linux/include/linux/memcontrol.h
<<
>>
Prefs
   1/* SPDX-License-Identifier: GPL-2.0-or-later */
   2/* memcontrol.h - Memory Controller
   3 *
   4 * Copyright IBM Corporation, 2007
   5 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
   6 *
   7 * Copyright 2007 OpenVZ SWsoft Inc
   8 * Author: Pavel Emelianov <xemul@openvz.org>
   9 */
  10
  11#ifndef _LINUX_MEMCONTROL_H
  12#define _LINUX_MEMCONTROL_H
  13#include <linux/cgroup.h>
  14#include <linux/vm_event_item.h>
  15#include <linux/hardirq.h>
  16#include <linux/jump_label.h>
  17#include <linux/page_counter.h>
  18#include <linux/vmpressure.h>
  19#include <linux/eventfd.h>
  20#include <linux/mm.h>
  21#include <linux/vmstat.h>
  22#include <linux/writeback.h>
  23#include <linux/page-flags.h>
  24
  25struct mem_cgroup;
  26struct obj_cgroup;
  27struct page;
  28struct mm_struct;
  29struct kmem_cache;
  30
  31/* Cgroup-specific page state, on top of universal node page state */
  32enum memcg_stat_item {
  33        MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS,
  34        MEMCG_SOCK,
  35        MEMCG_PERCPU_B,
  36        MEMCG_NR_STAT,
  37};
  38
  39enum memcg_memory_event {
  40        MEMCG_LOW,
  41        MEMCG_HIGH,
  42        MEMCG_MAX,
  43        MEMCG_OOM,
  44        MEMCG_OOM_KILL,
  45        MEMCG_SWAP_HIGH,
  46        MEMCG_SWAP_MAX,
  47        MEMCG_SWAP_FAIL,
  48        MEMCG_NR_MEMORY_EVENTS,
  49};
  50
  51struct mem_cgroup_reclaim_cookie {
  52        pg_data_t *pgdat;
  53        unsigned int generation;
  54};
  55
  56#ifdef CONFIG_MEMCG
  57
  58#define MEM_CGROUP_ID_SHIFT     16
  59#define MEM_CGROUP_ID_MAX       USHRT_MAX
  60
  61struct mem_cgroup_id {
  62        int id;
  63        refcount_t ref;
  64};
  65
  66/*
  67 * Per memcg event counter is incremented at every pagein/pageout. With THP,
  68 * it will be incremented by the number of pages. This counter is used
  69 * to trigger some periodic events. This is straightforward and better
  70 * than using jiffies etc. to handle periodic memcg event.
  71 */
  72enum mem_cgroup_events_target {
  73        MEM_CGROUP_TARGET_THRESH,
  74        MEM_CGROUP_TARGET_SOFTLIMIT,
  75        MEM_CGROUP_NTARGETS,
  76};
  77
  78struct memcg_vmstats_percpu {
  79        /* Local (CPU and cgroup) page state & events */
  80        long                    state[MEMCG_NR_STAT];
  81        unsigned long           events[NR_VM_EVENT_ITEMS];
  82
  83        /* Delta calculation for lockless upward propagation */
  84        long                    state_prev[MEMCG_NR_STAT];
  85        unsigned long           events_prev[NR_VM_EVENT_ITEMS];
  86
  87        /* Cgroup1: threshold notifications & softlimit tree updates */
  88        unsigned long           nr_page_events;
  89        unsigned long           targets[MEM_CGROUP_NTARGETS];
  90};
  91
  92struct memcg_vmstats {
  93        /* Aggregated (CPU and subtree) page state & events */
  94        long                    state[MEMCG_NR_STAT];
  95        unsigned long           events[NR_VM_EVENT_ITEMS];
  96
  97        /* Pending child counts during tree propagation */
  98        long                    state_pending[MEMCG_NR_STAT];
  99        unsigned long           events_pending[NR_VM_EVENT_ITEMS];
 100};
 101
 102struct mem_cgroup_reclaim_iter {
 103        struct mem_cgroup *position;
 104        /* scan generation, increased every round-trip */
 105        unsigned int generation;
 106};
 107
 108/*
 109 * Bitmap and deferred work of shrinker::id corresponding to memcg-aware
 110 * shrinkers, which have elements charged to this memcg.
 111 */
 112struct shrinker_info {
 113        struct rcu_head rcu;
 114        atomic_long_t *nr_deferred;
 115        unsigned long *map;
 116};
 117
 118struct lruvec_stats_percpu {
 119        /* Local (CPU and cgroup) state */
 120        long state[NR_VM_NODE_STAT_ITEMS];
 121
 122        /* Delta calculation for lockless upward propagation */
 123        long state_prev[NR_VM_NODE_STAT_ITEMS];
 124};
 125
 126struct lruvec_stats {
 127        /* Aggregated (CPU and subtree) state */
 128        long state[NR_VM_NODE_STAT_ITEMS];
 129
 130        /* Pending child counts during tree propagation */
 131        long state_pending[NR_VM_NODE_STAT_ITEMS];
 132};
 133
 134/*
 135 * per-node information in memory controller.
 136 */
 137struct mem_cgroup_per_node {
 138        struct lruvec           lruvec;
 139
 140        struct lruvec_stats_percpu __percpu     *lruvec_stats_percpu;
 141        struct lruvec_stats                     lruvec_stats;
 142
 143        unsigned long           lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
 144
 145        struct mem_cgroup_reclaim_iter  iter;
 146
 147        struct shrinker_info __rcu      *shrinker_info;
 148
 149        struct rb_node          tree_node;      /* RB tree node */
 150        unsigned long           usage_in_excess;/* Set to the value by which */
 151                                                /* the soft limit is exceeded*/
 152        bool                    on_tree;
 153        struct mem_cgroup       *memcg;         /* Back pointer, we cannot */
 154                                                /* use container_of        */
 155};
 156
 157struct mem_cgroup_threshold {
 158        struct eventfd_ctx *eventfd;
 159        unsigned long threshold;
 160};
 161
 162/* For threshold */
 163struct mem_cgroup_threshold_ary {
 164        /* An array index points to threshold just below or equal to usage. */
 165        int current_threshold;
 166        /* Size of entries[] */
 167        unsigned int size;
 168        /* Array of thresholds */
 169        struct mem_cgroup_threshold entries[];
 170};
 171
 172struct mem_cgroup_thresholds {
 173        /* Primary thresholds array */
 174        struct mem_cgroup_threshold_ary *primary;
 175        /*
 176         * Spare threshold array.
 177         * This is needed to make mem_cgroup_unregister_event() "never fail".
 178         * It must be able to store at least primary->size - 1 entries.
 179         */
 180        struct mem_cgroup_threshold_ary *spare;
 181};
 182
 183enum memcg_kmem_state {
 184        KMEM_NONE,
 185        KMEM_ALLOCATED,
 186        KMEM_ONLINE,
 187};
 188
 189#if defined(CONFIG_SMP)
 190struct memcg_padding {
 191        char x[0];
 192} ____cacheline_internodealigned_in_smp;
 193#define MEMCG_PADDING(name)      struct memcg_padding name
 194#else
 195#define MEMCG_PADDING(name)
 196#endif
 197
 198/*
 199 * Remember four most recent foreign writebacks with dirty pages in this
 200 * cgroup.  Inode sharing is expected to be uncommon and, even if we miss
 201 * one in a given round, we're likely to catch it later if it keeps
 202 * foreign-dirtying, so a fairly low count should be enough.
 203 *
 204 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
 205 */
 206#define MEMCG_CGWB_FRN_CNT      4
 207
 208struct memcg_cgwb_frn {
 209        u64 bdi_id;                     /* bdi->id of the foreign inode */
 210        int memcg_id;                   /* memcg->css.id of foreign inode */
 211        u64 at;                         /* jiffies_64 at the time of dirtying */
 212        struct wb_completion done;      /* tracks in-flight foreign writebacks */
 213};
 214
 215/*
 216 * Bucket for arbitrarily byte-sized objects charged to a memory
 217 * cgroup. The bucket can be reparented in one piece when the cgroup
 218 * is destroyed, without having to round up the individual references
 219 * of all live memory objects in the wild.
 220 */
 221struct obj_cgroup {
 222        struct percpu_ref refcnt;
 223        struct mem_cgroup *memcg;
 224        atomic_t nr_charged_bytes;
 225        union {
 226                struct list_head list;
 227                struct rcu_head rcu;
 228        };
 229};
 230
 231/*
 232 * The memory controller data structure. The memory controller controls both
 233 * page cache and RSS per cgroup. We would eventually like to provide
 234 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
 235 * to help the administrator determine what knobs to tune.
 236 */
 237struct mem_cgroup {
 238        struct cgroup_subsys_state css;
 239
 240        /* Private memcg ID. Used to ID objects that outlive the cgroup */
 241        struct mem_cgroup_id id;
 242
 243        /* Accounted resources */
 244        struct page_counter memory;             /* Both v1 & v2 */
 245
 246        union {
 247                struct page_counter swap;       /* v2 only */
 248                struct page_counter memsw;      /* v1 only */
 249        };
 250
 251        /* Legacy consumer-oriented counters */
 252        struct page_counter kmem;               /* v1 only */
 253        struct page_counter tcpmem;             /* v1 only */
 254
 255        /* Range enforcement for interrupt charges */
 256        struct work_struct high_work;
 257
 258        unsigned long soft_limit;
 259
 260        /* vmpressure notifications */
 261        struct vmpressure vmpressure;
 262
 263        /*
 264         * Should the OOM killer kill all belonging tasks, had it kill one?
 265         */
 266        bool oom_group;
 267
 268        /* protected by memcg_oom_lock */
 269        bool            oom_lock;
 270        int             under_oom;
 271
 272        int     swappiness;
 273        /* OOM-Killer disable */
 274        int             oom_kill_disable;
 275
 276        /* memory.events and memory.events.local */
 277        struct cgroup_file events_file;
 278        struct cgroup_file events_local_file;
 279
 280        /* handle for "memory.swap.events" */
 281        struct cgroup_file swap_events_file;
 282
 283        /* protect arrays of thresholds */
 284        struct mutex thresholds_lock;
 285
 286        /* thresholds for memory usage. RCU-protected */
 287        struct mem_cgroup_thresholds thresholds;
 288
 289        /* thresholds for mem+swap usage. RCU-protected */
 290        struct mem_cgroup_thresholds memsw_thresholds;
 291
 292        /* For oom notifier event fd */
 293        struct list_head oom_notify;
 294
 295        /*
 296         * Should we move charges of a task when a task is moved into this
 297         * mem_cgroup ? And what type of charges should we move ?
 298         */
 299        unsigned long move_charge_at_immigrate;
 300        /* taken only while moving_account > 0 */
 301        spinlock_t              move_lock;
 302        unsigned long           move_lock_flags;
 303
 304        MEMCG_PADDING(_pad1_);
 305
 306        /* memory.stat */
 307        struct memcg_vmstats    vmstats;
 308
 309        /* memory.events */
 310        atomic_long_t           memory_events[MEMCG_NR_MEMORY_EVENTS];
 311        atomic_long_t           memory_events_local[MEMCG_NR_MEMORY_EVENTS];
 312
 313        unsigned long           socket_pressure;
 314
 315        /* Legacy tcp memory accounting */
 316        bool                    tcpmem_active;
 317        int                     tcpmem_pressure;
 318
 319#ifdef CONFIG_MEMCG_KMEM
 320        int kmemcg_id;
 321        enum memcg_kmem_state kmem_state;
 322        struct obj_cgroup __rcu *objcg;
 323        struct list_head objcg_list; /* list of inherited objcgs */
 324#endif
 325
 326        MEMCG_PADDING(_pad2_);
 327
 328        /*
 329         * set > 0 if pages under this cgroup are moving to other cgroup.
 330         */
 331        atomic_t                moving_account;
 332        struct task_struct      *move_lock_task;
 333
 334        struct memcg_vmstats_percpu __percpu *vmstats_percpu;
 335
 336#ifdef CONFIG_CGROUP_WRITEBACK
 337        struct list_head cgwb_list;
 338        struct wb_domain cgwb_domain;
 339        struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
 340#endif
 341
 342        /* List of events which userspace want to receive */
 343        struct list_head event_list;
 344        spinlock_t event_list_lock;
 345
 346#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 347        struct deferred_split deferred_split_queue;
 348#endif
 349
 350        struct mem_cgroup_per_node *nodeinfo[];
 351};
 352
 353/*
 354 * size of first charge trial. "32" comes from vmscan.c's magic value.
 355 * TODO: maybe necessary to use big numbers in big irons.
 356 */
 357#define MEMCG_CHARGE_BATCH 32U
 358
 359extern struct mem_cgroup *root_mem_cgroup;
 360
 361enum page_memcg_data_flags {
 362        /* page->memcg_data is a pointer to an objcgs vector */
 363        MEMCG_DATA_OBJCGS = (1UL << 0),
 364        /* page has been accounted as a non-slab kernel page */
 365        MEMCG_DATA_KMEM = (1UL << 1),
 366        /* the next bit after the last actual flag */
 367        __NR_MEMCG_DATA_FLAGS  = (1UL << 2),
 368};
 369
 370#define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1)
 371
 372static inline bool PageMemcgKmem(struct page *page);
 373
 374/*
 375 * After the initialization objcg->memcg is always pointing at
 376 * a valid memcg, but can be atomically swapped to the parent memcg.
 377 *
 378 * The caller must ensure that the returned memcg won't be released:
 379 * e.g. acquire the rcu_read_lock or css_set_lock.
 380 */
 381static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
 382{
 383        return READ_ONCE(objcg->memcg);
 384}
 385
 386/*
 387 * __page_memcg - get the memory cgroup associated with a non-kmem page
 388 * @page: a pointer to the page struct
 389 *
 390 * Returns a pointer to the memory cgroup associated with the page,
 391 * or NULL. This function assumes that the page is known to have a
 392 * proper memory cgroup pointer. It's not safe to call this function
 393 * against some type of pages, e.g. slab pages or ex-slab pages or
 394 * kmem pages.
 395 */
 396static inline struct mem_cgroup *__page_memcg(struct page *page)
 397{
 398        unsigned long memcg_data = page->memcg_data;
 399
 400        VM_BUG_ON_PAGE(PageSlab(page), page);
 401        VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_OBJCGS, page);
 402        VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
 403
 404        return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 405}
 406
 407/*
 408 * __page_objcg - get the object cgroup associated with a kmem page
 409 * @page: a pointer to the page struct
 410 *
 411 * Returns a pointer to the object cgroup associated with the page,
 412 * or NULL. This function assumes that the page is known to have a
 413 * proper object cgroup pointer. It's not safe to call this function
 414 * against some type of pages, e.g. slab pages or ex-slab pages or
 415 * LRU pages.
 416 */
 417static inline struct obj_cgroup *__page_objcg(struct page *page)
 418{
 419        unsigned long memcg_data = page->memcg_data;
 420
 421        VM_BUG_ON_PAGE(PageSlab(page), page);
 422        VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_OBJCGS, page);
 423        VM_BUG_ON_PAGE(!(memcg_data & MEMCG_DATA_KMEM), page);
 424
 425        return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 426}
 427
 428/*
 429 * page_memcg - get the memory cgroup associated with a page
 430 * @page: a pointer to the page struct
 431 *
 432 * Returns a pointer to the memory cgroup associated with the page,
 433 * or NULL. This function assumes that the page is known to have a
 434 * proper memory cgroup pointer. It's not safe to call this function
 435 * against some type of pages, e.g. slab pages or ex-slab pages.
 436 *
 437 * For a non-kmem page any of the following ensures page and memcg binding
 438 * stability:
 439 *
 440 * - the page lock
 441 * - LRU isolation
 442 * - lock_page_memcg()
 443 * - exclusive reference
 444 *
 445 * For a kmem page a caller should hold an rcu read lock to protect memcg
 446 * associated with a kmem page from being released.
 447 */
 448static inline struct mem_cgroup *page_memcg(struct page *page)
 449{
 450        if (PageMemcgKmem(page))
 451                return obj_cgroup_memcg(__page_objcg(page));
 452        else
 453                return __page_memcg(page);
 454}
 455
 456/*
 457 * page_memcg_rcu - locklessly get the memory cgroup associated with a page
 458 * @page: a pointer to the page struct
 459 *
 460 * Returns a pointer to the memory cgroup associated with the page,
 461 * or NULL. This function assumes that the page is known to have a
 462 * proper memory cgroup pointer. It's not safe to call this function
 463 * against some type of pages, e.g. slab pages or ex-slab pages.
 464 */
 465static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
 466{
 467        unsigned long memcg_data = READ_ONCE(page->memcg_data);
 468
 469        VM_BUG_ON_PAGE(PageSlab(page), page);
 470        WARN_ON_ONCE(!rcu_read_lock_held());
 471
 472        if (memcg_data & MEMCG_DATA_KMEM) {
 473                struct obj_cgroup *objcg;
 474
 475                objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 476                return obj_cgroup_memcg(objcg);
 477        }
 478
 479        return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 480}
 481
 482/*
 483 * page_memcg_check - get the memory cgroup associated with a page
 484 * @page: a pointer to the page struct
 485 *
 486 * Returns a pointer to the memory cgroup associated with the page,
 487 * or NULL. This function unlike page_memcg() can take any page
 488 * as an argument. It has to be used in cases when it's not known if a page
 489 * has an associated memory cgroup pointer or an object cgroups vector or
 490 * an object cgroup.
 491 *
 492 * For a non-kmem page any of the following ensures page and memcg binding
 493 * stability:
 494 *
 495 * - the page lock
 496 * - LRU isolation
 497 * - lock_page_memcg()
 498 * - exclusive reference
 499 *
 500 * For a kmem page a caller should hold an rcu read lock to protect memcg
 501 * associated with a kmem page from being released.
 502 */
 503static inline struct mem_cgroup *page_memcg_check(struct page *page)
 504{
 505        /*
 506         * Because page->memcg_data might be changed asynchronously
 507         * for slab pages, READ_ONCE() should be used here.
 508         */
 509        unsigned long memcg_data = READ_ONCE(page->memcg_data);
 510
 511        if (memcg_data & MEMCG_DATA_OBJCGS)
 512                return NULL;
 513
 514        if (memcg_data & MEMCG_DATA_KMEM) {
 515                struct obj_cgroup *objcg;
 516
 517                objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 518                return obj_cgroup_memcg(objcg);
 519        }
 520
 521        return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 522}
 523
 524#ifdef CONFIG_MEMCG_KMEM
 525/*
 526 * PageMemcgKmem - check if the page has MemcgKmem flag set
 527 * @page: a pointer to the page struct
 528 *
 529 * Checks if the page has MemcgKmem flag set. The caller must ensure that
 530 * the page has an associated memory cgroup. It's not safe to call this function
 531 * against some types of pages, e.g. slab pages.
 532 */
 533static inline bool PageMemcgKmem(struct page *page)
 534{
 535        VM_BUG_ON_PAGE(page->memcg_data & MEMCG_DATA_OBJCGS, page);
 536        return page->memcg_data & MEMCG_DATA_KMEM;
 537}
 538
 539/*
 540 * page_objcgs - get the object cgroups vector associated with a page
 541 * @page: a pointer to the page struct
 542 *
 543 * Returns a pointer to the object cgroups vector associated with the page,
 544 * or NULL. This function assumes that the page is known to have an
 545 * associated object cgroups vector. It's not safe to call this function
 546 * against pages, which might have an associated memory cgroup: e.g.
 547 * kernel stack pages.
 548 */
 549static inline struct obj_cgroup **page_objcgs(struct page *page)
 550{
 551        unsigned long memcg_data = READ_ONCE(page->memcg_data);
 552
 553        VM_BUG_ON_PAGE(memcg_data && !(memcg_data & MEMCG_DATA_OBJCGS), page);
 554        VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
 555
 556        return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 557}
 558
 559/*
 560 * page_objcgs_check - get the object cgroups vector associated with a page
 561 * @page: a pointer to the page struct
 562 *
 563 * Returns a pointer to the object cgroups vector associated with the page,
 564 * or NULL. This function is safe to use if the page can be directly associated
 565 * with a memory cgroup.
 566 */
 567static inline struct obj_cgroup **page_objcgs_check(struct page *page)
 568{
 569        unsigned long memcg_data = READ_ONCE(page->memcg_data);
 570
 571        if (!memcg_data || !(memcg_data & MEMCG_DATA_OBJCGS))
 572                return NULL;
 573
 574        VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
 575
 576        return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
 577}
 578
 579#else
 580static inline bool PageMemcgKmem(struct page *page)
 581{
 582        return false;
 583}
 584
 585static inline struct obj_cgroup **page_objcgs(struct page *page)
 586{
 587        return NULL;
 588}
 589
 590static inline struct obj_cgroup **page_objcgs_check(struct page *page)
 591{
 592        return NULL;
 593}
 594#endif
 595
 596static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
 597{
 598        return (memcg == root_mem_cgroup);
 599}
 600
 601static inline bool mem_cgroup_disabled(void)
 602{
 603        return !cgroup_subsys_enabled(memory_cgrp_subsys);
 604}
 605
 606static inline void mem_cgroup_protection(struct mem_cgroup *root,
 607                                         struct mem_cgroup *memcg,
 608                                         unsigned long *min,
 609                                         unsigned long *low)
 610{
 611        *min = *low = 0;
 612
 613        if (mem_cgroup_disabled())
 614                return;
 615
 616        /*
 617         * There is no reclaim protection applied to a targeted reclaim.
 618         * We are special casing this specific case here because
 619         * mem_cgroup_protected calculation is not robust enough to keep
 620         * the protection invariant for calculated effective values for
 621         * parallel reclaimers with different reclaim target. This is
 622         * especially a problem for tail memcgs (as they have pages on LRU)
 623         * which would want to have effective values 0 for targeted reclaim
 624         * but a different value for external reclaim.
 625         *
 626         * Example
 627         * Let's have global and A's reclaim in parallel:
 628         *  |
 629         *  A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
 630         *  |\
 631         *  | C (low = 1G, usage = 2.5G)
 632         *  B (low = 1G, usage = 0.5G)
 633         *
 634         * For the global reclaim
 635         * A.elow = A.low
 636         * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
 637         * C.elow = min(C.usage, C.low)
 638         *
 639         * With the effective values resetting we have A reclaim
 640         * A.elow = 0
 641         * B.elow = B.low
 642         * C.elow = C.low
 643         *
 644         * If the global reclaim races with A's reclaim then
 645         * B.elow = C.elow = 0 because children_low_usage > A.elow)
 646         * is possible and reclaiming B would be violating the protection.
 647         *
 648         */
 649        if (root == memcg)
 650                return;
 651
 652        *min = READ_ONCE(memcg->memory.emin);
 653        *low = READ_ONCE(memcg->memory.elow);
 654}
 655
 656void mem_cgroup_calculate_protection(struct mem_cgroup *root,
 657                                     struct mem_cgroup *memcg);
 658
 659static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg)
 660{
 661        /*
 662         * The root memcg doesn't account charges, and doesn't support
 663         * protection.
 664         */
 665        return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg);
 666
 667}
 668
 669static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
 670{
 671        if (!mem_cgroup_supports_protection(memcg))
 672                return false;
 673
 674        return READ_ONCE(memcg->memory.elow) >=
 675                page_counter_read(&memcg->memory);
 676}
 677
 678static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
 679{
 680        if (!mem_cgroup_supports_protection(memcg))
 681                return false;
 682
 683        return READ_ONCE(memcg->memory.emin) >=
 684                page_counter_read(&memcg->memory);
 685}
 686
 687int __mem_cgroup_charge(struct page *page, struct mm_struct *mm,
 688                        gfp_t gfp_mask);
 689static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
 690                                    gfp_t gfp_mask)
 691{
 692        if (mem_cgroup_disabled())
 693                return 0;
 694        return __mem_cgroup_charge(page, mm, gfp_mask);
 695}
 696
 697int mem_cgroup_swapin_charge_page(struct page *page, struct mm_struct *mm,
 698                                  gfp_t gfp, swp_entry_t entry);
 699void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry);
 700
 701void __mem_cgroup_uncharge(struct page *page);
 702static inline void mem_cgroup_uncharge(struct page *page)
 703{
 704        if (mem_cgroup_disabled())
 705                return;
 706        __mem_cgroup_uncharge(page);
 707}
 708
 709void __mem_cgroup_uncharge_list(struct list_head *page_list);
 710static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
 711{
 712        if (mem_cgroup_disabled())
 713                return;
 714        __mem_cgroup_uncharge_list(page_list);
 715}
 716
 717void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
 718
 719/**
 720 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
 721 * @memcg: memcg of the wanted lruvec
 722 * @pgdat: pglist_data
 723 *
 724 * Returns the lru list vector holding pages for a given @memcg &
 725 * @pgdat combination. This can be the node lruvec, if the memory
 726 * controller is disabled.
 727 */
 728static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
 729                                               struct pglist_data *pgdat)
 730{
 731        struct mem_cgroup_per_node *mz;
 732        struct lruvec *lruvec;
 733
 734        if (mem_cgroup_disabled()) {
 735                lruvec = &pgdat->__lruvec;
 736                goto out;
 737        }
 738
 739        if (!memcg)
 740                memcg = root_mem_cgroup;
 741
 742        mz = memcg->nodeinfo[pgdat->node_id];
 743        lruvec = &mz->lruvec;
 744out:
 745        /*
 746         * Since a node can be onlined after the mem_cgroup was created,
 747         * we have to be prepared to initialize lruvec->pgdat here;
 748         * and if offlined then reonlined, we need to reinitialize it.
 749         */
 750        if (unlikely(lruvec->pgdat != pgdat))
 751                lruvec->pgdat = pgdat;
 752        return lruvec;
 753}
 754
 755/**
 756 * mem_cgroup_page_lruvec - return lruvec for isolating/putting an LRU page
 757 * @page: the page
 758 *
 759 * This function relies on page->mem_cgroup being stable.
 760 */
 761static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page)
 762{
 763        pg_data_t *pgdat = page_pgdat(page);
 764        struct mem_cgroup *memcg = page_memcg(page);
 765
 766        VM_WARN_ON_ONCE_PAGE(!memcg && !mem_cgroup_disabled(), page);
 767        return mem_cgroup_lruvec(memcg, pgdat);
 768}
 769
 770struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
 771
 772struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
 773
 774struct lruvec *lock_page_lruvec(struct page *page);
 775struct lruvec *lock_page_lruvec_irq(struct page *page);
 776struct lruvec *lock_page_lruvec_irqsave(struct page *page,
 777                                                unsigned long *flags);
 778
 779#ifdef CONFIG_DEBUG_VM
 780void lruvec_memcg_debug(struct lruvec *lruvec, struct page *page);
 781#else
 782static inline void lruvec_memcg_debug(struct lruvec *lruvec, struct page *page)
 783{
 784}
 785#endif
 786
 787static inline
 788struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
 789        return css ? container_of(css, struct mem_cgroup, css) : NULL;
 790}
 791
 792static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
 793{
 794        return percpu_ref_tryget(&objcg->refcnt);
 795}
 796
 797static inline void obj_cgroup_get(struct obj_cgroup *objcg)
 798{
 799        percpu_ref_get(&objcg->refcnt);
 800}
 801
 802static inline void obj_cgroup_get_many(struct obj_cgroup *objcg,
 803                                       unsigned long nr)
 804{
 805        percpu_ref_get_many(&objcg->refcnt, nr);
 806}
 807
 808static inline void obj_cgroup_put(struct obj_cgroup *objcg)
 809{
 810        percpu_ref_put(&objcg->refcnt);
 811}
 812
 813static inline void mem_cgroup_put(struct mem_cgroup *memcg)
 814{
 815        if (memcg)
 816                css_put(&memcg->css);
 817}
 818
 819#define mem_cgroup_from_counter(counter, member)        \
 820        container_of(counter, struct mem_cgroup, member)
 821
 822struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 823                                   struct mem_cgroup *,
 824                                   struct mem_cgroup_reclaim_cookie *);
 825void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 826int mem_cgroup_scan_tasks(struct mem_cgroup *,
 827                          int (*)(struct task_struct *, void *), void *);
 828
 829static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
 830{
 831        if (mem_cgroup_disabled())
 832                return 0;
 833
 834        return memcg->id.id;
 835}
 836struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
 837
 838static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
 839{
 840        return mem_cgroup_from_css(seq_css(m));
 841}
 842
 843static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
 844{
 845        struct mem_cgroup_per_node *mz;
 846
 847        if (mem_cgroup_disabled())
 848                return NULL;
 849
 850        mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
 851        return mz->memcg;
 852}
 853
 854/**
 855 * parent_mem_cgroup - find the accounting parent of a memcg
 856 * @memcg: memcg whose parent to find
 857 *
 858 * Returns the parent memcg, or NULL if this is the root or the memory
 859 * controller is in legacy no-hierarchy mode.
 860 */
 861static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
 862{
 863        if (!memcg->memory.parent)
 864                return NULL;
 865        return mem_cgroup_from_counter(memcg->memory.parent, memory);
 866}
 867
 868static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
 869                              struct mem_cgroup *root)
 870{
 871        if (root == memcg)
 872                return true;
 873        return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
 874}
 875
 876static inline bool mm_match_cgroup(struct mm_struct *mm,
 877                                   struct mem_cgroup *memcg)
 878{
 879        struct mem_cgroup *task_memcg;
 880        bool match = false;
 881
 882        rcu_read_lock();
 883        task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 884        if (task_memcg)
 885                match = mem_cgroup_is_descendant(task_memcg, memcg);
 886        rcu_read_unlock();
 887        return match;
 888}
 889
 890struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
 891ino_t page_cgroup_ino(struct page *page);
 892
 893static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
 894{
 895        if (mem_cgroup_disabled())
 896                return true;
 897        return !!(memcg->css.flags & CSS_ONLINE);
 898}
 899
 900void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 901                int zid, int nr_pages);
 902
 903static inline
 904unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
 905                enum lru_list lru, int zone_idx)
 906{
 907        struct mem_cgroup_per_node *mz;
 908
 909        mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
 910        return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
 911}
 912
 913void mem_cgroup_handle_over_high(void);
 914
 915unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
 916
 917unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
 918
 919void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
 920                                struct task_struct *p);
 921
 922void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
 923
 924static inline void mem_cgroup_enter_user_fault(void)
 925{
 926        WARN_ON(current->in_user_fault);
 927        current->in_user_fault = 1;
 928}
 929
 930static inline void mem_cgroup_exit_user_fault(void)
 931{
 932        WARN_ON(!current->in_user_fault);
 933        current->in_user_fault = 0;
 934}
 935
 936static inline bool task_in_memcg_oom(struct task_struct *p)
 937{
 938        return p->memcg_in_oom;
 939}
 940
 941bool mem_cgroup_oom_synchronize(bool wait);
 942struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
 943                                            struct mem_cgroup *oom_domain);
 944void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
 945
 946#ifdef CONFIG_MEMCG_SWAP
 947extern bool cgroup_memory_noswap;
 948#endif
 949
 950void lock_page_memcg(struct page *page);
 951void unlock_page_memcg(struct page *page);
 952
 953void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
 954
 955/* idx can be of type enum memcg_stat_item or node_stat_item */
 956static inline void mod_memcg_state(struct mem_cgroup *memcg,
 957                                   int idx, int val)
 958{
 959        unsigned long flags;
 960
 961        local_irq_save(flags);
 962        __mod_memcg_state(memcg, idx, val);
 963        local_irq_restore(flags);
 964}
 965
 966static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
 967{
 968        return READ_ONCE(memcg->vmstats.state[idx]);
 969}
 970
 971static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
 972                                              enum node_stat_item idx)
 973{
 974        struct mem_cgroup_per_node *pn;
 975
 976        if (mem_cgroup_disabled())
 977                return node_page_state(lruvec_pgdat(lruvec), idx);
 978
 979        pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
 980        return READ_ONCE(pn->lruvec_stats.state[idx]);
 981}
 982
 983static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
 984                                                    enum node_stat_item idx)
 985{
 986        struct mem_cgroup_per_node *pn;
 987        long x = 0;
 988        int cpu;
 989
 990        if (mem_cgroup_disabled())
 991                return node_page_state(lruvec_pgdat(lruvec), idx);
 992
 993        pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
 994        for_each_possible_cpu(cpu)
 995                x += per_cpu(pn->lruvec_stats_percpu->state[idx], cpu);
 996#ifdef CONFIG_SMP
 997        if (x < 0)
 998                x = 0;
 999#endif
1000        return x;
1001}
1002
1003void mem_cgroup_flush_stats(void);
1004
1005void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
1006                              int val);
1007void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
1008
1009static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1010                                         int val)
1011{
1012        unsigned long flags;
1013
1014        local_irq_save(flags);
1015        __mod_lruvec_kmem_state(p, idx, val);
1016        local_irq_restore(flags);
1017}
1018
1019static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
1020                                          enum node_stat_item idx, int val)
1021{
1022        unsigned long flags;
1023
1024        local_irq_save(flags);
1025        __mod_memcg_lruvec_state(lruvec, idx, val);
1026        local_irq_restore(flags);
1027}
1028
1029void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
1030                          unsigned long count);
1031
1032static inline void count_memcg_events(struct mem_cgroup *memcg,
1033                                      enum vm_event_item idx,
1034                                      unsigned long count)
1035{
1036        unsigned long flags;
1037
1038        local_irq_save(flags);
1039        __count_memcg_events(memcg, idx, count);
1040        local_irq_restore(flags);
1041}
1042
1043static inline void count_memcg_page_event(struct page *page,
1044                                          enum vm_event_item idx)
1045{
1046        struct mem_cgroup *memcg = page_memcg(page);
1047
1048        if (memcg)
1049                count_memcg_events(memcg, idx, 1);
1050}
1051
1052static inline void count_memcg_event_mm(struct mm_struct *mm,
1053                                        enum vm_event_item idx)
1054{
1055        struct mem_cgroup *memcg;
1056
1057        if (mem_cgroup_disabled())
1058                return;
1059
1060        rcu_read_lock();
1061        memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1062        if (likely(memcg))
1063                count_memcg_events(memcg, idx, 1);
1064        rcu_read_unlock();
1065}
1066
1067static inline void memcg_memory_event(struct mem_cgroup *memcg,
1068                                      enum memcg_memory_event event)
1069{
1070        bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
1071                          event == MEMCG_SWAP_FAIL;
1072
1073        atomic_long_inc(&memcg->memory_events_local[event]);
1074        if (!swap_event)
1075                cgroup_file_notify(&memcg->events_local_file);
1076
1077        do {
1078                atomic_long_inc(&memcg->memory_events[event]);
1079                if (swap_event)
1080                        cgroup_file_notify(&memcg->swap_events_file);
1081                else
1082                        cgroup_file_notify(&memcg->events_file);
1083
1084                if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1085                        break;
1086                if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1087                        break;
1088        } while ((memcg = parent_mem_cgroup(memcg)) &&
1089                 !mem_cgroup_is_root(memcg));
1090}
1091
1092static inline void memcg_memory_event_mm(struct mm_struct *mm,
1093                                         enum memcg_memory_event event)
1094{
1095        struct mem_cgroup *memcg;
1096
1097        if (mem_cgroup_disabled())
1098                return;
1099
1100        rcu_read_lock();
1101        memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1102        if (likely(memcg))
1103                memcg_memory_event(memcg, event);
1104        rcu_read_unlock();
1105}
1106
1107void split_page_memcg(struct page *head, unsigned int nr);
1108
1109unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1110                                                gfp_t gfp_mask,
1111                                                unsigned long *total_scanned);
1112
1113#else /* CONFIG_MEMCG */
1114
1115#define MEM_CGROUP_ID_SHIFT     0
1116#define MEM_CGROUP_ID_MAX       0
1117
1118static inline struct mem_cgroup *page_memcg(struct page *page)
1119{
1120        return NULL;
1121}
1122
1123static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1124{
1125        WARN_ON_ONCE(!rcu_read_lock_held());
1126        return NULL;
1127}
1128
1129static inline struct mem_cgroup *page_memcg_check(struct page *page)
1130{
1131        return NULL;
1132}
1133
1134static inline bool PageMemcgKmem(struct page *page)
1135{
1136        return false;
1137}
1138
1139static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1140{
1141        return true;
1142}
1143
1144static inline bool mem_cgroup_disabled(void)
1145{
1146        return true;
1147}
1148
1149static inline void memcg_memory_event(struct mem_cgroup *memcg,
1150                                      enum memcg_memory_event event)
1151{
1152}
1153
1154static inline void memcg_memory_event_mm(struct mm_struct *mm,
1155                                         enum memcg_memory_event event)
1156{
1157}
1158
1159static inline void mem_cgroup_protection(struct mem_cgroup *root,
1160                                         struct mem_cgroup *memcg,
1161                                         unsigned long *min,
1162                                         unsigned long *low)
1163{
1164        *min = *low = 0;
1165}
1166
1167static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1168                                                   struct mem_cgroup *memcg)
1169{
1170}
1171
1172static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
1173{
1174        return false;
1175}
1176
1177static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
1178{
1179        return false;
1180}
1181
1182static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
1183                                    gfp_t gfp_mask)
1184{
1185        return 0;
1186}
1187
1188static inline int mem_cgroup_swapin_charge_page(struct page *page,
1189                        struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
1190{
1191        return 0;
1192}
1193
1194static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry)
1195{
1196}
1197
1198static inline void mem_cgroup_uncharge(struct page *page)
1199{
1200}
1201
1202static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
1203{
1204}
1205
1206static inline void mem_cgroup_migrate(struct page *old, struct page *new)
1207{
1208}
1209
1210static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1211                                               struct pglist_data *pgdat)
1212{
1213        return &pgdat->__lruvec;
1214}
1215
1216static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page)
1217{
1218        pg_data_t *pgdat = page_pgdat(page);
1219
1220        return &pgdat->__lruvec;
1221}
1222
1223static inline void lruvec_memcg_debug(struct lruvec *lruvec, struct page *page)
1224{
1225}
1226
1227static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1228{
1229        return NULL;
1230}
1231
1232static inline bool mm_match_cgroup(struct mm_struct *mm,
1233                struct mem_cgroup *memcg)
1234{
1235        return true;
1236}
1237
1238static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1239{
1240        return NULL;
1241}
1242
1243static inline
1244struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css)
1245{
1246        return NULL;
1247}
1248
1249static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1250{
1251}
1252
1253static inline struct lruvec *lock_page_lruvec(struct page *page)
1254{
1255        struct pglist_data *pgdat = page_pgdat(page);
1256
1257        spin_lock(&pgdat->__lruvec.lru_lock);
1258        return &pgdat->__lruvec;
1259}
1260
1261static inline struct lruvec *lock_page_lruvec_irq(struct page *page)
1262{
1263        struct pglist_data *pgdat = page_pgdat(page);
1264
1265        spin_lock_irq(&pgdat->__lruvec.lru_lock);
1266        return &pgdat->__lruvec;
1267}
1268
1269static inline struct lruvec *lock_page_lruvec_irqsave(struct page *page,
1270                unsigned long *flagsp)
1271{
1272        struct pglist_data *pgdat = page_pgdat(page);
1273
1274        spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1275        return &pgdat->__lruvec;
1276}
1277
1278static inline struct mem_cgroup *
1279mem_cgroup_iter(struct mem_cgroup *root,
1280                struct mem_cgroup *prev,
1281                struct mem_cgroup_reclaim_cookie *reclaim)
1282{
1283        return NULL;
1284}
1285
1286static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1287                                         struct mem_cgroup *prev)
1288{
1289}
1290
1291static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1292                int (*fn)(struct task_struct *, void *), void *arg)
1293{
1294        return 0;
1295}
1296
1297static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1298{
1299        return 0;
1300}
1301
1302static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1303{
1304        WARN_ON_ONCE(id);
1305        /* XXX: This should always return root_mem_cgroup */
1306        return NULL;
1307}
1308
1309static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1310{
1311        return NULL;
1312}
1313
1314static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1315{
1316        return NULL;
1317}
1318
1319static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1320{
1321        return true;
1322}
1323
1324static inline
1325unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1326                enum lru_list lru, int zone_idx)
1327{
1328        return 0;
1329}
1330
1331static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1332{
1333        return 0;
1334}
1335
1336static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1337{
1338        return 0;
1339}
1340
1341static inline void
1342mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1343{
1344}
1345
1346static inline void
1347mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1348{
1349}
1350
1351static inline void lock_page_memcg(struct page *page)
1352{
1353}
1354
1355static inline void unlock_page_memcg(struct page *page)
1356{
1357}
1358
1359static inline void mem_cgroup_handle_over_high(void)
1360{
1361}
1362
1363static inline void mem_cgroup_enter_user_fault(void)
1364{
1365}
1366
1367static inline void mem_cgroup_exit_user_fault(void)
1368{
1369}
1370
1371static inline bool task_in_memcg_oom(struct task_struct *p)
1372{
1373        return false;
1374}
1375
1376static inline bool mem_cgroup_oom_synchronize(bool wait)
1377{
1378        return false;
1379}
1380
1381static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1382        struct task_struct *victim, struct mem_cgroup *oom_domain)
1383{
1384        return NULL;
1385}
1386
1387static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1388{
1389}
1390
1391static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1392                                     int idx,
1393                                     int nr)
1394{
1395}
1396
1397static inline void mod_memcg_state(struct mem_cgroup *memcg,
1398                                   int idx,
1399                                   int nr)
1400{
1401}
1402
1403static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1404{
1405        return 0;
1406}
1407
1408static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1409                                              enum node_stat_item idx)
1410{
1411        return node_page_state(lruvec_pgdat(lruvec), idx);
1412}
1413
1414static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1415                                                    enum node_stat_item idx)
1416{
1417        return node_page_state(lruvec_pgdat(lruvec), idx);
1418}
1419
1420static inline void mem_cgroup_flush_stats(void)
1421{
1422}
1423
1424static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1425                                            enum node_stat_item idx, int val)
1426{
1427}
1428
1429static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1430                                           int val)
1431{
1432        struct page *page = virt_to_head_page(p);
1433
1434        __mod_node_page_state(page_pgdat(page), idx, val);
1435}
1436
1437static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1438                                         int val)
1439{
1440        struct page *page = virt_to_head_page(p);
1441
1442        mod_node_page_state(page_pgdat(page), idx, val);
1443}
1444
1445static inline void count_memcg_events(struct mem_cgroup *memcg,
1446                                      enum vm_event_item idx,
1447                                      unsigned long count)
1448{
1449}
1450
1451static inline void __count_memcg_events(struct mem_cgroup *memcg,
1452                                        enum vm_event_item idx,
1453                                        unsigned long count)
1454{
1455}
1456
1457static inline void count_memcg_page_event(struct page *page,
1458                                          int idx)
1459{
1460}
1461
1462static inline
1463void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1464{
1465}
1466
1467static inline void split_page_memcg(struct page *head, unsigned int nr)
1468{
1469}
1470
1471static inline
1472unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1473                                            gfp_t gfp_mask,
1474                                            unsigned long *total_scanned)
1475{
1476        return 0;
1477}
1478#endif /* CONFIG_MEMCG */
1479
1480static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1481{
1482        __mod_lruvec_kmem_state(p, idx, 1);
1483}
1484
1485static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1486{
1487        __mod_lruvec_kmem_state(p, idx, -1);
1488}
1489
1490static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1491{
1492        struct mem_cgroup *memcg;
1493
1494        memcg = lruvec_memcg(lruvec);
1495        if (!memcg)
1496                return NULL;
1497        memcg = parent_mem_cgroup(memcg);
1498        if (!memcg)
1499                return NULL;
1500        return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1501}
1502
1503static inline void unlock_page_lruvec(struct lruvec *lruvec)
1504{
1505        spin_unlock(&lruvec->lru_lock);
1506}
1507
1508static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1509{
1510        spin_unlock_irq(&lruvec->lru_lock);
1511}
1512
1513static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1514                unsigned long flags)
1515{
1516        spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1517}
1518
1519/* Test requires a stable page->memcg binding, see page_memcg() */
1520static inline bool page_matches_lruvec(struct page *page, struct lruvec *lruvec)
1521{
1522        return lruvec_pgdat(lruvec) == page_pgdat(page) &&
1523               lruvec_memcg(lruvec) == page_memcg(page);
1524}
1525
1526/* Don't lock again iff page's lruvec locked */
1527static inline struct lruvec *relock_page_lruvec_irq(struct page *page,
1528                struct lruvec *locked_lruvec)
1529{
1530        if (locked_lruvec) {
1531                if (page_matches_lruvec(page, locked_lruvec))
1532                        return locked_lruvec;
1533
1534                unlock_page_lruvec_irq(locked_lruvec);
1535        }
1536
1537        return lock_page_lruvec_irq(page);
1538}
1539
1540/* Don't lock again iff page's lruvec locked */
1541static inline struct lruvec *relock_page_lruvec_irqsave(struct page *page,
1542                struct lruvec *locked_lruvec, unsigned long *flags)
1543{
1544        if (locked_lruvec) {
1545                if (page_matches_lruvec(page, locked_lruvec))
1546                        return locked_lruvec;
1547
1548                unlock_page_lruvec_irqrestore(locked_lruvec, *flags);
1549        }
1550
1551        return lock_page_lruvec_irqsave(page, flags);
1552}
1553
1554#ifdef CONFIG_CGROUP_WRITEBACK
1555
1556struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1557void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1558                         unsigned long *pheadroom, unsigned long *pdirty,
1559                         unsigned long *pwriteback);
1560
1561void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1562                                             struct bdi_writeback *wb);
1563
1564static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1565                                                  struct bdi_writeback *wb)
1566{
1567        if (mem_cgroup_disabled())
1568                return;
1569
1570        if (unlikely(&page_memcg(page)->css != wb->memcg_css))
1571                mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1572}
1573
1574void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1575
1576#else   /* CONFIG_CGROUP_WRITEBACK */
1577
1578static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1579{
1580        return NULL;
1581}
1582
1583static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1584                                       unsigned long *pfilepages,
1585                                       unsigned long *pheadroom,
1586                                       unsigned long *pdirty,
1587                                       unsigned long *pwriteback)
1588{
1589}
1590
1591static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1592                                                  struct bdi_writeback *wb)
1593{
1594}
1595
1596static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1597{
1598}
1599
1600#endif  /* CONFIG_CGROUP_WRITEBACK */
1601
1602struct sock;
1603bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages,
1604                             gfp_t gfp_mask);
1605void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1606#ifdef CONFIG_MEMCG
1607extern struct static_key_false memcg_sockets_enabled_key;
1608#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1609void mem_cgroup_sk_alloc(struct sock *sk);
1610void mem_cgroup_sk_free(struct sock *sk);
1611static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1612{
1613        if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1614                return true;
1615        do {
1616                if (time_before(jiffies, memcg->socket_pressure))
1617                        return true;
1618        } while ((memcg = parent_mem_cgroup(memcg)));
1619        return false;
1620}
1621
1622int alloc_shrinker_info(struct mem_cgroup *memcg);
1623void free_shrinker_info(struct mem_cgroup *memcg);
1624void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id);
1625void reparent_shrinker_deferred(struct mem_cgroup *memcg);
1626#else
1627#define mem_cgroup_sockets_enabled 0
1628static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1629static inline void mem_cgroup_sk_free(struct sock *sk) { };
1630static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1631{
1632        return false;
1633}
1634
1635static inline void set_shrinker_bit(struct mem_cgroup *memcg,
1636                                    int nid, int shrinker_id)
1637{
1638}
1639#endif
1640
1641#ifdef CONFIG_MEMCG_KMEM
1642bool mem_cgroup_kmem_disabled(void);
1643int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1644void __memcg_kmem_uncharge_page(struct page *page, int order);
1645
1646struct obj_cgroup *get_obj_cgroup_from_current(void);
1647
1648int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1649void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1650
1651extern struct static_key_false memcg_kmem_enabled_key;
1652
1653extern int memcg_nr_cache_ids;
1654void memcg_get_cache_ids(void);
1655void memcg_put_cache_ids(void);
1656
1657/*
1658 * Helper macro to loop through all memcg-specific caches. Callers must still
1659 * check if the cache is valid (it is either valid or NULL).
1660 * the slab_mutex must be held when looping through those caches
1661 */
1662#define for_each_memcg_cache_index(_idx)        \
1663        for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1664
1665static inline bool memcg_kmem_enabled(void)
1666{
1667        return static_branch_likely(&memcg_kmem_enabled_key);
1668}
1669
1670static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1671                                         int order)
1672{
1673        if (memcg_kmem_enabled())
1674                return __memcg_kmem_charge_page(page, gfp, order);
1675        return 0;
1676}
1677
1678static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1679{
1680        if (memcg_kmem_enabled())
1681                __memcg_kmem_uncharge_page(page, order);
1682}
1683
1684/*
1685 * A helper for accessing memcg's kmem_id, used for getting
1686 * corresponding LRU lists.
1687 */
1688static inline int memcg_cache_id(struct mem_cgroup *memcg)
1689{
1690        return memcg ? memcg->kmemcg_id : -1;
1691}
1692
1693struct mem_cgroup *mem_cgroup_from_obj(void *p);
1694
1695#else
1696static inline bool mem_cgroup_kmem_disabled(void)
1697{
1698        return true;
1699}
1700
1701static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1702                                         int order)
1703{
1704        return 0;
1705}
1706
1707static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1708{
1709}
1710
1711static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1712                                           int order)
1713{
1714        return 0;
1715}
1716
1717static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1718{
1719}
1720
1721#define for_each_memcg_cache_index(_idx)        \
1722        for (; NULL; )
1723
1724static inline bool memcg_kmem_enabled(void)
1725{
1726        return false;
1727}
1728
1729static inline int memcg_cache_id(struct mem_cgroup *memcg)
1730{
1731        return -1;
1732}
1733
1734static inline void memcg_get_cache_ids(void)
1735{
1736}
1737
1738static inline void memcg_put_cache_ids(void)
1739{
1740}
1741
1742static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1743{
1744       return NULL;
1745}
1746
1747#endif /* CONFIG_MEMCG_KMEM */
1748
1749#endif /* _LINUX_MEMCONTROL_H */
1750