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_check - Get the memory cgroup associated with a folio. 448 * @folio: Pointer to the folio. 449 * 450 * Returns a pointer to the memory cgroup associated with the folio, 451 * or NULL. This function unlike folio_memcg() can take any folio 452 * as an argument. It has to be used in cases when it's not known if a folio 453 * has an associated memory cgroup pointer or an object cgroups vector or 454 * an object cgroup. 455 * 456 * For a non-kmem folio any of the following ensures folio and memcg binding 457 * stability: 458 * 459 * - the folio lock 460 * - LRU isolation 461 * - exclusive reference 462 * 463 * For a kmem folio a caller should hold an rcu read lock to protect memcg 464 * associated with a kmem folio from being released. 465 */ 466 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 467 { 468 /* 469 * Because folio->memcg_data might be changed asynchronously 470 * for slabs, READ_ONCE() should be used here. 471 */ 472 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 473 474 if (memcg_data & MEMCG_DATA_OBJEXTS) 475 return NULL; 476 477 if (memcg_data & MEMCG_DATA_KMEM) { 478 struct obj_cgroup *objcg; 479 480 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 481 return obj_cgroup_memcg(objcg); 482 } 483 484 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK); 485 } 486 487 static inline struct mem_cgroup *page_memcg_check(struct page *page) 488 { 489 if (PageTail(page)) 490 return NULL; 491 return folio_memcg_check((struct folio *)page); 492 } 493 494 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 495 { 496 struct mem_cgroup *memcg; 497 498 rcu_read_lock(); 499 retry: 500 memcg = obj_cgroup_memcg(objcg); 501 if (unlikely(!css_tryget(&memcg->css))) 502 goto retry; 503 rcu_read_unlock(); 504 505 return memcg; 506 } 507 508 /* 509 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set. 510 * @folio: Pointer to the folio. 511 * 512 * Checks if the folio has MemcgKmem flag set. The caller must ensure 513 * that the folio has an associated memory cgroup. It's not safe to call 514 * this function against some types of folios, e.g. slab folios. 515 */ 516 static inline bool folio_memcg_kmem(struct folio *folio) 517 { 518 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page); 519 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJEXTS, folio); 520 return folio->memcg_data & MEMCG_DATA_KMEM; 521 } 522 523 static inline bool PageMemcgKmem(struct page *page) 524 { 525 return folio_memcg_kmem(page_folio(page)); 526 } 527 528 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 529 { 530 return (memcg == root_mem_cgroup); 531 } 532 533 static inline bool mem_cgroup_disabled(void) 534 { 535 return !cgroup_subsys_enabled(memory_cgrp_subsys); 536 } 537 538 static inline void mem_cgroup_protection(struct mem_cgroup *root, 539 struct mem_cgroup *memcg, 540 unsigned long *min, 541 unsigned long *low) 542 { 543 *min = *low = 0; 544 545 if (mem_cgroup_disabled()) 546 return; 547 548 /* 549 * There is no reclaim protection applied to a targeted reclaim. 550 * We are special casing this specific case here because 551 * mem_cgroup_calculate_protection is not robust enough to keep 552 * the protection invariant for calculated effective values for 553 * parallel reclaimers with different reclaim target. This is 554 * especially a problem for tail memcgs (as they have pages on LRU) 555 * which would want to have effective values 0 for targeted reclaim 556 * but a different value for external reclaim. 557 * 558 * Example 559 * Let's have global and A's reclaim in parallel: 560 * | 561 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) 562 * |\ 563 * | C (low = 1G, usage = 2.5G) 564 * B (low = 1G, usage = 0.5G) 565 * 566 * For the global reclaim 567 * A.elow = A.low 568 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow 569 * C.elow = min(C.usage, C.low) 570 * 571 * With the effective values resetting we have A reclaim 572 * A.elow = 0 573 * B.elow = B.low 574 * C.elow = C.low 575 * 576 * If the global reclaim races with A's reclaim then 577 * B.elow = C.elow = 0 because children_low_usage > A.elow) 578 * is possible and reclaiming B would be violating the protection. 579 * 580 */ 581 if (root == memcg) 582 return; 583 584 *min = READ_ONCE(memcg->memory.emin); 585 *low = READ_ONCE(memcg->memory.elow); 586 } 587 588 void mem_cgroup_calculate_protection(struct mem_cgroup *root, 589 struct mem_cgroup *memcg); 590 591 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 592 struct mem_cgroup *memcg) 593 { 594 /* 595 * The root memcg doesn't account charges, and doesn't support 596 * protection. The target memcg's protection is ignored, see 597 * mem_cgroup_calculate_protection() and mem_cgroup_protection() 598 */ 599 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) || 600 memcg == target; 601 } 602 603 static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 604 struct mem_cgroup *memcg) 605 { 606 if (mem_cgroup_unprotected(target, memcg)) 607 return false; 608 609 return READ_ONCE(memcg->memory.elow) >= 610 page_counter_read(&memcg->memory); 611 } 612 613 static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 614 struct mem_cgroup *memcg) 615 { 616 if (mem_cgroup_unprotected(target, memcg)) 617 return false; 618 619 return READ_ONCE(memcg->memory.emin) >= 620 page_counter_read(&memcg->memory); 621 } 622 623 int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp); 624 625 /** 626 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup. 627 * @folio: Folio to charge. 628 * @mm: mm context of the allocating task. 629 * @gfp: Reclaim mode. 630 * 631 * Try to charge @folio to the memcg that @mm belongs to, reclaiming 632 * pages according to @gfp if necessary. If @mm is NULL, try to 633 * charge to the active memcg. 634 * 635 * Do not use this for folios allocated for swapin. 636 * 637 * Return: 0 on success. Otherwise, an error code is returned. 638 */ 639 static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, 640 gfp_t gfp) 641 { 642 if (mem_cgroup_disabled()) 643 return 0; 644 return __mem_cgroup_charge(folio, mm, gfp); 645 } 646 647 int mem_cgroup_charge_hugetlb(struct folio* folio, gfp_t gfp); 648 649 int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm, 650 gfp_t gfp, swp_entry_t entry); 651 652 void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry, unsigned int nr_pages); 653 654 void __mem_cgroup_uncharge(struct folio *folio); 655 656 /** 657 * mem_cgroup_uncharge - Uncharge a folio. 658 * @folio: Folio to uncharge. 659 * 660 * Uncharge a folio previously charged with mem_cgroup_charge(). 661 */ 662 static inline void mem_cgroup_uncharge(struct folio *folio) 663 { 664 if (mem_cgroup_disabled()) 665 return; 666 __mem_cgroup_uncharge(folio); 667 } 668 669 void __mem_cgroup_uncharge_folios(struct folio_batch *folios); 670 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 671 { 672 if (mem_cgroup_disabled()) 673 return; 674 __mem_cgroup_uncharge_folios(folios); 675 } 676 677 void mem_cgroup_replace_folio(struct folio *old, struct folio *new); 678 void mem_cgroup_migrate(struct folio *old, struct folio *new); 679 680 /** 681 * mem_cgroup_lruvec - get the lru list vector for a memcg & node 682 * @memcg: memcg of the wanted lruvec 683 * @pgdat: pglist_data 684 * 685 * Returns the lru list vector holding pages for a given @memcg & 686 * @pgdat combination. This can be the node lruvec, if the memory 687 * controller is disabled. 688 */ 689 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 690 struct pglist_data *pgdat) 691 { 692 struct mem_cgroup_per_node *mz; 693 struct lruvec *lruvec; 694 695 if (mem_cgroup_disabled()) { 696 lruvec = &pgdat->__lruvec; 697 goto out; 698 } 699 700 if (!memcg) 701 memcg = root_mem_cgroup; 702 703 mz = memcg->nodeinfo[pgdat->node_id]; 704 lruvec = &mz->lruvec; 705 out: 706 /* 707 * Since a node can be onlined after the mem_cgroup was created, 708 * we have to be prepared to initialize lruvec->pgdat here; 709 * and if offlined then reonlined, we need to reinitialize it. 710 */ 711 if (unlikely(lruvec->pgdat != pgdat)) 712 lruvec->pgdat = pgdat; 713 return lruvec; 714 } 715 716 /** 717 * folio_lruvec - return lruvec for isolating/putting an LRU folio 718 * @folio: Pointer to the folio. 719 * 720 * This function relies on folio->mem_cgroup being stable. 721 */ 722 static inline struct lruvec *folio_lruvec(struct folio *folio) 723 { 724 struct mem_cgroup *memcg = folio_memcg(folio); 725 726 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio); 727 return mem_cgroup_lruvec(memcg, folio_pgdat(folio)); 728 } 729 730 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); 731 732 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); 733 734 struct mem_cgroup *get_mem_cgroup_from_current(void); 735 736 struct mem_cgroup *get_mem_cgroup_from_folio(struct folio *folio); 737 738 struct lruvec *folio_lruvec_lock(struct folio *folio); 739 struct lruvec *folio_lruvec_lock_irq(struct folio *folio); 740 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 741 unsigned long *flags); 742 743 #ifdef CONFIG_DEBUG_VM 744 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio); 745 #else 746 static inline 747 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 748 { 749 } 750 #endif 751 752 static inline 753 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ 754 return css ? container_of(css, struct mem_cgroup, css) : NULL; 755 } 756 757 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) 758 { 759 return percpu_ref_tryget(&objcg->refcnt); 760 } 761 762 static inline void obj_cgroup_get(struct obj_cgroup *objcg) 763 { 764 percpu_ref_get(&objcg->refcnt); 765 } 766 767 static inline void obj_cgroup_get_many(struct obj_cgroup *objcg, 768 unsigned long nr) 769 { 770 percpu_ref_get_many(&objcg->refcnt, nr); 771 } 772 773 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 774 { 775 if (objcg) 776 percpu_ref_put(&objcg->refcnt); 777 } 778 779 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 780 { 781 return !memcg || css_tryget(&memcg->css); 782 } 783 784 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 785 { 786 return !memcg || css_tryget_online(&memcg->css); 787 } 788 789 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 790 { 791 if (memcg) 792 css_put(&memcg->css); 793 } 794 795 #define mem_cgroup_from_counter(counter, member) \ 796 container_of(counter, struct mem_cgroup, member) 797 798 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, 799 struct mem_cgroup *, 800 struct mem_cgroup_reclaim_cookie *); 801 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); 802 void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 803 int (*)(struct task_struct *, void *), void *arg); 804 805 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 806 { 807 if (mem_cgroup_disabled()) 808 return 0; 809 810 return memcg->id.id; 811 } 812 struct mem_cgroup *mem_cgroup_from_id(unsigned short id); 813 814 #ifdef CONFIG_SHRINKER_DEBUG 815 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 816 { 817 return memcg ? cgroup_ino(memcg->css.cgroup) : 0; 818 } 819 820 struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino); 821 #endif 822 823 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 824 { 825 return mem_cgroup_from_css(seq_css(m)); 826 } 827 828 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 829 { 830 struct mem_cgroup_per_node *mz; 831 832 if (mem_cgroup_disabled()) 833 return NULL; 834 835 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 836 return mz->memcg; 837 } 838 839 /** 840 * parent_mem_cgroup - find the accounting parent of a memcg 841 * @memcg: memcg whose parent to find 842 * 843 * Returns the parent memcg, or NULL if this is the root. 844 */ 845 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 846 { 847 return mem_cgroup_from_css(memcg->css.parent); 848 } 849 850 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, 851 struct mem_cgroup *root) 852 { 853 if (root == memcg) 854 return true; 855 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); 856 } 857 858 static inline bool mm_match_cgroup(struct mm_struct *mm, 859 struct mem_cgroup *memcg) 860 { 861 struct mem_cgroup *task_memcg; 862 bool match = false; 863 864 rcu_read_lock(); 865 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 866 if (task_memcg) 867 match = mem_cgroup_is_descendant(task_memcg, memcg); 868 rcu_read_unlock(); 869 return match; 870 } 871 872 struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio); 873 ino_t page_cgroup_ino(struct page *page); 874 875 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 876 { 877 if (mem_cgroup_disabled()) 878 return true; 879 return !!(memcg->css.flags & CSS_ONLINE); 880 } 881 882 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, 883 int zid, int nr_pages); 884 885 static inline 886 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 887 enum lru_list lru, int zone_idx) 888 { 889 struct mem_cgroup_per_node *mz; 890 891 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 892 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); 893 } 894 895 void mem_cgroup_handle_over_high(gfp_t gfp_mask); 896 897 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); 898 899 unsigned long mem_cgroup_size(struct mem_cgroup *memcg); 900 901 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, 902 struct task_struct *p); 903 904 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); 905 906 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, 907 struct mem_cgroup *oom_domain); 908 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); 909 910 void __mod_memcg_state(struct mem_cgroup *memcg, enum memcg_stat_item idx, 911 int val); 912 913 /* idx can be of type enum memcg_stat_item or node_stat_item */ 914 static inline void mod_memcg_state(struct mem_cgroup *memcg, 915 enum memcg_stat_item idx, int val) 916 { 917 unsigned long flags; 918 919 local_irq_save(flags); 920 __mod_memcg_state(memcg, idx, val); 921 local_irq_restore(flags); 922 } 923 924 static inline void mod_memcg_page_state(struct page *page, 925 enum memcg_stat_item idx, int val) 926 { 927 struct mem_cgroup *memcg; 928 929 if (mem_cgroup_disabled()) 930 return; 931 932 rcu_read_lock(); 933 memcg = folio_memcg(page_folio(page)); 934 if (memcg) 935 mod_memcg_state(memcg, idx, val); 936 rcu_read_unlock(); 937 } 938 939 unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx); 940 unsigned long lruvec_page_state(struct lruvec *lruvec, enum node_stat_item idx); 941 unsigned long lruvec_page_state_local(struct lruvec *lruvec, 942 enum node_stat_item idx); 943 944 void mem_cgroup_flush_stats(struct mem_cgroup *memcg); 945 void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg); 946 947 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val); 948 949 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 950 int val) 951 { 952 unsigned long flags; 953 954 local_irq_save(flags); 955 __mod_lruvec_kmem_state(p, idx, val); 956 local_irq_restore(flags); 957 } 958 959 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, 960 unsigned long count); 961 962 static inline void count_memcg_events(struct mem_cgroup *memcg, 963 enum vm_event_item idx, 964 unsigned long count) 965 { 966 unsigned long flags; 967 968 local_irq_save(flags); 969 __count_memcg_events(memcg, idx, count); 970 local_irq_restore(flags); 971 } 972 973 static inline void count_memcg_folio_events(struct folio *folio, 974 enum vm_event_item idx, unsigned long nr) 975 { 976 struct mem_cgroup *memcg = folio_memcg(folio); 977 978 if (memcg) 979 count_memcg_events(memcg, idx, nr); 980 } 981 982 static inline void count_memcg_events_mm(struct mm_struct *mm, 983 enum vm_event_item idx, unsigned long count) 984 { 985 struct mem_cgroup *memcg; 986 987 if (mem_cgroup_disabled()) 988 return; 989 990 rcu_read_lock(); 991 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 992 if (likely(memcg)) 993 count_memcg_events(memcg, idx, count); 994 rcu_read_unlock(); 995 } 996 997 static inline void count_memcg_event_mm(struct mm_struct *mm, 998 enum vm_event_item idx) 999 { 1000 count_memcg_events_mm(mm, idx, 1); 1001 } 1002 1003 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1004 enum memcg_memory_event event) 1005 { 1006 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX || 1007 event == MEMCG_SWAP_FAIL; 1008 1009 atomic_long_inc(&memcg->memory_events_local[event]); 1010 if (!swap_event) 1011 cgroup_file_notify(&memcg->events_local_file); 1012 1013 do { 1014 atomic_long_inc(&memcg->memory_events[event]); 1015 if (swap_event) 1016 cgroup_file_notify(&memcg->swap_events_file); 1017 else 1018 cgroup_file_notify(&memcg->events_file); 1019 1020 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1021 break; 1022 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) 1023 break; 1024 } while ((memcg = parent_mem_cgroup(memcg)) && 1025 !mem_cgroup_is_root(memcg)); 1026 } 1027 1028 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1029 enum memcg_memory_event event) 1030 { 1031 struct mem_cgroup *memcg; 1032 1033 if (mem_cgroup_disabled()) 1034 return; 1035 1036 rcu_read_lock(); 1037 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1038 if (likely(memcg)) 1039 memcg_memory_event(memcg, event); 1040 rcu_read_unlock(); 1041 } 1042 1043 void split_page_memcg(struct page *head, int old_order, int new_order); 1044 1045 static inline u64 cgroup_id_from_mm(struct mm_struct *mm) 1046 { 1047 struct mem_cgroup *memcg; 1048 u64 id; 1049 1050 if (mem_cgroup_disabled()) 1051 return 0; 1052 1053 rcu_read_lock(); 1054 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1055 if (!memcg) 1056 memcg = root_mem_cgroup; 1057 id = cgroup_id(memcg->css.cgroup); 1058 rcu_read_unlock(); 1059 return id; 1060 } 1061 1062 #else /* CONFIG_MEMCG */ 1063 1064 #define MEM_CGROUP_ID_SHIFT 0 1065 1066 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 1067 { 1068 return NULL; 1069 } 1070 1071 static inline bool folio_memcg_charged(struct folio *folio) 1072 { 1073 return false; 1074 } 1075 1076 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio) 1077 { 1078 return NULL; 1079 } 1080 1081 static inline struct mem_cgroup *page_memcg_check(struct page *page) 1082 { 1083 return NULL; 1084 } 1085 1086 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) 1087 { 1088 return NULL; 1089 } 1090 1091 static inline bool folio_memcg_kmem(struct folio *folio) 1092 { 1093 return false; 1094 } 1095 1096 static inline bool PageMemcgKmem(struct page *page) 1097 { 1098 return false; 1099 } 1100 1101 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 1102 { 1103 return true; 1104 } 1105 1106 static inline bool mem_cgroup_disabled(void) 1107 { 1108 return true; 1109 } 1110 1111 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1112 enum memcg_memory_event event) 1113 { 1114 } 1115 1116 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1117 enum memcg_memory_event event) 1118 { 1119 } 1120 1121 static inline void mem_cgroup_protection(struct mem_cgroup *root, 1122 struct mem_cgroup *memcg, 1123 unsigned long *min, 1124 unsigned long *low) 1125 { 1126 *min = *low = 0; 1127 } 1128 1129 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, 1130 struct mem_cgroup *memcg) 1131 { 1132 } 1133 1134 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, 1135 struct mem_cgroup *memcg) 1136 { 1137 return true; 1138 } 1139 static inline bool mem_cgroup_below_low(struct mem_cgroup *target, 1140 struct mem_cgroup *memcg) 1141 { 1142 return false; 1143 } 1144 1145 static inline bool mem_cgroup_below_min(struct mem_cgroup *target, 1146 struct mem_cgroup *memcg) 1147 { 1148 return false; 1149 } 1150 1151 static inline int mem_cgroup_charge(struct folio *folio, 1152 struct mm_struct *mm, gfp_t gfp) 1153 { 1154 return 0; 1155 } 1156 1157 static inline int mem_cgroup_charge_hugetlb(struct folio* folio, gfp_t gfp) 1158 { 1159 return 0; 1160 } 1161 1162 static inline int mem_cgroup_swapin_charge_folio(struct folio *folio, 1163 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry) 1164 { 1165 return 0; 1166 } 1167 1168 static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry, unsigned int nr) 1169 { 1170 } 1171 1172 static inline void mem_cgroup_uncharge(struct folio *folio) 1173 { 1174 } 1175 1176 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios) 1177 { 1178 } 1179 1180 static inline void mem_cgroup_replace_folio(struct folio *old, 1181 struct folio *new) 1182 { 1183 } 1184 1185 static inline void mem_cgroup_migrate(struct folio *old, struct folio *new) 1186 { 1187 } 1188 1189 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 1190 struct pglist_data *pgdat) 1191 { 1192 return &pgdat->__lruvec; 1193 } 1194 1195 static inline struct lruvec *folio_lruvec(struct folio *folio) 1196 { 1197 struct pglist_data *pgdat = folio_pgdat(folio); 1198 return &pgdat->__lruvec; 1199 } 1200 1201 static inline 1202 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 1203 { 1204 } 1205 1206 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 1207 { 1208 return NULL; 1209 } 1210 1211 static inline bool mm_match_cgroup(struct mm_struct *mm, 1212 struct mem_cgroup *memcg) 1213 { 1214 return true; 1215 } 1216 1217 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) 1218 { 1219 return NULL; 1220 } 1221 1222 static inline struct mem_cgroup *get_mem_cgroup_from_current(void) 1223 { 1224 return NULL; 1225 } 1226 1227 static inline struct mem_cgroup *get_mem_cgroup_from_folio(struct folio *folio) 1228 { 1229 return NULL; 1230 } 1231 1232 static inline 1233 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css) 1234 { 1235 return NULL; 1236 } 1237 1238 static inline void obj_cgroup_get(struct obj_cgroup *objcg) 1239 { 1240 } 1241 1242 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 1243 { 1244 } 1245 1246 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg) 1247 { 1248 return true; 1249 } 1250 1251 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg) 1252 { 1253 return true; 1254 } 1255 1256 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 1257 { 1258 } 1259 1260 static inline struct lruvec *folio_lruvec_lock(struct folio *folio) 1261 { 1262 struct pglist_data *pgdat = folio_pgdat(folio); 1263 1264 spin_lock(&pgdat->__lruvec.lru_lock); 1265 return &pgdat->__lruvec; 1266 } 1267 1268 static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio) 1269 { 1270 struct pglist_data *pgdat = folio_pgdat(folio); 1271 1272 spin_lock_irq(&pgdat->__lruvec.lru_lock); 1273 return &pgdat->__lruvec; 1274 } 1275 1276 static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 1277 unsigned long *flagsp) 1278 { 1279 struct pglist_data *pgdat = folio_pgdat(folio); 1280 1281 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp); 1282 return &pgdat->__lruvec; 1283 } 1284 1285 static inline struct mem_cgroup * 1286 mem_cgroup_iter(struct mem_cgroup *root, 1287 struct mem_cgroup *prev, 1288 struct mem_cgroup_reclaim_cookie *reclaim) 1289 { 1290 return NULL; 1291 } 1292 1293 static inline void mem_cgroup_iter_break(struct mem_cgroup *root, 1294 struct mem_cgroup *prev) 1295 { 1296 } 1297 1298 static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 1299 int (*fn)(struct task_struct *, void *), void *arg) 1300 { 1301 } 1302 1303 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 1304 { 1305 return 0; 1306 } 1307 1308 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) 1309 { 1310 WARN_ON_ONCE(id); 1311 /* XXX: This should always return root_mem_cgroup */ 1312 return NULL; 1313 } 1314 1315 #ifdef CONFIG_SHRINKER_DEBUG 1316 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) 1317 { 1318 return 0; 1319 } 1320 1321 static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino) 1322 { 1323 return NULL; 1324 } 1325 #endif 1326 1327 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 1328 { 1329 return NULL; 1330 } 1331 1332 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 1333 { 1334 return NULL; 1335 } 1336 1337 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 1338 { 1339 return true; 1340 } 1341 1342 static inline 1343 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 1344 enum lru_list lru, int zone_idx) 1345 { 1346 return 0; 1347 } 1348 1349 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) 1350 { 1351 return 0; 1352 } 1353 1354 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) 1355 { 1356 return 0; 1357 } 1358 1359 static inline void 1360 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) 1361 { 1362 } 1363 1364 static inline void 1365 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) 1366 { 1367 } 1368 1369 static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask) 1370 { 1371 } 1372 1373 static inline struct mem_cgroup *mem_cgroup_get_oom_group( 1374 struct task_struct *victim, struct mem_cgroup *oom_domain) 1375 { 1376 return NULL; 1377 } 1378 1379 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) 1380 { 1381 } 1382 1383 static inline void __mod_memcg_state(struct mem_cgroup *memcg, 1384 enum memcg_stat_item idx, 1385 int nr) 1386 { 1387 } 1388 1389 static inline void mod_memcg_state(struct mem_cgroup *memcg, 1390 enum memcg_stat_item idx, 1391 int nr) 1392 { 1393 } 1394 1395 static inline void mod_memcg_page_state(struct page *page, 1396 enum memcg_stat_item idx, int val) 1397 { 1398 } 1399 1400 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) 1401 { 1402 return 0; 1403 } 1404 1405 static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1406 enum node_stat_item idx) 1407 { 1408 return node_page_state(lruvec_pgdat(lruvec), idx); 1409 } 1410 1411 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1412 enum node_stat_item idx) 1413 { 1414 return node_page_state(lruvec_pgdat(lruvec), idx); 1415 } 1416 1417 static inline void mem_cgroup_flush_stats(struct mem_cgroup *memcg) 1418 { 1419 } 1420 1421 static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg) 1422 { 1423 } 1424 1425 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1426 int val) 1427 { 1428 struct page *page = virt_to_head_page(p); 1429 1430 __mod_node_page_state(page_pgdat(page), idx, val); 1431 } 1432 1433 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1434 int val) 1435 { 1436 struct page *page = virt_to_head_page(p); 1437 1438 mod_node_page_state(page_pgdat(page), idx, val); 1439 } 1440 1441 static inline void count_memcg_events(struct mem_cgroup *memcg, 1442 enum vm_event_item idx, 1443 unsigned long count) 1444 { 1445 } 1446 1447 static inline void __count_memcg_events(struct mem_cgroup *memcg, 1448 enum vm_event_item idx, 1449 unsigned long count) 1450 { 1451 } 1452 1453 static inline void count_memcg_folio_events(struct folio *folio, 1454 enum vm_event_item idx, unsigned long nr) 1455 { 1456 } 1457 1458 static inline void count_memcg_events_mm(struct mm_struct *mm, 1459 enum vm_event_item idx, unsigned long count) 1460 { 1461 } 1462 1463 static inline 1464 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) 1465 { 1466 } 1467 1468 static inline void split_page_memcg(struct page *head, int old_order, int new_order) 1469 { 1470 } 1471 1472 static inline u64 cgroup_id_from_mm(struct mm_struct *mm) 1473 { 1474 return 0; 1475 } 1476 #endif /* CONFIG_MEMCG */ 1477 1478 /* 1479 * Extended information for slab objects stored as an array in page->memcg_data 1480 * if MEMCG_DATA_OBJEXTS is set. 1481 */ 1482 struct slabobj_ext { 1483 #ifdef CONFIG_MEMCG 1484 struct obj_cgroup *objcg; 1485 #endif 1486 #ifdef CONFIG_MEM_ALLOC_PROFILING 1487 union codetag_ref ref; 1488 #endif 1489 } __aligned(8); 1490 1491 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx) 1492 { 1493 __mod_lruvec_kmem_state(p, idx, 1); 1494 } 1495 1496 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx) 1497 { 1498 __mod_lruvec_kmem_state(p, idx, -1); 1499 } 1500 1501 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) 1502 { 1503 struct mem_cgroup *memcg; 1504 1505 memcg = lruvec_memcg(lruvec); 1506 if (!memcg) 1507 return NULL; 1508 memcg = parent_mem_cgroup(memcg); 1509 if (!memcg) 1510 return NULL; 1511 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); 1512 } 1513 1514 static inline void unlock_page_lruvec(struct lruvec *lruvec) 1515 { 1516 spin_unlock(&lruvec->lru_lock); 1517 } 1518 1519 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec) 1520 { 1521 spin_unlock_irq(&lruvec->lru_lock); 1522 } 1523 1524 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec, 1525 unsigned long flags) 1526 { 1527 spin_unlock_irqrestore(&lruvec->lru_lock, flags); 1528 } 1529 1530 /* Test requires a stable folio->memcg binding, see folio_memcg() */ 1531 static inline bool folio_matches_lruvec(struct folio *folio, 1532 struct lruvec *lruvec) 1533 { 1534 return lruvec_pgdat(lruvec) == folio_pgdat(folio) && 1535 lruvec_memcg(lruvec) == folio_memcg(folio); 1536 } 1537 1538 /* Don't lock again iff page's lruvec locked */ 1539 static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio, 1540 struct lruvec *locked_lruvec) 1541 { 1542 if (locked_lruvec) { 1543 if (folio_matches_lruvec(folio, locked_lruvec)) 1544 return locked_lruvec; 1545 1546 unlock_page_lruvec_irq(locked_lruvec); 1547 } 1548 1549 return folio_lruvec_lock_irq(folio); 1550 } 1551 1552 /* Don't lock again iff folio's lruvec locked */ 1553 static inline void folio_lruvec_relock_irqsave(struct folio *folio, 1554 struct lruvec **lruvecp, unsigned long *flags) 1555 { 1556 if (*lruvecp) { 1557 if (folio_matches_lruvec(folio, *lruvecp)) 1558 return; 1559 1560 unlock_page_lruvec_irqrestore(*lruvecp, *flags); 1561 } 1562 1563 *lruvecp = folio_lruvec_lock_irqsave(folio, flags); 1564 } 1565 1566 #ifdef CONFIG_CGROUP_WRITEBACK 1567 1568 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); 1569 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, 1570 unsigned long *pheadroom, unsigned long *pdirty, 1571 unsigned long *pwriteback); 1572 1573 void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio, 1574 struct bdi_writeback *wb); 1575 1576 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1577 struct bdi_writeback *wb) 1578 { 1579 struct mem_cgroup *memcg; 1580 1581 if (mem_cgroup_disabled()) 1582 return; 1583 1584 memcg = folio_memcg(folio); 1585 if (unlikely(memcg && &memcg->css != wb->memcg_css)) 1586 mem_cgroup_track_foreign_dirty_slowpath(folio, wb); 1587 } 1588 1589 void mem_cgroup_flush_foreign(struct bdi_writeback *wb); 1590 1591 #else /* CONFIG_CGROUP_WRITEBACK */ 1592 1593 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) 1594 { 1595 return NULL; 1596 } 1597 1598 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, 1599 unsigned long *pfilepages, 1600 unsigned long *pheadroom, 1601 unsigned long *pdirty, 1602 unsigned long *pwriteback) 1603 { 1604 } 1605 1606 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1607 struct bdi_writeback *wb) 1608 { 1609 } 1610 1611 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) 1612 { 1613 } 1614 1615 #endif /* CONFIG_CGROUP_WRITEBACK */ 1616 1617 struct sock; 1618 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages, 1619 gfp_t gfp_mask); 1620 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); 1621 #ifdef CONFIG_MEMCG 1622 extern struct static_key_false memcg_sockets_enabled_key; 1623 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) 1624 void mem_cgroup_sk_alloc(struct sock *sk); 1625 void mem_cgroup_sk_free(struct sock *sk); 1626 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1627 { 1628 #ifdef CONFIG_MEMCG_V1 1629 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1630 return !!memcg->tcpmem_pressure; 1631 #endif /* CONFIG_MEMCG_V1 */ 1632 do { 1633 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure))) 1634 return true; 1635 } while ((memcg = parent_mem_cgroup(memcg))); 1636 return false; 1637 } 1638 1639 int alloc_shrinker_info(struct mem_cgroup *memcg); 1640 void free_shrinker_info(struct mem_cgroup *memcg); 1641 void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id); 1642 void reparent_shrinker_deferred(struct mem_cgroup *memcg); 1643 #else 1644 #define mem_cgroup_sockets_enabled 0 1645 static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; 1646 static inline void mem_cgroup_sk_free(struct sock *sk) { }; 1647 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1648 { 1649 return false; 1650 } 1651 1652 static inline void set_shrinker_bit(struct mem_cgroup *memcg, 1653 int nid, int shrinker_id) 1654 { 1655 } 1656 #endif 1657 1658 #ifdef CONFIG_MEMCG 1659 bool mem_cgroup_kmem_disabled(void); 1660 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); 1661 void __memcg_kmem_uncharge_page(struct page *page, int order); 1662 1663 /* 1664 * The returned objcg pointer is safe to use without additional 1665 * protection within a scope. The scope is defined either by 1666 * the current task (similar to the "current" global variable) 1667 * or by set_active_memcg() pair. 1668 * Please, use obj_cgroup_get() to get a reference if the pointer 1669 * needs to be used outside of the local scope. 1670 */ 1671 struct obj_cgroup *current_obj_cgroup(void); 1672 struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio); 1673 1674 static inline struct obj_cgroup *get_obj_cgroup_from_current(void) 1675 { 1676 struct obj_cgroup *objcg = current_obj_cgroup(); 1677 1678 if (objcg) 1679 obj_cgroup_get(objcg); 1680 1681 return objcg; 1682 } 1683 1684 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); 1685 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); 1686 1687 extern struct static_key_false memcg_bpf_enabled_key; 1688 static inline bool memcg_bpf_enabled(void) 1689 { 1690 return static_branch_likely(&memcg_bpf_enabled_key); 1691 } 1692 1693 extern struct static_key_false memcg_kmem_online_key; 1694 1695 static inline bool memcg_kmem_online(void) 1696 { 1697 return static_branch_likely(&memcg_kmem_online_key); 1698 } 1699 1700 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1701 int order) 1702 { 1703 if (memcg_kmem_online()) 1704 return __memcg_kmem_charge_page(page, gfp, order); 1705 return 0; 1706 } 1707 1708 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1709 { 1710 if (memcg_kmem_online()) 1711 __memcg_kmem_uncharge_page(page, order); 1712 } 1713 1714 /* 1715 * A helper for accessing memcg's kmem_id, used for getting 1716 * corresponding LRU lists. 1717 */ 1718 static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1719 { 1720 return memcg ? memcg->kmemcg_id : -1; 1721 } 1722 1723 struct mem_cgroup *mem_cgroup_from_slab_obj(void *p); 1724 1725 static inline void count_objcg_events(struct obj_cgroup *objcg, 1726 enum vm_event_item idx, 1727 unsigned long count) 1728 { 1729 struct mem_cgroup *memcg; 1730 1731 if (!memcg_kmem_online()) 1732 return; 1733 1734 rcu_read_lock(); 1735 memcg = obj_cgroup_memcg(objcg); 1736 count_memcg_events(memcg, idx, count); 1737 rcu_read_unlock(); 1738 } 1739 1740 #else 1741 static inline bool mem_cgroup_kmem_disabled(void) 1742 { 1743 return true; 1744 } 1745 1746 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1747 int order) 1748 { 1749 return 0; 1750 } 1751 1752 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1753 { 1754 } 1755 1756 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1757 int order) 1758 { 1759 return 0; 1760 } 1761 1762 static inline void __memcg_kmem_uncharge_page(struct page *page, int order) 1763 { 1764 } 1765 1766 static inline struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio) 1767 { 1768 return NULL; 1769 } 1770 1771 static inline bool memcg_bpf_enabled(void) 1772 { 1773 return false; 1774 } 1775 1776 static inline bool memcg_kmem_online(void) 1777 { 1778 return false; 1779 } 1780 1781 static inline int memcg_kmem_id(struct mem_cgroup *memcg) 1782 { 1783 return -1; 1784 } 1785 1786 static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p) 1787 { 1788 return NULL; 1789 } 1790 1791 static inline void count_objcg_events(struct obj_cgroup *objcg, 1792 enum vm_event_item idx, 1793 unsigned long count) 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 static inline void mem_cgroup_enter_user_fault(void) 1840 { 1841 WARN_ON(current->in_user_fault); 1842 current->in_user_fault = 1; 1843 } 1844 1845 static inline void mem_cgroup_exit_user_fault(void) 1846 { 1847 WARN_ON(!current->in_user_fault); 1848 current->in_user_fault = 0; 1849 } 1850 1851 #else /* CONFIG_MEMCG_V1 */ 1852 static inline 1853 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order, 1854 gfp_t gfp_mask, 1855 unsigned long *total_scanned) 1856 { 1857 return 0; 1858 } 1859 1860 static inline bool task_in_memcg_oom(struct task_struct *p) 1861 { 1862 return false; 1863 } 1864 1865 static inline bool mem_cgroup_oom_synchronize(bool wait) 1866 { 1867 return false; 1868 } 1869 1870 static inline void mem_cgroup_enter_user_fault(void) 1871 { 1872 } 1873 1874 static inline void mem_cgroup_exit_user_fault(void) 1875 { 1876 } 1877 1878 #endif /* CONFIG_MEMCG_V1 */ 1879 1880 #endif /* _LINUX_MEMCONTROL_H */ 1881