Line data Source code
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_VMALLOC,
37 : MEMCG_KMEM,
38 : MEMCG_ZSWAP_B,
39 : MEMCG_ZSWAPPED,
40 : MEMCG_NR_STAT,
41 : };
42 :
43 : enum memcg_memory_event {
44 : MEMCG_LOW,
45 : MEMCG_HIGH,
46 : MEMCG_MAX,
47 : MEMCG_OOM,
48 : MEMCG_OOM_KILL,
49 : MEMCG_OOM_GROUP_KILL,
50 : MEMCG_SWAP_HIGH,
51 : MEMCG_SWAP_MAX,
52 : MEMCG_SWAP_FAIL,
53 : MEMCG_NR_MEMORY_EVENTS,
54 : };
55 :
56 : struct mem_cgroup_reclaim_cookie {
57 : pg_data_t *pgdat;
58 : unsigned int generation;
59 : };
60 :
61 : #ifdef CONFIG_MEMCG
62 :
63 : #define MEM_CGROUP_ID_SHIFT 16
64 : #define MEM_CGROUP_ID_MAX USHRT_MAX
65 :
66 : struct mem_cgroup_id {
67 : int id;
68 : refcount_t ref;
69 : };
70 :
71 : /*
72 : * Per memcg event counter is incremented at every pagein/pageout. With THP,
73 : * it will be incremented by the number of pages. This counter is used
74 : * to trigger some periodic events. This is straightforward and better
75 : * than using jiffies etc. to handle periodic memcg event.
76 : */
77 : enum mem_cgroup_events_target {
78 : MEM_CGROUP_TARGET_THRESH,
79 : MEM_CGROUP_TARGET_SOFTLIMIT,
80 : MEM_CGROUP_NTARGETS,
81 : };
82 :
83 : struct memcg_vmstats_percpu;
84 : struct memcg_vmstats;
85 :
86 : struct mem_cgroup_reclaim_iter {
87 : struct mem_cgroup *position;
88 : /* scan generation, increased every round-trip */
89 : unsigned int generation;
90 : };
91 :
92 : /*
93 : * Bitmap and deferred work of shrinker::id corresponding to memcg-aware
94 : * shrinkers, which have elements charged to this memcg.
95 : */
96 : struct shrinker_info {
97 : struct rcu_head rcu;
98 : atomic_long_t *nr_deferred;
99 : unsigned long *map;
100 : int map_nr_max;
101 : };
102 :
103 : struct lruvec_stats_percpu {
104 : /* Local (CPU and cgroup) state */
105 : long state[NR_VM_NODE_STAT_ITEMS];
106 :
107 : /* Delta calculation for lockless upward propagation */
108 : long state_prev[NR_VM_NODE_STAT_ITEMS];
109 : };
110 :
111 : struct lruvec_stats {
112 : /* Aggregated (CPU and subtree) state */
113 : long state[NR_VM_NODE_STAT_ITEMS];
114 :
115 : /* Pending child counts during tree propagation */
116 : long state_pending[NR_VM_NODE_STAT_ITEMS];
117 : };
118 :
119 : /*
120 : * per-node information in memory controller.
121 : */
122 : struct mem_cgroup_per_node {
123 : struct lruvec lruvec;
124 :
125 : struct lruvec_stats_percpu __percpu *lruvec_stats_percpu;
126 : struct lruvec_stats lruvec_stats;
127 :
128 : unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
129 :
130 : struct mem_cgroup_reclaim_iter iter;
131 :
132 : struct shrinker_info __rcu *shrinker_info;
133 :
134 : struct rb_node tree_node; /* RB tree node */
135 : unsigned long usage_in_excess;/* Set to the value by which */
136 : /* the soft limit is exceeded*/
137 : bool on_tree;
138 : struct mem_cgroup *memcg; /* Back pointer, we cannot */
139 : /* use container_of */
140 : };
141 :
142 : struct mem_cgroup_threshold {
143 : struct eventfd_ctx *eventfd;
144 : unsigned long threshold;
145 : };
146 :
147 : /* For threshold */
148 : struct mem_cgroup_threshold_ary {
149 : /* An array index points to threshold just below or equal to usage. */
150 : int current_threshold;
151 : /* Size of entries[] */
152 : unsigned int size;
153 : /* Array of thresholds */
154 : struct mem_cgroup_threshold entries[];
155 : };
156 :
157 : struct mem_cgroup_thresholds {
158 : /* Primary thresholds array */
159 : struct mem_cgroup_threshold_ary *primary;
160 : /*
161 : * Spare threshold array.
162 : * This is needed to make mem_cgroup_unregister_event() "never fail".
163 : * It must be able to store at least primary->size - 1 entries.
164 : */
165 : struct mem_cgroup_threshold_ary *spare;
166 : };
167 :
168 : /*
169 : * Remember four most recent foreign writebacks with dirty pages in this
170 : * cgroup. Inode sharing is expected to be uncommon and, even if we miss
171 : * one in a given round, we're likely to catch it later if it keeps
172 : * foreign-dirtying, so a fairly low count should be enough.
173 : *
174 : * See mem_cgroup_track_foreign_dirty_slowpath() for details.
175 : */
176 : #define MEMCG_CGWB_FRN_CNT 4
177 :
178 : struct memcg_cgwb_frn {
179 : u64 bdi_id; /* bdi->id of the foreign inode */
180 : int memcg_id; /* memcg->css.id of foreign inode */
181 : u64 at; /* jiffies_64 at the time of dirtying */
182 : struct wb_completion done; /* tracks in-flight foreign writebacks */
183 : };
184 :
185 : /*
186 : * Bucket for arbitrarily byte-sized objects charged to a memory
187 : * cgroup. The bucket can be reparented in one piece when the cgroup
188 : * is destroyed, without having to round up the individual references
189 : * of all live memory objects in the wild.
190 : */
191 : struct obj_cgroup {
192 : struct percpu_ref refcnt;
193 : struct mem_cgroup *memcg;
194 : atomic_t nr_charged_bytes;
195 : union {
196 : struct list_head list; /* protected by objcg_lock */
197 : struct rcu_head rcu;
198 : };
199 : };
200 :
201 : /*
202 : * The memory controller data structure. The memory controller controls both
203 : * page cache and RSS per cgroup. We would eventually like to provide
204 : * statistics based on the statistics developed by Rik Van Riel for clock-pro,
205 : * to help the administrator determine what knobs to tune.
206 : */
207 : struct mem_cgroup {
208 : struct cgroup_subsys_state css;
209 :
210 : /* Private memcg ID. Used to ID objects that outlive the cgroup */
211 : struct mem_cgroup_id id;
212 :
213 : /* Accounted resources */
214 : struct page_counter memory; /* Both v1 & v2 */
215 :
216 : union {
217 : struct page_counter swap; /* v2 only */
218 : struct page_counter memsw; /* v1 only */
219 : };
220 :
221 : /* Legacy consumer-oriented counters */
222 : struct page_counter kmem; /* v1 only */
223 : struct page_counter tcpmem; /* v1 only */
224 :
225 : /* Range enforcement for interrupt charges */
226 : struct work_struct high_work;
227 :
228 : #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP)
229 : unsigned long zswap_max;
230 : #endif
231 :
232 : unsigned long soft_limit;
233 :
234 : /* vmpressure notifications */
235 : struct vmpressure vmpressure;
236 :
237 : /*
238 : * Should the OOM killer kill all belonging tasks, had it kill one?
239 : */
240 : bool oom_group;
241 :
242 : /* protected by memcg_oom_lock */
243 : bool oom_lock;
244 : int under_oom;
245 :
246 : int swappiness;
247 : /* OOM-Killer disable */
248 : int oom_kill_disable;
249 :
250 : /* memory.events and memory.events.local */
251 : struct cgroup_file events_file;
252 : struct cgroup_file events_local_file;
253 :
254 : /* handle for "memory.swap.events" */
255 : struct cgroup_file swap_events_file;
256 :
257 : /* protect arrays of thresholds */
258 : struct mutex thresholds_lock;
259 :
260 : /* thresholds for memory usage. RCU-protected */
261 : struct mem_cgroup_thresholds thresholds;
262 :
263 : /* thresholds for mem+swap usage. RCU-protected */
264 : struct mem_cgroup_thresholds memsw_thresholds;
265 :
266 : /* For oom notifier event fd */
267 : struct list_head oom_notify;
268 :
269 : /*
270 : * Should we move charges of a task when a task is moved into this
271 : * mem_cgroup ? And what type of charges should we move ?
272 : */
273 : unsigned long move_charge_at_immigrate;
274 : /* taken only while moving_account > 0 */
275 : spinlock_t move_lock;
276 : unsigned long move_lock_flags;
277 :
278 : CACHELINE_PADDING(_pad1_);
279 :
280 : /* memory.stat */
281 : struct memcg_vmstats *vmstats;
282 :
283 : /* memory.events */
284 : atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
285 : atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS];
286 :
287 : unsigned long socket_pressure;
288 :
289 : /* Legacy tcp memory accounting */
290 : bool tcpmem_active;
291 : int tcpmem_pressure;
292 :
293 : #ifdef CONFIG_MEMCG_KMEM
294 : int kmemcg_id;
295 : struct obj_cgroup __rcu *objcg;
296 : /* list of inherited objcgs, protected by objcg_lock */
297 : struct list_head objcg_list;
298 : #endif
299 :
300 : CACHELINE_PADDING(_pad2_);
301 :
302 : /*
303 : * set > 0 if pages under this cgroup are moving to other cgroup.
304 : */
305 : atomic_t moving_account;
306 : struct task_struct *move_lock_task;
307 :
308 : struct memcg_vmstats_percpu __percpu *vmstats_percpu;
309 :
310 : #ifdef CONFIG_CGROUP_WRITEBACK
311 : struct list_head cgwb_list;
312 : struct wb_domain cgwb_domain;
313 : struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT];
314 : #endif
315 :
316 : /* List of events which userspace want to receive */
317 : struct list_head event_list;
318 : spinlock_t event_list_lock;
319 :
320 : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
321 : struct deferred_split deferred_split_queue;
322 : #endif
323 :
324 : #ifdef CONFIG_LRU_GEN
325 : /* per-memcg mm_struct list */
326 : struct lru_gen_mm_list mm_list;
327 : #endif
328 :
329 : struct mem_cgroup_per_node *nodeinfo[];
330 : };
331 :
332 : /*
333 : * size of first charge trial.
334 : * TODO: maybe necessary to use big numbers in big irons or dynamic based of the
335 : * workload.
336 : */
337 : #define MEMCG_CHARGE_BATCH 64U
338 :
339 : extern struct mem_cgroup *root_mem_cgroup;
340 :
341 : enum page_memcg_data_flags {
342 : /* page->memcg_data is a pointer to an objcgs vector */
343 : MEMCG_DATA_OBJCGS = (1UL << 0),
344 : /* page has been accounted as a non-slab kernel page */
345 : MEMCG_DATA_KMEM = (1UL << 1),
346 : /* the next bit after the last actual flag */
347 : __NR_MEMCG_DATA_FLAGS = (1UL << 2),
348 : };
349 :
350 : #define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1)
351 :
352 : static inline bool folio_memcg_kmem(struct folio *folio);
353 :
354 : /*
355 : * After the initialization objcg->memcg is always pointing at
356 : * a valid memcg, but can be atomically swapped to the parent memcg.
357 : *
358 : * The caller must ensure that the returned memcg won't be released:
359 : * e.g. acquire the rcu_read_lock or css_set_lock.
360 : */
361 : static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg)
362 : {
363 : return READ_ONCE(objcg->memcg);
364 : }
365 :
366 : /*
367 : * __folio_memcg - Get the memory cgroup associated with a non-kmem folio
368 : * @folio: Pointer to the folio.
369 : *
370 : * Returns a pointer to the memory cgroup associated with the folio,
371 : * or NULL. This function assumes that the folio is known to have a
372 : * proper memory cgroup pointer. It's not safe to call this function
373 : * against some type of folios, e.g. slab folios or ex-slab folios or
374 : * kmem folios.
375 : */
376 : static inline struct mem_cgroup *__folio_memcg(struct folio *folio)
377 : {
378 : unsigned long memcg_data = folio->memcg_data;
379 :
380 : VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
381 : VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio);
382 : VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio);
383 :
384 : return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
385 : }
386 :
387 : /*
388 : * __folio_objcg - get the object cgroup associated with a kmem folio.
389 : * @folio: Pointer to the folio.
390 : *
391 : * Returns a pointer to the object cgroup associated with the folio,
392 : * or NULL. This function assumes that the folio is known to have a
393 : * proper object cgroup pointer. It's not safe to call this function
394 : * against some type of folios, e.g. slab folios or ex-slab folios or
395 : * LRU folios.
396 : */
397 : static inline struct obj_cgroup *__folio_objcg(struct folio *folio)
398 : {
399 : unsigned long memcg_data = folio->memcg_data;
400 :
401 : VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
402 : VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio);
403 : VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio);
404 :
405 : return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
406 : }
407 :
408 : /*
409 : * folio_memcg - Get the memory cgroup associated with a folio.
410 : * @folio: Pointer to the folio.
411 : *
412 : * Returns a pointer to the memory cgroup associated with the folio,
413 : * or NULL. This function assumes that the folio is known to have a
414 : * proper memory cgroup pointer. It's not safe to call this function
415 : * against some type of folios, e.g. slab folios or ex-slab folios.
416 : *
417 : * For a non-kmem folio any of the following ensures folio and memcg binding
418 : * stability:
419 : *
420 : * - the folio lock
421 : * - LRU isolation
422 : * - folio_memcg_lock()
423 : * - exclusive reference
424 : * - mem_cgroup_trylock_pages()
425 : *
426 : * For a kmem folio a caller should hold an rcu read lock to protect memcg
427 : * associated with a kmem folio from being released.
428 : */
429 : static inline struct mem_cgroup *folio_memcg(struct folio *folio)
430 : {
431 : if (folio_memcg_kmem(folio))
432 : return obj_cgroup_memcg(__folio_objcg(folio));
433 : return __folio_memcg(folio);
434 : }
435 :
436 : static inline struct mem_cgroup *page_memcg(struct page *page)
437 : {
438 : return folio_memcg(page_folio(page));
439 : }
440 :
441 : /**
442 : * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio.
443 : * @folio: Pointer to the folio.
444 : *
445 : * This function assumes that the folio is known to have a
446 : * proper memory cgroup pointer. It's not safe to call this function
447 : * against some type of folios, e.g. slab folios or ex-slab folios.
448 : *
449 : * Return: A pointer to the memory cgroup associated with the folio,
450 : * or NULL.
451 : */
452 : static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
453 : {
454 : unsigned long memcg_data = READ_ONCE(folio->memcg_data);
455 :
456 : VM_BUG_ON_FOLIO(folio_test_slab(folio), folio);
457 : WARN_ON_ONCE(!rcu_read_lock_held());
458 :
459 : if (memcg_data & MEMCG_DATA_KMEM) {
460 : struct obj_cgroup *objcg;
461 :
462 : objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
463 : return obj_cgroup_memcg(objcg);
464 : }
465 :
466 : return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
467 : }
468 :
469 : /*
470 : * folio_memcg_check - Get the memory cgroup associated with a folio.
471 : * @folio: Pointer to the folio.
472 : *
473 : * Returns a pointer to the memory cgroup associated with the folio,
474 : * or NULL. This function unlike folio_memcg() can take any folio
475 : * as an argument. It has to be used in cases when it's not known if a folio
476 : * has an associated memory cgroup pointer or an object cgroups vector or
477 : * an object cgroup.
478 : *
479 : * For a non-kmem folio any of the following ensures folio and memcg binding
480 : * stability:
481 : *
482 : * - the folio lock
483 : * - LRU isolation
484 : * - lock_folio_memcg()
485 : * - exclusive reference
486 : * - mem_cgroup_trylock_pages()
487 : *
488 : * For a kmem folio a caller should hold an rcu read lock to protect memcg
489 : * associated with a kmem folio from being released.
490 : */
491 : static inline struct mem_cgroup *folio_memcg_check(struct folio *folio)
492 : {
493 : /*
494 : * Because folio->memcg_data might be changed asynchronously
495 : * for slabs, READ_ONCE() should be used here.
496 : */
497 : unsigned long memcg_data = READ_ONCE(folio->memcg_data);
498 :
499 : if (memcg_data & MEMCG_DATA_OBJCGS)
500 : return NULL;
501 :
502 : if (memcg_data & MEMCG_DATA_KMEM) {
503 : struct obj_cgroup *objcg;
504 :
505 : objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
506 : return obj_cgroup_memcg(objcg);
507 : }
508 :
509 : return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK);
510 : }
511 :
512 : static inline struct mem_cgroup *page_memcg_check(struct page *page)
513 : {
514 : if (PageTail(page))
515 : return NULL;
516 : return folio_memcg_check((struct folio *)page);
517 : }
518 :
519 : static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg)
520 : {
521 : struct mem_cgroup *memcg;
522 :
523 : rcu_read_lock();
524 : retry:
525 : memcg = obj_cgroup_memcg(objcg);
526 : if (unlikely(!css_tryget(&memcg->css)))
527 : goto retry;
528 : rcu_read_unlock();
529 :
530 : return memcg;
531 : }
532 :
533 : #ifdef CONFIG_MEMCG_KMEM
534 : /*
535 : * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set.
536 : * @folio: Pointer to the folio.
537 : *
538 : * Checks if the folio has MemcgKmem flag set. The caller must ensure
539 : * that the folio has an associated memory cgroup. It's not safe to call
540 : * this function against some types of folios, e.g. slab folios.
541 : */
542 : static inline bool folio_memcg_kmem(struct folio *folio)
543 : {
544 : VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page);
545 : VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJCGS, folio);
546 : return folio->memcg_data & MEMCG_DATA_KMEM;
547 : }
548 :
549 :
550 : #else
551 : static inline bool folio_memcg_kmem(struct folio *folio)
552 : {
553 : return false;
554 : }
555 :
556 : #endif
557 :
558 : static inline bool PageMemcgKmem(struct page *page)
559 : {
560 : return folio_memcg_kmem(page_folio(page));
561 : }
562 :
563 : static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
564 : {
565 : return (memcg == root_mem_cgroup);
566 : }
567 :
568 : static inline bool mem_cgroup_disabled(void)
569 : {
570 : return !cgroup_subsys_enabled(memory_cgrp_subsys);
571 : }
572 :
573 : static inline void mem_cgroup_protection(struct mem_cgroup *root,
574 : struct mem_cgroup *memcg,
575 : unsigned long *min,
576 : unsigned long *low)
577 : {
578 : *min = *low = 0;
579 :
580 : if (mem_cgroup_disabled())
581 : return;
582 :
583 : /*
584 : * There is no reclaim protection applied to a targeted reclaim.
585 : * We are special casing this specific case here because
586 : * mem_cgroup_protected calculation is not robust enough to keep
587 : * the protection invariant for calculated effective values for
588 : * parallel reclaimers with different reclaim target. This is
589 : * especially a problem for tail memcgs (as they have pages on LRU)
590 : * which would want to have effective values 0 for targeted reclaim
591 : * but a different value for external reclaim.
592 : *
593 : * Example
594 : * Let's have global and A's reclaim in parallel:
595 : * |
596 : * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
597 : * |\
598 : * | C (low = 1G, usage = 2.5G)
599 : * B (low = 1G, usage = 0.5G)
600 : *
601 : * For the global reclaim
602 : * A.elow = A.low
603 : * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
604 : * C.elow = min(C.usage, C.low)
605 : *
606 : * With the effective values resetting we have A reclaim
607 : * A.elow = 0
608 : * B.elow = B.low
609 : * C.elow = C.low
610 : *
611 : * If the global reclaim races with A's reclaim then
612 : * B.elow = C.elow = 0 because children_low_usage > A.elow)
613 : * is possible and reclaiming B would be violating the protection.
614 : *
615 : */
616 : if (root == memcg)
617 : return;
618 :
619 : *min = READ_ONCE(memcg->memory.emin);
620 : *low = READ_ONCE(memcg->memory.elow);
621 : }
622 :
623 : void mem_cgroup_calculate_protection(struct mem_cgroup *root,
624 : struct mem_cgroup *memcg);
625 :
626 : static inline bool mem_cgroup_unprotected(struct mem_cgroup *target,
627 : struct mem_cgroup *memcg)
628 : {
629 : /*
630 : * The root memcg doesn't account charges, and doesn't support
631 : * protection. The target memcg's protection is ignored, see
632 : * mem_cgroup_calculate_protection() and mem_cgroup_protection()
633 : */
634 : return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) ||
635 : memcg == target;
636 : }
637 :
638 : static inline bool mem_cgroup_below_low(struct mem_cgroup *target,
639 : struct mem_cgroup *memcg)
640 : {
641 : if (mem_cgroup_unprotected(target, memcg))
642 : return false;
643 :
644 : return READ_ONCE(memcg->memory.elow) >=
645 : page_counter_read(&memcg->memory);
646 : }
647 :
648 : static inline bool mem_cgroup_below_min(struct mem_cgroup *target,
649 : struct mem_cgroup *memcg)
650 : {
651 : if (mem_cgroup_unprotected(target, memcg))
652 : return false;
653 :
654 : return READ_ONCE(memcg->memory.emin) >=
655 : page_counter_read(&memcg->memory);
656 : }
657 :
658 : int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp);
659 :
660 : /**
661 : * mem_cgroup_charge - Charge a newly allocated folio to a cgroup.
662 : * @folio: Folio to charge.
663 : * @mm: mm context of the allocating task.
664 : * @gfp: Reclaim mode.
665 : *
666 : * Try to charge @folio to the memcg that @mm belongs to, reclaiming
667 : * pages according to @gfp if necessary. If @mm is NULL, try to
668 : * charge to the active memcg.
669 : *
670 : * Do not use this for folios allocated for swapin.
671 : *
672 : * Return: 0 on success. Otherwise, an error code is returned.
673 : */
674 : static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm,
675 : gfp_t gfp)
676 : {
677 : if (mem_cgroup_disabled())
678 : return 0;
679 : return __mem_cgroup_charge(folio, mm, gfp);
680 : }
681 :
682 : int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm,
683 : gfp_t gfp, swp_entry_t entry);
684 : void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry);
685 :
686 : void __mem_cgroup_uncharge(struct folio *folio);
687 :
688 : /**
689 : * mem_cgroup_uncharge - Uncharge a folio.
690 : * @folio: Folio to uncharge.
691 : *
692 : * Uncharge a folio previously charged with mem_cgroup_charge().
693 : */
694 : static inline void mem_cgroup_uncharge(struct folio *folio)
695 : {
696 : if (mem_cgroup_disabled())
697 : return;
698 : __mem_cgroup_uncharge(folio);
699 : }
700 :
701 : void __mem_cgroup_uncharge_list(struct list_head *page_list);
702 : static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
703 : {
704 : if (mem_cgroup_disabled())
705 : return;
706 : __mem_cgroup_uncharge_list(page_list);
707 : }
708 :
709 : void mem_cgroup_migrate(struct folio *old, struct folio *new);
710 :
711 : /**
712 : * mem_cgroup_lruvec - get the lru list vector for a memcg & node
713 : * @memcg: memcg of the wanted lruvec
714 : * @pgdat: pglist_data
715 : *
716 : * Returns the lru list vector holding pages for a given @memcg &
717 : * @pgdat combination. This can be the node lruvec, if the memory
718 : * controller is disabled.
719 : */
720 : static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
721 : struct pglist_data *pgdat)
722 : {
723 : struct mem_cgroup_per_node *mz;
724 : struct lruvec *lruvec;
725 :
726 : if (mem_cgroup_disabled()) {
727 : lruvec = &pgdat->__lruvec;
728 : goto out;
729 : }
730 :
731 : if (!memcg)
732 : memcg = root_mem_cgroup;
733 :
734 : mz = memcg->nodeinfo[pgdat->node_id];
735 : lruvec = &mz->lruvec;
736 : out:
737 : /*
738 : * Since a node can be onlined after the mem_cgroup was created,
739 : * we have to be prepared to initialize lruvec->pgdat here;
740 : * and if offlined then reonlined, we need to reinitialize it.
741 : */
742 : if (unlikely(lruvec->pgdat != pgdat))
743 : lruvec->pgdat = pgdat;
744 : return lruvec;
745 : }
746 :
747 : /**
748 : * folio_lruvec - return lruvec for isolating/putting an LRU folio
749 : * @folio: Pointer to the folio.
750 : *
751 : * This function relies on folio->mem_cgroup being stable.
752 : */
753 : static inline struct lruvec *folio_lruvec(struct folio *folio)
754 : {
755 : struct mem_cgroup *memcg = folio_memcg(folio);
756 :
757 : VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio);
758 : return mem_cgroup_lruvec(memcg, folio_pgdat(folio));
759 : }
760 :
761 : struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
762 :
763 : struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm);
764 :
765 : struct lruvec *folio_lruvec_lock(struct folio *folio);
766 : struct lruvec *folio_lruvec_lock_irq(struct folio *folio);
767 : struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
768 : unsigned long *flags);
769 :
770 : #ifdef CONFIG_DEBUG_VM
771 : void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio);
772 : #else
773 : static inline
774 : void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
775 : {
776 : }
777 : #endif
778 :
779 : static inline
780 : struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
781 : return css ? container_of(css, struct mem_cgroup, css) : NULL;
782 : }
783 :
784 : static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg)
785 : {
786 : return percpu_ref_tryget(&objcg->refcnt);
787 : }
788 :
789 : static inline void obj_cgroup_get(struct obj_cgroup *objcg)
790 : {
791 : percpu_ref_get(&objcg->refcnt);
792 : }
793 :
794 : static inline void obj_cgroup_get_many(struct obj_cgroup *objcg,
795 : unsigned long nr)
796 : {
797 : percpu_ref_get_many(&objcg->refcnt, nr);
798 : }
799 :
800 : static inline void obj_cgroup_put(struct obj_cgroup *objcg)
801 : {
802 : percpu_ref_put(&objcg->refcnt);
803 : }
804 :
805 : static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg)
806 : {
807 : return !memcg || css_tryget(&memcg->css);
808 : }
809 :
810 : static inline void mem_cgroup_put(struct mem_cgroup *memcg)
811 : {
812 : if (memcg)
813 : css_put(&memcg->css);
814 : }
815 :
816 : #define mem_cgroup_from_counter(counter, member) \
817 : container_of(counter, struct mem_cgroup, member)
818 :
819 : struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
820 : struct mem_cgroup *,
821 : struct mem_cgroup_reclaim_cookie *);
822 : void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
823 : void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
824 : int (*)(struct task_struct *, void *), void *arg);
825 :
826 : static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
827 : {
828 : if (mem_cgroup_disabled())
829 : return 0;
830 :
831 : return memcg->id.id;
832 : }
833 : struct mem_cgroup *mem_cgroup_from_id(unsigned short id);
834 :
835 : #ifdef CONFIG_SHRINKER_DEBUG
836 : static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
837 : {
838 : return memcg ? cgroup_ino(memcg->css.cgroup) : 0;
839 : }
840 :
841 : struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino);
842 : #endif
843 :
844 : static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
845 : {
846 : return mem_cgroup_from_css(seq_css(m));
847 : }
848 :
849 : static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
850 : {
851 : struct mem_cgroup_per_node *mz;
852 :
853 : if (mem_cgroup_disabled())
854 : return NULL;
855 :
856 : mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
857 : return mz->memcg;
858 : }
859 :
860 : /**
861 : * parent_mem_cgroup - find the accounting parent of a memcg
862 : * @memcg: memcg whose parent to find
863 : *
864 : * Returns the parent memcg, or NULL if this is the root or the memory
865 : * controller is in legacy no-hierarchy mode.
866 : */
867 : static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
868 : {
869 : return mem_cgroup_from_css(memcg->css.parent);
870 : }
871 :
872 : static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
873 : struct mem_cgroup *root)
874 : {
875 : if (root == memcg)
876 : return true;
877 : return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
878 : }
879 :
880 : static inline bool mm_match_cgroup(struct mm_struct *mm,
881 : struct mem_cgroup *memcg)
882 : {
883 : struct mem_cgroup *task_memcg;
884 : bool match = false;
885 :
886 : rcu_read_lock();
887 : task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
888 : if (task_memcg)
889 : match = mem_cgroup_is_descendant(task_memcg, memcg);
890 : rcu_read_unlock();
891 : return match;
892 : }
893 :
894 : struct cgroup_subsys_state *mem_cgroup_css_from_folio(struct folio *folio);
895 : ino_t page_cgroup_ino(struct page *page);
896 :
897 : static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
898 : {
899 : if (mem_cgroup_disabled())
900 : return true;
901 : return !!(memcg->css.flags & CSS_ONLINE);
902 : }
903 :
904 : void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
905 : int zid, int nr_pages);
906 :
907 : static inline
908 : unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
909 : enum lru_list lru, int zone_idx)
910 : {
911 : struct mem_cgroup_per_node *mz;
912 :
913 : mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
914 : return READ_ONCE(mz->lru_zone_size[zone_idx][lru]);
915 : }
916 :
917 : void mem_cgroup_handle_over_high(void);
918 :
919 : unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg);
920 :
921 : unsigned long mem_cgroup_size(struct mem_cgroup *memcg);
922 :
923 : void mem_cgroup_print_oom_context(struct mem_cgroup *memcg,
924 : struct task_struct *p);
925 :
926 : void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg);
927 :
928 : static inline void mem_cgroup_enter_user_fault(void)
929 : {
930 : WARN_ON(current->in_user_fault);
931 : current->in_user_fault = 1;
932 : }
933 :
934 : static inline void mem_cgroup_exit_user_fault(void)
935 : {
936 : WARN_ON(!current->in_user_fault);
937 : current->in_user_fault = 0;
938 : }
939 :
940 : static inline bool task_in_memcg_oom(struct task_struct *p)
941 : {
942 : return p->memcg_in_oom;
943 : }
944 :
945 : bool mem_cgroup_oom_synchronize(bool wait);
946 : struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
947 : struct mem_cgroup *oom_domain);
948 : void mem_cgroup_print_oom_group(struct mem_cgroup *memcg);
949 :
950 : void folio_memcg_lock(struct folio *folio);
951 : void folio_memcg_unlock(struct folio *folio);
952 :
953 : void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val);
954 :
955 : /* try to stablize folio_memcg() for all the pages in a memcg */
956 : static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
957 : {
958 : rcu_read_lock();
959 :
960 : if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account))
961 : return true;
962 :
963 : rcu_read_unlock();
964 : return false;
965 : }
966 :
967 : static inline void mem_cgroup_unlock_pages(void)
968 : {
969 : rcu_read_unlock();
970 : }
971 :
972 : /* idx can be of type enum memcg_stat_item or node_stat_item */
973 : static inline void mod_memcg_state(struct mem_cgroup *memcg,
974 : int idx, int val)
975 : {
976 : unsigned long flags;
977 :
978 : local_irq_save(flags);
979 : __mod_memcg_state(memcg, idx, val);
980 : local_irq_restore(flags);
981 : }
982 :
983 : static inline void mod_memcg_page_state(struct page *page,
984 : int idx, int val)
985 : {
986 : struct mem_cgroup *memcg;
987 :
988 : if (mem_cgroup_disabled())
989 : return;
990 :
991 : rcu_read_lock();
992 : memcg = page_memcg(page);
993 : if (memcg)
994 : mod_memcg_state(memcg, idx, val);
995 : rcu_read_unlock();
996 : }
997 :
998 : unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx);
999 :
1000 : static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1001 : enum node_stat_item idx)
1002 : {
1003 : struct mem_cgroup_per_node *pn;
1004 : long x;
1005 :
1006 : if (mem_cgroup_disabled())
1007 : return node_page_state(lruvec_pgdat(lruvec), idx);
1008 :
1009 : pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
1010 : x = READ_ONCE(pn->lruvec_stats.state[idx]);
1011 : #ifdef CONFIG_SMP
1012 : if (x < 0)
1013 : x = 0;
1014 : #endif
1015 : return x;
1016 : }
1017 :
1018 : static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1019 : enum node_stat_item idx)
1020 : {
1021 : struct mem_cgroup_per_node *pn;
1022 : long x = 0;
1023 : int cpu;
1024 :
1025 : if (mem_cgroup_disabled())
1026 : return node_page_state(lruvec_pgdat(lruvec), idx);
1027 :
1028 : pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
1029 : for_each_possible_cpu(cpu)
1030 : x += per_cpu(pn->lruvec_stats_percpu->state[idx], cpu);
1031 : #ifdef CONFIG_SMP
1032 : if (x < 0)
1033 : x = 0;
1034 : #endif
1035 : return x;
1036 : }
1037 :
1038 : void mem_cgroup_flush_stats(void);
1039 : void mem_cgroup_flush_stats_ratelimited(void);
1040 :
1041 : void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
1042 : int val);
1043 : void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val);
1044 :
1045 : static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1046 : int val)
1047 : {
1048 : unsigned long flags;
1049 :
1050 : local_irq_save(flags);
1051 : __mod_lruvec_kmem_state(p, idx, val);
1052 : local_irq_restore(flags);
1053 : }
1054 :
1055 : static inline void mod_memcg_lruvec_state(struct lruvec *lruvec,
1056 : enum node_stat_item idx, int val)
1057 : {
1058 : unsigned long flags;
1059 :
1060 : local_irq_save(flags);
1061 : __mod_memcg_lruvec_state(lruvec, idx, val);
1062 : local_irq_restore(flags);
1063 : }
1064 :
1065 : void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
1066 : unsigned long count);
1067 :
1068 : static inline void count_memcg_events(struct mem_cgroup *memcg,
1069 : enum vm_event_item idx,
1070 : unsigned long count)
1071 : {
1072 : unsigned long flags;
1073 :
1074 : local_irq_save(flags);
1075 : __count_memcg_events(memcg, idx, count);
1076 : local_irq_restore(flags);
1077 : }
1078 :
1079 : static inline void count_memcg_page_event(struct page *page,
1080 : enum vm_event_item idx)
1081 : {
1082 : struct mem_cgroup *memcg = page_memcg(page);
1083 :
1084 : if (memcg)
1085 : count_memcg_events(memcg, idx, 1);
1086 : }
1087 :
1088 : static inline void count_memcg_folio_events(struct folio *folio,
1089 : enum vm_event_item idx, unsigned long nr)
1090 : {
1091 : struct mem_cgroup *memcg = folio_memcg(folio);
1092 :
1093 : if (memcg)
1094 : count_memcg_events(memcg, idx, nr);
1095 : }
1096 :
1097 : static inline void count_memcg_event_mm(struct mm_struct *mm,
1098 : enum vm_event_item idx)
1099 : {
1100 : struct mem_cgroup *memcg;
1101 :
1102 : if (mem_cgroup_disabled())
1103 : return;
1104 :
1105 : rcu_read_lock();
1106 : memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1107 : if (likely(memcg))
1108 : count_memcg_events(memcg, idx, 1);
1109 : rcu_read_unlock();
1110 : }
1111 :
1112 : static inline void memcg_memory_event(struct mem_cgroup *memcg,
1113 : enum memcg_memory_event event)
1114 : {
1115 : bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX ||
1116 : event == MEMCG_SWAP_FAIL;
1117 :
1118 : atomic_long_inc(&memcg->memory_events_local[event]);
1119 : if (!swap_event)
1120 : cgroup_file_notify(&memcg->events_local_file);
1121 :
1122 : do {
1123 : atomic_long_inc(&memcg->memory_events[event]);
1124 : if (swap_event)
1125 : cgroup_file_notify(&memcg->swap_events_file);
1126 : else
1127 : cgroup_file_notify(&memcg->events_file);
1128 :
1129 : if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
1130 : break;
1131 : if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS)
1132 : break;
1133 : } while ((memcg = parent_mem_cgroup(memcg)) &&
1134 : !mem_cgroup_is_root(memcg));
1135 : }
1136 :
1137 : static inline void memcg_memory_event_mm(struct mm_struct *mm,
1138 : enum memcg_memory_event event)
1139 : {
1140 : struct mem_cgroup *memcg;
1141 :
1142 : if (mem_cgroup_disabled())
1143 : return;
1144 :
1145 : rcu_read_lock();
1146 : memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
1147 : if (likely(memcg))
1148 : memcg_memory_event(memcg, event);
1149 : rcu_read_unlock();
1150 : }
1151 :
1152 : void split_page_memcg(struct page *head, unsigned int nr);
1153 :
1154 : unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1155 : gfp_t gfp_mask,
1156 : unsigned long *total_scanned);
1157 :
1158 : #else /* CONFIG_MEMCG */
1159 :
1160 : #define MEM_CGROUP_ID_SHIFT 0
1161 : #define MEM_CGROUP_ID_MAX 0
1162 :
1163 : static inline struct mem_cgroup *folio_memcg(struct folio *folio)
1164 : {
1165 : return NULL;
1166 : }
1167 :
1168 : static inline struct mem_cgroup *page_memcg(struct page *page)
1169 : {
1170 : return NULL;
1171 : }
1172 :
1173 : static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio)
1174 : {
1175 0 : WARN_ON_ONCE(!rcu_read_lock_held());
1176 : return NULL;
1177 : }
1178 :
1179 : static inline struct mem_cgroup *folio_memcg_check(struct folio *folio)
1180 : {
1181 : return NULL;
1182 : }
1183 :
1184 : static inline struct mem_cgroup *page_memcg_check(struct page *page)
1185 : {
1186 : return NULL;
1187 : }
1188 :
1189 : static inline bool folio_memcg_kmem(struct folio *folio)
1190 : {
1191 : return false;
1192 : }
1193 :
1194 : static inline bool PageMemcgKmem(struct page *page)
1195 : {
1196 : return false;
1197 : }
1198 :
1199 : static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
1200 : {
1201 : return true;
1202 : }
1203 :
1204 : static inline bool mem_cgroup_disabled(void)
1205 : {
1206 : return true;
1207 : }
1208 :
1209 : static inline void memcg_memory_event(struct mem_cgroup *memcg,
1210 : enum memcg_memory_event event)
1211 : {
1212 : }
1213 :
1214 : static inline void memcg_memory_event_mm(struct mm_struct *mm,
1215 : enum memcg_memory_event event)
1216 : {
1217 : }
1218 :
1219 : static inline void mem_cgroup_protection(struct mem_cgroup *root,
1220 : struct mem_cgroup *memcg,
1221 : unsigned long *min,
1222 : unsigned long *low)
1223 : {
1224 0 : *min = *low = 0;
1225 : }
1226 :
1227 : static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root,
1228 : struct mem_cgroup *memcg)
1229 : {
1230 : }
1231 :
1232 : static inline bool mem_cgroup_unprotected(struct mem_cgroup *target,
1233 : struct mem_cgroup *memcg)
1234 : {
1235 : return true;
1236 : }
1237 : static inline bool mem_cgroup_below_low(struct mem_cgroup *target,
1238 : struct mem_cgroup *memcg)
1239 : {
1240 : return false;
1241 : }
1242 :
1243 : static inline bool mem_cgroup_below_min(struct mem_cgroup *target,
1244 : struct mem_cgroup *memcg)
1245 : {
1246 : return false;
1247 : }
1248 :
1249 : static inline int mem_cgroup_charge(struct folio *folio,
1250 : struct mm_struct *mm, gfp_t gfp)
1251 : {
1252 : return 0;
1253 : }
1254 :
1255 : static inline int mem_cgroup_swapin_charge_folio(struct folio *folio,
1256 : struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
1257 : {
1258 : return 0;
1259 : }
1260 :
1261 : static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry)
1262 : {
1263 : }
1264 :
1265 : static inline void mem_cgroup_uncharge(struct folio *folio)
1266 : {
1267 : }
1268 :
1269 : static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
1270 : {
1271 : }
1272 :
1273 : static inline void mem_cgroup_migrate(struct folio *old, struct folio *new)
1274 : {
1275 : }
1276 :
1277 : static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
1278 : struct pglist_data *pgdat)
1279 : {
1280 0 : return &pgdat->__lruvec;
1281 : }
1282 :
1283 : static inline struct lruvec *folio_lruvec(struct folio *folio)
1284 : {
1285 0 : struct pglist_data *pgdat = folio_pgdat(folio);
1286 : return &pgdat->__lruvec;
1287 : }
1288 :
1289 : static inline
1290 : void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio)
1291 : {
1292 : }
1293 :
1294 : static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
1295 : {
1296 : return NULL;
1297 : }
1298 :
1299 : static inline bool mm_match_cgroup(struct mm_struct *mm,
1300 : struct mem_cgroup *memcg)
1301 : {
1302 : return true;
1303 : }
1304 :
1305 : static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1306 : {
1307 : return NULL;
1308 : }
1309 :
1310 : static inline
1311 : struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css)
1312 : {
1313 : return NULL;
1314 : }
1315 :
1316 : static inline void obj_cgroup_put(struct obj_cgroup *objcg)
1317 : {
1318 : }
1319 :
1320 : static inline bool mem_cgroup_tryget(struct mem_cgroup *memcg)
1321 : {
1322 : return true;
1323 : }
1324 :
1325 : static inline void mem_cgroup_put(struct mem_cgroup *memcg)
1326 : {
1327 : }
1328 :
1329 : static inline struct lruvec *folio_lruvec_lock(struct folio *folio)
1330 : {
1331 : struct pglist_data *pgdat = folio_pgdat(folio);
1332 :
1333 : spin_lock(&pgdat->__lruvec.lru_lock);
1334 : return &pgdat->__lruvec;
1335 : }
1336 :
1337 : static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio)
1338 : {
1339 0 : struct pglist_data *pgdat = folio_pgdat(folio);
1340 :
1341 0 : spin_lock_irq(&pgdat->__lruvec.lru_lock);
1342 : return &pgdat->__lruvec;
1343 : }
1344 :
1345 : static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
1346 : unsigned long *flagsp)
1347 : {
1348 0 : struct pglist_data *pgdat = folio_pgdat(folio);
1349 :
1350 0 : spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp);
1351 : return &pgdat->__lruvec;
1352 : }
1353 :
1354 : static inline struct mem_cgroup *
1355 : mem_cgroup_iter(struct mem_cgroup *root,
1356 : struct mem_cgroup *prev,
1357 : struct mem_cgroup_reclaim_cookie *reclaim)
1358 : {
1359 : return NULL;
1360 : }
1361 :
1362 : static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
1363 : struct mem_cgroup *prev)
1364 : {
1365 : }
1366 :
1367 : static inline void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
1368 : int (*fn)(struct task_struct *, void *), void *arg)
1369 : {
1370 : }
1371 :
1372 : static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
1373 : {
1374 : return 0;
1375 : }
1376 :
1377 0 : static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
1378 : {
1379 0 : WARN_ON_ONCE(id);
1380 : /* XXX: This should always return root_mem_cgroup */
1381 0 : return NULL;
1382 : }
1383 :
1384 : #ifdef CONFIG_SHRINKER_DEBUG
1385 : static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg)
1386 : {
1387 : return 0;
1388 : }
1389 :
1390 : static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino)
1391 : {
1392 : return NULL;
1393 : }
1394 : #endif
1395 :
1396 : static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m)
1397 : {
1398 : return NULL;
1399 : }
1400 :
1401 : static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec)
1402 : {
1403 : return NULL;
1404 : }
1405 :
1406 : static inline bool mem_cgroup_online(struct mem_cgroup *memcg)
1407 : {
1408 : return true;
1409 : }
1410 :
1411 : static inline
1412 : unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec,
1413 : enum lru_list lru, int zone_idx)
1414 : {
1415 : return 0;
1416 : }
1417 :
1418 : static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
1419 : {
1420 : return 0;
1421 : }
1422 :
1423 : static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
1424 : {
1425 : return 0;
1426 : }
1427 :
1428 : static inline void
1429 : mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1430 : {
1431 : }
1432 :
1433 : static inline void
1434 : mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
1435 : {
1436 : }
1437 :
1438 : static inline void folio_memcg_lock(struct folio *folio)
1439 : {
1440 : }
1441 :
1442 : static inline void folio_memcg_unlock(struct folio *folio)
1443 : {
1444 : }
1445 :
1446 : static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg)
1447 : {
1448 : /* to match folio_memcg_rcu() */
1449 : rcu_read_lock();
1450 : return true;
1451 : }
1452 :
1453 : static inline void mem_cgroup_unlock_pages(void)
1454 : {
1455 : rcu_read_unlock();
1456 : }
1457 :
1458 : static inline void mem_cgroup_handle_over_high(void)
1459 : {
1460 : }
1461 :
1462 : static inline void mem_cgroup_enter_user_fault(void)
1463 : {
1464 : }
1465 :
1466 : static inline void mem_cgroup_exit_user_fault(void)
1467 : {
1468 : }
1469 :
1470 : static inline bool task_in_memcg_oom(struct task_struct *p)
1471 : {
1472 : return false;
1473 : }
1474 :
1475 : static inline bool mem_cgroup_oom_synchronize(bool wait)
1476 : {
1477 : return false;
1478 : }
1479 :
1480 : static inline struct mem_cgroup *mem_cgroup_get_oom_group(
1481 : struct task_struct *victim, struct mem_cgroup *oom_domain)
1482 : {
1483 : return NULL;
1484 : }
1485 :
1486 : static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
1487 : {
1488 : }
1489 :
1490 : static inline void __mod_memcg_state(struct mem_cgroup *memcg,
1491 : int idx,
1492 : int nr)
1493 : {
1494 : }
1495 :
1496 : static inline void mod_memcg_state(struct mem_cgroup *memcg,
1497 : int idx,
1498 : int nr)
1499 : {
1500 : }
1501 :
1502 : static inline void mod_memcg_page_state(struct page *page,
1503 : int idx, int val)
1504 : {
1505 : }
1506 :
1507 : static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
1508 : {
1509 : return 0;
1510 : }
1511 :
1512 : static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
1513 : enum node_stat_item idx)
1514 : {
1515 0 : return node_page_state(lruvec_pgdat(lruvec), idx);
1516 : }
1517 :
1518 : static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,
1519 : enum node_stat_item idx)
1520 : {
1521 : return node_page_state(lruvec_pgdat(lruvec), idx);
1522 : }
1523 :
1524 : static inline void mem_cgroup_flush_stats(void)
1525 : {
1526 : }
1527 :
1528 : static inline void mem_cgroup_flush_stats_ratelimited(void)
1529 : {
1530 : }
1531 :
1532 : static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec,
1533 : enum node_stat_item idx, int val)
1534 : {
1535 : }
1536 :
1537 : static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1538 : int val)
1539 : {
1540 0 : struct page *page = virt_to_head_page(p);
1541 :
1542 0 : __mod_node_page_state(page_pgdat(page), idx, val);
1543 : }
1544 :
1545 : static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx,
1546 : int val)
1547 : {
1548 : struct page *page = virt_to_head_page(p);
1549 :
1550 : mod_node_page_state(page_pgdat(page), idx, val);
1551 : }
1552 :
1553 : static inline void count_memcg_events(struct mem_cgroup *memcg,
1554 : enum vm_event_item idx,
1555 : unsigned long count)
1556 : {
1557 : }
1558 :
1559 : static inline void __count_memcg_events(struct mem_cgroup *memcg,
1560 : enum vm_event_item idx,
1561 : unsigned long count)
1562 : {
1563 : }
1564 :
1565 : static inline void count_memcg_page_event(struct page *page,
1566 : int idx)
1567 : {
1568 : }
1569 :
1570 : static inline void count_memcg_folio_events(struct folio *folio,
1571 : enum vm_event_item idx, unsigned long nr)
1572 : {
1573 : }
1574 :
1575 : static inline
1576 : void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx)
1577 : {
1578 : }
1579 :
1580 : static inline void split_page_memcg(struct page *head, unsigned int nr)
1581 : {
1582 : }
1583 :
1584 : static inline
1585 : unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
1586 : gfp_t gfp_mask,
1587 : unsigned long *total_scanned)
1588 : {
1589 : return 0;
1590 : }
1591 : #endif /* CONFIG_MEMCG */
1592 :
1593 : static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx)
1594 : {
1595 0 : __mod_lruvec_kmem_state(p, idx, 1);
1596 : }
1597 :
1598 : static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx)
1599 : {
1600 0 : __mod_lruvec_kmem_state(p, idx, -1);
1601 : }
1602 :
1603 : static inline struct lruvec *parent_lruvec(struct lruvec *lruvec)
1604 : {
1605 : struct mem_cgroup *memcg;
1606 :
1607 0 : memcg = lruvec_memcg(lruvec);
1608 : if (!memcg)
1609 : return NULL;
1610 : memcg = parent_mem_cgroup(memcg);
1611 : if (!memcg)
1612 : return NULL;
1613 : return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec));
1614 : }
1615 :
1616 : static inline void unlock_page_lruvec(struct lruvec *lruvec)
1617 : {
1618 : spin_unlock(&lruvec->lru_lock);
1619 : }
1620 :
1621 : static inline void unlock_page_lruvec_irq(struct lruvec *lruvec)
1622 : {
1623 0 : spin_unlock_irq(&lruvec->lru_lock);
1624 : }
1625 :
1626 : static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec,
1627 : unsigned long flags)
1628 : {
1629 0 : spin_unlock_irqrestore(&lruvec->lru_lock, flags);
1630 : }
1631 :
1632 : /* Test requires a stable page->memcg binding, see page_memcg() */
1633 : static inline bool folio_matches_lruvec(struct folio *folio,
1634 : struct lruvec *lruvec)
1635 : {
1636 0 : return lruvec_pgdat(lruvec) == folio_pgdat(folio) &&
1637 : lruvec_memcg(lruvec) == folio_memcg(folio);
1638 : }
1639 :
1640 : /* Don't lock again iff page's lruvec locked */
1641 0 : static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio,
1642 : struct lruvec *locked_lruvec)
1643 : {
1644 0 : if (locked_lruvec) {
1645 0 : if (folio_matches_lruvec(folio, locked_lruvec))
1646 : return locked_lruvec;
1647 :
1648 0 : unlock_page_lruvec_irq(locked_lruvec);
1649 : }
1650 :
1651 0 : return folio_lruvec_lock_irq(folio);
1652 : }
1653 :
1654 : /* Don't lock again iff page's lruvec locked */
1655 0 : static inline struct lruvec *folio_lruvec_relock_irqsave(struct folio *folio,
1656 : struct lruvec *locked_lruvec, unsigned long *flags)
1657 : {
1658 0 : if (locked_lruvec) {
1659 0 : if (folio_matches_lruvec(folio, locked_lruvec))
1660 : return locked_lruvec;
1661 :
1662 0 : unlock_page_lruvec_irqrestore(locked_lruvec, *flags);
1663 : }
1664 :
1665 0 : return folio_lruvec_lock_irqsave(folio, flags);
1666 : }
1667 :
1668 : #ifdef CONFIG_CGROUP_WRITEBACK
1669 :
1670 : struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
1671 : void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
1672 : unsigned long *pheadroom, unsigned long *pdirty,
1673 : unsigned long *pwriteback);
1674 :
1675 : void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio,
1676 : struct bdi_writeback *wb);
1677 :
1678 : static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1679 : struct bdi_writeback *wb)
1680 : {
1681 : struct mem_cgroup *memcg;
1682 :
1683 : if (mem_cgroup_disabled())
1684 : return;
1685 :
1686 : memcg = folio_memcg(folio);
1687 : if (unlikely(memcg && &memcg->css != wb->memcg_css))
1688 : mem_cgroup_track_foreign_dirty_slowpath(folio, wb);
1689 : }
1690 :
1691 : void mem_cgroup_flush_foreign(struct bdi_writeback *wb);
1692 :
1693 : #else /* CONFIG_CGROUP_WRITEBACK */
1694 :
1695 : static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
1696 : {
1697 : return NULL;
1698 : }
1699 :
1700 : static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
1701 : unsigned long *pfilepages,
1702 : unsigned long *pheadroom,
1703 : unsigned long *pdirty,
1704 : unsigned long *pwriteback)
1705 : {
1706 : }
1707 :
1708 : static inline void mem_cgroup_track_foreign_dirty(struct folio *folio,
1709 : struct bdi_writeback *wb)
1710 : {
1711 : }
1712 :
1713 : static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
1714 : {
1715 : }
1716 :
1717 : #endif /* CONFIG_CGROUP_WRITEBACK */
1718 :
1719 : struct sock;
1720 : bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages,
1721 : gfp_t gfp_mask);
1722 : void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages);
1723 : #ifdef CONFIG_MEMCG
1724 : extern struct static_key_false memcg_sockets_enabled_key;
1725 : #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1726 : void mem_cgroup_sk_alloc(struct sock *sk);
1727 : void mem_cgroup_sk_free(struct sock *sk);
1728 : static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1729 : {
1730 : if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
1731 : return true;
1732 : do {
1733 : if (time_before(jiffies, READ_ONCE(memcg->socket_pressure)))
1734 : return true;
1735 : } while ((memcg = parent_mem_cgroup(memcg)));
1736 : return false;
1737 : }
1738 :
1739 : int alloc_shrinker_info(struct mem_cgroup *memcg);
1740 : void free_shrinker_info(struct mem_cgroup *memcg);
1741 : void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id);
1742 : void reparent_shrinker_deferred(struct mem_cgroup *memcg);
1743 : #else
1744 : #define mem_cgroup_sockets_enabled 0
1745 : static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
1746 : static inline void mem_cgroup_sk_free(struct sock *sk) { };
1747 : static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
1748 : {
1749 : return false;
1750 : }
1751 :
1752 : static inline void set_shrinker_bit(struct mem_cgroup *memcg,
1753 : int nid, int shrinker_id)
1754 : {
1755 : }
1756 : #endif
1757 :
1758 : #ifdef CONFIG_MEMCG_KMEM
1759 : bool mem_cgroup_kmem_disabled(void);
1760 : int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order);
1761 : void __memcg_kmem_uncharge_page(struct page *page, int order);
1762 :
1763 : struct obj_cgroup *get_obj_cgroup_from_current(void);
1764 : struct obj_cgroup *get_obj_cgroup_from_page(struct page *page);
1765 :
1766 : int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size);
1767 : void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size);
1768 :
1769 : extern struct static_key_false memcg_bpf_enabled_key;
1770 : static inline bool memcg_bpf_enabled(void)
1771 : {
1772 : return static_branch_likely(&memcg_bpf_enabled_key);
1773 : }
1774 :
1775 : extern struct static_key_false memcg_kmem_online_key;
1776 :
1777 : static inline bool memcg_kmem_online(void)
1778 : {
1779 : return static_branch_likely(&memcg_kmem_online_key);
1780 : }
1781 :
1782 : static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1783 : int order)
1784 : {
1785 : if (memcg_kmem_online())
1786 : return __memcg_kmem_charge_page(page, gfp, order);
1787 : return 0;
1788 : }
1789 :
1790 : static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1791 : {
1792 : if (memcg_kmem_online())
1793 : __memcg_kmem_uncharge_page(page, order);
1794 : }
1795 :
1796 : /*
1797 : * A helper for accessing memcg's kmem_id, used for getting
1798 : * corresponding LRU lists.
1799 : */
1800 : static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1801 : {
1802 : return memcg ? memcg->kmemcg_id : -1;
1803 : }
1804 :
1805 : struct mem_cgroup *mem_cgroup_from_obj(void *p);
1806 : struct mem_cgroup *mem_cgroup_from_slab_obj(void *p);
1807 :
1808 : static inline void count_objcg_event(struct obj_cgroup *objcg,
1809 : enum vm_event_item idx)
1810 : {
1811 : struct mem_cgroup *memcg;
1812 :
1813 : if (!memcg_kmem_online())
1814 : return;
1815 :
1816 : rcu_read_lock();
1817 : memcg = obj_cgroup_memcg(objcg);
1818 : count_memcg_events(memcg, idx, 1);
1819 : rcu_read_unlock();
1820 : }
1821 :
1822 : #else
1823 : static inline bool mem_cgroup_kmem_disabled(void)
1824 : {
1825 : return true;
1826 : }
1827 :
1828 : static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1829 : int order)
1830 : {
1831 : return 0;
1832 : }
1833 :
1834 : static inline void memcg_kmem_uncharge_page(struct page *page, int order)
1835 : {
1836 : }
1837 :
1838 : static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp,
1839 : int order)
1840 : {
1841 : return 0;
1842 : }
1843 :
1844 : static inline void __memcg_kmem_uncharge_page(struct page *page, int order)
1845 : {
1846 : }
1847 :
1848 : static inline struct obj_cgroup *get_obj_cgroup_from_page(struct page *page)
1849 : {
1850 : return NULL;
1851 : }
1852 :
1853 : static inline bool memcg_bpf_enabled(void)
1854 : {
1855 : return false;
1856 : }
1857 :
1858 : static inline bool memcg_kmem_online(void)
1859 : {
1860 : return false;
1861 : }
1862 :
1863 : static inline int memcg_kmem_id(struct mem_cgroup *memcg)
1864 : {
1865 : return -1;
1866 : }
1867 :
1868 : static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
1869 : {
1870 : return NULL;
1871 : }
1872 :
1873 : static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p)
1874 : {
1875 : return NULL;
1876 : }
1877 :
1878 : static inline void count_objcg_event(struct obj_cgroup *objcg,
1879 : enum vm_event_item idx)
1880 : {
1881 : }
1882 :
1883 : #endif /* CONFIG_MEMCG_KMEM */
1884 :
1885 : #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP)
1886 : bool obj_cgroup_may_zswap(struct obj_cgroup *objcg);
1887 : void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size);
1888 : void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size);
1889 : #else
1890 : static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg)
1891 : {
1892 : return true;
1893 : }
1894 : static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg,
1895 : size_t size)
1896 : {
1897 : }
1898 : static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg,
1899 : size_t size)
1900 : {
1901 : }
1902 : #endif
1903 :
1904 : #endif /* _LINUX_MEMCONTROL_H */
|
