xref: /linux-6.15/include/linux/memcontrol.h (revision 151f4e2b)
1 /* memcontrol.h - Memory Controller
2  *
3  * Copyright IBM Corporation, 2007
4  * Author Balbir Singh <[email protected]>
5  *
6  * Copyright 2007 OpenVZ SWsoft Inc
7  * Author: Pavel Emelianov <[email protected]>
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of the GNU General Public License as published by
11  * the Free Software Foundation; either version 2 of the License, or
12  * (at your option) any later version.
13  *
14  * This program is distributed in the hope that it will be useful,
15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17  * GNU General Public License for more details.
18  */
19 
20 #ifndef _LINUX_MEMCONTROL_H
21 #define _LINUX_MEMCONTROL_H
22 #include <linux/cgroup.h>
23 #include <linux/vm_event_item.h>
24 #include <linux/hardirq.h>
25 #include <linux/jump_label.h>
26 #include <linux/page_counter.h>
27 #include <linux/vmpressure.h>
28 #include <linux/eventfd.h>
29 #include <linux/mm.h>
30 #include <linux/vmstat.h>
31 #include <linux/writeback.h>
32 #include <linux/page-flags.h>
33 
34 struct mem_cgroup;
35 struct page;
36 struct mm_struct;
37 struct kmem_cache;
38 
39 /* Cgroup-specific page state, on top of universal node page state */
40 enum memcg_stat_item {
41 	MEMCG_CACHE = NR_VM_NODE_STAT_ITEMS,
42 	MEMCG_RSS,
43 	MEMCG_RSS_HUGE,
44 	MEMCG_SWAP,
45 	MEMCG_SOCK,
46 	/* XXX: why are these zone and not node counters? */
47 	MEMCG_KERNEL_STACK_KB,
48 	MEMCG_NR_STAT,
49 };
50 
51 enum memcg_memory_event {
52 	MEMCG_LOW,
53 	MEMCG_HIGH,
54 	MEMCG_MAX,
55 	MEMCG_OOM,
56 	MEMCG_OOM_KILL,
57 	MEMCG_SWAP_MAX,
58 	MEMCG_SWAP_FAIL,
59 	MEMCG_NR_MEMORY_EVENTS,
60 };
61 
62 enum mem_cgroup_protection {
63 	MEMCG_PROT_NONE,
64 	MEMCG_PROT_LOW,
65 	MEMCG_PROT_MIN,
66 };
67 
68 struct mem_cgroup_reclaim_cookie {
69 	pg_data_t *pgdat;
70 	int priority;
71 	unsigned int generation;
72 };
73 
74 #ifdef CONFIG_MEMCG
75 
76 #define MEM_CGROUP_ID_SHIFT	16
77 #define MEM_CGROUP_ID_MAX	USHRT_MAX
78 
79 struct mem_cgroup_id {
80 	int id;
81 	refcount_t ref;
82 };
83 
84 /*
85  * Per memcg event counter is incremented at every pagein/pageout. With THP,
86  * it will be incremated by the number of pages. This counter is used for
87  * for trigger some periodic events. This is straightforward and better
88  * than using jiffies etc. to handle periodic memcg event.
89  */
90 enum mem_cgroup_events_target {
91 	MEM_CGROUP_TARGET_THRESH,
92 	MEM_CGROUP_TARGET_SOFTLIMIT,
93 	MEM_CGROUP_TARGET_NUMAINFO,
94 	MEM_CGROUP_NTARGETS,
95 };
96 
97 struct memcg_vmstats_percpu {
98 	long stat[MEMCG_NR_STAT];
99 	unsigned long events[NR_VM_EVENT_ITEMS];
100 	unsigned long nr_page_events;
101 	unsigned long targets[MEM_CGROUP_NTARGETS];
102 };
103 
104 struct mem_cgroup_reclaim_iter {
105 	struct mem_cgroup *position;
106 	/* scan generation, increased every round-trip */
107 	unsigned int generation;
108 };
109 
110 struct lruvec_stat {
111 	long count[NR_VM_NODE_STAT_ITEMS];
112 };
113 
114 /*
115  * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
116  * which have elements charged to this memcg.
117  */
118 struct memcg_shrinker_map {
119 	struct rcu_head rcu;
120 	unsigned long map[0];
121 };
122 
123 /*
124  * per-zone information in memory controller.
125  */
126 struct mem_cgroup_per_node {
127 	struct lruvec		lruvec;
128 
129 	struct lruvec_stat __percpu *lruvec_stat_cpu;
130 	atomic_long_t		lruvec_stat[NR_VM_NODE_STAT_ITEMS];
131 	atomic_long_t		lruvec_stat_local[NR_VM_NODE_STAT_ITEMS];
132 
133 	unsigned long		lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
134 
135 	struct mem_cgroup_reclaim_iter	iter[DEF_PRIORITY + 1];
136 
137 #ifdef CONFIG_MEMCG_KMEM
138 	struct memcg_shrinker_map __rcu	*shrinker_map;
139 #endif
140 	struct rb_node		tree_node;	/* RB tree node */
141 	unsigned long		usage_in_excess;/* Set to the value by which */
142 						/* the soft limit is exceeded*/
143 	bool			on_tree;
144 	bool			congested;	/* memcg has many dirty pages */
145 						/* backed by a congested BDI */
146 
147 	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
148 						/* use container_of	   */
149 };
150 
151 struct mem_cgroup_threshold {
152 	struct eventfd_ctx *eventfd;
153 	unsigned long threshold;
154 };
155 
156 /* For threshold */
157 struct mem_cgroup_threshold_ary {
158 	/* An array index points to threshold just below or equal to usage. */
159 	int current_threshold;
160 	/* Size of entries[] */
161 	unsigned int size;
162 	/* Array of thresholds */
163 	struct mem_cgroup_threshold entries[0];
164 };
165 
166 struct mem_cgroup_thresholds {
167 	/* Primary thresholds array */
168 	struct mem_cgroup_threshold_ary *primary;
169 	/*
170 	 * Spare threshold array.
171 	 * This is needed to make mem_cgroup_unregister_event() "never fail".
172 	 * It must be able to store at least primary->size - 1 entries.
173 	 */
174 	struct mem_cgroup_threshold_ary *spare;
175 };
176 
177 enum memcg_kmem_state {
178 	KMEM_NONE,
179 	KMEM_ALLOCATED,
180 	KMEM_ONLINE,
181 };
182 
183 #if defined(CONFIG_SMP)
184 struct 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  * The memory controller data structure. The memory controller controls both
194  * page cache and RSS per cgroup. We would eventually like to provide
195  * statistics based on the statistics developed by Rik Van Riel for clock-pro,
196  * to help the administrator determine what knobs to tune.
197  */
198 struct mem_cgroup {
199 	struct cgroup_subsys_state css;
200 
201 	/* Private memcg ID. Used to ID objects that outlive the cgroup */
202 	struct mem_cgroup_id id;
203 
204 	/* Accounted resources */
205 	struct page_counter memory;
206 	struct page_counter swap;
207 
208 	/* Legacy consumer-oriented counters */
209 	struct page_counter memsw;
210 	struct page_counter kmem;
211 	struct page_counter tcpmem;
212 
213 	/* Upper bound of normal memory consumption range */
214 	unsigned long high;
215 
216 	/* Range enforcement for interrupt charges */
217 	struct work_struct high_work;
218 
219 	unsigned long soft_limit;
220 
221 	/* vmpressure notifications */
222 	struct vmpressure vmpressure;
223 
224 	/*
225 	 * Should the accounting and control be hierarchical, per subtree?
226 	 */
227 	bool use_hierarchy;
228 
229 	/*
230 	 * Should the OOM killer kill all belonging tasks, had it kill one?
231 	 */
232 	bool oom_group;
233 
234 	/* protected by memcg_oom_lock */
235 	bool		oom_lock;
236 	int		under_oom;
237 
238 	int	swappiness;
239 	/* OOM-Killer disable */
240 	int		oom_kill_disable;
241 
242 	/* memory.events */
243 	struct cgroup_file events_file;
244 
245 	/* handle for "memory.swap.events" */
246 	struct cgroup_file swap_events_file;
247 
248 	/* protect arrays of thresholds */
249 	struct mutex thresholds_lock;
250 
251 	/* thresholds for memory usage. RCU-protected */
252 	struct mem_cgroup_thresholds thresholds;
253 
254 	/* thresholds for mem+swap usage. RCU-protected */
255 	struct mem_cgroup_thresholds memsw_thresholds;
256 
257 	/* For oom notifier event fd */
258 	struct list_head oom_notify;
259 
260 	/*
261 	 * Should we move charges of a task when a task is moved into this
262 	 * mem_cgroup ? And what type of charges should we move ?
263 	 */
264 	unsigned long move_charge_at_immigrate;
265 	/* taken only while moving_account > 0 */
266 	spinlock_t		move_lock;
267 	unsigned long		move_lock_flags;
268 
269 	MEMCG_PADDING(_pad1_);
270 
271 	/*
272 	 * set > 0 if pages under this cgroup are moving to other cgroup.
273 	 */
274 	atomic_t		moving_account;
275 	struct task_struct	*move_lock_task;
276 
277 	/* memory.stat */
278 	struct memcg_vmstats_percpu __percpu *vmstats_percpu;
279 
280 	MEMCG_PADDING(_pad2_);
281 
282 	atomic_long_t		vmstats[MEMCG_NR_STAT];
283 	atomic_long_t		vmstats_local[MEMCG_NR_STAT];
284 
285 	atomic_long_t		vmevents[NR_VM_EVENT_ITEMS];
286 	atomic_long_t		vmevents_local[NR_VM_EVENT_ITEMS];
287 
288 	atomic_long_t		memory_events[MEMCG_NR_MEMORY_EVENTS];
289 
290 	unsigned long		socket_pressure;
291 
292 	/* Legacy tcp memory accounting */
293 	bool			tcpmem_active;
294 	int			tcpmem_pressure;
295 
296 #ifdef CONFIG_MEMCG_KMEM
297         /* Index in the kmem_cache->memcg_params.memcg_caches array */
298 	int kmemcg_id;
299 	enum memcg_kmem_state kmem_state;
300 	struct list_head kmem_caches;
301 #endif
302 
303 	int last_scanned_node;
304 #if MAX_NUMNODES > 1
305 	nodemask_t	scan_nodes;
306 	atomic_t	numainfo_events;
307 	atomic_t	numainfo_updating;
308 #endif
309 
310 #ifdef CONFIG_CGROUP_WRITEBACK
311 	struct list_head cgwb_list;
312 	struct wb_domain cgwb_domain;
313 #endif
314 
315 	/* List of events which userspace want to receive */
316 	struct list_head event_list;
317 	spinlock_t event_list_lock;
318 
319 	struct mem_cgroup_per_node *nodeinfo[0];
320 	/* WARNING: nodeinfo must be the last member here */
321 };
322 
323 /*
324  * size of first charge trial. "32" comes from vmscan.c's magic value.
325  * TODO: maybe necessary to use big numbers in big irons.
326  */
327 #define MEMCG_CHARGE_BATCH 32U
328 
329 extern struct mem_cgroup *root_mem_cgroup;
330 
331 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
332 {
333 	return (memcg == root_mem_cgroup);
334 }
335 
336 static inline bool mem_cgroup_disabled(void)
337 {
338 	return !cgroup_subsys_enabled(memory_cgrp_subsys);
339 }
340 
341 enum mem_cgroup_protection mem_cgroup_protected(struct mem_cgroup *root,
342 						struct mem_cgroup *memcg);
343 
344 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
345 			  gfp_t gfp_mask, struct mem_cgroup **memcgp,
346 			  bool compound);
347 int mem_cgroup_try_charge_delay(struct page *page, struct mm_struct *mm,
348 			  gfp_t gfp_mask, struct mem_cgroup **memcgp,
349 			  bool compound);
350 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
351 			      bool lrucare, bool compound);
352 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
353 		bool compound);
354 void mem_cgroup_uncharge(struct page *page);
355 void mem_cgroup_uncharge_list(struct list_head *page_list);
356 
357 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
358 
359 static struct mem_cgroup_per_node *
360 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
361 {
362 	return memcg->nodeinfo[nid];
363 }
364 
365 /**
366  * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
367  * @node: node of the wanted lruvec
368  * @memcg: memcg of the wanted lruvec
369  *
370  * Returns the lru list vector holding pages for a given @node or a given
371  * @memcg and @zone. This can be the node lruvec, if the memory controller
372  * is disabled.
373  */
374 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
375 				struct mem_cgroup *memcg)
376 {
377 	struct mem_cgroup_per_node *mz;
378 	struct lruvec *lruvec;
379 
380 	if (mem_cgroup_disabled()) {
381 		lruvec = node_lruvec(pgdat);
382 		goto out;
383 	}
384 
385 	mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
386 	lruvec = &mz->lruvec;
387 out:
388 	/*
389 	 * Since a node can be onlined after the mem_cgroup was created,
390 	 * we have to be prepared to initialize lruvec->pgdat here;
391 	 * and if offlined then reonlined, we need to reinitialize it.
392 	 */
393 	if (unlikely(lruvec->pgdat != pgdat))
394 		lruvec->pgdat = pgdat;
395 	return lruvec;
396 }
397 
398 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
399 
400 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
401 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
402 
403 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
404 
405 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
406 
407 static inline
408 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
409 	return css ? container_of(css, struct mem_cgroup, css) : NULL;
410 }
411 
412 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
413 {
414 	if (memcg)
415 		css_put(&memcg->css);
416 }
417 
418 #define mem_cgroup_from_counter(counter, member)	\
419 	container_of(counter, struct mem_cgroup, member)
420 
421 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
422 				   struct mem_cgroup *,
423 				   struct mem_cgroup_reclaim_cookie *);
424 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
425 int mem_cgroup_scan_tasks(struct mem_cgroup *,
426 			  int (*)(struct task_struct *, void *), void *);
427 
428 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
429 {
430 	if (mem_cgroup_disabled())
431 		return 0;
432 
433 	return memcg->id.id;
434 }
435 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
436 
437 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
438 {
439 	return mem_cgroup_from_css(seq_css(m));
440 }
441 
442 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
443 {
444 	struct mem_cgroup_per_node *mz;
445 
446 	if (mem_cgroup_disabled())
447 		return NULL;
448 
449 	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
450 	return mz->memcg;
451 }
452 
453 /**
454  * parent_mem_cgroup - find the accounting parent of a memcg
455  * @memcg: memcg whose parent to find
456  *
457  * Returns the parent memcg, or NULL if this is the root or the memory
458  * controller is in legacy no-hierarchy mode.
459  */
460 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
461 {
462 	if (!memcg->memory.parent)
463 		return NULL;
464 	return mem_cgroup_from_counter(memcg->memory.parent, memory);
465 }
466 
467 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
468 			      struct mem_cgroup *root)
469 {
470 	if (root == memcg)
471 		return true;
472 	if (!root->use_hierarchy)
473 		return false;
474 	return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
475 }
476 
477 static inline bool mm_match_cgroup(struct mm_struct *mm,
478 				   struct mem_cgroup *memcg)
479 {
480 	struct mem_cgroup *task_memcg;
481 	bool match = false;
482 
483 	rcu_read_lock();
484 	task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
485 	if (task_memcg)
486 		match = mem_cgroup_is_descendant(task_memcg, memcg);
487 	rcu_read_unlock();
488 	return match;
489 }
490 
491 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
492 ino_t page_cgroup_ino(struct page *page);
493 
494 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
495 {
496 	if (mem_cgroup_disabled())
497 		return true;
498 	return !!(memcg->css.flags & CSS_ONLINE);
499 }
500 
501 /*
502  * For memory reclaim.
503  */
504 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
505 
506 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
507 		int zid, int nr_pages);
508 
509 static inline
510 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
511 		enum lru_list lru, int zone_idx)
512 {
513 	struct mem_cgroup_per_node *mz;
514 
515 	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
516 	return mz->lru_zone_size[zone_idx][lru];
517 }
518 
519 void mem_cgroup_handle_over_high(void);
520 
521 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
522 
523 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
524 				struct task_struct *p);
525 
526 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
527 
528 static inline void mem_cgroup_enter_user_fault(void)
529 {
530 	WARN_ON(current->in_user_fault);
531 	current->in_user_fault = 1;
532 }
533 
534 static inline void mem_cgroup_exit_user_fault(void)
535 {
536 	WARN_ON(!current->in_user_fault);
537 	current->in_user_fault = 0;
538 }
539 
540 static inline bool task_in_memcg_oom(struct task_struct *p)
541 {
542 	return p->memcg_in_oom;
543 }
544 
545 bool mem_cgroup_oom_synchronize(bool wait);
546 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
547 					    struct mem_cgroup *oom_domain);
548 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
549 
550 #ifdef CONFIG_MEMCG_SWAP
551 extern int do_swap_account;
552 #endif
553 
554 struct mem_cgroup *lock_page_memcg(struct page *page);
555 void __unlock_page_memcg(struct mem_cgroup *memcg);
556 void unlock_page_memcg(struct page *page);
557 
558 /*
559  * idx can be of type enum memcg_stat_item or node_stat_item.
560  * Keep in sync with memcg_exact_page_state().
561  */
562 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
563 {
564 	long x = atomic_long_read(&memcg->vmstats[idx]);
565 #ifdef CONFIG_SMP
566 	if (x < 0)
567 		x = 0;
568 #endif
569 	return x;
570 }
571 
572 /*
573  * idx can be of type enum memcg_stat_item or node_stat_item.
574  * Keep in sync with memcg_exact_page_state().
575  */
576 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
577 						   int idx)
578 {
579 	long x = atomic_long_read(&memcg->vmstats_local[idx]);
580 #ifdef CONFIG_SMP
581 	if (x < 0)
582 		x = 0;
583 #endif
584 	return x;
585 }
586 
587 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
588 
589 /* idx can be of type enum memcg_stat_item or node_stat_item */
590 static inline void mod_memcg_state(struct mem_cgroup *memcg,
591 				   int idx, int val)
592 {
593 	unsigned long flags;
594 
595 	local_irq_save(flags);
596 	__mod_memcg_state(memcg, idx, val);
597 	local_irq_restore(flags);
598 }
599 
600 /**
601  * mod_memcg_page_state - update page state statistics
602  * @page: the page
603  * @idx: page state item to account
604  * @val: number of pages (positive or negative)
605  *
606  * The @page must be locked or the caller must use lock_page_memcg()
607  * to prevent double accounting when the page is concurrently being
608  * moved to another memcg:
609  *
610  *   lock_page(page) or lock_page_memcg(page)
611  *   if (TestClearPageState(page))
612  *     mod_memcg_page_state(page, state, -1);
613  *   unlock_page(page) or unlock_page_memcg(page)
614  *
615  * Kernel pages are an exception to this, since they'll never move.
616  */
617 static inline void __mod_memcg_page_state(struct page *page,
618 					  int idx, int val)
619 {
620 	if (page->mem_cgroup)
621 		__mod_memcg_state(page->mem_cgroup, idx, val);
622 }
623 
624 static inline void mod_memcg_page_state(struct page *page,
625 					int idx, int val)
626 {
627 	if (page->mem_cgroup)
628 		mod_memcg_state(page->mem_cgroup, idx, val);
629 }
630 
631 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
632 					      enum node_stat_item idx)
633 {
634 	struct mem_cgroup_per_node *pn;
635 	long x;
636 
637 	if (mem_cgroup_disabled())
638 		return node_page_state(lruvec_pgdat(lruvec), idx);
639 
640 	pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
641 	x = atomic_long_read(&pn->lruvec_stat[idx]);
642 #ifdef CONFIG_SMP
643 	if (x < 0)
644 		x = 0;
645 #endif
646 	return x;
647 }
648 
649 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
650 						    enum node_stat_item idx)
651 {
652 	struct mem_cgroup_per_node *pn;
653 	long x;
654 
655 	if (mem_cgroup_disabled())
656 		return node_page_state(lruvec_pgdat(lruvec), idx);
657 
658 	pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
659 	x = atomic_long_read(&pn->lruvec_stat_local[idx]);
660 #ifdef CONFIG_SMP
661 	if (x < 0)
662 		x = 0;
663 #endif
664 	return x;
665 }
666 
667 void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
668 			int val);
669 
670 static inline void mod_lruvec_state(struct lruvec *lruvec,
671 				    enum node_stat_item idx, int val)
672 {
673 	unsigned long flags;
674 
675 	local_irq_save(flags);
676 	__mod_lruvec_state(lruvec, idx, val);
677 	local_irq_restore(flags);
678 }
679 
680 static inline void __mod_lruvec_page_state(struct page *page,
681 					   enum node_stat_item idx, int val)
682 {
683 	pg_data_t *pgdat = page_pgdat(page);
684 	struct lruvec *lruvec;
685 
686 	/* Untracked pages have no memcg, no lruvec. Update only the node */
687 	if (!page->mem_cgroup) {
688 		__mod_node_page_state(pgdat, idx, val);
689 		return;
690 	}
691 
692 	lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
693 	__mod_lruvec_state(lruvec, idx, val);
694 }
695 
696 static inline void mod_lruvec_page_state(struct page *page,
697 					 enum node_stat_item idx, int val)
698 {
699 	unsigned long flags;
700 
701 	local_irq_save(flags);
702 	__mod_lruvec_page_state(page, idx, val);
703 	local_irq_restore(flags);
704 }
705 
706 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
707 						gfp_t gfp_mask,
708 						unsigned long *total_scanned);
709 
710 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
711 			  unsigned long count);
712 
713 static inline void count_memcg_events(struct mem_cgroup *memcg,
714 				      enum vm_event_item idx,
715 				      unsigned long count)
716 {
717 	unsigned long flags;
718 
719 	local_irq_save(flags);
720 	__count_memcg_events(memcg, idx, count);
721 	local_irq_restore(flags);
722 }
723 
724 static inline void count_memcg_page_event(struct page *page,
725 					  enum vm_event_item idx)
726 {
727 	if (page->mem_cgroup)
728 		count_memcg_events(page->mem_cgroup, idx, 1);
729 }
730 
731 static inline void count_memcg_event_mm(struct mm_struct *mm,
732 					enum vm_event_item idx)
733 {
734 	struct mem_cgroup *memcg;
735 
736 	if (mem_cgroup_disabled())
737 		return;
738 
739 	rcu_read_lock();
740 	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
741 	if (likely(memcg))
742 		count_memcg_events(memcg, idx, 1);
743 	rcu_read_unlock();
744 }
745 
746 static inline void memcg_memory_event(struct mem_cgroup *memcg,
747 				      enum memcg_memory_event event)
748 {
749 	atomic_long_inc(&memcg->memory_events[event]);
750 	cgroup_file_notify(&memcg->events_file);
751 }
752 
753 static inline void memcg_memory_event_mm(struct mm_struct *mm,
754 					 enum memcg_memory_event event)
755 {
756 	struct mem_cgroup *memcg;
757 
758 	if (mem_cgroup_disabled())
759 		return;
760 
761 	rcu_read_lock();
762 	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
763 	if (likely(memcg))
764 		memcg_memory_event(memcg, event);
765 	rcu_read_unlock();
766 }
767 
768 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
769 void mem_cgroup_split_huge_fixup(struct page *head);
770 #endif
771 
772 #else /* CONFIG_MEMCG */
773 
774 #define MEM_CGROUP_ID_SHIFT	0
775 #define MEM_CGROUP_ID_MAX	0
776 
777 struct mem_cgroup;
778 
779 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
780 {
781 	return true;
782 }
783 
784 static inline bool mem_cgroup_disabled(void)
785 {
786 	return true;
787 }
788 
789 static inline void memcg_memory_event(struct mem_cgroup *memcg,
790 				      enum memcg_memory_event event)
791 {
792 }
793 
794 static inline void memcg_memory_event_mm(struct mm_struct *mm,
795 					 enum memcg_memory_event event)
796 {
797 }
798 
799 static inline enum mem_cgroup_protection mem_cgroup_protected(
800 	struct mem_cgroup *root, struct mem_cgroup *memcg)
801 {
802 	return MEMCG_PROT_NONE;
803 }
804 
805 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
806 					gfp_t gfp_mask,
807 					struct mem_cgroup **memcgp,
808 					bool compound)
809 {
810 	*memcgp = NULL;
811 	return 0;
812 }
813 
814 static inline int mem_cgroup_try_charge_delay(struct page *page,
815 					      struct mm_struct *mm,
816 					      gfp_t gfp_mask,
817 					      struct mem_cgroup **memcgp,
818 					      bool compound)
819 {
820 	*memcgp = NULL;
821 	return 0;
822 }
823 
824 static inline void mem_cgroup_commit_charge(struct page *page,
825 					    struct mem_cgroup *memcg,
826 					    bool lrucare, bool compound)
827 {
828 }
829 
830 static inline void mem_cgroup_cancel_charge(struct page *page,
831 					    struct mem_cgroup *memcg,
832 					    bool compound)
833 {
834 }
835 
836 static inline void mem_cgroup_uncharge(struct page *page)
837 {
838 }
839 
840 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
841 {
842 }
843 
844 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
845 {
846 }
847 
848 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
849 				struct mem_cgroup *memcg)
850 {
851 	return node_lruvec(pgdat);
852 }
853 
854 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
855 						    struct pglist_data *pgdat)
856 {
857 	return &pgdat->lruvec;
858 }
859 
860 static inline bool mm_match_cgroup(struct mm_struct *mm,
861 		struct mem_cgroup *memcg)
862 {
863 	return true;
864 }
865 
866 static inline bool task_in_mem_cgroup(struct task_struct *task,
867 				      const struct mem_cgroup *memcg)
868 {
869 	return true;
870 }
871 
872 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
873 {
874 	return NULL;
875 }
876 
877 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
878 {
879 	return NULL;
880 }
881 
882 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
883 {
884 }
885 
886 static inline struct mem_cgroup *
887 mem_cgroup_iter(struct mem_cgroup *root,
888 		struct mem_cgroup *prev,
889 		struct mem_cgroup_reclaim_cookie *reclaim)
890 {
891 	return NULL;
892 }
893 
894 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
895 					 struct mem_cgroup *prev)
896 {
897 }
898 
899 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
900 		int (*fn)(struct task_struct *, void *), void *arg)
901 {
902 	return 0;
903 }
904 
905 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
906 {
907 	return 0;
908 }
909 
910 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
911 {
912 	WARN_ON_ONCE(id);
913 	/* XXX: This should always return root_mem_cgroup */
914 	return NULL;
915 }
916 
917 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
918 {
919 	return NULL;
920 }
921 
922 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
923 {
924 	return NULL;
925 }
926 
927 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
928 {
929 	return true;
930 }
931 
932 static inline
933 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
934 		enum lru_list lru, int zone_idx)
935 {
936 	return 0;
937 }
938 
939 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
940 {
941 	return 0;
942 }
943 
944 static inline void
945 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
946 {
947 }
948 
949 static inline void
950 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
951 {
952 }
953 
954 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
955 {
956 	return NULL;
957 }
958 
959 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
960 {
961 }
962 
963 static inline void unlock_page_memcg(struct page *page)
964 {
965 }
966 
967 static inline void mem_cgroup_handle_over_high(void)
968 {
969 }
970 
971 static inline void mem_cgroup_enter_user_fault(void)
972 {
973 }
974 
975 static inline void mem_cgroup_exit_user_fault(void)
976 {
977 }
978 
979 static inline bool task_in_memcg_oom(struct task_struct *p)
980 {
981 	return false;
982 }
983 
984 static inline bool mem_cgroup_oom_synchronize(bool wait)
985 {
986 	return false;
987 }
988 
989 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
990 	struct task_struct *victim, struct mem_cgroup *oom_domain)
991 {
992 	return NULL;
993 }
994 
995 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
996 {
997 }
998 
999 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1000 {
1001 	return 0;
1002 }
1003 
1004 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1005 						   int idx)
1006 {
1007 	return 0;
1008 }
1009 
1010 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1011 				     int idx,
1012 				     int nr)
1013 {
1014 }
1015 
1016 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1017 				   int idx,
1018 				   int nr)
1019 {
1020 }
1021 
1022 static inline void __mod_memcg_page_state(struct page *page,
1023 					  int idx,
1024 					  int nr)
1025 {
1026 }
1027 
1028 static inline void mod_memcg_page_state(struct page *page,
1029 					int idx,
1030 					int nr)
1031 {
1032 }
1033 
1034 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1035 					      enum node_stat_item idx)
1036 {
1037 	return node_page_state(lruvec_pgdat(lruvec), idx);
1038 }
1039 
1040 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1041 						    enum node_stat_item idx)
1042 {
1043 	return node_page_state(lruvec_pgdat(lruvec), idx);
1044 }
1045 
1046 static inline void __mod_lruvec_state(struct lruvec *lruvec,
1047 				      enum node_stat_item idx, int val)
1048 {
1049 	__mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1050 }
1051 
1052 static inline void mod_lruvec_state(struct lruvec *lruvec,
1053 				    enum node_stat_item idx, int val)
1054 {
1055 	mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
1056 }
1057 
1058 static inline void __mod_lruvec_page_state(struct page *page,
1059 					   enum node_stat_item idx, int val)
1060 {
1061 	__mod_node_page_state(page_pgdat(page), idx, val);
1062 }
1063 
1064 static inline void mod_lruvec_page_state(struct page *page,
1065 					 enum node_stat_item idx, int val)
1066 {
1067 	mod_node_page_state(page_pgdat(page), idx, val);
1068 }
1069 
1070 static inline
1071 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1072 					    gfp_t gfp_mask,
1073 					    unsigned long *total_scanned)
1074 {
1075 	return 0;
1076 }
1077 
1078 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1079 {
1080 }
1081 
1082 static inline void count_memcg_events(struct mem_cgroup *memcg,
1083 				      enum vm_event_item idx,
1084 				      unsigned long count)
1085 {
1086 }
1087 
1088 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1089 					enum vm_event_item idx,
1090 					unsigned long count)
1091 {
1092 }
1093 
1094 static inline void count_memcg_page_event(struct page *page,
1095 					  int idx)
1096 {
1097 }
1098 
1099 static inline
1100 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1101 {
1102 }
1103 #endif /* CONFIG_MEMCG */
1104 
1105 /* idx can be of type enum memcg_stat_item or node_stat_item */
1106 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
1107 				     int idx)
1108 {
1109 	__mod_memcg_state(memcg, idx, 1);
1110 }
1111 
1112 /* idx can be of type enum memcg_stat_item or node_stat_item */
1113 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
1114 				     int idx)
1115 {
1116 	__mod_memcg_state(memcg, idx, -1);
1117 }
1118 
1119 /* idx can be of type enum memcg_stat_item or node_stat_item */
1120 static inline void __inc_memcg_page_state(struct page *page,
1121 					  int idx)
1122 {
1123 	__mod_memcg_page_state(page, idx, 1);
1124 }
1125 
1126 /* idx can be of type enum memcg_stat_item or node_stat_item */
1127 static inline void __dec_memcg_page_state(struct page *page,
1128 					  int idx)
1129 {
1130 	__mod_memcg_page_state(page, idx, -1);
1131 }
1132 
1133 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1134 				      enum node_stat_item idx)
1135 {
1136 	__mod_lruvec_state(lruvec, idx, 1);
1137 }
1138 
1139 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1140 				      enum node_stat_item idx)
1141 {
1142 	__mod_lruvec_state(lruvec, idx, -1);
1143 }
1144 
1145 static inline void __inc_lruvec_page_state(struct page *page,
1146 					   enum node_stat_item idx)
1147 {
1148 	__mod_lruvec_page_state(page, idx, 1);
1149 }
1150 
1151 static inline void __dec_lruvec_page_state(struct page *page,
1152 					   enum node_stat_item idx)
1153 {
1154 	__mod_lruvec_page_state(page, idx, -1);
1155 }
1156 
1157 /* idx can be of type enum memcg_stat_item or node_stat_item */
1158 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1159 				   int idx)
1160 {
1161 	mod_memcg_state(memcg, idx, 1);
1162 }
1163 
1164 /* idx can be of type enum memcg_stat_item or node_stat_item */
1165 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1166 				   int idx)
1167 {
1168 	mod_memcg_state(memcg, idx, -1);
1169 }
1170 
1171 /* idx can be of type enum memcg_stat_item or node_stat_item */
1172 static inline void inc_memcg_page_state(struct page *page,
1173 					int idx)
1174 {
1175 	mod_memcg_page_state(page, idx, 1);
1176 }
1177 
1178 /* idx can be of type enum memcg_stat_item or node_stat_item */
1179 static inline void dec_memcg_page_state(struct page *page,
1180 					int idx)
1181 {
1182 	mod_memcg_page_state(page, idx, -1);
1183 }
1184 
1185 static inline void inc_lruvec_state(struct lruvec *lruvec,
1186 				    enum node_stat_item idx)
1187 {
1188 	mod_lruvec_state(lruvec, idx, 1);
1189 }
1190 
1191 static inline void dec_lruvec_state(struct lruvec *lruvec,
1192 				    enum node_stat_item idx)
1193 {
1194 	mod_lruvec_state(lruvec, idx, -1);
1195 }
1196 
1197 static inline void inc_lruvec_page_state(struct page *page,
1198 					 enum node_stat_item idx)
1199 {
1200 	mod_lruvec_page_state(page, idx, 1);
1201 }
1202 
1203 static inline void dec_lruvec_page_state(struct page *page,
1204 					 enum node_stat_item idx)
1205 {
1206 	mod_lruvec_page_state(page, idx, -1);
1207 }
1208 
1209 #ifdef CONFIG_CGROUP_WRITEBACK
1210 
1211 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1212 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1213 			 unsigned long *pheadroom, unsigned long *pdirty,
1214 			 unsigned long *pwriteback);
1215 
1216 #else	/* CONFIG_CGROUP_WRITEBACK */
1217 
1218 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1219 {
1220 	return NULL;
1221 }
1222 
1223 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1224 				       unsigned long *pfilepages,
1225 				       unsigned long *pheadroom,
1226 				       unsigned long *pdirty,
1227 				       unsigned long *pwriteback)
1228 {
1229 }
1230 
1231 #endif	/* CONFIG_CGROUP_WRITEBACK */
1232 
1233 struct sock;
1234 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1235 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1236 #ifdef CONFIG_MEMCG
1237 extern struct static_key_false memcg_sockets_enabled_key;
1238 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1239 void mem_cgroup_sk_alloc(struct sock *sk);
1240 void mem_cgroup_sk_free(struct sock *sk);
1241 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1242 {
1243 	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1244 		return true;
1245 	do {
1246 		if (time_before(jiffies, memcg->socket_pressure))
1247 			return true;
1248 	} while ((memcg = parent_mem_cgroup(memcg)));
1249 	return false;
1250 }
1251 #else
1252 #define mem_cgroup_sockets_enabled 0
1253 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1254 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1255 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1256 {
1257 	return false;
1258 }
1259 #endif
1260 
1261 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1262 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1263 
1264 #ifdef CONFIG_MEMCG_KMEM
1265 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1266 void __memcg_kmem_uncharge(struct page *page, int order);
1267 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1268 			      struct mem_cgroup *memcg);
1269 
1270 extern struct static_key_false memcg_kmem_enabled_key;
1271 extern struct workqueue_struct *memcg_kmem_cache_wq;
1272 
1273 extern int memcg_nr_cache_ids;
1274 void memcg_get_cache_ids(void);
1275 void memcg_put_cache_ids(void);
1276 
1277 /*
1278  * Helper macro to loop through all memcg-specific caches. Callers must still
1279  * check if the cache is valid (it is either valid or NULL).
1280  * the slab_mutex must be held when looping through those caches
1281  */
1282 #define for_each_memcg_cache_index(_idx)	\
1283 	for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1284 
1285 static inline bool memcg_kmem_enabled(void)
1286 {
1287 	return static_branch_unlikely(&memcg_kmem_enabled_key);
1288 }
1289 
1290 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1291 {
1292 	if (memcg_kmem_enabled())
1293 		return __memcg_kmem_charge(page, gfp, order);
1294 	return 0;
1295 }
1296 
1297 static inline void memcg_kmem_uncharge(struct page *page, int order)
1298 {
1299 	if (memcg_kmem_enabled())
1300 		__memcg_kmem_uncharge(page, order);
1301 }
1302 
1303 static inline int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp,
1304 					  int order, struct mem_cgroup *memcg)
1305 {
1306 	if (memcg_kmem_enabled())
1307 		return __memcg_kmem_charge_memcg(page, gfp, order, memcg);
1308 	return 0;
1309 }
1310 /*
1311  * helper for accessing a memcg's index. It will be used as an index in the
1312  * child cache array in kmem_cache, and also to derive its name. This function
1313  * will return -1 when this is not a kmem-limited memcg.
1314  */
1315 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1316 {
1317 	return memcg ? memcg->kmemcg_id : -1;
1318 }
1319 
1320 extern int memcg_expand_shrinker_maps(int new_id);
1321 
1322 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1323 				   int nid, int shrinker_id);
1324 #else
1325 
1326 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1327 {
1328 	return 0;
1329 }
1330 
1331 static inline void memcg_kmem_uncharge(struct page *page, int order)
1332 {
1333 }
1334 
1335 static inline int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
1336 {
1337 	return 0;
1338 }
1339 
1340 static inline void __memcg_kmem_uncharge(struct page *page, int order)
1341 {
1342 }
1343 
1344 #define for_each_memcg_cache_index(_idx)	\
1345 	for (; NULL; )
1346 
1347 static inline bool memcg_kmem_enabled(void)
1348 {
1349 	return false;
1350 }
1351 
1352 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1353 {
1354 	return -1;
1355 }
1356 
1357 static inline void memcg_get_cache_ids(void)
1358 {
1359 }
1360 
1361 static inline void memcg_put_cache_ids(void)
1362 {
1363 }
1364 
1365 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1366 					  int nid, int shrinker_id) { }
1367 #endif /* CONFIG_MEMCG_KMEM */
1368 
1369 #endif /* _LINUX_MEMCONTROL_H */
1370