1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* memcontrol.h - Memory Controller 3 * 4 * Copyright IBM Corporation, 2007 5 * Author Balbir Singh <[email protected]> 6 * 7 * Copyright 2007 OpenVZ SWsoft Inc 8 * Author: Pavel Emelianov <[email protected]> 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 27 struct mem_cgroup; 28 struct obj_cgroup; 29 struct page; 30 struct mm_struct; 31 struct kmem_cache; 32 33 /* Cgroup-specific page state, on top of universal node page state */ 34 enum 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 45 enum 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 58 struct 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 67 struct mem_cgroup_id { 68 int id; 69 refcount_t ref; 70 }; 71 72 struct memcg_vmstats_percpu; 73 struct memcg_vmstats; 74 struct lruvec_stats_percpu; 75 struct lruvec_stats; 76 77 struct mem_cgroup_reclaim_iter { 78 struct mem_cgroup *position; 79 /* scan generation, increased every round-trip */ 80 unsigned int generation; 81 }; 82 83 /* 84 * per-node information in memory controller. 85 */ 86 struct mem_cgroup_per_node { 87 /* Keep the read-only fields at the start */ 88 struct mem_cgroup *memcg; /* Back pointer, we cannot */ 89 /* use container_of */ 90 91 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu; 92 struct lruvec_stats *lruvec_stats; 93 struct shrinker_info __rcu *shrinker_info; 94 95 #ifdef CONFIG_MEMCG_V1 96 /* 97 * Memcg-v1 only stuff in middle as buffer between read mostly fields 98 * and update often fields to avoid false sharing. If v1 stuff is 99 * not present, an explicit padding is needed. 100 */ 101 102 struct rb_node tree_node; /* RB tree node */ 103 unsigned long usage_in_excess;/* Set to the value by which */ 104 /* the soft limit is exceeded*/ 105 bool on_tree; 106 #else 107 CACHELINE_PADDING(_pad1_); 108 #endif 109 110 /* Fields which get updated often at the end. */ 111 struct lruvec lruvec; 112 CACHELINE_PADDING(_pad2_); 113 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; 114 struct mem_cgroup_reclaim_iter iter; 115 }; 116 117 struct mem_cgroup_threshold { 118 struct eventfd_ctx *eventfd; 119 unsigned long threshold; 120 }; 121 122 /* For threshold */ 123 struct mem_cgroup_threshold_ary { 124 /* An array index points to threshold just below or equal to usage. */ 125 int current_threshold; 126 /* Size of entries[] */ 127 unsigned int size; 128 /* Array of thresholds */ 129 struct mem_cgroup_threshold entries[] __counted_by(size); 130 }; 131 132 struct mem_cgroup_thresholds { 133 /* Primary thresholds array */ 134 struct mem_cgroup_threshold_ary *primary; 135 /* 136 * Spare threshold array. 137 * This is needed to make mem_cgroup_unregister_event() "never fail". 138 * It must be able to store at least primary->size - 1 entries. 139 */ 140 struct mem_cgroup_threshold_ary *spare; 141 }; 142 143 /* 144 * Remember four most recent foreign writebacks with dirty pages in this 145 * cgroup. Inode sharing is expected to be uncommon and, even if we miss 146 * one in a given round, we're likely to catch it later if it keeps 147 * foreign-dirtying, so a fairly low count should be enough. 148 * 149 * See mem_cgroup_track_foreign_dirty_slowpath() for details. 150 */ 151 #define MEMCG_CGWB_FRN_CNT 4 152 153 struct memcg_cgwb_frn { 154 u64 bdi_id; /* bdi->id of the foreign inode */ 155 int memcg_id; /* memcg->css.id of foreign inode */ 156 u64 at; /* jiffies_64 at the time of dirtying */ 157 struct wb_completion done; /* tracks in-flight foreign writebacks */ 158 }; 159 160 /* 161 * Bucket for arbitrarily byte-sized objects charged to a memory 162 * cgroup. The bucket can be reparented in one piece when the cgroup 163 * is destroyed, without having to round up the individual references 164 * of all live memory objects in the wild. 165 */ 166 struct obj_cgroup { 167 struct percpu_ref refcnt; 168 struct mem_cgroup *memcg; 169 atomic_t nr_charged_bytes; 170 union { 171 struct list_head list; /* protected by objcg_lock */ 172 struct rcu_head rcu; 173 }; 174 }; 175 176 /* 177 * The memory controller data structure. The memory controller controls both 178 * page cache and RSS per cgroup. We would eventually like to provide 179 * statistics based on the statistics developed by Rik Van Riel for clock-pro, 180 * to help the administrator determine what knobs to tune. 181 */ 182 struct mem_cgroup { 183 struct cgroup_subsys_state css; 184 185 /* Private memcg ID. Used to ID objects that outlive the cgroup */ 186 struct mem_cgroup_id id; 187 188 /* Accounted resources */ 189 struct page_counter memory; /* Both v1 & v2 */ 190 191 union { 192 struct page_counter swap; /* v2 only */ 193 struct page_counter memsw; /* v1 only */ 194 }; 195 196 /* registered local peak watchers */ 197 struct list_head memory_peaks; 198 struct list_head swap_peaks; 199 spinlock_t peaks_lock; 200 201 /* Range enforcement for interrupt charges */ 202 struct work_struct high_work; 203 204 #ifdef CONFIG_ZSWAP 205 unsigned long zswap_max; 206 207 /* 208 * Prevent pages from this memcg from being written back from zswap to 209 * swap, and from being swapped out on zswap store failures. 210 */ 211 bool zswap_writeback; 212 #endif 213 214 /* vmpressure notifications */ 215 struct vmpressure vmpressure; 216 217 /* 218 * Should the OOM killer kill all belonging tasks, had it kill one? 219 */ 220 bool oom_group; 221 222 int swappiness; 223 224 /* memory.events and memory.events.local */ 225 struct cgroup_file events_file; 226 struct cgroup_file events_local_file; 227 228 /* handle for "memory.swap.events" */ 229 struct cgroup_file swap_events_file; 230 231 /* memory.stat */ 232 struct memcg_vmstats *vmstats; 233 234 /* memory.events */ 235 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS]; 236 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS]; 237 238 /* 239 * Hint of reclaim pressure for socket memroy management. Note 240 * that this indicator should NOT be used in legacy cgroup mode 241 * where socket memory is accounted/charged separately. 242 */ 243 unsigned long socket_pressure; 244 245 int kmemcg_id; 246 /* 247 * memcg->objcg is wiped out as a part of the objcg repaprenting 248 * process. memcg->orig_objcg preserves a pointer (and a reference) 249 * to the original objcg until the end of live of memcg. 250 */ 251 struct obj_cgroup __rcu *objcg; 252 struct obj_cgroup *orig_objcg; 253 /* list of inherited objcgs, protected by objcg_lock */ 254 struct list_head objcg_list; 255 256 struct memcg_vmstats_percpu __percpu *vmstats_percpu; 257 258 #ifdef CONFIG_CGROUP_WRITEBACK 259 struct list_head cgwb_list; 260 struct wb_domain cgwb_domain; 261 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT]; 262 #endif 263 264 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 265 struct deferred_split deferred_split_queue; 266 #endif 267 268 #ifdef CONFIG_LRU_GEN_WALKS_MMU 269 /* per-memcg mm_struct list */ 270 struct lru_gen_mm_list mm_list; 271 #endif 272 273 #ifdef CONFIG_MEMCG_V1 274 /* Legacy consumer-oriented counters */ 275 struct page_counter kmem; /* v1 only */ 276 struct page_counter tcpmem; /* v1 only */ 277 278 unsigned long soft_limit; 279 280 /* protected by memcg_oom_lock */ 281 bool oom_lock; 282 int under_oom; 283 284 /* OOM-Killer disable */ 285 int oom_kill_disable; 286 287 /* protect arrays of thresholds */ 288 struct mutex thresholds_lock; 289 290 /* thresholds for memory usage. RCU-protected */ 291 struct mem_cgroup_thresholds thresholds; 292 293 /* thresholds for mem+swap usage. RCU-protected */ 294 struct mem_cgroup_thresholds memsw_thresholds; 295 296 /* For oom notifier event fd */ 297 struct list_head oom_notify; 298 299 /* 300 * Should we move charges of a task when a task is moved into this 301 * mem_cgroup ? And what type of charges should we move ? 302 */ 303 unsigned long move_charge_at_immigrate; 304 /* taken only while moving_account > 0 */ 305 spinlock_t move_lock; 306 unsigned long move_lock_flags; 307 308 /* Legacy tcp memory accounting */ 309 bool tcpmem_active; 310 int tcpmem_pressure; 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 /* List of events which userspace want to receive */ 319 struct list_head event_list; 320 spinlock_t event_list_lock; 321 #endif /* CONFIG_MEMCG_V1 */ 322 323 struct mem_cgroup_per_node *nodeinfo[]; 324 }; 325 326 /* 327 * size of first charge trial. 328 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the 329 * workload. 330 */ 331 #define MEMCG_CHARGE_BATCH 64U 332 333 extern struct mem_cgroup *root_mem_cgroup; 334 335 enum page_memcg_data_flags { 336 /* page->memcg_data is a pointer to an slabobj_ext vector */ 337 MEMCG_DATA_OBJEXTS = (1UL << 0), 338 /* page has been accounted as a non-slab kernel page */ 339 MEMCG_DATA_KMEM = (1UL << 1), 340 /* the next bit after the last actual flag */ 341 __NR_MEMCG_DATA_FLAGS = (1UL << 2), 342 }; 343 344 #define __FIRST_OBJEXT_FLAG __NR_MEMCG_DATA_FLAGS 345 346 #else /* CONFIG_MEMCG */ 347 348 #define __FIRST_OBJEXT_FLAG (1UL << 0) 349 350 #endif /* CONFIG_MEMCG */ 351 352 enum objext_flags { 353 /* slabobj_ext vector failed to allocate */ 354 OBJEXTS_ALLOC_FAIL = __FIRST_OBJEXT_FLAG, 355 /* the next bit after the last actual flag */ 356 __NR_OBJEXTS_FLAGS = (__FIRST_OBJEXT_FLAG << 1), 357 }; 358 359 #define OBJEXTS_FLAGS_MASK (__NR_OBJEXTS_FLAGS - 1) 360 361 #ifdef CONFIG_MEMCG 362 363 static inline bool folio_memcg_kmem(struct folio *folio); 364 365 /* 366 * After the initialization objcg->memcg is always pointing at 367 * a valid memcg, but can be atomically swapped to the parent memcg. 368 * 369 * The caller must ensure that the returned memcg won't be released: 370 * e.g. acquire the rcu_read_lock or css_set_lock. 371 */ 372 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg) 373 { 374 return READ_ONCE(objcg->memcg); 375 } 376 377 /* 378 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio 379 * @folio: Pointer to the folio. 380 * 381 * Returns a pointer to the memory cgroup associated with the folio, 382 * or NULL. This function assumes that the folio is known to have a 383 * proper memory cgroup pointer. It's not safe to call this function 384 * against some type of folios, e.g. slab folios or ex-slab folios or 385 * kmem folios. 386 */ 387 static inline struct mem_cgroup *__folio_memcg(struct folio *folio) 388 { 389 unsigned long memcg_data = folio->memcg_data; 390 391 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 392 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio); 393 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio); 394 395 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 396 } 397 398 /* 399 * __folio_objcg - get the object cgroup associated with a kmem folio. 400 * @folio: Pointer to the folio. 401 * 402 * Returns a pointer to the object cgroup associated with the folio, 403 * or NULL. This function assumes that the folio is known to have a 404 * proper object cgroup pointer. It's not safe to call this function 405 * against some type of folios, e.g. slab folios or ex-slab folios or 406 * LRU folios. 407 */ 408 static inline struct obj_cgroup *__folio_objcg(struct folio *folio) 409 { 410 unsigned long memcg_data = folio->memcg_data; 411 412 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 413 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio); 414 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio); 415 416 return (struct obj_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 417 } 418 419 /* 420 * folio_memcg - Get the memory cgroup associated with a folio. 421 * @folio: Pointer to the folio. 422 * 423 * Returns a pointer to the memory cgroup associated with the folio, 424 * or NULL. This function assumes that the folio is known to have a 425 * proper memory cgroup pointer. It's not safe to call this function 426 * against some type of folios, e.g. slab folios or ex-slab folios. 427 * 428 * For a non-kmem folio any of the following ensures folio and memcg binding 429 * stability: 430 * 431 * - the folio lock 432 * - LRU isolation 433 * - folio_memcg_lock() 434 * - exclusive reference 435 * - mem_cgroup_trylock_pages() 436 * 437 * For a kmem folio a caller should hold an rcu read lock to protect memcg 438 * associated with a kmem folio from being released. 439 */ 440 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 441 { 442 if (folio_memcg_kmem(folio)) 443 return obj_cgroup_memcg(__folio_objcg(folio)); 444 return __folio_memcg(folio); 445 } 446 447 /** 448 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio. 449 * @folio: Pointer to the folio. 450 * 451 * This function assumes that the folio is known to have a 452 * proper memory cgroup pointer. It's not safe to call this function 453 * against some type of folios, e.g. slab folios or ex-slab folios. 454 * 455 * Return: A pointer to the memory cgroup associated with the folio, 456 * or NULL. 457 */ 458 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 459 { 460 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 461 462 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 463 WARN_ON_ONCE(!rcu_read_lock_held()); 464 465 if (memcg_data & MEMCG_DATA_KMEM) { 466 struct obj_cgroup *objcg; 467 468 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 469 return obj_cgroup_memcg(objcg); 470 } 471 472 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 473 } 474 475 /* 476 * folio_memcg_check - Get the memory cgroup associated with a folio. 477 * @folio: Pointer to the folio. 478 * 479 * Returns a pointer to the memory cgroup associated with the folio, 480 * or NULL. This function unlike folio_memcg() can take any folio 481 * as an argument. It has to be used in cases when it's not known if a folio 482 * has an associated memory cgroup pointer or an object cgroups vector or 483 * an object cgroup. 484 * 485 * For a non-kmem folio any of the following ensures folio and memcg binding 486 * stability: 487 * 488 * - the folio lock 489 * - LRU isolation 490 * - lock_folio_memcg() 491 * - exclusive reference 492 * - mem_cgroup_trylock_pages() 493 * 494 * For a kmem folio a caller should hold an rcu read lock to protect memcg 495 * associated with a kmem folio from being released. 496 */ 497 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 498 { 499 /* 500 * Because folio->memcg_data might be changed asynchronously 501 * for slabs, READ_ONCE() should be used here. 502 */ 503 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 504 505 if (memcg_data & MEMCG_DATA_OBJEXTS) 506 return NULL; 507 508 if (memcg_data & MEMCG_DATA_KMEM) { 509 struct obj_cgroup *objcg; 510 511 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 512 return obj_cgroup_memcg(objcg); 513 } 514 515 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 516 } 517 518 static inline struct mem_cgroup *page_memcg_check(struct page *page) 519 { 520 if (PageTail(page)) 521 return NULL; 522 return folio_memcg_check((struct folio *)page); 523 } 524 525 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 526 { 527 struct mem_cgroup *memcg; 528 529 rcu_read_lock(); 530 retry: 531 memcg = obj_cgroup_memcg(objcg); 532 if (unlikely(!css_tryget(&memcg->css))) 533 goto retry; 534 rcu_read_unlock(); 535 536 return memcg; 537 } 538 539 /* 540 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set. 541 * @folio: Pointer to the folio. 542 * 543 * Checks if the folio has MemcgKmem flag set. The caller must ensure 544 * that the folio has an associated memory cgroup. It's not safe to call 545 * this function against some types of folios, e.g. slab folios. 546 */ 547 static inline bool folio_memcg_kmem(struct folio *folio) 548 { 549 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page); 550 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJEXTS, folio); 551 return folio->memcg_data & MEMCG_DATA_KMEM; 552 } 553 554 static inline bool PageMemcgKmem(struct page *page) 555 { 556 return folio_memcg_kmem(page_folio(page)); 557 } 558 559 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 560 { 561 return (memcg == root_mem_cgroup); 562 } 563 564 static inline bool mem_cgroup_disabled(void) 565 { 566 return !cgroup_subsys_enabled(memory_cgrp_subsys); 567 } 568 569 static inline void mem_cgroup_protection(struct mem_cgroup *root, 570 struct mem_cgroup *memcg, 571 unsigned long *min, 572 unsigned long *low) 573 { 574 *min = *low = 0; 575 576 if (mem_cgroup_disabled()) 577 return; 578 579 /* 580 * There is no reclaim protection applied to a targeted reclaim. 581 * We are special casing this specific case here because 582 * mem_cgroup_calculate_protection is not robust enough to keep 583 * the protection invariant for calculated effective values for 584 * parallel reclaimers with different reclaim target. This is 585 * especially a problem for tail memcgs (as they have pages on LRU) 586 * which would want to have effective values 0 for targeted reclaim 587 * but a different value for external reclaim. 588 * 589 * Example 590 * Let's have global and A's reclaim in parallel: 591 * | 592 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) 593 * |\ 594 * | C (low = 1G, usage = 2.5G) 595 * B (low = 1G, usage = 0.5G) 596 * 597 * For the global reclaim 598 * A.elow = A.low 599 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow 600 * C.elow = min(C.usage, C.low) 601 * 602 * With the effective values resetting we have A reclaim 603 * A.elow = 0 604 * B.elow = B.low 605 * C.elow = C.low 606 * 607 * If the global reclaim races with A's reclaim then 608 * B.elow = C.elow = 0 because children_low_usage > A.elow) 609 * is possible and reclaiming B would be violating the protection. 610 * 611 */ 612 if (root == memcg) 613 return; 614 615 *min = READ_ONCE(memcg->memory.emin); 616 *low = READ_ONCE(memcg->memory.elow); 617 } 618 619 void mem_cgroup_calculate_protection(struct mem_cgroup *root, 620 struct mem_cgroup *memcg); 621 622 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 623 struct mem_cgroup *memcg) 624 { 625 /* 626 * The root memcg doesn't account charges, and doesn't support 627 * protection. The target memcg's protection is ignored, see 628 * mem_cgroup_calculate_protection() and mem_cgroup_protection() 629 */ 630 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) || 631 memcg == target; 632 } 633 634 static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 635 struct mem_cgroup *memcg) 636 { 637 if (mem_cgroup_unprotected(target, memcg)) 638 return false; 639 640 return READ_ONCE(memcg->memory.elow) >= 641 page_counter_read(&memcg->memory); 642 } 643 644 static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 645 struct mem_cgroup *memcg) 646 { 647 if (mem_cgroup_unprotected(target, memcg)) 648 return false; 649 650 return READ_ONCE(memcg->memory.emin) >= 651 page_counter_read(&memcg->memory); 652 } 653 654 void mem_cgroup_commit_charge(struct folio *folio, struct mem_cgroup *memcg); 655 656 int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp); 657 658 /** 659 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup. 660 * @folio: Folio to charge. 661 * @mm: mm context of the allocating task. 662 * @gfp: Reclaim mode. 663 * 664 * Try to charge @folio to the memcg that @mm belongs to, reclaiming 665 * pages according to @gfp if necessary. If @mm is NULL, try to 666 * charge to the active memcg. 667 * 668 * Do not use this for folios allocated for swapin. 669 * 670 * Return: 0 on success. Otherwise, an error code is returned. 671 */ 672 static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, 673 gfp_t gfp) 674 { 675 if (mem_cgroup_disabled()) 676 return 0; 677 return __mem_cgroup_charge(folio, mm, gfp); 678 } 679 680 int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, gfp_t gfp, 681 long nr_pages); 682 683 int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm, 684 gfp_t gfp, swp_entry_t entry); 685 void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry); 686 687 void __mem_cgroup_uncharge(struct folio *folio); 688 689 /** 690 * mem_cgroup_uncharge - Uncharge a folio. 691 * @folio: Folio to uncharge. 692 * 693 * Uncharge a folio previously charged with mem_cgroup_charge(). 694 */ 695 static inline void mem_cgroup_uncharge(struct folio *folio) 696 { 697 if (mem_cgroup_disabled()) 698 return; 699 __mem_cgroup_uncharge(folio); 700 } 701 702 void __mem_cgroup_uncharge_folios(struct folio_batch *folios); 703 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 704 { 705 if (mem_cgroup_disabled()) 706 return; 707 __mem_cgroup_uncharge_folios(folios); 708 } 709 710 void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages); 711 void mem_cgroup_replace_folio(struct folio *old, struct folio *new); 712 void mem_cgroup_migrate(struct folio *old, struct folio *new); 713 714 /** 715 * mem_cgroup_lruvec - get the lru list vector for a memcg & node 716 * @memcg: memcg of the wanted lruvec 717 * @pgdat: pglist_data 718 * 719 * Returns the lru list vector holding pages for a given @memcg & 720 * @pgdat combination. This can be the node lruvec, if the memory 721 * controller is disabled. 722 */ 723 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 724 struct pglist_data *pgdat) 725 { 726 struct mem_cgroup_per_node *mz; 727 struct lruvec *lruvec; 728 729 if (mem_cgroup_disabled()) { 730 lruvec = &pgdat->__lruvec; 731 goto out; 732 } 733 734 if (!memcg) 735 memcg = root_mem_cgroup; 736 737 mz = memcg->nodeinfo[pgdat->node_id]; 738 lruvec = &mz->lruvec; 739 out: 740 /* 741 * Since a node can be onlined after the mem_cgroup was created, 742 * we have to be prepared to initialize lruvec->pgdat here; 743 * and if offlined then reonlined, we need to reinitialize it. 744 */ 745 if (unlikely(lruvec->pgdat != pgdat)) 746 lruvec->pgdat = pgdat; 747 return lruvec; 748 } 749 750 /** 751 * folio_lruvec - return lruvec for isolating/putting an LRU folio 752 * @folio: Pointer to the folio. 753 * 754 * This function relies on folio->mem_cgroup being stable. 755 */ 756 static inline struct lruvec *folio_lruvec(struct folio *folio) 757 { 758 struct mem_cgroup *memcg = folio_memcg(folio); 759 760 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio); 761 return mem_cgroup_lruvec(memcg, folio_pgdat(folio)); 762 } 763 764 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); 765 766 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); 767 768 struct mem_cgroup *get_mem_cgroup_from_current(void); 769 770 struct lruvec *folio_lruvec_lock(struct folio *folio); 771 struct lruvec *folio_lruvec_lock_irq(struct folio *folio); 772 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 773 unsigned long *flags); 774 775 #ifdef CONFIG_DEBUG_VM 776 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio); 777 #else 778 static inline 779 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 780 { 781 } 782 #endif 783 784 static inline 785 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ 786 return css ? container_of(css, struct mem_cgroup, css) : NULL; 787 } 788 789 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) 790 { 791 return percpu_ref_tryget(&objcg->refcnt); 792 } 793 794 static inline void obj_cgroup_get(struct obj_cgroup *objcg) 795 { 796 percpu_ref_get(&objcg->refcnt); 797 } 798 799 static inline void obj_cgroup_get_many(struct obj_cgroup *objcg, 800 unsigned long nr) 801 { 802 percpu_ref_get_many(&objcg->refcnt, nr); 803 } 804 805 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 806 { 807 if (objcg) 808 percpu_ref_put(&objcg->refcnt); 809 } 810 811 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 812 { 813 return !memcg || css_tryget(&memcg->css); 814 } 815 816 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 817 { 818 return !memcg || css_tryget_online(&memcg->css); 819 } 820 821 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 822 { 823 if (memcg) 824 css_put(&memcg->css); 825 } 826 827 #define mem_cgroup_from_counter(counter, member) \ 828 container_of(counter, struct mem_cgroup, member) 829 830 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, 831 struct mem_cgroup *, 832 struct mem_cgroup_reclaim_cookie *); 833 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); 834 void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 835 int (*)(struct task_struct *, void *), void *arg); 836 837 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 838 { 839 if (mem_cgroup_disabled()) 840 return 0; 841 842 return memcg->id.id; 843 } 844 struct mem_cgroup *mem_cgroup_from_id(unsigned short id); 845 846 #ifdef CONFIG_SHRINKER_DEBUG 847 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 848 { 849 return memcg ? cgroup_ino(memcg->css.cgroup) : 0; 850 } 851 852 struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino); 853 #endif 854 855 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 856 { 857 return mem_cgroup_from_css(seq_css(m)); 858 } 859 860 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 861 { 862 struct mem_cgroup_per_node *mz; 863 864 if (mem_cgroup_disabled()) 865 return NULL; 866 867 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 868 return mz->memcg; 869 } 870 871 /** 872 * parent_mem_cgroup - find the accounting parent of a memcg 873 * @memcg: memcg whose parent to find 874 * 875 * Returns the parent memcg, or NULL if this is the root. 876 */ 877 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 878 { 879 return mem_cgroup_from_css(memcg->css.parent); 880 } 881 882 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, 883 struct mem_cgroup *root) 884 { 885 if (root == memcg) 886 return true; 887 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); 888 } 889 890 static inline bool mm_match_cgroup(struct mm_struct *mm, 891 struct mem_cgroup *memcg) 892 { 893 struct mem_cgroup *task_memcg; 894 bool match = false; 895 896 rcu_read_lock(); 897 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 898 if (task_memcg) 899 match = mem_cgroup_is_descendant(task_memcg, memcg); 900 rcu_read_unlock(); 901 return match; 902 } 903 904 struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio); 905 ino_t page_cgroup_ino(struct page *page); 906 907 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 908 { 909 if (mem_cgroup_disabled()) 910 return true; 911 return !!(memcg->css.flags & CSS_ONLINE); 912 } 913 914 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, 915 int zid, int nr_pages); 916 917 static inline 918 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 919 enum lru_list lru, int zone_idx) 920 { 921 struct mem_cgroup_per_node *mz; 922 923 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 924 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); 925 } 926 927 void mem_cgroup_handle_over_high(gfp_t gfp_mask); 928 929 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); 930 931 unsigned long mem_cgroup_size(struct mem_cgroup *memcg); 932 933 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, 934 struct task_struct *p); 935 936 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); 937 938 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, 939 struct mem_cgroup *oom_domain); 940 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); 941 942 void __mod_memcg_state(struct mem_cgroup *memcg, enum memcg_stat_item idx, 943 int val); 944 945 /* idx can be of type enum memcg_stat_item or node_stat_item */ 946 static inline void mod_memcg_state(struct mem_cgroup *memcg, 947 enum memcg_stat_item idx, int val) 948 { 949 unsigned long flags; 950 951 local_irq_save(flags); 952 __mod_memcg_state(memcg, idx, val); 953 local_irq_restore(flags); 954 } 955 956 static inline void mod_memcg_page_state(struct page *page, 957 enum memcg_stat_item idx, int val) 958 { 959 struct mem_cgroup *memcg; 960 961 if (mem_cgroup_disabled()) 962 return; 963 964 rcu_read_lock(); 965 memcg = folio_memcg(page_folio(page)); 966 if (memcg) 967 mod_memcg_state(memcg, idx, val); 968 rcu_read_unlock(); 969 } 970 971 unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx); 972 unsigned long lruvec_page_state(struct lruvec *lruvec, enum node_stat_item idx); 973 unsigned long lruvec_page_state_local(struct lruvec *lruvec, 974 enum node_stat_item idx); 975 976 void mem_cgroup_flush_stats(struct mem_cgroup *memcg); 977 void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg); 978 979 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val); 980 981 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 982 int val) 983 { 984 unsigned long flags; 985 986 local_irq_save(flags); 987 __mod_lruvec_kmem_state(p, idx, val); 988 local_irq_restore(flags); 989 } 990 991 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, 992 unsigned long count); 993 994 static inline void count_memcg_events(struct mem_cgroup *memcg, 995 enum vm_event_item idx, 996 unsigned long count) 997 { 998 unsigned long flags; 999 1000 local_irq_save(flags); 1001 __count_memcg_events(memcg, idx, count); 1002 local_irq_restore(flags); 1003 } 1004 1005 static inline void count_memcg_folio_events(struct folio *folio, 1006 enum vm_event_item idx, unsigned long nr) 1007 { 1008 struct mem_cgroup *memcg = folio_memcg(folio); 1009 1010 if (memcg) 1011 count_memcg_events(memcg, idx, nr); 1012 } 1013 1014 static inline void count_memcg_event_mm(struct mm_struct *mm, 1015 enum vm_event_item idx) 1016 { 1017 struct mem_cgroup *memcg; 1018 1019 if (mem_cgroup_disabled()) 1020 return; 1021 1022 rcu_read_lock(); 1023 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1024 if (likely(memcg)) 1025 count_memcg_events(memcg, idx, 1); 1026 rcu_read_unlock(); 1027 } 1028 1029 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1030 enum memcg_memory_event event) 1031 { 1032 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX || 1033 event == MEMCG_SWAP_FAIL; 1034 1035 atomic_long_inc(&memcg->memory_events_local[event]); 1036 if (!swap_event) 1037 cgroup_file_notify(&memcg->events_local_file); 1038 1039 do { 1040 atomic_long_inc(&memcg->memory_events[event]); 1041 if (swap_event) 1042 cgroup_file_notify(&memcg->swap_events_file); 1043 else 1044 cgroup_file_notify(&memcg->events_file); 1045 1046 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1047 break; 1048 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) 1049 break; 1050 } while ((memcg = parent_mem_cgroup(memcg)) && 1051 !mem_cgroup_is_root(memcg)); 1052 } 1053 1054 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1055 enum memcg_memory_event event) 1056 { 1057 struct mem_cgroup *memcg; 1058 1059 if (mem_cgroup_disabled()) 1060 return; 1061 1062 rcu_read_lock(); 1063 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1064 if (likely(memcg)) 1065 memcg_memory_event(memcg, event); 1066 rcu_read_unlock(); 1067 } 1068 1069 void split_page_memcg(struct page *head, int old_order, int new_order); 1070 1071 #else /* CONFIG_MEMCG */ 1072 1073 #define MEM_CGROUP_ID_SHIFT 0 1074 1075 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 1076 { 1077 return NULL; 1078 } 1079 1080 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 1081 { 1082 WARN_ON_ONCE(!rcu_read_lock_held()); 1083 return NULL; 1084 } 1085 1086 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 1087 { 1088 return NULL; 1089 } 1090 1091 static inline struct mem_cgroup *page_memcg_check(struct page *page) 1092 { 1093 return NULL; 1094 } 1095 1096 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 1097 { 1098 return NULL; 1099 } 1100 1101 static inline bool folio_memcg_kmem(struct folio *folio) 1102 { 1103 return false; 1104 } 1105 1106 static inline bool PageMemcgKmem(struct page *page) 1107 { 1108 return false; 1109 } 1110 1111 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 1112 { 1113 return true; 1114 } 1115 1116 static inline bool mem_cgroup_disabled(void) 1117 { 1118 return true; 1119 } 1120 1121 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1122 enum memcg_memory_event event) 1123 { 1124 } 1125 1126 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1127 enum memcg_memory_event event) 1128 { 1129 } 1130 1131 static inline void mem_cgroup_protection(struct mem_cgroup *root, 1132 struct mem_cgroup *memcg, 1133 unsigned long *min, 1134 unsigned long *low) 1135 { 1136 *min = *low = 0; 1137 } 1138 1139 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, 1140 struct mem_cgroup *memcg) 1141 { 1142 } 1143 1144 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 1145 struct mem_cgroup *memcg) 1146 { 1147 return true; 1148 } 1149 static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 1150 struct mem_cgroup *memcg) 1151 { 1152 return false; 1153 } 1154 1155 static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 1156 struct mem_cgroup *memcg) 1157 { 1158 return false; 1159 } 1160 1161 static inline void mem_cgroup_commit_charge(struct folio *folio, 1162 struct mem_cgroup *memcg) 1163 { 1164 } 1165 1166 static inline int mem_cgroup_charge(struct folio *folio, 1167 struct mm_struct *mm, gfp_t gfp) 1168 { 1169 return 0; 1170 } 1171 1172 static inline int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, 1173 gfp_t gfp, long nr_pages) 1174 { 1175 return 0; 1176 } 1177 1178 static inline int mem_cgroup_swapin_charge_folio(struct folio *folio, 1179 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry) 1180 { 1181 return 0; 1182 } 1183 1184 static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry) 1185 { 1186 } 1187 1188 static inline void mem_cgroup_uncharge(struct folio *folio) 1189 { 1190 } 1191 1192 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 1193 { 1194 } 1195 1196 static inline void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, 1197 unsigned int nr_pages) 1198 { 1199 } 1200 1201 static inline void mem_cgroup_replace_folio(struct folio *old, 1202 struct folio *new) 1203 { 1204 } 1205 1206 static inline void mem_cgroup_migrate(struct folio *old, struct folio *new) 1207 { 1208 } 1209 1210 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 1211 struct pglist_data *pgdat) 1212 { 1213 return &pgdat->__lruvec; 1214 } 1215 1216 static inline struct lruvec *folio_lruvec(struct folio *folio) 1217 { 1218 struct pglist_data *pgdat = folio_pgdat(folio); 1219 return &pgdat->__lruvec; 1220 } 1221 1222 static inline 1223 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 1224 { 1225 } 1226 1227 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 1228 { 1229 return NULL; 1230 } 1231 1232 static inline bool mm_match_cgroup(struct mm_struct *mm, 1233 struct mem_cgroup *memcg) 1234 { 1235 return true; 1236 } 1237 1238 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) 1239 { 1240 return NULL; 1241 } 1242 1243 static inline struct mem_cgroup *get_mem_cgroup_from_current(void) 1244 { 1245 return NULL; 1246 } 1247 1248 static inline 1249 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css) 1250 { 1251 return NULL; 1252 } 1253 1254 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 1255 { 1256 } 1257 1258 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 1259 { 1260 return true; 1261 } 1262 1263 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 1264 { 1265 return true; 1266 } 1267 1268 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 1269 { 1270 } 1271 1272 static inline struct lruvec *folio_lruvec_lock(struct folio *folio) 1273 { 1274 struct pglist_data *pgdat = folio_pgdat(folio); 1275 1276 spin_lock(&pgdat->__lruvec.lru_lock); 1277 return &pgdat->__lruvec; 1278 } 1279 1280 static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio) 1281 { 1282 struct pglist_data *pgdat = folio_pgdat(folio); 1283 1284 spin_lock_irq(&pgdat->__lruvec.lru_lock); 1285 return &pgdat->__lruvec; 1286 } 1287 1288 static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 1289 unsigned long *flagsp) 1290 { 1291 struct pglist_data *pgdat = folio_pgdat(folio); 1292 1293 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp); 1294 return &pgdat->__lruvec; 1295 } 1296 1297 static inline struct mem_cgroup * 1298 mem_cgroup_iter(struct mem_cgroup *root, 1299 struct mem_cgroup *prev, 1300 struct mem_cgroup_reclaim_cookie *reclaim) 1301 { 1302 return NULL; 1303 } 1304 1305 static inline void mem_cgroup_iter_break(struct mem_cgroup *root, 1306 struct mem_cgroup *prev) 1307 { 1308 } 1309 1310 static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 1311 int (*fn)(struct task_struct *, void *), void *arg) 1312 { 1313 } 1314 1315 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 1316 { 1317 return 0; 1318 } 1319 1320 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) 1321 { 1322 WARN_ON_ONCE(id); 1323 /* XXX: This should always return root_mem_cgroup */ 1324 return NULL; 1325 } 1326 1327 #ifdef CONFIG_SHRINKER_DEBUG 1328 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 1329 { 1330 return 0; 1331 } 1332 1333 static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino) 1334 { 1335 return NULL; 1336 } 1337 #endif 1338 1339 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 1340 { 1341 return NULL; 1342 } 1343 1344 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 1345 { 1346 return NULL; 1347 } 1348 1349 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 1350 { 1351 return true; 1352 } 1353 1354 static inline 1355 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 1356 enum lru_list lru, int zone_idx) 1357 { 1358 return 0; 1359 } 1360 1361 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) 1362 { 1363 return 0; 1364 } 1365 1366 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) 1367 { 1368 return 0; 1369 } 1370 1371 static inline void 1372 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) 1373 { 1374 } 1375 1376 static inline void 1377 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) 1378 { 1379 } 1380 1381 static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask) 1382 { 1383 } 1384 1385 static inline struct mem_cgroup *mem_cgroup_get_oom_group( 1386 struct task_struct *victim, struct mem_cgroup *oom_domain) 1387 { 1388 return NULL; 1389 } 1390 1391 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) 1392 { 1393 } 1394 1395 static inline void __mod_memcg_state(struct mem_cgroup *memcg, 1396 enum memcg_stat_item idx, 1397 int nr) 1398 { 1399 } 1400 1401 static inline void mod_memcg_state(struct mem_cgroup *memcg, 1402 enum memcg_stat_item idx, 1403 int nr) 1404 { 1405 } 1406 1407 static inline void mod_memcg_page_state(struct page *page, 1408 enum memcg_stat_item idx, int val) 1409 { 1410 } 1411 1412 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) 1413 { 1414 return 0; 1415 } 1416 1417 static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1418 enum node_stat_item idx) 1419 { 1420 return node_page_state(lruvec_pgdat(lruvec), idx); 1421 } 1422 1423 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1424 enum node_stat_item idx) 1425 { 1426 return node_page_state(lruvec_pgdat(lruvec), idx); 1427 } 1428 1429 static inline void mem_cgroup_flush_stats(struct mem_cgroup *memcg) 1430 { 1431 } 1432 1433 static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg) 1434 { 1435 } 1436 1437 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1438 int val) 1439 { 1440 struct page *page = virt_to_head_page(p); 1441 1442 __mod_node_page_state(page_pgdat(page), idx, val); 1443 } 1444 1445 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1446 int val) 1447 { 1448 struct page *page = virt_to_head_page(p); 1449 1450 mod_node_page_state(page_pgdat(page), idx, val); 1451 } 1452 1453 static inline void count_memcg_events(struct mem_cgroup *memcg, 1454 enum vm_event_item idx, 1455 unsigned long count) 1456 { 1457 } 1458 1459 static inline void __count_memcg_events(struct mem_cgroup *memcg, 1460 enum vm_event_item idx, 1461 unsigned long count) 1462 { 1463 } 1464 1465 static inline void count_memcg_folio_events(struct folio *folio, 1466 enum vm_event_item idx, unsigned long nr) 1467 { 1468 } 1469 1470 static inline 1471 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) 1472 { 1473 } 1474 1475 static inline void split_page_memcg(struct page *head, int old_order, int new_order) 1476 { 1477 } 1478 #endif /* CONFIG_MEMCG */ 1479 1480 /* 1481 * Extended information for slab objects stored as an array in page->memcg_data 1482 * if MEMCG_DATA_OBJEXTS is set. 1483 */ 1484 struct slabobj_ext { 1485 #ifdef CONFIG_MEMCG 1486 struct obj_cgroup *objcg; 1487 #endif 1488 #ifdef CONFIG_MEM_ALLOC_PROFILING 1489 union codetag_ref ref; 1490 #endif 1491 } __aligned(8); 1492 1493 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx) 1494 { 1495 __mod_lruvec_kmem_state(p, idx, 1); 1496 } 1497 1498 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx) 1499 { 1500 __mod_lruvec_kmem_state(p, idx, -1); 1501 } 1502 1503 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) 1504 { 1505 struct mem_cgroup *memcg; 1506 1507 memcg = lruvec_memcg(lruvec); 1508 if (!memcg) 1509 return NULL; 1510 memcg = parent_mem_cgroup(memcg); 1511 if (!memcg) 1512 return NULL; 1513 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); 1514 } 1515 1516 static inline void unlock_page_lruvec(struct lruvec *lruvec) 1517 { 1518 spin_unlock(&lruvec->lru_lock); 1519 } 1520 1521 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec) 1522 { 1523 spin_unlock_irq(&lruvec->lru_lock); 1524 } 1525 1526 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec, 1527 unsigned long flags) 1528 { 1529 spin_unlock_irqrestore(&lruvec->lru_lock, flags); 1530 } 1531 1532 /* Test requires a stable folio->memcg binding, see folio_memcg() */ 1533 static inline bool folio_matches_lruvec(struct folio *folio, 1534 struct lruvec *lruvec) 1535 { 1536 return lruvec_pgdat(lruvec) == folio_pgdat(folio) && 1537 lruvec_memcg(lruvec) == folio_memcg(folio); 1538 } 1539 1540 /* Don't lock again iff page's lruvec locked */ 1541 static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio, 1542 struct lruvec *locked_lruvec) 1543 { 1544 if (locked_lruvec) { 1545 if (folio_matches_lruvec(folio, locked_lruvec)) 1546 return locked_lruvec; 1547 1548 unlock_page_lruvec_irq(locked_lruvec); 1549 } 1550 1551 return folio_lruvec_lock_irq(folio); 1552 } 1553 1554 /* Don't lock again iff folio's lruvec locked */ 1555 static inline void folio_lruvec_relock_irqsave(struct folio *folio, 1556 struct lruvec **lruvecp, unsigned long *flags) 1557 { 1558 if (*lruvecp) { 1559 if (folio_matches_lruvec(folio, *lruvecp)) 1560 return; 1561 1562 unlock_page_lruvec_irqrestore(*lruvecp, *flags); 1563 } 1564 1565 *lruvecp = folio_lruvec_lock_irqsave(folio, flags); 1566 } 1567 1568 #ifdef CONFIG_CGROUP_WRITEBACK 1569 1570 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); 1571 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, 1572 unsigned long *pheadroom, unsigned long *pdirty, 1573 unsigned long *pwriteback); 1574 1575 void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio, 1576 struct bdi_writeback *wb); 1577 1578 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1579 struct bdi_writeback *wb) 1580 { 1581 struct mem_cgroup *memcg; 1582 1583 if (mem_cgroup_disabled()) 1584 return; 1585 1586 memcg = folio_memcg(folio); 1587 if (unlikely(memcg && &memcg->css != wb->memcg_css)) 1588 mem_cgroup_track_foreign_dirty_slowpath(folio, wb); 1589 } 1590 1591 void mem_cgroup_flush_foreign(struct bdi_writeback *wb); 1592 1593 #else /* CONFIG_CGROUP_WRITEBACK */ 1594 1595 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) 1596 { 1597 return NULL; 1598 } 1599 1600 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, 1601 unsigned long *pfilepages, 1602 unsigned long *pheadroom, 1603 unsigned long *pdirty, 1604 unsigned long *pwriteback) 1605 { 1606 } 1607 1608 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1609 struct bdi_writeback *wb) 1610 { 1611 } 1612 1613 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) 1614 { 1615 } 1616 1617 #endif /* CONFIG_CGROUP_WRITEBACK */ 1618 1619 struct sock; 1620 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages, 1621 gfp_t gfp_mask); 1622 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); 1623 #ifdef CONFIG_MEMCG 1624 extern struct static_key_false memcg_sockets_enabled_key; 1625 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) 1626 void mem_cgroup_sk_alloc(struct sock *sk); 1627 void mem_cgroup_sk_free(struct sock *sk); 1628 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1629 { 1630 #ifdef CONFIG_MEMCG_V1 1631 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1632 return !!memcg->tcpmem_pressure; 1633 #endif /* CONFIG_MEMCG_V1 */ 1634 do { 1635 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure))) 1636 return true; 1637 } while ((memcg = parent_mem_cgroup(memcg))); 1638 return false; 1639 } 1640 1641 int alloc_shrinker_info(struct mem_cgroup *memcg); 1642 void free_shrinker_info(struct mem_cgroup *memcg); 1643 void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id); 1644 void reparent_shrinker_deferred(struct mem_cgroup *memcg); 1645 #else 1646 #define mem_cgroup_sockets_enabled 0 1647 static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; 1648 static inline void mem_cgroup_sk_free(struct sock *sk) { }; 1649 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1650 { 1651 return false; 1652 } 1653 1654 static inline void set_shrinker_bit(struct mem_cgroup *memcg, 1655 int nid, int shrinker_id) 1656 { 1657 } 1658 #endif 1659 1660 #ifdef CONFIG_MEMCG 1661 bool mem_cgroup_kmem_disabled(void); 1662 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); 1663 void __memcg_kmem_uncharge_page(struct page *page, int order); 1664 1665 /* 1666 * The returned objcg pointer is safe to use without additional 1667 * protection within a scope. The scope is defined either by 1668 * the current task (similar to the "current" global variable) 1669 * or by set_active_memcg() pair. 1670 * Please, use obj_cgroup_get() to get a reference if the pointer 1671 * needs to be used outside of the local scope. 1672 */ 1673 struct obj_cgroup *current_obj_cgroup(void); 1674 struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio); 1675 1676 static inline struct obj_cgroup *get_obj_cgroup_from_current(void) 1677 { 1678 struct obj_cgroup *objcg = current_obj_cgroup(); 1679 1680 if (objcg) 1681 obj_cgroup_get(objcg); 1682 1683 return objcg; 1684 } 1685 1686 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); 1687 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); 1688 1689 extern struct static_key_false memcg_bpf_enabled_key; 1690 static inline bool memcg_bpf_enabled(void) 1691 { 1692 return static_branch_likely(&memcg_bpf_enabled_key); 1693 } 1694 1695 extern struct static_key_false memcg_kmem_online_key; 1696 1697 static inline bool memcg_kmem_online(void) 1698 { 1699 return static_branch_likely(&memcg_kmem_online_key); 1700 } 1701 1702 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1703 int order) 1704 { 1705 if (memcg_kmem_online()) 1706 return __memcg_kmem_charge_page(page, gfp, order); 1707 return 0; 1708 } 1709 1710 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1711 { 1712 if (memcg_kmem_online()) 1713 __memcg_kmem_uncharge_page(page, order); 1714 } 1715 1716 /* 1717 * A helper for accessing memcg's kmem_id, used for getting 1718 * corresponding LRU lists. 1719 */ 1720 static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1721 { 1722 return memcg ? memcg->kmemcg_id : -1; 1723 } 1724 1725 struct mem_cgroup *mem_cgroup_from_slab_obj(void *p); 1726 1727 static inline void count_objcg_event(struct obj_cgroup *objcg, 1728 enum vm_event_item idx) 1729 { 1730 struct mem_cgroup *memcg; 1731 1732 if (!memcg_kmem_online()) 1733 return; 1734 1735 rcu_read_lock(); 1736 memcg = obj_cgroup_memcg(objcg); 1737 count_memcg_events(memcg, idx, 1); 1738 rcu_read_unlock(); 1739 } 1740 1741 #else 1742 static inline bool mem_cgroup_kmem_disabled(void) 1743 { 1744 return true; 1745 } 1746 1747 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1748 int order) 1749 { 1750 return 0; 1751 } 1752 1753 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1754 { 1755 } 1756 1757 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1758 int order) 1759 { 1760 return 0; 1761 } 1762 1763 static inline void __memcg_kmem_uncharge_page(struct page *page, int order) 1764 { 1765 } 1766 1767 static inline struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio) 1768 { 1769 return NULL; 1770 } 1771 1772 static inline bool memcg_bpf_enabled(void) 1773 { 1774 return false; 1775 } 1776 1777 static inline bool memcg_kmem_online(void) 1778 { 1779 return false; 1780 } 1781 1782 static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1783 { 1784 return -1; 1785 } 1786 1787 static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p) 1788 { 1789 return NULL; 1790 } 1791 1792 static inline void count_objcg_event(struct obj_cgroup *objcg, 1793 enum vm_event_item idx) 1794 { 1795 } 1796 1797 #endif /* CONFIG_MEMCG */ 1798 1799 #if defined(CONFIG_MEMCG) && defined(CONFIG_ZSWAP) 1800 bool obj_cgroup_may_zswap(struct obj_cgroup *objcg); 1801 void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size); 1802 void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size); 1803 bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg); 1804 #else 1805 static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg) 1806 { 1807 return true; 1808 } 1809 static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, 1810 size_t size) 1811 { 1812 } 1813 static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, 1814 size_t size) 1815 { 1816 } 1817 static inline bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg) 1818 { 1819 /* if zswap is disabled, do not block pages going to the swapping device */ 1820 return true; 1821 } 1822 #endif 1823 1824 1825 /* Cgroup v1-related declarations */ 1826 1827 #ifdef CONFIG_MEMCG_V1 1828 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order, 1829 gfp_t gfp_mask, 1830 unsigned long *total_scanned); 1831 1832 bool mem_cgroup_oom_synchronize(bool wait); 1833 1834 static inline bool task_in_memcg_oom(struct task_struct *p) 1835 { 1836 return p->memcg_in_oom; 1837 } 1838 1839 void folio_memcg_lock(struct folio *folio); 1840 void folio_memcg_unlock(struct folio *folio); 1841 1842 /* try to stablize folio_memcg() for all the pages in a memcg */ 1843 static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg) 1844 { 1845 rcu_read_lock(); 1846 1847 if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account)) 1848 return true; 1849 1850 rcu_read_unlock(); 1851 return false; 1852 } 1853 1854 static inline void mem_cgroup_unlock_pages(void) 1855 { 1856 rcu_read_unlock(); 1857 } 1858 1859 static inline void mem_cgroup_enter_user_fault(void) 1860 { 1861 WARN_ON(current->in_user_fault); 1862 current->in_user_fault = 1; 1863 } 1864 1865 static inline void mem_cgroup_exit_user_fault(void) 1866 { 1867 WARN_ON(!current->in_user_fault); 1868 current->in_user_fault = 0; 1869 } 1870 1871 #else /* CONFIG_MEMCG_V1 */ 1872 static inline 1873 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order, 1874 gfp_t gfp_mask, 1875 unsigned long *total_scanned) 1876 { 1877 return 0; 1878 } 1879 1880 static inline void folio_memcg_lock(struct folio *folio) 1881 { 1882 } 1883 1884 static inline void folio_memcg_unlock(struct folio *folio) 1885 { 1886 } 1887 1888 static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg) 1889 { 1890 /* to match folio_memcg_rcu() */ 1891 rcu_read_lock(); 1892 return true; 1893 } 1894 1895 static inline void mem_cgroup_unlock_pages(void) 1896 { 1897 rcu_read_unlock(); 1898 } 1899 1900 static inline bool task_in_memcg_oom(struct task_struct *p) 1901 { 1902 return false; 1903 } 1904 1905 static inline bool mem_cgroup_oom_synchronize(bool wait) 1906 { 1907 return false; 1908 } 1909 1910 static inline void mem_cgroup_enter_user_fault(void) 1911 { 1912 } 1913 1914 static inline void mem_cgroup_exit_user_fault(void) 1915 { 1916 } 1917 1918 #endif /* CONFIG_MEMCG_V1 */ 1919 1920 #endif /* _LINUX_MEMCONTROL_H */ 1921