| 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
| 2 | /* memcontrol.h - Memory Controller |
| 3 | * |
| 4 | * Copyright IBM Corporation, 2007 |
| 5 | * Author Balbir Singh <balbir@linux.vnet.ibm.com> |
| 6 | * |
| 7 | * Copyright 2007 OpenVZ SWsoft Inc |
| 8 | * Author: Pavel Emelianov <xemul@openvz.org> |
| 9 | */ |
| 10 | |
| 11 | #ifndef _LINUX_MEMCONTROL_H |
| 12 | #define _LINUX_MEMCONTROL_H |
| 13 | #include <linux/cgroup.h> |
| 14 | #include <linux/vm_event_item.h> |
| 15 | #include <linux/hardirq.h> |
| 16 | #include <linux/jump_label.h> |
| 17 | #include <linux/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 | long stat[MEMCG_NR_STAT]; |
| 80 | unsigned long events[NR_VM_EVENT_ITEMS]; |
| 81 | unsigned long nr_page_events; |
| 82 | unsigned long targets[MEM_CGROUP_NTARGETS]; |
| 83 | }; |
| 84 | |
| 85 | struct mem_cgroup_reclaim_iter { |
| 86 | struct mem_cgroup *position; |
| 87 | /* scan generation, increased every round-trip */ |
| 88 | unsigned int generation; |
| 89 | }; |
| 90 | |
| 91 | struct lruvec_stat { |
| 92 | long count[NR_VM_NODE_STAT_ITEMS]; |
| 93 | }; |
| 94 | |
| 95 | /* |
| 96 | * Bitmap of shrinker::id corresponding to memcg-aware shrinkers, |
| 97 | * which have elements charged to this memcg. |
| 98 | */ |
| 99 | struct memcg_shrinker_map { |
| 100 | struct rcu_head rcu; |
| 101 | unsigned long map[]; |
| 102 | }; |
| 103 | |
| 104 | /* |
| 105 | * per-node information in memory controller. |
| 106 | */ |
| 107 | struct mem_cgroup_per_node { |
| 108 | struct lruvec lruvec; |
| 109 | |
| 110 | /* Legacy local VM stats */ |
| 111 | struct lruvec_stat __percpu *lruvec_stat_local; |
| 112 | |
| 113 | /* Subtree VM stats (batched updates) */ |
| 114 | struct lruvec_stat __percpu *lruvec_stat_cpu; |
| 115 | atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS]; |
| 116 | |
| 117 | unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; |
| 118 | |
| 119 | struct mem_cgroup_reclaim_iter iter; |
| 120 | |
| 121 | struct memcg_shrinker_map __rcu *shrinker_map; |
| 122 | |
| 123 | struct rb_node tree_node; /* RB tree node */ |
| 124 | unsigned long usage_in_excess;/* Set to the value by which */ |
| 125 | /* the soft limit is exceeded*/ |
| 126 | bool on_tree; |
| 127 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ |
| 128 | /* use container_of */ |
| 129 | }; |
| 130 | |
| 131 | struct mem_cgroup_threshold { |
| 132 | struct eventfd_ctx *eventfd; |
| 133 | unsigned long threshold; |
| 134 | }; |
| 135 | |
| 136 | /* For threshold */ |
| 137 | struct mem_cgroup_threshold_ary { |
| 138 | /* An array index points to threshold just below or equal to usage. */ |
| 139 | int current_threshold; |
| 140 | /* Size of entries[] */ |
| 141 | unsigned int size; |
| 142 | /* Array of thresholds */ |
| 143 | struct mem_cgroup_threshold entries[]; |
| 144 | }; |
| 145 | |
| 146 | struct mem_cgroup_thresholds { |
| 147 | /* Primary thresholds array */ |
| 148 | struct mem_cgroup_threshold_ary *primary; |
| 149 | /* |
| 150 | * Spare threshold array. |
| 151 | * This is needed to make mem_cgroup_unregister_event() "never fail". |
| 152 | * It must be able to store at least primary->size - 1 entries. |
| 153 | */ |
| 154 | struct mem_cgroup_threshold_ary *spare; |
| 155 | }; |
| 156 | |
| 157 | enum memcg_kmem_state { |
| 158 | KMEM_NONE, |
| 159 | KMEM_ALLOCATED, |
| 160 | KMEM_ONLINE, |
| 161 | }; |
| 162 | |
| 163 | #if defined(CONFIG_SMP) |
| 164 | struct memcg_padding { |
| 165 | char x[0]; |
| 166 | } ____cacheline_internodealigned_in_smp; |
| 167 | #define MEMCG_PADDING(name) struct memcg_padding name; |
| 168 | #else |
| 169 | #define MEMCG_PADDING(name) |
| 170 | #endif |
| 171 | |
| 172 | /* |
| 173 | * Remember four most recent foreign writebacks with dirty pages in this |
| 174 | * cgroup. Inode sharing is expected to be uncommon and, even if we miss |
| 175 | * one in a given round, we're likely to catch it later if it keeps |
| 176 | * foreign-dirtying, so a fairly low count should be enough. |
| 177 | * |
| 178 | * See mem_cgroup_track_foreign_dirty_slowpath() for details. |
| 179 | */ |
| 180 | #define MEMCG_CGWB_FRN_CNT 4 |
| 181 | |
| 182 | struct memcg_cgwb_frn { |
| 183 | u64 bdi_id; /* bdi->id of the foreign inode */ |
| 184 | int memcg_id; /* memcg->css.id of foreign inode */ |
| 185 | u64 at; /* jiffies_64 at the time of dirtying */ |
| 186 | struct wb_completion done; /* tracks in-flight foreign writebacks */ |
| 187 | }; |
| 188 | |
| 189 | /* |
| 190 | * Bucket for arbitrarily byte-sized objects charged to a memory |
| 191 | * cgroup. The bucket can be reparented in one piece when the cgroup |
| 192 | * is destroyed, without having to round up the individual references |
| 193 | * of all live memory objects in the wild. |
| 194 | */ |
| 195 | struct obj_cgroup { |
| 196 | struct percpu_ref refcnt; |
| 197 | struct mem_cgroup *memcg; |
| 198 | atomic_t nr_charged_bytes; |
| 199 | union { |
| 200 | struct list_head list; |
| 201 | struct rcu_head rcu; |
| 202 | }; |
| 203 | }; |
| 204 | |
| 205 | /* |
| 206 | * The memory controller data structure. The memory controller controls both |
| 207 | * page cache and RSS per cgroup. We would eventually like to provide |
| 208 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, |
| 209 | * to help the administrator determine what knobs to tune. |
| 210 | */ |
| 211 | struct mem_cgroup { |
| 212 | struct cgroup_subsys_state css; |
| 213 | |
| 214 | /* Private memcg ID. Used to ID objects that outlive the cgroup */ |
| 215 | struct mem_cgroup_id id; |
| 216 | |
| 217 | /* Accounted resources */ |
| 218 | struct page_counter memory; |
| 219 | struct page_counter swap; |
| 220 | |
| 221 | /* Legacy consumer-oriented counters */ |
| 222 | struct page_counter memsw; |
| 223 | struct page_counter kmem; |
| 224 | struct page_counter tcpmem; |
| 225 | |
| 226 | /* Range enforcement for interrupt charges */ |
| 227 | struct work_struct high_work; |
| 228 | |
| 229 | unsigned long soft_limit; |
| 230 | |
| 231 | /* vmpressure notifications */ |
| 232 | struct vmpressure vmpressure; |
| 233 | |
| 234 | /* |
| 235 | * Should the accounting and control be hierarchical, per subtree? |
| 236 | */ |
| 237 | bool use_hierarchy; |
| 238 | |
| 239 | /* |
| 240 | * Should the OOM killer kill all belonging tasks, had it kill one? |
| 241 | */ |
| 242 | bool oom_group; |
| 243 | |
| 244 | /* protected by memcg_oom_lock */ |
| 245 | bool oom_lock; |
| 246 | int under_oom; |
| 247 | |
| 248 | int swappiness; |
| 249 | /* OOM-Killer disable */ |
| 250 | int oom_kill_disable; |
| 251 | |
| 252 | /* memory.events and memory.events.local */ |
| 253 | struct cgroup_file events_file; |
| 254 | struct cgroup_file events_local_file; |
| 255 | |
| 256 | /* handle for "memory.swap.events" */ |
| 257 | struct cgroup_file swap_events_file; |
| 258 | |
| 259 | /* protect arrays of thresholds */ |
| 260 | struct mutex thresholds_lock; |
| 261 | |
| 262 | /* thresholds for memory usage. RCU-protected */ |
| 263 | struct mem_cgroup_thresholds thresholds; |
| 264 | |
| 265 | /* thresholds for mem+swap usage. RCU-protected */ |
| 266 | struct mem_cgroup_thresholds memsw_thresholds; |
| 267 | |
| 268 | /* For oom notifier event fd */ |
| 269 | struct list_head oom_notify; |
| 270 | |
| 271 | /* |
| 272 | * Should we move charges of a task when a task is moved into this |
| 273 | * mem_cgroup ? And what type of charges should we move ? |
| 274 | */ |
| 275 | unsigned long move_charge_at_immigrate; |
| 276 | /* taken only while moving_account > 0 */ |
| 277 | spinlock_t move_lock; |
| 278 | unsigned long move_lock_flags; |
| 279 | |
| 280 | MEMCG_PADDING(_pad1_); |
| 281 | |
| 282 | /* |
| 283 | * set > 0 if pages under this cgroup are moving to other cgroup. |
| 284 | */ |
| 285 | atomic_t moving_account; |
| 286 | struct task_struct *move_lock_task; |
| 287 | |
| 288 | /* Legacy local VM stats and events */ |
| 289 | struct memcg_vmstats_percpu __percpu *vmstats_local; |
| 290 | |
| 291 | /* Subtree VM stats and events (batched updates) */ |
| 292 | struct memcg_vmstats_percpu __percpu *vmstats_percpu; |
| 293 | |
| 294 | MEMCG_PADDING(_pad2_); |
| 295 | |
| 296 | atomic_long_t vmstats[MEMCG_NR_STAT]; |
| 297 | atomic_long_t vmevents[NR_VM_EVENT_ITEMS]; |
| 298 | |
| 299 | /* memory.events */ |
| 300 | atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS]; |
| 301 | atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS]; |
| 302 | |
| 303 | unsigned long socket_pressure; |
| 304 | |
| 305 | /* Legacy tcp memory accounting */ |
| 306 | bool tcpmem_active; |
| 307 | int tcpmem_pressure; |
| 308 | |
| 309 | #ifdef CONFIG_MEMCG_KMEM |
| 310 | /* Index in the kmem_cache->memcg_params.memcg_caches array */ |
| 311 | int kmemcg_id; |
| 312 | enum memcg_kmem_state kmem_state; |
| 313 | struct obj_cgroup __rcu *objcg; |
| 314 | struct list_head objcg_list; /* list of inherited objcgs */ |
| 315 | #endif |
| 316 | |
| 317 | #ifdef CONFIG_CGROUP_WRITEBACK |
| 318 | struct list_head cgwb_list; |
| 319 | struct wb_domain cgwb_domain; |
| 320 | struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT]; |
| 321 | #endif |
| 322 | |
| 323 | /* List of events which userspace want to receive */ |
| 324 | struct list_head event_list; |
| 325 | spinlock_t event_list_lock; |
| 326 | |
| 327 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 328 | struct deferred_split deferred_split_queue; |
| 329 | #endif |
| 330 | |
| 331 | struct mem_cgroup_per_node *nodeinfo[0]; |
| 332 | /* WARNING: nodeinfo must be the last member here */ |
| 333 | }; |
| 334 | |
| 335 | /* |
| 336 | * size of first charge trial. "32" comes from vmscan.c's magic value. |
| 337 | * TODO: maybe necessary to use big numbers in big irons. |
| 338 | */ |
| 339 | #define MEMCG_CHARGE_BATCH 32U |
| 340 | |
| 341 | extern struct mem_cgroup *root_mem_cgroup; |
| 342 | |
| 343 | static __always_inline bool memcg_stat_item_in_bytes(int idx) |
| 344 | { |
| 345 | if (idx == MEMCG_PERCPU_B) |
| 346 | return true; |
| 347 | return vmstat_item_in_bytes(idx); |
| 348 | } |
| 349 | |
| 350 | static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) |
| 351 | { |
| 352 | return (memcg == root_mem_cgroup); |
| 353 | } |
| 354 | |
| 355 | static inline bool mem_cgroup_disabled(void) |
| 356 | { |
| 357 | return !cgroup_subsys_enabled(memory_cgrp_subsys); |
| 358 | } |
| 359 | |
| 360 | static inline unsigned long mem_cgroup_protection(struct mem_cgroup *root, |
| 361 | struct mem_cgroup *memcg, |
| 362 | bool in_low_reclaim) |
| 363 | { |
| 364 | if (mem_cgroup_disabled()) |
| 365 | return 0; |
| 366 | |
| 367 | /* |
| 368 | * There is no reclaim protection applied to a targeted reclaim. |
| 369 | * We are special casing this specific case here because |
| 370 | * mem_cgroup_protected calculation is not robust enough to keep |
| 371 | * the protection invariant for calculated effective values for |
| 372 | * parallel reclaimers with different reclaim target. This is |
| 373 | * especially a problem for tail memcgs (as they have pages on LRU) |
| 374 | * which would want to have effective values 0 for targeted reclaim |
| 375 | * but a different value for external reclaim. |
| 376 | * |
| 377 | * Example |
| 378 | * Let's have global and A's reclaim in parallel: |
| 379 | * | |
| 380 | * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) |
| 381 | * |\ |
| 382 | * | C (low = 1G, usage = 2.5G) |
| 383 | * B (low = 1G, usage = 0.5G) |
| 384 | * |
| 385 | * For the global reclaim |
| 386 | * A.elow = A.low |
| 387 | * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow |
| 388 | * C.elow = min(C.usage, C.low) |
| 389 | * |
| 390 | * With the effective values resetting we have A reclaim |
| 391 | * A.elow = 0 |
| 392 | * B.elow = B.low |
| 393 | * C.elow = C.low |
| 394 | * |
| 395 | * If the global reclaim races with A's reclaim then |
| 396 | * B.elow = C.elow = 0 because children_low_usage > A.elow) |
| 397 | * is possible and reclaiming B would be violating the protection. |
| 398 | * |
| 399 | */ |
| 400 | if (root == memcg) |
| 401 | return 0; |
| 402 | |
| 403 | if (in_low_reclaim) |
| 404 | return READ_ONCE(memcg->memory.emin); |
| 405 | |
| 406 | return max(READ_ONCE(memcg->memory.emin), |
| 407 | READ_ONCE(memcg->memory.elow)); |
| 408 | } |
| 409 | |
| 410 | void mem_cgroup_calculate_protection(struct mem_cgroup *root, |
| 411 | struct mem_cgroup *memcg); |
| 412 | |
| 413 | static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg) |
| 414 | { |
| 415 | /* |
| 416 | * The root memcg doesn't account charges, and doesn't support |
| 417 | * protection. |
| 418 | */ |
| 419 | return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg); |
| 420 | |
| 421 | } |
| 422 | |
| 423 | static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg) |
| 424 | { |
| 425 | if (!mem_cgroup_supports_protection(memcg)) |
| 426 | return false; |
| 427 | |
| 428 | return READ_ONCE(memcg->memory.elow) >= |
| 429 | page_counter_read(&memcg->memory); |
| 430 | } |
| 431 | |
| 432 | static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg) |
| 433 | { |
| 434 | if (!mem_cgroup_supports_protection(memcg)) |
| 435 | return false; |
| 436 | |
| 437 | return READ_ONCE(memcg->memory.emin) >= |
| 438 | page_counter_read(&memcg->memory); |
| 439 | } |
| 440 | |
| 441 | int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask); |
| 442 | |
| 443 | void mem_cgroup_uncharge(struct page *page); |
| 444 | void mem_cgroup_uncharge_list(struct list_head *page_list); |
| 445 | |
| 446 | void mem_cgroup_migrate(struct page *oldpage, struct page *newpage); |
| 447 | |
| 448 | static struct mem_cgroup_per_node * |
| 449 | mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid) |
| 450 | { |
| 451 | return memcg->nodeinfo[nid]; |
| 452 | } |
| 453 | |
| 454 | /** |
| 455 | * mem_cgroup_lruvec - get the lru list vector for a memcg & node |
| 456 | * @memcg: memcg of the wanted lruvec |
| 457 | * |
| 458 | * Returns the lru list vector holding pages for a given @memcg & |
| 459 | * @node combination. This can be the node lruvec, if the memory |
| 460 | * controller is disabled. |
| 461 | */ |
| 462 | static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, |
| 463 | struct pglist_data *pgdat) |
| 464 | { |
| 465 | struct mem_cgroup_per_node *mz; |
| 466 | struct lruvec *lruvec; |
| 467 | |
| 468 | if (mem_cgroup_disabled()) { |
| 469 | lruvec = &pgdat->__lruvec; |
| 470 | goto out; |
| 471 | } |
| 472 | |
| 473 | if (!memcg) |
| 474 | memcg = root_mem_cgroup; |
| 475 | |
| 476 | mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id); |
| 477 | lruvec = &mz->lruvec; |
| 478 | out: |
| 479 | /* |
| 480 | * Since a node can be onlined after the mem_cgroup was created, |
| 481 | * we have to be prepared to initialize lruvec->pgdat here; |
| 482 | * and if offlined then reonlined, we need to reinitialize it. |
| 483 | */ |
| 484 | if (unlikely(lruvec->pgdat != pgdat)) |
| 485 | lruvec->pgdat = pgdat; |
| 486 | return lruvec; |
| 487 | } |
| 488 | |
| 489 | struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *); |
| 490 | |
| 491 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
| 492 | |
| 493 | struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); |
| 494 | |
| 495 | struct mem_cgroup *get_mem_cgroup_from_page(struct page *page); |
| 496 | |
| 497 | static inline |
| 498 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ |
| 499 | return css ? container_of(css, struct mem_cgroup, css) : NULL; |
| 500 | } |
| 501 | |
| 502 | static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) |
| 503 | { |
| 504 | return percpu_ref_tryget(&objcg->refcnt); |
| 505 | } |
| 506 | |
| 507 | static inline void obj_cgroup_get(struct obj_cgroup *objcg) |
| 508 | { |
| 509 | percpu_ref_get(&objcg->refcnt); |
| 510 | } |
| 511 | |
| 512 | static inline void obj_cgroup_put(struct obj_cgroup *objcg) |
| 513 | { |
| 514 | percpu_ref_put(&objcg->refcnt); |
| 515 | } |
| 516 | |
| 517 | /* |
| 518 | * After the initialization objcg->memcg is always pointing at |
| 519 | * a valid memcg, but can be atomically swapped to the parent memcg. |
| 520 | * |
| 521 | * The caller must ensure that the returned memcg won't be released: |
| 522 | * e.g. acquire the rcu_read_lock or css_set_lock. |
| 523 | */ |
| 524 | static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg) |
| 525 | { |
| 526 | return READ_ONCE(objcg->memcg); |
| 527 | } |
| 528 | |
| 529 | static inline void mem_cgroup_put(struct mem_cgroup *memcg) |
| 530 | { |
| 531 | if (memcg) |
| 532 | css_put(&memcg->css); |
| 533 | } |
| 534 | |
| 535 | #define mem_cgroup_from_counter(counter, member) \ |
| 536 | container_of(counter, struct mem_cgroup, member) |
| 537 | |
| 538 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
| 539 | struct mem_cgroup *, |
| 540 | struct mem_cgroup_reclaim_cookie *); |
| 541 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); |
| 542 | int mem_cgroup_scan_tasks(struct mem_cgroup *, |
| 543 | int (*)(struct task_struct *, void *), void *); |
| 544 | |
| 545 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
| 546 | { |
| 547 | if (mem_cgroup_disabled()) |
| 548 | return 0; |
| 549 | |
| 550 | return memcg->id.id; |
| 551 | } |
| 552 | struct mem_cgroup *mem_cgroup_from_id(unsigned short id); |
| 553 | |
| 554 | static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) |
| 555 | { |
| 556 | return mem_cgroup_from_css(seq_css(m)); |
| 557 | } |
| 558 | |
| 559 | static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) |
| 560 | { |
| 561 | struct mem_cgroup_per_node *mz; |
| 562 | |
| 563 | if (mem_cgroup_disabled()) |
| 564 | return NULL; |
| 565 | |
| 566 | mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); |
| 567 | return mz->memcg; |
| 568 | } |
| 569 | |
| 570 | /** |
| 571 | * parent_mem_cgroup - find the accounting parent of a memcg |
| 572 | * @memcg: memcg whose parent to find |
| 573 | * |
| 574 | * Returns the parent memcg, or NULL if this is the root or the memory |
| 575 | * controller is in legacy no-hierarchy mode. |
| 576 | */ |
| 577 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) |
| 578 | { |
| 579 | if (!memcg->memory.parent) |
| 580 | return NULL; |
| 581 | return mem_cgroup_from_counter(memcg->memory.parent, memory); |
| 582 | } |
| 583 | |
| 584 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
| 585 | struct mem_cgroup *root) |
| 586 | { |
| 587 | if (root == memcg) |
| 588 | return true; |
| 589 | if (!root->use_hierarchy) |
| 590 | return false; |
| 591 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); |
| 592 | } |
| 593 | |
| 594 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
| 595 | struct mem_cgroup *memcg) |
| 596 | { |
| 597 | struct mem_cgroup *task_memcg; |
| 598 | bool match = false; |
| 599 | |
| 600 | rcu_read_lock(); |
| 601 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
| 602 | if (task_memcg) |
| 603 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
| 604 | rcu_read_unlock(); |
| 605 | return match; |
| 606 | } |
| 607 | |
| 608 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
| 609 | ino_t page_cgroup_ino(struct page *page); |
| 610 | |
| 611 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
| 612 | { |
| 613 | if (mem_cgroup_disabled()) |
| 614 | return true; |
| 615 | return !!(memcg->css.flags & CSS_ONLINE); |
| 616 | } |
| 617 | |
| 618 | /* |
| 619 | * For memory reclaim. |
| 620 | */ |
| 621 | int mem_cgroup_select_victim_node(struct mem_cgroup *memcg); |
| 622 | |
| 623 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, |
| 624 | int zid, int nr_pages); |
| 625 | |
| 626 | static inline |
| 627 | unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, |
| 628 | enum lru_list lru, int zone_idx) |
| 629 | { |
| 630 | struct mem_cgroup_per_node *mz; |
| 631 | |
| 632 | mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); |
| 633 | return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); |
| 634 | } |
| 635 | |
| 636 | void mem_cgroup_handle_over_high(void); |
| 637 | |
| 638 | unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); |
| 639 | |
| 640 | unsigned long mem_cgroup_size(struct mem_cgroup *memcg); |
| 641 | |
| 642 | void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, |
| 643 | struct task_struct *p); |
| 644 | |
| 645 | void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); |
| 646 | |
| 647 | static inline void mem_cgroup_enter_user_fault(void) |
| 648 | { |
| 649 | WARN_ON(current->in_user_fault); |
| 650 | current->in_user_fault = 1; |
| 651 | } |
| 652 | |
| 653 | static inline void mem_cgroup_exit_user_fault(void) |
| 654 | { |
| 655 | WARN_ON(!current->in_user_fault); |
| 656 | current->in_user_fault = 0; |
| 657 | } |
| 658 | |
| 659 | static inline bool task_in_memcg_oom(struct task_struct *p) |
| 660 | { |
| 661 | return p->memcg_in_oom; |
| 662 | } |
| 663 | |
| 664 | bool mem_cgroup_oom_synchronize(bool wait); |
| 665 | struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, |
| 666 | struct mem_cgroup *oom_domain); |
| 667 | void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); |
| 668 | |
| 669 | #ifdef CONFIG_MEMCG_SWAP |
| 670 | extern bool cgroup_memory_noswap; |
| 671 | #endif |
| 672 | |
| 673 | struct mem_cgroup *lock_page_memcg(struct page *page); |
| 674 | void __unlock_page_memcg(struct mem_cgroup *memcg); |
| 675 | void unlock_page_memcg(struct page *page); |
| 676 | |
| 677 | /* |
| 678 | * idx can be of type enum memcg_stat_item or node_stat_item. |
| 679 | * Keep in sync with memcg_exact_page_state(). |
| 680 | */ |
| 681 | static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) |
| 682 | { |
| 683 | long x = atomic_long_read(&memcg->vmstats[idx]); |
| 684 | #ifdef CONFIG_SMP |
| 685 | if (x < 0) |
| 686 | x = 0; |
| 687 | #endif |
| 688 | return x; |
| 689 | } |
| 690 | |
| 691 | /* |
| 692 | * idx can be of type enum memcg_stat_item or node_stat_item. |
| 693 | * Keep in sync with memcg_exact_page_state(). |
| 694 | */ |
| 695 | static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg, |
| 696 | int idx) |
| 697 | { |
| 698 | long x = 0; |
| 699 | int cpu; |
| 700 | |
| 701 | for_each_possible_cpu(cpu) |
| 702 | x += per_cpu(memcg->vmstats_local->stat[idx], cpu); |
| 703 | #ifdef CONFIG_SMP |
| 704 | if (x < 0) |
| 705 | x = 0; |
| 706 | #endif |
| 707 | return x; |
| 708 | } |
| 709 | |
| 710 | void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val); |
| 711 | |
| 712 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 713 | static inline void mod_memcg_state(struct mem_cgroup *memcg, |
| 714 | int idx, int val) |
| 715 | { |
| 716 | unsigned long flags; |
| 717 | |
| 718 | local_irq_save(flags); |
| 719 | __mod_memcg_state(memcg, idx, val); |
| 720 | local_irq_restore(flags); |
| 721 | } |
| 722 | |
| 723 | /** |
| 724 | * mod_memcg_page_state - update page state statistics |
| 725 | * @page: the page |
| 726 | * @idx: page state item to account |
| 727 | * @val: number of pages (positive or negative) |
| 728 | * |
| 729 | * The @page must be locked or the caller must use lock_page_memcg() |
| 730 | * to prevent double accounting when the page is concurrently being |
| 731 | * moved to another memcg: |
| 732 | * |
| 733 | * lock_page(page) or lock_page_memcg(page) |
| 734 | * if (TestClearPageState(page)) |
| 735 | * mod_memcg_page_state(page, state, -1); |
| 736 | * unlock_page(page) or unlock_page_memcg(page) |
| 737 | * |
| 738 | * Kernel pages are an exception to this, since they'll never move. |
| 739 | */ |
| 740 | static inline void __mod_memcg_page_state(struct page *page, |
| 741 | int idx, int val) |
| 742 | { |
| 743 | if (page->mem_cgroup) |
| 744 | __mod_memcg_state(page->mem_cgroup, idx, val); |
| 745 | } |
| 746 | |
| 747 | static inline void mod_memcg_page_state(struct page *page, |
| 748 | int idx, int val) |
| 749 | { |
| 750 | if (page->mem_cgroup) |
| 751 | mod_memcg_state(page->mem_cgroup, idx, val); |
| 752 | } |
| 753 | |
| 754 | static inline unsigned long lruvec_page_state(struct lruvec *lruvec, |
| 755 | enum node_stat_item idx) |
| 756 | { |
| 757 | struct mem_cgroup_per_node *pn; |
| 758 | long x; |
| 759 | |
| 760 | if (mem_cgroup_disabled()) |
| 761 | return node_page_state(lruvec_pgdat(lruvec), idx); |
| 762 | |
| 763 | pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); |
| 764 | x = atomic_long_read(&pn->lruvec_stat[idx]); |
| 765 | #ifdef CONFIG_SMP |
| 766 | if (x < 0) |
| 767 | x = 0; |
| 768 | #endif |
| 769 | return x; |
| 770 | } |
| 771 | |
| 772 | static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, |
| 773 | enum node_stat_item idx) |
| 774 | { |
| 775 | struct mem_cgroup_per_node *pn; |
| 776 | long x = 0; |
| 777 | int cpu; |
| 778 | |
| 779 | if (mem_cgroup_disabled()) |
| 780 | return node_page_state(lruvec_pgdat(lruvec), idx); |
| 781 | |
| 782 | pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); |
| 783 | for_each_possible_cpu(cpu) |
| 784 | x += per_cpu(pn->lruvec_stat_local->count[idx], cpu); |
| 785 | #ifdef CONFIG_SMP |
| 786 | if (x < 0) |
| 787 | x = 0; |
| 788 | #endif |
| 789 | return x; |
| 790 | } |
| 791 | |
| 792 | void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx, |
| 793 | int val); |
| 794 | void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx, |
| 795 | int val); |
| 796 | void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val); |
| 797 | |
| 798 | void mod_memcg_obj_state(void *p, int idx, int val); |
| 799 | |
| 800 | static inline void mod_lruvec_slab_state(void *p, enum node_stat_item idx, |
| 801 | int val) |
| 802 | { |
| 803 | unsigned long flags; |
| 804 | |
| 805 | local_irq_save(flags); |
| 806 | __mod_lruvec_slab_state(p, idx, val); |
| 807 | local_irq_restore(flags); |
| 808 | } |
| 809 | |
| 810 | static inline void mod_memcg_lruvec_state(struct lruvec *lruvec, |
| 811 | enum node_stat_item idx, int val) |
| 812 | { |
| 813 | unsigned long flags; |
| 814 | |
| 815 | local_irq_save(flags); |
| 816 | __mod_memcg_lruvec_state(lruvec, idx, val); |
| 817 | local_irq_restore(flags); |
| 818 | } |
| 819 | |
| 820 | static inline void mod_lruvec_state(struct lruvec *lruvec, |
| 821 | enum node_stat_item idx, int val) |
| 822 | { |
| 823 | unsigned long flags; |
| 824 | |
| 825 | local_irq_save(flags); |
| 826 | __mod_lruvec_state(lruvec, idx, val); |
| 827 | local_irq_restore(flags); |
| 828 | } |
| 829 | |
| 830 | static inline void __mod_lruvec_page_state(struct page *page, |
| 831 | enum node_stat_item idx, int val) |
| 832 | { |
| 833 | struct page *head = compound_head(page); /* rmap on tail pages */ |
| 834 | pg_data_t *pgdat = page_pgdat(page); |
| 835 | struct lruvec *lruvec; |
| 836 | |
| 837 | /* Untracked pages have no memcg, no lruvec. Update only the node */ |
| 838 | if (!head->mem_cgroup) { |
| 839 | __mod_node_page_state(pgdat, idx, val); |
| 840 | return; |
| 841 | } |
| 842 | |
| 843 | lruvec = mem_cgroup_lruvec(head->mem_cgroup, pgdat); |
| 844 | __mod_lruvec_state(lruvec, idx, val); |
| 845 | } |
| 846 | |
| 847 | static inline void mod_lruvec_page_state(struct page *page, |
| 848 | enum node_stat_item idx, int val) |
| 849 | { |
| 850 | unsigned long flags; |
| 851 | |
| 852 | local_irq_save(flags); |
| 853 | __mod_lruvec_page_state(page, idx, val); |
| 854 | local_irq_restore(flags); |
| 855 | } |
| 856 | |
| 857 | unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, |
| 858 | gfp_t gfp_mask, |
| 859 | unsigned long *total_scanned); |
| 860 | |
| 861 | void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, |
| 862 | unsigned long count); |
| 863 | |
| 864 | static inline void count_memcg_events(struct mem_cgroup *memcg, |
| 865 | enum vm_event_item idx, |
| 866 | unsigned long count) |
| 867 | { |
| 868 | unsigned long flags; |
| 869 | |
| 870 | local_irq_save(flags); |
| 871 | __count_memcg_events(memcg, idx, count); |
| 872 | local_irq_restore(flags); |
| 873 | } |
| 874 | |
| 875 | static inline void count_memcg_page_event(struct page *page, |
| 876 | enum vm_event_item idx) |
| 877 | { |
| 878 | if (page->mem_cgroup) |
| 879 | count_memcg_events(page->mem_cgroup, idx, 1); |
| 880 | } |
| 881 | |
| 882 | static inline void count_memcg_event_mm(struct mm_struct *mm, |
| 883 | enum vm_event_item idx) |
| 884 | { |
| 885 | struct mem_cgroup *memcg; |
| 886 | |
| 887 | if (mem_cgroup_disabled()) |
| 888 | return; |
| 889 | |
| 890 | rcu_read_lock(); |
| 891 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
| 892 | if (likely(memcg)) |
| 893 | count_memcg_events(memcg, idx, 1); |
| 894 | rcu_read_unlock(); |
| 895 | } |
| 896 | |
| 897 | static inline void memcg_memory_event(struct mem_cgroup *memcg, |
| 898 | enum memcg_memory_event event) |
| 899 | { |
| 900 | atomic_long_inc(&memcg->memory_events_local[event]); |
| 901 | cgroup_file_notify(&memcg->events_local_file); |
| 902 | |
| 903 | do { |
| 904 | atomic_long_inc(&memcg->memory_events[event]); |
| 905 | cgroup_file_notify(&memcg->events_file); |
| 906 | |
| 907 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) |
| 908 | break; |
| 909 | if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) |
| 910 | break; |
| 911 | } while ((memcg = parent_mem_cgroup(memcg)) && |
| 912 | !mem_cgroup_is_root(memcg)); |
| 913 | } |
| 914 | |
| 915 | static inline void memcg_memory_event_mm(struct mm_struct *mm, |
| 916 | enum memcg_memory_event event) |
| 917 | { |
| 918 | struct mem_cgroup *memcg; |
| 919 | |
| 920 | if (mem_cgroup_disabled()) |
| 921 | return; |
| 922 | |
| 923 | rcu_read_lock(); |
| 924 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
| 925 | if (likely(memcg)) |
| 926 | memcg_memory_event(memcg, event); |
| 927 | rcu_read_unlock(); |
| 928 | } |
| 929 | |
| 930 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 931 | void mem_cgroup_split_huge_fixup(struct page *head); |
| 932 | #endif |
| 933 | |
| 934 | #else /* CONFIG_MEMCG */ |
| 935 | |
| 936 | #define MEM_CGROUP_ID_SHIFT 0 |
| 937 | #define MEM_CGROUP_ID_MAX 0 |
| 938 | |
| 939 | struct mem_cgroup; |
| 940 | |
| 941 | static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) |
| 942 | { |
| 943 | return true; |
| 944 | } |
| 945 | |
| 946 | static inline bool mem_cgroup_disabled(void) |
| 947 | { |
| 948 | return true; |
| 949 | } |
| 950 | |
| 951 | static inline void memcg_memory_event(struct mem_cgroup *memcg, |
| 952 | enum memcg_memory_event event) |
| 953 | { |
| 954 | } |
| 955 | |
| 956 | static inline void memcg_memory_event_mm(struct mm_struct *mm, |
| 957 | enum memcg_memory_event event) |
| 958 | { |
| 959 | } |
| 960 | |
| 961 | static inline unsigned long mem_cgroup_protection(struct mem_cgroup *root, |
| 962 | struct mem_cgroup *memcg, |
| 963 | bool in_low_reclaim) |
| 964 | { |
| 965 | return 0; |
| 966 | } |
| 967 | |
| 968 | static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, |
| 969 | struct mem_cgroup *memcg) |
| 970 | { |
| 971 | } |
| 972 | |
| 973 | static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg) |
| 974 | { |
| 975 | return false; |
| 976 | } |
| 977 | |
| 978 | static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg) |
| 979 | { |
| 980 | return false; |
| 981 | } |
| 982 | |
| 983 | static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm, |
| 984 | gfp_t gfp_mask) |
| 985 | { |
| 986 | return 0; |
| 987 | } |
| 988 | |
| 989 | static inline void mem_cgroup_uncharge(struct page *page) |
| 990 | { |
| 991 | } |
| 992 | |
| 993 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
| 994 | { |
| 995 | } |
| 996 | |
| 997 | static inline void mem_cgroup_migrate(struct page *old, struct page *new) |
| 998 | { |
| 999 | } |
| 1000 | |
| 1001 | static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, |
| 1002 | struct pglist_data *pgdat) |
| 1003 | { |
| 1004 | return &pgdat->__lruvec; |
| 1005 | } |
| 1006 | |
| 1007 | static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page, |
| 1008 | struct pglist_data *pgdat) |
| 1009 | { |
| 1010 | return &pgdat->__lruvec; |
| 1011 | } |
| 1012 | |
| 1013 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) |
| 1014 | { |
| 1015 | return NULL; |
| 1016 | } |
| 1017 | |
| 1018 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
| 1019 | struct mem_cgroup *memcg) |
| 1020 | { |
| 1021 | return true; |
| 1022 | } |
| 1023 | |
| 1024 | static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) |
| 1025 | { |
| 1026 | return NULL; |
| 1027 | } |
| 1028 | |
| 1029 | static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page) |
| 1030 | { |
| 1031 | return NULL; |
| 1032 | } |
| 1033 | |
| 1034 | static inline void mem_cgroup_put(struct mem_cgroup *memcg) |
| 1035 | { |
| 1036 | } |
| 1037 | |
| 1038 | static inline struct mem_cgroup * |
| 1039 | mem_cgroup_iter(struct mem_cgroup *root, |
| 1040 | struct mem_cgroup *prev, |
| 1041 | struct mem_cgroup_reclaim_cookie *reclaim) |
| 1042 | { |
| 1043 | return NULL; |
| 1044 | } |
| 1045 | |
| 1046 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, |
| 1047 | struct mem_cgroup *prev) |
| 1048 | { |
| 1049 | } |
| 1050 | |
| 1051 | static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg, |
| 1052 | int (*fn)(struct task_struct *, void *), void *arg) |
| 1053 | { |
| 1054 | return 0; |
| 1055 | } |
| 1056 | |
| 1057 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
| 1058 | { |
| 1059 | return 0; |
| 1060 | } |
| 1061 | |
| 1062 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) |
| 1063 | { |
| 1064 | WARN_ON_ONCE(id); |
| 1065 | /* XXX: This should always return root_mem_cgroup */ |
| 1066 | return NULL; |
| 1067 | } |
| 1068 | |
| 1069 | static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) |
| 1070 | { |
| 1071 | return NULL; |
| 1072 | } |
| 1073 | |
| 1074 | static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) |
| 1075 | { |
| 1076 | return NULL; |
| 1077 | } |
| 1078 | |
| 1079 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
| 1080 | { |
| 1081 | return true; |
| 1082 | } |
| 1083 | |
| 1084 | static inline |
| 1085 | unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, |
| 1086 | enum lru_list lru, int zone_idx) |
| 1087 | { |
| 1088 | return 0; |
| 1089 | } |
| 1090 | |
| 1091 | static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) |
| 1092 | { |
| 1093 | return 0; |
| 1094 | } |
| 1095 | |
| 1096 | static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) |
| 1097 | { |
| 1098 | return 0; |
| 1099 | } |
| 1100 | |
| 1101 | static inline void |
| 1102 | mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) |
| 1103 | { |
| 1104 | } |
| 1105 | |
| 1106 | static inline void |
| 1107 | mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) |
| 1108 | { |
| 1109 | } |
| 1110 | |
| 1111 | static inline struct mem_cgroup *lock_page_memcg(struct page *page) |
| 1112 | { |
| 1113 | return NULL; |
| 1114 | } |
| 1115 | |
| 1116 | static inline void __unlock_page_memcg(struct mem_cgroup *memcg) |
| 1117 | { |
| 1118 | } |
| 1119 | |
| 1120 | static inline void unlock_page_memcg(struct page *page) |
| 1121 | { |
| 1122 | } |
| 1123 | |
| 1124 | static inline void mem_cgroup_handle_over_high(void) |
| 1125 | { |
| 1126 | } |
| 1127 | |
| 1128 | static inline void mem_cgroup_enter_user_fault(void) |
| 1129 | { |
| 1130 | } |
| 1131 | |
| 1132 | static inline void mem_cgroup_exit_user_fault(void) |
| 1133 | { |
| 1134 | } |
| 1135 | |
| 1136 | static inline bool task_in_memcg_oom(struct task_struct *p) |
| 1137 | { |
| 1138 | return false; |
| 1139 | } |
| 1140 | |
| 1141 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
| 1142 | { |
| 1143 | return false; |
| 1144 | } |
| 1145 | |
| 1146 | static inline struct mem_cgroup *mem_cgroup_get_oom_group( |
| 1147 | struct task_struct *victim, struct mem_cgroup *oom_domain) |
| 1148 | { |
| 1149 | return NULL; |
| 1150 | } |
| 1151 | |
| 1152 | static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) |
| 1153 | { |
| 1154 | } |
| 1155 | |
| 1156 | static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) |
| 1157 | { |
| 1158 | return 0; |
| 1159 | } |
| 1160 | |
| 1161 | static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg, |
| 1162 | int idx) |
| 1163 | { |
| 1164 | return 0; |
| 1165 | } |
| 1166 | |
| 1167 | static inline void __mod_memcg_state(struct mem_cgroup *memcg, |
| 1168 | int idx, |
| 1169 | int nr) |
| 1170 | { |
| 1171 | } |
| 1172 | |
| 1173 | static inline void mod_memcg_state(struct mem_cgroup *memcg, |
| 1174 | int idx, |
| 1175 | int nr) |
| 1176 | { |
| 1177 | } |
| 1178 | |
| 1179 | static inline void __mod_memcg_page_state(struct page *page, |
| 1180 | int idx, |
| 1181 | int nr) |
| 1182 | { |
| 1183 | } |
| 1184 | |
| 1185 | static inline void mod_memcg_page_state(struct page *page, |
| 1186 | int idx, |
| 1187 | int nr) |
| 1188 | { |
| 1189 | } |
| 1190 | |
| 1191 | static inline unsigned long lruvec_page_state(struct lruvec *lruvec, |
| 1192 | enum node_stat_item idx) |
| 1193 | { |
| 1194 | return node_page_state(lruvec_pgdat(lruvec), idx); |
| 1195 | } |
| 1196 | |
| 1197 | static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, |
| 1198 | enum node_stat_item idx) |
| 1199 | { |
| 1200 | return node_page_state(lruvec_pgdat(lruvec), idx); |
| 1201 | } |
| 1202 | |
| 1203 | static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec, |
| 1204 | enum node_stat_item idx, int val) |
| 1205 | { |
| 1206 | } |
| 1207 | |
| 1208 | static inline void __mod_lruvec_state(struct lruvec *lruvec, |
| 1209 | enum node_stat_item idx, int val) |
| 1210 | { |
| 1211 | __mod_node_page_state(lruvec_pgdat(lruvec), idx, val); |
| 1212 | } |
| 1213 | |
| 1214 | static inline void mod_lruvec_state(struct lruvec *lruvec, |
| 1215 | enum node_stat_item idx, int val) |
| 1216 | { |
| 1217 | mod_node_page_state(lruvec_pgdat(lruvec), idx, val); |
| 1218 | } |
| 1219 | |
| 1220 | static inline void __mod_lruvec_page_state(struct page *page, |
| 1221 | enum node_stat_item idx, int val) |
| 1222 | { |
| 1223 | __mod_node_page_state(page_pgdat(page), idx, val); |
| 1224 | } |
| 1225 | |
| 1226 | static inline void mod_lruvec_page_state(struct page *page, |
| 1227 | enum node_stat_item idx, int val) |
| 1228 | { |
| 1229 | mod_node_page_state(page_pgdat(page), idx, val); |
| 1230 | } |
| 1231 | |
| 1232 | static inline void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, |
| 1233 | int val) |
| 1234 | { |
| 1235 | struct page *page = virt_to_head_page(p); |
| 1236 | |
| 1237 | __mod_node_page_state(page_pgdat(page), idx, val); |
| 1238 | } |
| 1239 | |
| 1240 | static inline void mod_lruvec_slab_state(void *p, enum node_stat_item idx, |
| 1241 | int val) |
| 1242 | { |
| 1243 | struct page *page = virt_to_head_page(p); |
| 1244 | |
| 1245 | mod_node_page_state(page_pgdat(page), idx, val); |
| 1246 | } |
| 1247 | |
| 1248 | static inline void mod_memcg_obj_state(void *p, int idx, int val) |
| 1249 | { |
| 1250 | } |
| 1251 | |
| 1252 | static inline |
| 1253 | unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, |
| 1254 | gfp_t gfp_mask, |
| 1255 | unsigned long *total_scanned) |
| 1256 | { |
| 1257 | return 0; |
| 1258 | } |
| 1259 | |
| 1260 | static inline void mem_cgroup_split_huge_fixup(struct page *head) |
| 1261 | { |
| 1262 | } |
| 1263 | |
| 1264 | static inline void count_memcg_events(struct mem_cgroup *memcg, |
| 1265 | enum vm_event_item idx, |
| 1266 | unsigned long count) |
| 1267 | { |
| 1268 | } |
| 1269 | |
| 1270 | static inline void __count_memcg_events(struct mem_cgroup *memcg, |
| 1271 | enum vm_event_item idx, |
| 1272 | unsigned long count) |
| 1273 | { |
| 1274 | } |
| 1275 | |
| 1276 | static inline void count_memcg_page_event(struct page *page, |
| 1277 | int idx) |
| 1278 | { |
| 1279 | } |
| 1280 | |
| 1281 | static inline |
| 1282 | void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) |
| 1283 | { |
| 1284 | } |
| 1285 | #endif /* CONFIG_MEMCG */ |
| 1286 | |
| 1287 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1288 | static inline void __inc_memcg_state(struct mem_cgroup *memcg, |
| 1289 | int idx) |
| 1290 | { |
| 1291 | __mod_memcg_state(memcg, idx, 1); |
| 1292 | } |
| 1293 | |
| 1294 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1295 | static inline void __dec_memcg_state(struct mem_cgroup *memcg, |
| 1296 | int idx) |
| 1297 | { |
| 1298 | __mod_memcg_state(memcg, idx, -1); |
| 1299 | } |
| 1300 | |
| 1301 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1302 | static inline void __inc_memcg_page_state(struct page *page, |
| 1303 | int idx) |
| 1304 | { |
| 1305 | __mod_memcg_page_state(page, idx, 1); |
| 1306 | } |
| 1307 | |
| 1308 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1309 | static inline void __dec_memcg_page_state(struct page *page, |
| 1310 | int idx) |
| 1311 | { |
| 1312 | __mod_memcg_page_state(page, idx, -1); |
| 1313 | } |
| 1314 | |
| 1315 | static inline void __inc_lruvec_state(struct lruvec *lruvec, |
| 1316 | enum node_stat_item idx) |
| 1317 | { |
| 1318 | __mod_lruvec_state(lruvec, idx, 1); |
| 1319 | } |
| 1320 | |
| 1321 | static inline void __dec_lruvec_state(struct lruvec *lruvec, |
| 1322 | enum node_stat_item idx) |
| 1323 | { |
| 1324 | __mod_lruvec_state(lruvec, idx, -1); |
| 1325 | } |
| 1326 | |
| 1327 | static inline void __inc_lruvec_page_state(struct page *page, |
| 1328 | enum node_stat_item idx) |
| 1329 | { |
| 1330 | __mod_lruvec_page_state(page, idx, 1); |
| 1331 | } |
| 1332 | |
| 1333 | static inline void __dec_lruvec_page_state(struct page *page, |
| 1334 | enum node_stat_item idx) |
| 1335 | { |
| 1336 | __mod_lruvec_page_state(page, idx, -1); |
| 1337 | } |
| 1338 | |
| 1339 | static inline void __inc_lruvec_slab_state(void *p, enum node_stat_item idx) |
| 1340 | { |
| 1341 | __mod_lruvec_slab_state(p, idx, 1); |
| 1342 | } |
| 1343 | |
| 1344 | static inline void __dec_lruvec_slab_state(void *p, enum node_stat_item idx) |
| 1345 | { |
| 1346 | __mod_lruvec_slab_state(p, idx, -1); |
| 1347 | } |
| 1348 | |
| 1349 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1350 | static inline void inc_memcg_state(struct mem_cgroup *memcg, |
| 1351 | int idx) |
| 1352 | { |
| 1353 | mod_memcg_state(memcg, idx, 1); |
| 1354 | } |
| 1355 | |
| 1356 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1357 | static inline void dec_memcg_state(struct mem_cgroup *memcg, |
| 1358 | int idx) |
| 1359 | { |
| 1360 | mod_memcg_state(memcg, idx, -1); |
| 1361 | } |
| 1362 | |
| 1363 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1364 | static inline void inc_memcg_page_state(struct page *page, |
| 1365 | int idx) |
| 1366 | { |
| 1367 | mod_memcg_page_state(page, idx, 1); |
| 1368 | } |
| 1369 | |
| 1370 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
| 1371 | static inline void dec_memcg_page_state(struct page *page, |
| 1372 | int idx) |
| 1373 | { |
| 1374 | mod_memcg_page_state(page, idx, -1); |
| 1375 | } |
| 1376 | |
| 1377 | static inline void inc_lruvec_state(struct lruvec *lruvec, |
| 1378 | enum node_stat_item idx) |
| 1379 | { |
| 1380 | mod_lruvec_state(lruvec, idx, 1); |
| 1381 | } |
| 1382 | |
| 1383 | static inline void dec_lruvec_state(struct lruvec *lruvec, |
| 1384 | enum node_stat_item idx) |
| 1385 | { |
| 1386 | mod_lruvec_state(lruvec, idx, -1); |
| 1387 | } |
| 1388 | |
| 1389 | static inline void inc_lruvec_page_state(struct page *page, |
| 1390 | enum node_stat_item idx) |
| 1391 | { |
| 1392 | mod_lruvec_page_state(page, idx, 1); |
| 1393 | } |
| 1394 | |
| 1395 | static inline void dec_lruvec_page_state(struct page *page, |
| 1396 | enum node_stat_item idx) |
| 1397 | { |
| 1398 | mod_lruvec_page_state(page, idx, -1); |
| 1399 | } |
| 1400 | |
| 1401 | static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) |
| 1402 | { |
| 1403 | struct mem_cgroup *memcg; |
| 1404 | |
| 1405 | memcg = lruvec_memcg(lruvec); |
| 1406 | if (!memcg) |
| 1407 | return NULL; |
| 1408 | memcg = parent_mem_cgroup(memcg); |
| 1409 | if (!memcg) |
| 1410 | return NULL; |
| 1411 | return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); |
| 1412 | } |
| 1413 | |
| 1414 | #ifdef CONFIG_CGROUP_WRITEBACK |
| 1415 | |
| 1416 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
| 1417 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
| 1418 | unsigned long *pheadroom, unsigned long *pdirty, |
| 1419 | unsigned long *pwriteback); |
| 1420 | |
| 1421 | void mem_cgroup_track_foreign_dirty_slowpath(struct page *page, |
| 1422 | struct bdi_writeback *wb); |
| 1423 | |
| 1424 | static inline void mem_cgroup_track_foreign_dirty(struct page *page, |
| 1425 | struct bdi_writeback *wb) |
| 1426 | { |
| 1427 | if (mem_cgroup_disabled()) |
| 1428 | return; |
| 1429 | |
| 1430 | if (unlikely(&page->mem_cgroup->css != wb->memcg_css)) |
| 1431 | mem_cgroup_track_foreign_dirty_slowpath(page, wb); |
| 1432 | } |
| 1433 | |
| 1434 | void mem_cgroup_flush_foreign(struct bdi_writeback *wb); |
| 1435 | |
| 1436 | #else /* CONFIG_CGROUP_WRITEBACK */ |
| 1437 | |
| 1438 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) |
| 1439 | { |
| 1440 | return NULL; |
| 1441 | } |
| 1442 | |
| 1443 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
| 1444 | unsigned long *pfilepages, |
| 1445 | unsigned long *pheadroom, |
| 1446 | unsigned long *pdirty, |
| 1447 | unsigned long *pwriteback) |
| 1448 | { |
| 1449 | } |
| 1450 | |
| 1451 | static inline void mem_cgroup_track_foreign_dirty(struct page *page, |
| 1452 | struct bdi_writeback *wb) |
| 1453 | { |
| 1454 | } |
| 1455 | |
| 1456 | static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) |
| 1457 | { |
| 1458 | } |
| 1459 | |
| 1460 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
| 1461 | |
| 1462 | struct sock; |
| 1463 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
| 1464 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
| 1465 | #ifdef CONFIG_MEMCG |
| 1466 | extern struct static_key_false memcg_sockets_enabled_key; |
| 1467 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) |
| 1468 | void mem_cgroup_sk_alloc(struct sock *sk); |
| 1469 | void mem_cgroup_sk_free(struct sock *sk); |
| 1470 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
| 1471 | { |
| 1472 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
| 1473 | return true; |
| 1474 | do { |
| 1475 | if (time_before(jiffies, memcg->socket_pressure)) |
| 1476 | return true; |
| 1477 | } while ((memcg = parent_mem_cgroup(memcg))); |
| 1478 | return false; |
| 1479 | } |
| 1480 | |
| 1481 | extern int memcg_expand_shrinker_maps(int new_id); |
| 1482 | |
| 1483 | extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg, |
| 1484 | int nid, int shrinker_id); |
| 1485 | #else |
| 1486 | #define mem_cgroup_sockets_enabled 0 |
| 1487 | static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; |
| 1488 | static inline void mem_cgroup_sk_free(struct sock *sk) { }; |
| 1489 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
| 1490 | { |
| 1491 | return false; |
| 1492 | } |
| 1493 | |
| 1494 | static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg, |
| 1495 | int nid, int shrinker_id) |
| 1496 | { |
| 1497 | } |
| 1498 | #endif |
| 1499 | |
| 1500 | #ifdef CONFIG_MEMCG_KMEM |
| 1501 | int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp, |
| 1502 | unsigned int nr_pages); |
| 1503 | void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages); |
| 1504 | int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); |
| 1505 | void __memcg_kmem_uncharge_page(struct page *page, int order); |
| 1506 | |
| 1507 | struct obj_cgroup *get_obj_cgroup_from_current(void); |
| 1508 | |
| 1509 | int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); |
| 1510 | void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); |
| 1511 | |
| 1512 | extern struct static_key_false memcg_kmem_enabled_key; |
| 1513 | |
| 1514 | extern int memcg_nr_cache_ids; |
| 1515 | void memcg_get_cache_ids(void); |
| 1516 | void memcg_put_cache_ids(void); |
| 1517 | |
| 1518 | /* |
| 1519 | * Helper macro to loop through all memcg-specific caches. Callers must still |
| 1520 | * check if the cache is valid (it is either valid or NULL). |
| 1521 | * the slab_mutex must be held when looping through those caches |
| 1522 | */ |
| 1523 | #define for_each_memcg_cache_index(_idx) \ |
| 1524 | for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++) |
| 1525 | |
| 1526 | static inline bool memcg_kmem_enabled(void) |
| 1527 | { |
| 1528 | return static_branch_likely(&memcg_kmem_enabled_key); |
| 1529 | } |
| 1530 | |
| 1531 | static inline bool memcg_kmem_bypass(void) |
| 1532 | { |
| 1533 | if (in_interrupt()) |
| 1534 | return true; |
| 1535 | |
| 1536 | /* Allow remote memcg charging in kthread contexts. */ |
| 1537 | if ((!current->mm || (current->flags & PF_KTHREAD)) && |
| 1538 | !current->active_memcg) |
| 1539 | return true; |
| 1540 | return false; |
| 1541 | } |
| 1542 | |
| 1543 | static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, |
| 1544 | int order) |
| 1545 | { |
| 1546 | if (memcg_kmem_enabled()) |
| 1547 | return __memcg_kmem_charge_page(page, gfp, order); |
| 1548 | return 0; |
| 1549 | } |
| 1550 | |
| 1551 | static inline void memcg_kmem_uncharge_page(struct page *page, int order) |
| 1552 | { |
| 1553 | if (memcg_kmem_enabled()) |
| 1554 | __memcg_kmem_uncharge_page(page, order); |
| 1555 | } |
| 1556 | |
| 1557 | static inline int memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp, |
| 1558 | unsigned int nr_pages) |
| 1559 | { |
| 1560 | if (memcg_kmem_enabled()) |
| 1561 | return __memcg_kmem_charge(memcg, gfp, nr_pages); |
| 1562 | return 0; |
| 1563 | } |
| 1564 | |
| 1565 | static inline void memcg_kmem_uncharge(struct mem_cgroup *memcg, |
| 1566 | unsigned int nr_pages) |
| 1567 | { |
| 1568 | if (memcg_kmem_enabled()) |
| 1569 | __memcg_kmem_uncharge(memcg, nr_pages); |
| 1570 | } |
| 1571 | |
| 1572 | /* |
| 1573 | * helper for accessing a memcg's index. It will be used as an index in the |
| 1574 | * child cache array in kmem_cache, and also to derive its name. This function |
| 1575 | * will return -1 when this is not a kmem-limited memcg. |
| 1576 | */ |
| 1577 | static inline int memcg_cache_id(struct mem_cgroup *memcg) |
| 1578 | { |
| 1579 | return memcg ? memcg->kmemcg_id : -1; |
| 1580 | } |
| 1581 | |
| 1582 | struct mem_cgroup *mem_cgroup_from_obj(void *p); |
| 1583 | |
| 1584 | #else |
| 1585 | |
| 1586 | static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, |
| 1587 | int order) |
| 1588 | { |
| 1589 | return 0; |
| 1590 | } |
| 1591 | |
| 1592 | static inline void memcg_kmem_uncharge_page(struct page *page, int order) |
| 1593 | { |
| 1594 | } |
| 1595 | |
| 1596 | static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, |
| 1597 | int order) |
| 1598 | { |
| 1599 | return 0; |
| 1600 | } |
| 1601 | |
| 1602 | static inline void __memcg_kmem_uncharge_page(struct page *page, int order) |
| 1603 | { |
| 1604 | } |
| 1605 | |
| 1606 | #define for_each_memcg_cache_index(_idx) \ |
| 1607 | for (; NULL; ) |
| 1608 | |
| 1609 | static inline bool memcg_kmem_enabled(void) |
| 1610 | { |
| 1611 | return false; |
| 1612 | } |
| 1613 | |
| 1614 | static inline int memcg_cache_id(struct mem_cgroup *memcg) |
| 1615 | { |
| 1616 | return -1; |
| 1617 | } |
| 1618 | |
| 1619 | static inline void memcg_get_cache_ids(void) |
| 1620 | { |
| 1621 | } |
| 1622 | |
| 1623 | static inline void memcg_put_cache_ids(void) |
| 1624 | { |
| 1625 | } |
| 1626 | |
| 1627 | static inline struct mem_cgroup *mem_cgroup_from_obj(void *p) |
| 1628 | { |
| 1629 | return NULL; |
| 1630 | } |
| 1631 | |
| 1632 | #endif /* CONFIG_MEMCG_KMEM */ |
| 1633 | |
| 1634 | #endif /* _LINUX_MEMCONTROL_H */ |