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/page_counter.h> 18 #include <linux/vmpressure.h> 19 #include <linux/eventfd.h> 20 #include <linux/mm.h> 21 #include <linux/vmstat.h> 22 #include <linux/writeback.h> 23 #include <linux/page-flags.h> 24 25 struct mem_cgroup; 26 struct obj_cgroup; 27 struct page; 28 struct mm_struct; 29 struct kmem_cache; 30 31 /* Cgroup-specific page state, on top of universal node page state */ 32 enum memcg_stat_item { 33 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS, 34 MEMCG_SOCK, 35 MEMCG_PERCPU_B, 36 MEMCG_NR_STAT, 37 }; 38 39 enum memcg_memory_event { 40 MEMCG_LOW, 41 MEMCG_HIGH, 42 MEMCG_MAX, 43 MEMCG_OOM, 44 MEMCG_OOM_KILL, 45 MEMCG_SWAP_HIGH, 46 MEMCG_SWAP_MAX, 47 MEMCG_SWAP_FAIL, 48 MEMCG_NR_MEMORY_EVENTS, 49 }; 50 51 struct mem_cgroup_reclaim_cookie { 52 pg_data_t *pgdat; 53 unsigned int generation; 54 }; 55 56 #ifdef CONFIG_MEMCG 57 58 #define MEM_CGROUP_ID_SHIFT 16 59 #define MEM_CGROUP_ID_MAX USHRT_MAX 60 61 struct mem_cgroup_id { 62 int id; 63 refcount_t ref; 64 }; 65 66 /* 67 * Per memcg event counter is incremented at every pagein/pageout. With THP, 68 * it will be incremented by the number of pages. This counter is used 69 * to trigger some periodic events. This is straightforward and better 70 * than using jiffies etc. to handle periodic memcg event. 71 */ 72 enum mem_cgroup_events_target { 73 MEM_CGROUP_TARGET_THRESH, 74 MEM_CGROUP_TARGET_SOFTLIMIT, 75 MEM_CGROUP_NTARGETS, 76 }; 77 78 struct memcg_vmstats_percpu { 79 /* Local (CPU and cgroup) page state & events */ 80 long state[MEMCG_NR_STAT]; 81 unsigned long events[NR_VM_EVENT_ITEMS]; 82 83 /* Delta calculation for lockless upward propagation */ 84 long state_prev[MEMCG_NR_STAT]; 85 unsigned long events_prev[NR_VM_EVENT_ITEMS]; 86 87 /* Cgroup1: threshold notifications & softlimit tree updates */ 88 unsigned long nr_page_events; 89 unsigned long targets[MEM_CGROUP_NTARGETS]; 90 }; 91 92 struct memcg_vmstats { 93 /* Aggregated (CPU and subtree) page state & events */ 94 long state[MEMCG_NR_STAT]; 95 unsigned long events[NR_VM_EVENT_ITEMS]; 96 97 /* Pending child counts during tree propagation */ 98 long state_pending[MEMCG_NR_STAT]; 99 unsigned long events_pending[NR_VM_EVENT_ITEMS]; 100 }; 101 102 struct mem_cgroup_reclaim_iter { 103 struct mem_cgroup *position; 104 /* scan generation, increased every round-trip */ 105 unsigned int generation; 106 }; 107 108 /* 109 * Bitmap and deferred work of shrinker::id corresponding to memcg-aware 110 * shrinkers, which have elements charged to this memcg. 111 */ 112 struct shrinker_info { 113 struct rcu_head rcu; 114 atomic_long_t *nr_deferred; 115 unsigned long *map; 116 }; 117 118 struct lruvec_stats_percpu { 119 /* Local (CPU and cgroup) state */ 120 long state[NR_VM_NODE_STAT_ITEMS]; 121 122 /* Delta calculation for lockless upward propagation */ 123 long state_prev[NR_VM_NODE_STAT_ITEMS]; 124 }; 125 126 struct lruvec_stats { 127 /* Aggregated (CPU and subtree) state */ 128 long state[NR_VM_NODE_STAT_ITEMS]; 129 130 /* Pending child counts during tree propagation */ 131 long state_pending[NR_VM_NODE_STAT_ITEMS]; 132 }; 133 134 /* 135 * per-node information in memory controller. 136 */ 137 struct mem_cgroup_per_node { 138 struct lruvec lruvec; 139 140 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu; 141 struct lruvec_stats lruvec_stats; 142 143 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; 144 145 struct mem_cgroup_reclaim_iter iter; 146 147 struct shrinker_info __rcu *shrinker_info; 148 149 struct rb_node tree_node; /* RB tree node */ 150 unsigned long usage_in_excess;/* Set to the value by which */ 151 /* the soft limit is exceeded*/ 152 bool on_tree; 153 struct mem_cgroup *memcg; /* Back pointer, we cannot */ 154 /* use container_of */ 155 }; 156 157 struct mem_cgroup_threshold { 158 struct eventfd_ctx *eventfd; 159 unsigned long threshold; 160 }; 161 162 /* For threshold */ 163 struct mem_cgroup_threshold_ary { 164 /* An array index points to threshold just below or equal to usage. */ 165 int current_threshold; 166 /* Size of entries[] */ 167 unsigned int size; 168 /* Array of thresholds */ 169 struct mem_cgroup_threshold entries[]; 170 }; 171 172 struct mem_cgroup_thresholds { 173 /* Primary thresholds array */ 174 struct mem_cgroup_threshold_ary *primary; 175 /* 176 * Spare threshold array. 177 * This is needed to make mem_cgroup_unregister_event() "never fail". 178 * It must be able to store at least primary->size - 1 entries. 179 */ 180 struct mem_cgroup_threshold_ary *spare; 181 }; 182 183 enum memcg_kmem_state { 184 KMEM_NONE, 185 KMEM_ALLOCATED, 186 KMEM_ONLINE, 187 }; 188 189 #if defined(CONFIG_SMP) 190 struct memcg_padding { 191 char x[0]; 192 } ____cacheline_internodealigned_in_smp; 193 #define MEMCG_PADDING(name) struct memcg_padding name 194 #else 195 #define MEMCG_PADDING(name) 196 #endif 197 198 /* 199 * Remember four most recent foreign writebacks with dirty pages in this 200 * cgroup. Inode sharing is expected to be uncommon and, even if we miss 201 * one in a given round, we're likely to catch it later if it keeps 202 * foreign-dirtying, so a fairly low count should be enough. 203 * 204 * See mem_cgroup_track_foreign_dirty_slowpath() for details. 205 */ 206 #define MEMCG_CGWB_FRN_CNT 4 207 208 struct memcg_cgwb_frn { 209 u64 bdi_id; /* bdi->id of the foreign inode */ 210 int memcg_id; /* memcg->css.id of foreign inode */ 211 u64 at; /* jiffies_64 at the time of dirtying */ 212 struct wb_completion done; /* tracks in-flight foreign writebacks */ 213 }; 214 215 /* 216 * Bucket for arbitrarily byte-sized objects charged to a memory 217 * cgroup. The bucket can be reparented in one piece when the cgroup 218 * is destroyed, without having to round up the individual references 219 * of all live memory objects in the wild. 220 */ 221 struct obj_cgroup { 222 struct percpu_ref refcnt; 223 struct mem_cgroup *memcg; 224 atomic_t nr_charged_bytes; 225 union { 226 struct list_head list; 227 struct rcu_head rcu; 228 }; 229 }; 230 231 /* 232 * The memory controller data structure. The memory controller controls both 233 * page cache and RSS per cgroup. We would eventually like to provide 234 * statistics based on the statistics developed by Rik Van Riel for clock-pro, 235 * to help the administrator determine what knobs to tune. 236 */ 237 struct mem_cgroup { 238 struct cgroup_subsys_state css; 239 240 /* Private memcg ID. Used to ID objects that outlive the cgroup */ 241 struct mem_cgroup_id id; 242 243 /* Accounted resources */ 244 struct page_counter memory; /* Both v1 & v2 */ 245 246 union { 247 struct page_counter swap; /* v2 only */ 248 struct page_counter memsw; /* v1 only */ 249 }; 250 251 /* Legacy consumer-oriented counters */ 252 struct page_counter kmem; /* v1 only */ 253 struct page_counter tcpmem; /* v1 only */ 254 255 /* Range enforcement for interrupt charges */ 256 struct work_struct high_work; 257 258 unsigned long soft_limit; 259 260 /* vmpressure notifications */ 261 struct vmpressure vmpressure; 262 263 /* 264 * Should the OOM killer kill all belonging tasks, had it kill one? 265 */ 266 bool oom_group; 267 268 /* protected by memcg_oom_lock */ 269 bool oom_lock; 270 int under_oom; 271 272 int swappiness; 273 /* OOM-Killer disable */ 274 int oom_kill_disable; 275 276 /* memory.events and memory.events.local */ 277 struct cgroup_file events_file; 278 struct cgroup_file events_local_file; 279 280 /* handle for "memory.swap.events" */ 281 struct cgroup_file swap_events_file; 282 283 /* protect arrays of thresholds */ 284 struct mutex thresholds_lock; 285 286 /* thresholds for memory usage. RCU-protected */ 287 struct mem_cgroup_thresholds thresholds; 288 289 /* thresholds for mem+swap usage. RCU-protected */ 290 struct mem_cgroup_thresholds memsw_thresholds; 291 292 /* For oom notifier event fd */ 293 struct list_head oom_notify; 294 295 /* 296 * Should we move charges of a task when a task is moved into this 297 * mem_cgroup ? And what type of charges should we move ? 298 */ 299 unsigned long move_charge_at_immigrate; 300 /* taken only while moving_account > 0 */ 301 spinlock_t move_lock; 302 unsigned long move_lock_flags; 303 304 MEMCG_PADDING(_pad1_); 305 306 /* memory.stat */ 307 struct memcg_vmstats vmstats; 308 309 /* memory.events */ 310 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS]; 311 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS]; 312 313 unsigned long socket_pressure; 314 315 /* Legacy tcp memory accounting */ 316 bool tcpmem_active; 317 int tcpmem_pressure; 318 319 #ifdef CONFIG_MEMCG_KMEM 320 int kmemcg_id; 321 enum memcg_kmem_state kmem_state; 322 struct obj_cgroup __rcu *objcg; 323 struct list_head objcg_list; /* list of inherited objcgs */ 324 #endif 325 326 MEMCG_PADDING(_pad2_); 327 328 /* 329 * set > 0 if pages under this cgroup are moving to other cgroup. 330 */ 331 atomic_t moving_account; 332 struct task_struct *move_lock_task; 333 334 struct memcg_vmstats_percpu __percpu *vmstats_percpu; 335 336 #ifdef CONFIG_CGROUP_WRITEBACK 337 struct list_head cgwb_list; 338 struct wb_domain cgwb_domain; 339 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT]; 340 #endif 341 342 /* List of events which userspace want to receive */ 343 struct list_head event_list; 344 spinlock_t event_list_lock; 345 346 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 347 struct deferred_split deferred_split_queue; 348 #endif 349 350 struct mem_cgroup_per_node *nodeinfo[]; 351 }; 352 353 /* 354 * size of first charge trial. "32" comes from vmscan.c's magic value. 355 * TODO: maybe necessary to use big numbers in big irons. 356 */ 357 #define MEMCG_CHARGE_BATCH 32U 358 359 extern struct mem_cgroup *root_mem_cgroup; 360 361 enum page_memcg_data_flags { 362 /* page->memcg_data is a pointer to an objcgs vector */ 363 MEMCG_DATA_OBJCGS = (1UL << 0), 364 /* page has been accounted as a non-slab kernel page */ 365 MEMCG_DATA_KMEM = (1UL << 1), 366 /* the next bit after the last actual flag */ 367 __NR_MEMCG_DATA_FLAGS = (1UL << 2), 368 }; 369 370 #define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1) 371 372 static inline bool folio_memcg_kmem(struct folio *folio); 373 374 /* 375 * After the initialization objcg->memcg is always pointing at 376 * a valid memcg, but can be atomically swapped to the parent memcg. 377 * 378 * The caller must ensure that the returned memcg won't be released: 379 * e.g. acquire the rcu_read_lock or css_set_lock. 380 */ 381 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg) 382 { 383 return READ_ONCE(objcg->memcg); 384 } 385 386 /* 387 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio 388 * @folio: Pointer to the folio. 389 * 390 * Returns a pointer to the memory cgroup associated with the folio, 391 * or NULL. This function assumes that the folio is known to have a 392 * proper memory 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 * kmem folios. 395 */ 396 static inline struct mem_cgroup *__folio_memcg(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_OBJCGS, folio); 402 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio); 403 404 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 405 } 406 407 /* 408 * __folio_objcg - get the object cgroup associated with a kmem folio. 409 * @folio: Pointer to the folio. 410 * 411 * Returns a pointer to the object cgroup associated with the folio, 412 * or NULL. This function assumes that the folio is known to have a 413 * proper object cgroup pointer. It's not safe to call this function 414 * against some type of folios, e.g. slab folios or ex-slab folios or 415 * LRU folios. 416 */ 417 static inline struct obj_cgroup *__folio_objcg(struct folio *folio) 418 { 419 unsigned long memcg_data = folio->memcg_data; 420 421 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 422 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio); 423 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio); 424 425 return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 426 } 427 428 /* 429 * folio_memcg - Get the memory cgroup associated with a folio. 430 * @folio: Pointer to the folio. 431 * 432 * Returns a pointer to the memory cgroup associated with the folio, 433 * or NULL. This function assumes that the folio is known to have a 434 * proper memory cgroup pointer. It's not safe to call this function 435 * against some type of folios, e.g. slab folios or ex-slab folios. 436 * 437 * For a non-kmem folio any of the following ensures folio and memcg binding 438 * stability: 439 * 440 * - the folio lock 441 * - LRU isolation 442 * - lock_page_memcg() 443 * - exclusive reference 444 * 445 * For a kmem folio a caller should hold an rcu read lock to protect memcg 446 * associated with a kmem folio from being released. 447 */ 448 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 449 { 450 if (folio_memcg_kmem(folio)) 451 return obj_cgroup_memcg(__folio_objcg(folio)); 452 return __folio_memcg(folio); 453 } 454 455 static inline struct mem_cgroup *page_memcg(struct page *page) 456 { 457 return folio_memcg(page_folio(page)); 458 } 459 460 /** 461 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio. 462 * @folio: Pointer to the folio. 463 * 464 * This function assumes that the folio is known to have a 465 * proper memory cgroup pointer. It's not safe to call this function 466 * against some type of folios, e.g. slab folios or ex-slab folios. 467 * 468 * Return: A pointer to the memory cgroup associated with the folio, 469 * or NULL. 470 */ 471 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 472 { 473 unsigned long memcg_data = READ_ONCE(folio->memcg_data); 474 475 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); 476 WARN_ON_ONCE(!rcu_read_lock_held()); 477 478 if (memcg_data & MEMCG_DATA_KMEM) { 479 struct obj_cgroup *objcg; 480 481 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 482 return obj_cgroup_memcg(objcg); 483 } 484 485 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 486 } 487 488 /* 489 * page_memcg_check - get the memory cgroup associated with a page 490 * @page: a pointer to the page struct 491 * 492 * Returns a pointer to the memory cgroup associated with the page, 493 * or NULL. This function unlike page_memcg() can take any page 494 * as an argument. It has to be used in cases when it's not known if a page 495 * has an associated memory cgroup pointer or an object cgroups vector or 496 * an object cgroup. 497 * 498 * For a non-kmem page any of the following ensures page and memcg binding 499 * stability: 500 * 501 * - the page lock 502 * - LRU isolation 503 * - lock_page_memcg() 504 * - exclusive reference 505 * 506 * For a kmem page a caller should hold an rcu read lock to protect memcg 507 * associated with a kmem page from being released. 508 */ 509 static inline struct mem_cgroup *page_memcg_check(struct page *page) 510 { 511 /* 512 * Because page->memcg_data might be changed asynchronously 513 * for slab pages, READ_ONCE() should be used here. 514 */ 515 unsigned long memcg_data = READ_ONCE(page->memcg_data); 516 517 if (memcg_data & MEMCG_DATA_OBJCGS) 518 return NULL; 519 520 if (memcg_data & MEMCG_DATA_KMEM) { 521 struct obj_cgroup *objcg; 522 523 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 524 return obj_cgroup_memcg(objcg); 525 } 526 527 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 528 } 529 530 #ifdef CONFIG_MEMCG_KMEM 531 /* 532 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set. 533 * @folio: Pointer to the folio. 534 * 535 * Checks if the folio has MemcgKmem flag set. The caller must ensure 536 * that the folio has an associated memory cgroup. It's not safe to call 537 * this function against some types of folios, e.g. slab folios. 538 */ 539 static inline bool folio_memcg_kmem(struct folio *folio) 540 { 541 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page); 542 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJCGS, folio); 543 return folio->memcg_data & MEMCG_DATA_KMEM; 544 } 545 546 /* 547 * page_objcgs - get the object cgroups vector associated with a page 548 * @page: a pointer to the page struct 549 * 550 * Returns a pointer to the object cgroups vector associated with the page, 551 * or NULL. This function assumes that the page is known to have an 552 * associated object cgroups vector. It's not safe to call this function 553 * against pages, which might have an associated memory cgroup: e.g. 554 * kernel stack pages. 555 */ 556 static inline struct obj_cgroup **page_objcgs(struct page *page) 557 { 558 unsigned long memcg_data = READ_ONCE(page->memcg_data); 559 560 VM_BUG_ON_PAGE(memcg_data && !(memcg_data & MEMCG_DATA_OBJCGS), page); 561 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page); 562 563 return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 564 } 565 566 /* 567 * page_objcgs_check - get the object cgroups vector associated with a page 568 * @page: a pointer to the page struct 569 * 570 * Returns a pointer to the object cgroups vector associated with the page, 571 * or NULL. This function is safe to use if the page can be directly associated 572 * with a memory cgroup. 573 */ 574 static inline struct obj_cgroup **page_objcgs_check(struct page *page) 575 { 576 unsigned long memcg_data = READ_ONCE(page->memcg_data); 577 578 if (!memcg_data || !(memcg_data & MEMCG_DATA_OBJCGS)) 579 return NULL; 580 581 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page); 582 583 return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); 584 } 585 586 #else 587 static inline bool folio_memcg_kmem(struct folio *folio) 588 { 589 return false; 590 } 591 592 static inline struct obj_cgroup **page_objcgs(struct page *page) 593 { 594 return NULL; 595 } 596 597 static inline struct obj_cgroup **page_objcgs_check(struct page *page) 598 { 599 return NULL; 600 } 601 #endif 602 603 static inline bool PageMemcgKmem(struct page *page) 604 { 605 return folio_memcg_kmem(page_folio(page)); 606 } 607 608 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 609 { 610 return (memcg == root_mem_cgroup); 611 } 612 613 static inline bool mem_cgroup_disabled(void) 614 { 615 return !cgroup_subsys_enabled(memory_cgrp_subsys); 616 } 617 618 static inline void mem_cgroup_protection(struct mem_cgroup *root, 619 struct mem_cgroup *memcg, 620 unsigned long *min, 621 unsigned long *low) 622 { 623 *min = *low = 0; 624 625 if (mem_cgroup_disabled()) 626 return; 627 628 /* 629 * There is no reclaim protection applied to a targeted reclaim. 630 * We are special casing this specific case here because 631 * mem_cgroup_protected calculation is not robust enough to keep 632 * the protection invariant for calculated effective values for 633 * parallel reclaimers with different reclaim target. This is 634 * especially a problem for tail memcgs (as they have pages on LRU) 635 * which would want to have effective values 0 for targeted reclaim 636 * but a different value for external reclaim. 637 * 638 * Example 639 * Let's have global and A's reclaim in parallel: 640 * | 641 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) 642 * |\ 643 * | C (low = 1G, usage = 2.5G) 644 * B (low = 1G, usage = 0.5G) 645 * 646 * For the global reclaim 647 * A.elow = A.low 648 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow 649 * C.elow = min(C.usage, C.low) 650 * 651 * With the effective values resetting we have A reclaim 652 * A.elow = 0 653 * B.elow = B.low 654 * C.elow = C.low 655 * 656 * If the global reclaim races with A's reclaim then 657 * B.elow = C.elow = 0 because children_low_usage > A.elow) 658 * is possible and reclaiming B would be violating the protection. 659 * 660 */ 661 if (root == memcg) 662 return; 663 664 *min = READ_ONCE(memcg->memory.emin); 665 *low = READ_ONCE(memcg->memory.elow); 666 } 667 668 void mem_cgroup_calculate_protection(struct mem_cgroup *root, 669 struct mem_cgroup *memcg); 670 671 static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg) 672 { 673 /* 674 * The root memcg doesn't account charges, and doesn't support 675 * protection. 676 */ 677 return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg); 678 679 } 680 681 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg) 682 { 683 if (!mem_cgroup_supports_protection(memcg)) 684 return false; 685 686 return READ_ONCE(memcg->memory.elow) >= 687 page_counter_read(&memcg->memory); 688 } 689 690 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg) 691 { 692 if (!mem_cgroup_supports_protection(memcg)) 693 return false; 694 695 return READ_ONCE(memcg->memory.emin) >= 696 page_counter_read(&memcg->memory); 697 } 698 699 int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp); 700 701 /** 702 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup. 703 * @folio: Folio to charge. 704 * @mm: mm context of the allocating task. 705 * @gfp: Reclaim mode. 706 * 707 * Try to charge @folio to the memcg that @mm belongs to, reclaiming 708 * pages according to @gfp if necessary. If @mm is NULL, try to 709 * charge to the active memcg. 710 * 711 * Do not use this for folios allocated for swapin. 712 * 713 * Return: 0 on success. Otherwise, an error code is returned. 714 */ 715 static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, 716 gfp_t gfp) 717 { 718 if (mem_cgroup_disabled()) 719 return 0; 720 return __mem_cgroup_charge(folio, mm, gfp); 721 } 722 723 int mem_cgroup_swapin_charge_page(struct page *page, struct mm_struct *mm, 724 gfp_t gfp, swp_entry_t entry); 725 void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry); 726 727 void __mem_cgroup_uncharge(struct folio *folio); 728 729 /** 730 * mem_cgroup_uncharge - Uncharge a folio. 731 * @folio: Folio to uncharge. 732 * 733 * Uncharge a folio previously charged with mem_cgroup_charge(). 734 */ 735 static inline void mem_cgroup_uncharge(struct folio *folio) 736 { 737 if (mem_cgroup_disabled()) 738 return; 739 __mem_cgroup_uncharge(folio); 740 } 741 742 void __mem_cgroup_uncharge_list(struct list_head *page_list); 743 static inline void mem_cgroup_uncharge_list(struct list_head *page_list) 744 { 745 if (mem_cgroup_disabled()) 746 return; 747 __mem_cgroup_uncharge_list(page_list); 748 } 749 750 void mem_cgroup_migrate(struct folio *old, struct folio *new); 751 752 /** 753 * mem_cgroup_lruvec - get the lru list vector for a memcg & node 754 * @memcg: memcg of the wanted lruvec 755 * @pgdat: pglist_data 756 * 757 * Returns the lru list vector holding pages for a given @memcg & 758 * @pgdat combination. This can be the node lruvec, if the memory 759 * controller is disabled. 760 */ 761 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 762 struct pglist_data *pgdat) 763 { 764 struct mem_cgroup_per_node *mz; 765 struct lruvec *lruvec; 766 767 if (mem_cgroup_disabled()) { 768 lruvec = &pgdat->__lruvec; 769 goto out; 770 } 771 772 if (!memcg) 773 memcg = root_mem_cgroup; 774 775 mz = memcg->nodeinfo[pgdat->node_id]; 776 lruvec = &mz->lruvec; 777 out: 778 /* 779 * Since a node can be onlined after the mem_cgroup was created, 780 * we have to be prepared to initialize lruvec->pgdat here; 781 * and if offlined then reonlined, we need to reinitialize it. 782 */ 783 if (unlikely(lruvec->pgdat != pgdat)) 784 lruvec->pgdat = pgdat; 785 return lruvec; 786 } 787 788 /** 789 * folio_lruvec - return lruvec for isolating/putting an LRU folio 790 * @folio: Pointer to the folio. 791 * 792 * This function relies on folio->mem_cgroup being stable. 793 */ 794 static inline struct lruvec *folio_lruvec(struct folio *folio) 795 { 796 struct mem_cgroup *memcg = folio_memcg(folio); 797 798 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio); 799 return mem_cgroup_lruvec(memcg, folio_pgdat(folio)); 800 } 801 802 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); 803 804 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); 805 806 struct lruvec *folio_lruvec_lock(struct folio *folio); 807 struct lruvec *folio_lruvec_lock_irq(struct folio *folio); 808 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 809 unsigned long *flags); 810 811 #ifdef CONFIG_DEBUG_VM 812 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio); 813 #else 814 static inline 815 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 816 { 817 } 818 #endif 819 820 static inline 821 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ 822 return css ? container_of(css, struct mem_cgroup, css) : NULL; 823 } 824 825 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) 826 { 827 return percpu_ref_tryget(&objcg->refcnt); 828 } 829 830 static inline void obj_cgroup_get(struct obj_cgroup *objcg) 831 { 832 percpu_ref_get(&objcg->refcnt); 833 } 834 835 static inline void obj_cgroup_get_many(struct obj_cgroup *objcg, 836 unsigned long nr) 837 { 838 percpu_ref_get_many(&objcg->refcnt, nr); 839 } 840 841 static inline void obj_cgroup_put(struct obj_cgroup *objcg) 842 { 843 percpu_ref_put(&objcg->refcnt); 844 } 845 846 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 847 { 848 if (memcg) 849 css_put(&memcg->css); 850 } 851 852 #define mem_cgroup_from_counter(counter, member) \ 853 container_of(counter, struct mem_cgroup, member) 854 855 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, 856 struct mem_cgroup *, 857 struct mem_cgroup_reclaim_cookie *); 858 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); 859 int mem_cgroup_scan_tasks(struct mem_cgroup *, 860 int (*)(struct task_struct *, void *), void *); 861 862 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 863 { 864 if (mem_cgroup_disabled()) 865 return 0; 866 867 return memcg->id.id; 868 } 869 struct mem_cgroup *mem_cgroup_from_id(unsigned short id); 870 871 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 872 { 873 return mem_cgroup_from_css(seq_css(m)); 874 } 875 876 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 877 { 878 struct mem_cgroup_per_node *mz; 879 880 if (mem_cgroup_disabled()) 881 return NULL; 882 883 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 884 return mz->memcg; 885 } 886 887 /** 888 * parent_mem_cgroup - find the accounting parent of a memcg 889 * @memcg: memcg whose parent to find 890 * 891 * Returns the parent memcg, or NULL if this is the root or the memory 892 * controller is in legacy no-hierarchy mode. 893 */ 894 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 895 { 896 if (!memcg->memory.parent) 897 return NULL; 898 return mem_cgroup_from_counter(memcg->memory.parent, memory); 899 } 900 901 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, 902 struct mem_cgroup *root) 903 { 904 if (root == memcg) 905 return true; 906 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); 907 } 908 909 static inline bool mm_match_cgroup(struct mm_struct *mm, 910 struct mem_cgroup *memcg) 911 { 912 struct mem_cgroup *task_memcg; 913 bool match = false; 914 915 rcu_read_lock(); 916 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 917 if (task_memcg) 918 match = mem_cgroup_is_descendant(task_memcg, memcg); 919 rcu_read_unlock(); 920 return match; 921 } 922 923 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); 924 ino_t page_cgroup_ino(struct page *page); 925 926 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 927 { 928 if (mem_cgroup_disabled()) 929 return true; 930 return !!(memcg->css.flags & CSS_ONLINE); 931 } 932 933 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, 934 int zid, int nr_pages); 935 936 static inline 937 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 938 enum lru_list lru, int zone_idx) 939 { 940 struct mem_cgroup_per_node *mz; 941 942 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 943 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); 944 } 945 946 void mem_cgroup_handle_over_high(void); 947 948 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); 949 950 unsigned long mem_cgroup_size(struct mem_cgroup *memcg); 951 952 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, 953 struct task_struct *p); 954 955 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); 956 957 static inline void mem_cgroup_enter_user_fault(void) 958 { 959 WARN_ON(current->in_user_fault); 960 current->in_user_fault = 1; 961 } 962 963 static inline void mem_cgroup_exit_user_fault(void) 964 { 965 WARN_ON(!current->in_user_fault); 966 current->in_user_fault = 0; 967 } 968 969 static inline bool task_in_memcg_oom(struct task_struct *p) 970 { 971 return p->memcg_in_oom; 972 } 973 974 bool mem_cgroup_oom_synchronize(bool wait); 975 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, 976 struct mem_cgroup *oom_domain); 977 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); 978 979 #ifdef CONFIG_MEMCG_SWAP 980 extern bool cgroup_memory_noswap; 981 #endif 982 983 void folio_memcg_lock(struct folio *folio); 984 void folio_memcg_unlock(struct folio *folio); 985 void lock_page_memcg(struct page *page); 986 void unlock_page_memcg(struct page *page); 987 988 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val); 989 990 /* idx can be of type enum memcg_stat_item or node_stat_item */ 991 static inline void mod_memcg_state(struct mem_cgroup *memcg, 992 int idx, int val) 993 { 994 unsigned long flags; 995 996 local_irq_save(flags); 997 __mod_memcg_state(memcg, idx, val); 998 local_irq_restore(flags); 999 } 1000 1001 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) 1002 { 1003 return READ_ONCE(memcg->vmstats.state[idx]); 1004 } 1005 1006 static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1007 enum node_stat_item idx) 1008 { 1009 struct mem_cgroup_per_node *pn; 1010 1011 if (mem_cgroup_disabled()) 1012 return node_page_state(lruvec_pgdat(lruvec), idx); 1013 1014 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 1015 return READ_ONCE(pn->lruvec_stats.state[idx]); 1016 } 1017 1018 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1019 enum node_stat_item idx) 1020 { 1021 struct mem_cgroup_per_node *pn; 1022 long x = 0; 1023 int cpu; 1024 1025 if (mem_cgroup_disabled()) 1026 return node_page_state(lruvec_pgdat(lruvec), idx); 1027 1028 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); 1029 for_each_possible_cpu(cpu) 1030 x += per_cpu(pn->lruvec_stats_percpu->state[idx], cpu); 1031 #ifdef CONFIG_SMP 1032 if (x < 0) 1033 x = 0; 1034 #endif 1035 return x; 1036 } 1037 1038 void mem_cgroup_flush_stats(void); 1039 1040 void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx, 1041 int val); 1042 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val); 1043 1044 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1045 int val) 1046 { 1047 unsigned long flags; 1048 1049 local_irq_save(flags); 1050 __mod_lruvec_kmem_state(p, idx, val); 1051 local_irq_restore(flags); 1052 } 1053 1054 static inline void mod_memcg_lruvec_state(struct lruvec *lruvec, 1055 enum node_stat_item idx, int val) 1056 { 1057 unsigned long flags; 1058 1059 local_irq_save(flags); 1060 __mod_memcg_lruvec_state(lruvec, idx, val); 1061 local_irq_restore(flags); 1062 } 1063 1064 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, 1065 unsigned long count); 1066 1067 static inline void count_memcg_events(struct mem_cgroup *memcg, 1068 enum vm_event_item idx, 1069 unsigned long count) 1070 { 1071 unsigned long flags; 1072 1073 local_irq_save(flags); 1074 __count_memcg_events(memcg, idx, count); 1075 local_irq_restore(flags); 1076 } 1077 1078 static inline void count_memcg_page_event(struct page *page, 1079 enum vm_event_item idx) 1080 { 1081 struct mem_cgroup *memcg = page_memcg(page); 1082 1083 if (memcg) 1084 count_memcg_events(memcg, idx, 1); 1085 } 1086 1087 static inline void count_memcg_event_mm(struct mm_struct *mm, 1088 enum vm_event_item idx) 1089 { 1090 struct mem_cgroup *memcg; 1091 1092 if (mem_cgroup_disabled()) 1093 return; 1094 1095 rcu_read_lock(); 1096 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1097 if (likely(memcg)) 1098 count_memcg_events(memcg, idx, 1); 1099 rcu_read_unlock(); 1100 } 1101 1102 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1103 enum memcg_memory_event event) 1104 { 1105 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX || 1106 event == MEMCG_SWAP_FAIL; 1107 1108 atomic_long_inc(&memcg->memory_events_local[event]); 1109 if (!swap_event) 1110 cgroup_file_notify(&memcg->events_local_file); 1111 1112 do { 1113 atomic_long_inc(&memcg->memory_events[event]); 1114 if (swap_event) 1115 cgroup_file_notify(&memcg->swap_events_file); 1116 else 1117 cgroup_file_notify(&memcg->events_file); 1118 1119 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) 1120 break; 1121 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) 1122 break; 1123 } while ((memcg = parent_mem_cgroup(memcg)) && 1124 !mem_cgroup_is_root(memcg)); 1125 } 1126 1127 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1128 enum memcg_memory_event event) 1129 { 1130 struct mem_cgroup *memcg; 1131 1132 if (mem_cgroup_disabled()) 1133 return; 1134 1135 rcu_read_lock(); 1136 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); 1137 if (likely(memcg)) 1138 memcg_memory_event(memcg, event); 1139 rcu_read_unlock(); 1140 } 1141 1142 void split_page_memcg(struct page *head, unsigned int nr); 1143 1144 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, 1145 gfp_t gfp_mask, 1146 unsigned long *total_scanned); 1147 1148 #else /* CONFIG_MEMCG */ 1149 1150 #define MEM_CGROUP_ID_SHIFT 0 1151 #define MEM_CGROUP_ID_MAX 0 1152 1153 static inline struct mem_cgroup *folio_memcg(struct folio *folio) 1154 { 1155 return NULL; 1156 } 1157 1158 static inline struct mem_cgroup *page_memcg(struct page *page) 1159 { 1160 return NULL; 1161 } 1162 1163 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) 1164 { 1165 WARN_ON_ONCE(!rcu_read_lock_held()); 1166 return NULL; 1167 } 1168 1169 static inline struct mem_cgroup *page_memcg_check(struct page *page) 1170 { 1171 return NULL; 1172 } 1173 1174 static inline bool folio_memcg_kmem(struct folio *folio) 1175 { 1176 return false; 1177 } 1178 1179 static inline bool PageMemcgKmem(struct page *page) 1180 { 1181 return false; 1182 } 1183 1184 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) 1185 { 1186 return true; 1187 } 1188 1189 static inline bool mem_cgroup_disabled(void) 1190 { 1191 return true; 1192 } 1193 1194 static inline void memcg_memory_event(struct mem_cgroup *memcg, 1195 enum memcg_memory_event event) 1196 { 1197 } 1198 1199 static inline void memcg_memory_event_mm(struct mm_struct *mm, 1200 enum memcg_memory_event event) 1201 { 1202 } 1203 1204 static inline void mem_cgroup_protection(struct mem_cgroup *root, 1205 struct mem_cgroup *memcg, 1206 unsigned long *min, 1207 unsigned long *low) 1208 { 1209 *min = *low = 0; 1210 } 1211 1212 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, 1213 struct mem_cgroup *memcg) 1214 { 1215 } 1216 1217 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg) 1218 { 1219 return false; 1220 } 1221 1222 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg) 1223 { 1224 return false; 1225 } 1226 1227 static inline int mem_cgroup_charge(struct folio *folio, 1228 struct mm_struct *mm, gfp_t gfp) 1229 { 1230 return 0; 1231 } 1232 1233 static inline int mem_cgroup_swapin_charge_page(struct page *page, 1234 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry) 1235 { 1236 return 0; 1237 } 1238 1239 static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry) 1240 { 1241 } 1242 1243 static inline void mem_cgroup_uncharge(struct folio *folio) 1244 { 1245 } 1246 1247 static inline void mem_cgroup_uncharge_list(struct list_head *page_list) 1248 { 1249 } 1250 1251 static inline void mem_cgroup_migrate(struct folio *old, struct folio *new) 1252 { 1253 } 1254 1255 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, 1256 struct pglist_data *pgdat) 1257 { 1258 return &pgdat->__lruvec; 1259 } 1260 1261 static inline struct lruvec *folio_lruvec(struct folio *folio) 1262 { 1263 struct pglist_data *pgdat = folio_pgdat(folio); 1264 return &pgdat->__lruvec; 1265 } 1266 1267 static inline 1268 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) 1269 { 1270 } 1271 1272 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) 1273 { 1274 return NULL; 1275 } 1276 1277 static inline bool mm_match_cgroup(struct mm_struct *mm, 1278 struct mem_cgroup *memcg) 1279 { 1280 return true; 1281 } 1282 1283 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) 1284 { 1285 return NULL; 1286 } 1287 1288 static inline 1289 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css) 1290 { 1291 return NULL; 1292 } 1293 1294 static inline void mem_cgroup_put(struct mem_cgroup *memcg) 1295 { 1296 } 1297 1298 static inline struct lruvec *folio_lruvec_lock(struct folio *folio) 1299 { 1300 struct pglist_data *pgdat = folio_pgdat(folio); 1301 1302 spin_lock(&pgdat->__lruvec.lru_lock); 1303 return &pgdat->__lruvec; 1304 } 1305 1306 static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio) 1307 { 1308 struct pglist_data *pgdat = folio_pgdat(folio); 1309 1310 spin_lock_irq(&pgdat->__lruvec.lru_lock); 1311 return &pgdat->__lruvec; 1312 } 1313 1314 static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, 1315 unsigned long *flagsp) 1316 { 1317 struct pglist_data *pgdat = folio_pgdat(folio); 1318 1319 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp); 1320 return &pgdat->__lruvec; 1321 } 1322 1323 static inline struct mem_cgroup * 1324 mem_cgroup_iter(struct mem_cgroup *root, 1325 struct mem_cgroup *prev, 1326 struct mem_cgroup_reclaim_cookie *reclaim) 1327 { 1328 return NULL; 1329 } 1330 1331 static inline void mem_cgroup_iter_break(struct mem_cgroup *root, 1332 struct mem_cgroup *prev) 1333 { 1334 } 1335 1336 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg, 1337 int (*fn)(struct task_struct *, void *), void *arg) 1338 { 1339 return 0; 1340 } 1341 1342 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) 1343 { 1344 return 0; 1345 } 1346 1347 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) 1348 { 1349 WARN_ON_ONCE(id); 1350 /* XXX: This should always return root_mem_cgroup */ 1351 return NULL; 1352 } 1353 1354 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) 1355 { 1356 return NULL; 1357 } 1358 1359 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) 1360 { 1361 return NULL; 1362 } 1363 1364 static inline bool mem_cgroup_online(struct mem_cgroup *memcg) 1365 { 1366 return true; 1367 } 1368 1369 static inline 1370 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, 1371 enum lru_list lru, int zone_idx) 1372 { 1373 return 0; 1374 } 1375 1376 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) 1377 { 1378 return 0; 1379 } 1380 1381 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) 1382 { 1383 return 0; 1384 } 1385 1386 static inline void 1387 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) 1388 { 1389 } 1390 1391 static inline void 1392 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) 1393 { 1394 } 1395 1396 static inline void lock_page_memcg(struct page *page) 1397 { 1398 } 1399 1400 static inline void unlock_page_memcg(struct page *page) 1401 { 1402 } 1403 1404 static inline void folio_memcg_lock(struct folio *folio) 1405 { 1406 } 1407 1408 static inline void folio_memcg_unlock(struct folio *folio) 1409 { 1410 } 1411 1412 static inline void mem_cgroup_handle_over_high(void) 1413 { 1414 } 1415 1416 static inline void mem_cgroup_enter_user_fault(void) 1417 { 1418 } 1419 1420 static inline void mem_cgroup_exit_user_fault(void) 1421 { 1422 } 1423 1424 static inline bool task_in_memcg_oom(struct task_struct *p) 1425 { 1426 return false; 1427 } 1428 1429 static inline bool mem_cgroup_oom_synchronize(bool wait) 1430 { 1431 return false; 1432 } 1433 1434 static inline struct mem_cgroup *mem_cgroup_get_oom_group( 1435 struct task_struct *victim, struct mem_cgroup *oom_domain) 1436 { 1437 return NULL; 1438 } 1439 1440 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) 1441 { 1442 } 1443 1444 static inline void __mod_memcg_state(struct mem_cgroup *memcg, 1445 int idx, 1446 int nr) 1447 { 1448 } 1449 1450 static inline void mod_memcg_state(struct mem_cgroup *memcg, 1451 int idx, 1452 int nr) 1453 { 1454 } 1455 1456 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) 1457 { 1458 return 0; 1459 } 1460 1461 static inline unsigned long lruvec_page_state(struct lruvec *lruvec, 1462 enum node_stat_item idx) 1463 { 1464 return node_page_state(lruvec_pgdat(lruvec), idx); 1465 } 1466 1467 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, 1468 enum node_stat_item idx) 1469 { 1470 return node_page_state(lruvec_pgdat(lruvec), idx); 1471 } 1472 1473 static inline void mem_cgroup_flush_stats(void) 1474 { 1475 } 1476 1477 static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec, 1478 enum node_stat_item idx, int val) 1479 { 1480 } 1481 1482 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1483 int val) 1484 { 1485 struct page *page = virt_to_head_page(p); 1486 1487 __mod_node_page_state(page_pgdat(page), idx, val); 1488 } 1489 1490 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, 1491 int val) 1492 { 1493 struct page *page = virt_to_head_page(p); 1494 1495 mod_node_page_state(page_pgdat(page), idx, val); 1496 } 1497 1498 static inline void count_memcg_events(struct mem_cgroup *memcg, 1499 enum vm_event_item idx, 1500 unsigned long count) 1501 { 1502 } 1503 1504 static inline void __count_memcg_events(struct mem_cgroup *memcg, 1505 enum vm_event_item idx, 1506 unsigned long count) 1507 { 1508 } 1509 1510 static inline void count_memcg_page_event(struct page *page, 1511 int idx) 1512 { 1513 } 1514 1515 static inline 1516 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) 1517 { 1518 } 1519 1520 static inline void split_page_memcg(struct page *head, unsigned int nr) 1521 { 1522 } 1523 1524 static inline 1525 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, 1526 gfp_t gfp_mask, 1527 unsigned long *total_scanned) 1528 { 1529 return 0; 1530 } 1531 #endif /* CONFIG_MEMCG */ 1532 1533 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx) 1534 { 1535 __mod_lruvec_kmem_state(p, idx, 1); 1536 } 1537 1538 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx) 1539 { 1540 __mod_lruvec_kmem_state(p, idx, -1); 1541 } 1542 1543 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) 1544 { 1545 struct mem_cgroup *memcg; 1546 1547 memcg = lruvec_memcg(lruvec); 1548 if (!memcg) 1549 return NULL; 1550 memcg = parent_mem_cgroup(memcg); 1551 if (!memcg) 1552 return NULL; 1553 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); 1554 } 1555 1556 static inline void unlock_page_lruvec(struct lruvec *lruvec) 1557 { 1558 spin_unlock(&lruvec->lru_lock); 1559 } 1560 1561 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec) 1562 { 1563 spin_unlock_irq(&lruvec->lru_lock); 1564 } 1565 1566 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec, 1567 unsigned long flags) 1568 { 1569 spin_unlock_irqrestore(&lruvec->lru_lock, flags); 1570 } 1571 1572 /* Test requires a stable page->memcg binding, see page_memcg() */ 1573 static inline bool folio_matches_lruvec(struct folio *folio, 1574 struct lruvec *lruvec) 1575 { 1576 return lruvec_pgdat(lruvec) == folio_pgdat(folio) && 1577 lruvec_memcg(lruvec) == folio_memcg(folio); 1578 } 1579 1580 /* Don't lock again iff page's lruvec locked */ 1581 static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio, 1582 struct lruvec *locked_lruvec) 1583 { 1584 if (locked_lruvec) { 1585 if (folio_matches_lruvec(folio, locked_lruvec)) 1586 return locked_lruvec; 1587 1588 unlock_page_lruvec_irq(locked_lruvec); 1589 } 1590 1591 return folio_lruvec_lock_irq(folio); 1592 } 1593 1594 /* Don't lock again iff page's lruvec locked */ 1595 static inline struct lruvec *folio_lruvec_relock_irqsave(struct folio *folio, 1596 struct lruvec *locked_lruvec, unsigned long *flags) 1597 { 1598 if (locked_lruvec) { 1599 if (folio_matches_lruvec(folio, locked_lruvec)) 1600 return locked_lruvec; 1601 1602 unlock_page_lruvec_irqrestore(locked_lruvec, *flags); 1603 } 1604 1605 return folio_lruvec_lock_irqsave(folio, flags); 1606 } 1607 1608 #ifdef CONFIG_CGROUP_WRITEBACK 1609 1610 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); 1611 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, 1612 unsigned long *pheadroom, unsigned long *pdirty, 1613 unsigned long *pwriteback); 1614 1615 void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio, 1616 struct bdi_writeback *wb); 1617 1618 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1619 struct bdi_writeback *wb) 1620 { 1621 if (mem_cgroup_disabled()) 1622 return; 1623 1624 if (unlikely(&folio_memcg(folio)->css != wb->memcg_css)) 1625 mem_cgroup_track_foreign_dirty_slowpath(folio, wb); 1626 } 1627 1628 void mem_cgroup_flush_foreign(struct bdi_writeback *wb); 1629 1630 #else /* CONFIG_CGROUP_WRITEBACK */ 1631 1632 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) 1633 { 1634 return NULL; 1635 } 1636 1637 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, 1638 unsigned long *pfilepages, 1639 unsigned long *pheadroom, 1640 unsigned long *pdirty, 1641 unsigned long *pwriteback) 1642 { 1643 } 1644 1645 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, 1646 struct bdi_writeback *wb) 1647 { 1648 } 1649 1650 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) 1651 { 1652 } 1653 1654 #endif /* CONFIG_CGROUP_WRITEBACK */ 1655 1656 struct sock; 1657 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages, 1658 gfp_t gfp_mask); 1659 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); 1660 #ifdef CONFIG_MEMCG 1661 extern struct static_key_false memcg_sockets_enabled_key; 1662 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) 1663 void mem_cgroup_sk_alloc(struct sock *sk); 1664 void mem_cgroup_sk_free(struct sock *sk); 1665 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1666 { 1667 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) 1668 return true; 1669 do { 1670 if (time_before(jiffies, memcg->socket_pressure)) 1671 return true; 1672 } while ((memcg = parent_mem_cgroup(memcg))); 1673 return false; 1674 } 1675 1676 int alloc_shrinker_info(struct mem_cgroup *memcg); 1677 void free_shrinker_info(struct mem_cgroup *memcg); 1678 void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id); 1679 void reparent_shrinker_deferred(struct mem_cgroup *memcg); 1680 #else 1681 #define mem_cgroup_sockets_enabled 0 1682 static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; 1683 static inline void mem_cgroup_sk_free(struct sock *sk) { }; 1684 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) 1685 { 1686 return false; 1687 } 1688 1689 static inline void set_shrinker_bit(struct mem_cgroup *memcg, 1690 int nid, int shrinker_id) 1691 { 1692 } 1693 #endif 1694 1695 #ifdef CONFIG_MEMCG_KMEM 1696 bool mem_cgroup_kmem_disabled(void); 1697 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); 1698 void __memcg_kmem_uncharge_page(struct page *page, int order); 1699 1700 struct obj_cgroup *get_obj_cgroup_from_current(void); 1701 1702 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); 1703 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); 1704 1705 extern struct static_key_false memcg_kmem_enabled_key; 1706 1707 extern int memcg_nr_cache_ids; 1708 void memcg_get_cache_ids(void); 1709 void memcg_put_cache_ids(void); 1710 1711 /* 1712 * Helper macro to loop through all memcg-specific caches. Callers must still 1713 * check if the cache is valid (it is either valid or NULL). 1714 * the slab_mutex must be held when looping through those caches 1715 */ 1716 #define for_each_memcg_cache_index(_idx) \ 1717 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) 1718 1719 static inline bool memcg_kmem_enabled(void) 1720 { 1721 return static_branch_likely(&memcg_kmem_enabled_key); 1722 } 1723 1724 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1725 int order) 1726 { 1727 if (memcg_kmem_enabled()) 1728 return __memcg_kmem_charge_page(page, gfp, order); 1729 return 0; 1730 } 1731 1732 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1733 { 1734 if (memcg_kmem_enabled()) 1735 __memcg_kmem_uncharge_page(page, order); 1736 } 1737 1738 /* 1739 * A helper for accessing memcg's kmem_id, used for getting 1740 * corresponding LRU lists. 1741 */ 1742 static inline int memcg_cache_id(struct mem_cgroup *memcg) 1743 { 1744 return memcg ? memcg->kmemcg_id : -1; 1745 } 1746 1747 struct mem_cgroup *mem_cgroup_from_obj(void *p); 1748 1749 #else 1750 static inline bool mem_cgroup_kmem_disabled(void) 1751 { 1752 return true; 1753 } 1754 1755 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1756 int order) 1757 { 1758 return 0; 1759 } 1760 1761 static inline void memcg_kmem_uncharge_page(struct page *page, int order) 1762 { 1763 } 1764 1765 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, 1766 int order) 1767 { 1768 return 0; 1769 } 1770 1771 static inline void __memcg_kmem_uncharge_page(struct page *page, int order) 1772 { 1773 } 1774 1775 #define for_each_memcg_cache_index(_idx) \ 1776 for (; NULL; ) 1777 1778 static inline bool memcg_kmem_enabled(void) 1779 { 1780 return false; 1781 } 1782 1783 static inline int memcg_cache_id(struct mem_cgroup *memcg) 1784 { 1785 return -1; 1786 } 1787 1788 static inline void memcg_get_cache_ids(void) 1789 { 1790 } 1791 1792 static inline void memcg_put_cache_ids(void) 1793 { 1794 } 1795 1796 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p) 1797 { 1798 return NULL; 1799 } 1800 1801 #endif /* CONFIG_MEMCG_KMEM */ 1802 1803 #endif /* _LINUX_MEMCONTROL_H */ 1804