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 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 memcg1_events_percpu; 74 struct memcg_vmstats; 75 struct lruvec_stats_percpu; 76 struct lruvec_stats; 77 78 struct mem_cgroup_reclaim_iter { 79 struct mem_cgroup *position; 80 /* scan generation, increased every round-trip */ 81 atomic_t generation; 82 }; 83 84 /* 85 * per-node information in memory controller. 86 */ 87 struct mem_cgroup_per_node { 88 /* Keep the read-only fields at the start */ 89 struct mem_cgroup *memcg; /* Back pointer, we cannot */ 90 /* use container_of */ 91 92 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu; 93 struct lruvec_stats *lruvec_stats; 94 struct shrinker_info __rcu *shrinker_info; 95 96 #ifdef CONFIG_MEMCG_V1 97 /* 98 * Memcg-v1 only stuff in middle as buffer between read mostly fields 99 * and update often fields to avoid false sharing. If v1 stuff is 100 * not present, an explicit padding is needed. 101 */ 102 103 struct rb_node tree_node; /* RB tree node */ 104 unsigned long usage_in_excess;/* Set to the value by which */ 105 /* the soft limit is exceeded*/ 106 bool on_tree; 107 #else 108 CACHELINE_PADDING(_pad1_); 109 #endif 110 111 /* Fields which get updated often at the end. */ 112 struct lruvec lruvec; 113 CACHELINE_PADDING(_pad2_); 114 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; 115 struct mem_cgroup_reclaim_iter iter; 116 }; 117 118 struct mem_cgroup_threshold { 119 struct eventfd_ctx *eventfd; 120 unsigned long threshold; 121 }; 122 123 /* For threshold */ 124 struct mem_cgroup_threshold_ary { 125 /* An array index points to threshold just below or equal to usage. */ 126 int current_threshold; 127 /* Size of entries[] */ 128 unsigned int size; 129 /* Array of thresholds */ 130 struct mem_cgroup_threshold entries[] __counted_by(size); 131 }; 132 133 struct mem_cgroup_thresholds { 134 /* Primary thresholds array */ 135 struct mem_cgroup_threshold_ary *primary; 136 /* 137 * Spare threshold array. 138 * This is needed to make mem_cgroup_unregister_event() "never fail". 139 * It must be able to store at least primary->size - 1 entries. 140 */ 141 struct mem_cgroup_threshold_ary *spare; 142 }; 143 144 /* 145 * Remember four most recent foreign writebacks with dirty pages in this 146 * cgroup. Inode sharing is expected to be uncommon and, even if we miss 147 * one in a given round, we're likely to catch it later if it keeps 148 * foreign-dirtying, so a fairly low count should be enough. 149 * 150 * See mem_cgroup_track_foreign_dirty_slowpath() for details. 151 */ 152 #define MEMCG_CGWB_FRN_CNT 4 153 154 struct memcg_cgwb_frn { 155 u64 bdi_id; /* bdi->id of the foreign inode */ 156 int memcg_id; /* memcg->css.id of foreign inode */ 157 u64 at; /* jiffies_64 at the time of dirtying */ 158 struct wb_completion done; /* tracks in-flight foreign writebacks */ 159 }; 160 161 /* 162 * Bucket for arbitrarily byte-sized objects charged to a memory 163 * cgroup. The bucket can be reparented in one piece when the cgroup 164 * is destroyed, without having to round up the individual references 165 * of all live memory objects in the wild. 166 */ 167 struct obj_cgroup { 168 struct percpu_ref refcnt; 169 struct mem_cgroup *memcg; 170 atomic_t nr_charged_bytes; 171 union { 172 struct list_head list; /* protected by objcg_lock */ 173 struct rcu_head rcu; 174 }; 175 }; 176 177 /* 178 * The memory controller data structure. The memory controller controls both 179 * page cache and RSS per cgroup. We would eventually like to provide 180 * statistics based on the statistics developed by Rik Van Riel for clock-pro, 181 * to help the administrator determine what knobs to tune. 182 */ 183 struct mem_cgroup { 184 struct cgroup_subsys_state css; 185 186 /* Private memcg ID. Used to ID objects that outlive the cgroup */ 187 struct mem_cgroup_id id; 188 189 /* Accounted resources */ 190 struct page_counter memory; /* Both v1 & v2 */ 191 192 union { 193 struct page_counter swap; /* v2 only */ 194 struct page_counter memsw; /* v1 only */ 195 }; 196 197 /* registered local peak watchers */ 198 struct list_head memory_peaks; 199 struct list_head swap_peaks; 200 spinlock_t peaks_lock; 201 202 /* Range enforcement for interrupt charges */ 203 struct work_struct high_work; 204 205 #ifdef CONFIG_ZSWAP 206 unsigned long zswap_max; 207 208 /* 209 * Prevent pages from this memcg from being written back from zswap to 210 * swap, and from being swapped out on zswap store failures. 211 */ 212 bool zswap_writeback; 213 #endif 214 215 /* vmpressure notifications */ 216 struct vmpressure vmpressure; 217 218 /* 219 * Should the OOM killer kill all belonging tasks, had it kill one? 220 */ 221 bool oom_group; 222 223 int swappiness; 224 225 /* memory.events and memory.events.local */ 226 struct cgroup_file events_file; 227 struct cgroup_file events_local_file; 228 229 /* handle for "memory.swap.events" */ 230 struct cgroup_file swap_events_file; 231 232 /* memory.stat */ 233 struct memcg_vmstats *vmstats; 234 235 /* memory.events */ 236 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS]; 237 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS]; 238 239 /* 240 * Hint of reclaim pressure for socket memroy management. Note 241 * that this indicator should NOT be used in legacy cgroup mode 242 * where socket memory is accounted/charged separately. 243 */ 244 unsigned long socket_pressure; 245 246 int kmemcg_id; 247 /* 248 * memcg->objcg is wiped out as a part of the objcg repaprenting 249 * process. memcg->orig_objcg preserves a pointer (and a reference) 250 * to the original objcg until the end of live of memcg. 251 */ 252 struct obj_cgroup __rcu *objcg; 253 struct obj_cgroup *orig_objcg; 254 /* list of inherited objcgs, protected by objcg_lock */ 255 struct list_head objcg_list; 256 257 struct memcg_vmstats_percpu __percpu *vmstats_percpu; 258 259 #ifdef CONFIG_CGROUP_WRITEBACK 260 struct list_head cgwb_list; 261 struct wb_domain cgwb_domain; 262 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT]; 263 #endif 264 265 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 266 struct deferred_split deferred_split_queue; 267 #endif 268 269 #ifdef CONFIG_LRU_GEN_WALKS_MMU 270 /* per-memcg mm_struct list */ 271 struct lru_gen_mm_list mm_list; 272 #endif 273 274 #ifdef CONFIG_MEMCG_V1 275 /* Legacy consumer-oriented counters */ 276 struct page_counter kmem; /* v1 only */ 277 struct page_counter tcpmem; /* v1 only */ 278 279 struct memcg1_events_percpu __percpu *events_percpu; 280 281 unsigned long soft_limit; 282 283 /* protected by memcg_oom_lock */ 284 bool oom_lock; 285 int under_oom; 286 287 /* OOM-Killer disable */ 288 int oom_kill_disable; 289 290 /* protect arrays of thresholds */ 291 struct mutex thresholds_lock; 292 293 /* thresholds for memory usage. RCU-protected */ 294 struct mem_cgroup_thresholds thresholds; 295 296 /* thresholds for mem+swap usage. RCU-protected */ 297 struct mem_cgroup_thresholds memsw_thresholds; 298 299 /* For oom notifier event fd */ 300 struct list_head oom_notify; 301 302 /* Legacy tcp memory accounting */ 303 bool tcpmem_active; 304 int tcpmem_pressure; 305 306 /* List of events which userspace want to receive */ 307 struct list_head event_list; 308 spinlock_t event_list_lock; 309 #endif /* CONFIG_MEMCG_V1 */ 310 311 struct mem_cgroup_per_node *nodeinfo[]; 312 }; 313 314 /* 315 * size of first charge trial. 316 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the 317 * workload. 318 */ 319 #define MEMCG_CHARGE_BATCH 64U 320 321 extern struct mem_cgroup *root_mem_cgroup; 322 323 enum page_memcg_data_flags { 324 /* page->memcg_data is a pointer to an slabobj_ext vector */ 325 MEMCG_DATA_OBJEXTS = (1UL << 0), 326 /* page has been accounted as a non-slab kernel page */ 327 MEMCG_DATA_KMEM = (1UL << 1), 328 /* the next bit after the last actual flag */ 329 __NR_MEMCG_DATA_FLAGS = (1UL << 2), 330 }; 331 332 #define __FIRST_OBJEXT_FLAG __NR_MEMCG_DATA_FLAGS 333 334 #else /* CONFIG_MEMCG */ 335 336 #define __FIRST_OBJEXT_FLAG (1UL << 0) 337 338 #endif /* CONFIG_MEMCG */ 339 340 enum objext_flags { 341 /* slabobj_ext vector failed to allocate */ 342 OBJEXTS_ALLOC_FAIL = __FIRST_OBJEXT_FLAG, 343 /* the next bit after the last actual flag */ 344 __NR_OBJEXTS_FLAGS = (__FIRST_OBJEXT_FLAG << 1), 345 }; 346 347 #define OBJEXTS_FLAGS_MASK (__NR_OBJEXTS_FLAGS - 1) 348 349 #ifdef CONFIG_MEMCG 350 351 static inline bool folio_memcg_kmem(struct folio *folio); 352 353 /* 354 * After the initialization objcg->memcg is always pointing at 355 * a valid memcg, but can be atomically swapped to the parent memcg. 356 * 357 * The caller must ensure that the returned memcg won't be released. 358 */ 359 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg) 360 { 361 lockdep_assert_once(rcu_read_lock_held() || lockdep_is_held(&cgroup_mutex)); 362 return READ_ONCE(objcg->memcg); 363 } 364 365 /* 366 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio 367 * @folio: Pointer to the folio. 368 * 369 * Returns a pointer to the memory cgroup associated with the folio, 370 * or NULL. This function assumes that the folio is known to have a 371 * proper memory cgroup pointer. It's not safe to call this function 372 * against some type of folios, e.g. slab folios or ex-slab folios or 373 * kmem folios. 374 */ 375 static inline struct mem_cgroup *__folio_memcg(struct folio *folio) 376 { 377 unsigned long memcg_data = folio->memcg_data; 378 379 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 380 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio); 381 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio); 382 383 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 384 } 385 386 /* 387 * __folio_objcg - get the object cgroup associated with a kmem folio. 388 * @folio: Pointer to the folio. 389 * 390 * Returns a pointer to the object cgroup associated with the folio, 391 * or NULL. This function assumes that the folio is known to have a 392 * proper object cgroup pointer. It's not safe to call this function 393 * against some type of folios, e.g. slab folios or ex-slab folios or 394 * LRU folios. 395 */ 396 static inline struct obj_cgroup *__folio_objcg(struct folio *folio) 397 { 398 unsigned long memcg_data = folio->memcg_data; 399 400 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 401 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio); 402 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio); 403 404 return (struct obj_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 405 } 406 407 /* 408 * folio_memcg - Get the memory cgroup associated with a folio. 409 * @folio: Pointer to the folio. 410 * 411 * Returns a pointer to the memory cgroup associated with the folio, 412 * or NULL. This function assumes that the folio is known to have a 413 * proper memory cgroup pointer. It's not safe to call this function 414 * against some type of folios, e.g. slab folios or ex-slab folios. 415 * 416 * For a non-kmem folio any of the following ensures folio and memcg binding 417 * stability: 418 * 419 * - the folio lock 420 * - LRU isolation 421 * - exclusive reference 422 * 423 * For a kmem folio a caller should hold an rcu read lock to protect memcg 424 * associated with a kmem folio from being released. 425 */ 426 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 427 { 428 if (folio_memcg_kmem(folio)) 429 return obj_cgroup_memcg(__folio_objcg(folio)); 430 return __folio_memcg(folio); 431 } 432 433 /* 434 * folio_memcg_charged - If a folio is charged to a memory cgroup. 435 * @folio: Pointer to the folio. 436 * 437 * Returns true if folio is charged to a memory cgroup, otherwise returns false. 438 */ 439 static inline bool folio_memcg_charged(struct folio *folio) 440 { 441 if (folio_memcg_kmem(folio)) 442 return __folio_objcg(folio) != NULL; 443 return __folio_memcg(folio) != NULL; 444 } 445 446 /** 447 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio. 448 * @folio: Pointer to the folio. 449 * 450 * This function assumes that the folio is known to have a 451 * proper memory cgroup pointer. It's not safe to call this function 452 * against some type of folios, e.g. slab folios or ex-slab folios. 453 * 454 * Return: A pointer to the memory cgroup associated with the folio, 455 * or NULL. 456 */ 457 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 458 { 459 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 460 461 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 462 463 if (memcg_data & MEMCG_DATA_KMEM) { 464 struct obj_cgroup *objcg; 465 466 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 467 return obj_cgroup_memcg(objcg); 468 } 469 470 WARN_ON_ONCE(!rcu_read_lock_held()); 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 * - exclusive reference 491 * 492 * For a kmem folio a caller should hold an rcu read lock to protect memcg 493 * associated with a kmem folio from being released. 494 */ 495 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 496 { 497 /* 498 * Because folio->memcg_data might be changed asynchronously 499 * for slabs, READ_ONCE() should be used here. 500 */ 501 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 502 503 if (memcg_data & MEMCG_DATA_OBJEXTS) 504 return NULL; 505 506 if (memcg_data & MEMCG_DATA_KMEM) { 507 struct obj_cgroup *objcg; 508 509 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 510 return obj_cgroup_memcg(objcg); 511 } 512 513 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 514 } 515 516 static inline struct mem_cgroup *page_memcg_check(struct page *page) 517 { 518 if (PageTail(page)) 519 return NULL; 520 return folio_memcg_check((struct folio *)page); 521 } 522 523 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 524 { 525 struct mem_cgroup *memcg; 526 527 rcu_read_lock(); 528 retry: 529 memcg = obj_cgroup_memcg(objcg); 530 if (unlikely(!css_tryget(&memcg->css))) 531 goto retry; 532 rcu_read_unlock(); 533 534 return memcg; 535 } 536 537 /* 538 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set. 539 * @folio: Pointer to the folio. 540 * 541 * Checks if the folio has MemcgKmem flag set. The caller must ensure 542 * that the folio has an associated memory cgroup. It's not safe to call 543 * this function against some types of folios, e.g. slab folios. 544 */ 545 static inline bool folio_memcg_kmem(struct folio *folio) 546 { 547 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page); 548 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJEXTS, folio); 549 return folio->memcg_data & MEMCG_DATA_KMEM; 550 } 551 552 static inline bool PageMemcgKmem(struct page *page) 553 { 554 return folio_memcg_kmem(page_folio(page)); 555 } 556 557 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 558 { 559 return (memcg == root_mem_cgroup); 560 } 561 562 static inline bool mem_cgroup_disabled(void) 563 { 564 return !cgroup_subsys_enabled(memory_cgrp_subsys); 565 } 566 567 static inline void mem_cgroup_protection(struct mem_cgroup *root, 568 struct mem_cgroup *memcg, 569 unsigned long *min, 570 unsigned long *low) 571 { 572 *min = *low = 0; 573 574 if (mem_cgroup_disabled()) 575 return; 576 577 /* 578 * There is no reclaim protection applied to a targeted reclaim. 579 * We are special casing this specific case here because 580 * mem_cgroup_calculate_protection is not robust enough to keep 581 * the protection invariant for calculated effective values for 582 * parallel reclaimers with different reclaim target. This is 583 * especially a problem for tail memcgs (as they have pages on LRU) 584 * which would want to have effective values 0 for targeted reclaim 585 * but a different value for external reclaim. 586 * 587 * Example 588 * Let's have global and A's reclaim in parallel: 589 * | 590 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) 591 * |\ 592 * | C (low = 1G, usage = 2.5G) 593 * B (low = 1G, usage = 0.5G) 594 * 595 * For the global reclaim 596 * A.elow = A.low 597 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow 598 * C.elow = min(C.usage, C.low) 599 * 600 * With the effective values resetting we have A reclaim 601 * A.elow = 0 602 * B.elow = B.low 603 * C.elow = C.low 604 * 605 * If the global reclaim races with A's reclaim then 606 * B.elow = C.elow = 0 because children_low_usage > A.elow) 607 * is possible and reclaiming B would be violating the protection. 608 * 609 */ 610 if (root == memcg) 611 return; 612 613 *min = READ_ONCE(memcg->memory.emin); 614 *low = READ_ONCE(memcg->memory.elow); 615 } 616 617 void mem_cgroup_calculate_protection(struct mem_cgroup *root, 618 struct mem_cgroup *memcg); 619 620 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 621 struct mem_cgroup *memcg) 622 { 623 /* 624 * The root memcg doesn't account charges, and doesn't support 625 * protection. The target memcg's protection is ignored, see 626 * mem_cgroup_calculate_protection() and mem_cgroup_protection() 627 */ 628 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) || 629 memcg == target; 630 } 631 632 static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 633 struct mem_cgroup *memcg) 634 { 635 if (mem_cgroup_unprotected(target, memcg)) 636 return false; 637 638 return READ_ONCE(memcg->memory.elow) >= 639 page_counter_read(&memcg->memory); 640 } 641 642 static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 643 struct mem_cgroup *memcg) 644 { 645 if (mem_cgroup_unprotected(target, memcg)) 646 return false; 647 648 return READ_ONCE(memcg->memory.emin) >= 649 page_counter_read(&memcg->memory); 650 } 651 652 void mem_cgroup_commit_charge(struct folio *folio, struct mem_cgroup *memcg); 653 654 int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp); 655 656 /** 657 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup. 658 * @folio: Folio to charge. 659 * @mm: mm context of the allocating task. 660 * @gfp: Reclaim mode. 661 * 662 * Try to charge @folio to the memcg that @mm belongs to, reclaiming 663 * pages according to @gfp if necessary. If @mm is NULL, try to 664 * charge to the active memcg. 665 * 666 * Do not use this for folios allocated for swapin. 667 * 668 * Return: 0 on success. Otherwise, an error code is returned. 669 */ 670 static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, 671 gfp_t gfp) 672 { 673 if (mem_cgroup_disabled()) 674 return 0; 675 return __mem_cgroup_charge(folio, mm, gfp); 676 } 677 678 int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, gfp_t gfp, 679 long nr_pages); 680 681 int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm, 682 gfp_t gfp, swp_entry_t entry); 683 684 void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry, unsigned int nr_pages); 685 686 void __mem_cgroup_uncharge(struct folio *folio); 687 688 /** 689 * mem_cgroup_uncharge - Uncharge a folio. 690 * @folio: Folio to uncharge. 691 * 692 * Uncharge a folio previously charged with mem_cgroup_charge(). 693 */ 694 static inline void mem_cgroup_uncharge(struct folio *folio) 695 { 696 if (mem_cgroup_disabled()) 697 return; 698 __mem_cgroup_uncharge(folio); 699 } 700 701 void __mem_cgroup_uncharge_folios(struct folio_batch *folios); 702 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 703 { 704 if (mem_cgroup_disabled()) 705 return; 706 __mem_cgroup_uncharge_folios(folios); 707 } 708 709 void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages); 710 void mem_cgroup_replace_folio(struct folio *old, struct folio *new); 711 void mem_cgroup_migrate(struct folio *old, struct folio *new); 712 713 /** 714 * mem_cgroup_lruvec - get the lru list vector for a memcg & node 715 * @memcg: memcg of the wanted lruvec 716 * @pgdat: pglist_data 717 * 718 * Returns the lru list vector holding pages for a given @memcg & 719 * @pgdat combination. This can be the node lruvec, if the memory 720 * controller is disabled. 721 */ 722 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 723 struct pglist_data *pgdat) 724 { 725 struct mem_cgroup_per_node *mz; 726 struct lruvec *lruvec; 727 728 if (mem_cgroup_disabled()) { 729 lruvec = &pgdat->__lruvec; 730 goto out; 731 } 732 733 if (!memcg) 734 memcg = root_mem_cgroup; 735 736 mz = memcg->nodeinfo[pgdat->node_id]; 737 lruvec = &mz->lruvec; 738 out: 739 /* 740 * Since a node can be onlined after the mem_cgroup was created, 741 * we have to be prepared to initialize lruvec->pgdat here; 742 * and if offlined then reonlined, we need to reinitialize it. 743 */ 744 if (unlikely(lruvec->pgdat != pgdat)) 745 lruvec->pgdat = pgdat; 746 return lruvec; 747 } 748 749 /** 750 * folio_lruvec - return lruvec for isolating/putting an LRU folio 751 * @folio: Pointer to the folio. 752 * 753 * This function relies on folio->mem_cgroup being stable. 754 */ 755 static inline struct lruvec *folio_lruvec(struct folio *folio) 756 { 757 struct mem_cgroup *memcg = folio_memcg(folio); 758 759 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio); 760 return mem_cgroup_lruvec(memcg, folio_pgdat(folio)); 761 } 762 763 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); 764 765 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); 766 767 struct mem_cgroup *get_mem_cgroup_from_current(void); 768 769 struct mem_cgroup *get_mem_cgroup_from_folio(struct folio *folio); 770 771 struct lruvec *folio_lruvec_lock(struct folio *folio); 772 struct lruvec *folio_lruvec_lock_irq(struct folio *folio); 773 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 774 unsigned long *flags); 775 776 #ifdef CONFIG_DEBUG_VM 777 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio); 778 #else 779 static inline 780 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 781 { 782 } 783 #endif 784 785 static inline 786 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ 787 return css ? container_of(css, struct mem_cgroup, css) : NULL; 788 } 789 790 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) 791 { 792 return percpu_ref_tryget(&objcg->refcnt); 793 } 794 795 static inline void obj_cgroup_get(struct obj_cgroup *objcg) 796 { 797 percpu_ref_get(&objcg->refcnt); 798 } 799 800 static inline void obj_cgroup_get_many(struct obj_cgroup *objcg, 801 unsigned long nr) 802 { 803 percpu_ref_get_many(&objcg->refcnt, nr); 804 } 805 806 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 807 { 808 if (objcg) 809 percpu_ref_put(&objcg->refcnt); 810 } 811 812 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 813 { 814 return !memcg || css_tryget(&memcg->css); 815 } 816 817 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 818 { 819 return !memcg || css_tryget_online(&memcg->css); 820 } 821 822 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 823 { 824 if (memcg) 825 css_put(&memcg->css); 826 } 827 828 #define mem_cgroup_from_counter(counter, member) \ 829 container_of(counter, struct mem_cgroup, member) 830 831 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, 832 struct mem_cgroup *, 833 struct mem_cgroup_reclaim_cookie *); 834 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); 835 void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 836 int (*)(struct task_struct *, void *), void *arg); 837 838 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 839 { 840 if (mem_cgroup_disabled()) 841 return 0; 842 843 return memcg->id.id; 844 } 845 struct mem_cgroup *mem_cgroup_from_id(unsigned short id); 846 847 #ifdef CONFIG_SHRINKER_DEBUG 848 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 849 { 850 return memcg ? cgroup_ino(memcg->css.cgroup) : 0; 851 } 852 853 struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino); 854 #endif 855 856 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 857 { 858 return mem_cgroup_from_css(seq_css(m)); 859 } 860 861 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 862 { 863 struct mem_cgroup_per_node *mz; 864 865 if (mem_cgroup_disabled()) 866 return NULL; 867 868 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 869 return mz->memcg; 870 } 871 872 /** 873 * parent_mem_cgroup - find the accounting parent of a memcg 874 * @memcg: memcg whose parent to find 875 * 876 * Returns the parent memcg, or NULL if this is the root. 877 */ 878 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 879 { 880 return mem_cgroup_from_css(memcg->css.parent); 881 } 882 883 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, 884 struct mem_cgroup *root) 885 { 886 if (root == memcg) 887 return true; 888 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); 889 } 890 891 static inline bool mm_match_cgroup(struct mm_struct *mm, 892 struct mem_cgroup *memcg) 893 { 894 struct mem_cgroup *task_memcg; 895 bool match = false; 896 897 rcu_read_lock(); 898 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 899 if (task_memcg) 900 match = mem_cgroup_is_descendant(task_memcg, memcg); 901 rcu_read_unlock(); 902 return match; 903 } 904 905 struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio); 906 ino_t page_cgroup_ino(struct page *page); 907 908 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 909 { 910 if (mem_cgroup_disabled()) 911 return true; 912 return !!(memcg->css.flags & CSS_ONLINE); 913 } 914 915 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, 916 int zid, int nr_pages); 917 918 static inline 919 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 920 enum lru_list lru, int zone_idx) 921 { 922 struct mem_cgroup_per_node *mz; 923 924 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 925 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); 926 } 927 928 void mem_cgroup_handle_over_high(gfp_t gfp_mask); 929 930 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); 931 932 unsigned long mem_cgroup_size(struct mem_cgroup *memcg); 933 934 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, 935 struct task_struct *p); 936 937 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); 938 939 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, 940 struct mem_cgroup *oom_domain); 941 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); 942 943 void __mod_memcg_state(struct mem_cgroup *memcg, enum memcg_stat_item idx, 944 int val); 945 946 /* idx can be of type enum memcg_stat_item or node_stat_item */ 947 static inline void mod_memcg_state(struct mem_cgroup *memcg, 948 enum memcg_stat_item idx, int val) 949 { 950 unsigned long flags; 951 952 local_irq_save(flags); 953 __mod_memcg_state(memcg, idx, val); 954 local_irq_restore(flags); 955 } 956 957 static inline void mod_memcg_page_state(struct page *page, 958 enum memcg_stat_item idx, int val) 959 { 960 struct mem_cgroup *memcg; 961 962 if (mem_cgroup_disabled()) 963 return; 964 965 rcu_read_lock(); 966 memcg = folio_memcg(page_folio(page)); 967 if (memcg) 968 mod_memcg_state(memcg, idx, val); 969 rcu_read_unlock(); 970 } 971 972 unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx); 973 unsigned long lruvec_page_state(struct lruvec *lruvec, enum node_stat_item idx); 974 unsigned long lruvec_page_state_local(struct lruvec *lruvec, 975 enum node_stat_item idx); 976 977 void mem_cgroup_flush_stats(struct mem_cgroup *memcg); 978 void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg); 979 980 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val); 981 982 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 983 int val) 984 { 985 unsigned long flags; 986 987 local_irq_save(flags); 988 __mod_lruvec_kmem_state(p, idx, val); 989 local_irq_restore(flags); 990 } 991 992 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, 993 unsigned long count); 994 995 static inline void count_memcg_events(struct mem_cgroup *memcg, 996 enum vm_event_item idx, 997 unsigned long count) 998 { 999 unsigned long flags; 1000 1001 local_irq_save(flags); 1002 __count_memcg_events(memcg, idx, count); 1003 local_irq_restore(flags); 1004 } 1005 1006 static inline void count_memcg_folio_events(struct folio *folio, 1007 enum vm_event_item idx, unsigned long nr) 1008 { 1009 struct mem_cgroup *memcg = folio_memcg(folio); 1010 1011 if (memcg) 1012 count_memcg_events(memcg, idx, nr); 1013 } 1014 1015 static inline void count_memcg_events_mm(struct mm_struct *mm, 1016 enum vm_event_item idx, unsigned long count) 1017 { 1018 struct mem_cgroup *memcg; 1019 1020 if (mem_cgroup_disabled()) 1021 return; 1022 1023 rcu_read_lock(); 1024 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1025 if (likely(memcg)) 1026 count_memcg_events(memcg, idx, count); 1027 rcu_read_unlock(); 1028 } 1029 1030 static inline void count_memcg_event_mm(struct mm_struct *mm, 1031 enum vm_event_item idx) 1032 { 1033 count_memcg_events_mm(mm, idx, 1); 1034 } 1035 1036 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1037 enum memcg_memory_event event) 1038 { 1039 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX || 1040 event == MEMCG_SWAP_FAIL; 1041 1042 atomic_long_inc(&memcg->memory_events_local[event]); 1043 if (!swap_event) 1044 cgroup_file_notify(&memcg->events_local_file); 1045 1046 do { 1047 atomic_long_inc(&memcg->memory_events[event]); 1048 if (swap_event) 1049 cgroup_file_notify(&memcg->swap_events_file); 1050 else 1051 cgroup_file_notify(&memcg->events_file); 1052 1053 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1054 break; 1055 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) 1056 break; 1057 } while ((memcg = parent_mem_cgroup(memcg)) && 1058 !mem_cgroup_is_root(memcg)); 1059 } 1060 1061 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1062 enum memcg_memory_event event) 1063 { 1064 struct mem_cgroup *memcg; 1065 1066 if (mem_cgroup_disabled()) 1067 return; 1068 1069 rcu_read_lock(); 1070 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1071 if (likely(memcg)) 1072 memcg_memory_event(memcg, event); 1073 rcu_read_unlock(); 1074 } 1075 1076 void split_page_memcg(struct page *head, int old_order, int new_order); 1077 1078 #else /* CONFIG_MEMCG */ 1079 1080 #define MEM_CGROUP_ID_SHIFT 0 1081 1082 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 1083 { 1084 return NULL; 1085 } 1086 1087 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 1088 { 1089 WARN_ON_ONCE(!rcu_read_lock_held()); 1090 return NULL; 1091 } 1092 1093 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 1094 { 1095 return NULL; 1096 } 1097 1098 static inline struct mem_cgroup *page_memcg_check(struct page *page) 1099 { 1100 return NULL; 1101 } 1102 1103 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 1104 { 1105 return NULL; 1106 } 1107 1108 static inline bool folio_memcg_kmem(struct folio *folio) 1109 { 1110 return false; 1111 } 1112 1113 static inline bool PageMemcgKmem(struct page *page) 1114 { 1115 return false; 1116 } 1117 1118 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 1119 { 1120 return true; 1121 } 1122 1123 static inline bool mem_cgroup_disabled(void) 1124 { 1125 return true; 1126 } 1127 1128 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1129 enum memcg_memory_event event) 1130 { 1131 } 1132 1133 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1134 enum memcg_memory_event event) 1135 { 1136 } 1137 1138 static inline void mem_cgroup_protection(struct mem_cgroup *root, 1139 struct mem_cgroup *memcg, 1140 unsigned long *min, 1141 unsigned long *low) 1142 { 1143 *min = *low = 0; 1144 } 1145 1146 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, 1147 struct mem_cgroup *memcg) 1148 { 1149 } 1150 1151 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 1152 struct mem_cgroup *memcg) 1153 { 1154 return true; 1155 } 1156 static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 1157 struct mem_cgroup *memcg) 1158 { 1159 return false; 1160 } 1161 1162 static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 1163 struct mem_cgroup *memcg) 1164 { 1165 return false; 1166 } 1167 1168 static inline void mem_cgroup_commit_charge(struct folio *folio, 1169 struct mem_cgroup *memcg) 1170 { 1171 } 1172 1173 static inline int mem_cgroup_charge(struct folio *folio, 1174 struct mm_struct *mm, gfp_t gfp) 1175 { 1176 return 0; 1177 } 1178 1179 static inline int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, 1180 gfp_t gfp, long nr_pages) 1181 { 1182 return 0; 1183 } 1184 1185 static inline int mem_cgroup_swapin_charge_folio(struct folio *folio, 1186 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry) 1187 { 1188 return 0; 1189 } 1190 1191 static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry, unsigned int nr) 1192 { 1193 } 1194 1195 static inline void mem_cgroup_uncharge(struct folio *folio) 1196 { 1197 } 1198 1199 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 1200 { 1201 } 1202 1203 static inline void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, 1204 unsigned int nr_pages) 1205 { 1206 } 1207 1208 static inline void mem_cgroup_replace_folio(struct folio *old, 1209 struct folio *new) 1210 { 1211 } 1212 1213 static inline void mem_cgroup_migrate(struct folio *old, struct folio *new) 1214 { 1215 } 1216 1217 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 1218 struct pglist_data *pgdat) 1219 { 1220 return &pgdat->__lruvec; 1221 } 1222 1223 static inline struct lruvec *folio_lruvec(struct folio *folio) 1224 { 1225 struct pglist_data *pgdat = folio_pgdat(folio); 1226 return &pgdat->__lruvec; 1227 } 1228 1229 static inline 1230 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 1231 { 1232 } 1233 1234 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 1235 { 1236 return NULL; 1237 } 1238 1239 static inline bool mm_match_cgroup(struct mm_struct *mm, 1240 struct mem_cgroup *memcg) 1241 { 1242 return true; 1243 } 1244 1245 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) 1246 { 1247 return NULL; 1248 } 1249 1250 static inline struct mem_cgroup *get_mem_cgroup_from_current(void) 1251 { 1252 return NULL; 1253 } 1254 1255 static inline struct mem_cgroup *get_mem_cgroup_from_folio(struct folio *folio) 1256 { 1257 return NULL; 1258 } 1259 1260 static inline 1261 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css) 1262 { 1263 return NULL; 1264 } 1265 1266 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 1267 { 1268 } 1269 1270 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 1271 { 1272 return true; 1273 } 1274 1275 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 1276 { 1277 return true; 1278 } 1279 1280 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 1281 { 1282 } 1283 1284 static inline struct lruvec *folio_lruvec_lock(struct folio *folio) 1285 { 1286 struct pglist_data *pgdat = folio_pgdat(folio); 1287 1288 spin_lock(&pgdat->__lruvec.lru_lock); 1289 return &pgdat->__lruvec; 1290 } 1291 1292 static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio) 1293 { 1294 struct pglist_data *pgdat = folio_pgdat(folio); 1295 1296 spin_lock_irq(&pgdat->__lruvec.lru_lock); 1297 return &pgdat->__lruvec; 1298 } 1299 1300 static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 1301 unsigned long *flagsp) 1302 { 1303 struct pglist_data *pgdat = folio_pgdat(folio); 1304 1305 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp); 1306 return &pgdat->__lruvec; 1307 } 1308 1309 static inline struct mem_cgroup * 1310 mem_cgroup_iter(struct mem_cgroup *root, 1311 struct mem_cgroup *prev, 1312 struct mem_cgroup_reclaim_cookie *reclaim) 1313 { 1314 return NULL; 1315 } 1316 1317 static inline void mem_cgroup_iter_break(struct mem_cgroup *root, 1318 struct mem_cgroup *prev) 1319 { 1320 } 1321 1322 static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 1323 int (*fn)(struct task_struct *, void *), void *arg) 1324 { 1325 } 1326 1327 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 1328 { 1329 return 0; 1330 } 1331 1332 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) 1333 { 1334 WARN_ON_ONCE(id); 1335 /* XXX: This should always return root_mem_cgroup */ 1336 return NULL; 1337 } 1338 1339 #ifdef CONFIG_SHRINKER_DEBUG 1340 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 1341 { 1342 return 0; 1343 } 1344 1345 static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino) 1346 { 1347 return NULL; 1348 } 1349 #endif 1350 1351 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 1352 { 1353 return NULL; 1354 } 1355 1356 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 1357 { 1358 return NULL; 1359 } 1360 1361 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 1362 { 1363 return true; 1364 } 1365 1366 static inline 1367 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 1368 enum lru_list lru, int zone_idx) 1369 { 1370 return 0; 1371 } 1372 1373 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) 1374 { 1375 return 0; 1376 } 1377 1378 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) 1379 { 1380 return 0; 1381 } 1382 1383 static inline void 1384 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) 1385 { 1386 } 1387 1388 static inline void 1389 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) 1390 { 1391 } 1392 1393 static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask) 1394 { 1395 } 1396 1397 static inline struct mem_cgroup *mem_cgroup_get_oom_group( 1398 struct task_struct *victim, struct mem_cgroup *oom_domain) 1399 { 1400 return NULL; 1401 } 1402 1403 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) 1404 { 1405 } 1406 1407 static inline void __mod_memcg_state(struct mem_cgroup *memcg, 1408 enum memcg_stat_item idx, 1409 int nr) 1410 { 1411 } 1412 1413 static inline void mod_memcg_state(struct mem_cgroup *memcg, 1414 enum memcg_stat_item idx, 1415 int nr) 1416 { 1417 } 1418 1419 static inline void mod_memcg_page_state(struct page *page, 1420 enum memcg_stat_item idx, int val) 1421 { 1422 } 1423 1424 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) 1425 { 1426 return 0; 1427 } 1428 1429 static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1430 enum node_stat_item idx) 1431 { 1432 return node_page_state(lruvec_pgdat(lruvec), idx); 1433 } 1434 1435 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1436 enum node_stat_item idx) 1437 { 1438 return node_page_state(lruvec_pgdat(lruvec), idx); 1439 } 1440 1441 static inline void mem_cgroup_flush_stats(struct mem_cgroup *memcg) 1442 { 1443 } 1444 1445 static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg) 1446 { 1447 } 1448 1449 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1450 int val) 1451 { 1452 struct page *page = virt_to_head_page(p); 1453 1454 __mod_node_page_state(page_pgdat(page), idx, val); 1455 } 1456 1457 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1458 int val) 1459 { 1460 struct page *page = virt_to_head_page(p); 1461 1462 mod_node_page_state(page_pgdat(page), idx, val); 1463 } 1464 1465 static inline void count_memcg_events(struct mem_cgroup *memcg, 1466 enum vm_event_item idx, 1467 unsigned long count) 1468 { 1469 } 1470 1471 static inline void __count_memcg_events(struct mem_cgroup *memcg, 1472 enum vm_event_item idx, 1473 unsigned long count) 1474 { 1475 } 1476 1477 static inline void count_memcg_folio_events(struct folio *folio, 1478 enum vm_event_item idx, unsigned long nr) 1479 { 1480 } 1481 1482 static inline void count_memcg_events_mm(struct mm_struct *mm, 1483 enum vm_event_item idx, unsigned long count) 1484 { 1485 } 1486 1487 static inline 1488 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) 1489 { 1490 } 1491 1492 static inline void split_page_memcg(struct page *head, int old_order, int new_order) 1493 { 1494 } 1495 #endif /* CONFIG_MEMCG */ 1496 1497 /* 1498 * Extended information for slab objects stored as an array in page->memcg_data 1499 * if MEMCG_DATA_OBJEXTS is set. 1500 */ 1501 struct slabobj_ext { 1502 #ifdef CONFIG_MEMCG 1503 struct obj_cgroup *objcg; 1504 #endif 1505 #ifdef CONFIG_MEM_ALLOC_PROFILING 1506 union codetag_ref ref; 1507 #endif 1508 } __aligned(8); 1509 1510 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx) 1511 { 1512 __mod_lruvec_kmem_state(p, idx, 1); 1513 } 1514 1515 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx) 1516 { 1517 __mod_lruvec_kmem_state(p, idx, -1); 1518 } 1519 1520 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) 1521 { 1522 struct mem_cgroup *memcg; 1523 1524 memcg = lruvec_memcg(lruvec); 1525 if (!memcg) 1526 return NULL; 1527 memcg = parent_mem_cgroup(memcg); 1528 if (!memcg) 1529 return NULL; 1530 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); 1531 } 1532 1533 static inline void unlock_page_lruvec(struct lruvec *lruvec) 1534 { 1535 spin_unlock(&lruvec->lru_lock); 1536 } 1537 1538 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec) 1539 { 1540 spin_unlock_irq(&lruvec->lru_lock); 1541 } 1542 1543 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec, 1544 unsigned long flags) 1545 { 1546 spin_unlock_irqrestore(&lruvec->lru_lock, flags); 1547 } 1548 1549 /* Test requires a stable folio->memcg binding, see folio_memcg() */ 1550 static inline bool folio_matches_lruvec(struct folio *folio, 1551 struct lruvec *lruvec) 1552 { 1553 return lruvec_pgdat(lruvec) == folio_pgdat(folio) && 1554 lruvec_memcg(lruvec) == folio_memcg(folio); 1555 } 1556 1557 /* Don't lock again iff page's lruvec locked */ 1558 static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio, 1559 struct lruvec *locked_lruvec) 1560 { 1561 if (locked_lruvec) { 1562 if (folio_matches_lruvec(folio, locked_lruvec)) 1563 return locked_lruvec; 1564 1565 unlock_page_lruvec_irq(locked_lruvec); 1566 } 1567 1568 return folio_lruvec_lock_irq(folio); 1569 } 1570 1571 /* Don't lock again iff folio's lruvec locked */ 1572 static inline void folio_lruvec_relock_irqsave(struct folio *folio, 1573 struct lruvec **lruvecp, unsigned long *flags) 1574 { 1575 if (*lruvecp) { 1576 if (folio_matches_lruvec(folio, *lruvecp)) 1577 return; 1578 1579 unlock_page_lruvec_irqrestore(*lruvecp, *flags); 1580 } 1581 1582 *lruvecp = folio_lruvec_lock_irqsave(folio, flags); 1583 } 1584 1585 #ifdef CONFIG_CGROUP_WRITEBACK 1586 1587 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); 1588 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, 1589 unsigned long *pheadroom, unsigned long *pdirty, 1590 unsigned long *pwriteback); 1591 1592 void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio, 1593 struct bdi_writeback *wb); 1594 1595 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1596 struct bdi_writeback *wb) 1597 { 1598 struct mem_cgroup *memcg; 1599 1600 if (mem_cgroup_disabled()) 1601 return; 1602 1603 memcg = folio_memcg(folio); 1604 if (unlikely(memcg && &memcg->css != wb->memcg_css)) 1605 mem_cgroup_track_foreign_dirty_slowpath(folio, wb); 1606 } 1607 1608 void mem_cgroup_flush_foreign(struct bdi_writeback *wb); 1609 1610 #else /* CONFIG_CGROUP_WRITEBACK */ 1611 1612 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) 1613 { 1614 return NULL; 1615 } 1616 1617 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, 1618 unsigned long *pfilepages, 1619 unsigned long *pheadroom, 1620 unsigned long *pdirty, 1621 unsigned long *pwriteback) 1622 { 1623 } 1624 1625 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1626 struct bdi_writeback *wb) 1627 { 1628 } 1629 1630 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) 1631 { 1632 } 1633 1634 #endif /* CONFIG_CGROUP_WRITEBACK */ 1635 1636 struct sock; 1637 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages, 1638 gfp_t gfp_mask); 1639 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); 1640 #ifdef CONFIG_MEMCG 1641 extern struct static_key_false memcg_sockets_enabled_key; 1642 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) 1643 void mem_cgroup_sk_alloc(struct sock *sk); 1644 void mem_cgroup_sk_free(struct sock *sk); 1645 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1646 { 1647 #ifdef CONFIG_MEMCG_V1 1648 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1649 return !!memcg->tcpmem_pressure; 1650 #endif /* CONFIG_MEMCG_V1 */ 1651 do { 1652 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure))) 1653 return true; 1654 } while ((memcg = parent_mem_cgroup(memcg))); 1655 return false; 1656 } 1657 1658 int alloc_shrinker_info(struct mem_cgroup *memcg); 1659 void free_shrinker_info(struct mem_cgroup *memcg); 1660 void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id); 1661 void reparent_shrinker_deferred(struct mem_cgroup *memcg); 1662 #else 1663 #define mem_cgroup_sockets_enabled 0 1664 static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; 1665 static inline void mem_cgroup_sk_free(struct sock *sk) { }; 1666 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1667 { 1668 return false; 1669 } 1670 1671 static inline void set_shrinker_bit(struct mem_cgroup *memcg, 1672 int nid, int shrinker_id) 1673 { 1674 } 1675 #endif 1676 1677 #ifdef CONFIG_MEMCG 1678 bool mem_cgroup_kmem_disabled(void); 1679 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); 1680 void __memcg_kmem_uncharge_page(struct page *page, int order); 1681 1682 /* 1683 * The returned objcg pointer is safe to use without additional 1684 * protection within a scope. The scope is defined either by 1685 * the current task (similar to the "current" global variable) 1686 * or by set_active_memcg() pair. 1687 * Please, use obj_cgroup_get() to get a reference if the pointer 1688 * needs to be used outside of the local scope. 1689 */ 1690 struct obj_cgroup *current_obj_cgroup(void); 1691 struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio); 1692 1693 static inline struct obj_cgroup *get_obj_cgroup_from_current(void) 1694 { 1695 struct obj_cgroup *objcg = current_obj_cgroup(); 1696 1697 if (objcg) 1698 obj_cgroup_get(objcg); 1699 1700 return objcg; 1701 } 1702 1703 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); 1704 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); 1705 1706 extern struct static_key_false memcg_bpf_enabled_key; 1707 static inline bool memcg_bpf_enabled(void) 1708 { 1709 return static_branch_likely(&memcg_bpf_enabled_key); 1710 } 1711 1712 extern struct static_key_false memcg_kmem_online_key; 1713 1714 static inline bool memcg_kmem_online(void) 1715 { 1716 return static_branch_likely(&memcg_kmem_online_key); 1717 } 1718 1719 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1720 int order) 1721 { 1722 if (memcg_kmem_online()) 1723 return __memcg_kmem_charge_page(page, gfp, order); 1724 return 0; 1725 } 1726 1727 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1728 { 1729 if (memcg_kmem_online()) 1730 __memcg_kmem_uncharge_page(page, order); 1731 } 1732 1733 /* 1734 * A helper for accessing memcg's kmem_id, used for getting 1735 * corresponding LRU lists. 1736 */ 1737 static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1738 { 1739 return memcg ? memcg->kmemcg_id : -1; 1740 } 1741 1742 struct mem_cgroup *mem_cgroup_from_slab_obj(void *p); 1743 1744 static inline void count_objcg_event(struct obj_cgroup *objcg, 1745 enum vm_event_item idx) 1746 { 1747 struct mem_cgroup *memcg; 1748 1749 if (!memcg_kmem_online()) 1750 return; 1751 1752 rcu_read_lock(); 1753 memcg = obj_cgroup_memcg(objcg); 1754 count_memcg_events(memcg, idx, 1); 1755 rcu_read_unlock(); 1756 } 1757 1758 #else 1759 static inline bool mem_cgroup_kmem_disabled(void) 1760 { 1761 return true; 1762 } 1763 1764 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1765 int order) 1766 { 1767 return 0; 1768 } 1769 1770 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1771 { 1772 } 1773 1774 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1775 int order) 1776 { 1777 return 0; 1778 } 1779 1780 static inline void __memcg_kmem_uncharge_page(struct page *page, int order) 1781 { 1782 } 1783 1784 static inline struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio) 1785 { 1786 return NULL; 1787 } 1788 1789 static inline bool memcg_bpf_enabled(void) 1790 { 1791 return false; 1792 } 1793 1794 static inline bool memcg_kmem_online(void) 1795 { 1796 return false; 1797 } 1798 1799 static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1800 { 1801 return -1; 1802 } 1803 1804 static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p) 1805 { 1806 return NULL; 1807 } 1808 1809 static inline void count_objcg_event(struct obj_cgroup *objcg, 1810 enum vm_event_item idx) 1811 { 1812 } 1813 1814 #endif /* CONFIG_MEMCG */ 1815 1816 #if defined(CONFIG_MEMCG) && defined(CONFIG_ZSWAP) 1817 bool obj_cgroup_may_zswap(struct obj_cgroup *objcg); 1818 void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size); 1819 void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size); 1820 bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg); 1821 #else 1822 static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg) 1823 { 1824 return true; 1825 } 1826 static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, 1827 size_t size) 1828 { 1829 } 1830 static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, 1831 size_t size) 1832 { 1833 } 1834 static inline bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg) 1835 { 1836 /* if zswap is disabled, do not block pages going to the swapping device */ 1837 return true; 1838 } 1839 #endif 1840 1841 1842 /* Cgroup v1-related declarations */ 1843 1844 #ifdef CONFIG_MEMCG_V1 1845 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order, 1846 gfp_t gfp_mask, 1847 unsigned long *total_scanned); 1848 1849 bool mem_cgroup_oom_synchronize(bool wait); 1850 1851 static inline bool task_in_memcg_oom(struct task_struct *p) 1852 { 1853 return p->memcg_in_oom; 1854 } 1855 1856 static inline void mem_cgroup_enter_user_fault(void) 1857 { 1858 WARN_ON(current->in_user_fault); 1859 current->in_user_fault = 1; 1860 } 1861 1862 static inline void mem_cgroup_exit_user_fault(void) 1863 { 1864 WARN_ON(!current->in_user_fault); 1865 current->in_user_fault = 0; 1866 } 1867 1868 #else /* CONFIG_MEMCG_V1 */ 1869 static inline 1870 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order, 1871 gfp_t gfp_mask, 1872 unsigned long *total_scanned) 1873 { 1874 return 0; 1875 } 1876 1877 static inline bool task_in_memcg_oom(struct task_struct *p) 1878 { 1879 return false; 1880 } 1881 1882 static inline bool mem_cgroup_oom_synchronize(bool wait) 1883 { 1884 return false; 1885 } 1886 1887 static inline void mem_cgroup_enter_user_fault(void) 1888 { 1889 } 1890 1891 static inline void mem_cgroup_exit_user_fault(void) 1892 { 1893 } 1894 1895 #endif /* CONFIG_MEMCG_V1 */ 1896 1897 #endif /* _LINUX_MEMCONTROL_H */ 1898