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