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         unsigned long soft_limit;
 174
 175         /* vmpressure notifications */
 176         struct vmpressure vmpressure;
 177
 178         /* css_online() has been completed */
 179         int initialized;
 180
 181         /*
 182          * Should the accounting and control be hierarchical, per subtree?
 183          */
 184         bool use_hierarchy;
 185
 186         /* protected by memcg_oom_lock */
 187         bool            oom_lock;
 188         int             under_oom;
 189
 190         int     swappiness;
 191         /* OOM-Killer disable */
 192         int             oom_kill_disable;
 193
 194         /* handle for "memory.events" */
 195         struct cgroup_file events_file;
 196
 197         /* protect arrays of thresholds */
 198         struct mutex thresholds_lock;
 199
 200         /* thresholds for memory usage. RCU-protected */
 201         struct mem_cgroup_thresholds thresholds;
 202
 203         /* thresholds for mem+swap usage. RCU-protected */
 204         struct mem_cgroup_thresholds memsw_thresholds;
 205
 206         /* For oom notifier event fd */
 207         struct list_head oom_notify;
 208
 209         /*
 210          * Should we move charges of a task when a task is moved into this
 211          * mem_cgroup ? And what type of charges should we move ?
 212          */
 213         unsigned long move_charge_at_immigrate;
 214         /*
 215          * set > 0 if pages under this cgroup are moving to other cgroup.
 216          */
 217         atomic_t                moving_account;
 218         /* taken only while moving_account > 0 */
 219         spinlock_t              move_lock;
 220         struct task_struct      *move_lock_task;
 221         unsigned long           move_lock_flags;
 222         /*
 223          * percpu counter.
 224          */
 225         struct mem_cgroup_stat_cpu __percpu *stat;
 226
 227 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
 228         struct cg_proto tcp_mem;
 229 #endif
 230 #if defined(CONFIG_MEMCG_KMEM)
 231         /* Index in the kmem_cache->memcg_params.memcg_caches array */
 232         int kmemcg_id;
 233         bool kmem_acct_activated;
 234         bool kmem_acct_active;
 235 #endif
 236
 237         int last_scanned_node;
 238 #if MAX_NUMNODES > 1
 239         nodemask_t      scan_nodes;
 240         atomic_t        numainfo_events;
 241         atomic_t        numainfo_updating;
 242 #endif
 243
 244 #ifdef CONFIG_CGROUP_WRITEBACK
 245         struct list_head cgwb_list;
 246         struct wb_domain cgwb_domain;
 247 #endif
 248
 249         /* List of events which userspace want to receive */
 250         struct list_head event_list;
 251         spinlock_t event_list_lock;
 252
 253         struct mem_cgroup_per_node *nodeinfo[0];
 254         /* WARNING: nodeinfo must be the last member here */
 255 };
 256
 257 extern struct mem_cgroup *root_mem_cgroup;
 258
 259 /**
 260  * mem_cgroup_events - count memory events against a cgroup
 261  * @memcg: the memory cgroup
 262  * @idx: the event index
 263  * @nr: the number of events to account for
 264  */
 265 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 266                        enum mem_cgroup_events_index idx,
 267                        unsigned int nr)
 268 {
 269         this_cpu_add(memcg->stat->events[idx], nr);
 270         cgroup_file_notify(&memcg->events_file);
 271 }
 272
 273 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
 274
 275 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 276                           gfp_t gfp_mask, struct mem_cgroup **memcgp);
 277 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
 278                               bool lrucare);
 279 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
 280 void mem_cgroup_uncharge(struct page *page);
 281 void mem_cgroup_uncharge_list(struct list_head *page_list);
 282
 283 void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
 284
 285 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
 286 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
 287
 288 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
 289 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 290 struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
 291
 292 static inline
 293 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 294         return css ? container_of(css, struct mem_cgroup, css) : NULL;
 295 }
 296
 297 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 298                                    struct mem_cgroup *,
 299                                    struct mem_cgroup_reclaim_cookie *);
 300 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 301
 302 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 303                               struct mem_cgroup *root)
 304 {
 305         if (root == memcg)
 306                 return true;
 307         if (!root->use_hierarchy)
 308                 return false;
 309         return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 310 }
 311
 312 static inline bool mm_match_cgroup(struct mm_struct *mm,
 313                                    struct mem_cgroup *memcg)
 314 {
 315         struct mem_cgroup *task_memcg;
 316         bool match = false;
 317
 318         rcu_read_lock();
 319         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 320         if (task_memcg)
 321                 match = mem_cgroup_is_descendant(task_memcg, memcg);
 322         rcu_read_unlock();
 323         return match;
 324 }
 325
 326 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 327 ino_t page_cgroup_ino(struct page *page);
 328
 329 static inline bool mem_cgroup_disabled(void)
 330 {
 331         return !cgroup_subsys_enabled(memory_cgrp_subsys);
 332 }
 333
 334 /*
 335  * For memory reclaim.
 336  */
 337 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 338
 339 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 340                 int nr_pages);
 341
 342 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
 343 {
 344         struct mem_cgroup_per_zone *mz;
 345         struct mem_cgroup *memcg;
 346
 347         if (mem_cgroup_disabled())
 348                 return true;
 349
 350         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 351         memcg = mz->memcg;
 352
 353         return !!(memcg->css.flags & CSS_ONLINE);
 354 }
 355
 356 static inline
 357 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 358 {
 359         struct mem_cgroup_per_zone *mz;
 360
 361         mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
 362         return mz->lru_size[lru];
 363 }
 364
 365 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 366 {
 367         unsigned long inactive_ratio;
 368         unsigned long inactive;
 369         unsigned long active;
 370         unsigned long gb;
 371
 372         inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
 373         active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
 374
 375         gb = (inactive + active) >> (30 - PAGE_SHIFT);
 376         if (gb)
 377                 inactive_ratio = int_sqrt(10 * gb);
 378         else
 379                 inactive_ratio = 1;
 380
 381         return inactive * inactive_ratio < active;
 382 }
 383
 384 void mem_cgroup_handle_over_high(void);
 385
 386 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 387                                 struct task_struct *p);
 388
 389 static inline void mem_cgroup_oom_enable(void)
 390 {
 391         WARN_ON(current->memcg_may_oom);
 392         current->memcg_may_oom = 1;
 393 }
 394
 395 static inline void mem_cgroup_oom_disable(void)
 396 {
 397         WARN_ON(!current->memcg_may_oom);
 398         current->memcg_may_oom = 0;
 399 }
 400
 401 static inline bool task_in_memcg_oom(struct task_struct *p)
 402 {
 403         return p->memcg_in_oom;
 404 }
 405
 406 bool mem_cgroup_oom_synchronize(bool wait);
 407
 408 #ifdef CONFIG_MEMCG_SWAP
 409 extern int do_swap_account;
 410 #endif
 411
 412 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
 413 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
 414
 415 /**
 416  * mem_cgroup_update_page_stat - update page state statistics
 417  * @memcg: memcg to account against
 418  * @idx: page state item to account
 419  * @val: number of pages (positive or negative)
 420  *
 421  * See mem_cgroup_begin_page_stat() for locking requirements.
 422  */
 423 static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
 424                                  enum mem_cgroup_stat_index idx, int val)
 425 {
 426         VM_BUG_ON(!rcu_read_lock_held());
 427
 428         if (memcg)
 429                 this_cpu_add(memcg->stat->count[idx], val);
 430 }
 431
 432 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 433                                             enum mem_cgroup_stat_index idx)
 434 {
 435         mem_cgroup_update_page_stat(memcg, idx, 1);
 436 }
 437
 438 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 439                                             enum mem_cgroup_stat_index idx)
 440 {
 441         mem_cgroup_update_page_stat(memcg, idx, -1);
 442 }
 443
 444 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 445                                                 gfp_t gfp_mask,
 446                                                 unsigned long *total_scanned);
 447
 448 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 449                                              enum vm_event_item idx)
 450 {
 451         struct mem_cgroup *memcg;
 452
 453         if (mem_cgroup_disabled())
 454                 return;
 455
 456         rcu_read_lock();
 457         memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 458         if (unlikely(!memcg))
 459                 goto out;
 460
 461         switch (idx) {
 462         case PGFAULT:
 463                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
 464                 break;
 465         case PGMAJFAULT:
 466                 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
 467                 break;
 468         default:
 469                 BUG();
 470         }
 471 out:
 472         rcu_read_unlock();
 473 }
 474 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 475 void mem_cgroup_split_huge_fixup(struct page *head);
 476 #endif
 477
 478 #else /* CONFIG_MEMCG */
 479 struct mem_cgroup;
 480
 481 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
 482                                      enum mem_cgroup_events_index idx,
 483                                      unsigned int nr)
 484 {
 485 }
 486
 487 static inline bool mem_cgroup_low(struct mem_cgroup *root,
 488                                   struct mem_cgroup *memcg)
 489 {
 490         return false;
 491 }
 492
 493 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
 494                                         gfp_t gfp_mask,
 495                                         struct mem_cgroup **memcgp)
 496 {
 497         *memcgp = NULL;
 498         return 0;
 499 }
 500
 501 static inline void mem_cgroup_commit_charge(struct page *page,
 502                                             struct mem_cgroup *memcg,
 503                                             bool lrucare)
 504 {
 505 }
 506
 507 static inline void mem_cgroup_cancel_charge(struct page *page,
 508                                             struct mem_cgroup *memcg)
 509 {
 510 }
 511
 512 static inline void mem_cgroup_uncharge(struct page *page)
 513 {
 514 }
 515
 516 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 517 {
 518 }
 519
 520 static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
 521 {
 522 }
 523
 524 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 525                                                     struct mem_cgroup *memcg)
 526 {
 527         return &zone->lruvec;
 528 }
 529
 530 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 531                                                     struct zone *zone)
 532 {
 533         return &zone->lruvec;
 534 }
 535
 536 static inline bool mm_match_cgroup(struct mm_struct *mm,
 537                 struct mem_cgroup *memcg)
 538 {
 539         return true;
 540 }
 541
 542 static inline bool task_in_mem_cgroup(struct task_struct *task,
 543                                       const struct mem_cgroup *memcg)
 544 {
 545         return true;
 546 }
 547
 548 static inline struct mem_cgroup *
 549 mem_cgroup_iter(struct mem_cgroup *root,
 550                 struct mem_cgroup *prev,
 551                 struct mem_cgroup_reclaim_cookie *reclaim)
 552 {
 553         return NULL;
 554 }
 555
 556 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 557                                          struct mem_cgroup *prev)
 558 {
 559 }
 560
 561 static inline bool mem_cgroup_disabled(void)
 562 {
 563         return true;
 564 }
 565
 566 static inline bool
 567 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 568 {
 569         return true;
 570 }
 571
 572 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
 573 {
 574         return true;
 575 }
 576
 577 static inline unsigned long
 578 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 579 {
 580         return 0;
 581 }
 582
 583 static inline void
 584 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 585                               int increment)
 586 {
 587 }
 588
 589 static inline void
 590 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 591 {
 592 }
 593
 594 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
 595 {
 596         return NULL;
 597 }
 598
 599 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
 600 {
 601 }
 602
 603 static inline void mem_cgroup_handle_over_high(void)
 604 {
 605 }
 606
 607 static inline void mem_cgroup_oom_enable(void)
 608 {
 609 }
 610
 611 static inline void mem_cgroup_oom_disable(void)
 612 {
 613 }
 614
 615 static inline bool task_in_memcg_oom(struct task_struct *p)
 616 {
 617         return false;
 618 }
 619
 620 static inline bool mem_cgroup_oom_synchronize(bool wait)
 621 {
 622         return false;
 623 }
 624
 625 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
 626                                             enum mem_cgroup_stat_index idx)
 627 {
 628 }
 629
 630 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
 631                                             enum mem_cgroup_stat_index idx)
 632 {
 633 }
 634
 635 static inline
 636 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 637                                             gfp_t gfp_mask,
 638                                             unsigned long *total_scanned)
 639 {
 640         return 0;
 641 }
 642
 643 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 644 {
 645 }
 646
 647 static inline
 648 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 649 {
 650 }
 651 #endif /* CONFIG_MEMCG */
 652
 653 #ifdef CONFIG_CGROUP_WRITEBACK
 654
 655 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
 656 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
 657 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
 658                          unsigned long *pheadroom, unsigned long *pdirty,
 659                          unsigned long *pwriteback);
 660
 661 #else   /* CONFIG_CGROUP_WRITEBACK */
 662
 663 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
 664 {
 665         return NULL;
 666 }
 667
 668 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
 669                                        unsigned long *pfilepages,
 670                                        unsigned long *pheadroom,
 671                                        unsigned long *pdirty,
 672                                        unsigned long *pwriteback)
 673 {
 674 }
 675
 676 #endif  /* CONFIG_CGROUP_WRITEBACK */
 677
 678 struct sock;
 679 void sock_update_memcg(struct sock *sk);
 680 void sock_release_memcg(struct sock *sk);
 681 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 682 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
 683 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
 684 extern struct static_key memcg_sockets_enabled_key;
 685 #define mem_cgroup_sockets_enabled static_key_false(&memcg_sockets_enabled_key)
 686 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 687 {
 688         return memcg->tcp_mem.memory_pressure;
 689 }
 690 #else
 691 #define mem_cgroup_sockets_enabled 0
 692 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 693 {
 694         return false;
 695 }
 696 #endif
 697
 698 #ifdef CONFIG_MEMCG_KMEM
 699 extern struct static_key memcg_kmem_enabled_key;
 700
 701 extern int memcg_nr_cache_ids;
 702 void memcg_get_cache_ids(void);
 703 void memcg_put_cache_ids(void);
 704
 705 /*
 706  * Helper macro to loop through all memcg-specific caches. Callers must still
 707  * check if the cache is valid (it is either valid or NULL).
 708  * the slab_mutex must be held when looping through those caches
 709  */
 710 #define for_each_memcg_cache_index(_idx)        \
 711         for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
 712
 713 static inline bool memcg_kmem_enabled(void)
 714 {
 715         return static_key_false(&memcg_kmem_enabled_key);
 716 }
 717
 718 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
 719 {
 720         return memcg->kmem_acct_active;
 721 }
 722
 723 /*
 724  * In general, we'll do everything in our power to not incur in any overhead
 725  * for non-memcg users for the kmem functions. Not even a function call, if we
 726  * can avoid it.
 727  *
 728  * Therefore, we'll inline all those functions so that in the best case, we'll
 729  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 730  * we'll still do most of the flag checking inline. We check a lot of
 731  * conditions, but because they are pretty simple, they are expected to be
 732  * fast.
 733  */
 734 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
 735                               struct mem_cgroup *memcg);
 736 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
 737 void __memcg_kmem_uncharge(struct page *page, int order);
 738
 739 /*
 740  * helper for acessing a memcg's index. It will be used as an index in the
 741  * child cache array in kmem_cache, and also to derive its name. This function
 742  * will return -1 when this is not a kmem-limited memcg.
 743  */
 744 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 745 {
 746         return memcg ? memcg->kmemcg_id : -1;
 747 }
 748
 749 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
 750 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
 751
 752 static inline bool __memcg_kmem_bypass(void)
 753 {
 754         if (!memcg_kmem_enabled())
 755                 return true;
 756         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 757                 return true;
 758         return false;
 759 }
 760
 761 /**
 762  * memcg_kmem_charge: charge a kmem page
 763  * @page: page to charge
 764  * @gfp: reclaim mode
 765  * @order: allocation order
 766  *
 767  * Returns 0 on success, an error code on failure.
 768  */
 769 static __always_inline int memcg_kmem_charge(struct page *page,
 770                                              gfp_t gfp, int order)
 771 {
 772         if (__memcg_kmem_bypass())
 773                 return 0;
 774         if (!(gfp & __GFP_ACCOUNT))
 775                 return 0;
 776         return __memcg_kmem_charge(page, gfp, order);
 777 }
 778
 779 /**
 780  * memcg_kmem_uncharge: uncharge a kmem page
 781  * @page: page to uncharge
 782  * @order: allocation order
 783  */
 784 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
 785 {
 786         if (memcg_kmem_enabled())
 787                 __memcg_kmem_uncharge(page, order);
 788 }
 789
 790 /**
 791  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 792  * @cachep: the original global kmem cache
 793  *
 794  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 795  */
 796 static __always_inline struct kmem_cache *
 797 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 798 {
 799         if (__memcg_kmem_bypass())
 800                 return cachep;
 801         return __memcg_kmem_get_cache(cachep, gfp);
 802 }
 803
 804 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 805 {
 806         if (memcg_kmem_enabled())
 807                 __memcg_kmem_put_cache(cachep);
 808 }
 809 #else
 810 #define for_each_memcg_cache_index(_idx)        \
 811         for (; NULL; )
 812
 813 static inline bool memcg_kmem_enabled(void)
 814 {
 815         return false;
 816 }
 817
 818 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
 819 {
 820         return false;
 821 }
 822
 823 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
 824 {
 825         return 0;
 826 }
 827
 828 static inline void memcg_kmem_uncharge(struct page *page, int order)
 829 {
 830 }
 831
 832 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 833 {
 834         return -1;
 835 }
 836
 837 static inline void memcg_get_cache_ids(void)
 838 {
 839 }
 840
 841 static inline void memcg_put_cache_ids(void)
 842 {
 843 }
 844
 845 static inline struct kmem_cache *
 846 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 847 {
 848         return cachep;
 849 }
 850
 851 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
 852 {
 853 }
 854 #endif /* CONFIG_MEMCG_KMEM */
 855 #endif /* _LINUX_MEMCONTROL_H */