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/kernel.h>
18 #include <linux/page_counter.h>
19 #include <linux/vmpressure.h>
20 #include <linux/eventfd.h>
21 #include <linux/mm.h>
22 #include <linux/vmstat.h>
23 #include <linux/writeback.h>
24 #include <linux/page-flags.h>
25 #include <linux/shrinker.h>
26
27 struct mem_cgroup;
28 struct obj_cgroup;
29 struct page;
30 struct mm_struct;
31 struct kmem_cache;
32
33 /* Cgroup-specific page state, on top of universal node page state */
34 enum memcg_stat_item {
35 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS,
36 MEMCG_SOCK,
37 MEMCG_PERCPU_B,
38 MEMCG_VMALLOC,
39 MEMCG_KMEM,
40 MEMCG_ZSWAP_B,
41 MEMCG_ZSWAPPED,
42 MEMCG_NR_STAT,
43 };
44
45 enum memcg_memory_event {
46 MEMCG_LOW,
47 MEMCG_HIGH,
48 MEMCG_MAX,
49 MEMCG_OOM,
50 MEMCG_OOM_KILL,
51 MEMCG_OOM_GROUP_KILL,
52 MEMCG_SWAP_HIGH,
53 MEMCG_SWAP_MAX,
54 MEMCG_SWAP_FAIL,
55 MEMCG_NR_MEMORY_EVENTS,
56 };
57
58 struct mem_cgroup_reclaim_cookie {
59 pg_data_t *pgdat;
60 unsigned int generation;
61 };
62
63 #ifdef CONFIG_MEMCG
64
65 #define MEM_CGROUP_ID_SHIFT 16
66
67 struct mem_cgroup_id {
68 int id;
69 refcount_t ref;
70 };
71
72 struct memcg_vmstats_percpu;
73 struct memcg_vmstats;
74 struct lruvec_stats_percpu;
75 struct lruvec_stats;
76
77 struct mem_cgroup_reclaim_iter {
78 struct mem_cgroup *position;
79 /* scan generation, increased every round-trip */
80 unsigned int generation;
81 };
82
83 /*
84 * per-node information in memory controller.
85 */
86 struct mem_cgroup_per_node {
87 /* Keep the read-only fields at the start */
88 struct mem_cgroup *memcg; /* Back pointer, we cannot */
89 /* use container_of */
90
91 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu;
92 struct lruvec_stats *lruvec_stats;
93 struct shrinker_info __rcu *shrinker_info;
94
95 #ifdef CONFIG_MEMCG_V1
96 /*
97 * Memcg-v1 only stuff in middle as buffer between read mostly fields
98 * and update often fields to avoid false sharing. If v1 stuff is
99 * not present, an explicit padding is needed.
100 */
101
102 struct rb_node tree_node; /* RB tree node */
103 unsigned long usage_in_excess;/* Set to the value by which */
104 /* the soft limit is exceeded*/
105 bool on_tree;
106 #else
107 CACHELINE_PADDING(_pad1_);
108 #endif
109
110 /* Fields which get updated often at the end. */
111 struct lruvec lruvec;
112 CACHELINE_PADDING(_pad2_);
113 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
114 struct mem_cgroup_reclaim_iter iter;
115 };
116
117 struct mem_cgroup_threshold {
118 struct eventfd_ctx *eventfd;
119 unsigned long threshold;
120 };
121
122 /* For threshold */
123 struct mem_cgroup_threshold_ary {
124 /* An array index points to threshold just below or equal to usage. */
125 int current_threshold;
126 /* Size of entries[] */
127 unsigned int size;
128 /* Array of thresholds */
129 struct mem_cgroup_threshold entries[] __counted_by(size);
130 };
131
132 struct mem_cgroup_thresholds {
133 /* Primary thresholds array */
134 struct mem_cgroup_threshold_ary *primary;
135 /*
136 * Spare threshold array.
137 * This is needed to make mem_cgroup_unregister_event() "never fail".
138 * It must be able to store at least primary->size - 1 entries.
139 */
140 struct mem_cgroup_threshold_ary *spare;
141 };
142
143 /*
144 * Remember four most recent foreign writebacks with dirty pages in this
145 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
146 * one in a given round, we're likely to catch it later if it keeps
147 * foreign-dirtying, so a fairly low count should be enough.
148 *
149 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
150 */
151 #define MEMCG_CGWB_FRN_CNT 4
152
153 struct memcg_cgwb_frn {
154 u64 bdi_id; /* bdi->id of the foreign inode */
155 int memcg_id; /* memcg->css.id of foreign inode */
156 u64 at; /* jiffies_64 at the time of dirtying */
157 struct wb_completion done; /* tracks in-flight foreign writebacks */
158 };
159
160 /*
161 * Bucket for arbitrarily byte-sized objects charged to a memory
162 * cgroup. The bucket can be reparented in one piece when the cgroup
163 * is destroyed, without having to round up the individual references
164 * of all live memory objects in the wild.
165 */
166 struct obj_cgroup {
167 struct percpu_ref refcnt;
168 struct mem_cgroup *memcg;
169 atomic_t nr_charged_bytes;
170 union {
171 struct list_head list; /* protected by objcg_lock */
172 struct rcu_head rcu;
173 };
174 };
175
176 /*
177 * The memory controller data structure. The memory controller controls both
178 * page cache and RSS per cgroup. We would eventually like to provide
179 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
180 * to help the administrator determine what knobs to tune.
181 */
182 struct mem_cgroup {
183 struct cgroup_subsys_state css;
184
185 /* Private memcg ID. Used to ID objects that outlive the cgroup */
186 struct mem_cgroup_id id;
187
188 /* Accounted resources */
189 struct page_counter memory; /* Both v1 & v2 */
190
191 union {
192 struct page_counter swap; /* v2 only */
193 struct page_counter memsw; /* v1 only */
194 };
195
196 /* Range enforcement for interrupt charges */
197 struct work_struct high_work;
198
199 #ifdef CONFIG_ZSWAP
200 unsigned long zswap_max;
201
202 /*
203 * Prevent pages from this memcg from being written back from zswap to
204 * swap, and from being swapped out on zswap store failures.
205 */
206 bool zswap_writeback;
207 #endif
208
209 /* vmpressure notifications */
210 struct vmpressure vmpressure;
211
212 /*
213 * Should the OOM killer kill all belonging tasks, had it kill one?
214 */
215 bool oom_group;
216
217 int swappiness;
218
219 /* memory.events and memory.events.local */
220 struct cgroup_file events_file;
221 struct cgroup_file events_local_file;
222
223 /* handle for "memory.swap.events" */
224 struct cgroup_file swap_events_file;
225
226 /* memory.stat */
227 struct memcg_vmstats *vmstats;
228
229 /* memory.events */
230 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
231 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
232
233 /*
234 * Hint of reclaim pressure for socket memroy management. Note
235 * that this indicator should NOT be used in legacy cgroup mode
236 * where socket memory is accounted/charged separately.
237 */
238 unsigned long socket_pressure;
239
240 int kmemcg_id;
241 /*
242 * memcg->objcg is wiped out as a part of the objcg repaprenting
243 * process. memcg->orig_objcg preserves a pointer (and a reference)
244 * to the original objcg until the end of live of memcg.
245 */
246 struct obj_cgroup __rcu *objcg;
247 struct obj_cgroup *orig_objcg;
248 /* list of inherited objcgs, protected by objcg_lock */
249 struct list_head objcg_list;
250
251 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
252
253 #ifdef CONFIG_CGROUP_WRITEBACK
254 struct list_head cgwb_list;
255 struct wb_domain cgwb_domain;
256 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
257 #endif
258
259 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
260 struct deferred_split deferred_split_queue;
261 #endif
262
263 #ifdef CONFIG_LRU_GEN_WALKS_MMU
264 /* per-memcg mm_struct list */
265 struct lru_gen_mm_list mm_list;
266 #endif
267
268 #ifdef CONFIG_MEMCG_V1
269 /* Legacy consumer-oriented counters */
270 struct page_counter kmem; /* v1 only */
271 struct page_counter tcpmem; /* v1 only */
272
273 unsigned long soft_limit;
274
275 /* protected by memcg_oom_lock */
276 bool oom_lock;
277 int under_oom;
278
279 /* OOM-Killer disable */
280 int oom_kill_disable;
281
282 /* protect arrays of thresholds */
283 struct mutex thresholds_lock;
284
285 /* thresholds for memory usage. RCU-protected */
286 struct mem_cgroup_thresholds thresholds;
287
288 /* thresholds for mem+swap usage. RCU-protected */
289 struct mem_cgroup_thresholds memsw_thresholds;
290
291 /* For oom notifier event fd */
292 struct list_head oom_notify;
293
294 /*
295 * Should we move charges of a task when a task is moved into this
296 * mem_cgroup ? And what type of charges should we move ?
297 */
298 unsigned long move_charge_at_immigrate;
299 /* taken only while moving_account > 0 */
300 spinlock_t move_lock;
301 unsigned long move_lock_flags;
302
303 /* Legacy tcp memory accounting */
304 bool tcpmem_active;
305 int tcpmem_pressure;
306
307 /*
308 * set > 0 if pages under this cgroup are moving to other cgroup.
309 */
310 atomic_t moving_account;
311 struct task_struct *move_lock_task;
312
313 /* List of events which userspace want to receive */
314 struct list_head event_list;
315 spinlock_t event_list_lock;
316 #endif /* CONFIG_MEMCG_V1 */
317
318 struct mem_cgroup_per_node *nodeinfo[];
319 };
320
321 /*
322 * size of first charge trial.
323 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the
324 * workload.
325 */
326 #define MEMCG_CHARGE_BATCH 64U
327
328 extern struct mem_cgroup *root_mem_cgroup;
329
330 enum page_memcg_data_flags {
331 /* page->memcg_data is a pointer to an slabobj_ext vector */
332 MEMCG_DATA_OBJEXTS = (1UL << 0),
333 /* page has been accounted as a non-slab kernel page */
334 MEMCG_DATA_KMEM = (1UL << 1),
335 /* the next bit after the last actual flag */
336 __NR_MEMCG_DATA_FLAGS = (1UL << 2),
337 };
338
339 #define __FIRST_OBJEXT_FLAG __NR_MEMCG_DATA_FLAGS
340
341 #else /* CONFIG_MEMCG */
342
343 #define __FIRST_OBJEXT_FLAG (1UL << 0)
344
345 #endif /* CONFIG_MEMCG */
346
347 enum objext_flags {
348 /* slabobj_ext vector failed to allocate */
349 OBJEXTS_ALLOC_FAIL = __FIRST_OBJEXT_FLAG,
350 /* the next bit after the last actual flag */
351 __NR_OBJEXTS_FLAGS = (__FIRST_OBJEXT_FLAG << 1),
352 };
353
354 #define OBJEXTS_FLAGS_MASK (__NR_OBJEXTS_FLAGS - 1)
355
356 #ifdef CONFIG_MEMCG
357
358 static inline bool folio_memcg_kmem(struct folio *folio);
359
360 /*
361 * After the initialization objcg->memcg is always pointing at
362 * a valid memcg, but can be atomically swapped to the parent memcg.
363 *
364 * The caller must ensure that the returned memcg won't be released:
365 * e.g. acquire the rcu_read_lock or css_set_lock.
366 */
367 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
368 {
369 return READ_ONCE(objcg->memcg);
370 }
371
372 /*
373 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio
374 * @folio: Pointer to the folio.
375 *
376 * Returns a pointer to the memory cgroup associated with the folio,
377 * or NULL. This function assumes that the folio is known to have a
378 * proper memory cgroup pointer. It's not safe to call this function
379 * against some type of folios, e.g. slab folios or ex-slab folios or
380 * kmem folios.
381 */
382 static inline struct mem_cgroup *__folio_memcg(struct folio *folio)
383 {
384 unsigned long memcg_data = folio->memcg_data;
385
386 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
387 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio);
388 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio);
389
390 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
391 }
392
393 /*
394 * __folio_objcg - get the object cgroup associated with a kmem folio.
395 * @folio: Pointer to the folio.
396 *
397 * Returns a pointer to the object cgroup associated with the folio,
398 * or NULL. This function assumes that the folio is known to have a
399 * proper object cgroup pointer. It's not safe to call this function
400 * against some type of folios, e.g. slab folios or ex-slab folios or
401 * LRU folios.
402 */
403 static inline struct obj_cgroup *__folio_objcg(struct folio *folio)
404 {
405 unsigned long memcg_data = folio->memcg_data;
406
407 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
408 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio);
409 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio);
410
411 return (struct obj_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
412 }
413
414 /*
415 * folio_memcg - Get the memory cgroup associated with a folio.
416 * @folio: Pointer to the folio.
417 *
418 * Returns a pointer to the memory cgroup associated with the folio,
419 * or NULL. This function assumes that the folio is known to have a
420 * proper memory cgroup pointer. It's not safe to call this function
421 * against some type of folios, e.g. slab folios or ex-slab folios.
422 *
423 * For a non-kmem folio any of the following ensures folio and memcg binding
424 * stability:
425 *
426 * - the folio lock
427 * - LRU isolation
428 * - folio_memcg_lock()
429 * - exclusive reference
430 * - mem_cgroup_trylock_pages()
431 *
432 * For a kmem folio a caller should hold an rcu read lock to protect memcg
433 * associated with a kmem folio from being released.
434 */
435 static inline struct mem_cgroup *folio_memcg(struct folio *folio)
436 {
437 if (folio_memcg_kmem(folio))
438 return obj_cgroup_memcg(__folio_objcg(folio));
439 return __folio_memcg(folio);
440 }
441
442 /**
443 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio.
444 * @folio: Pointer to the folio.
445 *
446 * This function assumes that the folio is known to have a
447 * proper memory cgroup pointer. It's not safe to call this function
448 * against some type of folios, e.g. slab folios or ex-slab folios.
449 *
450 * Return: A pointer to the memory cgroup associated with the folio,
451 * or NULL.
452 */
453 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
454 {
455 unsigned long memcg_data = READ_ONCE(folio->memcg_data);
456
457 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
458 WARN_ON_ONCE(!rcu_read_lock_held());
459
460 if (memcg_data & MEMCG_DATA_KMEM) {
461 struct obj_cgroup *objcg;
462
463 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
464 return obj_cgroup_memcg(objcg);
465 }
466
467 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
468 }
469
470 /*
471 * folio_memcg_check - Get the memory cgroup associated with a folio.
472 * @folio: Pointer to the folio.
473 *
474 * Returns a pointer to the memory cgroup associated with the folio,
475 * or NULL. This function unlike folio_memcg() can take any folio
476 * as an argument. It has to be used in cases when it's not known if a folio
477 * has an associated memory cgroup pointer or an object cgroups vector or
478 * an object cgroup.
479 *
480 * For a non-kmem folio any of the following ensures folio and memcg binding
481 * stability:
482 *
483 * - the folio lock
484 * - LRU isolation
485 * - lock_folio_memcg()
486 * - exclusive reference
487 * - mem_cgroup_trylock_pages()
488 *
489 * For a kmem folio a caller should hold an rcu read lock to protect memcg
490 * associated with a kmem folio from being released.
491 */
492 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio)
493 {
494 /*
495 * Because folio->memcg_data might be changed asynchronously
496 * for slabs, READ_ONCE() should be used here.
497 */
498 unsigned long memcg_data = READ_ONCE(folio->memcg_data);
499
500 if (memcg_data & MEMCG_DATA_OBJEXTS)
501 return NULL;
502
503 if (memcg_data & MEMCG_DATA_KMEM) {
504 struct obj_cgroup *objcg;
505
506 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
507 return obj_cgroup_memcg(objcg);
508 }
509
510 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
511 }
512
513 static inline struct mem_cgroup *page_memcg_check(struct page *page)
514 {
515 if (PageTail(page))
516 return NULL;
517 return folio_memcg_check((struct folio *)page);
518 }
519
520 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg)
521 {
522 struct mem_cgroup *memcg;
523
524 rcu_read_lock();
525 retry:
526 memcg = obj_cgroup_memcg(objcg);
527 if (unlikely(!css_tryget(&memcg->css)))
528 goto retry;
529 rcu_read_unlock();
530
531 return memcg;
532 }
533
534 /*
535 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set.
536 * @folio: Pointer to the folio.
537 *
538 * Checks if the folio has MemcgKmem flag set. The caller must ensure
539 * that the folio has an associated memory cgroup. It's not safe to call
540 * this function against some types of folios, e.g. slab folios.
541 */
542 static inline bool folio_memcg_kmem(struct folio *folio)
543 {
544 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page);
545 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJEXTS, folio);
546 return folio->memcg_data & MEMCG_DATA_KMEM;
547 }
548
549 static inline bool PageMemcgKmem(struct page *page)
550 {
551 return folio_memcg_kmem(page_folio(page));
552 }
553
554 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
555 {
556 return (memcg == root_mem_cgroup);
557 }
558
559 static inline bool mem_cgroup_disabled(void)
560 {
561 return !cgroup_subsys_enabled(memory_cgrp_subsys);
562 }
563
564 static inline void mem_cgroup_protection(struct mem_cgroup *root,
565 struct mem_cgroup *memcg,
566 unsigned long *min,
567 unsigned long *low)
568 {
569 *min = *low = 0;
570
571 if (mem_cgroup_disabled())
572 return;
573
574 /*
575 * There is no reclaim protection applied to a targeted reclaim.
576 * We are special casing this specific case here because
577 * mem_cgroup_calculate_protection is not robust enough to keep
578 * the protection invariant for calculated effective values for
579 * parallel reclaimers with different reclaim target. This is
580 * especially a problem for tail memcgs (as they have pages on LRU)
581 * which would want to have effective values 0 for targeted reclaim
582 * but a different value for external reclaim.
583 *
584 * Example
585 * Let's have global and A's reclaim in parallel:
586 * |
587 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
588 * |\
589 * | C (low = 1G, usage = 2.5G)
590 * B (low = 1G, usage = 0.5G)
591 *
592 * For the global reclaim
593 * A.elow = A.low
594 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
595 * C.elow = min(C.usage, C.low)
596 *
597 * With the effective values resetting we have A reclaim
598 * A.elow = 0
599 * B.elow = B.low
600 * C.elow = C.low
601 *
602 * If the global reclaim races with A's reclaim then
603 * B.elow = C.elow = 0 because children_low_usage > A.elow)
604 * is possible and reclaiming B would be violating the protection.
605 *
606 */
607 if (root == memcg)
608 return;
609
610 *min = READ_ONCE(memcg->memory.emin);
611 *low = READ_ONCE(memcg->memory.elow);
612 }
613
614 void mem_cgroup_calculate_protection(struct mem_cgroup *root,
615 struct mem_cgroup *memcg);
616
617 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target,
618 struct mem_cgroup *memcg)
619 {
620 /*
621 * The root memcg doesn't account charges, and doesn't support
622 * protection. The target memcg's protection is ignored, see
623 * mem_cgroup_calculate_protection() and mem_cgroup_protection()
624 */
625 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) ||
626 memcg == target;
627 }
628
629 static inline bool mem_cgroup_below_low(struct mem_cgroup *target,
630 struct mem_cgroup *memcg)
631 {
632 if (mem_cgroup_unprotected(target, memcg))
633 return false;
634
635 return READ_ONCE(memcg->memory.elow) >=
636 page_counter_read(&memcg->memory);
637 }
638
639 static inline bool mem_cgroup_below_min(struct mem_cgroup *target,
640 struct mem_cgroup *memcg)
641 {
642 if (mem_cgroup_unprotected(target, memcg))
643 return false;
644
645 return READ_ONCE(memcg->memory.emin) >=
646 page_counter_read(&memcg->memory);
647 }
648
649 void mem_cgroup_commit_charge(struct folio *folio, struct mem_cgroup *memcg);
650
651 int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp);
652
653 /**
654 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup.
655 * @folio: Folio to charge.
656 * @mm: mm context of the allocating task.
657 * @gfp: Reclaim mode.
658 *
659 * Try to charge @folio to the memcg that @mm belongs to, reclaiming
660 * pages according to @gfp if necessary. If @mm is NULL, try to
661 * charge to the active memcg.
662 *
663 * Do not use this for folios allocated for swapin.
664 *
665 * Return: 0 on success. Otherwise, an error code is returned.
666 */
667 static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm,
668 gfp_t gfp)
669 {
670 if (mem_cgroup_disabled())
671 return 0;
672 return __mem_cgroup_charge(folio, mm, gfp);
673 }
674
675 int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg, gfp_t gfp,
676 long nr_pages);
677
678 int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm,
679 gfp_t gfp, swp_entry_t entry);
680 void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry);
681
682 void __mem_cgroup_uncharge(struct folio *folio);
683
684 /**
685 * mem_cgroup_uncharge - Uncharge a folio.
686 * @folio: Folio to uncharge.
687 *
688 * Uncharge a folio previously charged with mem_cgroup_charge().
689 */
690 static inline void mem_cgroup_uncharge(struct folio *folio)
691 {
692 if (mem_cgroup_disabled())
693 return;
694 __mem_cgroup_uncharge(folio);
695 }
696
697 void __mem_cgroup_uncharge_folios(struct folio_batch *folios);
698 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios)
699 {
700 if (mem_cgroup_disabled())
701 return;
702 __mem_cgroup_uncharge_folios(folios);
703 }
704
705 void mem_cgroup_cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages);
706 void mem_cgroup_replace_folio(struct folio *old, struct folio *new);
707 void mem_cgroup_migrate(struct folio *old, struct folio *new);
708
709 /**
710 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
711 * @memcg: memcg of the wanted lruvec
712 * @pgdat: pglist_data
713 *
714 * Returns the lru list vector holding pages for a given @memcg &
715 * @pgdat combination. This can be the node lruvec, if the memory
716 * controller is disabled.
717 */
718 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
719 struct pglist_data *pgdat)
720 {
721 struct mem_cgroup_per_node *mz;
722 struct lruvec *lruvec;
723
724 if (mem_cgroup_disabled()) {
725 lruvec = &pgdat->__lruvec;
726 goto out;
727 }
728
729 if (!memcg)
730 memcg = root_mem_cgroup;
731
732 mz = memcg->nodeinfo[pgdat->node_id];
733 lruvec = &mz->lruvec;
734 out:
735 /*
736 * Since a node can be onlined after the mem_cgroup was created,
737 * we have to be prepared to initialize lruvec->pgdat here;
738 * and if offlined then reonlined, we need to reinitialize it.
739 */
740 if (unlikely(lruvec->pgdat != pgdat))
741 lruvec->pgdat = pgdat;
742 return lruvec;
743 }
744
745 /**
746 * folio_lruvec - return lruvec for isolating/putting an LRU folio
747 * @folio: Pointer to the folio.
748 *
749 * This function relies on folio->mem_cgroup being stable.
750 */
751 static inline struct lruvec *folio_lruvec(struct folio *folio)
752 {
753 struct mem_cgroup *memcg = folio_memcg(folio);
754
755 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio);
756 return mem_cgroup_lruvec(memcg, folio_pgdat(folio));
757 }
758
759 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
760
761 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
762
763 struct mem_cgroup *get_mem_cgroup_from_current(void);
764
765 struct lruvec *folio_lruvec_lock(struct folio *folio);
766 struct lruvec *folio_lruvec_lock_irq(struct folio *folio);
767 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
768 unsigned long *flags);
769
770 #ifdef CONFIG_DEBUG_VM
771 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio);
772 #else
773 static inline
774 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
775 {
776 }
777 #endif
778
779 static inline
780 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
781 return css ? container_of(css, struct mem_cgroup, css) : NULL;
782 }
783
784 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
785 {
786 return percpu_ref_tryget(&objcg->refcnt);
787 }
788
789 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
790 {
791 percpu_ref_get(&objcg->refcnt);
792 }
793
794 static inline void obj_cgroup_get_many(struct obj_cgroup *objcg,
795 unsigned long nr)
796 {
797 percpu_ref_get_many(&objcg->refcnt, nr);
798 }
799
800 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
801 {
802 if (objcg)
803 percpu_ref_put(&objcg->refcnt);
804 }
805
806 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg)
807 {
808 return !memcg || css_tryget(&memcg->css);
809 }
810
811 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg)
812 {
813 return !memcg || css_tryget_online(&memcg->css);
814 }
815
816 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
817 {
818 if (memcg)
819 css_put(&memcg->css);
820 }
821
822 #define mem_cgroup_from_counter(counter, member) \
823 container_of(counter, struct mem_cgroup, member)
824
825 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
826 struct mem_cgroup *,
827 struct mem_cgroup_reclaim_cookie *);
828 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
829 void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
830 int (*)(struct task_struct *, void *), void *arg);
831
832 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
833 {
834 if (mem_cgroup_disabled())
835 return 0;
836
837 return memcg->id.id;
838 }
839 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
840
841 #ifdef CONFIG_SHRINKER_DEBUG
842 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
843 {
844 return memcg ? cgroup_ino(memcg->css.cgroup) : 0;
845 }
846
847 struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino);
848 #endif
849
850 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
851 {
852 return mem_cgroup_from_css(seq_css(m));
853 }
854
855 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
856 {
857 struct mem_cgroup_per_node *mz;
858
859 if (mem_cgroup_disabled())
860 return NULL;
861
862 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
863 return mz->memcg;
864 }
865
866 /**
867 * parent_mem_cgroup - find the accounting parent of a memcg
868 * @memcg: memcg whose parent to find
869 *
870 * Returns the parent memcg, or NULL if this is the root.
871 */
872 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
873 {
874 return mem_cgroup_from_css(memcg->css.parent);
875 }
876
877 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
878 struct mem_cgroup *root)
879 {
880 if (root == memcg)
881 return true;
882 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
883 }
884
885 static inline bool mm_match_cgroup(struct mm_struct *mm,
886 struct mem_cgroup *memcg)
887 {
888 struct mem_cgroup *task_memcg;
889 bool match = false;
890
891 rcu_read_lock();
892 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
893 if (task_memcg)
894 match = mem_cgroup_is_descendant(task_memcg, memcg);
895 rcu_read_unlock();
896 return match;
897 }
898
899 struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio);
900 ino_t page_cgroup_ino(struct page *page);
901
902 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
903 {
904 if (mem_cgroup_disabled())
905 return true;
906 return !!(memcg->css.flags & CSS_ONLINE);
907 }
908
909 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
910 int zid, int nr_pages);
911
912 static inline
913 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
914 enum lru_list lru, int zone_idx)
915 {
916 struct mem_cgroup_per_node *mz;
917
918 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
919 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
920 }
921
922 void mem_cgroup_handle_over_high(gfp_t gfp_mask);
923
924 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
925
926 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
927
928 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
929 struct task_struct *p);
930
931 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
932
933 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
934 struct mem_cgroup *oom_domain);
935 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
936
937 void __mod_memcg_state(struct mem_cgroup *memcg, enum memcg_stat_item idx,
938 int val);
939
940 /* idx can be of type enum memcg_stat_item or node_stat_item */
941 static inline void mod_memcg_state(struct mem_cgroup *memcg,
942 enum memcg_stat_item idx, int val)
943 {
944 unsigned long flags;
945
946 local_irq_save(flags);
947 __mod_memcg_state(memcg, idx, val);
948 local_irq_restore(flags);
949 }
950
951 static inline void mod_memcg_page_state(struct page *page,
952 enum memcg_stat_item idx, int val)
953 {
954 struct mem_cgroup *memcg;
955
956 if (mem_cgroup_disabled())
957 return;
958
959 rcu_read_lock();
960 memcg = folio_memcg(page_folio(page));
961 if (memcg)
962 mod_memcg_state(memcg, idx, val);
963 rcu_read_unlock();
964 }
965
966 unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx);
967 unsigned long lruvec_page_state(struct lruvec *lruvec, enum node_stat_item idx);
968 unsigned long lruvec_page_state_local(struct lruvec *lruvec,
969 enum node_stat_item idx);
970
971 void mem_cgroup_flush_stats(struct mem_cgroup *memcg);
972 void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg);
973
974 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
975
976 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
977 int val)
978 {
979 unsigned long flags;
980
981 local_irq_save(flags);
982 __mod_lruvec_kmem_state(p, idx, val);
983 local_irq_restore(flags);
984 }
985
986 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
987 unsigned long count);
988
989 static inline void count_memcg_events(struct mem_cgroup *memcg,
990 enum vm_event_item idx,
991 unsigned long count)
992 {
993 unsigned long flags;
994
995 local_irq_save(flags);
996 __count_memcg_events(memcg, idx, count);
997 local_irq_restore(flags);
998 }
999
1000 static inline void count_memcg_folio_events(struct folio *folio,
1001 enum vm_event_item idx, unsigned long nr)
1002 {
1003 struct mem_cgroup *memcg = folio_memcg(folio);
1004
1005 if (memcg)
1006 count_memcg_events(memcg, idx, nr);
1007 }
1008
1009 static inline void count_memcg_event_mm(struct mm_struct *mm,
1010 enum vm_event_item idx)
1011 {
1012 struct mem_cgroup *memcg;
1013
1014 if (mem_cgroup_disabled())
1015 return;
1016
1017 rcu_read_lock();
1018 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1019 if (likely(memcg))
1020 count_memcg_events(memcg, idx, 1);
1021 rcu_read_unlock();
1022 }
1023
1024 static inline void memcg_memory_event(struct mem_cgroup *memcg,
1025 enum memcg_memory_event event)
1026 {
1027 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
1028 event == MEMCG_SWAP_FAIL;
1029
1030 atomic_long_inc(&memcg->memory_events_local[event]);
1031 if (!swap_event)
1032 cgroup_file_notify(&memcg->events_local_file);
1033
1034 do {
1035 atomic_long_inc(&memcg->memory_events[event]);
1036 if (swap_event)
1037 cgroup_file_notify(&memcg->swap_events_file);
1038 else
1039 cgroup_file_notify(&memcg->events_file);
1040
1041 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1042 break;
1043 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1044 break;
1045 } while ((memcg = parent_mem_cgroup(memcg)) &&
1046 !mem_cgroup_is_root(memcg));
1047 }
1048
1049 static inline void memcg_memory_event_mm(struct mm_struct *mm,
1050 enum memcg_memory_event event)
1051 {
1052 struct mem_cgroup *memcg;
1053
1054 if (mem_cgroup_disabled())
1055 return;
1056
1057 rcu_read_lock();
1058 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1059 if (likely(memcg))
1060 memcg_memory_event(memcg, event);
1061 rcu_read_unlock();
1062 }
1063
1064 void split_page_memcg(struct page *head, int old_order, int new_order);
1065
1066 #else /* CONFIG_MEMCG */
1067
1068 #define MEM_CGROUP_ID_SHIFT 0
1069
1070 static inline struct mem_cgroup *folio_memcg(struct folio *folio)
1071 {
1072 return NULL;
1073 }
1074
1075 static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
1076 {
1077 WARN_ON_ONCE(!rcu_read_lock_held());
1078 return NULL;
1079 }
1080
1081 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio)
1082 {
1083 return NULL;
1084 }
1085
1086 static inline struct mem_cgroup *page_memcg_check(struct page *page)
1087 {
1088 return NULL;
1089 }
1090
1091 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg)
1092 {
1093 return NULL;
1094 }
1095
1096 static inline bool folio_memcg_kmem(struct folio *folio)
1097 {
1098 return false;
1099 }
1100
1101 static inline bool PageMemcgKmem(struct page *page)
1102 {
1103 return false;
1104 }
1105
1106 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1107 {
1108 return true;
1109 }
1110
1111 static inline bool mem_cgroup_disabled(void)
1112 {
1113 return true;
1114 }
1115
1116 static inline void memcg_memory_event(struct mem_cgroup *memcg,
1117 enum memcg_memory_event event)
1118 {
1119 }
1120
1121 static inline void memcg_memory_event_mm(struct mm_struct *mm,
1122 enum memcg_memory_event event)
1123 {
1124 }
1125
1126 static inline void mem_cgroup_protection(struct mem_cgroup *root,
1127 struct mem_cgroup *memcg,
1128 unsigned long *min,
1129 unsigned long *low)
1130 {
1131 *min = *low = 0;
1132 }
1133
1134 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1135 struct mem_cgroup *memcg)
1136 {
1137 }
1138
1139 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target,
1140 struct mem_cgroup *memcg)
1141 {
1142 return true;
1143 }
1144 static inline bool mem_cgroup_below_low(struct mem_cgroup *target,
1145 struct mem_cgroup *memcg)
1146 {
1147 return false;
1148 }
1149
1150 static inline bool mem_cgroup_below_min(struct mem_cgroup *target,
1151 struct mem_cgroup *memcg)
1152 {
1153 return false;
1154 }
1155
1156 static inline void mem_cgroup_commit_charge(struct folio *folio,
1157 struct mem_cgroup *memcg)
1158 {
1159 }
1160
1161 static inline int mem_cgroup_charge(struct folio *folio,
1162 struct mm_struct *mm, gfp_t gfp)
1163 {
1164 return 0;
1165 }
1166
1167 static inline int mem_cgroup_hugetlb_try_charge(struct mem_cgroup *memcg,
1168 gfp_t gfp, long nr_pages)
1169 {
1170 return 0;
1171 }
1172
1173 static inline int mem_cgroup_swapin_charge_folio(struct folio *folio,
1174 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
1175 {
1176 return 0;
1177 }
1178
1179 static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry)
1180 {
1181 }
1182
1183 static inline void mem_cgroup_uncharge(struct folio *folio)
1184 {
1185 }
1186
1187 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios)
1188 {
1189 }
1190
1191 static inline void mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
1192 unsigned int nr_pages)
1193 {
1194 }
1195
1196 static inline void mem_cgroup_replace_folio(struct folio *old,
1197 struct folio *new)
1198 {
1199 }
1200
1201 static inline void mem_cgroup_migrate(struct folio *old, struct folio *new)
1202 {
1203 }
1204
1205 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1206 struct pglist_data *pgdat)
1207 {
1208 return &pgdat->__lruvec;
1209 }
1210
1211 static inline struct lruvec *folio_lruvec(struct folio *folio)
1212 {
1213 struct pglist_data *pgdat = folio_pgdat(folio);
1214 return &pgdat->__lruvec;
1215 }
1216
1217 static inline
1218 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
1219 {
1220 }
1221
1222 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1223 {
1224 return NULL;
1225 }
1226
1227 static inline bool mm_match_cgroup(struct mm_struct *mm,
1228 struct mem_cgroup *memcg)
1229 {
1230 return true;
1231 }
1232
1233 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1234 {
1235 return NULL;
1236 }
1237
1238 static inline struct mem_cgroup *get_mem_cgroup_from_current(void)
1239 {
1240 return NULL;
1241 }
1242
1243 static inline
1244 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css)
1245 {
1246 return NULL;
1247 }
1248
1249 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
1250 {
1251 }
1252
1253 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg)
1254 {
1255 return true;
1256 }
1257
1258 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg)
1259 {
1260 return true;
1261 }
1262
1263 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1264 {
1265 }
1266
1267 static inline struct lruvec *folio_lruvec_lock(struct folio *folio)
1268 {
1269 struct pglist_data *pgdat = folio_pgdat(folio);
1270
1271 spin_lock(&pgdat->__lruvec.lru_lock);
1272 return &pgdat->__lruvec;
1273 }
1274
1275 static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio)
1276 {
1277 struct pglist_data *pgdat = folio_pgdat(folio);
1278
1279 spin_lock_irq(&pgdat->__lruvec.lru_lock);
1280 return &pgdat->__lruvec;
1281 }
1282
1283 static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
1284 unsigned long *flagsp)
1285 {
1286 struct pglist_data *pgdat = folio_pgdat(folio);
1287
1288 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1289 return &pgdat->__lruvec;
1290 }
1291
1292 static inline struct mem_cgroup *
1293 mem_cgroup_iter(struct mem_cgroup *root,
1294 struct mem_cgroup *prev,
1295 struct mem_cgroup_reclaim_cookie *reclaim)
1296 {
1297 return NULL;
1298 }
1299
1300 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1301 struct mem_cgroup *prev)
1302 {
1303 }
1304
1305 static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1306 int (*fn)(struct task_struct *, void *), void *arg)
1307 {
1308 }
1309
1310 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1311 {
1312 return 0;
1313 }
1314
1315 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1316 {
1317 WARN_ON_ONCE(id);
1318 /* XXX: This should always return root_mem_cgroup */
1319 return NULL;
1320 }
1321
1322 #ifdef CONFIG_SHRINKER_DEBUG
1323 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
1324 {
1325 return 0;
1326 }
1327
1328 static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino)
1329 {
1330 return NULL;
1331 }
1332 #endif
1333
1334 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1335 {
1336 return NULL;
1337 }
1338
1339 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1340 {
1341 return NULL;
1342 }
1343
1344 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1345 {
1346 return true;
1347 }
1348
1349 static inline
1350 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1351 enum lru_list lru, int zone_idx)
1352 {
1353 return 0;
1354 }
1355
1356 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1357 {
1358 return 0;
1359 }
1360
1361 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1362 {
1363 return 0;
1364 }
1365
1366 static inline void
1367 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1368 {
1369 }
1370
1371 static inline void
1372 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1373 {
1374 }
1375
1376 static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask)
1377 {
1378 }
1379
1380 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1381 struct task_struct *victim, struct mem_cgroup *oom_domain)
1382 {
1383 return NULL;
1384 }
1385
1386 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1387 {
1388 }
1389
1390 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1391 enum memcg_stat_item idx,
1392 int nr)
1393 {
1394 }
1395
1396 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1397 enum memcg_stat_item idx,
1398 int nr)
1399 {
1400 }
1401
1402 static inline void mod_memcg_page_state(struct page *page,
1403 enum memcg_stat_item idx, int val)
1404 {
1405 }
1406
1407 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1408 {
1409 return 0;
1410 }
1411
1412 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1413 enum node_stat_item idx)
1414 {
1415 return node_page_state(lruvec_pgdat(lruvec), idx);
1416 }
1417
1418 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1419 enum node_stat_item idx)
1420 {
1421 return node_page_state(lruvec_pgdat(lruvec), idx);
1422 }
1423
1424 static inline void mem_cgroup_flush_stats(struct mem_cgroup *memcg)
1425 {
1426 }
1427
1428 static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg)
1429 {
1430 }
1431
1432 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1433 int val)
1434 {
1435 struct page *page = virt_to_head_page(p);
1436
1437 __mod_node_page_state(page_pgdat(page), idx, val);
1438 }
1439
1440 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1441 int val)
1442 {
1443 struct page *page = virt_to_head_page(p);
1444
1445 mod_node_page_state(page_pgdat(page), idx, val);
1446 }
1447
1448 static inline void count_memcg_events(struct mem_cgroup *memcg,
1449 enum vm_event_item idx,
1450 unsigned long count)
1451 {
1452 }
1453
1454 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1455 enum vm_event_item idx,
1456 unsigned long count)
1457 {
1458 }
1459
1460 static inline void count_memcg_folio_events(struct folio *folio,
1461 enum vm_event_item idx, unsigned long nr)
1462 {
1463 }
1464
1465 static inline
1466 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1467 {
1468 }
1469
1470 static inline void split_page_memcg(struct page *head, int old_order, int new_order)
1471 {
1472 }
1473 #endif /* CONFIG_MEMCG */
1474
1475 /*
1476 * Extended information for slab objects stored as an array in page->memcg_data
1477 * if MEMCG_DATA_OBJEXTS is set.
1478 */
1479 struct slabobj_ext {
1480 #ifdef CONFIG_MEMCG
1481 struct obj_cgroup *objcg;
1482 #endif
1483 #ifdef CONFIG_MEM_ALLOC_PROFILING
1484 union codetag_ref ref;
1485 #endif
1486 } __aligned(8);
1487
1488 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1489 {
1490 __mod_lruvec_kmem_state(p, idx, 1);
1491 }
1492
1493 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1494 {
1495 __mod_lruvec_kmem_state(p, idx, -1);
1496 }
1497
1498 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1499 {
1500 struct mem_cgroup *memcg;
1501
1502 memcg = lruvec_memcg(lruvec);
1503 if (!memcg)
1504 return NULL;
1505 memcg = parent_mem_cgroup(memcg);
1506 if (!memcg)
1507 return NULL;
1508 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1509 }
1510
1511 static inline void unlock_page_lruvec(struct lruvec *lruvec)
1512 {
1513 spin_unlock(&lruvec->lru_lock);
1514 }
1515
1516 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1517 {
1518 spin_unlock_irq(&lruvec->lru_lock);
1519 }
1520
1521 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1522 unsigned long flags)
1523 {
1524 spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1525 }
1526
1527 /* Test requires a stable folio->memcg binding, see folio_memcg() */
1528 static inline bool folio_matches_lruvec(struct folio *folio,
1529 struct lruvec *lruvec)
1530 {
1531 return lruvec_pgdat(lruvec) == folio_pgdat(folio) &&
1532 lruvec_memcg(lruvec) == folio_memcg(folio);
1533 }
1534
1535 /* Don't lock again iff page's lruvec locked */
1536 static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio,
1537 struct lruvec *locked_lruvec)
1538 {
1539 if (locked_lruvec) {
1540 if (folio_matches_lruvec(folio, locked_lruvec))
1541 return locked_lruvec;
1542
1543 unlock_page_lruvec_irq(locked_lruvec);
1544 }
1545
1546 return folio_lruvec_lock_irq(folio);
1547 }
1548
1549 /* Don't lock again iff folio's lruvec locked */
1550 static inline void folio_lruvec_relock_irqsave(struct folio *folio,
1551 struct lruvec **lruvecp, unsigned long *flags)
1552 {
1553 if (*lruvecp) {
1554 if (folio_matches_lruvec(folio, *lruvecp))
1555 return;
1556
1557 unlock_page_lruvec_irqrestore(*lruvecp, *flags);
1558 }
1559
1560 *lruvecp = folio_lruvec_lock_irqsave(folio, flags);
1561 }
1562
1563 #ifdef CONFIG_CGROUP_WRITEBACK
1564
1565 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1566 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1567 unsigned long *pheadroom, unsigned long *pdirty,
1568 unsigned long *pwriteback);
1569
1570 void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio,
1571 struct bdi_writeback *wb);
1572
1573 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1574 struct bdi_writeback *wb)
1575 {
1576 struct mem_cgroup *memcg;
1577
1578 if (mem_cgroup_disabled())
1579 return;
1580
1581 memcg = folio_memcg(folio);
1582 if (unlikely(memcg && &memcg->css != wb->memcg_css))
1583 mem_cgroup_track_foreign_dirty_slowpath(folio, wb);
1584 }
1585
1586 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1587
1588 #else /* CONFIG_CGROUP_WRITEBACK */
1589
1590 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1591 {
1592 return NULL;
1593 }
1594
1595 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1596 unsigned long *pfilepages,
1597 unsigned long *pheadroom,
1598 unsigned long *pdirty,
1599 unsigned long *pwriteback)
1600 {
1601 }
1602
1603 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1604 struct bdi_writeback *wb)
1605 {
1606 }
1607
1608 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1609 {
1610 }
1611
1612 #endif /* CONFIG_CGROUP_WRITEBACK */
1613
1614 struct sock;
1615 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages,
1616 gfp_t gfp_mask);
1617 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1618 #ifdef CONFIG_MEMCG
1619 extern struct static_key_false memcg_sockets_enabled_key;
1620 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1621 void mem_cgroup_sk_alloc(struct sock *sk);
1622 void mem_cgroup_sk_free(struct sock *sk);
1623 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1624 {
1625 #ifdef CONFIG_MEMCG_V1
1626 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1627 return !!memcg->tcpmem_pressure;
1628 #endif /* CONFIG_MEMCG_V1 */
1629 do {
1630 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure)))
1631 return true;
1632 } while ((memcg = parent_mem_cgroup(memcg)));
1633 return false;
1634 }
1635
1636 int alloc_shrinker_info(struct mem_cgroup *memcg);
1637 void free_shrinker_info(struct mem_cgroup *memcg);
1638 void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id);
1639 void reparent_shrinker_deferred(struct mem_cgroup *memcg);
1640 #else
1641 #define mem_cgroup_sockets_enabled 0
1642 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1643 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1644 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1645 {
1646 return false;
1647 }
1648
1649 static inline void set_shrinker_bit(struct mem_cgroup *memcg,
1650 int nid, int shrinker_id)
1651 {
1652 }
1653 #endif
1654
1655 #ifdef CONFIG_MEMCG
1656 bool mem_cgroup_kmem_disabled(void);
1657 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1658 void __memcg_kmem_uncharge_page(struct page *page, int order);
1659
1660 /*
1661 * The returned objcg pointer is safe to use without additional
1662 * protection within a scope. The scope is defined either by
1663 * the current task (similar to the "current" global variable)
1664 * or by set_active_memcg() pair.
1665 * Please, use obj_cgroup_get() to get a reference if the pointer
1666 * needs to be used outside of the local scope.
1667 */
1668 struct obj_cgroup *current_obj_cgroup(void);
1669 struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio);
1670
1671 static inline struct obj_cgroup *get_obj_cgroup_from_current(void)
1672 {
1673 struct obj_cgroup *objcg = current_obj_cgroup();
1674
1675 if (objcg)
1676 obj_cgroup_get(objcg);
1677
1678 return objcg;
1679 }
1680
1681 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1682 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1683
1684 extern struct static_key_false memcg_bpf_enabled_key;
1685 static inline bool memcg_bpf_enabled(void)
1686 {
1687 return static_branch_likely(&memcg_bpf_enabled_key);
1688 }
1689
1690 extern struct static_key_false memcg_kmem_online_key;
1691
1692 static inline bool memcg_kmem_online(void)
1693 {
1694 return static_branch_likely(&memcg_kmem_online_key);
1695 }
1696
1697 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1698 int order)
1699 {
1700 if (memcg_kmem_online())
1701 return __memcg_kmem_charge_page(page, gfp, order);
1702 return 0;
1703 }
1704
1705 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1706 {
1707 if (memcg_kmem_online())
1708 __memcg_kmem_uncharge_page(page, order);
1709 }
1710
1711 /*
1712 * A helper for accessing memcg's kmem_id, used for getting
1713 * corresponding LRU lists.
1714 */
1715 static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1716 {
1717 return memcg ? memcg->kmemcg_id : -1;
1718 }
1719
1720 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1721 struct mem_cgroup *mem_cgroup_from_slab_obj(void *p);
1722
1723 static inline void count_objcg_event(struct obj_cgroup *objcg,
1724 enum vm_event_item idx)
1725 {
1726 struct mem_cgroup *memcg;
1727
1728 if (!memcg_kmem_online())
1729 return;
1730
1731 rcu_read_lock();
1732 memcg = obj_cgroup_memcg(objcg);
1733 count_memcg_events(memcg, idx, 1);
1734 rcu_read_unlock();
1735 }
1736
1737 #else
1738 static inline bool mem_cgroup_kmem_disabled(void)
1739 {
1740 return true;
1741 }
1742
1743 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1744 int order)
1745 {
1746 return 0;
1747 }
1748
1749 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1750 {
1751 }
1752
1753 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1754 int order)
1755 {
1756 return 0;
1757 }
1758
1759 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1760 {
1761 }
1762
1763 static inline struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio)
1764 {
1765 return NULL;
1766 }
1767
1768 static inline bool memcg_bpf_enabled(void)
1769 {
1770 return false;
1771 }
1772
1773 static inline bool memcg_kmem_online(void)
1774 {
1775 return false;
1776 }
1777
1778 static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1779 {
1780 return -1;
1781 }
1782
1783 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1784 {
1785 return NULL;
1786 }
1787
1788 static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p)
1789 {
1790 return NULL;
1791 }
1792
1793 static inline void count_objcg_event(struct obj_cgroup *objcg,
1794 enum vm_event_item idx)
1795 {
1796 }
1797
1798 #endif /* CONFIG_MEMCG */
1799
1800 #if defined(CONFIG_MEMCG) && defined(CONFIG_ZSWAP)
1801 bool obj_cgroup_may_zswap(struct obj_cgroup *objcg);
1802 void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size);
1803 void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size);
1804 bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg);
1805 #else
1806 static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg)
1807 {
1808 return true;
1809 }
1810 static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg,
1811 size_t size)
1812 {
1813 }
1814 static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg,
1815 size_t size)
1816 {
1817 }
1818 static inline bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg)
1819 {
1820 /* if zswap is disabled, do not block pages going to the swapping device */
1821 return true;
1822 }
1823 #endif
1824
1825
1826 /* Cgroup v1-related declarations */
1827
1828 #ifdef CONFIG_MEMCG_V1
1829 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order,
1830 gfp_t gfp_mask,
1831 unsigned long *total_scanned);
1832
1833 bool mem_cgroup_oom_synchronize(bool wait);
1834
1835 static inline bool task_in_memcg_oom(struct task_struct *p)
1836 {
1837 return p->memcg_in_oom;
1838 }
1839
1840 void folio_memcg_lock(struct folio *folio);
1841 void folio_memcg_unlock(struct folio *folio);
1842
1843 /* try to stablize folio_memcg() for all the pages in a memcg */
1844 static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
1845 {
1846 rcu_read_lock();
1847
1848 if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account))
1849 return true;
1850
1851 rcu_read_unlock();
1852 return false;
1853 }
1854
1855 static inline void mem_cgroup_unlock_pages(void)
1856 {
1857 rcu_read_unlock();
1858 }
1859
1860 static inline void mem_cgroup_enter_user_fault(void)
1861 {
1862 WARN_ON(current->in_user_fault);
1863 current->in_user_fault = 1;
1864 }
1865
1866 static inline void mem_cgroup_exit_user_fault(void)
1867 {
1868 WARN_ON(!current->in_user_fault);
1869 current->in_user_fault = 0;
1870 }
1871
1872 #else /* CONFIG_MEMCG_V1 */
1873 static inline
1874 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order,
1875 gfp_t gfp_mask,
1876 unsigned long *total_scanned)
1877 {
1878 return 0;
1879 }
1880
1881 static inline void folio_memcg_lock(struct folio *folio)
1882 {
1883 }
1884
1885 static inline void folio_memcg_unlock(struct folio *folio)
1886 {
1887 }
1888
1889 static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
1890 {
1891 /* to match folio_memcg_rcu() */
1892 rcu_read_lock();
1893 return true;
1894 }
1895
1896 static inline void mem_cgroup_unlock_pages(void)
1897 {
1898 rcu_read_unlock();
1899 }
1900
1901 static inline bool task_in_memcg_oom(struct task_struct *p)
1902 {
1903 return false;
1904 }
1905
1906 static inline bool mem_cgroup_oom_synchronize(bool wait)
1907 {
1908 return false;
1909 }
1910
1911 static inline void mem_cgroup_enter_user_fault(void)
1912 {
1913 }
1914
1915 static inline void mem_cgroup_exit_user_fault(void)
1916 {
1917 }
1918
1919 #endif /* CONFIG_MEMCG_V1 */
1920
1921 #endif /* _LINUX_MEMCONTROL_H */
1922
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.