xref: /linux-6.15/include/linux/memcontrol.h (revision ecdf06e1)
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 /* Cgroup-specific events, on top of universal VM events */
52 enum memcg_event_item {
53 	MEMCG_LOW = NR_VM_EVENT_ITEMS,
54 	MEMCG_HIGH,
55 	MEMCG_MAX,
56 	MEMCG_OOM,
57 	MEMCG_NR_EVENTS,
58 };
59 
60 struct mem_cgroup_reclaim_cookie {
61 	pg_data_t *pgdat;
62 	int priority;
63 	unsigned int generation;
64 };
65 
66 #ifdef CONFIG_MEMCG
67 
68 #define MEM_CGROUP_ID_SHIFT	16
69 #define MEM_CGROUP_ID_MAX	USHRT_MAX
70 
71 struct mem_cgroup_id {
72 	int id;
73 	atomic_t ref;
74 };
75 
76 /*
77  * Per memcg event counter is incremented at every pagein/pageout. With THP,
78  * it will be incremated by the number of pages. This counter is used for
79  * for trigger some periodic events. This is straightforward and better
80  * than using jiffies etc. to handle periodic memcg event.
81  */
82 enum mem_cgroup_events_target {
83 	MEM_CGROUP_TARGET_THRESH,
84 	MEM_CGROUP_TARGET_SOFTLIMIT,
85 	MEM_CGROUP_TARGET_NUMAINFO,
86 	MEM_CGROUP_NTARGETS,
87 };
88 
89 struct mem_cgroup_stat_cpu {
90 	long count[MEMCG_NR_STAT];
91 	unsigned long events[MEMCG_NR_EVENTS];
92 	unsigned long nr_page_events;
93 	unsigned long targets[MEM_CGROUP_NTARGETS];
94 };
95 
96 struct mem_cgroup_reclaim_iter {
97 	struct mem_cgroup *position;
98 	/* scan generation, increased every round-trip */
99 	unsigned int generation;
100 };
101 
102 struct lruvec_stat {
103 	long count[NR_VM_NODE_STAT_ITEMS];
104 };
105 
106 /*
107  * per-zone information in memory controller.
108  */
109 struct mem_cgroup_per_node {
110 	struct lruvec		lruvec;
111 
112 	struct lruvec_stat __percpu *lruvec_stat_cpu;
113 	atomic_long_t		lruvec_stat[NR_VM_NODE_STAT_ITEMS];
114 
115 	unsigned long		lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
116 
117 	struct mem_cgroup_reclaim_iter	iter[DEF_PRIORITY + 1];
118 
119 	struct rb_node		tree_node;	/* RB tree node */
120 	unsigned long		usage_in_excess;/* Set to the value by which */
121 						/* the soft limit is exceeded*/
122 	bool			on_tree;
123 	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
124 						/* use container_of	   */
125 };
126 
127 struct mem_cgroup_threshold {
128 	struct eventfd_ctx *eventfd;
129 	unsigned long threshold;
130 };
131 
132 /* For threshold */
133 struct mem_cgroup_threshold_ary {
134 	/* An array index points to threshold just below or equal to usage. */
135 	int current_threshold;
136 	/* Size of entries[] */
137 	unsigned int size;
138 	/* Array of thresholds */
139 	struct mem_cgroup_threshold entries[0];
140 };
141 
142 struct mem_cgroup_thresholds {
143 	/* Primary thresholds array */
144 	struct mem_cgroup_threshold_ary *primary;
145 	/*
146 	 * Spare threshold array.
147 	 * This is needed to make mem_cgroup_unregister_event() "never fail".
148 	 * It must be able to store at least primary->size - 1 entries.
149 	 */
150 	struct mem_cgroup_threshold_ary *spare;
151 };
152 
153 enum memcg_kmem_state {
154 	KMEM_NONE,
155 	KMEM_ALLOCATED,
156 	KMEM_ONLINE,
157 };
158 
159 /*
160  * The memory controller data structure. The memory controller controls both
161  * page cache and RSS per cgroup. We would eventually like to provide
162  * statistics based on the statistics developed by Rik Van Riel for clock-pro,
163  * to help the administrator determine what knobs to tune.
164  */
165 struct mem_cgroup {
166 	struct cgroup_subsys_state css;
167 
168 	/* Private memcg ID. Used to ID objects that outlive the cgroup */
169 	struct mem_cgroup_id id;
170 
171 	/* Accounted resources */
172 	struct page_counter memory;
173 	struct page_counter swap;
174 
175 	/* Legacy consumer-oriented counters */
176 	struct page_counter memsw;
177 	struct page_counter kmem;
178 	struct page_counter tcpmem;
179 
180 	/* Normal memory consumption range */
181 	unsigned long low;
182 	unsigned long high;
183 
184 	/* Range enforcement for interrupt charges */
185 	struct work_struct high_work;
186 
187 	unsigned long soft_limit;
188 
189 	/* vmpressure notifications */
190 	struct vmpressure vmpressure;
191 
192 	/*
193 	 * Should the accounting and control be hierarchical, per subtree?
194 	 */
195 	bool use_hierarchy;
196 
197 	/* protected by memcg_oom_lock */
198 	bool		oom_lock;
199 	int		under_oom;
200 
201 	int	swappiness;
202 	/* OOM-Killer disable */
203 	int		oom_kill_disable;
204 
205 	/* handle for "memory.events" */
206 	struct cgroup_file events_file;
207 
208 	/* protect arrays of thresholds */
209 	struct mutex thresholds_lock;
210 
211 	/* thresholds for memory usage. RCU-protected */
212 	struct mem_cgroup_thresholds thresholds;
213 
214 	/* thresholds for mem+swap usage. RCU-protected */
215 	struct mem_cgroup_thresholds memsw_thresholds;
216 
217 	/* For oom notifier event fd */
218 	struct list_head oom_notify;
219 
220 	/*
221 	 * Should we move charges of a task when a task is moved into this
222 	 * mem_cgroup ? And what type of charges should we move ?
223 	 */
224 	unsigned long move_charge_at_immigrate;
225 	/*
226 	 * set > 0 if pages under this cgroup are moving to other cgroup.
227 	 */
228 	atomic_t		moving_account;
229 	/* taken only while moving_account > 0 */
230 	spinlock_t		move_lock;
231 	struct task_struct	*move_lock_task;
232 	unsigned long		move_lock_flags;
233 
234 	struct mem_cgroup_stat_cpu __percpu *stat_cpu;
235 	atomic_long_t		stat[MEMCG_NR_STAT];
236 	atomic_long_t		events[MEMCG_NR_EVENTS];
237 
238 	unsigned long		socket_pressure;
239 
240 	/* Legacy tcp memory accounting */
241 	bool			tcpmem_active;
242 	int			tcpmem_pressure;
243 
244 #ifndef CONFIG_SLOB
245         /* Index in the kmem_cache->memcg_params.memcg_caches array */
246 	int kmemcg_id;
247 	enum memcg_kmem_state kmem_state;
248 	struct list_head kmem_caches;
249 #endif
250 
251 	int last_scanned_node;
252 #if MAX_NUMNODES > 1
253 	nodemask_t	scan_nodes;
254 	atomic_t	numainfo_events;
255 	atomic_t	numainfo_updating;
256 #endif
257 
258 #ifdef CONFIG_CGROUP_WRITEBACK
259 	struct list_head cgwb_list;
260 	struct wb_domain cgwb_domain;
261 #endif
262 
263 	/* List of events which userspace want to receive */
264 	struct list_head event_list;
265 	spinlock_t event_list_lock;
266 
267 	struct mem_cgroup_per_node *nodeinfo[0];
268 	/* WARNING: nodeinfo must be the last member here */
269 };
270 
271 /*
272  * size of first charge trial. "32" comes from vmscan.c's magic value.
273  * TODO: maybe necessary to use big numbers in big irons.
274  */
275 #define MEMCG_CHARGE_BATCH 32U
276 
277 extern struct mem_cgroup *root_mem_cgroup;
278 
279 static inline bool mem_cgroup_disabled(void)
280 {
281 	return !cgroup_subsys_enabled(memory_cgrp_subsys);
282 }
283 
284 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
285 
286 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
287 			  gfp_t gfp_mask, struct mem_cgroup **memcgp,
288 			  bool compound);
289 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
290 			      bool lrucare, bool compound);
291 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
292 		bool compound);
293 void mem_cgroup_uncharge(struct page *page);
294 void mem_cgroup_uncharge_list(struct list_head *page_list);
295 
296 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
297 
298 static struct mem_cgroup_per_node *
299 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
300 {
301 	return memcg->nodeinfo[nid];
302 }
303 
304 /**
305  * mem_cgroup_lruvec - get the lru list vector for a node or a memcg zone
306  * @node: node of the wanted lruvec
307  * @memcg: memcg of the wanted lruvec
308  *
309  * Returns the lru list vector holding pages for a given @node or a given
310  * @memcg and @zone. This can be the node lruvec, if the memory controller
311  * is disabled.
312  */
313 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
314 				struct mem_cgroup *memcg)
315 {
316 	struct mem_cgroup_per_node *mz;
317 	struct lruvec *lruvec;
318 
319 	if (mem_cgroup_disabled()) {
320 		lruvec = node_lruvec(pgdat);
321 		goto out;
322 	}
323 
324 	mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
325 	lruvec = &mz->lruvec;
326 out:
327 	/*
328 	 * Since a node can be onlined after the mem_cgroup was created,
329 	 * we have to be prepared to initialize lruvec->pgdat here;
330 	 * and if offlined then reonlined, we need to reinitialize it.
331 	 */
332 	if (unlikely(lruvec->pgdat != pgdat))
333 		lruvec->pgdat = pgdat;
334 	return lruvec;
335 }
336 
337 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct pglist_data *);
338 
339 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
340 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
341 
342 static inline
343 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
344 	return css ? container_of(css, struct mem_cgroup, css) : NULL;
345 }
346 
347 #define mem_cgroup_from_counter(counter, member)	\
348 	container_of(counter, struct mem_cgroup, member)
349 
350 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
351 				   struct mem_cgroup *,
352 				   struct mem_cgroup_reclaim_cookie *);
353 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
354 int mem_cgroup_scan_tasks(struct mem_cgroup *,
355 			  int (*)(struct task_struct *, void *), void *);
356 
357 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
358 {
359 	if (mem_cgroup_disabled())
360 		return 0;
361 
362 	return memcg->id.id;
363 }
364 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
365 
366 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
367 {
368 	struct mem_cgroup_per_node *mz;
369 
370 	if (mem_cgroup_disabled())
371 		return NULL;
372 
373 	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
374 	return mz->memcg;
375 }
376 
377 /**
378  * parent_mem_cgroup - find the accounting parent of a memcg
379  * @memcg: memcg whose parent to find
380  *
381  * Returns the parent memcg, or NULL if this is the root or the memory
382  * controller is in legacy no-hierarchy mode.
383  */
384 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
385 {
386 	if (!memcg->memory.parent)
387 		return NULL;
388 	return mem_cgroup_from_counter(memcg->memory.parent, memory);
389 }
390 
391 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
392 			      struct mem_cgroup *root)
393 {
394 	if (root == memcg)
395 		return true;
396 	if (!root->use_hierarchy)
397 		return false;
398 	return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
399 }
400 
401 static inline bool mm_match_cgroup(struct mm_struct *mm,
402 				   struct mem_cgroup *memcg)
403 {
404 	struct mem_cgroup *task_memcg;
405 	bool match = false;
406 
407 	rcu_read_lock();
408 	task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
409 	if (task_memcg)
410 		match = mem_cgroup_is_descendant(task_memcg, memcg);
411 	rcu_read_unlock();
412 	return match;
413 }
414 
415 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
416 ino_t page_cgroup_ino(struct page *page);
417 
418 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
419 {
420 	if (mem_cgroup_disabled())
421 		return true;
422 	return !!(memcg->css.flags & CSS_ONLINE);
423 }
424 
425 /*
426  * For memory reclaim.
427  */
428 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
429 
430 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
431 		int zid, int nr_pages);
432 
433 unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
434 					   int nid, unsigned int lru_mask);
435 
436 static inline
437 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
438 {
439 	struct mem_cgroup_per_node *mz;
440 	unsigned long nr_pages = 0;
441 	int zid;
442 
443 	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
444 	for (zid = 0; zid < MAX_NR_ZONES; zid++)
445 		nr_pages += mz->lru_zone_size[zid][lru];
446 	return nr_pages;
447 }
448 
449 static inline
450 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
451 		enum lru_list lru, int zone_idx)
452 {
453 	struct mem_cgroup_per_node *mz;
454 
455 	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
456 	return mz->lru_zone_size[zone_idx][lru];
457 }
458 
459 void mem_cgroup_handle_over_high(void);
460 
461 unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg);
462 
463 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
464 				struct task_struct *p);
465 
466 static inline void mem_cgroup_oom_enable(void)
467 {
468 	WARN_ON(current->memcg_may_oom);
469 	current->memcg_may_oom = 1;
470 }
471 
472 static inline void mem_cgroup_oom_disable(void)
473 {
474 	WARN_ON(!current->memcg_may_oom);
475 	current->memcg_may_oom = 0;
476 }
477 
478 static inline bool task_in_memcg_oom(struct task_struct *p)
479 {
480 	return p->memcg_in_oom;
481 }
482 
483 bool mem_cgroup_oom_synchronize(bool wait);
484 
485 #ifdef CONFIG_MEMCG_SWAP
486 extern int do_swap_account;
487 #endif
488 
489 struct mem_cgroup *lock_page_memcg(struct page *page);
490 void __unlock_page_memcg(struct mem_cgroup *memcg);
491 void unlock_page_memcg(struct page *page);
492 
493 /* idx can be of type enum memcg_stat_item or node_stat_item */
494 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
495 					     int idx)
496 {
497 	long x = atomic_long_read(&memcg->stat[idx]);
498 #ifdef CONFIG_SMP
499 	if (x < 0)
500 		x = 0;
501 #endif
502 	return x;
503 }
504 
505 /* idx can be of type enum memcg_stat_item or node_stat_item */
506 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
507 				     int idx, int val)
508 {
509 	long x;
510 
511 	if (mem_cgroup_disabled())
512 		return;
513 
514 	x = val + __this_cpu_read(memcg->stat_cpu->count[idx]);
515 	if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
516 		atomic_long_add(x, &memcg->stat[idx]);
517 		x = 0;
518 	}
519 	__this_cpu_write(memcg->stat_cpu->count[idx], x);
520 }
521 
522 /* idx can be of type enum memcg_stat_item or node_stat_item */
523 static inline void mod_memcg_state(struct mem_cgroup *memcg,
524 				   int idx, int val)
525 {
526 	preempt_disable();
527 	__mod_memcg_state(memcg, idx, val);
528 	preempt_enable();
529 }
530 
531 /**
532  * mod_memcg_page_state - update page state statistics
533  * @page: the page
534  * @idx: page state item to account
535  * @val: number of pages (positive or negative)
536  *
537  * The @page must be locked or the caller must use lock_page_memcg()
538  * to prevent double accounting when the page is concurrently being
539  * moved to another memcg:
540  *
541  *   lock_page(page) or lock_page_memcg(page)
542  *   if (TestClearPageState(page))
543  *     mod_memcg_page_state(page, state, -1);
544  *   unlock_page(page) or unlock_page_memcg(page)
545  *
546  * Kernel pages are an exception to this, since they'll never move.
547  */
548 static inline void __mod_memcg_page_state(struct page *page,
549 					  int idx, int val)
550 {
551 	if (page->mem_cgroup)
552 		__mod_memcg_state(page->mem_cgroup, idx, val);
553 }
554 
555 static inline void mod_memcg_page_state(struct page *page,
556 					int idx, int val)
557 {
558 	if (page->mem_cgroup)
559 		mod_memcg_state(page->mem_cgroup, idx, val);
560 }
561 
562 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
563 					      enum node_stat_item idx)
564 {
565 	struct mem_cgroup_per_node *pn;
566 	long x;
567 
568 	if (mem_cgroup_disabled())
569 		return node_page_state(lruvec_pgdat(lruvec), idx);
570 
571 	pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
572 	x = atomic_long_read(&pn->lruvec_stat[idx]);
573 #ifdef CONFIG_SMP
574 	if (x < 0)
575 		x = 0;
576 #endif
577 	return x;
578 }
579 
580 static inline void __mod_lruvec_state(struct lruvec *lruvec,
581 				      enum node_stat_item idx, int val)
582 {
583 	struct mem_cgroup_per_node *pn;
584 	long x;
585 
586 	/* Update node */
587 	__mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
588 
589 	if (mem_cgroup_disabled())
590 		return;
591 
592 	pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
593 
594 	/* Update memcg */
595 	__mod_memcg_state(pn->memcg, idx, val);
596 
597 	/* Update lruvec */
598 	x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
599 	if (unlikely(abs(x) > MEMCG_CHARGE_BATCH)) {
600 		atomic_long_add(x, &pn->lruvec_stat[idx]);
601 		x = 0;
602 	}
603 	__this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
604 }
605 
606 static inline void mod_lruvec_state(struct lruvec *lruvec,
607 				    enum node_stat_item idx, int val)
608 {
609 	preempt_disable();
610 	__mod_lruvec_state(lruvec, idx, val);
611 	preempt_enable();
612 }
613 
614 static inline void __mod_lruvec_page_state(struct page *page,
615 					   enum node_stat_item idx, int val)
616 {
617 	pg_data_t *pgdat = page_pgdat(page);
618 	struct lruvec *lruvec;
619 
620 	/* Untracked pages have no memcg, no lruvec. Update only the node */
621 	if (!page->mem_cgroup) {
622 		__mod_node_page_state(pgdat, idx, val);
623 		return;
624 	}
625 
626 	lruvec = mem_cgroup_lruvec(pgdat, page->mem_cgroup);
627 	__mod_lruvec_state(lruvec, idx, val);
628 }
629 
630 static inline void mod_lruvec_page_state(struct page *page,
631 					 enum node_stat_item idx, int val)
632 {
633 	preempt_disable();
634 	__mod_lruvec_page_state(page, idx, val);
635 	preempt_enable();
636 }
637 
638 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
639 						gfp_t gfp_mask,
640 						unsigned long *total_scanned);
641 
642 /* idx can be of type enum memcg_event_item or vm_event_item */
643 static inline void __count_memcg_events(struct mem_cgroup *memcg,
644 					int idx, unsigned long count)
645 {
646 	unsigned long x;
647 
648 	if (mem_cgroup_disabled())
649 		return;
650 
651 	x = count + __this_cpu_read(memcg->stat_cpu->events[idx]);
652 	if (unlikely(x > MEMCG_CHARGE_BATCH)) {
653 		atomic_long_add(x, &memcg->events[idx]);
654 		x = 0;
655 	}
656 	__this_cpu_write(memcg->stat_cpu->events[idx], x);
657 }
658 
659 static inline void count_memcg_events(struct mem_cgroup *memcg,
660 				      int idx, unsigned long count)
661 {
662 	preempt_disable();
663 	__count_memcg_events(memcg, idx, count);
664 	preempt_enable();
665 }
666 
667 /* idx can be of type enum memcg_event_item or vm_event_item */
668 static inline void count_memcg_page_event(struct page *page,
669 					  int idx)
670 {
671 	if (page->mem_cgroup)
672 		count_memcg_events(page->mem_cgroup, idx, 1);
673 }
674 
675 static inline void count_memcg_event_mm(struct mm_struct *mm,
676 					enum vm_event_item idx)
677 {
678 	struct mem_cgroup *memcg;
679 
680 	if (mem_cgroup_disabled())
681 		return;
682 
683 	rcu_read_lock();
684 	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
685 	if (likely(memcg)) {
686 		count_memcg_events(memcg, idx, 1);
687 		if (idx == OOM_KILL)
688 			cgroup_file_notify(&memcg->events_file);
689 	}
690 	rcu_read_unlock();
691 }
692 
693 static inline void mem_cgroup_event(struct mem_cgroup *memcg,
694 				    enum memcg_event_item event)
695 {
696 	count_memcg_events(memcg, event, 1);
697 	cgroup_file_notify(&memcg->events_file);
698 }
699 
700 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
701 void mem_cgroup_split_huge_fixup(struct page *head);
702 #endif
703 
704 #else /* CONFIG_MEMCG */
705 
706 #define MEM_CGROUP_ID_SHIFT	0
707 #define MEM_CGROUP_ID_MAX	0
708 
709 struct mem_cgroup;
710 
711 static inline bool mem_cgroup_disabled(void)
712 {
713 	return true;
714 }
715 
716 static inline void mem_cgroup_event(struct mem_cgroup *memcg,
717 				    enum memcg_event_item event)
718 {
719 }
720 
721 static inline bool mem_cgroup_low(struct mem_cgroup *root,
722 				  struct mem_cgroup *memcg)
723 {
724 	return false;
725 }
726 
727 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
728 					gfp_t gfp_mask,
729 					struct mem_cgroup **memcgp,
730 					bool compound)
731 {
732 	*memcgp = NULL;
733 	return 0;
734 }
735 
736 static inline void mem_cgroup_commit_charge(struct page *page,
737 					    struct mem_cgroup *memcg,
738 					    bool lrucare, bool compound)
739 {
740 }
741 
742 static inline void mem_cgroup_cancel_charge(struct page *page,
743 					    struct mem_cgroup *memcg,
744 					    bool compound)
745 {
746 }
747 
748 static inline void mem_cgroup_uncharge(struct page *page)
749 {
750 }
751 
752 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
753 {
754 }
755 
756 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
757 {
758 }
759 
760 static inline struct lruvec *mem_cgroup_lruvec(struct pglist_data *pgdat,
761 				struct mem_cgroup *memcg)
762 {
763 	return node_lruvec(pgdat);
764 }
765 
766 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
767 						    struct pglist_data *pgdat)
768 {
769 	return &pgdat->lruvec;
770 }
771 
772 static inline bool mm_match_cgroup(struct mm_struct *mm,
773 		struct mem_cgroup *memcg)
774 {
775 	return true;
776 }
777 
778 static inline bool task_in_mem_cgroup(struct task_struct *task,
779 				      const struct mem_cgroup *memcg)
780 {
781 	return true;
782 }
783 
784 static inline struct mem_cgroup *
785 mem_cgroup_iter(struct mem_cgroup *root,
786 		struct mem_cgroup *prev,
787 		struct mem_cgroup_reclaim_cookie *reclaim)
788 {
789 	return NULL;
790 }
791 
792 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
793 					 struct mem_cgroup *prev)
794 {
795 }
796 
797 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
798 		int (*fn)(struct task_struct *, void *), void *arg)
799 {
800 	return 0;
801 }
802 
803 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
804 {
805 	return 0;
806 }
807 
808 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
809 {
810 	WARN_ON_ONCE(id);
811 	/* XXX: This should always return root_mem_cgroup */
812 	return NULL;
813 }
814 
815 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
816 {
817 	return NULL;
818 }
819 
820 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
821 {
822 	return true;
823 }
824 
825 static inline unsigned long
826 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
827 {
828 	return 0;
829 }
830 static inline
831 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
832 		enum lru_list lru, int zone_idx)
833 {
834 	return 0;
835 }
836 
837 static inline unsigned long
838 mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
839 			     int nid, unsigned int lru_mask)
840 {
841 	return 0;
842 }
843 
844 static inline unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
845 {
846 	return 0;
847 }
848 
849 static inline void
850 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
851 {
852 }
853 
854 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
855 {
856 	return NULL;
857 }
858 
859 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
860 {
861 }
862 
863 static inline void unlock_page_memcg(struct page *page)
864 {
865 }
866 
867 static inline void mem_cgroup_handle_over_high(void)
868 {
869 }
870 
871 static inline void mem_cgroup_oom_enable(void)
872 {
873 }
874 
875 static inline void mem_cgroup_oom_disable(void)
876 {
877 }
878 
879 static inline bool task_in_memcg_oom(struct task_struct *p)
880 {
881 	return false;
882 }
883 
884 static inline bool mem_cgroup_oom_synchronize(bool wait)
885 {
886 	return false;
887 }
888 
889 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg,
890 					     int idx)
891 {
892 	return 0;
893 }
894 
895 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
896 				     int idx,
897 				     int nr)
898 {
899 }
900 
901 static inline void mod_memcg_state(struct mem_cgroup *memcg,
902 				   int idx,
903 				   int nr)
904 {
905 }
906 
907 static inline void __mod_memcg_page_state(struct page *page,
908 					  int idx,
909 					  int nr)
910 {
911 }
912 
913 static inline void mod_memcg_page_state(struct page *page,
914 					int idx,
915 					int nr)
916 {
917 }
918 
919 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
920 					      enum node_stat_item idx)
921 {
922 	return node_page_state(lruvec_pgdat(lruvec), idx);
923 }
924 
925 static inline void __mod_lruvec_state(struct lruvec *lruvec,
926 				      enum node_stat_item idx, int val)
927 {
928 	__mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
929 }
930 
931 static inline void mod_lruvec_state(struct lruvec *lruvec,
932 				    enum node_stat_item idx, int val)
933 {
934 	mod_node_page_state(lruvec_pgdat(lruvec), idx, val);
935 }
936 
937 static inline void __mod_lruvec_page_state(struct page *page,
938 					   enum node_stat_item idx, int val)
939 {
940 	__mod_node_page_state(page_pgdat(page), idx, val);
941 }
942 
943 static inline void mod_lruvec_page_state(struct page *page,
944 					 enum node_stat_item idx, int val)
945 {
946 	mod_node_page_state(page_pgdat(page), idx, val);
947 }
948 
949 static inline
950 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
951 					    gfp_t gfp_mask,
952 					    unsigned long *total_scanned)
953 {
954 	return 0;
955 }
956 
957 static inline void mem_cgroup_split_huge_fixup(struct page *head)
958 {
959 }
960 
961 static inline void count_memcg_events(struct mem_cgroup *memcg,
962 				      enum vm_event_item idx,
963 				      unsigned long count)
964 {
965 }
966 
967 static inline void count_memcg_page_event(struct page *page,
968 					  int idx)
969 {
970 }
971 
972 static inline
973 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
974 {
975 }
976 #endif /* CONFIG_MEMCG */
977 
978 /* idx can be of type enum memcg_stat_item or node_stat_item */
979 static inline void __inc_memcg_state(struct mem_cgroup *memcg,
980 				     int idx)
981 {
982 	__mod_memcg_state(memcg, idx, 1);
983 }
984 
985 /* idx can be of type enum memcg_stat_item or node_stat_item */
986 static inline void __dec_memcg_state(struct mem_cgroup *memcg,
987 				     int idx)
988 {
989 	__mod_memcg_state(memcg, idx, -1);
990 }
991 
992 /* idx can be of type enum memcg_stat_item or node_stat_item */
993 static inline void __inc_memcg_page_state(struct page *page,
994 					  int idx)
995 {
996 	__mod_memcg_page_state(page, idx, 1);
997 }
998 
999 /* idx can be of type enum memcg_stat_item or node_stat_item */
1000 static inline void __dec_memcg_page_state(struct page *page,
1001 					  int idx)
1002 {
1003 	__mod_memcg_page_state(page, idx, -1);
1004 }
1005 
1006 static inline void __inc_lruvec_state(struct lruvec *lruvec,
1007 				      enum node_stat_item idx)
1008 {
1009 	__mod_lruvec_state(lruvec, idx, 1);
1010 }
1011 
1012 static inline void __dec_lruvec_state(struct lruvec *lruvec,
1013 				      enum node_stat_item idx)
1014 {
1015 	__mod_lruvec_state(lruvec, idx, -1);
1016 }
1017 
1018 static inline void __inc_lruvec_page_state(struct page *page,
1019 					   enum node_stat_item idx)
1020 {
1021 	__mod_lruvec_page_state(page, idx, 1);
1022 }
1023 
1024 static inline void __dec_lruvec_page_state(struct page *page,
1025 					   enum node_stat_item idx)
1026 {
1027 	__mod_lruvec_page_state(page, idx, -1);
1028 }
1029 
1030 /* idx can be of type enum memcg_stat_item or node_stat_item */
1031 static inline void inc_memcg_state(struct mem_cgroup *memcg,
1032 				   int idx)
1033 {
1034 	mod_memcg_state(memcg, idx, 1);
1035 }
1036 
1037 /* idx can be of type enum memcg_stat_item or node_stat_item */
1038 static inline void dec_memcg_state(struct mem_cgroup *memcg,
1039 				   int idx)
1040 {
1041 	mod_memcg_state(memcg, idx, -1);
1042 }
1043 
1044 /* idx can be of type enum memcg_stat_item or node_stat_item */
1045 static inline void inc_memcg_page_state(struct page *page,
1046 					int idx)
1047 {
1048 	mod_memcg_page_state(page, idx, 1);
1049 }
1050 
1051 /* idx can be of type enum memcg_stat_item or node_stat_item */
1052 static inline void dec_memcg_page_state(struct page *page,
1053 					int idx)
1054 {
1055 	mod_memcg_page_state(page, idx, -1);
1056 }
1057 
1058 static inline void inc_lruvec_state(struct lruvec *lruvec,
1059 				    enum node_stat_item idx)
1060 {
1061 	mod_lruvec_state(lruvec, idx, 1);
1062 }
1063 
1064 static inline void dec_lruvec_state(struct lruvec *lruvec,
1065 				    enum node_stat_item idx)
1066 {
1067 	mod_lruvec_state(lruvec, idx, -1);
1068 }
1069 
1070 static inline void inc_lruvec_page_state(struct page *page,
1071 					 enum node_stat_item idx)
1072 {
1073 	mod_lruvec_page_state(page, idx, 1);
1074 }
1075 
1076 static inline void dec_lruvec_page_state(struct page *page,
1077 					 enum node_stat_item idx)
1078 {
1079 	mod_lruvec_page_state(page, idx, -1);
1080 }
1081 
1082 #ifdef CONFIG_CGROUP_WRITEBACK
1083 
1084 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
1085 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1086 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1087 			 unsigned long *pheadroom, unsigned long *pdirty,
1088 			 unsigned long *pwriteback);
1089 
1090 #else	/* CONFIG_CGROUP_WRITEBACK */
1091 
1092 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1093 {
1094 	return NULL;
1095 }
1096 
1097 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1098 				       unsigned long *pfilepages,
1099 				       unsigned long *pheadroom,
1100 				       unsigned long *pdirty,
1101 				       unsigned long *pwriteback)
1102 {
1103 }
1104 
1105 #endif	/* CONFIG_CGROUP_WRITEBACK */
1106 
1107 struct sock;
1108 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1109 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1110 #ifdef CONFIG_MEMCG
1111 extern struct static_key_false memcg_sockets_enabled_key;
1112 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1113 void mem_cgroup_sk_alloc(struct sock *sk);
1114 void mem_cgroup_sk_free(struct sock *sk);
1115 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1116 {
1117 	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1118 		return true;
1119 	do {
1120 		if (time_before(jiffies, memcg->socket_pressure))
1121 			return true;
1122 	} while ((memcg = parent_mem_cgroup(memcg)));
1123 	return false;
1124 }
1125 #else
1126 #define mem_cgroup_sockets_enabled 0
1127 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1128 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1129 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1130 {
1131 	return false;
1132 }
1133 #endif
1134 
1135 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
1136 void memcg_kmem_put_cache(struct kmem_cache *cachep);
1137 int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
1138 			    struct mem_cgroup *memcg);
1139 int memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
1140 void memcg_kmem_uncharge(struct page *page, int order);
1141 
1142 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
1143 extern struct static_key_false memcg_kmem_enabled_key;
1144 extern struct workqueue_struct *memcg_kmem_cache_wq;
1145 
1146 extern int memcg_nr_cache_ids;
1147 void memcg_get_cache_ids(void);
1148 void memcg_put_cache_ids(void);
1149 
1150 /*
1151  * Helper macro to loop through all memcg-specific caches. Callers must still
1152  * check if the cache is valid (it is either valid or NULL).
1153  * the slab_mutex must be held when looping through those caches
1154  */
1155 #define for_each_memcg_cache_index(_idx)	\
1156 	for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1157 
1158 static inline bool memcg_kmem_enabled(void)
1159 {
1160 	return static_branch_unlikely(&memcg_kmem_enabled_key);
1161 }
1162 
1163 /*
1164  * helper for accessing a memcg's index. It will be used as an index in the
1165  * child cache array in kmem_cache, and also to derive its name. This function
1166  * will return -1 when this is not a kmem-limited memcg.
1167  */
1168 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1169 {
1170 	return memcg ? memcg->kmemcg_id : -1;
1171 }
1172 
1173 #else
1174 #define for_each_memcg_cache_index(_idx)	\
1175 	for (; NULL; )
1176 
1177 static inline bool memcg_kmem_enabled(void)
1178 {
1179 	return false;
1180 }
1181 
1182 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1183 {
1184 	return -1;
1185 }
1186 
1187 static inline void memcg_get_cache_ids(void)
1188 {
1189 }
1190 
1191 static inline void memcg_put_cache_ids(void)
1192 {
1193 }
1194 
1195 #endif /* CONFIG_MEMCG && !CONFIG_SLOB */
1196 
1197 #endif /* _LINUX_MEMCONTROL_H */
1198