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/kernel.h> 18#include <linux/page_counter.h> 19#include <linux/vmpressure.h> 20#include <linux/eventfd.h> 21#include <linux/mm.h> 22#include <linux/vmstat.h> 23#include <linux/writeback.h> 24#include <linux/page-flags.h> 25#include <linux/shrinker.h> 26 27struct mem_cgroup; 28struct obj_cgroup; 29struct page; 30struct mm_struct; 31struct kmem_cache; 32 33/* Cgroup-specific page state, on top of universal node page state */ 34enum memcg_stat_item { 35 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS, 36 MEMCG_SOCK, 37 MEMCG_PERCPU_B, 38 MEMCG_VMALLOC, 39 MEMCG_KMEM, 40 MEMCG_ZSWAP_B, 41 MEMCG_ZSWAPPED, 42 MEMCG_NR_STAT, 43}; 44 45enum memcg_memory_event { 46 MEMCG_LOW, 47 MEMCG_HIGH, 48 MEMCG_MAX, 49 MEMCG_OOM, 50 MEMCG_OOM_KILL, 51 MEMCG_OOM_GROUP_KILL, 52 MEMCG_SWAP_HIGH, 53 MEMCG_SWAP_MAX, 54 MEMCG_SWAP_FAIL, 55 MEMCG_NR_MEMORY_EVENTS, 56}; 57 58struct mem_cgroup_reclaim_cookie { 59 pg_data_t *pgdat; 60 unsigned int generation; 61}; 62 63#ifdef CONFIG_MEMCG 64 65#define MEM_CGROUP_ID_SHIFT 16 66 67struct mem_cgroup_id { 68 int id; 69 refcount_t ref; 70}; 71 72/* 73 * Per memcg event counter is incremented at every pagein/pageout. With THP, 74 * it will be incremented by the number of pages. This counter is used 75 * to trigger some periodic events. This is straightforward and better 76 * than using jiffies etc. to handle periodic memcg event. 77 */ 78enum mem_cgroup_events_target { 79 MEM_CGROUP_TARGET_THRESH, 80 MEM_CGROUP_TARGET_SOFTLIMIT, 81 MEM_CGROUP_NTARGETS, 82}; 83 84struct memcg_vmstats_percpu; 85struct memcg_vmstats; 86 87struct mem_cgroup_reclaim_iter { 88 struct mem_cgroup *position; 89 /* scan generation, increased every round-trip */ 90 unsigned int generation; 91}; 92 93struct lruvec_stats_percpu { 94 /* Local (CPU and cgroup) state */ 95 long state[NR_VM_NODE_STAT_ITEMS]; 96 97 /* Delta calculation for lockless upward propagation */ 98 long state_prev[NR_VM_NODE_STAT_ITEMS]; 99}; 100 101struct lruvec_stats { 102 /* Aggregated (CPU and subtree) state */ 103 long state[NR_VM_NODE_STAT_ITEMS]; 104 105 /* Non-hierarchical (CPU aggregated) state */ 106 long state_local[NR_VM_NODE_STAT_ITEMS]; 107 108 /* Pending child counts during tree propagation */ 109 long state_pending[NR_VM_NODE_STAT_ITEMS]; 110}; 111 112/* 113 * per-node information in memory controller. 114 */ 115struct mem_cgroup_per_node { 116 struct lruvec lruvec; 117 118 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu; 119 struct lruvec_stats lruvec_stats; 120 121 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; 122 123 struct mem_cgroup_reclaim_iter iter; 124 125 struct shrinker_info __rcu *shrinker_info; 126 127 struct rb_node tree_node; /* RB tree node */ 128 unsigned long usage_in_excess;/* Set to the value by which */ 129 /* the soft limit is exceeded*/ 130 bool on_tree; 131 struct mem_cgroup *memcg; /* Back pointer, we cannot */ 132 /* use container_of */ 133}; 134 135struct mem_cgroup_threshold { 136 struct eventfd_ctx *eventfd; 137 unsigned long threshold; 138}; 139 140/* For threshold */ 141struct mem_cgroup_threshold_ary { 142 /* An array index points to threshold just below or equal to usage. */ 143 int current_threshold; 144 /* Size of entries[] */ 145 unsigned int size; 146 /* Array of thresholds */ 147 struct mem_cgroup_threshold entries[] __counted_by(size); 148}; 149 150struct mem_cgroup_thresholds { 151 /* Primary thresholds array */ 152 struct mem_cgroup_threshold_ary *primary; 153 /* 154 * Spare threshold array. 155 * This is needed to make mem_cgroup_unregister_event() "never fail". 156 * It must be able to store at least primary->size - 1 entries. 157 */ 158 struct mem_cgroup_threshold_ary *spare; 159}; 160 161/* 162 * Remember four most recent foreign writebacks with dirty pages in this 163 * cgroup. Inode sharing is expected to be uncommon and, even if we miss 164 * one in a given round, we're likely to catch it later if it keeps 165 * foreign-dirtying, so a fairly low count should be enough. 166 * 167 * See mem_cgroup_track_foreign_dirty_slowpath() for details. 168 */ 169#define MEMCG_CGWB_FRN_CNT 4 170 171struct memcg_cgwb_frn { 172 u64 bdi_id; /* bdi->id of the foreign inode */ 173 int memcg_id; /* memcg->css.id of foreign inode */ 174 u64 at; /* jiffies_64 at the time of dirtying */ 175 struct wb_completion done; /* tracks in-flight foreign writebacks */ 176}; 177 178/* 179 * Bucket for arbitrarily byte-sized objects charged to a memory 180 * cgroup. The bucket can be reparented in one piece when the cgroup 181 * is destroyed, without having to round up the individual references 182 * of all live memory objects in the wild. 183 */ 184struct obj_cgroup { 185 struct percpu_ref refcnt; 186 struct mem_cgroup *memcg; 187 atomic_t nr_charged_bytes; 188 union { 189 struct list_head list; /* protected by objcg_lock */ 190 struct rcu_head rcu; 191 }; 192}; 193 194/* 195 * The memory controller data structure. The memory controller controls both 196 * page cache and RSS per cgroup. We would eventually like to provide 197 * statistics based on the statistics developed by Rik Van Riel for clock-pro, 198 * to help the administrator determine what knobs to tune. 199 */ 200struct mem_cgroup { 201 struct cgroup_subsys_state css; 202 203 /* Private memcg ID. Used to ID objects that outlive the cgroup */ 204 struct mem_cgroup_id id; 205 206 /* Accounted resources */ 207 struct page_counter memory; /* Both v1 & v2 */ 208 209 union { 210 struct page_counter swap; /* v2 only */ 211 struct page_counter memsw; /* v1 only */ 212 }; 213 214 /* Legacy consumer-oriented counters */ 215 struct page_counter kmem; /* v1 only */ 216 struct page_counter tcpmem; /* v1 only */ 217 218 /* Range enforcement for interrupt charges */ 219 struct work_struct high_work; 220 221#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP) 222 unsigned long zswap_max; 223 224 /* 225 * Prevent pages from this memcg from being written back from zswap to 226 * swap, and from being swapped out on zswap store failures. 227 */ 228 bool zswap_writeback; 229#endif 230 231 unsigned long soft_limit; 232 233 /* vmpressure notifications */ 234 struct vmpressure vmpressure; 235 236 /* 237 * Should the OOM killer kill all belonging tasks, had it kill one? 238 */ 239 bool oom_group; 240 241 /* protected by memcg_oom_lock */ 242 bool oom_lock; 243 int under_oom; 244 245 int swappiness; 246 /* OOM-Killer disable */ 247 int oom_kill_disable; 248 249 /* memory.events and memory.events.local */ 250 struct cgroup_file events_file; 251 struct cgroup_file events_local_file; 252 253 /* handle for "memory.swap.events" */ 254 struct cgroup_file swap_events_file; 255 256 /* protect arrays of thresholds */ 257 struct mutex thresholds_lock; 258 259 /* thresholds for memory usage. RCU-protected */ 260 struct mem_cgroup_thresholds thresholds; 261 262 /* thresholds for mem+swap usage. RCU-protected */ 263 struct mem_cgroup_thresholds memsw_thresholds; 264 265 /* For oom notifier event fd */ 266 struct list_head oom_notify; 267 268 /* 269 * Should we move charges of a task when a task is moved into this 270 * mem_cgroup ? And what type of charges should we move ? 271 */ 272 unsigned long move_charge_at_immigrate; 273 /* taken only while moving_account > 0 */ 274 spinlock_t move_lock; 275 unsigned long move_lock_flags; 276 277 CACHELINE_PADDING(_pad1_); 278 279 /* memory.stat */ 280 struct memcg_vmstats *vmstats; 281 282 /* memory.events */ 283 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS]; 284 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS]; 285 286 /* 287 * Hint of reclaim pressure for socket memroy management. Note 288 * that this indicator should NOT be used in legacy cgroup mode 289 * where socket memory is accounted/charged separately. 290 */ 291 unsigned long socket_pressure; 292 293 /* Legacy tcp memory accounting */ 294 bool tcpmem_active; 295 int tcpmem_pressure; 296 297#ifdef CONFIG_MEMCG_KMEM 298 int kmemcg_id; 299 /* 300 * memcg->objcg is wiped out as a part of the objcg repaprenting 301 * process. memcg->orig_objcg preserves a pointer (and a reference) 302 * to the original objcg until the end of live of memcg. 303 */ 304 struct obj_cgroup __rcu *objcg; 305 struct obj_cgroup *orig_objcg; 306 /* list of inherited objcgs, protected by objcg_lock */ 307 struct list_head objcg_list; 308#endif 309 310 CACHELINE_PADDING(_pad2_); 311 312 /* 313 * set > 0 if pages under this cgroup are moving to other cgroup. 314 */ 315 atomic_t moving_account; 316 struct task_struct *move_lock_task; 317 318 struct memcg_vmstats_percpu __percpu *vmstats_percpu; 319 320#ifdef CONFIG_CGROUP_WRITEBACK 321 struct list_head cgwb_list; 322 struct wb_domain cgwb_domain; 323 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT]; 324#endif 325 326 /* List of events which userspace want to receive */ 327 struct list_head event_list; 328 spinlock_t event_list_lock; 329 330#ifdef CONFIG_TRANSPARENT_HUGEPAGE 331 struct deferred_split deferred_split_queue; 332#endif 333 334#ifdef CONFIG_LRU_GEN_WALKS_MMU 335 /* per-memcg mm_struct list */ 336 struct lru_gen_mm_list mm_list; 337#endif 338 339 struct mem_cgroup_per_node *nodeinfo[]; 340}; 341 342/* 343 * size of first charge trial. 344 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the 345 * workload. 346 */ 347#define MEMCG_CHARGE_BATCH 64U 348 349extern struct mem_cgroup *root_mem_cgroup; 350 351enum page_memcg_data_flags { 352 /* page->memcg_data is a pointer to an objcgs vector */ 353 MEMCG_DATA_OBJCGS = (1UL << 0), 354 /* page has been accounted as a non-slab kernel page */ 355 MEMCG_DATA_KMEM = (1UL << 1), 356 /* the next bit after the last actual flag */ 357 __NR_MEMCG_DATA_FLAGS = (1UL << 2), 358}; 359 360#define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1) 361 362static inline bool folio_memcg_kmem(struct folio *folio); 363 364/* 365 * After the initialization objcg->memcg is always pointing at 366 * a valid memcg, but can be atomically swapped to the parent memcg. 367 * 368 * The caller must ensure that the returned memcg won't be released: 369 * e.g. acquire the rcu_read_lock or css_set_lock. 370 */ 371static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg) 372{ 373 return READ_ONCE(objcg->memcg); 374} 375 376/* 377 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio 378 * @folio: Pointer to the folio. 379 * 380 * Returns a pointer to the memory cgroup associated with the folio, 381 * or NULL. This function assumes that the folio is known to have a 382 * proper memory cgroup pointer. It's not safe to call this function 383 * against some type of folios, e.g. slab folios or ex-slab folios or 384 * kmem folios. 385 */ 386static inline struct mem_cgroup *__folio_memcg(struct folio *folio) 387{ 388 unsigned long memcg_data = folio->memcg_data; 389 390 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 391 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio); 392 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio); 393 394 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 395} 396 397/* 398 * __folio_objcg - get the object cgroup associated with a kmem folio. 399 * @folio: Pointer to the folio. 400 * 401 * Returns a pointer to the object cgroup associated with the folio, 402 * or NULL. This function assumes that the folio is known to have a 403 * proper object cgroup pointer. It's not safe to call this function 404 * against some type of folios, e.g. slab folios or ex-slab folios or 405 * LRU folios. 406 */ 407static inline struct obj_cgroup *__folio_objcg(struct folio *folio) 408{ 409 unsigned long memcg_data = folio->memcg_data; 410 411 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 412 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio); 413 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio); 414 415 return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 416} 417 418/* 419 * folio_memcg - Get the memory cgroup associated with a folio. 420 * @folio: Pointer to the folio. 421 * 422 * Returns a pointer to the memory cgroup associated with the folio, 423 * or NULL. This function assumes that the folio is known to have a 424 * proper memory cgroup pointer. It's not safe to call this function 425 * against some type of folios, e.g. slab folios or ex-slab folios. 426 * 427 * For a non-kmem folio any of the following ensures folio and memcg binding 428 * stability: 429 * 430 * - the folio lock 431 * - LRU isolation 432 * - folio_memcg_lock() 433 * - exclusive reference 434 * - mem_cgroup_trylock_pages() 435 * 436 * For a kmem folio a caller should hold an rcu read lock to protect memcg 437 * associated with a kmem folio from being released. 438 */ 439static inline struct mem_cgroup *folio_memcg(struct folio *folio) 440{ 441 if (folio_memcg_kmem(folio)) 442 return obj_cgroup_memcg(__folio_objcg(folio)); 443 return __folio_memcg(folio); 444} 445 446static inline struct mem_cgroup *page_memcg(struct page *page) 447{ 448 return folio_memcg(page_folio(page)); 449} 450 451/** 452 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio. 453 * @folio: Pointer to the folio. 454 * 455 * This function assumes that the folio is known to have a 456 * proper memory cgroup pointer. It's not safe to call this function 457 * against some type of folios, e.g. slab folios or ex-slab folios. 458 * 459 * Return: A pointer to the memory cgroup associated with the folio, 460 * or NULL. 461 */ 462static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 463{ 464 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 465 466 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 467 WARN_ON_ONCE(!rcu_read_lock_held()); 468 469 if (memcg_data & MEMCG_DATA_KMEM) { 470 struct obj_cgroup *objcg; 471 472 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 473 return obj_cgroup_memcg(objcg); 474 } 475 476 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 477} 478 479/* 480 * folio_memcg_check - Get the memory cgroup associated with a folio. 481 * @folio: Pointer to the folio. 482 * 483 * Returns a pointer to the memory cgroup associated with the folio, 484 * or NULL. This function unlike folio_memcg() can take any folio 485 * as an argument. It has to be used in cases when it's not known if a folio 486 * has an associated memory cgroup pointer or an object cgroups vector or 487 * an object cgroup. 488 * 489 * For a non-kmem folio any of the following ensures folio and memcg binding 490 * stability: 491 * 492 * - the folio lock 493 * - LRU isolation 494 * - lock_folio_memcg() 495 * - exclusive reference 496 * - mem_cgroup_trylock_pages() 497 * 498 * For a kmem folio a caller should hold an rcu read lock to protect memcg 499 * associated with a kmem folio from being released. 500 */ 501static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 502{ 503 /* 504 * Because folio->memcg_data might be changed asynchronously 505 * for slabs, READ_ONCE() should be used here. 506 */ 507 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 508 509 if (memcg_data & MEMCG_DATA_OBJCGS) 510 return NULL; 511 512 if (memcg_data & MEMCG_DATA_KMEM) { 513 struct obj_cgroup *objcg; 514 515 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 516 return obj_cgroup_memcg(objcg); 517 } 518 519 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 520} 521 522static inline struct mem_cgroup *page_memcg_check(struct page *page) 523{ 524 if (PageTail(page)) 525 return NULL; 526 return folio_memcg_check((struct folio *)page); 527} 528 529static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 530{ 531 struct mem_cgroup *memcg; 532 533 rcu_read_lock(); 534retry: 535 memcg = obj_cgroup_memcg(objcg); 536 if (unlikely(!css_tryget(&memcg->css))) 537 goto retry; 538 rcu_read_unlock(); 539 540 return memcg; 541} 542 543#ifdef CONFIG_MEMCG_KMEM 544/* 545 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set. 546 * @folio: Pointer to the folio. 547 * 548 * Checks if the folio has MemcgKmem flag set. The caller must ensure 549 * that the folio has an associated memory cgroup. It's not safe to call 550 * this function against some types of folios, e.g. slab folios. 551 */ 552static inline bool folio_memcg_kmem(struct folio *folio) 553{ 554 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page); 555 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJCGS, folio); 556 return folio->memcg_data & MEMCG_DATA_KMEM; 557} 558 559 560#else 561static inline bool folio_memcg_kmem(struct folio *folio) 562{ 563 return false; 564} 565 566#endif 567 568static inline bool PageMemcgKmem(struct page *page) 569{ 570 return folio_memcg_kmem(page_folio(page)); 571} 572 573static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 574{ 575 return (memcg == root_mem_cgroup); 576} 577 578static inline bool mem_cgroup_disabled(void) 579{ 580 return !cgroup_subsys_enabled(memory_cgrp_subsys); 581} 582 583static inline void mem_cgroup_protection(struct mem_cgroup *root, 584 struct mem_cgroup *memcg, 585 unsigned long *min, 586 unsigned long *low) 587{ 588 *min = *low = 0; 589 590 if (mem_cgroup_disabled()) 591 return; 592 593 /* 594 * There is no reclaim protection applied to a targeted reclaim. 595 * We are special casing this specific case here because 596 * mem_cgroup_calculate_protection is not robust enough to keep 597 * the protection invariant for calculated effective values for 598 * parallel reclaimers with different reclaim target. This is 599 * especially a problem for tail memcgs (as they have pages on LRU) 600 * which would want to have effective values 0 for targeted reclaim 601 * but a different value for external reclaim. 602 * 603 * Example 604 * Let's have global and A's reclaim in parallel: 605 * | 606 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) 607 * |\ 608 * | C (low = 1G, usage = 2.5G) 609 * B (low = 1G, usage = 0.5G) 610 * 611 * For the global reclaim 612 * A.elow = A.low 613 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow 614 * C.elow = min(C.usage, C.low) 615 * 616 * With the effective values resetting we have A reclaim 617 * A.elow = 0 618 * B.elow = B.low 619 * C.elow = C.low 620 * 621 * If the global reclaim races with A's reclaim then 622 * B.elow = C.elow = 0 because children_low_usage > A.elow) 623 * is possible and reclaiming B would be violating the protection. 624 * 625 */ 626 if (root == memcg) 627 return; 628 629 *min = READ_ONCE(memcg->memory.emin); 630 *low = READ_ONCE(memcg->memory.elow); 631} 632 633void mem_cgroup_calculate_protection(struct mem_cgroup *root, 634 struct mem_cgroup *memcg); 635 636static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 637 struct mem_cgroup *memcg) 638{ 639 /* 640 * The root memcg doesn't account charges, and doesn't support 641 * protection. The target memcg's protection is ignored, see 642 * mem_cgroup_calculate_protection() and mem_cgroup_protection() 643 */ 644 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) || 645 memcg == target; 646} 647 648static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 649 struct mem_cgroup *memcg) 650{ 651 if (mem_cgroup_unprotected(target, memcg)) 652 return false; 653 654 return READ_ONCE(memcg->memory.elow) >= 655 page_counter_read(&memcg->memory); 656} 657 658static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 659 struct mem_cgroup *memcg) 660{ 661 if (mem_cgroup_unprotected(target, memcg)) 662 return false; 663 664 return READ_ONCE(memcg->memory.emin) >= 665 page_counter_read(&memcg->memory); 666} 667 668void mem_cgroup_commit_charge(struct folio *folio, struct mem_cgroup *memcg); 669 670int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp); 671 672/** 673 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup. 674 * @folio: Folio to charge. 675 * @mm: mm context of the allocating task. 676 * @gfp: Reclaim mode. 677 * 678 * Try to charge @folio to the memcg that @mm belongs to, reclaiming 679 * pages according to @gfp if necessary. If @mm is NULL, try to 680 * charge to the active memcg. 681 * 682 * Do not use this for folios allocated for swapin. 683 * 684 * Return: 0 on success. Otherwise, an error code is returned. 685 */ 686static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, 687 gfp_t gfp) 688{ 689 if (mem_cgroup_disabled()) 690 return 0; 691 return __mem_cgroup_charge(folio, mm, gfp); 692} 693 694int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, gfp_t gfp, 695 long nr_pages); 696 697int mem_cgroup_swapin_charge_folio(struct folio *folio, 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 folio *folio); 702 703/** 704 * mem_cgroup_uncharge - Uncharge a folio. 705 * @folio: Folio to uncharge. 706 * 707 * Uncharge a folio previously charged with mem_cgroup_charge(). 708 */ 709static inline void mem_cgroup_uncharge(struct folio *folio) 710{ 711 if (mem_cgroup_disabled()) 712 return; 713 __mem_cgroup_uncharge(folio); 714} 715 716void __mem_cgroup_uncharge_folios(struct folio_batch *folios); 717static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 718{ 719 if (mem_cgroup_disabled()) 720 return; 721 __mem_cgroup_uncharge_folios(folios); 722} 723 724void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages); 725void mem_cgroup_replace_folio(struct folio *old, struct folio *new); 726void mem_cgroup_migrate(struct folio *old, struct folio *new); 727 728/** 729 * mem_cgroup_lruvec - get the lru list vector for a memcg & node 730 * @memcg: memcg of the wanted lruvec 731 * @pgdat: pglist_data 732 * 733 * Returns the lru list vector holding pages for a given @memcg & 734 * @pgdat combination. This can be the node lruvec, if the memory 735 * controller is disabled. 736 */ 737static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 738 struct pglist_data *pgdat) 739{ 740 struct mem_cgroup_per_node *mz; 741 struct lruvec *lruvec; 742 743 if (mem_cgroup_disabled()) { 744 lruvec = &pgdat->__lruvec; 745 goto out; 746 } 747 748 if (!memcg) 749 memcg = root_mem_cgroup; 750 751 mz = memcg->nodeinfo[pgdat->node_id]; 752 lruvec = &mz->lruvec; 753out: 754 /* 755 * Since a node can be onlined after the mem_cgroup was created, 756 * we have to be prepared to initialize lruvec->pgdat here; 757 * and if offlined then reonlined, we need to reinitialize it. 758 */ 759 if (unlikely(lruvec->pgdat != pgdat)) 760 lruvec->pgdat = pgdat; 761 return lruvec; 762} 763 764/** 765 * folio_lruvec - return lruvec for isolating/putting an LRU folio 766 * @folio: Pointer to the folio. 767 * 768 * This function relies on folio->mem_cgroup being stable. 769 */ 770static inline struct lruvec *folio_lruvec(struct folio *folio) 771{ 772 struct mem_cgroup *memcg = folio_memcg(folio); 773 774 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio); 775 return mem_cgroup_lruvec(memcg, folio_pgdat(folio)); 776} 777 778struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); 779 780struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); 781 782struct mem_cgroup *get_mem_cgroup_from_current(void); 783 784struct lruvec *folio_lruvec_lock(struct folio *folio); 785struct lruvec *folio_lruvec_lock_irq(struct folio *folio); 786struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 787 unsigned long *flags); 788 789#ifdef CONFIG_DEBUG_VM 790void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio); 791#else 792static inline 793void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 794{ 795} 796#endif 797 798static inline 799struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ 800 return css ? container_of(css, struct mem_cgroup, css) : NULL; 801} 802 803static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) 804{ 805 return percpu_ref_tryget(&objcg->refcnt); 806} 807 808static inline void obj_cgroup_get(struct obj_cgroup *objcg) 809{ 810 percpu_ref_get(&objcg->refcnt); 811} 812 813static inline void obj_cgroup_get_many(struct obj_cgroup *objcg, 814 unsigned long nr) 815{ 816 percpu_ref_get_many(&objcg->refcnt, nr); 817} 818 819static inline void obj_cgroup_put(struct obj_cgroup *objcg) 820{ 821 percpu_ref_put(&objcg->refcnt); 822} 823 824static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 825{ 826 return !memcg || css_tryget(&memcg->css); 827} 828 829static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 830{ 831 return !memcg || css_tryget_online(&memcg->css); 832} 833 834static inline void mem_cgroup_put(struct mem_cgroup *memcg) 835{ 836 if (memcg) 837 css_put(&memcg->css); 838} 839 840#define mem_cgroup_from_counter(counter, member) \ 841 container_of(counter, struct mem_cgroup, member) 842 843struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, 844 struct mem_cgroup *, 845 struct mem_cgroup_reclaim_cookie *); 846void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); 847void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 848 int (*)(struct task_struct *, void *), void *arg); 849 850static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 851{ 852 if (mem_cgroup_disabled()) 853 return 0; 854 855 return memcg->id.id; 856} 857struct mem_cgroup *mem_cgroup_from_id(unsigned short id); 858 859#ifdef CONFIG_SHRINKER_DEBUG 860static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 861{ 862 return memcg ? cgroup_ino(memcg->css.cgroup) : 0; 863} 864 865struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino); 866#endif 867 868static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 869{ 870 return mem_cgroup_from_css(seq_css(m)); 871} 872 873static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 874{ 875 struct mem_cgroup_per_node *mz; 876 877 if (mem_cgroup_disabled()) 878 return NULL; 879 880 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 881 return mz->memcg; 882} 883 884/** 885 * parent_mem_cgroup - find the accounting parent of a memcg 886 * @memcg: memcg whose parent to find 887 * 888 * Returns the parent memcg, or NULL if this is the root. 889 */ 890static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 891{ 892 return mem_cgroup_from_css(memcg->css.parent); 893} 894 895static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, 896 struct mem_cgroup *root) 897{ 898 if (root == memcg) 899 return true; 900 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); 901} 902 903static inline bool mm_match_cgroup(struct mm_struct *mm, 904 struct mem_cgroup *memcg) 905{ 906 struct mem_cgroup *task_memcg; 907 bool match = false; 908 909 rcu_read_lock(); 910 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 911 if (task_memcg) 912 match = mem_cgroup_is_descendant(task_memcg, memcg); 913 rcu_read_unlock(); 914 return match; 915} 916 917struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio); 918ino_t page_cgroup_ino(struct page *page); 919 920static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 921{ 922 if (mem_cgroup_disabled()) 923 return true; 924 return !!(memcg->css.flags & CSS_ONLINE); 925} 926 927void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, 928 int zid, int nr_pages); 929 930static inline 931unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 932 enum lru_list lru, int zone_idx) 933{ 934 struct mem_cgroup_per_node *mz; 935 936 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 937 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); 938} 939 940void mem_cgroup_handle_over_high(gfp_t gfp_mask); 941 942unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); 943 944unsigned long mem_cgroup_size(struct mem_cgroup *memcg); 945 946void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, 947 struct task_struct *p); 948 949void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); 950 951static inline void mem_cgroup_enter_user_fault(void) 952{ 953 WARN_ON(current->in_user_fault); 954 current->in_user_fault = 1; 955} 956 957static inline void mem_cgroup_exit_user_fault(void) 958{ 959 WARN_ON(!current->in_user_fault); 960 current->in_user_fault = 0; 961} 962 963static inline bool task_in_memcg_oom(struct task_struct *p) 964{ 965 return p->memcg_in_oom; 966} 967 968bool mem_cgroup_oom_synchronize(bool wait); 969struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, 970 struct mem_cgroup *oom_domain); 971void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); 972 973void folio_memcg_lock(struct folio *folio); 974void folio_memcg_unlock(struct folio *folio); 975 976void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val); 977 978/* try to stablize folio_memcg() for all the pages in a memcg */ 979static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg) 980{ 981 rcu_read_lock(); 982 983 if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account)) 984 return true; 985 986 rcu_read_unlock(); 987 return false; 988} 989 990static inline void mem_cgroup_unlock_pages(void) 991{ 992 rcu_read_unlock(); 993} 994 995/* idx can be of type enum memcg_stat_item or node_stat_item */ 996static inline void mod_memcg_state(struct mem_cgroup *memcg, 997 int idx, int val) 998{ 999 unsigned long flags; 1000 1001 local_irq_save(flags); 1002 __mod_memcg_state(memcg, idx, val); 1003 local_irq_restore(flags); 1004} 1005 1006static inline void mod_memcg_page_state(struct page *page, 1007 int idx, int val) 1008{ 1009 struct mem_cgroup *memcg; 1010 1011 if (mem_cgroup_disabled()) 1012 return; 1013 1014 rcu_read_lock(); 1015 memcg = page_memcg(page); 1016 if (memcg) 1017 mod_memcg_state(memcg, idx, val); 1018 rcu_read_unlock(); 1019} 1020 1021unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx); 1022 1023static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1024 enum node_stat_item idx) 1025{ 1026 struct mem_cgroup_per_node *pn; 1027 long x; 1028 1029 if (mem_cgroup_disabled()) 1030 return node_page_state(lruvec_pgdat(lruvec), idx); 1031 1032 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 1033 x = READ_ONCE(pn->lruvec_stats.state[idx]); 1034#ifdef CONFIG_SMP 1035 if (x < 0) 1036 x = 0; 1037#endif 1038 return x; 1039} 1040 1041static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1042 enum node_stat_item idx) 1043{ 1044 struct mem_cgroup_per_node *pn; 1045 long x = 0; 1046 1047 if (mem_cgroup_disabled()) 1048 return node_page_state(lruvec_pgdat(lruvec), idx); 1049 1050 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 1051 x = READ_ONCE(pn->lruvec_stats.state_local[idx]); 1052#ifdef CONFIG_SMP 1053 if (x < 0) 1054 x = 0; 1055#endif 1056 return x; 1057} 1058 1059void mem_cgroup_flush_stats(struct mem_cgroup *memcg); 1060void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg); 1061 1062void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx, 1063 int val); 1064void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val); 1065 1066static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1067 int val) 1068{ 1069 unsigned long flags; 1070 1071 local_irq_save(flags); 1072 __mod_lruvec_kmem_state(p, idx, val); 1073 local_irq_restore(flags); 1074} 1075 1076static inline void mod_memcg_lruvec_state(struct lruvec *lruvec, 1077 enum node_stat_item idx, int val) 1078{ 1079 unsigned long flags; 1080 1081 local_irq_save(flags); 1082 __mod_memcg_lruvec_state(lruvec, idx, val); 1083 local_irq_restore(flags); 1084} 1085 1086void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, 1087 unsigned long count); 1088 1089static inline void count_memcg_events(struct mem_cgroup *memcg, 1090 enum vm_event_item idx, 1091 unsigned long count) 1092{ 1093 unsigned long flags; 1094 1095 local_irq_save(flags); 1096 __count_memcg_events(memcg, idx, count); 1097 local_irq_restore(flags); 1098} 1099 1100static inline void count_memcg_folio_events(struct folio *folio, 1101 enum vm_event_item idx, unsigned long nr) 1102{ 1103 struct mem_cgroup *memcg = folio_memcg(folio); 1104 1105 if (memcg) 1106 count_memcg_events(memcg, idx, nr); 1107} 1108 1109static inline void count_memcg_event_mm(struct mm_struct *mm, 1110 enum vm_event_item idx) 1111{ 1112 struct mem_cgroup *memcg; 1113 1114 if (mem_cgroup_disabled()) 1115 return; 1116 1117 rcu_read_lock(); 1118 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1119 if (likely(memcg)) 1120 count_memcg_events(memcg, idx, 1); 1121 rcu_read_unlock(); 1122} 1123 1124static inline void memcg_memory_event(struct mem_cgroup *memcg, 1125 enum memcg_memory_event event) 1126{ 1127 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX || 1128 event == MEMCG_SWAP_FAIL; 1129 1130 atomic_long_inc(&memcg->memory_events_local[event]); 1131 if (!swap_event) 1132 cgroup_file_notify(&memcg->events_local_file); 1133 1134 do { 1135 atomic_long_inc(&memcg->memory_events[event]); 1136 if (swap_event) 1137 cgroup_file_notify(&memcg->swap_events_file); 1138 else 1139 cgroup_file_notify(&memcg->events_file); 1140 1141 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1142 break; 1143 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) 1144 break; 1145 } while ((memcg = parent_mem_cgroup(memcg)) && 1146 !mem_cgroup_is_root(memcg)); 1147} 1148 1149static inline void memcg_memory_event_mm(struct mm_struct *mm, 1150 enum memcg_memory_event event) 1151{ 1152 struct mem_cgroup *memcg; 1153 1154 if (mem_cgroup_disabled()) 1155 return; 1156 1157 rcu_read_lock(); 1158 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1159 if (likely(memcg)) 1160 memcg_memory_event(memcg, event); 1161 rcu_read_unlock(); 1162} 1163 1164void split_page_memcg(struct page *head, int old_order, int new_order); 1165 1166unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, 1167 gfp_t gfp_mask, 1168 unsigned long *total_scanned); 1169 1170#else /* CONFIG_MEMCG */ 1171 1172#define MEM_CGROUP_ID_SHIFT 0 1173 1174static inline struct mem_cgroup *folio_memcg(struct folio *folio) 1175{ 1176 return NULL; 1177} 1178 1179static inline struct mem_cgroup *page_memcg(struct page *page) 1180{ 1181 return NULL; 1182} 1183 1184static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 1185{ 1186 WARN_ON_ONCE(!rcu_read_lock_held()); 1187 return NULL; 1188} 1189 1190static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 1191{ 1192 return NULL; 1193} 1194 1195static inline struct mem_cgroup *page_memcg_check(struct page *page) 1196{ 1197 return NULL; 1198} 1199 1200static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 1201{ 1202 return NULL; 1203} 1204 1205static inline bool folio_memcg_kmem(struct folio *folio) 1206{ 1207 return false; 1208} 1209 1210static inline bool PageMemcgKmem(struct page *page) 1211{ 1212 return false; 1213} 1214 1215static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 1216{ 1217 return true; 1218} 1219 1220static inline bool mem_cgroup_disabled(void) 1221{ 1222 return true; 1223} 1224 1225static inline void memcg_memory_event(struct mem_cgroup *memcg, 1226 enum memcg_memory_event event) 1227{ 1228} 1229 1230static inline void memcg_memory_event_mm(struct mm_struct *mm, 1231 enum memcg_memory_event event) 1232{ 1233} 1234 1235static inline void mem_cgroup_protection(struct mem_cgroup *root, 1236 struct mem_cgroup *memcg, 1237 unsigned long *min, 1238 unsigned long *low) 1239{ 1240 *min = *low = 0; 1241} 1242 1243static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, 1244 struct mem_cgroup *memcg) 1245{ 1246} 1247 1248static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 1249 struct mem_cgroup *memcg) 1250{ 1251 return true; 1252} 1253static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 1254 struct mem_cgroup *memcg) 1255{ 1256 return false; 1257} 1258 1259static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 1260 struct mem_cgroup *memcg) 1261{ 1262 return false; 1263} 1264 1265static inline void mem_cgroup_commit_charge(struct folio *folio, 1266 struct mem_cgroup *memcg) 1267{ 1268} 1269 1270static inline int mem_cgroup_charge(struct folio *folio, 1271 struct mm_struct *mm, gfp_t gfp) 1272{ 1273 return 0; 1274} 1275 1276static inline int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, 1277 gfp_t gfp, long nr_pages) 1278{ 1279 return 0; 1280} 1281 1282static inline int mem_cgroup_swapin_charge_folio(struct folio *folio, 1283 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry) 1284{ 1285 return 0; 1286} 1287 1288static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry) 1289{ 1290} 1291 1292static inline void mem_cgroup_uncharge(struct folio *folio) 1293{ 1294} 1295 1296static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 1297{ 1298} 1299 1300static inline void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, 1301 unsigned int nr_pages) 1302{ 1303} 1304 1305static inline void mem_cgroup_replace_folio(struct folio *old, 1306 struct folio *new) 1307{ 1308} 1309 1310static inline void mem_cgroup_migrate(struct folio *old, struct folio *new) 1311{ 1312} 1313 1314static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 1315 struct pglist_data *pgdat) 1316{ 1317 return &pgdat->__lruvec; 1318} 1319 1320static inline struct lruvec *folio_lruvec(struct folio *folio) 1321{ 1322 struct pglist_data *pgdat = folio_pgdat(folio); 1323 return &pgdat->__lruvec; 1324} 1325 1326static inline 1327void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 1328{ 1329} 1330 1331static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 1332{ 1333 return NULL; 1334} 1335 1336static inline bool mm_match_cgroup(struct mm_struct *mm, 1337 struct mem_cgroup *memcg) 1338{ 1339 return true; 1340} 1341 1342static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) 1343{ 1344 return NULL; 1345} 1346 1347static inline struct mem_cgroup *get_mem_cgroup_from_current(void) 1348{ 1349 return NULL; 1350} 1351 1352static inline 1353struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css) 1354{ 1355 return NULL; 1356} 1357 1358static inline void obj_cgroup_put(struct obj_cgroup *objcg) 1359{ 1360} 1361 1362static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 1363{ 1364 return true; 1365} 1366 1367static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 1368{ 1369 return true; 1370} 1371 1372static inline void mem_cgroup_put(struct mem_cgroup *memcg) 1373{ 1374} 1375 1376static inline struct lruvec *folio_lruvec_lock(struct folio *folio) 1377{ 1378 struct pglist_data *pgdat = folio_pgdat(folio); 1379 1380 spin_lock(&pgdat->__lruvec.lru_lock); 1381 return &pgdat->__lruvec; 1382} 1383 1384static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio) 1385{ 1386 struct pglist_data *pgdat = folio_pgdat(folio); 1387 1388 spin_lock_irq(&pgdat->__lruvec.lru_lock); 1389 return &pgdat->__lruvec; 1390} 1391 1392static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 1393 unsigned long *flagsp) 1394{ 1395 struct pglist_data *pgdat = folio_pgdat(folio); 1396 1397 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp); 1398 return &pgdat->__lruvec; 1399} 1400 1401static inline struct mem_cgroup * 1402mem_cgroup_iter(struct mem_cgroup *root, 1403 struct mem_cgroup *prev, 1404 struct mem_cgroup_reclaim_cookie *reclaim) 1405{ 1406 return NULL; 1407} 1408 1409static inline void mem_cgroup_iter_break(struct mem_cgroup *root, 1410 struct mem_cgroup *prev) 1411{ 1412} 1413 1414static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 1415 int (*fn)(struct task_struct *, void *), void *arg) 1416{ 1417} 1418 1419static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 1420{ 1421 return 0; 1422} 1423 1424static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) 1425{ 1426 WARN_ON_ONCE(id); 1427 /* XXX: This should always return root_mem_cgroup */ 1428 return NULL; 1429} 1430 1431#ifdef CONFIG_SHRINKER_DEBUG 1432static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 1433{ 1434 return 0; 1435} 1436 1437static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino) 1438{ 1439 return NULL; 1440} 1441#endif 1442 1443static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 1444{ 1445 return NULL; 1446} 1447 1448static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 1449{ 1450 return NULL; 1451} 1452 1453static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 1454{ 1455 return true; 1456} 1457 1458static inline 1459unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 1460 enum lru_list lru, int zone_idx) 1461{ 1462 return 0; 1463} 1464 1465static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) 1466{ 1467 return 0; 1468} 1469 1470static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) 1471{ 1472 return 0; 1473} 1474 1475static inline void 1476mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) 1477{ 1478} 1479 1480static inline void 1481mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) 1482{ 1483} 1484 1485static inline void folio_memcg_lock(struct folio *folio) 1486{ 1487} 1488 1489static inline void folio_memcg_unlock(struct folio *folio) 1490{ 1491} 1492 1493static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg) 1494{ 1495 /* to match folio_memcg_rcu() */ 1496 rcu_read_lock(); 1497 return true; 1498} 1499 1500static inline void mem_cgroup_unlock_pages(void) 1501{ 1502 rcu_read_unlock(); 1503} 1504 1505static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask) 1506{ 1507} 1508 1509static inline void mem_cgroup_enter_user_fault(void) 1510{ 1511} 1512 1513static inline void mem_cgroup_exit_user_fault(void) 1514{ 1515} 1516 1517static inline bool task_in_memcg_oom(struct task_struct *p) 1518{ 1519 return false; 1520} 1521 1522static inline bool mem_cgroup_oom_synchronize(bool wait) 1523{ 1524 return false; 1525} 1526 1527static inline struct mem_cgroup *mem_cgroup_get_oom_group( 1528 struct task_struct *victim, struct mem_cgroup *oom_domain) 1529{ 1530 return NULL; 1531} 1532 1533static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) 1534{ 1535} 1536 1537static inline void __mod_memcg_state(struct mem_cgroup *memcg, 1538 int idx, 1539 int nr) 1540{ 1541} 1542 1543static inline void mod_memcg_state(struct mem_cgroup *memcg, 1544 int idx, 1545 int nr) 1546{ 1547} 1548 1549static inline void mod_memcg_page_state(struct page *page, 1550 int idx, int val) 1551{ 1552} 1553 1554static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) 1555{ 1556 return 0; 1557} 1558 1559static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1560 enum node_stat_item idx) 1561{ 1562 return node_page_state(lruvec_pgdat(lruvec), idx); 1563} 1564 1565static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1566 enum node_stat_item idx) 1567{ 1568 return node_page_state(lruvec_pgdat(lruvec), idx); 1569} 1570 1571static inline void mem_cgroup_flush_stats(struct mem_cgroup *memcg) 1572{ 1573} 1574 1575static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg) 1576{ 1577} 1578 1579static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec, 1580 enum node_stat_item idx, int val) 1581{ 1582} 1583 1584static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1585 int val) 1586{ 1587 struct page *page = virt_to_head_page(p); 1588 1589 __mod_node_page_state(page_pgdat(page), idx, val); 1590} 1591 1592static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1593 int val) 1594{ 1595 struct page *page = virt_to_head_page(p); 1596 1597 mod_node_page_state(page_pgdat(page), idx, val); 1598} 1599 1600static inline void count_memcg_events(struct mem_cgroup *memcg, 1601 enum vm_event_item idx, 1602 unsigned long count) 1603{ 1604} 1605 1606static inline void __count_memcg_events(struct mem_cgroup *memcg, 1607 enum vm_event_item idx, 1608 unsigned long count) 1609{ 1610} 1611 1612static inline void count_memcg_folio_events(struct folio *folio, 1613 enum vm_event_item idx, unsigned long nr) 1614{ 1615} 1616 1617static inline 1618void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) 1619{ 1620} 1621 1622static inline void split_page_memcg(struct page *head, int old_order, int new_order) 1623{ 1624} 1625 1626static inline 1627unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, 1628 gfp_t gfp_mask, 1629 unsigned long *total_scanned) 1630{ 1631 return 0; 1632} 1633#endif /* CONFIG_MEMCG */ 1634 1635static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx) 1636{ 1637 __mod_lruvec_kmem_state(p, idx, 1); 1638} 1639 1640static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx) 1641{ 1642 __mod_lruvec_kmem_state(p, idx, -1); 1643} 1644 1645static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) 1646{ 1647 struct mem_cgroup *memcg; 1648 1649 memcg = lruvec_memcg(lruvec); 1650 if (!memcg) 1651 return NULL; 1652 memcg = parent_mem_cgroup(memcg); 1653 if (!memcg) 1654 return NULL; 1655 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); 1656} 1657 1658static inline void unlock_page_lruvec(struct lruvec *lruvec) 1659{ 1660 spin_unlock(&lruvec->lru_lock); 1661} 1662 1663static inline void unlock_page_lruvec_irq(struct lruvec *lruvec) 1664{ 1665 spin_unlock_irq(&lruvec->lru_lock); 1666} 1667 1668static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec, 1669 unsigned long flags) 1670{ 1671 spin_unlock_irqrestore(&lruvec->lru_lock, flags); 1672} 1673 1674/* Test requires a stable page->memcg binding, see page_memcg() */ 1675static inline bool folio_matches_lruvec(struct folio *folio, 1676 struct lruvec *lruvec) 1677{ 1678 return lruvec_pgdat(lruvec) == folio_pgdat(folio) && 1679 lruvec_memcg(lruvec) == folio_memcg(folio); 1680} 1681 1682/* Don't lock again iff page's lruvec locked */ 1683static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio, 1684 struct lruvec *locked_lruvec) 1685{ 1686 if (locked_lruvec) { 1687 if (folio_matches_lruvec(folio, locked_lruvec)) 1688 return locked_lruvec; 1689 1690 unlock_page_lruvec_irq(locked_lruvec); 1691 } 1692 1693 return folio_lruvec_lock_irq(folio); 1694} 1695 1696/* Don't lock again iff folio's lruvec locked */ 1697static inline void folio_lruvec_relock_irqsave(struct folio *folio, 1698 struct lruvec **lruvecp, unsigned long *flags) 1699{ 1700 if (*lruvecp) { 1701 if (folio_matches_lruvec(folio, *lruvecp)) 1702 return; 1703 1704 unlock_page_lruvec_irqrestore(*lruvecp, *flags); 1705 } 1706 1707 *lruvecp = folio_lruvec_lock_irqsave(folio, flags); 1708} 1709 1710#ifdef CONFIG_CGROUP_WRITEBACK 1711 1712struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); 1713void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, 1714 unsigned long *pheadroom, unsigned long *pdirty, 1715 unsigned long *pwriteback); 1716 1717void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio, 1718 struct bdi_writeback *wb); 1719 1720static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1721 struct bdi_writeback *wb) 1722{ 1723 struct mem_cgroup *memcg; 1724 1725 if (mem_cgroup_disabled()) 1726 return; 1727 1728 memcg = folio_memcg(folio); 1729 if (unlikely(memcg && &memcg->css != wb->memcg_css)) 1730 mem_cgroup_track_foreign_dirty_slowpath(folio, wb); 1731} 1732 1733void mem_cgroup_flush_foreign(struct bdi_writeback *wb); 1734 1735#else /* CONFIG_CGROUP_WRITEBACK */ 1736 1737static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) 1738{ 1739 return NULL; 1740} 1741 1742static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, 1743 unsigned long *pfilepages, 1744 unsigned long *pheadroom, 1745 unsigned long *pdirty, 1746 unsigned long *pwriteback) 1747{ 1748} 1749 1750static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1751 struct bdi_writeback *wb) 1752{ 1753} 1754 1755static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) 1756{ 1757} 1758 1759#endif /* CONFIG_CGROUP_WRITEBACK */ 1760 1761struct sock; 1762bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages, 1763 gfp_t gfp_mask); 1764void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); 1765#ifdef CONFIG_MEMCG 1766extern struct static_key_false memcg_sockets_enabled_key; 1767#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) 1768void mem_cgroup_sk_alloc(struct sock *sk); 1769void mem_cgroup_sk_free(struct sock *sk); 1770static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1771{ 1772 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1773 return !!memcg->tcpmem_pressure; 1774 do { 1775 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure))) 1776 return true; 1777 } while ((memcg = parent_mem_cgroup(memcg))); 1778 return false; 1779} 1780 1781int alloc_shrinker_info(struct mem_cgroup *memcg); 1782void free_shrinker_info(struct mem_cgroup *memcg); 1783void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id); 1784void reparent_shrinker_deferred(struct mem_cgroup *memcg); 1785#else 1786#define mem_cgroup_sockets_enabled 0 1787static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; 1788static inline void mem_cgroup_sk_free(struct sock *sk) { }; 1789static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1790{ 1791 return false; 1792} 1793 1794static inline void set_shrinker_bit(struct mem_cgroup *memcg, 1795 int nid, int shrinker_id) 1796{ 1797} 1798#endif 1799 1800#ifdef CONFIG_MEMCG_KMEM 1801bool mem_cgroup_kmem_disabled(void); 1802int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); 1803void __memcg_kmem_uncharge_page(struct page *page, int order); 1804 1805/* 1806 * The returned objcg pointer is safe to use without additional 1807 * protection within a scope. The scope is defined either by 1808 * the current task (similar to the "current" global variable) 1809 * or by set_active_memcg() pair. 1810 * Please, use obj_cgroup_get() to get a reference if the pointer 1811 * needs to be used outside of the local scope. 1812 */ 1813struct obj_cgroup *current_obj_cgroup(void); 1814struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio); 1815 1816static inline struct obj_cgroup *get_obj_cgroup_from_current(void) 1817{ 1818 struct obj_cgroup *objcg = current_obj_cgroup(); 1819 1820 if (objcg) 1821 obj_cgroup_get(objcg); 1822 1823 return objcg; 1824} 1825 1826int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); 1827void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); 1828 1829extern struct static_key_false memcg_bpf_enabled_key; 1830static inline bool memcg_bpf_enabled(void) 1831{ 1832 return static_branch_likely(&memcg_bpf_enabled_key); 1833} 1834 1835extern struct static_key_false memcg_kmem_online_key; 1836 1837static inline bool memcg_kmem_online(void) 1838{ 1839 return static_branch_likely(&memcg_kmem_online_key); 1840} 1841 1842static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1843 int order) 1844{ 1845 if (memcg_kmem_online()) 1846 return __memcg_kmem_charge_page(page, gfp, order); 1847 return 0; 1848} 1849 1850static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1851{ 1852 if (memcg_kmem_online()) 1853 __memcg_kmem_uncharge_page(page, order); 1854} 1855 1856/* 1857 * A helper for accessing memcg's kmem_id, used for getting 1858 * corresponding LRU lists. 1859 */ 1860static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1861{ 1862 return memcg ? memcg->kmemcg_id : -1; 1863} 1864 1865struct mem_cgroup *mem_cgroup_from_obj(void *p); 1866struct mem_cgroup *mem_cgroup_from_slab_obj(void *p); 1867 1868static inline void count_objcg_event(struct obj_cgroup *objcg, 1869 enum vm_event_item idx) 1870{ 1871 struct mem_cgroup *memcg; 1872 1873 if (!memcg_kmem_online()) 1874 return; 1875 1876 rcu_read_lock(); 1877 memcg = obj_cgroup_memcg(objcg); 1878 count_memcg_events(memcg, idx, 1); 1879 rcu_read_unlock(); 1880} 1881 1882#else 1883static inline bool mem_cgroup_kmem_disabled(void) 1884{ 1885 return true; 1886} 1887 1888static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1889 int order) 1890{ 1891 return 0; 1892} 1893 1894static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1895{ 1896} 1897 1898static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1899 int order) 1900{ 1901 return 0; 1902} 1903 1904static inline void __memcg_kmem_uncharge_page(struct page *page, int order) 1905{ 1906} 1907 1908static inline struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio) 1909{ 1910 return NULL; 1911} 1912 1913static inline bool memcg_bpf_enabled(void) 1914{ 1915 return false; 1916} 1917 1918static inline bool memcg_kmem_online(void) 1919{ 1920 return false; 1921} 1922 1923static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1924{ 1925 return -1; 1926} 1927 1928static inline struct mem_cgroup *mem_cgroup_from_obj(void *p) 1929{ 1930 return NULL; 1931} 1932 1933static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p) 1934{ 1935 return NULL; 1936} 1937 1938static inline void count_objcg_event(struct obj_cgroup *objcg, 1939 enum vm_event_item idx) 1940{ 1941} 1942 1943#endif /* CONFIG_MEMCG_KMEM */ 1944 1945#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP) 1946bool obj_cgroup_may_zswap(struct obj_cgroup *objcg); 1947void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size); 1948void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size); 1949bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg); 1950#else 1951static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg) 1952{ 1953 return true; 1954} 1955static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, 1956 size_t size) 1957{ 1958} 1959static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, 1960 size_t size) 1961{ 1962} 1963static inline bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg) 1964{ 1965 /* if zswap is disabled, do not block pages going to the swapping device */ 1966 return true; 1967} 1968#endif 1969 1970#endif /* _LINUX_MEMCONTROL_H */ 1971