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