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
  27 struct mem_cgroup;
  28 struct page_cgroup;
  29 struct page;
  30 struct mm_struct;
  31 struct kmem_cache;
  32
  33 /*
  34  * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
  35  * These two lists should keep in accord with each other.
  36  */
  37 enum mem_cgroup_stat_index {
  38         /*
  39          * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
  40          */
  41         MEM_CGROUP_STAT_CACHE,          /* # of pages charged as cache */
  42         MEM_CGROUP_STAT_RSS,            /* # of pages charged as anon rss */
  43         MEM_CGROUP_STAT_RSS_HUGE,       /* # of pages charged as anon huge */
  44         MEM_CGROUP_STAT_FILE_MAPPED,    /* # of pages charged as file rss */
  45         MEM_CGROUP_STAT_WRITEBACK,      /* # of pages under writeback */
  46         MEM_CGROUP_STAT_SWAP,           /* # of pages, swapped out */
  47         MEM_CGROUP_STAT_NSTATS,
  48 };
  49
  50 struct mem_cgroup_reclaim_cookie {
  51         struct zone *zone;
  52         int priority;
  53         unsigned int generation;
  54 };
  55
  56 #ifdef CONFIG_MEMCG
  57 /*
  58  * All "charge" functions with gfp_mask should use GFP_KERNEL or
  59  * (gfp_mask & GFP_RECLAIM_MASK). In current implementatin, memcg doesn't
  60  * alloc memory but reclaims memory from all available zones. So, "where I want
  61  * memory from" bits of gfp_mask has no meaning. So any bits of that field is
  62  * available but adding a rule is better. charge functions' gfp_mask should
  63  * be set to GFP_KERNEL or gfp_mask & GFP_RECLAIM_MASK for avoiding ambiguous
  64  * codes.
  65  * (Of course, if memcg does memory allocation in future, GFP_KERNEL is sane.)
  66  */
  67
  68 extern int mem_cgroup_charge_anon(struct page *page, struct mm_struct *mm,
  69                                 gfp_t gfp_mask);
  70 /* for swap handling */
  71 extern int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
  72                 struct page *page, gfp_t mask, struct mem_cgroup **memcgp);
  73 extern void mem_cgroup_commit_charge_swapin(struct page *page,
  74                                         struct mem_cgroup *memcg);
  75 extern void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg);
  76
  77 extern int mem_cgroup_charge_file(struct page *page, struct mm_struct *mm,
  78                                         gfp_t gfp_mask);
  79
  80 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
  81 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
  82
  83 /* For coalescing uncharge for reducing memcg' overhead*/
  84 extern void mem_cgroup_uncharge_start(void);
  85 extern void mem_cgroup_uncharge_end(void);
  86
  87 extern void mem_cgroup_uncharge_page(struct page *page);
  88 extern void mem_cgroup_uncharge_cache_page(struct page *page);
  89
  90 bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
  91                                   struct mem_cgroup *memcg);
  92 bool task_in_mem_cgroup(struct task_struct *task,
  93                         const struct mem_cgroup *memcg);
  94
  95 extern struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page);
  96 extern struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
  97
  98 extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
  99 extern struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css);
 100
 101 static inline
 102 bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg)
 103 {
 104         struct mem_cgroup *task_memcg;
 105         bool match;
 106
 107         rcu_read_lock();
 108         task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
 109         match = __mem_cgroup_same_or_subtree(memcg, task_memcg);
 110         rcu_read_unlock();
 111         return match;
 112 }
 113
 114 extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg);
 115
 116 extern void
 117 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
 118                              struct mem_cgroup **memcgp);
 119 extern void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 120         struct page *oldpage, struct page *newpage, bool migration_ok);
 121
 122 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
 123                                    struct mem_cgroup *,
 124                                    struct mem_cgroup_reclaim_cookie *);
 125 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 126
 127 /*
 128  * For memory reclaim.
 129  */
 130 int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec);
 131 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
 132 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list);
 133 void mem_cgroup_update_lru_size(struct lruvec *, enum lru_list, int);
 134 extern void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
 135                                         struct task_struct *p);
 136 extern void mem_cgroup_replace_page_cache(struct page *oldpage,
 137                                         struct page *newpage);
 138
 139 static inline void mem_cgroup_oom_enable(void)
 140 {
 141         WARN_ON(current->memcg_oom.may_oom);
 142         current->memcg_oom.may_oom = 1;
 143 }
 144
 145 static inline void mem_cgroup_oom_disable(void)
 146 {
 147         WARN_ON(!current->memcg_oom.may_oom);
 148         current->memcg_oom.may_oom = 0;
 149 }
 150
 151 static inline bool task_in_memcg_oom(struct task_struct *p)
 152 {
 153         return p->memcg_oom.memcg;
 154 }
 155
 156 bool mem_cgroup_oom_synchronize(bool wait);
 157
 158 #ifdef CONFIG_MEMCG_SWAP
 159 extern int do_swap_account;
 160 #endif
 161
 162 static inline bool mem_cgroup_disabled(void)
 163 {
 164         if (memory_cgrp_subsys.disabled)
 165                 return true;
 166         return false;
 167 }
 168
 169 void __mem_cgroup_begin_update_page_stat(struct page *page, bool *locked,
 170                                          unsigned long *flags);
 171
 172 extern atomic_t memcg_moving;
 173
 174 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
 175                                         bool *locked, unsigned long *flags)
 176 {
 177         if (mem_cgroup_disabled())
 178                 return;
 179         rcu_read_lock();
 180         *locked = false;
 181         if (atomic_read(&memcg_moving))
 182                 __mem_cgroup_begin_update_page_stat(page, locked, flags);
 183 }
 184
 185 void __mem_cgroup_end_update_page_stat(struct page *page,
 186                                 unsigned long *flags);
 187 static inline void mem_cgroup_end_update_page_stat(struct page *page,
 188                                         bool *locked, unsigned long *flags)
 189 {
 190         if (mem_cgroup_disabled())
 191                 return;
 192         if (*locked)
 193                 __mem_cgroup_end_update_page_stat(page, flags);
 194         rcu_read_unlock();
 195 }
 196
 197 void mem_cgroup_update_page_stat(struct page *page,
 198                                  enum mem_cgroup_stat_index idx,
 199                                  int val);
 200
 201 static inline void mem_cgroup_inc_page_stat(struct page *page,
 202                                             enum mem_cgroup_stat_index idx)
 203 {
 204         mem_cgroup_update_page_stat(page, idx, 1);
 205 }
 206
 207 static inline void mem_cgroup_dec_page_stat(struct page *page,
 208                                             enum mem_cgroup_stat_index idx)
 209 {
 210         mem_cgroup_update_page_stat(page, idx, -1);
 211 }
 212
 213 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 214                                                 gfp_t gfp_mask,
 215                                                 unsigned long *total_scanned);
 216
 217 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
 218 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
 219                                              enum vm_event_item idx)
 220 {
 221         if (mem_cgroup_disabled())
 222                 return;
 223         __mem_cgroup_count_vm_event(mm, idx);
 224 }
 225 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 226 void mem_cgroup_split_huge_fixup(struct page *head);
 227 #endif
 228
 229 #ifdef CONFIG_DEBUG_VM
 230 bool mem_cgroup_bad_page_check(struct page *page);
 231 void mem_cgroup_print_bad_page(struct page *page);
 232 #endif
 233 #else /* CONFIG_MEMCG */
 234 struct mem_cgroup;
 235
 236 static inline int mem_cgroup_charge_anon(struct page *page,
 237                                         struct mm_struct *mm, gfp_t gfp_mask)
 238 {
 239         return 0;
 240 }
 241
 242 static inline int mem_cgroup_charge_file(struct page *page,
 243                                         struct mm_struct *mm, gfp_t gfp_mask)
 244 {
 245         return 0;
 246 }
 247
 248 static inline int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
 249                 struct page *page, gfp_t gfp_mask, struct mem_cgroup **memcgp)
 250 {
 251         return 0;
 252 }
 253
 254 static inline void mem_cgroup_commit_charge_swapin(struct page *page,
 255                                           struct mem_cgroup *memcg)
 256 {
 257 }
 258
 259 static inline void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
 260 {
 261 }
 262
 263 static inline void mem_cgroup_uncharge_start(void)
 264 {
 265 }
 266
 267 static inline void mem_cgroup_uncharge_end(void)
 268 {
 269 }
 270
 271 static inline void mem_cgroup_uncharge_page(struct page *page)
 272 {
 273 }
 274
 275 static inline void mem_cgroup_uncharge_cache_page(struct page *page)
 276 {
 277 }
 278
 279 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
 280                                                     struct mem_cgroup *memcg)
 281 {
 282         return &zone->lruvec;
 283 }
 284
 285 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
 286                                                     struct zone *zone)
 287 {
 288         return &zone->lruvec;
 289 }
 290
 291 static inline struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
 292 {
 293         return NULL;
 294 }
 295
 296 static inline bool mm_match_cgroup(struct mm_struct *mm,
 297                 struct mem_cgroup *memcg)
 298 {
 299         return true;
 300 }
 301
 302 static inline bool task_in_mem_cgroup(struct task_struct *task,
 303                                       const struct mem_cgroup *memcg)
 304 {
 305         return true;
 306 }
 307
 308 static inline struct cgroup_subsys_state
 309                 *mem_cgroup_css(struct mem_cgroup *memcg)
 310 {
 311         return NULL;
 312 }
 313
 314 static inline void
 315 mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
 316                              struct mem_cgroup **memcgp)
 317 {
 318 }
 319
 320 static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 321                 struct page *oldpage, struct page *newpage, bool migration_ok)
 322 {
 323 }
 324
 325 static inline struct mem_cgroup *
 326 mem_cgroup_iter(struct mem_cgroup *root,
 327                 struct mem_cgroup *prev,
 328                 struct mem_cgroup_reclaim_cookie *reclaim)
 329 {
 330         return NULL;
 331 }
 332
 333 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
 334                                          struct mem_cgroup *prev)
 335 {
 336 }
 337
 338 static inline bool mem_cgroup_disabled(void)
 339 {
 340         return true;
 341 }
 342
 343 static inline int
 344 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
 345 {
 346         return 1;
 347 }
 348
 349 static inline unsigned long
 350 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
 351 {
 352         return 0;
 353 }
 354
 355 static inline void
 356 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
 357                               int increment)
 358 {
 359 }
 360
 361 static inline void
 362 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
 363 {
 364 }
 365
 366 static inline void mem_cgroup_begin_update_page_stat(struct page *page,
 367                                         bool *locked, unsigned long *flags)
 368 {
 369 }
 370
 371 static inline void mem_cgroup_end_update_page_stat(struct page *page,
 372                                         bool *locked, unsigned long *flags)
 373 {
 374 }
 375
 376 static inline void mem_cgroup_oom_enable(void)
 377 {
 378 }
 379
 380 static inline void mem_cgroup_oom_disable(void)
 381 {
 382 }
 383
 384 static inline bool task_in_memcg_oom(struct task_struct *p)
 385 {
 386         return false;
 387 }
 388
 389 static inline bool mem_cgroup_oom_synchronize(bool wait)
 390 {
 391         return false;
 392 }
 393
 394 static inline void mem_cgroup_inc_page_stat(struct page *page,
 395                                             enum mem_cgroup_stat_index idx)
 396 {
 397 }
 398
 399 static inline void mem_cgroup_dec_page_stat(struct page *page,
 400                                             enum mem_cgroup_stat_index idx)
 401 {
 402 }
 403
 404 static inline
 405 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
 406                                             gfp_t gfp_mask,
 407                                             unsigned long *total_scanned)
 408 {
 409         return 0;
 410 }
 411
 412 static inline void mem_cgroup_split_huge_fixup(struct page *head)
 413 {
 414 }
 415
 416 static inline
 417 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
 418 {
 419 }
 420 static inline void mem_cgroup_replace_page_cache(struct page *oldpage,
 421                                 struct page *newpage)
 422 {
 423 }
 424 #endif /* CONFIG_MEMCG */
 425
 426 #if !defined(CONFIG_MEMCG) || !defined(CONFIG_DEBUG_VM)
 427 static inline bool
 428 mem_cgroup_bad_page_check(struct page *page)
 429 {
 430         return false;
 431 }
 432
 433 static inline void
 434 mem_cgroup_print_bad_page(struct page *page)
 435 {
 436 }
 437 #endif
 438
 439 enum {
 440         UNDER_LIMIT,
 441         SOFT_LIMIT,
 442         OVER_LIMIT,
 443 };
 444
 445 struct sock;
 446 #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
 447 void sock_update_memcg(struct sock *sk);
 448 void sock_release_memcg(struct sock *sk);
 449 #else
 450 static inline void sock_update_memcg(struct sock *sk)
 451 {
 452 }
 453 static inline void sock_release_memcg(struct sock *sk)
 454 {
 455 }
 456 #endif /* CONFIG_INET && CONFIG_MEMCG_KMEM */
 457
 458 #ifdef CONFIG_MEMCG_KMEM
 459 extern struct static_key memcg_kmem_enabled_key;
 460
 461 extern int memcg_limited_groups_array_size;
 462
 463 /*
 464  * Helper macro to loop through all memcg-specific caches. Callers must still
 465  * check if the cache is valid (it is either valid or NULL).
 466  * the slab_mutex must be held when looping through those caches
 467  */
 468 #define for_each_memcg_cache_index(_idx)        \
 469         for ((_idx) = 0; (_idx) < memcg_limited_groups_array_size; (_idx)++)
 470
 471 static inline bool memcg_kmem_enabled(void)
 472 {
 473         return static_key_false(&memcg_kmem_enabled_key);
 474 }
 475
 476 /*
 477  * In general, we'll do everything in our power to not incur in any overhead
 478  * for non-memcg users for the kmem functions. Not even a function call, if we
 479  * can avoid it.
 480  *
 481  * Therefore, we'll inline all those functions so that in the best case, we'll
 482  * see that kmemcg is off for everybody and proceed quickly.  If it is on,
 483  * we'll still do most of the flag checking inline. We check a lot of
 484  * conditions, but because they are pretty simple, they are expected to be
 485  * fast.
 486  */
 487 bool __memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg,
 488                                         int order);
 489 void __memcg_kmem_commit_charge(struct page *page,
 490                                        struct mem_cgroup *memcg, int order);
 491 void __memcg_kmem_uncharge_pages(struct page *page, int order);
 492
 493 int memcg_cache_id(struct mem_cgroup *memcg);
 494
 495 int memcg_alloc_cache_params(struct mem_cgroup *memcg, struct kmem_cache *s,
 496                              struct kmem_cache *root_cache);
 497 void memcg_free_cache_params(struct kmem_cache *s);
 498
 499 int memcg_update_cache_size(struct kmem_cache *s, int num_groups);
 500 void memcg_update_array_size(int num_groups);
 501
 502 struct kmem_cache *
 503 __memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
 504
 505 int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order);
 506 void __memcg_uncharge_slab(struct kmem_cache *cachep, int order);
 507
 508 int __memcg_cleanup_cache_params(struct kmem_cache *s);
 509
 510 /**
 511  * memcg_kmem_newpage_charge: verify if a new kmem allocation is allowed.
 512  * @gfp: the gfp allocation flags.
 513  * @memcg: a pointer to the memcg this was charged against.
 514  * @order: allocation order.
 515  *
 516  * returns true if the memcg where the current task belongs can hold this
 517  * allocation.
 518  *
 519  * We return true automatically if this allocation is not to be accounted to
 520  * any memcg.
 521  */
 522 static inline bool
 523 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
 524 {
 525         if (!memcg_kmem_enabled())
 526                 return true;
 527
 528         /*
 529          * __GFP_NOFAIL allocations will move on even if charging is not
 530          * possible. Therefore we don't even try, and have this allocation
 531          * unaccounted. We could in theory charge it with
 532          * res_counter_charge_nofail, but we hope those allocations are rare,
 533          * and won't be worth the trouble.
 534          */
 535         if (gfp & __GFP_NOFAIL)
 536                 return true;
 537         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 538                 return true;
 539
 540         /* If the test is dying, just let it go. */
 541         if (unlikely(fatal_signal_pending(current)))
 542                 return true;
 543
 544         return __memcg_kmem_newpage_charge(gfp, memcg, order);
 545 }
 546
 547 /**
 548  * memcg_kmem_uncharge_pages: uncharge pages from memcg
 549  * @page: pointer to struct page being freed
 550  * @order: allocation order.
 551  *
 552  * there is no need to specify memcg here, since it is embedded in page_cgroup
 553  */
 554 static inline void
 555 memcg_kmem_uncharge_pages(struct page *page, int order)
 556 {
 557         if (memcg_kmem_enabled())
 558                 __memcg_kmem_uncharge_pages(page, order);
 559 }
 560
 561 /**
 562  * memcg_kmem_commit_charge: embeds correct memcg in a page
 563  * @page: pointer to struct page recently allocated
 564  * @memcg: the memcg structure we charged against
 565  * @order: allocation order.
 566  *
 567  * Needs to be called after memcg_kmem_newpage_charge, regardless of success or
 568  * failure of the allocation. if @page is NULL, this function will revert the
 569  * charges. Otherwise, it will commit the memcg given by @memcg to the
 570  * corresponding page_cgroup.
 571  */
 572 static inline void
 573 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
 574 {
 575         if (memcg_kmem_enabled() && memcg)
 576                 __memcg_kmem_commit_charge(page, memcg, order);
 577 }
 578
 579 /**
 580  * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
 581  * @cachep: the original global kmem cache
 582  * @gfp: allocation flags.
 583  *
 584  * All memory allocated from a per-memcg cache is charged to the owner memcg.
 585  */
 586 static __always_inline struct kmem_cache *
 587 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 588 {
 589         if (!memcg_kmem_enabled())
 590                 return cachep;
 591         if (gfp & __GFP_NOFAIL)
 592                 return cachep;
 593         if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
 594                 return cachep;
 595         if (unlikely(fatal_signal_pending(current)))
 596                 return cachep;
 597
 598         return __memcg_kmem_get_cache(cachep, gfp);
 599 }
 600 #else
 601 #define for_each_memcg_cache_index(_idx)        \
 602         for (; NULL; )
 603
 604 static inline bool memcg_kmem_enabled(void)
 605 {
 606         return false;
 607 }
 608
 609 static inline bool
 610 memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **memcg, int order)
 611 {
 612         return true;
 613 }
 614
 615 static inline void memcg_kmem_uncharge_pages(struct page *page, int order)
 616 {
 617 }
 618
 619 static inline void
 620 memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg, int order)
 621 {
 622 }
 623
 624 static inline int memcg_cache_id(struct mem_cgroup *memcg)
 625 {
 626         return -1;
 627 }
 628
 629 static inline int memcg_alloc_cache_params(struct mem_cgroup *memcg,
 630                 struct kmem_cache *s, struct kmem_cache *root_cache)
 631 {
 632         return 0;
 633 }
 634
 635 static inline void memcg_free_cache_params(struct kmem_cache *s)
 636 {
 637 }
 638
 639 static inline struct kmem_cache *
 640 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
 641 {
 642         return cachep;
 643 }
 644 #endif /* CONFIG_MEMCG_KMEM */
 645 #endif /* _LINUX_MEMCONTROL_H */
 646