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