projects/linux·blob v6.1 · refs

1867 lines · 45.0 KiB

1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/* memcontrol.h - Memory Controller
3 *
4 * Copyright IBM Corporation, 2007
5 * Author Balbir Singh <[email protected]>
6 *
7 * Copyright 2007 OpenVZ SWsoft Inc
8 * Author: Pavel Emelianov <[email protected]>
9 */
10
11#ifndef _LINUX_MEMCONTROL_H
12#define _LINUX_MEMCONTROL_H
13#include <linux/cgroup.h>
14#include <linux/vm_event_item.h>
15#include <linux/hardirq.h>
16#include <linux/jump_label.h>
17#include <linux/page_counter.h>
18#include <linux/vmpressure.h>
19#include <linux/eventfd.h>
20#include <linux/mm.h>
21#include <linux/vmstat.h>
22#include <linux/writeback.h>
23#include <linux/page-flags.h>
24
25struct mem_cgroup;
26struct obj_cgroup;
27struct page;
28struct mm_struct;
29struct kmem_cache;
30
31/* Cgroup-specific page state, on top of universal node page state */
32enum memcg_stat_item {
33 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS,
34 MEMCG_SOCK,
35 MEMCG_PERCPU_B,
36 MEMCG_VMALLOC,
37 MEMCG_KMEM,
38 MEMCG_ZSWAP_B,
39 MEMCG_ZSWAPPED,
40 MEMCG_NR_STAT,
41};
42
43enum memcg_memory_event {
44 MEMCG_LOW,
45 MEMCG_HIGH,
46 MEMCG_MAX,
47 MEMCG_OOM,
48 MEMCG_OOM_KILL,
49 MEMCG_OOM_GROUP_KILL,
50 MEMCG_SWAP_HIGH,
51 MEMCG_SWAP_MAX,
52 MEMCG_SWAP_FAIL,
53 MEMCG_NR_MEMORY_EVENTS,
54};
55
56struct mem_cgroup_reclaim_cookie {
57 pg_data_t *pgdat;
58 unsigned int generation;
59};
60
61#ifdef CONFIG_MEMCG
62
63#define MEM_CGROUP_ID_SHIFT 16
64#define MEM_CGROUP_ID_MAX USHRT_MAX
65
66struct mem_cgroup_id {
67 int id;
68 refcount_t ref;
69};
70
71/*
72 * Per memcg event counter is incremented at every pagein/pageout. With THP,
73 * it will be incremented by the number of pages. This counter is used
74 * to trigger some periodic events. This is straightforward and better
75 * than using jiffies etc. to handle periodic memcg event.
76 */
77enum mem_cgroup_events_target {
78 MEM_CGROUP_TARGET_THRESH,
79 MEM_CGROUP_TARGET_SOFTLIMIT,
80 MEM_CGROUP_NTARGETS,
81};
82
83struct memcg_vmstats_percpu;
84struct memcg_vmstats;
85
86struct mem_cgroup_reclaim_iter {
87 struct mem_cgroup *position;
88 /* scan generation, increased every round-trip */
89 unsigned int generation;
90};
91
92/*
93 * Bitmap and deferred work of shrinker::id corresponding to memcg-aware
94 * shrinkers, which have elements charged to this memcg.
95 */
96struct shrinker_info {
97 struct rcu_head rcu;
98 atomic_long_t *nr_deferred;
99 unsigned long *map;
100};
101
102struct lruvec_stats_percpu {
103 /* Local (CPU and cgroup) state */
104 long state[NR_VM_NODE_STAT_ITEMS];
105
106 /* Delta calculation for lockless upward propagation */
107 long state_prev[NR_VM_NODE_STAT_ITEMS];
108};
109
110struct lruvec_stats {
111 /* Aggregated (CPU and subtree) state */
112 long state[NR_VM_NODE_STAT_ITEMS];
113
114 /* Pending child counts during tree propagation */
115 long state_pending[NR_VM_NODE_STAT_ITEMS];
116};
117
118/*
119 * per-node information in memory controller.
120 */
121struct mem_cgroup_per_node {
122 struct lruvec lruvec;
123
124 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu;
125 struct lruvec_stats lruvec_stats;
126
127 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
128
129 struct mem_cgroup_reclaim_iter iter;
130
131 struct shrinker_info __rcu *shrinker_info;
132
133 struct rb_node tree_node; /* RB tree node */
134 unsigned long usage_in_excess;/* Set to the value by which */
135 /* the soft limit is exceeded*/
136 bool on_tree;
137 struct mem_cgroup *memcg; /* Back pointer, we cannot */
138 /* use container_of */
139};
140
141struct mem_cgroup_threshold {
142 struct eventfd_ctx *eventfd;
143 unsigned long threshold;
144};
145
146/* For threshold */
147struct mem_cgroup_threshold_ary {
148 /* An array index points to threshold just below or equal to usage. */
149 int current_threshold;
150 /* Size of entries[] */
151 unsigned int size;
152 /* Array of thresholds */
153 struct mem_cgroup_threshold entries[];
154};
155
156struct mem_cgroup_thresholds {
157 /* Primary thresholds array */
158 struct mem_cgroup_threshold_ary *primary;
159 /*
160 * Spare threshold array.
161 * This is needed to make mem_cgroup_unregister_event() "never fail".
162 * It must be able to store at least primary->size - 1 entries.
163 */
164 struct mem_cgroup_threshold_ary *spare;
165};
166
167/*
168 * Remember four most recent foreign writebacks with dirty pages in this
169 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
170 * one in a given round, we're likely to catch it later if it keeps
171 * foreign-dirtying, so a fairly low count should be enough.
172 *
173 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
174 */
175#define MEMCG_CGWB_FRN_CNT 4
176
177struct memcg_cgwb_frn {
178 u64 bdi_id; /* bdi->id of the foreign inode */
179 int memcg_id; /* memcg->css.id of foreign inode */
180 u64 at; /* jiffies_64 at the time of dirtying */
181 struct wb_completion done; /* tracks in-flight foreign writebacks */
182};
183
184/*
185 * Bucket for arbitrarily byte-sized objects charged to a memory
186 * cgroup. The bucket can be reparented in one piece when the cgroup
187 * is destroyed, without having to round up the individual references
188 * of all live memory objects in the wild.
189 */
190struct obj_cgroup {
191 struct percpu_ref refcnt;
192 struct mem_cgroup *memcg;
193 atomic_t nr_charged_bytes;
194 union {
195 struct list_head list; /* protected by objcg_lock */
196 struct rcu_head rcu;
197 };
198};
199
200/*
201 * The memory controller data structure. The memory controller controls both
202 * page cache and RSS per cgroup. We would eventually like to provide
203 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
204 * to help the administrator determine what knobs to tune.
205 */
206struct mem_cgroup {
207 struct cgroup_subsys_state css;
208
209 /* Private memcg ID. Used to ID objects that outlive the cgroup */
210 struct mem_cgroup_id id;
211
212 /* Accounted resources */
213 struct page_counter memory; /* Both v1 & v2 */
214
215 union {
216 struct page_counter swap; /* v2 only */
217 struct page_counter memsw; /* v1 only */
218 };
219
220 /* Legacy consumer-oriented counters */
221 struct page_counter kmem; /* v1 only */
222 struct page_counter tcpmem; /* v1 only */
223
224 /* Range enforcement for interrupt charges */
225 struct work_struct high_work;
226
227#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP)
228 unsigned long zswap_max;
229#endif
230
231 unsigned long soft_limit;
232
233 /* vmpressure notifications */
234 struct vmpressure vmpressure;
235
236 /*
237 * Should the OOM killer kill all belonging tasks, had it kill one?
238 */
239 bool oom_group;
240
241 /* protected by memcg_oom_lock */
242 bool oom_lock;
243 int under_oom;
244
245 int swappiness;
246 /* OOM-Killer disable */
247 int oom_kill_disable;
248
249 /* memory.events and memory.events.local */
250 struct cgroup_file events_file;
251 struct cgroup_file events_local_file;
252
253 /* handle for "memory.swap.events" */
254 struct cgroup_file swap_events_file;
255
256 /* protect arrays of thresholds */
257 struct mutex thresholds_lock;
258
259 /* thresholds for memory usage. RCU-protected */
260 struct mem_cgroup_thresholds thresholds;
261
262 /* thresholds for mem+swap usage. RCU-protected */
263 struct mem_cgroup_thresholds memsw_thresholds;
264
265 /* For oom notifier event fd */
266 struct list_head oom_notify;
267
268 /*
269 * Should we move charges of a task when a task is moved into this
270 * mem_cgroup ? And what type of charges should we move ?
271 */
272 unsigned long move_charge_at_immigrate;
273 /* taken only while moving_account > 0 */
274 spinlock_t move_lock;
275 unsigned long move_lock_flags;
276
277 CACHELINE_PADDING(_pad1_);
278
279 /* memory.stat */
280 struct memcg_vmstats *vmstats;
281
282 /* memory.events */
283 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
284 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
285
286 unsigned long socket_pressure;
287
288 /* Legacy tcp memory accounting */
289 bool tcpmem_active;
290 int tcpmem_pressure;
291
292#ifdef CONFIG_MEMCG_KMEM
293 int kmemcg_id;
294 struct obj_cgroup __rcu *objcg;
295 /* list of inherited objcgs, protected by objcg_lock */
296 struct list_head objcg_list;
297#endif
298
299 CACHELINE_PADDING(_pad2_);
300
301 /*
302 * set > 0 if pages under this cgroup are moving to other cgroup.
303 */
304 atomic_t moving_account;
305 struct task_struct *move_lock_task;
306
307 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
308
309#ifdef CONFIG_CGROUP_WRITEBACK
310 struct list_head cgwb_list;
311 struct wb_domain cgwb_domain;
312 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
313#endif
314
315 /* List of events which userspace want to receive */
316 struct list_head event_list;
317 spinlock_t event_list_lock;
318
319#ifdef CONFIG_TRANSPARENT_HUGEPAGE
320 struct deferred_split deferred_split_queue;
321#endif
322
323#ifdef CONFIG_LRU_GEN
324 /* per-memcg mm_struct list */
325 struct lru_gen_mm_list mm_list;
326#endif
327
328 struct mem_cgroup_per_node *nodeinfo[];
329};
330
331/*
332 * size of first charge trial.
333 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the
334 * workload.
335 */
336#define MEMCG_CHARGE_BATCH 64U
337
338extern struct mem_cgroup *root_mem_cgroup;
339
340enum page_memcg_data_flags {
341 /* page->memcg_data is a pointer to an objcgs vector */
342 MEMCG_DATA_OBJCGS = (1UL << 0),
343 /* page has been accounted as a non-slab kernel page */
344 MEMCG_DATA_KMEM = (1UL << 1),
345 /* the next bit after the last actual flag */
346 __NR_MEMCG_DATA_FLAGS = (1UL << 2),
347};
348
349#define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1)
350
351static inline bool folio_memcg_kmem(struct folio *folio);
352
353/*
354 * After the initialization objcg->memcg is always pointing at
355 * a valid memcg, but can be atomically swapped to the parent memcg.
356 *
357 * The caller must ensure that the returned memcg won't be released:
358 * e.g. acquire the rcu_read_lock or css_set_lock.
359 */
360static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
361{
362 return READ_ONCE(objcg->memcg);
363}
364
365/*
366 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio
367 * @folio: Pointer to the folio.
368 *
369 * Returns a pointer to the memory cgroup associated with the folio,
370 * or NULL. This function assumes that the folio is known to have a
371 * proper memory cgroup pointer. It's not safe to call this function
372 * against some type of folios, e.g. slab folios or ex-slab folios or
373 * kmem folios.
374 */
375static inline struct mem_cgroup *__folio_memcg(struct folio *folio)
376{
377 unsigned long memcg_data = folio->memcg_data;
378
379 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
380 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio);
381 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio);
382
383 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
384}
385
386/*
387 * __folio_objcg - get the object cgroup associated with a kmem folio.
388 * @folio: Pointer to the folio.
389 *
390 * Returns a pointer to the object cgroup associated with the folio,
391 * or NULL. This function assumes that the folio is known to have a
392 * proper object cgroup pointer. It's not safe to call this function
393 * against some type of folios, e.g. slab folios or ex-slab folios or
394 * LRU folios.
395 */
396static inline struct obj_cgroup *__folio_objcg(struct folio *folio)
397{
398 unsigned long memcg_data = folio->memcg_data;
399
400 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
401 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio);
402 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio);
403
404 return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
405}
406
407/*
408 * folio_memcg - Get the memory cgroup associated with a folio.
409 * @folio: Pointer to the folio.
410 *
411 * Returns a pointer to the memory cgroup associated with the folio,
412 * or NULL. This function assumes that the folio is known to have a
413 * proper memory cgroup pointer. It's not safe to call this function
414 * against some type of folios, e.g. slab folios or ex-slab folios.
415 *
416 * For a non-kmem folio any of the following ensures folio and memcg binding
417 * stability:
418 *
419 * - the folio lock
420 * - LRU isolation
421 * - lock_page_memcg()
422 * - exclusive reference
423 * - mem_cgroup_trylock_pages()
424 *
425 * For a kmem folio a caller should hold an rcu read lock to protect memcg
426 * associated with a kmem folio from being released.
427 */
428static inline struct mem_cgroup *folio_memcg(struct folio *folio)
429{
430 if (folio_memcg_kmem(folio))
431 return obj_cgroup_memcg(__folio_objcg(folio));
432 return __folio_memcg(folio);
433}
434
435static inline struct mem_cgroup *page_memcg(struct page *page)
436{
437 return folio_memcg(page_folio(page));
438}
439
440/**
441 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio.
442 * @folio: Pointer to the folio.
443 *
444 * This function assumes that the folio is known to have a
445 * proper memory cgroup pointer. It's not safe to call this function
446 * against some type of folios, e.g. slab folios or ex-slab folios.
447 *
448 * Return: A pointer to the memory cgroup associated with the folio,
449 * or NULL.
450 */
451static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
452{
453 unsigned long memcg_data = READ_ONCE(folio->memcg_data);
454
455 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
456 WARN_ON_ONCE(!rcu_read_lock_held());
457
458 if (memcg_data & MEMCG_DATA_KMEM) {
459 struct obj_cgroup *objcg;
460
461 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
462 return obj_cgroup_memcg(objcg);
463 }
464
465 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
466}
467
468/*
469 * page_memcg_check - get the memory cgroup associated with a page
470 * @page: a pointer to the page struct
471 *
472 * Returns a pointer to the memory cgroup associated with the page,
473 * or NULL. This function unlike page_memcg() can take any page
474 * as an argument. It has to be used in cases when it's not known if a page
475 * has an associated memory cgroup pointer or an object cgroups vector or
476 * an object cgroup.
477 *
478 * For a non-kmem page any of the following ensures page and memcg binding
479 * stability:
480 *
481 * - the page lock
482 * - LRU isolation
483 * - lock_page_memcg()
484 * - exclusive reference
485 * - mem_cgroup_trylock_pages()
486 *
487 * For a kmem page a caller should hold an rcu read lock to protect memcg
488 * associated with a kmem page from being released.
489 */
490static inline struct mem_cgroup *page_memcg_check(struct page *page)
491{
492 /*
493 * Because page->memcg_data might be changed asynchronously
494 * for slab pages, READ_ONCE() should be used here.
495 */
496 unsigned long memcg_data = READ_ONCE(page->memcg_data);
497
498 if (memcg_data & MEMCG_DATA_OBJCGS)
499 return NULL;
500
501 if (memcg_data & MEMCG_DATA_KMEM) {
502 struct obj_cgroup *objcg;
503
504 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
505 return obj_cgroup_memcg(objcg);
506 }
507
508 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
509}
510
511static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg)
512{
513 struct mem_cgroup *memcg;
514
515 rcu_read_lock();
516retry:
517 memcg = obj_cgroup_memcg(objcg);
518 if (unlikely(!css_tryget(&memcg->css)))
519 goto retry;
520 rcu_read_unlock();
521
522 return memcg;
523}
524
525#ifdef CONFIG_MEMCG_KMEM
526/*
527 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set.
528 * @folio: Pointer to the folio.
529 *
530 * Checks if the folio has MemcgKmem flag set. The caller must ensure
531 * that the folio has an associated memory cgroup. It's not safe to call
532 * this function against some types of folios, e.g. slab folios.
533 */
534static inline bool folio_memcg_kmem(struct folio *folio)
535{
536 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page);
537 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJCGS, folio);
538 return folio->memcg_data & MEMCG_DATA_KMEM;
539}
540
541
542#else
543static inline bool folio_memcg_kmem(struct folio *folio)
544{
545 return false;
546}
547
548#endif
549
550static inline bool PageMemcgKmem(struct page *page)
551{
552 return folio_memcg_kmem(page_folio(page));
553}
554
555static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
556{
557 return (memcg == root_mem_cgroup);
558}
559
560static inline bool mem_cgroup_disabled(void)
561{
562 return !cgroup_subsys_enabled(memory_cgrp_subsys);
563}
564
565static inline void mem_cgroup_protection(struct mem_cgroup *root,
566 struct mem_cgroup *memcg,
567 unsigned long *min,
568 unsigned long *low)
569{
570 *min = *low = 0;
571
572 if (mem_cgroup_disabled())
573 return;
574
575 /*
576 * There is no reclaim protection applied to a targeted reclaim.
577 * We are special casing this specific case here because
578 * mem_cgroup_protected calculation is not robust enough to keep
579 * the protection invariant for calculated effective values for
580 * parallel reclaimers with different reclaim target. This is
581 * especially a problem for tail memcgs (as they have pages on LRU)
582 * which would want to have effective values 0 for targeted reclaim
583 * but a different value for external reclaim.
584 *
585 * Example
586 * Let's have global and A's reclaim in parallel:
587 * |
588 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
589 * |\
590 * | C (low = 1G, usage = 2.5G)
591 * B (low = 1G, usage = 0.5G)
592 *
593 * For the global reclaim
594 * A.elow = A.low
595 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
596 * C.elow = min(C.usage, C.low)
597 *
598 * With the effective values resetting we have A reclaim
599 * A.elow = 0
600 * B.elow = B.low
601 * C.elow = C.low
602 *
603 * If the global reclaim races with A's reclaim then
604 * B.elow = C.elow = 0 because children_low_usage > A.elow)
605 * is possible and reclaiming B would be violating the protection.
606 *
607 */
608 if (root == memcg)
609 return;
610
611 *min = READ_ONCE(memcg->memory.emin);
612 *low = READ_ONCE(memcg->memory.elow);
613}
614
615void mem_cgroup_calculate_protection(struct mem_cgroup *root,
616 struct mem_cgroup *memcg);
617
618static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg)
619{
620 /*
621 * The root memcg doesn't account charges, and doesn't support
622 * protection.
623 */
624 return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg);
625
626}
627
628static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
629{
630 if (!mem_cgroup_supports_protection(memcg))
631 return false;
632
633 return READ_ONCE(memcg->memory.elow) >=
634 page_counter_read(&memcg->memory);
635}
636
637static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
638{
639 if (!mem_cgroup_supports_protection(memcg))
640 return false;
641
642 return READ_ONCE(memcg->memory.emin) >=
643 page_counter_read(&memcg->memory);
644}
645
646int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp);
647
648/**
649 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup.
650 * @folio: Folio to charge.
651 * @mm: mm context of the allocating task.
652 * @gfp: Reclaim mode.
653 *
654 * Try to charge @folio to the memcg that @mm belongs to, reclaiming
655 * pages according to @gfp if necessary. If @mm is NULL, try to
656 * charge to the active memcg.
657 *
658 * Do not use this for folios allocated for swapin.
659 *
660 * Return: 0 on success. Otherwise, an error code is returned.
661 */
662static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm,
663 gfp_t gfp)
664{
665 if (mem_cgroup_disabled())
666 return 0;
667 return __mem_cgroup_charge(folio, mm, gfp);
668}
669
670int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm,
671 gfp_t gfp, swp_entry_t entry);
672void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry);
673
674void __mem_cgroup_uncharge(struct folio *folio);
675
676/**
677 * mem_cgroup_uncharge - Uncharge a folio.
678 * @folio: Folio to uncharge.
679 *
680 * Uncharge a folio previously charged with mem_cgroup_charge().
681 */
682static inline void mem_cgroup_uncharge(struct folio *folio)
683{
684 if (mem_cgroup_disabled())
685 return;
686 __mem_cgroup_uncharge(folio);
687}
688
689void __mem_cgroup_uncharge_list(struct list_head *page_list);
690static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
691{
692 if (mem_cgroup_disabled())
693 return;
694 __mem_cgroup_uncharge_list(page_list);
695}
696
697void mem_cgroup_migrate(struct folio *old, struct folio *new);
698
699/**
700 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
701 * @memcg: memcg of the wanted lruvec
702 * @pgdat: pglist_data
703 *
704 * Returns the lru list vector holding pages for a given @memcg &
705 * @pgdat combination. This can be the node lruvec, if the memory
706 * controller is disabled.
707 */
708static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
709 struct pglist_data *pgdat)
710{
711 struct mem_cgroup_per_node *mz;
712 struct lruvec *lruvec;
713
714 if (mem_cgroup_disabled()) {
715 lruvec = &pgdat->__lruvec;
716 goto out;
717 }
718
719 if (!memcg)
720 memcg = root_mem_cgroup;
721
722 mz = memcg->nodeinfo[pgdat->node_id];
723 lruvec = &mz->lruvec;
724out:
725 /*
726 * Since a node can be onlined after the mem_cgroup was created,
727 * we have to be prepared to initialize lruvec->pgdat here;
728 * and if offlined then reonlined, we need to reinitialize it.
729 */
730 if (unlikely(lruvec->pgdat != pgdat))
731 lruvec->pgdat = pgdat;
732 return lruvec;
733}
734
735/**
736 * folio_lruvec - return lruvec for isolating/putting an LRU folio
737 * @folio: Pointer to the folio.
738 *
739 * This function relies on folio->mem_cgroup being stable.
740 */
741static inline struct lruvec *folio_lruvec(struct folio *folio)
742{
743 struct mem_cgroup *memcg = folio_memcg(folio);
744
745 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio);
746 return mem_cgroup_lruvec(memcg, folio_pgdat(folio));
747}
748
749struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
750
751struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
752
753struct lruvec *folio_lruvec_lock(struct folio *folio);
754struct lruvec *folio_lruvec_lock_irq(struct folio *folio);
755struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
756 unsigned long *flags);
757
758#ifdef CONFIG_DEBUG_VM
759void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio);
760#else
761static inline
762void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
763{
764}
765#endif
766
767static inline
768struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
769 return css ? container_of(css, struct mem_cgroup, css) : NULL;
770}
771
772static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
773{
774 return percpu_ref_tryget(&objcg->refcnt);
775}
776
777static inline void obj_cgroup_get(struct obj_cgroup *objcg)
778{
779 percpu_ref_get(&objcg->refcnt);
780}
781
782static inline void obj_cgroup_get_many(struct obj_cgroup *objcg,
783 unsigned long nr)
784{
785 percpu_ref_get_many(&objcg->refcnt, nr);
786}
787
788static inline void obj_cgroup_put(struct obj_cgroup *objcg)
789{
790 percpu_ref_put(&objcg->refcnt);
791}
792
793static inline void mem_cgroup_put(struct mem_cgroup *memcg)
794{
795 if (memcg)
796 css_put(&memcg->css);
797}
798
799#define mem_cgroup_from_counter(counter, member) \
800 container_of(counter, struct mem_cgroup, member)
801
802struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
803 struct mem_cgroup *,
804 struct mem_cgroup_reclaim_cookie *);
805void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
806int mem_cgroup_scan_tasks(struct mem_cgroup *,
807 int (*)(struct task_struct *, void *), void *);
808
809static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
810{
811 if (mem_cgroup_disabled())
812 return 0;
813
814 return memcg->id.id;
815}
816struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
817
818#ifdef CONFIG_SHRINKER_DEBUG
819static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
820{
821 return memcg ? cgroup_ino(memcg->css.cgroup) : 0;
822}
823
824struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino);
825#endif
826
827static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
828{
829 return mem_cgroup_from_css(seq_css(m));
830}
831
832static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
833{
834 struct mem_cgroup_per_node *mz;
835
836 if (mem_cgroup_disabled())
837 return NULL;
838
839 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
840 return mz->memcg;
841}
842
843/**
844 * parent_mem_cgroup - find the accounting parent of a memcg
845 * @memcg: memcg whose parent to find
846 *
847 * Returns the parent memcg, or NULL if this is the root or the memory
848 * controller is in legacy no-hierarchy mode.
849 */
850static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
851{
852 return mem_cgroup_from_css(memcg->css.parent);
853}
854
855static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
856 struct mem_cgroup *root)
857{
858 if (root == memcg)
859 return true;
860 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
861}
862
863static inline bool mm_match_cgroup(struct mm_struct *mm,
864 struct mem_cgroup *memcg)
865{
866 struct mem_cgroup *task_memcg;
867 bool match = false;
868
869 rcu_read_lock();
870 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
871 if (task_memcg)
872 match = mem_cgroup_is_descendant(task_memcg, memcg);
873 rcu_read_unlock();
874 return match;
875}
876
877struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
878ino_t page_cgroup_ino(struct page *page);
879
880static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
881{
882 if (mem_cgroup_disabled())
883 return true;
884 return !!(memcg->css.flags & CSS_ONLINE);
885}
886
887void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
888 int zid, int nr_pages);
889
890static inline
891unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
892 enum lru_list lru, int zone_idx)
893{
894 struct mem_cgroup_per_node *mz;
895
896 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
897 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
898}
899
900void mem_cgroup_handle_over_high(void);
901
902unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
903
904unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
905
906void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
907 struct task_struct *p);
908
909void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
910
911static inline void mem_cgroup_enter_user_fault(void)
912{
913 WARN_ON(current->in_user_fault);
914 current->in_user_fault = 1;
915}
916
917static inline void mem_cgroup_exit_user_fault(void)
918{
919 WARN_ON(!current->in_user_fault);
920 current->in_user_fault = 0;
921}
922
923static inline bool task_in_memcg_oom(struct task_struct *p)
924{
925 return p->memcg_in_oom;
926}
927
928bool mem_cgroup_oom_synchronize(bool wait);
929struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
930 struct mem_cgroup *oom_domain);
931void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
932
933void folio_memcg_lock(struct folio *folio);
934void folio_memcg_unlock(struct folio *folio);
935void lock_page_memcg(struct page *page);
936void unlock_page_memcg(struct page *page);
937
938void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
939
940/* try to stablize folio_memcg() for all the pages in a memcg */
941static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
942{
943 rcu_read_lock();
944
945 if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account))
946 return true;
947
948 rcu_read_unlock();
949 return false;
950}
951
952static inline void mem_cgroup_unlock_pages(void)
953{
954 rcu_read_unlock();
955}
956
957/* idx can be of type enum memcg_stat_item or node_stat_item */
958static inline void mod_memcg_state(struct mem_cgroup *memcg,
959 int idx, int val)
960{
961 unsigned long flags;
962
963 local_irq_save(flags);
964 __mod_memcg_state(memcg, idx, val);
965 local_irq_restore(flags);
966}
967
968static inline void mod_memcg_page_state(struct page *page,
969 int idx, int val)
970{
971 struct mem_cgroup *memcg;
972
973 if (mem_cgroup_disabled())
974 return;
975
976 rcu_read_lock();
977 memcg = page_memcg(page);
978 if (memcg)
979 mod_memcg_state(memcg, idx, val);
980 rcu_read_unlock();
981}
982
983unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx);
984
985static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
986 enum node_stat_item idx)
987{
988 struct mem_cgroup_per_node *pn;
989 long x;
990
991 if (mem_cgroup_disabled())
992 return node_page_state(lruvec_pgdat(lruvec), idx);
993
994 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
995 x = READ_ONCE(pn->lruvec_stats.state[idx]);
996#ifdef CONFIG_SMP
997 if (x < 0)
998 x = 0;
999#endif
1000 return x;
1001}
1002
1003static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1004 enum node_stat_item idx)
1005{
1006 struct mem_cgroup_per_node *pn;
1007 long x = 0;
1008 int cpu;
1009
1010 if (mem_cgroup_disabled())
1011 return node_page_state(lruvec_pgdat(lruvec), idx);
1012
1013 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
1014 for_each_possible_cpu(cpu)
1015 x += per_cpu(pn->lruvec_stats_percpu->state[idx], cpu);
1016#ifdef CONFIG_SMP
1017 if (x < 0)
1018 x = 0;
1019#endif
1020 return x;
1021}
1022
1023void mem_cgroup_flush_stats(void);
1024void mem_cgroup_flush_stats_delayed(void);
1025
1026void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
1027 int val);
1028void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
1029
1030static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1031 int val)
1032{
1033 unsigned long flags;
1034
1035 local_irq_save(flags);
1036 __mod_lruvec_kmem_state(p, idx, val);
1037 local_irq_restore(flags);
1038}
1039
1040static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
1041 enum node_stat_item idx, int val)
1042{
1043 unsigned long flags;
1044
1045 local_irq_save(flags);
1046 __mod_memcg_lruvec_state(lruvec, idx, val);
1047 local_irq_restore(flags);
1048}
1049
1050void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
1051 unsigned long count);
1052
1053static inline void count_memcg_events(struct mem_cgroup *memcg,
1054 enum vm_event_item idx,
1055 unsigned long count)
1056{
1057 unsigned long flags;
1058
1059 local_irq_save(flags);
1060 __count_memcg_events(memcg, idx, count);
1061 local_irq_restore(flags);
1062}
1063
1064static inline void count_memcg_page_event(struct page *page,
1065 enum vm_event_item idx)
1066{
1067 struct mem_cgroup *memcg = page_memcg(page);
1068
1069 if (memcg)
1070 count_memcg_events(memcg, idx, 1);
1071}
1072
1073static inline void count_memcg_folio_events(struct folio *folio,
1074 enum vm_event_item idx, unsigned long nr)
1075{
1076 struct mem_cgroup *memcg = folio_memcg(folio);
1077
1078 if (memcg)
1079 count_memcg_events(memcg, idx, nr);
1080}
1081
1082static inline void count_memcg_event_mm(struct mm_struct *mm,
1083 enum vm_event_item idx)
1084{
1085 struct mem_cgroup *memcg;
1086
1087 if (mem_cgroup_disabled())
1088 return;
1089
1090 rcu_read_lock();
1091 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1092 if (likely(memcg))
1093 count_memcg_events(memcg, idx, 1);
1094 rcu_read_unlock();
1095}
1096
1097static inline void memcg_memory_event(struct mem_cgroup *memcg,
1098 enum memcg_memory_event event)
1099{
1100 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
1101 event == MEMCG_SWAP_FAIL;
1102
1103 atomic_long_inc(&memcg->memory_events_local[event]);
1104 if (!swap_event)
1105 cgroup_file_notify(&memcg->events_local_file);
1106
1107 do {
1108 atomic_long_inc(&memcg->memory_events[event]);
1109 if (swap_event)
1110 cgroup_file_notify(&memcg->swap_events_file);
1111 else
1112 cgroup_file_notify(&memcg->events_file);
1113
1114 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1115 break;
1116 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1117 break;
1118 } while ((memcg = parent_mem_cgroup(memcg)) &&
1119 !mem_cgroup_is_root(memcg));
1120}
1121
1122static inline void memcg_memory_event_mm(struct mm_struct *mm,
1123 enum memcg_memory_event event)
1124{
1125 struct mem_cgroup *memcg;
1126
1127 if (mem_cgroup_disabled())
1128 return;
1129
1130 rcu_read_lock();
1131 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1132 if (likely(memcg))
1133 memcg_memory_event(memcg, event);
1134 rcu_read_unlock();
1135}
1136
1137void split_page_memcg(struct page *head, unsigned int nr);
1138
1139unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1140 gfp_t gfp_mask,
1141 unsigned long *total_scanned);
1142
1143#else /* CONFIG_MEMCG */
1144
1145#define MEM_CGROUP_ID_SHIFT 0
1146#define MEM_CGROUP_ID_MAX 0
1147
1148static inline struct mem_cgroup *folio_memcg(struct folio *folio)
1149{
1150 return NULL;
1151}
1152
1153static inline struct mem_cgroup *page_memcg(struct page *page)
1154{
1155 return NULL;
1156}
1157
1158static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
1159{
1160 WARN_ON_ONCE(!rcu_read_lock_held());
1161 return NULL;
1162}
1163
1164static inline struct mem_cgroup *page_memcg_check(struct page *page)
1165{
1166 return NULL;
1167}
1168
1169static inline bool folio_memcg_kmem(struct folio *folio)
1170{
1171 return false;
1172}
1173
1174static inline bool PageMemcgKmem(struct page *page)
1175{
1176 return false;
1177}
1178
1179static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1180{
1181 return true;
1182}
1183
1184static inline bool mem_cgroup_disabled(void)
1185{
1186 return true;
1187}
1188
1189static inline void memcg_memory_event(struct mem_cgroup *memcg,
1190 enum memcg_memory_event event)
1191{
1192}
1193
1194static inline void memcg_memory_event_mm(struct mm_struct *mm,
1195 enum memcg_memory_event event)
1196{
1197}
1198
1199static inline void mem_cgroup_protection(struct mem_cgroup *root,
1200 struct mem_cgroup *memcg,
1201 unsigned long *min,
1202 unsigned long *low)
1203{
1204 *min = *low = 0;
1205}
1206
1207static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1208 struct mem_cgroup *memcg)
1209{
1210}
1211
1212static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
1213{
1214 return false;
1215}
1216
1217static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
1218{
1219 return false;
1220}
1221
1222static inline int mem_cgroup_charge(struct folio *folio,
1223 struct mm_struct *mm, gfp_t gfp)
1224{
1225 return 0;
1226}
1227
1228static inline int mem_cgroup_swapin_charge_folio(struct folio *folio,
1229 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
1230{
1231 return 0;
1232}
1233
1234static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry)
1235{
1236}
1237
1238static inline void mem_cgroup_uncharge(struct folio *folio)
1239{
1240}
1241
1242static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
1243{
1244}
1245
1246static inline void mem_cgroup_migrate(struct folio *old, struct folio *new)
1247{
1248}
1249
1250static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1251 struct pglist_data *pgdat)
1252{
1253 return &pgdat->__lruvec;
1254}
1255
1256static inline struct lruvec *folio_lruvec(struct folio *folio)
1257{
1258 struct pglist_data *pgdat = folio_pgdat(folio);
1259 return &pgdat->__lruvec;
1260}
1261
1262static inline
1263void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
1264{
1265}
1266
1267static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1268{
1269 return NULL;
1270}
1271
1272static inline bool mm_match_cgroup(struct mm_struct *mm,
1273 struct mem_cgroup *memcg)
1274{
1275 return true;
1276}
1277
1278static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1279{
1280 return NULL;
1281}
1282
1283static inline
1284struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css)
1285{
1286 return NULL;
1287}
1288
1289static inline void obj_cgroup_put(struct obj_cgroup *objcg)
1290{
1291}
1292
1293static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1294{
1295}
1296
1297static inline struct lruvec *folio_lruvec_lock(struct folio *folio)
1298{
1299 struct pglist_data *pgdat = folio_pgdat(folio);
1300
1301 spin_lock(&pgdat->__lruvec.lru_lock);
1302 return &pgdat->__lruvec;
1303}
1304
1305static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio)
1306{
1307 struct pglist_data *pgdat = folio_pgdat(folio);
1308
1309 spin_lock_irq(&pgdat->__lruvec.lru_lock);
1310 return &pgdat->__lruvec;
1311}
1312
1313static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
1314 unsigned long *flagsp)
1315{
1316 struct pglist_data *pgdat = folio_pgdat(folio);
1317
1318 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1319 return &pgdat->__lruvec;
1320}
1321
1322static inline struct mem_cgroup *
1323mem_cgroup_iter(struct mem_cgroup *root,
1324 struct mem_cgroup *prev,
1325 struct mem_cgroup_reclaim_cookie *reclaim)
1326{
1327 return NULL;
1328}
1329
1330static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1331 struct mem_cgroup *prev)
1332{
1333}
1334
1335static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1336 int (*fn)(struct task_struct *, void *), void *arg)
1337{
1338 return 0;
1339}
1340
1341static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1342{
1343 return 0;
1344}
1345
1346static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1347{
1348 WARN_ON_ONCE(id);
1349 /* XXX: This should always return root_mem_cgroup */
1350 return NULL;
1351}
1352
1353#ifdef CONFIG_SHRINKER_DEBUG
1354static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
1355{
1356 return 0;
1357}
1358
1359static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino)
1360{
1361 return NULL;
1362}
1363#endif
1364
1365static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1366{
1367 return NULL;
1368}
1369
1370static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1371{
1372 return NULL;
1373}
1374
1375static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1376{
1377 return true;
1378}
1379
1380static inline
1381unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1382 enum lru_list lru, int zone_idx)
1383{
1384 return 0;
1385}
1386
1387static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1388{
1389 return 0;
1390}
1391
1392static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1393{
1394 return 0;
1395}
1396
1397static inline void
1398mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1399{
1400}
1401
1402static inline void
1403mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1404{
1405}
1406
1407static inline void lock_page_memcg(struct page *page)
1408{
1409}
1410
1411static inline void unlock_page_memcg(struct page *page)
1412{
1413}
1414
1415static inline void folio_memcg_lock(struct folio *folio)
1416{
1417}
1418
1419static inline void folio_memcg_unlock(struct folio *folio)
1420{
1421}
1422
1423static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
1424{
1425 /* to match folio_memcg_rcu() */
1426 rcu_read_lock();
1427 return true;
1428}
1429
1430static inline void mem_cgroup_unlock_pages(void)
1431{
1432 rcu_read_unlock();
1433}
1434
1435static inline void mem_cgroup_handle_over_high(void)
1436{
1437}
1438
1439static inline void mem_cgroup_enter_user_fault(void)
1440{
1441}
1442
1443static inline void mem_cgroup_exit_user_fault(void)
1444{
1445}
1446
1447static inline bool task_in_memcg_oom(struct task_struct *p)
1448{
1449 return false;
1450}
1451
1452static inline bool mem_cgroup_oom_synchronize(bool wait)
1453{
1454 return false;
1455}
1456
1457static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1458 struct task_struct *victim, struct mem_cgroup *oom_domain)
1459{
1460 return NULL;
1461}
1462
1463static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1464{
1465}
1466
1467static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1468 int idx,
1469 int nr)
1470{
1471}
1472
1473static inline void mod_memcg_state(struct mem_cgroup *memcg,
1474 int idx,
1475 int nr)
1476{
1477}
1478
1479static inline void mod_memcg_page_state(struct page *page,
1480 int idx, int val)
1481{
1482}
1483
1484static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1485{
1486 return 0;
1487}
1488
1489static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1490 enum node_stat_item idx)
1491{
1492 return node_page_state(lruvec_pgdat(lruvec), idx);
1493}
1494
1495static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1496 enum node_stat_item idx)
1497{
1498 return node_page_state(lruvec_pgdat(lruvec), idx);
1499}
1500
1501static inline void mem_cgroup_flush_stats(void)
1502{
1503}
1504
1505static inline void mem_cgroup_flush_stats_delayed(void)
1506{
1507}
1508
1509static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1510 enum node_stat_item idx, int val)
1511{
1512}
1513
1514static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1515 int val)
1516{
1517 struct page *page = virt_to_head_page(p);
1518
1519 __mod_node_page_state(page_pgdat(page), idx, val);
1520}
1521
1522static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1523 int val)
1524{
1525 struct page *page = virt_to_head_page(p);
1526
1527 mod_node_page_state(page_pgdat(page), idx, val);
1528}
1529
1530static inline void count_memcg_events(struct mem_cgroup *memcg,
1531 enum vm_event_item idx,
1532 unsigned long count)
1533{
1534}
1535
1536static inline void __count_memcg_events(struct mem_cgroup *memcg,
1537 enum vm_event_item idx,
1538 unsigned long count)
1539{
1540}
1541
1542static inline void count_memcg_page_event(struct page *page,
1543 int idx)
1544{
1545}
1546
1547static inline void count_memcg_folio_events(struct folio *folio,
1548 enum vm_event_item idx, unsigned long nr)
1549{
1550}
1551
1552static inline
1553void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1554{
1555}
1556
1557static inline void split_page_memcg(struct page *head, unsigned int nr)
1558{
1559}
1560
1561static inline
1562unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1563 gfp_t gfp_mask,
1564 unsigned long *total_scanned)
1565{
1566 return 0;
1567}
1568#endif /* CONFIG_MEMCG */
1569
1570static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1571{
1572 __mod_lruvec_kmem_state(p, idx, 1);
1573}
1574
1575static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1576{
1577 __mod_lruvec_kmem_state(p, idx, -1);
1578}
1579
1580static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1581{
1582 struct mem_cgroup *memcg;
1583
1584 memcg = lruvec_memcg(lruvec);
1585 if (!memcg)
1586 return NULL;
1587 memcg = parent_mem_cgroup(memcg);
1588 if (!memcg)
1589 return NULL;
1590 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1591}
1592
1593static inline void unlock_page_lruvec(struct lruvec *lruvec)
1594{
1595 spin_unlock(&lruvec->lru_lock);
1596}
1597
1598static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1599{
1600 spin_unlock_irq(&lruvec->lru_lock);
1601}
1602
1603static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1604 unsigned long flags)
1605{
1606 spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1607}
1608
1609/* Test requires a stable page->memcg binding, see page_memcg() */
1610static inline bool folio_matches_lruvec(struct folio *folio,
1611 struct lruvec *lruvec)
1612{
1613 return lruvec_pgdat(lruvec) == folio_pgdat(folio) &&
1614 lruvec_memcg(lruvec) == folio_memcg(folio);
1615}
1616
1617/* Don't lock again iff page's lruvec locked */
1618static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio,
1619 struct lruvec *locked_lruvec)
1620{
1621 if (locked_lruvec) {
1622 if (folio_matches_lruvec(folio, locked_lruvec))
1623 return locked_lruvec;
1624
1625 unlock_page_lruvec_irq(locked_lruvec);
1626 }
1627
1628 return folio_lruvec_lock_irq(folio);
1629}
1630
1631/* Don't lock again iff page's lruvec locked */
1632static inline struct lruvec *folio_lruvec_relock_irqsave(struct folio *folio,
1633 struct lruvec *locked_lruvec, unsigned long *flags)
1634{
1635 if (locked_lruvec) {
1636 if (folio_matches_lruvec(folio, locked_lruvec))
1637 return locked_lruvec;
1638
1639 unlock_page_lruvec_irqrestore(locked_lruvec, *flags);
1640 }
1641
1642 return folio_lruvec_lock_irqsave(folio, flags);
1643}
1644
1645#ifdef CONFIG_CGROUP_WRITEBACK
1646
1647struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1648void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1649 unsigned long *pheadroom, unsigned long *pdirty,
1650 unsigned long *pwriteback);
1651
1652void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio,
1653 struct bdi_writeback *wb);
1654
1655static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1656 struct bdi_writeback *wb)
1657{
1658 if (mem_cgroup_disabled())
1659 return;
1660
1661 if (unlikely(&folio_memcg(folio)->css != wb->memcg_css))
1662 mem_cgroup_track_foreign_dirty_slowpath(folio, wb);
1663}
1664
1665void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1666
1667#else /* CONFIG_CGROUP_WRITEBACK */
1668
1669static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1670{
1671 return NULL;
1672}
1673
1674static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1675 unsigned long *pfilepages,
1676 unsigned long *pheadroom,
1677 unsigned long *pdirty,
1678 unsigned long *pwriteback)
1679{
1680}
1681
1682static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1683 struct bdi_writeback *wb)
1684{
1685}
1686
1687static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1688{
1689}
1690
1691#endif /* CONFIG_CGROUP_WRITEBACK */
1692
1693struct sock;
1694bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages,
1695 gfp_t gfp_mask);
1696void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1697#ifdef CONFIG_MEMCG
1698extern struct static_key_false memcg_sockets_enabled_key;
1699#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1700void mem_cgroup_sk_alloc(struct sock *sk);
1701void mem_cgroup_sk_free(struct sock *sk);
1702static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1703{
1704 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1705 return true;
1706 do {
1707 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure)))
1708 return true;
1709 } while ((memcg = parent_mem_cgroup(memcg)));
1710 return false;
1711}
1712
1713int alloc_shrinker_info(struct mem_cgroup *memcg);
1714void free_shrinker_info(struct mem_cgroup *memcg);
1715void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id);
1716void reparent_shrinker_deferred(struct mem_cgroup *memcg);
1717#else
1718#define mem_cgroup_sockets_enabled 0
1719static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1720static inline void mem_cgroup_sk_free(struct sock *sk) { };
1721static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1722{
1723 return false;
1724}
1725
1726static inline void set_shrinker_bit(struct mem_cgroup *memcg,
1727 int nid, int shrinker_id)
1728{
1729}
1730#endif
1731
1732#ifdef CONFIG_MEMCG_KMEM
1733bool mem_cgroup_kmem_disabled(void);
1734int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1735void __memcg_kmem_uncharge_page(struct page *page, int order);
1736
1737struct obj_cgroup *get_obj_cgroup_from_current(void);
1738struct obj_cgroup *get_obj_cgroup_from_page(struct page *page);
1739
1740int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1741void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1742
1743extern struct static_key_false memcg_kmem_enabled_key;
1744
1745static inline bool memcg_kmem_enabled(void)
1746{
1747 return static_branch_likely(&memcg_kmem_enabled_key);
1748}
1749
1750static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1751 int order)
1752{
1753 if (memcg_kmem_enabled())
1754 return __memcg_kmem_charge_page(page, gfp, order);
1755 return 0;
1756}
1757
1758static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1759{
1760 if (memcg_kmem_enabled())
1761 __memcg_kmem_uncharge_page(page, order);
1762}
1763
1764/*
1765 * A helper for accessing memcg's kmem_id, used for getting
1766 * corresponding LRU lists.
1767 */
1768static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1769{
1770 return memcg ? memcg->kmemcg_id : -1;
1771}
1772
1773struct mem_cgroup *mem_cgroup_from_obj(void *p);
1774struct mem_cgroup *mem_cgroup_from_slab_obj(void *p);
1775
1776static inline void count_objcg_event(struct obj_cgroup *objcg,
1777 enum vm_event_item idx)
1778{
1779 struct mem_cgroup *memcg;
1780
1781 if (!memcg_kmem_enabled())
1782 return;
1783
1784 rcu_read_lock();
1785 memcg = obj_cgroup_memcg(objcg);
1786 count_memcg_events(memcg, idx, 1);
1787 rcu_read_unlock();
1788}
1789
1790#else
1791static inline bool mem_cgroup_kmem_disabled(void)
1792{
1793 return true;
1794}
1795
1796static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1797 int order)
1798{
1799 return 0;
1800}
1801
1802static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1803{
1804}
1805
1806static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1807 int order)
1808{
1809 return 0;
1810}
1811
1812static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1813{
1814}
1815
1816static inline struct obj_cgroup *get_obj_cgroup_from_page(struct page *page)
1817{
1818 return NULL;
1819}
1820
1821static inline bool memcg_kmem_enabled(void)
1822{
1823 return false;
1824}
1825
1826static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1827{
1828 return -1;
1829}
1830
1831static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1832{
1833 return NULL;
1834}
1835
1836static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p)
1837{
1838 return NULL;
1839}
1840
1841static inline void count_objcg_event(struct obj_cgroup *objcg,
1842 enum vm_event_item idx)
1843{
1844}
1845
1846#endif /* CONFIG_MEMCG_KMEM */
1847
1848#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP)
1849bool obj_cgroup_may_zswap(struct obj_cgroup *objcg);
1850void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size);
1851void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size);
1852#else
1853static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg)
1854{
1855 return true;
1856}
1857static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg,
1858 size_t size)
1859{
1860}
1861static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg,
1862 size_t size)
1863{
1864}
1865#endif
1866
1867#endif /* _LINUX_MEMCONTROL_H */