Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/* memcontrol.h - Memory Controller
3 *
4 * Copyright IBM Corporation, 2007
5 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
6 *
7 * Copyright 2007 OpenVZ SWsoft Inc
8 * Author: Pavel Emelianov <xemul@openvz.org>
9 */
10
11#ifndef _LINUX_MEMCONTROL_H
12#define _LINUX_MEMCONTROL_H
13#include <linux/cgroup.h>
14#include <linux/vm_event_item.h>
15#include <linux/hardirq.h>
16#include <linux/jump_label.h>
17#include <linux/page_counter.h>
18#include <linux/vmpressure.h>
19#include <linux/eventfd.h>
20#include <linux/mm.h>
21#include <linux/vmstat.h>
22#include <linux/writeback.h>
23#include <linux/page-flags.h>
24
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_NR_STAT,
37};
38
39enum memcg_memory_event {
40 MEMCG_LOW,
41 MEMCG_HIGH,
42 MEMCG_MAX,
43 MEMCG_OOM,
44 MEMCG_OOM_KILL,
45 MEMCG_SWAP_HIGH,
46 MEMCG_SWAP_MAX,
47 MEMCG_SWAP_FAIL,
48 MEMCG_NR_MEMORY_EVENTS,
49};
50
51struct mem_cgroup_reclaim_cookie {
52 pg_data_t *pgdat;
53 unsigned int generation;
54};
55
56#ifdef CONFIG_MEMCG
57
58#define MEM_CGROUP_ID_SHIFT 16
59#define MEM_CGROUP_ID_MAX USHRT_MAX
60
61struct mem_cgroup_id {
62 int id;
63 refcount_t ref;
64};
65
66/*
67 * Per memcg event counter is incremented at every pagein/pageout. With THP,
68 * it will be incremented by the number of pages. This counter is used
69 * to trigger some periodic events. This is straightforward and better
70 * than using jiffies etc. to handle periodic memcg event.
71 */
72enum mem_cgroup_events_target {
73 MEM_CGROUP_TARGET_THRESH,
74 MEM_CGROUP_TARGET_SOFTLIMIT,
75 MEM_CGROUP_NTARGETS,
76};
77
78struct memcg_vmstats_percpu {
79 /* Local (CPU and cgroup) page state & events */
80 long state[MEMCG_NR_STAT];
81 unsigned long events[NR_VM_EVENT_ITEMS];
82
83 /* Delta calculation for lockless upward propagation */
84 long state_prev[MEMCG_NR_STAT];
85 unsigned long events_prev[NR_VM_EVENT_ITEMS];
86
87 /* Cgroup1: threshold notifications & softlimit tree updates */
88 unsigned long nr_page_events;
89 unsigned long targets[MEM_CGROUP_NTARGETS];
90};
91
92struct memcg_vmstats {
93 /* Aggregated (CPU and subtree) page state & events */
94 long state[MEMCG_NR_STAT];
95 unsigned long events[NR_VM_EVENT_ITEMS];
96
97 /* Pending child counts during tree propagation */
98 long state_pending[MEMCG_NR_STAT];
99 unsigned long events_pending[NR_VM_EVENT_ITEMS];
100};
101
102struct mem_cgroup_reclaim_iter {
103 struct mem_cgroup *position;
104 /* scan generation, increased every round-trip */
105 unsigned int generation;
106};
107
108/*
109 * Bitmap and deferred work of shrinker::id corresponding to memcg-aware
110 * shrinkers, which have elements charged to this memcg.
111 */
112struct shrinker_info {
113 struct rcu_head rcu;
114 atomic_long_t *nr_deferred;
115 unsigned long *map;
116};
117
118struct lruvec_stats_percpu {
119 /* Local (CPU and cgroup) state */
120 long state[NR_VM_NODE_STAT_ITEMS];
121
122 /* Delta calculation for lockless upward propagation */
123 long state_prev[NR_VM_NODE_STAT_ITEMS];
124};
125
126struct lruvec_stats {
127 /* Aggregated (CPU and subtree) state */
128 long state[NR_VM_NODE_STAT_ITEMS];
129
130 /* Pending child counts during tree propagation */
131 long state_pending[NR_VM_NODE_STAT_ITEMS];
132};
133
134/*
135 * per-node information in memory controller.
136 */
137struct mem_cgroup_per_node {
138 struct lruvec lruvec;
139
140 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu;
141 struct lruvec_stats lruvec_stats;
142
143 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
144
145 struct mem_cgroup_reclaim_iter iter;
146
147 struct shrinker_info __rcu *shrinker_info;
148
149 struct rb_node tree_node; /* RB tree node */
150 unsigned long usage_in_excess;/* Set to the value by which */
151 /* the soft limit is exceeded*/
152 bool on_tree;
153 struct mem_cgroup *memcg; /* Back pointer, we cannot */
154 /* use container_of */
155};
156
157struct mem_cgroup_threshold {
158 struct eventfd_ctx *eventfd;
159 unsigned long threshold;
160};
161
162/* For threshold */
163struct mem_cgroup_threshold_ary {
164 /* An array index points to threshold just below or equal to usage. */
165 int current_threshold;
166 /* Size of entries[] */
167 unsigned int size;
168 /* Array of thresholds */
169 struct mem_cgroup_threshold entries[];
170};
171
172struct mem_cgroup_thresholds {
173 /* Primary thresholds array */
174 struct mem_cgroup_threshold_ary *primary;
175 /*
176 * Spare threshold array.
177 * This is needed to make mem_cgroup_unregister_event() "never fail".
178 * It must be able to store at least primary->size - 1 entries.
179 */
180 struct mem_cgroup_threshold_ary *spare;
181};
182
183#if defined(CONFIG_SMP)
184struct memcg_padding {
185 char x[0];
186} ____cacheline_internodealigned_in_smp;
187#define MEMCG_PADDING(name) struct memcg_padding name
188#else
189#define MEMCG_PADDING(name)
190#endif
191
192/*
193 * Remember four most recent foreign writebacks with dirty pages in this
194 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
195 * one in a given round, we're likely to catch it later if it keeps
196 * foreign-dirtying, so a fairly low count should be enough.
197 *
198 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
199 */
200#define MEMCG_CGWB_FRN_CNT 4
201
202struct memcg_cgwb_frn {
203 u64 bdi_id; /* bdi->id of the foreign inode */
204 int memcg_id; /* memcg->css.id of foreign inode */
205 u64 at; /* jiffies_64 at the time of dirtying */
206 struct wb_completion done; /* tracks in-flight foreign writebacks */
207};
208
209/*
210 * Bucket for arbitrarily byte-sized objects charged to a memory
211 * cgroup. The bucket can be reparented in one piece when the cgroup
212 * is destroyed, without having to round up the individual references
213 * of all live memory objects in the wild.
214 */
215struct obj_cgroup {
216 struct percpu_ref refcnt;
217 struct mem_cgroup *memcg;
218 atomic_t nr_charged_bytes;
219 union {
220 struct list_head list;
221 struct rcu_head rcu;
222 };
223};
224
225/*
226 * The memory controller data structure. The memory controller controls both
227 * page cache and RSS per cgroup. We would eventually like to provide
228 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
229 * to help the administrator determine what knobs to tune.
230 */
231struct mem_cgroup {
232 struct cgroup_subsys_state css;
233
234 /* Private memcg ID. Used to ID objects that outlive the cgroup */
235 struct mem_cgroup_id id;
236
237 /* Accounted resources */
238 struct page_counter memory; /* Both v1 & v2 */
239
240 union {
241 struct page_counter swap; /* v2 only */
242 struct page_counter memsw; /* v1 only */
243 };
244
245 /* Legacy consumer-oriented counters */
246 struct page_counter kmem; /* v1 only */
247 struct page_counter tcpmem; /* v1 only */
248
249 /* Range enforcement for interrupt charges */
250 struct work_struct high_work;
251
252 unsigned long soft_limit;
253
254 /* vmpressure notifications */
255 struct vmpressure vmpressure;
256
257 /*
258 * Should the OOM killer kill all belonging tasks, had it kill one?
259 */
260 bool oom_group;
261
262 /* protected by memcg_oom_lock */
263 bool oom_lock;
264 int under_oom;
265
266 int swappiness;
267 /* OOM-Killer disable */
268 int oom_kill_disable;
269
270 /* memory.events and memory.events.local */
271 struct cgroup_file events_file;
272 struct cgroup_file events_local_file;
273
274 /* handle for "memory.swap.events" */
275 struct cgroup_file swap_events_file;
276
277 /* protect arrays of thresholds */
278 struct mutex thresholds_lock;
279
280 /* thresholds for memory usage. RCU-protected */
281 struct mem_cgroup_thresholds thresholds;
282
283 /* thresholds for mem+swap usage. RCU-protected */
284 struct mem_cgroup_thresholds memsw_thresholds;
285
286 /* For oom notifier event fd */
287 struct list_head oom_notify;
288
289 /*
290 * Should we move charges of a task when a task is moved into this
291 * mem_cgroup ? And what type of charges should we move ?
292 */
293 unsigned long move_charge_at_immigrate;
294 /* taken only while moving_account > 0 */
295 spinlock_t move_lock;
296 unsigned long move_lock_flags;
297
298 MEMCG_PADDING(_pad1_);
299
300 /* memory.stat */
301 struct memcg_vmstats vmstats;
302
303 /* memory.events */
304 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
305 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
306
307 unsigned long socket_pressure;
308
309 /* Legacy tcp memory accounting */
310 bool tcpmem_active;
311 int tcpmem_pressure;
312
313#ifdef CONFIG_MEMCG_KMEM
314 int kmemcg_id;
315 struct obj_cgroup __rcu *objcg;
316 struct list_head objcg_list; /* list of inherited objcgs */
317#endif
318
319 MEMCG_PADDING(_pad2_);
320
321 /*
322 * set > 0 if pages under this cgroup are moving to other cgroup.
323 */
324 atomic_t moving_account;
325 struct task_struct *move_lock_task;
326
327 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
328
329#ifdef CONFIG_CGROUP_WRITEBACK
330 struct list_head cgwb_list;
331 struct wb_domain cgwb_domain;
332 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
333#endif
334
335 /* List of events which userspace want to receive */
336 struct list_head event_list;
337 spinlock_t event_list_lock;
338
339#ifdef CONFIG_TRANSPARENT_HUGEPAGE
340 struct deferred_split deferred_split_queue;
341#endif
342
343 struct mem_cgroup_per_node *nodeinfo[];
344};
345
346/*
347 * size of first charge trial. "32" comes from vmscan.c's magic value.
348 * TODO: maybe necessary to use big numbers in big irons.
349 */
350#define MEMCG_CHARGE_BATCH 32U
351
352extern struct mem_cgroup *root_mem_cgroup;
353
354enum page_memcg_data_flags {
355 /* page->memcg_data is a pointer to an objcgs vector */
356 MEMCG_DATA_OBJCGS = (1UL << 0),
357 /* page has been accounted as a non-slab kernel page */
358 MEMCG_DATA_KMEM = (1UL << 1),
359 /* the next bit after the last actual flag */
360 __NR_MEMCG_DATA_FLAGS = (1UL << 2),
361};
362
363#define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1)
364
365static inline bool folio_memcg_kmem(struct folio *folio);
366
367/*
368 * After the initialization objcg->memcg is always pointing at
369 * a valid memcg, but can be atomically swapped to the parent memcg.
370 *
371 * The caller must ensure that the returned memcg won't be released:
372 * e.g. acquire the rcu_read_lock or css_set_lock.
373 */
374static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
375{
376 return READ_ONCE(objcg->memcg);
377}
378
379/*
380 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio
381 * @folio: Pointer to the folio.
382 *
383 * Returns a pointer to the memory cgroup associated with the folio,
384 * or NULL. This function assumes that the folio is known to have a
385 * proper memory cgroup pointer. It's not safe to call this function
386 * against some type of folios, e.g. slab folios or ex-slab folios or
387 * kmem folios.
388 */
389static inline struct mem_cgroup *__folio_memcg(struct folio *folio)
390{
391 unsigned long memcg_data = folio->memcg_data;
392
393 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
394 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio);
395 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio);
396
397 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
398}
399
400/*
401 * __folio_objcg - get the object cgroup associated with a kmem folio.
402 * @folio: Pointer to the folio.
403 *
404 * Returns a pointer to the object cgroup associated with the folio,
405 * or NULL. This function assumes that the folio is known to have a
406 * proper object cgroup pointer. It's not safe to call this function
407 * against some type of folios, e.g. slab folios or ex-slab folios or
408 * LRU folios.
409 */
410static inline struct obj_cgroup *__folio_objcg(struct folio *folio)
411{
412 unsigned long memcg_data = folio->memcg_data;
413
414 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
415 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio);
416 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio);
417
418 return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
419}
420
421/*
422 * folio_memcg - Get the memory cgroup associated with a folio.
423 * @folio: Pointer to the folio.
424 *
425 * Returns a pointer to the memory cgroup associated with the folio,
426 * or NULL. This function assumes that the folio is known to have a
427 * proper memory cgroup pointer. It's not safe to call this function
428 * against some type of folios, e.g. slab folios or ex-slab folios.
429 *
430 * For a non-kmem folio any of the following ensures folio and memcg binding
431 * stability:
432 *
433 * - the folio lock
434 * - LRU isolation
435 * - lock_page_memcg()
436 * - exclusive reference
437 *
438 * For a kmem folio a caller should hold an rcu read lock to protect memcg
439 * associated with a kmem folio from being released.
440 */
441static inline struct mem_cgroup *folio_memcg(struct folio *folio)
442{
443 if (folio_memcg_kmem(folio))
444 return obj_cgroup_memcg(__folio_objcg(folio));
445 return __folio_memcg(folio);
446}
447
448static inline struct mem_cgroup *page_memcg(struct page *page)
449{
450 return folio_memcg(page_folio(page));
451}
452
453/**
454 * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio.
455 * @folio: Pointer to the folio.
456 *
457 * This function assumes that the folio is known to have a
458 * proper memory cgroup pointer. It's not safe to call this function
459 * against some type of folios, e.g. slab folios or ex-slab folios.
460 *
461 * Return: A pointer to the memory cgroup associated with the folio,
462 * or NULL.
463 */
464static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
465{
466 unsigned long memcg_data = READ_ONCE(folio->memcg_data);
467
468 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
469 WARN_ON_ONCE(!rcu_read_lock_held());
470
471 if (memcg_data & MEMCG_DATA_KMEM) {
472 struct obj_cgroup *objcg;
473
474 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
475 return obj_cgroup_memcg(objcg);
476 }
477
478 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
479}
480
481/*
482 * page_memcg_check - get the memory cgroup associated with a page
483 * @page: a pointer to the page struct
484 *
485 * Returns a pointer to the memory cgroup associated with the page,
486 * or NULL. This function unlike page_memcg() can take any page
487 * as an argument. It has to be used in cases when it's not known if a page
488 * has an associated memory cgroup pointer or an object cgroups vector or
489 * an object cgroup.
490 *
491 * For a non-kmem page any of the following ensures page and memcg binding
492 * stability:
493 *
494 * - the page lock
495 * - LRU isolation
496 * - lock_page_memcg()
497 * - exclusive reference
498 *
499 * For a kmem page a caller should hold an rcu read lock to protect memcg
500 * associated with a kmem page from being released.
501 */
502static inline struct mem_cgroup *page_memcg_check(struct page *page)
503{
504 /*
505 * Because page->memcg_data might be changed asynchronously
506 * for slab pages, READ_ONCE() should be used here.
507 */
508 unsigned long memcg_data = READ_ONCE(page->memcg_data);
509
510 if (memcg_data & MEMCG_DATA_OBJCGS)
511 return NULL;
512
513 if (memcg_data & MEMCG_DATA_KMEM) {
514 struct obj_cgroup *objcg;
515
516 objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
517 return obj_cgroup_memcg(objcg);
518 }
519
520 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
521}
522
523#ifdef CONFIG_MEMCG_KMEM
524/*
525 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set.
526 * @folio: Pointer to the folio.
527 *
528 * Checks if the folio has MemcgKmem flag set. The caller must ensure
529 * that the folio has an associated memory cgroup. It's not safe to call
530 * this function against some types of folios, e.g. slab folios.
531 */
532static inline bool folio_memcg_kmem(struct folio *folio)
533{
534 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page);
535 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJCGS, folio);
536 return folio->memcg_data & MEMCG_DATA_KMEM;
537}
538
539/*
540 * page_objcgs - get the object cgroups vector associated with a page
541 * @page: a pointer to the page struct
542 *
543 * Returns a pointer to the object cgroups vector associated with the page,
544 * or NULL. This function assumes that the page is known to have an
545 * associated object cgroups vector. It's not safe to call this function
546 * against pages, which might have an associated memory cgroup: e.g.
547 * kernel stack pages.
548 */
549static inline struct obj_cgroup **page_objcgs(struct page *page)
550{
551 unsigned long memcg_data = READ_ONCE(page->memcg_data);
552
553 VM_BUG_ON_PAGE(memcg_data && !(memcg_data & MEMCG_DATA_OBJCGS), page);
554 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
555
556 return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
557}
558
559/*
560 * page_objcgs_check - get the object cgroups vector associated with a page
561 * @page: a pointer to the page struct
562 *
563 * Returns a pointer to the object cgroups vector associated with the page,
564 * or NULL. This function is safe to use if the page can be directly associated
565 * with a memory cgroup.
566 */
567static inline struct obj_cgroup **page_objcgs_check(struct page *page)
568{
569 unsigned long memcg_data = READ_ONCE(page->memcg_data);
570
571 if (!memcg_data || !(memcg_data & MEMCG_DATA_OBJCGS))
572 return NULL;
573
574 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
575
576 return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
577}
578
579#else
580static inline bool folio_memcg_kmem(struct folio *folio)
581{
582 return false;
583}
584
585static inline struct obj_cgroup **page_objcgs(struct page *page)
586{
587 return NULL;
588}
589
590static inline struct obj_cgroup **page_objcgs_check(struct page *page)
591{
592 return NULL;
593}
594#endif
595
596static inline bool PageMemcgKmem(struct page *page)
597{
598 return folio_memcg_kmem(page_folio(page));
599}
600
601static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
602{
603 return (memcg == root_mem_cgroup);
604}
605
606static inline bool mem_cgroup_disabled(void)
607{
608 return !cgroup_subsys_enabled(memory_cgrp_subsys);
609}
610
611static inline void mem_cgroup_protection(struct mem_cgroup *root,
612 struct mem_cgroup *memcg,
613 unsigned long *min,
614 unsigned long *low)
615{
616 *min = *low = 0;
617
618 if (mem_cgroup_disabled())
619 return;
620
621 /*
622 * There is no reclaim protection applied to a targeted reclaim.
623 * We are special casing this specific case here because
624 * mem_cgroup_protected calculation is not robust enough to keep
625 * the protection invariant for calculated effective values for
626 * parallel reclaimers with different reclaim target. This is
627 * especially a problem for tail memcgs (as they have pages on LRU)
628 * which would want to have effective values 0 for targeted reclaim
629 * but a different value for external reclaim.
630 *
631 * Example
632 * Let's have global and A's reclaim in parallel:
633 * |
634 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
635 * |\
636 * | C (low = 1G, usage = 2.5G)
637 * B (low = 1G, usage = 0.5G)
638 *
639 * For the global reclaim
640 * A.elow = A.low
641 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
642 * C.elow = min(C.usage, C.low)
643 *
644 * With the effective values resetting we have A reclaim
645 * A.elow = 0
646 * B.elow = B.low
647 * C.elow = C.low
648 *
649 * If the global reclaim races with A's reclaim then
650 * B.elow = C.elow = 0 because children_low_usage > A.elow)
651 * is possible and reclaiming B would be violating the protection.
652 *
653 */
654 if (root == memcg)
655 return;
656
657 *min = READ_ONCE(memcg->memory.emin);
658 *low = READ_ONCE(memcg->memory.elow);
659}
660
661void mem_cgroup_calculate_protection(struct mem_cgroup *root,
662 struct mem_cgroup *memcg);
663
664static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg)
665{
666 /*
667 * The root memcg doesn't account charges, and doesn't support
668 * protection.
669 */
670 return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg);
671
672}
673
674static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
675{
676 if (!mem_cgroup_supports_protection(memcg))
677 return false;
678
679 return READ_ONCE(memcg->memory.elow) >=
680 page_counter_read(&memcg->memory);
681}
682
683static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
684{
685 if (!mem_cgroup_supports_protection(memcg))
686 return false;
687
688 return READ_ONCE(memcg->memory.emin) >=
689 page_counter_read(&memcg->memory);
690}
691
692int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp);
693
694/**
695 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup.
696 * @folio: Folio to charge.
697 * @mm: mm context of the allocating task.
698 * @gfp: Reclaim mode.
699 *
700 * Try to charge @folio to the memcg that @mm belongs to, reclaiming
701 * pages according to @gfp if necessary. If @mm is NULL, try to
702 * charge to the active memcg.
703 *
704 * Do not use this for folios allocated for swapin.
705 *
706 * Return: 0 on success. Otherwise, an error code is returned.
707 */
708static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm,
709 gfp_t gfp)
710{
711 if (mem_cgroup_disabled())
712 return 0;
713 return __mem_cgroup_charge(folio, mm, gfp);
714}
715
716int mem_cgroup_swapin_charge_page(struct page *page, struct mm_struct *mm,
717 gfp_t gfp, swp_entry_t entry);
718void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry);
719
720void __mem_cgroup_uncharge(struct folio *folio);
721
722/**
723 * mem_cgroup_uncharge - Uncharge a folio.
724 * @folio: Folio to uncharge.
725 *
726 * Uncharge a folio previously charged with mem_cgroup_charge().
727 */
728static inline void mem_cgroup_uncharge(struct folio *folio)
729{
730 if (mem_cgroup_disabled())
731 return;
732 __mem_cgroup_uncharge(folio);
733}
734
735void __mem_cgroup_uncharge_list(struct list_head *page_list);
736static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
737{
738 if (mem_cgroup_disabled())
739 return;
740 __mem_cgroup_uncharge_list(page_list);
741}
742
743void mem_cgroup_migrate(struct folio *old, struct folio *new);
744
745/**
746 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
747 * @memcg: memcg of the wanted lruvec
748 * @pgdat: pglist_data
749 *
750 * Returns the lru list vector holding pages for a given @memcg &
751 * @pgdat combination. This can be the node lruvec, if the memory
752 * controller is disabled.
753 */
754static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
755 struct pglist_data *pgdat)
756{
757 struct mem_cgroup_per_node *mz;
758 struct lruvec *lruvec;
759
760 if (mem_cgroup_disabled()) {
761 lruvec = &pgdat->__lruvec;
762 goto out;
763 }
764
765 if (!memcg)
766 memcg = root_mem_cgroup;
767
768 mz = memcg->nodeinfo[pgdat->node_id];
769 lruvec = &mz->lruvec;
770out:
771 /*
772 * Since a node can be onlined after the mem_cgroup was created,
773 * we have to be prepared to initialize lruvec->pgdat here;
774 * and if offlined then reonlined, we need to reinitialize it.
775 */
776 if (unlikely(lruvec->pgdat != pgdat))
777 lruvec->pgdat = pgdat;
778 return lruvec;
779}
780
781/**
782 * folio_lruvec - return lruvec for isolating/putting an LRU folio
783 * @folio: Pointer to the folio.
784 *
785 * This function relies on folio->mem_cgroup being stable.
786 */
787static inline struct lruvec *folio_lruvec(struct folio *folio)
788{
789 struct mem_cgroup *memcg = folio_memcg(folio);
790
791 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio);
792 return mem_cgroup_lruvec(memcg, folio_pgdat(folio));
793}
794
795struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
796
797struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
798
799struct lruvec *folio_lruvec_lock(struct folio *folio);
800struct lruvec *folio_lruvec_lock_irq(struct folio *folio);
801struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
802 unsigned long *flags);
803
804#ifdef CONFIG_DEBUG_VM
805void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio);
806#else
807static inline
808void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
809{
810}
811#endif
812
813static inline
814struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
815 return css ? container_of(css, struct mem_cgroup, css) : NULL;
816}
817
818static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
819{
820 return percpu_ref_tryget(&objcg->refcnt);
821}
822
823static inline void obj_cgroup_get(struct obj_cgroup *objcg)
824{
825 percpu_ref_get(&objcg->refcnt);
826}
827
828static inline void obj_cgroup_get_many(struct obj_cgroup *objcg,
829 unsigned long nr)
830{
831 percpu_ref_get_many(&objcg->refcnt, nr);
832}
833
834static inline void obj_cgroup_put(struct obj_cgroup *objcg)
835{
836 percpu_ref_put(&objcg->refcnt);
837}
838
839static inline void mem_cgroup_put(struct mem_cgroup *memcg)
840{
841 if (memcg)
842 css_put(&memcg->css);
843}
844
845#define mem_cgroup_from_counter(counter, member) \
846 container_of(counter, struct mem_cgroup, member)
847
848struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
849 struct mem_cgroup *,
850 struct mem_cgroup_reclaim_cookie *);
851void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
852int mem_cgroup_scan_tasks(struct mem_cgroup *,
853 int (*)(struct task_struct *, void *), void *);
854
855static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
856{
857 if (mem_cgroup_disabled())
858 return 0;
859
860 return memcg->id.id;
861}
862struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
863
864static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
865{
866 return mem_cgroup_from_css(seq_css(m));
867}
868
869static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
870{
871 struct mem_cgroup_per_node *mz;
872
873 if (mem_cgroup_disabled())
874 return NULL;
875
876 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
877 return mz->memcg;
878}
879
880/**
881 * parent_mem_cgroup - find the accounting parent of a memcg
882 * @memcg: memcg whose parent to find
883 *
884 * Returns the parent memcg, or NULL if this is the root or the memory
885 * controller is in legacy no-hierarchy mode.
886 */
887static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
888{
889 if (!memcg->memory.parent)
890 return NULL;
891 return mem_cgroup_from_counter(memcg->memory.parent, memory);
892}
893
894static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
895 struct mem_cgroup *root)
896{
897 if (root == memcg)
898 return true;
899 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
900}
901
902static inline bool mm_match_cgroup(struct mm_struct *mm,
903 struct mem_cgroup *memcg)
904{
905 struct mem_cgroup *task_memcg;
906 bool match = false;
907
908 rcu_read_lock();
909 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
910 if (task_memcg)
911 match = mem_cgroup_is_descendant(task_memcg, memcg);
912 rcu_read_unlock();
913 return match;
914}
915
916struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
917ino_t page_cgroup_ino(struct page *page);
918
919static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
920{
921 if (mem_cgroup_disabled())
922 return true;
923 return !!(memcg->css.flags & CSS_ONLINE);
924}
925
926void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
927 int zid, int nr_pages);
928
929static inline
930unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
931 enum lru_list lru, int zone_idx)
932{
933 struct mem_cgroup_per_node *mz;
934
935 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
936 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
937}
938
939void mem_cgroup_handle_over_high(void);
940
941unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
942
943unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
944
945void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
946 struct task_struct *p);
947
948void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
949
950static inline void mem_cgroup_enter_user_fault(void)
951{
952 WARN_ON(current->in_user_fault);
953 current->in_user_fault = 1;
954}
955
956static inline void mem_cgroup_exit_user_fault(void)
957{
958 WARN_ON(!current->in_user_fault);
959 current->in_user_fault = 0;
960}
961
962static inline bool task_in_memcg_oom(struct task_struct *p)
963{
964 return p->memcg_in_oom;
965}
966
967bool mem_cgroup_oom_synchronize(bool wait);
968struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
969 struct mem_cgroup *oom_domain);
970void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
971
972#ifdef CONFIG_MEMCG_SWAP
973extern bool cgroup_memory_noswap;
974#endif
975
976void folio_memcg_lock(struct folio *folio);
977void folio_memcg_unlock(struct folio *folio);
978void lock_page_memcg(struct page *page);
979void unlock_page_memcg(struct page *page);
980
981void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
982
983/* idx can be of type enum memcg_stat_item or node_stat_item */
984static inline void mod_memcg_state(struct mem_cgroup *memcg,
985 int idx, int val)
986{
987 unsigned long flags;
988
989 local_irq_save(flags);
990 __mod_memcg_state(memcg, idx, val);
991 local_irq_restore(flags);
992}
993
994static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
995{
996 return READ_ONCE(memcg->vmstats.state[idx]);
997}
998
999static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1000 enum node_stat_item idx)
1001{
1002 struct mem_cgroup_per_node *pn;
1003
1004 if (mem_cgroup_disabled())
1005 return node_page_state(lruvec_pgdat(lruvec), idx);
1006
1007 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
1008 return READ_ONCE(pn->lruvec_stats.state[idx]);
1009}
1010
1011static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1012 enum node_stat_item idx)
1013{
1014 struct mem_cgroup_per_node *pn;
1015 long x = 0;
1016 int cpu;
1017
1018 if (mem_cgroup_disabled())
1019 return node_page_state(lruvec_pgdat(lruvec), idx);
1020
1021 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
1022 for_each_possible_cpu(cpu)
1023 x += per_cpu(pn->lruvec_stats_percpu->state[idx], cpu);
1024#ifdef CONFIG_SMP
1025 if (x < 0)
1026 x = 0;
1027#endif
1028 return x;
1029}
1030
1031void mem_cgroup_flush_stats(void);
1032
1033void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
1034 int val);
1035void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
1036
1037static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1038 int val)
1039{
1040 unsigned long flags;
1041
1042 local_irq_save(flags);
1043 __mod_lruvec_kmem_state(p, idx, val);
1044 local_irq_restore(flags);
1045}
1046
1047static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
1048 enum node_stat_item idx, int val)
1049{
1050 unsigned long flags;
1051
1052 local_irq_save(flags);
1053 __mod_memcg_lruvec_state(lruvec, idx, val);
1054 local_irq_restore(flags);
1055}
1056
1057void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
1058 unsigned long count);
1059
1060static inline void count_memcg_events(struct mem_cgroup *memcg,
1061 enum vm_event_item idx,
1062 unsigned long count)
1063{
1064 unsigned long flags;
1065
1066 local_irq_save(flags);
1067 __count_memcg_events(memcg, idx, count);
1068 local_irq_restore(flags);
1069}
1070
1071static inline void count_memcg_page_event(struct page *page,
1072 enum vm_event_item idx)
1073{
1074 struct mem_cgroup *memcg = page_memcg(page);
1075
1076 if (memcg)
1077 count_memcg_events(memcg, idx, 1);
1078}
1079
1080static inline void count_memcg_event_mm(struct mm_struct *mm,
1081 enum vm_event_item idx)
1082{
1083 struct mem_cgroup *memcg;
1084
1085 if (mem_cgroup_disabled())
1086 return;
1087
1088 rcu_read_lock();
1089 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1090 if (likely(memcg))
1091 count_memcg_events(memcg, idx, 1);
1092 rcu_read_unlock();
1093}
1094
1095static inline void memcg_memory_event(struct mem_cgroup *memcg,
1096 enum memcg_memory_event event)
1097{
1098 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
1099 event == MEMCG_SWAP_FAIL;
1100
1101 atomic_long_inc(&memcg->memory_events_local[event]);
1102 if (!swap_event)
1103 cgroup_file_notify(&memcg->events_local_file);
1104
1105 do {
1106 atomic_long_inc(&memcg->memory_events[event]);
1107 if (swap_event)
1108 cgroup_file_notify(&memcg->swap_events_file);
1109 else
1110 cgroup_file_notify(&memcg->events_file);
1111
1112 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1113 break;
1114 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1115 break;
1116 } while ((memcg = parent_mem_cgroup(memcg)) &&
1117 !mem_cgroup_is_root(memcg));
1118}
1119
1120static inline void memcg_memory_event_mm(struct mm_struct *mm,
1121 enum memcg_memory_event event)
1122{
1123 struct mem_cgroup *memcg;
1124
1125 if (mem_cgroup_disabled())
1126 return;
1127
1128 rcu_read_lock();
1129 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1130 if (likely(memcg))
1131 memcg_memory_event(memcg, event);
1132 rcu_read_unlock();
1133}
1134
1135void split_page_memcg(struct page *head, unsigned int nr);
1136
1137unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1138 gfp_t gfp_mask,
1139 unsigned long *total_scanned);
1140
1141#else /* CONFIG_MEMCG */
1142
1143#define MEM_CGROUP_ID_SHIFT 0
1144#define MEM_CGROUP_ID_MAX 0
1145
1146static inline struct mem_cgroup *folio_memcg(struct folio *folio)
1147{
1148 return NULL;
1149}
1150
1151static inline struct mem_cgroup *page_memcg(struct page *page)
1152{
1153 return NULL;
1154}
1155
1156static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
1157{
1158 WARN_ON_ONCE(!rcu_read_lock_held());
1159 return NULL;
1160}
1161
1162static inline struct mem_cgroup *page_memcg_check(struct page *page)
1163{
1164 return NULL;
1165}
1166
1167static inline bool folio_memcg_kmem(struct folio *folio)
1168{
1169 return false;
1170}
1171
1172static inline bool PageMemcgKmem(struct page *page)
1173{
1174 return false;
1175}
1176
1177static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1178{
1179 return true;
1180}
1181
1182static inline bool mem_cgroup_disabled(void)
1183{
1184 return true;
1185}
1186
1187static inline void memcg_memory_event(struct mem_cgroup *memcg,
1188 enum memcg_memory_event event)
1189{
1190}
1191
1192static inline void memcg_memory_event_mm(struct mm_struct *mm,
1193 enum memcg_memory_event event)
1194{
1195}
1196
1197static inline void mem_cgroup_protection(struct mem_cgroup *root,
1198 struct mem_cgroup *memcg,
1199 unsigned long *min,
1200 unsigned long *low)
1201{
1202 *min = *low = 0;
1203}
1204
1205static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1206 struct mem_cgroup *memcg)
1207{
1208}
1209
1210static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
1211{
1212 return false;
1213}
1214
1215static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
1216{
1217 return false;
1218}
1219
1220static inline int mem_cgroup_charge(struct folio *folio,
1221 struct mm_struct *mm, gfp_t gfp)
1222{
1223 return 0;
1224}
1225
1226static inline int mem_cgroup_swapin_charge_page(struct page *page,
1227 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
1228{
1229 return 0;
1230}
1231
1232static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry)
1233{
1234}
1235
1236static inline void mem_cgroup_uncharge(struct folio *folio)
1237{
1238}
1239
1240static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
1241{
1242}
1243
1244static inline void mem_cgroup_migrate(struct folio *old, struct folio *new)
1245{
1246}
1247
1248static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1249 struct pglist_data *pgdat)
1250{
1251 return &pgdat->__lruvec;
1252}
1253
1254static inline struct lruvec *folio_lruvec(struct folio *folio)
1255{
1256 struct pglist_data *pgdat = folio_pgdat(folio);
1257 return &pgdat->__lruvec;
1258}
1259
1260static inline
1261void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
1262{
1263}
1264
1265static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1266{
1267 return NULL;
1268}
1269
1270static inline bool mm_match_cgroup(struct mm_struct *mm,
1271 struct mem_cgroup *memcg)
1272{
1273 return true;
1274}
1275
1276static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1277{
1278 return NULL;
1279}
1280
1281static inline
1282struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css)
1283{
1284 return NULL;
1285}
1286
1287static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1288{
1289}
1290
1291static inline struct lruvec *folio_lruvec_lock(struct folio *folio)
1292{
1293 struct pglist_data *pgdat = folio_pgdat(folio);
1294
1295 spin_lock(&pgdat->__lruvec.lru_lock);
1296 return &pgdat->__lruvec;
1297}
1298
1299static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio)
1300{
1301 struct pglist_data *pgdat = folio_pgdat(folio);
1302
1303 spin_lock_irq(&pgdat->__lruvec.lru_lock);
1304 return &pgdat->__lruvec;
1305}
1306
1307static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
1308 unsigned long *flagsp)
1309{
1310 struct pglist_data *pgdat = folio_pgdat(folio);
1311
1312 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1313 return &pgdat->__lruvec;
1314}
1315
1316static inline struct mem_cgroup *
1317mem_cgroup_iter(struct mem_cgroup *root,
1318 struct mem_cgroup *prev,
1319 struct mem_cgroup_reclaim_cookie *reclaim)
1320{
1321 return NULL;
1322}
1323
1324static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1325 struct mem_cgroup *prev)
1326{
1327}
1328
1329static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1330 int (*fn)(struct task_struct *, void *), void *arg)
1331{
1332 return 0;
1333}
1334
1335static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1336{
1337 return 0;
1338}
1339
1340static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1341{
1342 WARN_ON_ONCE(id);
1343 /* XXX: This should always return root_mem_cgroup */
1344 return NULL;
1345}
1346
1347static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1348{
1349 return NULL;
1350}
1351
1352static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1353{
1354 return NULL;
1355}
1356
1357static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1358{
1359 return true;
1360}
1361
1362static inline
1363unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1364 enum lru_list lru, int zone_idx)
1365{
1366 return 0;
1367}
1368
1369static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1370{
1371 return 0;
1372}
1373
1374static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1375{
1376 return 0;
1377}
1378
1379static inline void
1380mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1381{
1382}
1383
1384static inline void
1385mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1386{
1387}
1388
1389static inline void lock_page_memcg(struct page *page)
1390{
1391}
1392
1393static inline void unlock_page_memcg(struct page *page)
1394{
1395}
1396
1397static inline void folio_memcg_lock(struct folio *folio)
1398{
1399}
1400
1401static inline void folio_memcg_unlock(struct folio *folio)
1402{
1403}
1404
1405static inline void mem_cgroup_handle_over_high(void)
1406{
1407}
1408
1409static inline void mem_cgroup_enter_user_fault(void)
1410{
1411}
1412
1413static inline void mem_cgroup_exit_user_fault(void)
1414{
1415}
1416
1417static inline bool task_in_memcg_oom(struct task_struct *p)
1418{
1419 return false;
1420}
1421
1422static inline bool mem_cgroup_oom_synchronize(bool wait)
1423{
1424 return false;
1425}
1426
1427static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1428 struct task_struct *victim, struct mem_cgroup *oom_domain)
1429{
1430 return NULL;
1431}
1432
1433static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1434{
1435}
1436
1437static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1438 int idx,
1439 int nr)
1440{
1441}
1442
1443static inline void mod_memcg_state(struct mem_cgroup *memcg,
1444 int idx,
1445 int nr)
1446{
1447}
1448
1449static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1450{
1451 return 0;
1452}
1453
1454static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1455 enum node_stat_item idx)
1456{
1457 return node_page_state(lruvec_pgdat(lruvec), idx);
1458}
1459
1460static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1461 enum node_stat_item idx)
1462{
1463 return node_page_state(lruvec_pgdat(lruvec), idx);
1464}
1465
1466static inline void mem_cgroup_flush_stats(void)
1467{
1468}
1469
1470static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1471 enum node_stat_item idx, int val)
1472{
1473}
1474
1475static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1476 int val)
1477{
1478 struct page *page = virt_to_head_page(p);
1479
1480 __mod_node_page_state(page_pgdat(page), idx, val);
1481}
1482
1483static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1484 int val)
1485{
1486 struct page *page = virt_to_head_page(p);
1487
1488 mod_node_page_state(page_pgdat(page), idx, val);
1489}
1490
1491static inline void count_memcg_events(struct mem_cgroup *memcg,
1492 enum vm_event_item idx,
1493 unsigned long count)
1494{
1495}
1496
1497static inline void __count_memcg_events(struct mem_cgroup *memcg,
1498 enum vm_event_item idx,
1499 unsigned long count)
1500{
1501}
1502
1503static inline void count_memcg_page_event(struct page *page,
1504 int idx)
1505{
1506}
1507
1508static inline
1509void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1510{
1511}
1512
1513static inline void split_page_memcg(struct page *head, unsigned int nr)
1514{
1515}
1516
1517static inline
1518unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1519 gfp_t gfp_mask,
1520 unsigned long *total_scanned)
1521{
1522 return 0;
1523}
1524#endif /* CONFIG_MEMCG */
1525
1526static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1527{
1528 __mod_lruvec_kmem_state(p, idx, 1);
1529}
1530
1531static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1532{
1533 __mod_lruvec_kmem_state(p, idx, -1);
1534}
1535
1536static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1537{
1538 struct mem_cgroup *memcg;
1539
1540 memcg = lruvec_memcg(lruvec);
1541 if (!memcg)
1542 return NULL;
1543 memcg = parent_mem_cgroup(memcg);
1544 if (!memcg)
1545 return NULL;
1546 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1547}
1548
1549static inline void unlock_page_lruvec(struct lruvec *lruvec)
1550{
1551 spin_unlock(&lruvec->lru_lock);
1552}
1553
1554static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1555{
1556 spin_unlock_irq(&lruvec->lru_lock);
1557}
1558
1559static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1560 unsigned long flags)
1561{
1562 spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1563}
1564
1565/* Test requires a stable page->memcg binding, see page_memcg() */
1566static inline bool folio_matches_lruvec(struct folio *folio,
1567 struct lruvec *lruvec)
1568{
1569 return lruvec_pgdat(lruvec) == folio_pgdat(folio) &&
1570 lruvec_memcg(lruvec) == folio_memcg(folio);
1571}
1572
1573/* Don't lock again iff page's lruvec locked */
1574static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio,
1575 struct lruvec *locked_lruvec)
1576{
1577 if (locked_lruvec) {
1578 if (folio_matches_lruvec(folio, locked_lruvec))
1579 return locked_lruvec;
1580
1581 unlock_page_lruvec_irq(locked_lruvec);
1582 }
1583
1584 return folio_lruvec_lock_irq(folio);
1585}
1586
1587/* Don't lock again iff page's lruvec locked */
1588static inline struct lruvec *folio_lruvec_relock_irqsave(struct folio *folio,
1589 struct lruvec *locked_lruvec, unsigned long *flags)
1590{
1591 if (locked_lruvec) {
1592 if (folio_matches_lruvec(folio, locked_lruvec))
1593 return locked_lruvec;
1594
1595 unlock_page_lruvec_irqrestore(locked_lruvec, *flags);
1596 }
1597
1598 return folio_lruvec_lock_irqsave(folio, flags);
1599}
1600
1601#ifdef CONFIG_CGROUP_WRITEBACK
1602
1603struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1604void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1605 unsigned long *pheadroom, unsigned long *pdirty,
1606 unsigned long *pwriteback);
1607
1608void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio,
1609 struct bdi_writeback *wb);
1610
1611static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1612 struct bdi_writeback *wb)
1613{
1614 if (mem_cgroup_disabled())
1615 return;
1616
1617 if (unlikely(&folio_memcg(folio)->css != wb->memcg_css))
1618 mem_cgroup_track_foreign_dirty_slowpath(folio, wb);
1619}
1620
1621void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1622
1623#else /* CONFIG_CGROUP_WRITEBACK */
1624
1625static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1626{
1627 return NULL;
1628}
1629
1630static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1631 unsigned long *pfilepages,
1632 unsigned long *pheadroom,
1633 unsigned long *pdirty,
1634 unsigned long *pwriteback)
1635{
1636}
1637
1638static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1639 struct bdi_writeback *wb)
1640{
1641}
1642
1643static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1644{
1645}
1646
1647#endif /* CONFIG_CGROUP_WRITEBACK */
1648
1649struct sock;
1650bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages,
1651 gfp_t gfp_mask);
1652void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1653#ifdef CONFIG_MEMCG
1654extern struct static_key_false memcg_sockets_enabled_key;
1655#define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1656void mem_cgroup_sk_alloc(struct sock *sk);
1657void mem_cgroup_sk_free(struct sock *sk);
1658static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1659{
1660 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1661 return true;
1662 do {
1663 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure)))
1664 return true;
1665 } while ((memcg = parent_mem_cgroup(memcg)));
1666 return false;
1667}
1668
1669int alloc_shrinker_info(struct mem_cgroup *memcg);
1670void free_shrinker_info(struct mem_cgroup *memcg);
1671void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id);
1672void reparent_shrinker_deferred(struct mem_cgroup *memcg);
1673#else
1674#define mem_cgroup_sockets_enabled 0
1675static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1676static inline void mem_cgroup_sk_free(struct sock *sk) { };
1677static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1678{
1679 return false;
1680}
1681
1682static inline void set_shrinker_bit(struct mem_cgroup *memcg,
1683 int nid, int shrinker_id)
1684{
1685}
1686#endif
1687
1688#ifdef CONFIG_MEMCG_KMEM
1689bool mem_cgroup_kmem_disabled(void);
1690int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1691void __memcg_kmem_uncharge_page(struct page *page, int order);
1692
1693struct obj_cgroup *get_obj_cgroup_from_current(void);
1694
1695int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1696void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1697
1698extern struct static_key_false memcg_kmem_enabled_key;
1699
1700extern int memcg_nr_cache_ids;
1701void memcg_get_cache_ids(void);
1702void memcg_put_cache_ids(void);
1703
1704/*
1705 * Helper macro to loop through all memcg-specific caches. Callers must still
1706 * check if the cache is valid (it is either valid or NULL).
1707 * the slab_mutex must be held when looping through those caches
1708 */
1709#define for_each_memcg_cache_index(_idx) \
1710 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1711
1712static inline bool memcg_kmem_enabled(void)
1713{
1714 return static_branch_likely(&memcg_kmem_enabled_key);
1715}
1716
1717static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1718 int order)
1719{
1720 if (memcg_kmem_enabled())
1721 return __memcg_kmem_charge_page(page, gfp, order);
1722 return 0;
1723}
1724
1725static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1726{
1727 if (memcg_kmem_enabled())
1728 __memcg_kmem_uncharge_page(page, order);
1729}
1730
1731/*
1732 * A helper for accessing memcg's kmem_id, used for getting
1733 * corresponding LRU lists.
1734 */
1735static inline int memcg_cache_id(struct mem_cgroup *memcg)
1736{
1737 return memcg ? memcg->kmemcg_id : -1;
1738}
1739
1740struct mem_cgroup *mem_cgroup_from_obj(void *p);
1741
1742#else
1743static inline bool mem_cgroup_kmem_disabled(void)
1744{
1745 return true;
1746}
1747
1748static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1749 int order)
1750{
1751 return 0;
1752}
1753
1754static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1755{
1756}
1757
1758static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1759 int order)
1760{
1761 return 0;
1762}
1763
1764static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1765{
1766}
1767
1768#define for_each_memcg_cache_index(_idx) \
1769 for (; NULL; )
1770
1771static inline bool memcg_kmem_enabled(void)
1772{
1773 return false;
1774}
1775
1776static inline int memcg_cache_id(struct mem_cgroup *memcg)
1777{
1778 return -1;
1779}
1780
1781static inline void memcg_get_cache_ids(void)
1782{
1783}
1784
1785static inline void memcg_put_cache_ids(void)
1786{
1787}
1788
1789static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1790{
1791 return NULL;
1792}
1793
1794#endif /* CONFIG_MEMCG_KMEM */
1795
1796#endif /* _LINUX_MEMCONTROL_H */