmem_put_cache(cachep);
 827 }
 828 #else
 829 #define for_each_memcg_cache_index(_idx)        \
 830         for (; NULL; )
 831
 832 static inline bool memcg_kmem_enabled(void)
 833 {
 834         return false;
 835 }
 836
 837 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
 838 {
 839         return false;
 840 }
 841
 842 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
 843 {
 844         return 0;
 845 }
 846
 847 static inline void memcg_kmem_uncharge(struct page *page, int order)
 848 {
 849 }
 850
 851 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 852 {
 853         return -1;
 854 }
 855
 856 static inline void memcg_get_cache_ids(void)
 857 {
 858 }
 859
 860 static inline void memcg_put_cache_ids(void)
 861 {
 862 }
 863
 864 static inline struct kmem_cache *
 865 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 866 {
 867         return cachep;
 868 }
 869
 870 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 871 {
 872 }
 873 #endif /* CONFIG_MEMCG_KMEM */
 874 #endif /* _LINUX_MEMCONTROL_H */