]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - include/linux/memcontrol.h
87b6688f124a7d78135d1d9dd4c641d825446ac6
[thirdparty/kernel/stable.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/kernel.h>
18 #include <linux/page_counter.h>
19 #include <linux/vmpressure.h>
20 #include <linux/eventfd.h>
21 #include <linux/mm.h>
22 #include <linux/vmstat.h>
23 #include <linux/writeback.h>
24 #include <linux/page-flags.h>
25 #include <linux/shrinker.h>
26
27 struct mem_cgroup;
28 struct obj_cgroup;
29 struct page;
30 struct mm_struct;
31 struct kmem_cache;
32
33 /* Cgroup-specific page state, on top of universal node page state */
34 enum memcg_stat_item {
35 MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS,
36 MEMCG_SOCK,
37 MEMCG_PERCPU_B,
38 MEMCG_VMALLOC,
39 MEMCG_KMEM,
40 MEMCG_ZSWAP_B,
41 MEMCG_ZSWAPPED,
42 MEMCG_NR_STAT,
43 };
44
45 enum memcg_memory_event {
46 MEMCG_LOW,
47 MEMCG_HIGH,
48 MEMCG_MAX,
49 MEMCG_OOM,
50 MEMCG_OOM_KILL,
51 MEMCG_OOM_GROUP_KILL,
52 MEMCG_SWAP_HIGH,
53 MEMCG_SWAP_MAX,
54 MEMCG_SWAP_FAIL,
55 MEMCG_NR_MEMORY_EVENTS,
56 };
57
58 struct mem_cgroup_reclaim_cookie {
59 pg_data_t *pgdat;
60 int generation;
61 };
62
63 #ifdef CONFIG_MEMCG
64
65 #define MEM_CGROUP_ID_SHIFT 16
66
67 struct mem_cgroup_id {
68 int id;
69 refcount_t ref;
70 };
71
72 struct memcg_vmstats_percpu;
73 struct memcg1_events_percpu;
74 struct memcg_vmstats;
75 struct lruvec_stats_percpu;
76 struct lruvec_stats;
77
78 struct mem_cgroup_reclaim_iter {
79 struct mem_cgroup *position;
80 /* scan generation, increased every round-trip */
81 atomic_t generation;
82 };
83
84 /*
85 * per-node information in memory controller.
86 */
87 struct mem_cgroup_per_node {
88 /* Keep the read-only fields at the start */
89 struct mem_cgroup *memcg; /* Back pointer, we cannot */
90 /* use container_of */
91
92 struct lruvec_stats_percpu __percpu *lruvec_stats_percpu;
93 struct lruvec_stats *lruvec_stats;
94 struct shrinker_info __rcu *shrinker_info;
95
96 #ifdef CONFIG_MEMCG_V1
97 /*
98 * Memcg-v1 only stuff in middle as buffer between read mostly fields
99 * and update often fields to avoid false sharing. If v1 stuff is
100 * not present, an explicit padding is needed.
101 */
102
103 struct rb_node tree_node; /* RB tree node */
104 unsigned long usage_in_excess;/* Set to the value by which */
105 /* the soft limit is exceeded*/
106 bool on_tree;
107 #else
108 CACHELINE_PADDING(_pad1_);
109 #endif
110
111 /* Fields which get updated often at the end. */
112 struct lruvec lruvec;
113 CACHELINE_PADDING(_pad2_);
114 unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
115 struct mem_cgroup_reclaim_iter iter;
116
117 #ifdef CONFIG_MEMCG_NMI_SAFETY_REQUIRES_ATOMIC
118 /* slab stats for nmi context */
119 atomic_t slab_reclaimable;
120 atomic_t slab_unreclaimable;
121 #endif
122 };
123
124 struct mem_cgroup_threshold {
125 struct eventfd_ctx *eventfd;
126 unsigned long threshold;
127 };
128
129 /* For threshold */
130 struct mem_cgroup_threshold_ary {
131 /* An array index points to threshold just below or equal to usage. */
132 int current_threshold;
133 /* Size of entries[] */
134 unsigned int size;
135 /* Array of thresholds */
136 struct mem_cgroup_threshold entries[] __counted_by(size);
137 };
138
139 struct mem_cgroup_thresholds {
140 /* Primary thresholds array */
141 struct mem_cgroup_threshold_ary *primary;
142 /*
143 * Spare threshold array.
144 * This is needed to make mem_cgroup_unregister_event() "never fail".
145 * It must be able to store at least primary->size - 1 entries.
146 */
147 struct mem_cgroup_threshold_ary *spare;
148 };
149
150 /*
151 * Remember four most recent foreign writebacks with dirty pages in this
152 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
153 * one in a given round, we're likely to catch it later if it keeps
154 * foreign-dirtying, so a fairly low count should be enough.
155 *
156 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
157 */
158 #define MEMCG_CGWB_FRN_CNT 4
159
160 struct memcg_cgwb_frn {
161 u64 bdi_id; /* bdi->id of the foreign inode */
162 int memcg_id; /* memcg->css.id of foreign inode */
163 u64 at; /* jiffies_64 at the time of dirtying */
164 struct wb_completion done; /* tracks in-flight foreign writebacks */
165 };
166
167 /*
168 * Bucket for arbitrarily byte-sized objects charged to a memory
169 * cgroup. The bucket can be reparented in one piece when the cgroup
170 * is destroyed, without having to round up the individual references
171 * of all live memory objects in the wild.
172 */
173 struct obj_cgroup {
174 struct percpu_ref refcnt;
175 struct mem_cgroup *memcg;
176 atomic_t nr_charged_bytes;
177 union {
178 struct list_head list; /* protected by objcg_lock */
179 struct rcu_head rcu;
180 };
181 };
182
183 /*
184 * The memory controller data structure. The memory controller controls both
185 * page cache and RSS per cgroup. We would eventually like to provide
186 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
187 * to help the administrator determine what knobs to tune.
188 */
189 struct mem_cgroup {
190 struct cgroup_subsys_state css;
191
192 /* Private memcg ID. Used to ID objects that outlive the cgroup */
193 struct mem_cgroup_id id;
194
195 /* Accounted resources */
196 struct page_counter memory; /* Both v1 & v2 */
197
198 union {
199 struct page_counter swap; /* v2 only */
200 struct page_counter memsw; /* v1 only */
201 };
202
203 /* registered local peak watchers */
204 struct list_head memory_peaks;
205 struct list_head swap_peaks;
206 spinlock_t peaks_lock;
207
208 /* Range enforcement for interrupt charges */
209 struct work_struct high_work;
210
211 #ifdef CONFIG_ZSWAP
212 unsigned long zswap_max;
213
214 /*
215 * Prevent pages from this memcg from being written back from zswap to
216 * swap, and from being swapped out on zswap store failures.
217 */
218 bool zswap_writeback;
219 #endif
220
221 /* vmpressure notifications */
222 struct vmpressure vmpressure;
223
224 /*
225 * Should the OOM killer kill all belonging tasks, had it kill one?
226 */
227 bool oom_group;
228
229 int swappiness;
230
231 /* memory.events and memory.events.local */
232 struct cgroup_file events_file;
233 struct cgroup_file events_local_file;
234
235 /* handle for "memory.swap.events" */
236 struct cgroup_file swap_events_file;
237
238 /* memory.stat */
239 struct memcg_vmstats *vmstats;
240
241 /* memory.events */
242 atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
243 atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
244
245 #ifdef CONFIG_MEMCG_NMI_SAFETY_REQUIRES_ATOMIC
246 /* MEMCG_KMEM for nmi context */
247 atomic_t kmem_stat;
248 #endif
249 /*
250 * Hint of reclaim pressure for socket memroy management. Note
251 * that this indicator should NOT be used in legacy cgroup mode
252 * where socket memory is accounted/charged separately.
253 */
254 unsigned long socket_pressure;
255
256 int kmemcg_id;
257 /*
258 * memcg->objcg is wiped out as a part of the objcg repaprenting
259 * process. memcg->orig_objcg preserves a pointer (and a reference)
260 * to the original objcg until the end of live of memcg.
261 */
262 struct obj_cgroup __rcu *objcg;
263 struct obj_cgroup *orig_objcg;
264 /* list of inherited objcgs, protected by objcg_lock */
265 struct list_head objcg_list;
266
267 struct memcg_vmstats_percpu __percpu *vmstats_percpu;
268
269 #ifdef CONFIG_CGROUP_WRITEBACK
270 struct list_head cgwb_list;
271 struct wb_domain cgwb_domain;
272 struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
273 #endif
274
275 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
276 struct deferred_split deferred_split_queue;
277 #endif
278
279 #ifdef CONFIG_LRU_GEN_WALKS_MMU
280 /* per-memcg mm_struct list */
281 struct lru_gen_mm_list mm_list;
282 #endif
283
284 #ifdef CONFIG_MEMCG_V1
285 /* Legacy consumer-oriented counters */
286 struct page_counter kmem; /* v1 only */
287 struct page_counter tcpmem; /* v1 only */
288
289 struct memcg1_events_percpu __percpu *events_percpu;
290
291 unsigned long soft_limit;
292
293 /* protected by memcg_oom_lock */
294 bool oom_lock;
295 int under_oom;
296
297 /* OOM-Killer disable */
298 int oom_kill_disable;
299
300 /* protect arrays of thresholds */
301 struct mutex thresholds_lock;
302
303 /* thresholds for memory usage. RCU-protected */
304 struct mem_cgroup_thresholds thresholds;
305
306 /* thresholds for mem+swap usage. RCU-protected */
307 struct mem_cgroup_thresholds memsw_thresholds;
308
309 /* For oom notifier event fd */
310 struct list_head oom_notify;
311
312 /* Legacy tcp memory accounting */
313 bool tcpmem_active;
314 int tcpmem_pressure;
315
316 /* List of events which userspace want to receive */
317 struct list_head event_list;
318 spinlock_t event_list_lock;
319 #endif /* CONFIG_MEMCG_V1 */
320
321 struct mem_cgroup_per_node *nodeinfo[];
322 };
323
324 /*
325 * size of first charge trial.
326 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the
327 * workload.
328 */
329 #define MEMCG_CHARGE_BATCH 64U
330
331 extern struct mem_cgroup *root_mem_cgroup;
332
333 enum page_memcg_data_flags {
334 /* page->memcg_data is a pointer to an slabobj_ext vector */
335 MEMCG_DATA_OBJEXTS = (1UL << 0),
336 /* page has been accounted as a non-slab kernel page */
337 MEMCG_DATA_KMEM = (1UL << 1),
338 /* the next bit after the last actual flag */
339 __NR_MEMCG_DATA_FLAGS = (1UL << 2),
340 };
341
342 #define __FIRST_OBJEXT_FLAG __NR_MEMCG_DATA_FLAGS
343
344 #else /* CONFIG_MEMCG */
345
346 #define __FIRST_OBJEXT_FLAG (1UL << 0)
347
348 #endif /* CONFIG_MEMCG */
349
350 enum objext_flags {
351 /* slabobj_ext vector failed to allocate */
352 OBJEXTS_ALLOC_FAIL = __FIRST_OBJEXT_FLAG,
353 /* the next bit after the last actual flag */
354 __NR_OBJEXTS_FLAGS = (__FIRST_OBJEXT_FLAG << 1),
355 };
356
357 #define OBJEXTS_FLAGS_MASK (__NR_OBJEXTS_FLAGS - 1)
358
359 #ifdef CONFIG_MEMCG
360
361 static inline bool folio_memcg_kmem(struct folio *folio);
362
363 /*
364 * After the initialization objcg->memcg is always pointing at
365 * a valid memcg, but can be atomically swapped to the parent memcg.
366 *
367 * The caller must ensure that the returned memcg won't be released.
368 */
369 static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
370 {
371 lockdep_assert_once(rcu_read_lock_held() || lockdep_is_held(&cgroup_mutex));
372 return READ_ONCE(objcg->memcg);
373 }
374
375 /*
376 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio
377 * @folio: Pointer to the folio.
378 *
379 * Returns a pointer to the memory cgroup associated with the folio,
380 * or NULL. This function assumes that the folio is known to have a
381 * proper memory cgroup pointer. It's not safe to call this function
382 * against some type of folios, e.g. slab folios or ex-slab folios or
383 * kmem folios.
384 */
385 static inline struct mem_cgroup *__folio_memcg(struct folio *folio)
386 {
387 unsigned long memcg_data = folio->memcg_data;
388
389 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
390 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio);
391 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio);
392
393 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
394 }
395
396 /*
397 * __folio_objcg - get the object cgroup associated with a kmem folio.
398 * @folio: Pointer to the folio.
399 *
400 * Returns a pointer to the object cgroup associated with the folio,
401 * or NULL. This function assumes that the folio is known to have a
402 * proper object cgroup pointer. It's not safe to call this function
403 * against some type of folios, e.g. slab folios or ex-slab folios or
404 * LRU folios.
405 */
406 static inline struct obj_cgroup *__folio_objcg(struct folio *folio)
407 {
408 unsigned long memcg_data = folio->memcg_data;
409
410 VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
411 VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJEXTS, folio);
412 VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio);
413
414 return (struct obj_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
415 }
416
417 /*
418 * folio_memcg - Get the memory cgroup associated with a folio.
419 * @folio: Pointer to the folio.
420 *
421 * Returns a pointer to the memory cgroup associated with the folio,
422 * or NULL. This function assumes that the folio is known to have a
423 * proper memory cgroup pointer. It's not safe to call this function
424 * against some type of folios, e.g. slab folios or ex-slab folios.
425 *
426 * For a non-kmem folio any of the following ensures folio and memcg binding
427 * stability:
428 *
429 * - the folio lock
430 * - LRU isolation
431 * - exclusive reference
432 *
433 * For a kmem folio a caller should hold an rcu read lock to protect memcg
434 * associated with a kmem folio from being released.
435 */
436 static inline struct mem_cgroup *folio_memcg(struct folio *folio)
437 {
438 if (folio_memcg_kmem(folio))
439 return obj_cgroup_memcg(__folio_objcg(folio));
440 return __folio_memcg(folio);
441 }
442
443 /*
444 * folio_memcg_charged - If a folio is charged to a memory cgroup.
445 * @folio: Pointer to the folio.
446 *
447 * Returns true if folio is charged to a memory cgroup, otherwise returns false.
448 */
449 static inline bool folio_memcg_charged(struct folio *folio)
450 {
451 return folio->memcg_data != 0;
452 }
453
454 /*
455 * folio_memcg_check - Get the memory cgroup associated with a folio.
456 * @folio: Pointer to the folio.
457 *
458 * Returns a pointer to the memory cgroup associated with the folio,
459 * or NULL. This function unlike folio_memcg() can take any folio
460 * as an argument. It has to be used in cases when it's not known if a folio
461 * has an associated memory cgroup pointer or an object cgroups vector or
462 * an object cgroup.
463 *
464 * For a non-kmem folio any of the following ensures folio and memcg binding
465 * stability:
466 *
467 * - the folio lock
468 * - LRU isolation
469 * - exclusive reference
470 *
471 * For a kmem folio a caller should hold an rcu read lock to protect memcg
472 * associated with a kmem folio from being released.
473 */
474 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio)
475 {
476 /*
477 * Because folio->memcg_data might be changed asynchronously
478 * for slabs, READ_ONCE() should be used here.
479 */
480 unsigned long memcg_data = READ_ONCE(folio->memcg_data);
481
482 if (memcg_data & MEMCG_DATA_OBJEXTS)
483 return NULL;
484
485 if (memcg_data & MEMCG_DATA_KMEM) {
486 struct obj_cgroup *objcg;
487
488 objcg = (void *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
489 return obj_cgroup_memcg(objcg);
490 }
491
492 return (struct mem_cgroup *)(memcg_data & ~OBJEXTS_FLAGS_MASK);
493 }
494
495 static inline struct mem_cgroup *page_memcg_check(struct page *page)
496 {
497 if (PageTail(page))
498 return NULL;
499 return folio_memcg_check((struct folio *)page);
500 }
501
502 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg)
503 {
504 struct mem_cgroup *memcg;
505
506 rcu_read_lock();
507 retry:
508 memcg = obj_cgroup_memcg(objcg);
509 if (unlikely(!css_tryget(&memcg->css)))
510 goto retry;
511 rcu_read_unlock();
512
513 return memcg;
514 }
515
516 /*
517 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set.
518 * @folio: Pointer to the folio.
519 *
520 * Checks if the folio has MemcgKmem flag set. The caller must ensure
521 * that the folio has an associated memory cgroup. It's not safe to call
522 * this function against some types of folios, e.g. slab folios.
523 */
524 static inline bool folio_memcg_kmem(struct folio *folio)
525 {
526 VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page);
527 VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJEXTS, folio);
528 return folio->memcg_data & MEMCG_DATA_KMEM;
529 }
530
531 static inline bool PageMemcgKmem(struct page *page)
532 {
533 return folio_memcg_kmem(page_folio(page));
534 }
535
536 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
537 {
538 return (memcg == root_mem_cgroup);
539 }
540
541 static inline bool mem_cgroup_disabled(void)
542 {
543 return !cgroup_subsys_enabled(memory_cgrp_subsys);
544 }
545
546 static inline void mem_cgroup_protection(struct mem_cgroup *root,
547 struct mem_cgroup *memcg,
548 unsigned long *min,
549 unsigned long *low)
550 {
551 *min = *low = 0;
552
553 if (mem_cgroup_disabled())
554 return;
555
556 /*
557 * There is no reclaim protection applied to a targeted reclaim.
558 * We are special casing this specific case here because
559 * mem_cgroup_calculate_protection is not robust enough to keep
560 * the protection invariant for calculated effective values for
561 * parallel reclaimers with different reclaim target. This is
562 * especially a problem for tail memcgs (as they have pages on LRU)
563 * which would want to have effective values 0 for targeted reclaim
564 * but a different value for external reclaim.
565 *
566 * Example
567 * Let's have global and A's reclaim in parallel:
568 * |
569 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
570 * |\
571 * | C (low = 1G, usage = 2.5G)
572 * B (low = 1G, usage = 0.5G)
573 *
574 * For the global reclaim
575 * A.elow = A.low
576 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
577 * C.elow = min(C.usage, C.low)
578 *
579 * With the effective values resetting we have A reclaim
580 * A.elow = 0
581 * B.elow = B.low
582 * C.elow = C.low
583 *
584 * If the global reclaim races with A's reclaim then
585 * B.elow = C.elow = 0 because children_low_usage > A.elow)
586 * is possible and reclaiming B would be violating the protection.
587 *
588 */
589 if (root == memcg)
590 return;
591
592 *min = READ_ONCE(memcg->memory.emin);
593 *low = READ_ONCE(memcg->memory.elow);
594 }
595
596 void mem_cgroup_calculate_protection(struct mem_cgroup *root,
597 struct mem_cgroup *memcg);
598
599 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target,
600 struct mem_cgroup *memcg)
601 {
602 /*
603 * The root memcg doesn't account charges, and doesn't support
604 * protection. The target memcg's protection is ignored, see
605 * mem_cgroup_calculate_protection() and mem_cgroup_protection()
606 */
607 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) ||
608 memcg == target;
609 }
610
611 static inline bool mem_cgroup_below_low(struct mem_cgroup *target,
612 struct mem_cgroup *memcg)
613 {
614 if (mem_cgroup_unprotected(target, memcg))
615 return false;
616
617 return READ_ONCE(memcg->memory.elow) >=
618 page_counter_read(&memcg->memory);
619 }
620
621 static inline bool mem_cgroup_below_min(struct mem_cgroup *target,
622 struct mem_cgroup *memcg)
623 {
624 if (mem_cgroup_unprotected(target, memcg))
625 return false;
626
627 return READ_ONCE(memcg->memory.emin) >=
628 page_counter_read(&memcg->memory);
629 }
630
631 int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp);
632
633 /**
634 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup.
635 * @folio: Folio to charge.
636 * @mm: mm context of the allocating task.
637 * @gfp: Reclaim mode.
638 *
639 * Try to charge @folio to the memcg that @mm belongs to, reclaiming
640 * pages according to @gfp if necessary. If @mm is NULL, try to
641 * charge to the active memcg.
642 *
643 * Do not use this for folios allocated for swapin.
644 *
645 * Return: 0 on success. Otherwise, an error code is returned.
646 */
647 static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm,
648 gfp_t gfp)
649 {
650 if (mem_cgroup_disabled())
651 return 0;
652 return __mem_cgroup_charge(folio, mm, gfp);
653 }
654
655 int mem_cgroup_charge_hugetlb(struct folio* folio, gfp_t gfp);
656
657 int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm,
658 gfp_t gfp, swp_entry_t entry);
659
660 void __mem_cgroup_uncharge(struct folio *folio);
661
662 /**
663 * mem_cgroup_uncharge - Uncharge a folio.
664 * @folio: Folio to uncharge.
665 *
666 * Uncharge a folio previously charged with mem_cgroup_charge().
667 */
668 static inline void mem_cgroup_uncharge(struct folio *folio)
669 {
670 if (mem_cgroup_disabled())
671 return;
672 __mem_cgroup_uncharge(folio);
673 }
674
675 void __mem_cgroup_uncharge_folios(struct folio_batch *folios);
676 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios)
677 {
678 if (mem_cgroup_disabled())
679 return;
680 __mem_cgroup_uncharge_folios(folios);
681 }
682
683 void mem_cgroup_replace_folio(struct folio *old, struct folio *new);
684 void mem_cgroup_migrate(struct folio *old, struct folio *new);
685
686 /**
687 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
688 * @memcg: memcg of the wanted lruvec
689 * @pgdat: pglist_data
690 *
691 * Returns the lru list vector holding pages for a given @memcg &
692 * @pgdat combination. This can be the node lruvec, if the memory
693 * controller is disabled.
694 */
695 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
696 struct pglist_data *pgdat)
697 {
698 struct mem_cgroup_per_node *mz;
699 struct lruvec *lruvec;
700
701 if (mem_cgroup_disabled()) {
702 lruvec = &pgdat->__lruvec;
703 goto out;
704 }
705
706 if (!memcg)
707 memcg = root_mem_cgroup;
708
709 mz = memcg->nodeinfo[pgdat->node_id];
710 lruvec = &mz->lruvec;
711 out:
712 /*
713 * Since a node can be onlined after the mem_cgroup was created,
714 * we have to be prepared to initialize lruvec->pgdat here;
715 * and if offlined then reonlined, we need to reinitialize it.
716 */
717 if (unlikely(lruvec->pgdat != pgdat))
718 lruvec->pgdat = pgdat;
719 return lruvec;
720 }
721
722 /**
723 * folio_lruvec - return lruvec for isolating/putting an LRU folio
724 * @folio: Pointer to the folio.
725 *
726 * This function relies on folio->mem_cgroup being stable.
727 */
728 static inline struct lruvec *folio_lruvec(struct folio *folio)
729 {
730 struct mem_cgroup *memcg = folio_memcg(folio);
731
732 VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio);
733 return mem_cgroup_lruvec(memcg, folio_pgdat(folio));
734 }
735
736 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
737
738 struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
739
740 struct mem_cgroup *get_mem_cgroup_from_current(void);
741
742 struct mem_cgroup *get_mem_cgroup_from_folio(struct folio *folio);
743
744 struct lruvec *folio_lruvec_lock(struct folio *folio);
745 struct lruvec *folio_lruvec_lock_irq(struct folio *folio);
746 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
747 unsigned long *flags);
748
749 #ifdef CONFIG_DEBUG_VM
750 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio);
751 #else
752 static inline
753 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
754 {
755 }
756 #endif
757
758 static inline
759 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
760 return css ? container_of(css, struct mem_cgroup, css) : NULL;
761 }
762
763 static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
764 {
765 return percpu_ref_tryget(&objcg->refcnt);
766 }
767
768 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
769 {
770 percpu_ref_get(&objcg->refcnt);
771 }
772
773 static inline void obj_cgroup_get_many(struct obj_cgroup *objcg,
774 unsigned long nr)
775 {
776 percpu_ref_get_many(&objcg->refcnt, nr);
777 }
778
779 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
780 {
781 if (objcg)
782 percpu_ref_put(&objcg->refcnt);
783 }
784
785 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg)
786 {
787 return !memcg || css_tryget(&memcg->css);
788 }
789
790 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg)
791 {
792 return !memcg || css_tryget_online(&memcg->css);
793 }
794
795 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
796 {
797 if (memcg)
798 css_put(&memcg->css);
799 }
800
801 #define mem_cgroup_from_counter(counter, member) \
802 container_of(counter, struct mem_cgroup, member)
803
804 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
805 struct mem_cgroup *,
806 struct mem_cgroup_reclaim_cookie *);
807 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
808 void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
809 int (*)(struct task_struct *, void *), void *arg);
810
811 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
812 {
813 if (mem_cgroup_disabled())
814 return 0;
815
816 return memcg->id.id;
817 }
818 struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
819
820 #ifdef CONFIG_SHRINKER_DEBUG
821 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
822 {
823 return memcg ? cgroup_ino(memcg->css.cgroup) : 0;
824 }
825
826 struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino);
827 #endif
828
829 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
830 {
831 return mem_cgroup_from_css(seq_css(m));
832 }
833
834 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
835 {
836 struct mem_cgroup_per_node *mz;
837
838 if (mem_cgroup_disabled())
839 return NULL;
840
841 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
842 return mz->memcg;
843 }
844
845 /**
846 * parent_mem_cgroup - find the accounting parent of a memcg
847 * @memcg: memcg whose parent to find
848 *
849 * Returns the parent memcg, or NULL if this is the root.
850 */
851 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
852 {
853 return mem_cgroup_from_css(memcg->css.parent);
854 }
855
856 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
857 struct mem_cgroup *root)
858 {
859 if (root == memcg)
860 return true;
861 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
862 }
863
864 static inline bool mm_match_cgroup(struct mm_struct *mm,
865 struct mem_cgroup *memcg)
866 {
867 struct mem_cgroup *task_memcg;
868 bool match = false;
869
870 rcu_read_lock();
871 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
872 if (task_memcg)
873 match = mem_cgroup_is_descendant(task_memcg, memcg);
874 rcu_read_unlock();
875 return match;
876 }
877
878 struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio);
879 ino_t page_cgroup_ino(struct page *page);
880
881 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
882 {
883 if (mem_cgroup_disabled())
884 return true;
885 return !!(memcg->css.flags & CSS_ONLINE);
886 }
887
888 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
889 int zid, int nr_pages);
890
891 static inline
892 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
893 enum lru_list lru, int zone_idx)
894 {
895 struct mem_cgroup_per_node *mz;
896
897 mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
898 return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
899 }
900
901 void mem_cgroup_handle_over_high(gfp_t gfp_mask);
902
903 unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
904
905 unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
906
907 void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
908 struct task_struct *p);
909
910 void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
911
912 struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
913 struct mem_cgroup *oom_domain);
914 void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
915
916 /* idx can be of type enum memcg_stat_item or node_stat_item */
917 void mod_memcg_state(struct mem_cgroup *memcg,
918 enum memcg_stat_item idx, int val);
919
920 static inline void mod_memcg_page_state(struct page *page,
921 enum memcg_stat_item idx, int val)
922 {
923 struct mem_cgroup *memcg;
924
925 if (mem_cgroup_disabled())
926 return;
927
928 rcu_read_lock();
929 memcg = folio_memcg(page_folio(page));
930 if (memcg)
931 mod_memcg_state(memcg, idx, val);
932 rcu_read_unlock();
933 }
934
935 unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx);
936 unsigned long lruvec_page_state(struct lruvec *lruvec, enum node_stat_item idx);
937 unsigned long lruvec_page_state_local(struct lruvec *lruvec,
938 enum node_stat_item idx);
939
940 void mem_cgroup_flush_stats(struct mem_cgroup *memcg);
941 void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg);
942
943 void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
944
945 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
946 int val)
947 {
948 unsigned long flags;
949
950 local_irq_save(flags);
951 __mod_lruvec_kmem_state(p, idx, val);
952 local_irq_restore(flags);
953 }
954
955 void count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
956 unsigned long count);
957
958 static inline void count_memcg_folio_events(struct folio *folio,
959 enum vm_event_item idx, unsigned long nr)
960 {
961 struct mem_cgroup *memcg = folio_memcg(folio);
962
963 if (memcg)
964 count_memcg_events(memcg, idx, nr);
965 }
966
967 static inline void count_memcg_events_mm(struct mm_struct *mm,
968 enum vm_event_item idx, unsigned long count)
969 {
970 struct mem_cgroup *memcg;
971
972 if (mem_cgroup_disabled())
973 return;
974
975 rcu_read_lock();
976 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
977 if (likely(memcg))
978 count_memcg_events(memcg, idx, count);
979 rcu_read_unlock();
980 }
981
982 static inline void count_memcg_event_mm(struct mm_struct *mm,
983 enum vm_event_item idx)
984 {
985 count_memcg_events_mm(mm, idx, 1);
986 }
987
988 static inline void memcg_memory_event(struct mem_cgroup *memcg,
989 enum memcg_memory_event event)
990 {
991 bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
992 event == MEMCG_SWAP_FAIL;
993
994 atomic_long_inc(&memcg->memory_events_local[event]);
995 if (!swap_event)
996 cgroup_file_notify(&memcg->events_local_file);
997
998 do {
999 atomic_long_inc(&memcg->memory_events[event]);
1000 if (swap_event)
1001 cgroup_file_notify(&memcg->swap_events_file);
1002 else
1003 cgroup_file_notify(&memcg->events_file);
1004
1005 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1006 break;
1007 if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1008 break;
1009 } while ((memcg = parent_mem_cgroup(memcg)) &&
1010 !mem_cgroup_is_root(memcg));
1011 }
1012
1013 static inline void memcg_memory_event_mm(struct mm_struct *mm,
1014 enum memcg_memory_event event)
1015 {
1016 struct mem_cgroup *memcg;
1017
1018 if (mem_cgroup_disabled())
1019 return;
1020
1021 rcu_read_lock();
1022 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1023 if (likely(memcg))
1024 memcg_memory_event(memcg, event);
1025 rcu_read_unlock();
1026 }
1027
1028 void split_page_memcg(struct page *first, unsigned order);
1029 void folio_split_memcg_refs(struct folio *folio, unsigned old_order,
1030 unsigned new_order);
1031
1032 static inline u64 cgroup_id_from_mm(struct mm_struct *mm)
1033 {
1034 struct mem_cgroup *memcg;
1035 u64 id;
1036
1037 if (mem_cgroup_disabled())
1038 return 0;
1039
1040 rcu_read_lock();
1041 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1042 if (!memcg)
1043 memcg = root_mem_cgroup;
1044 id = cgroup_id(memcg->css.cgroup);
1045 rcu_read_unlock();
1046 return id;
1047 }
1048
1049 extern int mem_cgroup_init(void);
1050 #else /* CONFIG_MEMCG */
1051
1052 #define MEM_CGROUP_ID_SHIFT 0
1053
1054 static inline struct mem_cgroup *folio_memcg(struct folio *folio)
1055 {
1056 return NULL;
1057 }
1058
1059 static inline bool folio_memcg_charged(struct folio *folio)
1060 {
1061 return false;
1062 }
1063
1064 static inline struct mem_cgroup *folio_memcg_check(struct folio *folio)
1065 {
1066 return NULL;
1067 }
1068
1069 static inline struct mem_cgroup *page_memcg_check(struct page *page)
1070 {
1071 return NULL;
1072 }
1073
1074 static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg)
1075 {
1076 return NULL;
1077 }
1078
1079 static inline bool folio_memcg_kmem(struct folio *folio)
1080 {
1081 return false;
1082 }
1083
1084 static inline bool PageMemcgKmem(struct page *page)
1085 {
1086 return false;
1087 }
1088
1089 static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1090 {
1091 return true;
1092 }
1093
1094 static inline bool mem_cgroup_disabled(void)
1095 {
1096 return true;
1097 }
1098
1099 static inline void memcg_memory_event(struct mem_cgroup *memcg,
1100 enum memcg_memory_event event)
1101 {
1102 }
1103
1104 static inline void memcg_memory_event_mm(struct mm_struct *mm,
1105 enum memcg_memory_event event)
1106 {
1107 }
1108
1109 static inline void mem_cgroup_protection(struct mem_cgroup *root,
1110 struct mem_cgroup *memcg,
1111 unsigned long *min,
1112 unsigned long *low)
1113 {
1114 *min = *low = 0;
1115 }
1116
1117 static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1118 struct mem_cgroup *memcg)
1119 {
1120 }
1121
1122 static inline bool mem_cgroup_unprotected(struct mem_cgroup *target,
1123 struct mem_cgroup *memcg)
1124 {
1125 return true;
1126 }
1127 static inline bool mem_cgroup_below_low(struct mem_cgroup *target,
1128 struct mem_cgroup *memcg)
1129 {
1130 return false;
1131 }
1132
1133 static inline bool mem_cgroup_below_min(struct mem_cgroup *target,
1134 struct mem_cgroup *memcg)
1135 {
1136 return false;
1137 }
1138
1139 static inline int mem_cgroup_charge(struct folio *folio,
1140 struct mm_struct *mm, gfp_t gfp)
1141 {
1142 return 0;
1143 }
1144
1145 static inline int mem_cgroup_charge_hugetlb(struct folio* folio, gfp_t gfp)
1146 {
1147 return 0;
1148 }
1149
1150 static inline int mem_cgroup_swapin_charge_folio(struct folio *folio,
1151 struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
1152 {
1153 return 0;
1154 }
1155
1156 static inline void mem_cgroup_uncharge(struct folio *folio)
1157 {
1158 }
1159
1160 static inline void mem_cgroup_uncharge_folios(struct folio_batch *folios)
1161 {
1162 }
1163
1164 static inline void mem_cgroup_replace_folio(struct folio *old,
1165 struct folio *new)
1166 {
1167 }
1168
1169 static inline void mem_cgroup_migrate(struct folio *old, struct folio *new)
1170 {
1171 }
1172
1173 static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1174 struct pglist_data *pgdat)
1175 {
1176 return &pgdat->__lruvec;
1177 }
1178
1179 static inline struct lruvec *folio_lruvec(struct folio *folio)
1180 {
1181 struct pglist_data *pgdat = folio_pgdat(folio);
1182 return &pgdat->__lruvec;
1183 }
1184
1185 static inline
1186 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
1187 {
1188 }
1189
1190 static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1191 {
1192 return NULL;
1193 }
1194
1195 static inline bool mm_match_cgroup(struct mm_struct *mm,
1196 struct mem_cgroup *memcg)
1197 {
1198 return true;
1199 }
1200
1201 static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1202 {
1203 return NULL;
1204 }
1205
1206 static inline struct mem_cgroup *get_mem_cgroup_from_current(void)
1207 {
1208 return NULL;
1209 }
1210
1211 static inline struct mem_cgroup *get_mem_cgroup_from_folio(struct folio *folio)
1212 {
1213 return NULL;
1214 }
1215
1216 static inline
1217 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css)
1218 {
1219 return NULL;
1220 }
1221
1222 static inline void obj_cgroup_get(struct obj_cgroup *objcg)
1223 {
1224 }
1225
1226 static inline void obj_cgroup_put(struct obj_cgroup *objcg)
1227 {
1228 }
1229
1230 static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg)
1231 {
1232 return true;
1233 }
1234
1235 static inline bool mem_cgroup_tryget_online(struct mem_cgroup *memcg)
1236 {
1237 return true;
1238 }
1239
1240 static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1241 {
1242 }
1243
1244 static inline struct lruvec *folio_lruvec_lock(struct folio *folio)
1245 {
1246 struct pglist_data *pgdat = folio_pgdat(folio);
1247
1248 spin_lock(&pgdat->__lruvec.lru_lock);
1249 return &pgdat->__lruvec;
1250 }
1251
1252 static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio)
1253 {
1254 struct pglist_data *pgdat = folio_pgdat(folio);
1255
1256 spin_lock_irq(&pgdat->__lruvec.lru_lock);
1257 return &pgdat->__lruvec;
1258 }
1259
1260 static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
1261 unsigned long *flagsp)
1262 {
1263 struct pglist_data *pgdat = folio_pgdat(folio);
1264
1265 spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1266 return &pgdat->__lruvec;
1267 }
1268
1269 static inline struct mem_cgroup *
1270 mem_cgroup_iter(struct mem_cgroup *root,
1271 struct mem_cgroup *prev,
1272 struct mem_cgroup_reclaim_cookie *reclaim)
1273 {
1274 return NULL;
1275 }
1276
1277 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1278 struct mem_cgroup *prev)
1279 {
1280 }
1281
1282 static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1283 int (*fn)(struct task_struct *, void *), void *arg)
1284 {
1285 }
1286
1287 static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1288 {
1289 return 0;
1290 }
1291
1292 static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1293 {
1294 WARN_ON_ONCE(id);
1295 /* XXX: This should always return root_mem_cgroup */
1296 return NULL;
1297 }
1298
1299 #ifdef CONFIG_SHRINKER_DEBUG
1300 static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
1301 {
1302 return 0;
1303 }
1304
1305 static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino)
1306 {
1307 return NULL;
1308 }
1309 #endif
1310
1311 static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1312 {
1313 return NULL;
1314 }
1315
1316 static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1317 {
1318 return NULL;
1319 }
1320
1321 static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1322 {
1323 return true;
1324 }
1325
1326 static inline
1327 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1328 enum lru_list lru, int zone_idx)
1329 {
1330 return 0;
1331 }
1332
1333 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1334 {
1335 return 0;
1336 }
1337
1338 static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1339 {
1340 return 0;
1341 }
1342
1343 static inline void
1344 mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1345 {
1346 }
1347
1348 static inline void
1349 mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1350 {
1351 }
1352
1353 static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask)
1354 {
1355 }
1356
1357 static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1358 struct task_struct *victim, struct mem_cgroup *oom_domain)
1359 {
1360 return NULL;
1361 }
1362
1363 static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1364 {
1365 }
1366
1367 static inline void mod_memcg_state(struct mem_cgroup *memcg,
1368 enum memcg_stat_item idx,
1369 int nr)
1370 {
1371 }
1372
1373 static inline void mod_memcg_page_state(struct page *page,
1374 enum memcg_stat_item idx, int val)
1375 {
1376 }
1377
1378 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1379 {
1380 return 0;
1381 }
1382
1383 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1384 enum node_stat_item idx)
1385 {
1386 return node_page_state(lruvec_pgdat(lruvec), idx);
1387 }
1388
1389 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1390 enum node_stat_item idx)
1391 {
1392 return node_page_state(lruvec_pgdat(lruvec), idx);
1393 }
1394
1395 static inline void mem_cgroup_flush_stats(struct mem_cgroup *memcg)
1396 {
1397 }
1398
1399 static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup *memcg)
1400 {
1401 }
1402
1403 static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1404 int val)
1405 {
1406 struct page *page = virt_to_head_page(p);
1407
1408 __mod_node_page_state(page_pgdat(page), idx, val);
1409 }
1410
1411 static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1412 int val)
1413 {
1414 struct page *page = virt_to_head_page(p);
1415
1416 mod_node_page_state(page_pgdat(page), idx, val);
1417 }
1418
1419 static inline void count_memcg_events(struct mem_cgroup *memcg,
1420 enum vm_event_item idx,
1421 unsigned long count)
1422 {
1423 }
1424
1425 static inline void count_memcg_folio_events(struct folio *folio,
1426 enum vm_event_item idx, unsigned long nr)
1427 {
1428 }
1429
1430 static inline void count_memcg_events_mm(struct mm_struct *mm,
1431 enum vm_event_item idx, unsigned long count)
1432 {
1433 }
1434
1435 static inline
1436 void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1437 {
1438 }
1439
1440 static inline void split_page_memcg(struct page *first, unsigned order)
1441 {
1442 }
1443
1444 static inline void folio_split_memcg_refs(struct folio *folio,
1445 unsigned old_order, unsigned new_order)
1446 {
1447 }
1448
1449 static inline u64 cgroup_id_from_mm(struct mm_struct *mm)
1450 {
1451 return 0;
1452 }
1453
1454 static inline int mem_cgroup_init(void) { return 0; }
1455 #endif /* CONFIG_MEMCG */
1456
1457 /*
1458 * Extended information for slab objects stored as an array in page->memcg_data
1459 * if MEMCG_DATA_OBJEXTS is set.
1460 */
1461 struct slabobj_ext {
1462 #ifdef CONFIG_MEMCG
1463 struct obj_cgroup *objcg;
1464 #endif
1465 #ifdef CONFIG_MEM_ALLOC_PROFILING
1466 union codetag_ref ref;
1467 #endif
1468 } __aligned(8);
1469
1470 static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1471 {
1472 __mod_lruvec_kmem_state(p, idx, 1);
1473 }
1474
1475 static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1476 {
1477 __mod_lruvec_kmem_state(p, idx, -1);
1478 }
1479
1480 static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1481 {
1482 struct mem_cgroup *memcg;
1483
1484 memcg = lruvec_memcg(lruvec);
1485 if (!memcg)
1486 return NULL;
1487 memcg = parent_mem_cgroup(memcg);
1488 if (!memcg)
1489 return NULL;
1490 return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1491 }
1492
1493 static inline void unlock_page_lruvec(struct lruvec *lruvec)
1494 {
1495 spin_unlock(&lruvec->lru_lock);
1496 }
1497
1498 static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1499 {
1500 spin_unlock_irq(&lruvec->lru_lock);
1501 }
1502
1503 static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1504 unsigned long flags)
1505 {
1506 spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1507 }
1508
1509 /* Test requires a stable folio->memcg binding, see folio_memcg() */
1510 static inline bool folio_matches_lruvec(struct folio *folio,
1511 struct lruvec *lruvec)
1512 {
1513 return lruvec_pgdat(lruvec) == folio_pgdat(folio) &&
1514 lruvec_memcg(lruvec) == folio_memcg(folio);
1515 }
1516
1517 /* Don't lock again iff page's lruvec locked */
1518 static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio,
1519 struct lruvec *locked_lruvec)
1520 {
1521 if (locked_lruvec) {
1522 if (folio_matches_lruvec(folio, locked_lruvec))
1523 return locked_lruvec;
1524
1525 unlock_page_lruvec_irq(locked_lruvec);
1526 }
1527
1528 return folio_lruvec_lock_irq(folio);
1529 }
1530
1531 /* Don't lock again iff folio's lruvec locked */
1532 static inline void folio_lruvec_relock_irqsave(struct folio *folio,
1533 struct lruvec **lruvecp, unsigned long *flags)
1534 {
1535 if (*lruvecp) {
1536 if (folio_matches_lruvec(folio, *lruvecp))
1537 return;
1538
1539 unlock_page_lruvec_irqrestore(*lruvecp, *flags);
1540 }
1541
1542 *lruvecp = folio_lruvec_lock_irqsave(folio, flags);
1543 }
1544
1545 #ifdef CONFIG_CGROUP_WRITEBACK
1546
1547 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1548 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1549 unsigned long *pheadroom, unsigned long *pdirty,
1550 unsigned long *pwriteback);
1551
1552 void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio,
1553 struct bdi_writeback *wb);
1554
1555 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1556 struct bdi_writeback *wb)
1557 {
1558 struct mem_cgroup *memcg;
1559
1560 if (mem_cgroup_disabled())
1561 return;
1562
1563 memcg = folio_memcg(folio);
1564 if (unlikely(memcg && &memcg->css != wb->memcg_css))
1565 mem_cgroup_track_foreign_dirty_slowpath(folio, wb);
1566 }
1567
1568 void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1569
1570 #else /* CONFIG_CGROUP_WRITEBACK */
1571
1572 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1573 {
1574 return NULL;
1575 }
1576
1577 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1578 unsigned long *pfilepages,
1579 unsigned long *pheadroom,
1580 unsigned long *pdirty,
1581 unsigned long *pwriteback)
1582 {
1583 }
1584
1585 static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1586 struct bdi_writeback *wb)
1587 {
1588 }
1589
1590 static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1591 {
1592 }
1593
1594 #endif /* CONFIG_CGROUP_WRITEBACK */
1595
1596 struct sock;
1597 bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages,
1598 gfp_t gfp_mask);
1599 void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1600 #ifdef CONFIG_MEMCG
1601 extern struct static_key_false memcg_sockets_enabled_key;
1602 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1603 void mem_cgroup_sk_alloc(struct sock *sk);
1604 void mem_cgroup_sk_free(struct sock *sk);
1605 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1606 {
1607 #ifdef CONFIG_MEMCG_V1
1608 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1609 return !!memcg->tcpmem_pressure;
1610 #endif /* CONFIG_MEMCG_V1 */
1611 do {
1612 if (time_before(jiffies, READ_ONCE(memcg->socket_pressure)))
1613 return true;
1614 } while ((memcg = parent_mem_cgroup(memcg)));
1615 return false;
1616 }
1617
1618 int alloc_shrinker_info(struct mem_cgroup *memcg);
1619 void free_shrinker_info(struct mem_cgroup *memcg);
1620 void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id);
1621 void reparent_shrinker_deferred(struct mem_cgroup *memcg);
1622 #else
1623 #define mem_cgroup_sockets_enabled 0
1624 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1625 static inline void mem_cgroup_sk_free(struct sock *sk) { };
1626 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1627 {
1628 return false;
1629 }
1630
1631 static inline void set_shrinker_bit(struct mem_cgroup *memcg,
1632 int nid, int shrinker_id)
1633 {
1634 }
1635 #endif
1636
1637 #ifdef CONFIG_MEMCG
1638 bool mem_cgroup_kmem_disabled(void);
1639 int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1640 void __memcg_kmem_uncharge_page(struct page *page, int order);
1641
1642 /*
1643 * The returned objcg pointer is safe to use without additional
1644 * protection within a scope. The scope is defined either by
1645 * the current task (similar to the "current" global variable)
1646 * or by set_active_memcg() pair.
1647 * Please, use obj_cgroup_get() to get a reference if the pointer
1648 * needs to be used outside of the local scope.
1649 */
1650 struct obj_cgroup *current_obj_cgroup(void);
1651 struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio);
1652
1653 static inline struct obj_cgroup *get_obj_cgroup_from_current(void)
1654 {
1655 struct obj_cgroup *objcg = current_obj_cgroup();
1656
1657 if (objcg)
1658 obj_cgroup_get(objcg);
1659
1660 return objcg;
1661 }
1662
1663 int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1664 void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1665
1666 extern struct static_key_false memcg_bpf_enabled_key;
1667 static inline bool memcg_bpf_enabled(void)
1668 {
1669 return static_branch_likely(&memcg_bpf_enabled_key);
1670 }
1671
1672 extern struct static_key_false memcg_kmem_online_key;
1673
1674 static inline bool memcg_kmem_online(void)
1675 {
1676 return static_branch_likely(&memcg_kmem_online_key);
1677 }
1678
1679 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1680 int order)
1681 {
1682 if (memcg_kmem_online())
1683 return __memcg_kmem_charge_page(page, gfp, order);
1684 return 0;
1685 }
1686
1687 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1688 {
1689 if (memcg_kmem_online())
1690 __memcg_kmem_uncharge_page(page, order);
1691 }
1692
1693 /*
1694 * A helper for accessing memcg's kmem_id, used for getting
1695 * corresponding LRU lists.
1696 */
1697 static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1698 {
1699 return memcg ? memcg->kmemcg_id : -1;
1700 }
1701
1702 struct mem_cgroup *mem_cgroup_from_slab_obj(void *p);
1703
1704 static inline void count_objcg_events(struct obj_cgroup *objcg,
1705 enum vm_event_item idx,
1706 unsigned long count)
1707 {
1708 struct mem_cgroup *memcg;
1709
1710 if (!memcg_kmem_online())
1711 return;
1712
1713 rcu_read_lock();
1714 memcg = obj_cgroup_memcg(objcg);
1715 count_memcg_events(memcg, idx, count);
1716 rcu_read_unlock();
1717 }
1718
1719 bool mem_cgroup_node_allowed(struct mem_cgroup *memcg, int nid);
1720
1721 #else
1722 static inline bool mem_cgroup_kmem_disabled(void)
1723 {
1724 return true;
1725 }
1726
1727 static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1728 int order)
1729 {
1730 return 0;
1731 }
1732
1733 static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1734 {
1735 }
1736
1737 static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1738 int order)
1739 {
1740 return 0;
1741 }
1742
1743 static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1744 {
1745 }
1746
1747 static inline struct obj_cgroup *get_obj_cgroup_from_folio(struct folio *folio)
1748 {
1749 return NULL;
1750 }
1751
1752 static inline bool memcg_bpf_enabled(void)
1753 {
1754 return false;
1755 }
1756
1757 static inline bool memcg_kmem_online(void)
1758 {
1759 return false;
1760 }
1761
1762 static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1763 {
1764 return -1;
1765 }
1766
1767 static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p)
1768 {
1769 return NULL;
1770 }
1771
1772 static inline void count_objcg_events(struct obj_cgroup *objcg,
1773 enum vm_event_item idx,
1774 unsigned long count)
1775 {
1776 }
1777
1778 static inline ino_t page_cgroup_ino(struct page *page)
1779 {
1780 return 0;
1781 }
1782
1783 static inline bool mem_cgroup_node_allowed(struct mem_cgroup *memcg, int nid)
1784 {
1785 return true;
1786 }
1787 #endif /* CONFIG_MEMCG */
1788
1789 #if defined(CONFIG_MEMCG) && defined(CONFIG_ZSWAP)
1790 bool obj_cgroup_may_zswap(struct obj_cgroup *objcg);
1791 void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size);
1792 void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size);
1793 bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg);
1794 #else
1795 static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg)
1796 {
1797 return true;
1798 }
1799 static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg,
1800 size_t size)
1801 {
1802 }
1803 static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg,
1804 size_t size)
1805 {
1806 }
1807 static inline bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup *memcg)
1808 {
1809 /* if zswap is disabled, do not block pages going to the swapping device */
1810 return true;
1811 }
1812 #endif
1813
1814
1815 /* Cgroup v1-related declarations */
1816
1817 #ifdef CONFIG_MEMCG_V1
1818 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order,
1819 gfp_t gfp_mask,
1820 unsigned long *total_scanned);
1821
1822 bool mem_cgroup_oom_synchronize(bool wait);
1823
1824 static inline bool task_in_memcg_oom(struct task_struct *p)
1825 {
1826 return p->memcg_in_oom;
1827 }
1828
1829 static inline void mem_cgroup_enter_user_fault(void)
1830 {
1831 WARN_ON(current->in_user_fault);
1832 current->in_user_fault = 1;
1833 }
1834
1835 static inline void mem_cgroup_exit_user_fault(void)
1836 {
1837 WARN_ON(!current->in_user_fault);
1838 current->in_user_fault = 0;
1839 }
1840
1841 void memcg1_swapout(struct folio *folio, swp_entry_t entry);
1842 void memcg1_swapin(swp_entry_t entry, unsigned int nr_pages);
1843
1844 #else /* CONFIG_MEMCG_V1 */
1845 static inline
1846 unsigned long memcg1_soft_limit_reclaim(pg_data_t *pgdat, int order,
1847 gfp_t gfp_mask,
1848 unsigned long *total_scanned)
1849 {
1850 return 0;
1851 }
1852
1853 static inline bool task_in_memcg_oom(struct task_struct *p)
1854 {
1855 return false;
1856 }
1857
1858 static inline bool mem_cgroup_oom_synchronize(bool wait)
1859 {
1860 return false;
1861 }
1862
1863 static inline void mem_cgroup_enter_user_fault(void)
1864 {
1865 }
1866
1867 static inline void mem_cgroup_exit_user_fault(void)
1868 {
1869 }
1870
1871 static inline void memcg1_swapout(struct folio *folio, swp_entry_t entry)
1872 {
1873 }
1874
1875 static inline void memcg1_swapin(swp_entry_t entry, unsigned int nr_pages)
1876 {
1877 }
1878
1879 #endif /* CONFIG_MEMCG_V1 */
1880
1881 #endif /* _LINUX_MEMCONTROL_H */