]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/memcontrol.h
mm: memcontrol: convert NR_ANON_THPS account to pages
[mirror_ubuntu-jammy-kernel.git] / include / linux / memcontrol.h
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
25 struct mem_cgroup;
26 struct obj_cgroup;
27 struct page;
28 struct mm_struct;
29 struct kmem_cache;
30
31 /* Cgroup-specific page state, on top of universal node page state */
32 enum memcg_stat_item {
33 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS,
34 MEMCG_SOCK,
35 MEMCG_PERCPU_B,
36 MEMCG_NR_STAT,
37 };
38
39 enum memcg_memory_event {
40 MEMCG_LOW,
41 MEMCG_HIGH,
42 MEMCG_MAX,
43 MEMCG_OOM,
44 MEMCG_OOM_KILL,
45 MEMCG_SWAP_HIGH,
46 MEMCG_SWAP_MAX,
47 MEMCG_SWAP_FAIL,
48 MEMCG_NR_MEMORY_EVENTS,
49 };
50
51 struct mem_cgroup_reclaim_cookie {
52 pg_data_t *pgdat;
53 unsigned int generation;
54 };
55
56 #ifdef CONFIG_MEMCG
57
58 #define MEM_CGROUP_ID_SHIFT 16
59 #define MEM_CGROUP_ID_MAX USHRT_MAX
60
61 struct mem_cgroup_id {
62 int id;
63 refcount_t ref;
64 };
65
66 /*
67 * Per memcg event counter is incremented at every pagein/pageout. With THP,
68 * it will be incremented by the number of pages. This counter is used
69 * to trigger some periodic events. This is straightforward and better
70 * than using jiffies etc. to handle periodic memcg event.
71 */
72 enum mem_cgroup_events_target {
73 MEM_CGROUP_TARGET_THRESH,
74 MEM_CGROUP_TARGET_SOFTLIMIT,
75 MEM_CGROUP_NTARGETS,
76 };
77
78 struct memcg_vmstats_percpu {
79 long stat[MEMCG_NR_STAT];
80 unsigned long events[NR_VM_EVENT_ITEMS];
81 unsigned long nr_page_events;
82 unsigned long targets[MEM_CGROUP_NTARGETS];
83 };
84
85 struct mem_cgroup_reclaim_iter {
86 struct mem_cgroup *position;
87 /* scan generation, increased every round-trip */
88 unsigned int generation;
89 };
90
91 struct lruvec_stat {
92 long count[NR_VM_NODE_STAT_ITEMS];
93 };
94
95 struct batched_lruvec_stat {
96 s32 count[NR_VM_NODE_STAT_ITEMS];
97 };
98
99 /*
100 * Bitmap of shrinker::id corresponding to memcg-aware shrinkers,
101 * which have elements charged to this memcg.
102 */
103 struct memcg_shrinker_map {
104 struct rcu_head rcu;
105 unsigned long map[];
106 };
107
108 /*
109 * per-node information in memory controller.
110 */
111 struct mem_cgroup_per_node {
112 struct lruvec lruvec;
113
114 /*
115 * Legacy local VM stats. This should be struct lruvec_stat and
116 * cannot be optimized to struct batched_lruvec_stat. Because
117 * the threshold of the lruvec_stat_cpu can be as big as
118 * MEMCG_CHARGE_BATCH * PAGE_SIZE. It can fit into s32. But this
119 * filed has no upper limit.
120 */
121 struct lruvec_stat __percpu *lruvec_stat_local;
122
123 /* Subtree VM stats (batched updates) */
124 struct batched_lruvec_stat __percpu *lruvec_stat_cpu;
125 atomic_long_t lruvec_stat[NR_VM_NODE_STAT_ITEMS];
126
127 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
128
129 struct mem_cgroup_reclaim_iter iter;
130
131 struct memcg_shrinker_map __rcu *shrinker_map;
132
133 struct rb_node tree_node; /* RB tree node */
134 unsigned long usage_in_excess;/* Set to the value by which */
135 /* the soft limit is exceeded*/
136 bool on_tree;
137 struct mem_cgroup *memcg; /* Back pointer, we cannot */
138 /* use container_of */
139 };
140
141 struct mem_cgroup_threshold {
142 struct eventfd_ctx *eventfd;
143 unsigned long threshold;
144 };
145
146 /* For threshold */
147 struct mem_cgroup_threshold_ary {
148 /* An array index points to threshold just below or equal to usage. */
149 int current_threshold;
150 /* Size of entries[] */
151 unsigned int size;
152 /* Array of thresholds */
153 struct mem_cgroup_threshold entries[];
154 };
155
156 struct mem_cgroup_thresholds {
157 /* Primary thresholds array */
158 struct mem_cgroup_threshold_ary *primary;
159 /*
160 * Spare threshold array.
161 * This is needed to make mem_cgroup_unregister_event() "never fail".
162 * It must be able to store at least primary->size - 1 entries.
163 */
164 struct mem_cgroup_threshold_ary *spare;
165 };
166
167 enum memcg_kmem_state {
168 KMEM_NONE,
169 KMEM_ALLOCATED,
170 KMEM_ONLINE,
171 };
172
173 #if defined(CONFIG_SMP)
174 struct memcg_padding {
175 char x[0];
176 } ____cacheline_internodealigned_in_smp;
177 #define MEMCG_PADDING(name) struct memcg_padding name;
178 #else
179 #define MEMCG_PADDING(name)
180 #endif
181
182 /*
183 * Remember four most recent foreign writebacks with dirty pages in this
184 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
185 * one in a given round, we're likely to catch it later if it keeps
186 * foreign-dirtying, so a fairly low count should be enough.
187 *
188 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
189 */
190 #define MEMCG_CGWB_FRN_CNT 4
191
192 struct memcg_cgwb_frn {
193 u64 bdi_id; /* bdi->id of the foreign inode */
194 int memcg_id; /* memcg->css.id of foreign inode */
195 u64 at; /* jiffies_64 at the time of dirtying */
196 struct wb_completion done; /* tracks in-flight foreign writebacks */
197 };
198
199 /*
200 * Bucket for arbitrarily byte-sized objects charged to a memory
201 * cgroup. The bucket can be reparented in one piece when the cgroup
202 * is destroyed, without having to round up the individual references
203 * of all live memory objects in the wild.
204 */
205 struct obj_cgroup {
206 struct percpu_ref refcnt;
207 struct mem_cgroup *memcg;
208 atomic_t nr_charged_bytes;
209 union {
210 struct list_head list;
211 struct rcu_head rcu;
212 };
213 };
214
215 /*
216 * The memory controller data structure. The memory controller controls both
217 * page cache and RSS per cgroup. We would eventually like to provide
218 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
219 * to help the administrator determine what knobs to tune.
220 */
221 struct mem_cgroup {
222 struct cgroup_subsys_state css;
223
224 /* Private memcg ID. Used to ID objects that outlive the cgroup */
225 struct mem_cgroup_id id;
226
227 /* Accounted resources */
228 struct page_counter memory; /* Both v1 & v2 */
229
230 union {
231 struct page_counter swap; /* v2 only */
232 struct page_counter memsw; /* v1 only */
233 };
234
235 /* Legacy consumer-oriented counters */
236 struct page_counter kmem; /* v1 only */
237 struct page_counter tcpmem; /* v1 only */
238
239 /* Range enforcement for interrupt charges */
240 struct work_struct high_work;
241
242 unsigned long soft_limit;
243
244 /* vmpressure notifications */
245 struct vmpressure vmpressure;
246
247 /*
248 * Should the OOM killer kill all belonging tasks, had it kill one?
249 */
250 bool oom_group;
251
252 /* protected by memcg_oom_lock */
253 bool oom_lock;
254 int under_oom;
255
256 int swappiness;
257 /* OOM-Killer disable */
258 int oom_kill_disable;
259
260 /* memory.events and memory.events.local */
261 struct cgroup_file events_file;
262 struct cgroup_file events_local_file;
263
264 /* handle for "memory.swap.events" */
265 struct cgroup_file swap_events_file;
266
267 /* protect arrays of thresholds */
268 struct mutex thresholds_lock;
269
270 /* thresholds for memory usage. RCU-protected */
271 struct mem_cgroup_thresholds thresholds;
272
273 /* thresholds for mem+swap usage. RCU-protected */
274 struct mem_cgroup_thresholds memsw_thresholds;
275
276 /* For oom notifier event fd */
277 struct list_head oom_notify;
278
279 /*
280 * Should we move charges of a task when a task is moved into this
281 * mem_cgroup ? And what type of charges should we move ?
282 */
283 unsigned long move_charge_at_immigrate;
284 /* taken only while moving_account > 0 */
285 spinlock_t move_lock;
286 unsigned long move_lock_flags;
287
288 MEMCG_PADDING(_pad1_);
289
290 atomic_long_t vmstats[MEMCG_NR_STAT];
291 atomic_long_t vmevents[NR_VM_EVENT_ITEMS];
292
293 /* memory.events */
294 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
295 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
296
297 unsigned long socket_pressure;
298
299 /* Legacy tcp memory accounting */
300 bool tcpmem_active;
301 int tcpmem_pressure;
302
303 #ifdef CONFIG_MEMCG_KMEM
304 int kmemcg_id;
305 enum memcg_kmem_state kmem_state;
306 struct obj_cgroup __rcu *objcg;
307 struct list_head objcg_list; /* list of inherited objcgs */
308 #endif
309
310 MEMCG_PADDING(_pad2_);
311
312 /*
313 * set > 0 if pages under this cgroup are moving to other cgroup.
314 */
315 atomic_t moving_account;
316 struct task_struct *move_lock_task;
317
318 /* Legacy local VM stats and events */
319 struct memcg_vmstats_percpu __percpu *vmstats_local;
320
321 /* Subtree VM stats and events (batched updates) */
322 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
323
324 #ifdef CONFIG_CGROUP_WRITEBACK
325 struct list_head cgwb_list;
326 struct wb_domain cgwb_domain;
327 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
328 #endif
329
330 /* List of events which userspace want to receive */
331 struct list_head event_list;
332 spinlock_t event_list_lock;
333
334 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
335 struct deferred_split deferred_split_queue;
336 #endif
337
338 struct mem_cgroup_per_node *nodeinfo[0];
339 /* WARNING: nodeinfo must be the last member here */
340 };
341
342 /*
343 * size of first charge trial. "32" comes from vmscan.c's magic value.
344 * TODO: maybe necessary to use big numbers in big irons.
345 */
346 #define MEMCG_CHARGE_BATCH 32U
347
348 extern struct mem_cgroup *root_mem_cgroup;
349
350 enum page_memcg_data_flags {
351 /* page->memcg_data is a pointer to an objcgs vector */
352 MEMCG_DATA_OBJCGS = (1UL << 0),
353 /* page has been accounted as a non-slab kernel page */
354 MEMCG_DATA_KMEM = (1UL << 1),
355 /* the next bit after the last actual flag */
356 __NR_MEMCG_DATA_FLAGS = (1UL << 2),
357 };
358
359 #define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1)
360
361 /*
362 * page_memcg - get the memory cgroup associated with a page
363 * @page: a pointer to the page struct
364 *
365 * Returns a pointer to the memory cgroup associated with the page,
366 * or NULL. This function assumes that the page is known to have a
367 * proper memory cgroup pointer. It's not safe to call this function
368 * against some type of pages, e.g. slab pages or ex-slab pages.
369 *
370 * Any of the following ensures page and memcg binding stability:
371 * - the page lock
372 * - LRU isolation
373 * - lock_page_memcg()
374 * - exclusive reference
375 */
376 static inline struct mem_cgroup *page_memcg(struct page *page)
377 {
378 unsigned long memcg_data = page->memcg_data;
379
380 VM_BUG_ON_PAGE(PageSlab(page), page);
381 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_OBJCGS, page);
382
383 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
384 }
385
386 /*
387 * page_memcg_rcu - locklessly get the memory cgroup associated with a page
388 * @page: a pointer to the page struct
389 *
390 * Returns a pointer to the memory cgroup associated with the page,
391 * or NULL. This function assumes that the page is known to have a
392 * proper memory cgroup pointer. It's not safe to call this function
393 * against some type of pages, e.g. slab pages or ex-slab pages.
394 */
395 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
396 {
397 VM_BUG_ON_PAGE(PageSlab(page), page);
398 WARN_ON_ONCE(!rcu_read_lock_held());
399
400 return (struct mem_cgroup *)(READ_ONCE(page->memcg_data) &
401 ~MEMCG_DATA_FLAGS_MASK);
402 }
403
404 /*
405 * page_memcg_check - get the memory cgroup associated with a page
406 * @page: a pointer to the page struct
407 *
408 * Returns a pointer to the memory cgroup associated with the page,
409 * or NULL. This function unlike page_memcg() can take any page
410 * as an argument. It has to be used in cases when it's not known if a page
411 * has an associated memory cgroup pointer or an object cgroups vector.
412 *
413 * Any of the following ensures page and memcg binding stability:
414 * - the page lock
415 * - LRU isolation
416 * - lock_page_memcg()
417 * - exclusive reference
418 */
419 static inline struct mem_cgroup *page_memcg_check(struct page *page)
420 {
421 /*
422 * Because page->memcg_data might be changed asynchronously
423 * for slab pages, READ_ONCE() should be used here.
424 */
425 unsigned long memcg_data = READ_ONCE(page->memcg_data);
426
427 if (memcg_data & MEMCG_DATA_OBJCGS)
428 return NULL;
429
430 return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
431 }
432
433 /*
434 * PageMemcgKmem - check if the page has MemcgKmem flag set
435 * @page: a pointer to the page struct
436 *
437 * Checks if the page has MemcgKmem flag set. The caller must ensure that
438 * the page has an associated memory cgroup. It's not safe to call this function
439 * against some types of pages, e.g. slab pages.
440 */
441 static inline bool PageMemcgKmem(struct page *page)
442 {
443 VM_BUG_ON_PAGE(page->memcg_data & MEMCG_DATA_OBJCGS, page);
444 return page->memcg_data & MEMCG_DATA_KMEM;
445 }
446
447 #ifdef CONFIG_MEMCG_KMEM
448 /*
449 * page_objcgs - get the object cgroups vector associated with a page
450 * @page: a pointer to the page struct
451 *
452 * Returns a pointer to the object cgroups vector associated with the page,
453 * or NULL. This function assumes that the page is known to have an
454 * associated object cgroups vector. It's not safe to call this function
455 * against pages, which might have an associated memory cgroup: e.g.
456 * kernel stack pages.
457 */
458 static inline struct obj_cgroup **page_objcgs(struct page *page)
459 {
460 unsigned long memcg_data = READ_ONCE(page->memcg_data);
461
462 VM_BUG_ON_PAGE(memcg_data && !(memcg_data & MEMCG_DATA_OBJCGS), page);
463 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
464
465 return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
466 }
467
468 /*
469 * page_objcgs_check - get the object cgroups vector associated with a page
470 * @page: a pointer to the page struct
471 *
472 * Returns a pointer to the object cgroups vector associated with the page,
473 * or NULL. This function is safe to use if the page can be directly associated
474 * with a memory cgroup.
475 */
476 static inline struct obj_cgroup **page_objcgs_check(struct page *page)
477 {
478 unsigned long memcg_data = READ_ONCE(page->memcg_data);
479
480 if (!memcg_data || !(memcg_data & MEMCG_DATA_OBJCGS))
481 return NULL;
482
483 VM_BUG_ON_PAGE(memcg_data & MEMCG_DATA_KMEM, page);
484
485 return (struct obj_cgroup **)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
486 }
487
488 #else
489 static inline struct obj_cgroup **page_objcgs(struct page *page)
490 {
491 return NULL;
492 }
493
494 static inline struct obj_cgroup **page_objcgs_check(struct page *page)
495 {
496 return NULL;
497 }
498 #endif
499
500 static __always_inline bool memcg_stat_item_in_bytes(int idx)
501 {
502 if (idx == MEMCG_PERCPU_B)
503 return true;
504 return vmstat_item_in_bytes(idx);
505 }
506
507 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
508 {
509 return (memcg == root_mem_cgroup);
510 }
511
512 static inline bool mem_cgroup_disabled(void)
513 {
514 return !cgroup_subsys_enabled(memory_cgrp_subsys);
515 }
516
517 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *root,
518 struct mem_cgroup *memcg,
519 bool in_low_reclaim)
520 {
521 if (mem_cgroup_disabled())
522 return 0;
523
524 /*
525 * There is no reclaim protection applied to a targeted reclaim.
526 * We are special casing this specific case here because
527 * mem_cgroup_protected calculation is not robust enough to keep
528 * the protection invariant for calculated effective values for
529 * parallel reclaimers with different reclaim target. This is
530 * especially a problem for tail memcgs (as they have pages on LRU)
531 * which would want to have effective values 0 for targeted reclaim
532 * but a different value for external reclaim.
533 *
534 * Example
535 * Let's have global and A's reclaim in parallel:
536 * |
537 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
538 * |\
539 * | C (low = 1G, usage = 2.5G)
540 * B (low = 1G, usage = 0.5G)
541 *
542 * For the global reclaim
543 * A.elow = A.low
544 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
545 * C.elow = min(C.usage, C.low)
546 *
547 * With the effective values resetting we have A reclaim
548 * A.elow = 0
549 * B.elow = B.low
550 * C.elow = C.low
551 *
552 * If the global reclaim races with A's reclaim then
553 * B.elow = C.elow = 0 because children_low_usage > A.elow)
554 * is possible and reclaiming B would be violating the protection.
555 *
556 */
557 if (root == memcg)
558 return 0;
559
560 if (in_low_reclaim)
561 return READ_ONCE(memcg->memory.emin);
562
563 return max(READ_ONCE(memcg->memory.emin),
564 READ_ONCE(memcg->memory.elow));
565 }
566
567 void mem_cgroup_calculate_protection(struct mem_cgroup *root,
568 struct mem_cgroup *memcg);
569
570 static inline bool mem_cgroup_supports_protection(struct mem_cgroup *memcg)
571 {
572 /*
573 * The root memcg doesn't account charges, and doesn't support
574 * protection.
575 */
576 return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg);
577
578 }
579
580 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
581 {
582 if (!mem_cgroup_supports_protection(memcg))
583 return false;
584
585 return READ_ONCE(memcg->memory.elow) >=
586 page_counter_read(&memcg->memory);
587 }
588
589 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
590 {
591 if (!mem_cgroup_supports_protection(memcg))
592 return false;
593
594 return READ_ONCE(memcg->memory.emin) >=
595 page_counter_read(&memcg->memory);
596 }
597
598 int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask);
599
600 void mem_cgroup_uncharge(struct page *page);
601 void mem_cgroup_uncharge_list(struct list_head *page_list);
602
603 void mem_cgroup_migrate(struct page *oldpage, struct page *newpage);
604
605 static struct mem_cgroup_per_node *
606 mem_cgroup_nodeinfo(struct mem_cgroup *memcg, int nid)
607 {
608 return memcg->nodeinfo[nid];
609 }
610
611 /**
612 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
613 * @memcg: memcg of the wanted lruvec
614 * @pgdat: pglist_data
615 *
616 * Returns the lru list vector holding pages for a given @memcg &
617 * @pgdat combination. This can be the node lruvec, if the memory
618 * controller is disabled.
619 */
620 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
621 struct pglist_data *pgdat)
622 {
623 struct mem_cgroup_per_node *mz;
624 struct lruvec *lruvec;
625
626 if (mem_cgroup_disabled()) {
627 lruvec = &pgdat->__lruvec;
628 goto out;
629 }
630
631 if (!memcg)
632 memcg = root_mem_cgroup;
633
634 mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
635 lruvec = &mz->lruvec;
636 out:
637 /*
638 * Since a node can be onlined after the mem_cgroup was created,
639 * we have to be prepared to initialize lruvec->pgdat here;
640 * and if offlined then reonlined, we need to reinitialize it.
641 */
642 if (unlikely(lruvec->pgdat != pgdat))
643 lruvec->pgdat = pgdat;
644 return lruvec;
645 }
646
647 /**
648 * mem_cgroup_page_lruvec - return lruvec for isolating/putting an LRU page
649 * @page: the page
650 * @pgdat: pgdat of the page
651 *
652 * This function relies on page->mem_cgroup being stable.
653 */
654 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
655 struct pglist_data *pgdat)
656 {
657 struct mem_cgroup *memcg = page_memcg(page);
658
659 VM_WARN_ON_ONCE_PAGE(!memcg && !mem_cgroup_disabled(), page);
660 return mem_cgroup_lruvec(memcg, pgdat);
661 }
662
663 static inline bool lruvec_holds_page_lru_lock(struct page *page,
664 struct lruvec *lruvec)
665 {
666 pg_data_t *pgdat = page_pgdat(page);
667 const struct mem_cgroup *memcg;
668 struct mem_cgroup_per_node *mz;
669
670 if (mem_cgroup_disabled())
671 return lruvec == &pgdat->__lruvec;
672
673 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
674 memcg = page_memcg(page) ? : root_mem_cgroup;
675
676 return lruvec->pgdat == pgdat && mz->memcg == memcg;
677 }
678
679 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
680
681 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
682
683 struct mem_cgroup *get_mem_cgroup_from_page(struct page *page);
684
685 struct lruvec *lock_page_lruvec(struct page *page);
686 struct lruvec *lock_page_lruvec_irq(struct page *page);
687 struct lruvec *lock_page_lruvec_irqsave(struct page *page,
688 unsigned long *flags);
689
690 #ifdef CONFIG_DEBUG_VM
691 void lruvec_memcg_debug(struct lruvec *lruvec, struct page *page);
692 #else
693 static inline void lruvec_memcg_debug(struct lruvec *lruvec, struct page *page)
694 {
695 }
696 #endif
697
698 static inline
699 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
700 return css ? container_of(css, struct mem_cgroup, css) : NULL;
701 }
702
703 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
704 {
705 return percpu_ref_tryget(&objcg->refcnt);
706 }
707
708 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
709 {
710 percpu_ref_get(&objcg->refcnt);
711 }
712
713 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
714 {
715 percpu_ref_put(&objcg->refcnt);
716 }
717
718 /*
719 * After the initialization objcg->memcg is always pointing at
720 * a valid memcg, but can be atomically swapped to the parent memcg.
721 *
722 * The caller must ensure that the returned memcg won't be released:
723 * e.g. acquire the rcu_read_lock or css_set_lock.
724 */
725 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
726 {
727 return READ_ONCE(objcg->memcg);
728 }
729
730 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
731 {
732 if (memcg)
733 css_put(&memcg->css);
734 }
735
736 #define mem_cgroup_from_counter(counter, member) \
737 container_of(counter, struct mem_cgroup, member)
738
739 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
740 struct mem_cgroup *,
741 struct mem_cgroup_reclaim_cookie *);
742 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
743 int mem_cgroup_scan_tasks(struct mem_cgroup *,
744 int (*)(struct task_struct *, void *), void *);
745
746 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
747 {
748 if (mem_cgroup_disabled())
749 return 0;
750
751 return memcg->id.id;
752 }
753 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
754
755 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
756 {
757 return mem_cgroup_from_css(seq_css(m));
758 }
759
760 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
761 {
762 struct mem_cgroup_per_node *mz;
763
764 if (mem_cgroup_disabled())
765 return NULL;
766
767 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
768 return mz->memcg;
769 }
770
771 /**
772 * parent_mem_cgroup - find the accounting parent of a memcg
773 * @memcg: memcg whose parent to find
774 *
775 * Returns the parent memcg, or NULL if this is the root or the memory
776 * controller is in legacy no-hierarchy mode.
777 */
778 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
779 {
780 if (!memcg->memory.parent)
781 return NULL;
782 return mem_cgroup_from_counter(memcg->memory.parent, memory);
783 }
784
785 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
786 struct mem_cgroup *root)
787 {
788 if (root == memcg)
789 return true;
790 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
791 }
792
793 static inline bool mm_match_cgroup(struct mm_struct *mm,
794 struct mem_cgroup *memcg)
795 {
796 struct mem_cgroup *task_memcg;
797 bool match = false;
798
799 rcu_read_lock();
800 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
801 if (task_memcg)
802 match = mem_cgroup_is_descendant(task_memcg, memcg);
803 rcu_read_unlock();
804 return match;
805 }
806
807 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
808 ino_t page_cgroup_ino(struct page *page);
809
810 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
811 {
812 if (mem_cgroup_disabled())
813 return true;
814 return !!(memcg->css.flags & CSS_ONLINE);
815 }
816
817 /*
818 * For memory reclaim.
819 */
820 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
821
822 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
823 int zid, int nr_pages);
824
825 static inline
826 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
827 enum lru_list lru, int zone_idx)
828 {
829 struct mem_cgroup_per_node *mz;
830
831 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
832 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
833 }
834
835 void mem_cgroup_handle_over_high(void);
836
837 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
838
839 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
840
841 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
842 struct task_struct *p);
843
844 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
845
846 static inline void mem_cgroup_enter_user_fault(void)
847 {
848 WARN_ON(current->in_user_fault);
849 current->in_user_fault = 1;
850 }
851
852 static inline void mem_cgroup_exit_user_fault(void)
853 {
854 WARN_ON(!current->in_user_fault);
855 current->in_user_fault = 0;
856 }
857
858 static inline bool task_in_memcg_oom(struct task_struct *p)
859 {
860 return p->memcg_in_oom;
861 }
862
863 bool mem_cgroup_oom_synchronize(bool wait);
864 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
865 struct mem_cgroup *oom_domain);
866 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
867
868 #ifdef CONFIG_MEMCG_SWAP
869 extern bool cgroup_memory_noswap;
870 #endif
871
872 struct mem_cgroup *lock_page_memcg(struct page *page);
873 void __unlock_page_memcg(struct mem_cgroup *memcg);
874 void unlock_page_memcg(struct page *page);
875
876 /*
877 * idx can be of type enum memcg_stat_item or node_stat_item.
878 * Keep in sync with memcg_exact_page_state().
879 */
880 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
881 {
882 long x = atomic_long_read(&memcg->vmstats[idx]);
883 #ifdef CONFIG_SMP
884 if (x < 0)
885 x = 0;
886 #endif
887 return x;
888 }
889
890 /*
891 * idx can be of type enum memcg_stat_item or node_stat_item.
892 * Keep in sync with memcg_exact_page_state().
893 */
894 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
895 int idx)
896 {
897 long x = 0;
898 int cpu;
899
900 for_each_possible_cpu(cpu)
901 x += per_cpu(memcg->vmstats_local->stat[idx], cpu);
902 #ifdef CONFIG_SMP
903 if (x < 0)
904 x = 0;
905 #endif
906 return x;
907 }
908
909 void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
910
911 /* idx can be of type enum memcg_stat_item or node_stat_item */
912 static inline void mod_memcg_state(struct mem_cgroup *memcg,
913 int idx, int val)
914 {
915 unsigned long flags;
916
917 local_irq_save(flags);
918 __mod_memcg_state(memcg, idx, val);
919 local_irq_restore(flags);
920 }
921
922 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
923 enum node_stat_item idx)
924 {
925 struct mem_cgroup_per_node *pn;
926 long x;
927
928 if (mem_cgroup_disabled())
929 return node_page_state(lruvec_pgdat(lruvec), idx);
930
931 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
932 x = atomic_long_read(&pn->lruvec_stat[idx]);
933 #ifdef CONFIG_SMP
934 if (x < 0)
935 x = 0;
936 #endif
937 return x;
938 }
939
940 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
941 enum node_stat_item idx)
942 {
943 struct mem_cgroup_per_node *pn;
944 long x = 0;
945 int cpu;
946
947 if (mem_cgroup_disabled())
948 return node_page_state(lruvec_pgdat(lruvec), idx);
949
950 pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
951 for_each_possible_cpu(cpu)
952 x += per_cpu(pn->lruvec_stat_local->count[idx], cpu);
953 #ifdef CONFIG_SMP
954 if (x < 0)
955 x = 0;
956 #endif
957 return x;
958 }
959
960 void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
961 int val);
962 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
963
964 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
965 int val)
966 {
967 unsigned long flags;
968
969 local_irq_save(flags);
970 __mod_lruvec_kmem_state(p, idx, val);
971 local_irq_restore(flags);
972 }
973
974 static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
975 enum node_stat_item idx, int val)
976 {
977 unsigned long flags;
978
979 local_irq_save(flags);
980 __mod_memcg_lruvec_state(lruvec, idx, val);
981 local_irq_restore(flags);
982 }
983
984 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
985 gfp_t gfp_mask,
986 unsigned long *total_scanned);
987
988 void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
989 unsigned long count);
990
991 static inline void count_memcg_events(struct mem_cgroup *memcg,
992 enum vm_event_item idx,
993 unsigned long count)
994 {
995 unsigned long flags;
996
997 local_irq_save(flags);
998 __count_memcg_events(memcg, idx, count);
999 local_irq_restore(flags);
1000 }
1001
1002 static inline void count_memcg_page_event(struct page *page,
1003 enum vm_event_item idx)
1004 {
1005 struct mem_cgroup *memcg = page_memcg(page);
1006
1007 if (memcg)
1008 count_memcg_events(memcg, idx, 1);
1009 }
1010
1011 static inline void count_memcg_event_mm(struct mm_struct *mm,
1012 enum vm_event_item idx)
1013 {
1014 struct mem_cgroup *memcg;
1015
1016 if (mem_cgroup_disabled())
1017 return;
1018
1019 rcu_read_lock();
1020 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1021 if (likely(memcg))
1022 count_memcg_events(memcg, idx, 1);
1023 rcu_read_unlock();
1024 }
1025
1026 static inline void memcg_memory_event(struct mem_cgroup *memcg,
1027 enum memcg_memory_event event)
1028 {
1029 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
1030 event == MEMCG_SWAP_FAIL;
1031
1032 atomic_long_inc(&memcg->memory_events_local[event]);
1033 if (!swap_event)
1034 cgroup_file_notify(&memcg->events_local_file);
1035
1036 do {
1037 atomic_long_inc(&memcg->memory_events[event]);
1038 if (swap_event)
1039 cgroup_file_notify(&memcg->swap_events_file);
1040 else
1041 cgroup_file_notify(&memcg->events_file);
1042
1043 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1044 break;
1045 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1046 break;
1047 } while ((memcg = parent_mem_cgroup(memcg)) &&
1048 !mem_cgroup_is_root(memcg));
1049 }
1050
1051 static inline void memcg_memory_event_mm(struct mm_struct *mm,
1052 enum memcg_memory_event event)
1053 {
1054 struct mem_cgroup *memcg;
1055
1056 if (mem_cgroup_disabled())
1057 return;
1058
1059 rcu_read_lock();
1060 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1061 if (likely(memcg))
1062 memcg_memory_event(memcg, event);
1063 rcu_read_unlock();
1064 }
1065
1066 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1067 void mem_cgroup_split_huge_fixup(struct page *head);
1068 #endif
1069
1070 #else /* CONFIG_MEMCG */
1071
1072 #define MEM_CGROUP_ID_SHIFT 0
1073 #define MEM_CGROUP_ID_MAX 0
1074
1075 struct mem_cgroup;
1076
1077 static inline struct mem_cgroup *page_memcg(struct page *page)
1078 {
1079 return NULL;
1080 }
1081
1082 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1083 {
1084 WARN_ON_ONCE(!rcu_read_lock_held());
1085 return NULL;
1086 }
1087
1088 static inline struct mem_cgroup *page_memcg_check(struct page *page)
1089 {
1090 return NULL;
1091 }
1092
1093 static inline bool PageMemcgKmem(struct page *page)
1094 {
1095 return false;
1096 }
1097
1098 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1099 {
1100 return true;
1101 }
1102
1103 static inline bool mem_cgroup_disabled(void)
1104 {
1105 return true;
1106 }
1107
1108 static inline void memcg_memory_event(struct mem_cgroup *memcg,
1109 enum memcg_memory_event event)
1110 {
1111 }
1112
1113 static inline void memcg_memory_event_mm(struct mm_struct *mm,
1114 enum memcg_memory_event event)
1115 {
1116 }
1117
1118 static inline unsigned long mem_cgroup_protection(struct mem_cgroup *root,
1119 struct mem_cgroup *memcg,
1120 bool in_low_reclaim)
1121 {
1122 return 0;
1123 }
1124
1125 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1126 struct mem_cgroup *memcg)
1127 {
1128 }
1129
1130 static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
1131 {
1132 return false;
1133 }
1134
1135 static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
1136 {
1137 return false;
1138 }
1139
1140 static inline int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
1141 gfp_t gfp_mask)
1142 {
1143 return 0;
1144 }
1145
1146 static inline void mem_cgroup_uncharge(struct page *page)
1147 {
1148 }
1149
1150 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
1151 {
1152 }
1153
1154 static inline void mem_cgroup_migrate(struct page *old, struct page *new)
1155 {
1156 }
1157
1158 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1159 struct pglist_data *pgdat)
1160 {
1161 return &pgdat->__lruvec;
1162 }
1163
1164 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
1165 struct pglist_data *pgdat)
1166 {
1167 return &pgdat->__lruvec;
1168 }
1169
1170 static inline bool lruvec_holds_page_lru_lock(struct page *page,
1171 struct lruvec *lruvec)
1172 {
1173 pg_data_t *pgdat = page_pgdat(page);
1174
1175 return lruvec == &pgdat->__lruvec;
1176 }
1177
1178 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1179 {
1180 return NULL;
1181 }
1182
1183 static inline bool mm_match_cgroup(struct mm_struct *mm,
1184 struct mem_cgroup *memcg)
1185 {
1186 return true;
1187 }
1188
1189 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1190 {
1191 return NULL;
1192 }
1193
1194 static inline struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
1195 {
1196 return NULL;
1197 }
1198
1199 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1200 {
1201 }
1202
1203 static inline struct lruvec *lock_page_lruvec(struct page *page)
1204 {
1205 struct pglist_data *pgdat = page_pgdat(page);
1206
1207 spin_lock(&pgdat->__lruvec.lru_lock);
1208 return &pgdat->__lruvec;
1209 }
1210
1211 static inline struct lruvec *lock_page_lruvec_irq(struct page *page)
1212 {
1213 struct pglist_data *pgdat = page_pgdat(page);
1214
1215 spin_lock_irq(&pgdat->__lruvec.lru_lock);
1216 return &pgdat->__lruvec;
1217 }
1218
1219 static inline struct lruvec *lock_page_lruvec_irqsave(struct page *page,
1220 unsigned long *flagsp)
1221 {
1222 struct pglist_data *pgdat = page_pgdat(page);
1223
1224 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1225 return &pgdat->__lruvec;
1226 }
1227
1228 static inline struct mem_cgroup *
1229 mem_cgroup_iter(struct mem_cgroup *root,
1230 struct mem_cgroup *prev,
1231 struct mem_cgroup_reclaim_cookie *reclaim)
1232 {
1233 return NULL;
1234 }
1235
1236 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1237 struct mem_cgroup *prev)
1238 {
1239 }
1240
1241 static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1242 int (*fn)(struct task_struct *, void *), void *arg)
1243 {
1244 return 0;
1245 }
1246
1247 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1248 {
1249 return 0;
1250 }
1251
1252 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1253 {
1254 WARN_ON_ONCE(id);
1255 /* XXX: This should always return root_mem_cgroup */
1256 return NULL;
1257 }
1258
1259 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1260 {
1261 return NULL;
1262 }
1263
1264 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1265 {
1266 return NULL;
1267 }
1268
1269 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1270 {
1271 return true;
1272 }
1273
1274 static inline
1275 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1276 enum lru_list lru, int zone_idx)
1277 {
1278 return 0;
1279 }
1280
1281 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1282 {
1283 return 0;
1284 }
1285
1286 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1287 {
1288 return 0;
1289 }
1290
1291 static inline void
1292 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1293 {
1294 }
1295
1296 static inline void
1297 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1298 {
1299 }
1300
1301 static inline struct mem_cgroup *lock_page_memcg(struct page *page)
1302 {
1303 return NULL;
1304 }
1305
1306 static inline void __unlock_page_memcg(struct mem_cgroup *memcg)
1307 {
1308 }
1309
1310 static inline void unlock_page_memcg(struct page *page)
1311 {
1312 }
1313
1314 static inline void mem_cgroup_handle_over_high(void)
1315 {
1316 }
1317
1318 static inline void mem_cgroup_enter_user_fault(void)
1319 {
1320 }
1321
1322 static inline void mem_cgroup_exit_user_fault(void)
1323 {
1324 }
1325
1326 static inline bool task_in_memcg_oom(struct task_struct *p)
1327 {
1328 return false;
1329 }
1330
1331 static inline bool mem_cgroup_oom_synchronize(bool wait)
1332 {
1333 return false;
1334 }
1335
1336 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1337 struct task_struct *victim, struct mem_cgroup *oom_domain)
1338 {
1339 return NULL;
1340 }
1341
1342 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1343 {
1344 }
1345
1346 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1347 {
1348 return 0;
1349 }
1350
1351 static inline unsigned long memcg_page_state_local(struct mem_cgroup *memcg,
1352 int idx)
1353 {
1354 return 0;
1355 }
1356
1357 static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1358 int idx,
1359 int nr)
1360 {
1361 }
1362
1363 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1364 int idx,
1365 int nr)
1366 {
1367 }
1368
1369 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1370 enum node_stat_item idx)
1371 {
1372 return node_page_state(lruvec_pgdat(lruvec), idx);
1373 }
1374
1375 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1376 enum node_stat_item idx)
1377 {
1378 return node_page_state(lruvec_pgdat(lruvec), idx);
1379 }
1380
1381 static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1382 enum node_stat_item idx, int val)
1383 {
1384 }
1385
1386 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1387 int val)
1388 {
1389 struct page *page = virt_to_head_page(p);
1390
1391 __mod_node_page_state(page_pgdat(page), idx, val);
1392 }
1393
1394 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1395 int val)
1396 {
1397 struct page *page = virt_to_head_page(p);
1398
1399 mod_node_page_state(page_pgdat(page), idx, val);
1400 }
1401
1402 static inline
1403 unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1404 gfp_t gfp_mask,
1405 unsigned long *total_scanned)
1406 {
1407 return 0;
1408 }
1409
1410 static inline void mem_cgroup_split_huge_fixup(struct page *head)
1411 {
1412 }
1413
1414 static inline void count_memcg_events(struct mem_cgroup *memcg,
1415 enum vm_event_item idx,
1416 unsigned long count)
1417 {
1418 }
1419
1420 static inline void __count_memcg_events(struct mem_cgroup *memcg,
1421 enum vm_event_item idx,
1422 unsigned long count)
1423 {
1424 }
1425
1426 static inline void count_memcg_page_event(struct page *page,
1427 int idx)
1428 {
1429 }
1430
1431 static inline
1432 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1433 {
1434 }
1435
1436 static inline void lruvec_memcg_debug(struct lruvec *lruvec, struct page *page)
1437 {
1438 }
1439 #endif /* CONFIG_MEMCG */
1440
1441 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1442 {
1443 __mod_lruvec_kmem_state(p, idx, 1);
1444 }
1445
1446 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1447 {
1448 __mod_lruvec_kmem_state(p, idx, -1);
1449 }
1450
1451 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1452 {
1453 struct mem_cgroup *memcg;
1454
1455 memcg = lruvec_memcg(lruvec);
1456 if (!memcg)
1457 return NULL;
1458 memcg = parent_mem_cgroup(memcg);
1459 if (!memcg)
1460 return NULL;
1461 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1462 }
1463
1464 static inline void unlock_page_lruvec(struct lruvec *lruvec)
1465 {
1466 spin_unlock(&lruvec->lru_lock);
1467 }
1468
1469 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1470 {
1471 spin_unlock_irq(&lruvec->lru_lock);
1472 }
1473
1474 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1475 unsigned long flags)
1476 {
1477 spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1478 }
1479
1480 /* Don't lock again iff page's lruvec locked */
1481 static inline struct lruvec *relock_page_lruvec_irq(struct page *page,
1482 struct lruvec *locked_lruvec)
1483 {
1484 if (locked_lruvec) {
1485 if (lruvec_holds_page_lru_lock(page, locked_lruvec))
1486 return locked_lruvec;
1487
1488 unlock_page_lruvec_irq(locked_lruvec);
1489 }
1490
1491 return lock_page_lruvec_irq(page);
1492 }
1493
1494 /* Don't lock again iff page's lruvec locked */
1495 static inline struct lruvec *relock_page_lruvec_irqsave(struct page *page,
1496 struct lruvec *locked_lruvec, unsigned long *flags)
1497 {
1498 if (locked_lruvec) {
1499 if (lruvec_holds_page_lru_lock(page, locked_lruvec))
1500 return locked_lruvec;
1501
1502 unlock_page_lruvec_irqrestore(locked_lruvec, *flags);
1503 }
1504
1505 return lock_page_lruvec_irqsave(page, flags);
1506 }
1507
1508 #ifdef CONFIG_CGROUP_WRITEBACK
1509
1510 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1511 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1512 unsigned long *pheadroom, unsigned long *pdirty,
1513 unsigned long *pwriteback);
1514
1515 void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
1516 struct bdi_writeback *wb);
1517
1518 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1519 struct bdi_writeback *wb)
1520 {
1521 if (mem_cgroup_disabled())
1522 return;
1523
1524 if (unlikely(&page_memcg(page)->css != wb->memcg_css))
1525 mem_cgroup_track_foreign_dirty_slowpath(page, wb);
1526 }
1527
1528 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1529
1530 #else /* CONFIG_CGROUP_WRITEBACK */
1531
1532 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1533 {
1534 return NULL;
1535 }
1536
1537 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1538 unsigned long *pfilepages,
1539 unsigned long *pheadroom,
1540 unsigned long *pdirty,
1541 unsigned long *pwriteback)
1542 {
1543 }
1544
1545 static inline void mem_cgroup_track_foreign_dirty(struct page *page,
1546 struct bdi_writeback *wb)
1547 {
1548 }
1549
1550 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1551 {
1552 }
1553
1554 #endif /* CONFIG_CGROUP_WRITEBACK */
1555
1556 struct sock;
1557 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1558 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1559 #ifdef CONFIG_MEMCG
1560 extern struct static_key_false memcg_sockets_enabled_key;
1561 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1562 void mem_cgroup_sk_alloc(struct sock *sk);
1563 void mem_cgroup_sk_free(struct sock *sk);
1564 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1565 {
1566 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1567 return true;
1568 do {
1569 if (time_before(jiffies, memcg->socket_pressure))
1570 return true;
1571 } while ((memcg = parent_mem_cgroup(memcg)));
1572 return false;
1573 }
1574
1575 extern int memcg_expand_shrinker_maps(int new_id);
1576
1577 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1578 int nid, int shrinker_id);
1579 #else
1580 #define mem_cgroup_sockets_enabled 0
1581 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1582 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1583 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1584 {
1585 return false;
1586 }
1587
1588 static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
1589 int nid, int shrinker_id)
1590 {
1591 }
1592 #endif
1593
1594 #ifdef CONFIG_MEMCG_KMEM
1595 int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
1596 unsigned int nr_pages);
1597 void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages);
1598 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1599 void __memcg_kmem_uncharge_page(struct page *page, int order);
1600
1601 struct obj_cgroup *get_obj_cgroup_from_current(void);
1602
1603 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1604 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1605
1606 extern struct static_key_false memcg_kmem_enabled_key;
1607
1608 extern int memcg_nr_cache_ids;
1609 void memcg_get_cache_ids(void);
1610 void memcg_put_cache_ids(void);
1611
1612 /*
1613 * Helper macro to loop through all memcg-specific caches. Callers must still
1614 * check if the cache is valid (it is either valid or NULL).
1615 * the slab_mutex must be held when looping through those caches
1616 */
1617 #define for_each_memcg_cache_index(_idx) \
1618 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
1619
1620 static inline bool memcg_kmem_enabled(void)
1621 {
1622 return static_branch_likely(&memcg_kmem_enabled_key);
1623 }
1624
1625 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1626 int order)
1627 {
1628 if (memcg_kmem_enabled())
1629 return __memcg_kmem_charge_page(page, gfp, order);
1630 return 0;
1631 }
1632
1633 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1634 {
1635 if (memcg_kmem_enabled())
1636 __memcg_kmem_uncharge_page(page, order);
1637 }
1638
1639 /*
1640 * A helper for accessing memcg's kmem_id, used for getting
1641 * corresponding LRU lists.
1642 */
1643 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1644 {
1645 return memcg ? memcg->kmemcg_id : -1;
1646 }
1647
1648 struct mem_cgroup *mem_cgroup_from_obj(void *p);
1649
1650 #else
1651
1652 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1653 int order)
1654 {
1655 return 0;
1656 }
1657
1658 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1659 {
1660 }
1661
1662 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1663 int order)
1664 {
1665 return 0;
1666 }
1667
1668 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1669 {
1670 }
1671
1672 #define for_each_memcg_cache_index(_idx) \
1673 for (; NULL; )
1674
1675 static inline bool memcg_kmem_enabled(void)
1676 {
1677 return false;
1678 }
1679
1680 static inline int memcg_cache_id(struct mem_cgroup *memcg)
1681 {
1682 return -1;
1683 }
1684
1685 static inline void memcg_get_cache_ids(void)
1686 {
1687 }
1688
1689 static inline void memcg_put_cache_ids(void)
1690 {
1691 }
1692
1693 static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1694 {
1695 return NULL;
1696 }
1697
1698 #endif /* CONFIG_MEMCG_KMEM */
1699
1700 #endif /* _LINUX_MEMCONTROL_H */