Commit | Line | Data |
---|---|---|
c942fddf | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
8cdea7c0 BS |
2 | /* memcontrol.h - Memory Controller |
3 | * | |
4 | * Copyright IBM Corporation, 2007 | |
5 | * Author Balbir Singh <balbir@linux.vnet.ibm.com> | |
6 | * | |
78fb7466 PE |
7 | * Copyright 2007 OpenVZ SWsoft Inc |
8 | * Author: Pavel Emelianov <xemul@openvz.org> | |
8cdea7c0 BS |
9 | */ |
10 | ||
11 | #ifndef _LINUX_MEMCONTROL_H | |
12 | #define _LINUX_MEMCONTROL_H | |
f8d66542 | 13 | #include <linux/cgroup.h> |
456f998e | 14 | #include <linux/vm_event_item.h> |
7ae1e1d0 | 15 | #include <linux/hardirq.h> |
a8964b9b | 16 | #include <linux/jump_label.h> |
33398cf2 MH |
17 | #include <linux/page_counter.h> |
18 | #include <linux/vmpressure.h> | |
19 | #include <linux/eventfd.h> | |
00f3ca2c JW |
20 | #include <linux/mm.h> |
21 | #include <linux/vmstat.h> | |
33398cf2 | 22 | #include <linux/writeback.h> |
fdf1cdb9 | 23 | #include <linux/page-flags.h> |
456f998e | 24 | |
78fb7466 | 25 | struct mem_cgroup; |
bf4f0599 | 26 | struct obj_cgroup; |
8697d331 BS |
27 | struct page; |
28 | struct mm_struct; | |
2633d7a0 | 29 | struct kmem_cache; |
78fb7466 | 30 | |
71cd3113 JW |
31 | /* Cgroup-specific page state, on top of universal node page state */ |
32 | enum memcg_stat_item { | |
468c3982 | 33 | MEMCG_SWAP = NR_VM_NODE_STAT_ITEMS, |
71cd3113 | 34 | MEMCG_SOCK, |
772616b0 | 35 | MEMCG_PERCPU_B, |
4e5aa1f4 | 36 | MEMCG_VMALLOC, |
a8c49af3 | 37 | MEMCG_KMEM, |
f4840ccf JW |
38 | MEMCG_ZSWAP_B, |
39 | MEMCG_ZSWAPPED, | |
b2807f07 | 40 | MEMCG_NR_STAT, |
2a7106f2 GT |
41 | }; |
42 | ||
e27be240 JW |
43 | enum memcg_memory_event { |
44 | MEMCG_LOW, | |
71cd3113 JW |
45 | MEMCG_HIGH, |
46 | MEMCG_MAX, | |
47 | MEMCG_OOM, | |
fe6bdfc8 | 48 | MEMCG_OOM_KILL, |
b6bf9abb | 49 | MEMCG_OOM_GROUP_KILL, |
4b82ab4f | 50 | MEMCG_SWAP_HIGH, |
f3a53a3a TH |
51 | MEMCG_SWAP_MAX, |
52 | MEMCG_SWAP_FAIL, | |
e27be240 | 53 | MEMCG_NR_MEMORY_EVENTS, |
71cd3113 JW |
54 | }; |
55 | ||
5660048c | 56 | struct mem_cgroup_reclaim_cookie { |
ef8f2327 | 57 | pg_data_t *pgdat; |
5660048c JW |
58 | unsigned int generation; |
59 | }; | |
60 | ||
71cd3113 JW |
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; | |
1c2d479a | 68 | refcount_t ref; |
71cd3113 JW |
69 | }; |
70 | ||
33398cf2 MH |
71 | /* |
72 | * Per memcg event counter is incremented at every pagein/pageout. With THP, | |
0845f831 RD |
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 | |
33398cf2 MH |
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, | |
33398cf2 MH |
80 | MEM_CGROUP_NTARGETS, |
81 | }; | |
82 | ||
410f8e82 SB |
83 | struct memcg_vmstats_percpu; |
84 | struct memcg_vmstats; | |
33398cf2 MH |
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 | ||
0a4465d3 | 92 | /* |
3c6f17e6 YS |
93 | * Bitmap and deferred work of shrinker::id corresponding to memcg-aware |
94 | * shrinkers, which have elements charged to this memcg. | |
0a4465d3 | 95 | */ |
e4262c4f | 96 | struct shrinker_info { |
0a4465d3 | 97 | struct rcu_head rcu; |
3c6f17e6 YS |
98 | atomic_long_t *nr_deferred; |
99 | unsigned long *map; | |
0a4465d3 KT |
100 | }; |
101 | ||
7e1c0d6f SB |
102 | struct lruvec_stats_percpu { |
103 | /* Local (CPU and cgroup) state */ | |
104 | long state[NR_VM_NODE_STAT_ITEMS]; | |
105 | ||
106 | /* Delta calculation for lockless upward propagation */ | |
107 | long state_prev[NR_VM_NODE_STAT_ITEMS]; | |
108 | }; | |
109 | ||
110 | struct lruvec_stats { | |
111 | /* Aggregated (CPU and subtree) state */ | |
112 | long state[NR_VM_NODE_STAT_ITEMS]; | |
113 | ||
114 | /* Pending child counts during tree propagation */ | |
115 | long state_pending[NR_VM_NODE_STAT_ITEMS]; | |
116 | }; | |
117 | ||
33398cf2 | 118 | /* |
242c37b4 | 119 | * per-node information in memory controller. |
33398cf2 | 120 | */ |
ef8f2327 | 121 | struct mem_cgroup_per_node { |
33398cf2 | 122 | struct lruvec lruvec; |
a983b5eb | 123 | |
7e1c0d6f SB |
124 | struct lruvec_stats_percpu __percpu *lruvec_stats_percpu; |
125 | struct lruvec_stats lruvec_stats; | |
a983b5eb | 126 | |
b4536f0c | 127 | unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS]; |
33398cf2 | 128 | |
9da83f3f | 129 | struct mem_cgroup_reclaim_iter iter; |
33398cf2 | 130 | |
e4262c4f | 131 | struct shrinker_info __rcu *shrinker_info; |
0a432dcb | 132 | |
33398cf2 MH |
133 | struct rb_node tree_node; /* RB tree node */ |
134 | unsigned long usage_in_excess;/* Set to the value by which */ | |
135 | /* the soft limit is exceeded*/ | |
136 | bool on_tree; | |
137 | struct mem_cgroup *memcg; /* Back pointer, we cannot */ | |
138 | /* use container_of */ | |
139 | }; | |
140 | ||
33398cf2 MH |
141 | struct mem_cgroup_threshold { |
142 | struct eventfd_ctx *eventfd; | |
143 | unsigned long threshold; | |
144 | }; | |
145 | ||
146 | /* For threshold */ | |
147 | struct mem_cgroup_threshold_ary { | |
148 | /* An array index points to threshold just below or equal to usage. */ | |
149 | int current_threshold; | |
150 | /* Size of entries[] */ | |
151 | unsigned int size; | |
152 | /* Array of thresholds */ | |
307ed94c | 153 | struct mem_cgroup_threshold entries[]; |
33398cf2 MH |
154 | }; |
155 | ||
156 | struct mem_cgroup_thresholds { | |
157 | /* Primary thresholds array */ | |
158 | struct mem_cgroup_threshold_ary *primary; | |
159 | /* | |
160 | * Spare threshold array. | |
161 | * This is needed to make mem_cgroup_unregister_event() "never fail". | |
162 | * It must be able to store at least primary->size - 1 entries. | |
163 | */ | |
164 | struct mem_cgroup_threshold_ary *spare; | |
165 | }; | |
166 | ||
97b27821 TH |
167 | /* |
168 | * Remember four most recent foreign writebacks with dirty pages in this | |
169 | * cgroup. Inode sharing is expected to be uncommon and, even if we miss | |
170 | * one in a given round, we're likely to catch it later if it keeps | |
171 | * foreign-dirtying, so a fairly low count should be enough. | |
172 | * | |
173 | * See mem_cgroup_track_foreign_dirty_slowpath() for details. | |
174 | */ | |
175 | #define MEMCG_CGWB_FRN_CNT 4 | |
176 | ||
177 | struct memcg_cgwb_frn { | |
178 | u64 bdi_id; /* bdi->id of the foreign inode */ | |
179 | int memcg_id; /* memcg->css.id of foreign inode */ | |
180 | u64 at; /* jiffies_64 at the time of dirtying */ | |
181 | struct wb_completion done; /* tracks in-flight foreign writebacks */ | |
182 | }; | |
183 | ||
bf4f0599 RG |
184 | /* |
185 | * Bucket for arbitrarily byte-sized objects charged to a memory | |
186 | * cgroup. The bucket can be reparented in one piece when the cgroup | |
187 | * is destroyed, without having to round up the individual references | |
188 | * of all live memory objects in the wild. | |
189 | */ | |
190 | struct obj_cgroup { | |
191 | struct percpu_ref refcnt; | |
192 | struct mem_cgroup *memcg; | |
193 | atomic_t nr_charged_bytes; | |
194 | union { | |
0764db9b | 195 | struct list_head list; /* protected by objcg_lock */ |
bf4f0599 RG |
196 | struct rcu_head rcu; |
197 | }; | |
198 | }; | |
199 | ||
33398cf2 MH |
200 | /* |
201 | * The memory controller data structure. The memory controller controls both | |
202 | * page cache and RSS per cgroup. We would eventually like to provide | |
203 | * statistics based on the statistics developed by Rik Van Riel for clock-pro, | |
204 | * to help the administrator determine what knobs to tune. | |
205 | */ | |
206 | struct mem_cgroup { | |
207 | struct cgroup_subsys_state css; | |
208 | ||
73f576c0 JW |
209 | /* Private memcg ID. Used to ID objects that outlive the cgroup */ |
210 | struct mem_cgroup_id id; | |
211 | ||
33398cf2 | 212 | /* Accounted resources */ |
bd0b230f WL |
213 | struct page_counter memory; /* Both v1 & v2 */ |
214 | ||
215 | union { | |
216 | struct page_counter swap; /* v2 only */ | |
217 | struct page_counter memsw; /* v1 only */ | |
218 | }; | |
0db15298 JW |
219 | |
220 | /* Legacy consumer-oriented counters */ | |
bd0b230f WL |
221 | struct page_counter kmem; /* v1 only */ |
222 | struct page_counter tcpmem; /* v1 only */ | |
33398cf2 | 223 | |
f7e1cb6e JW |
224 | /* Range enforcement for interrupt charges */ |
225 | struct work_struct high_work; | |
226 | ||
f4840ccf JW |
227 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP) |
228 | unsigned long zswap_max; | |
229 | #endif | |
230 | ||
33398cf2 MH |
231 | unsigned long soft_limit; |
232 | ||
233 | /* vmpressure notifications */ | |
234 | struct vmpressure vmpressure; | |
235 | ||
3d8b38eb RG |
236 | /* |
237 | * Should the OOM killer kill all belonging tasks, had it kill one? | |
238 | */ | |
239 | bool oom_group; | |
240 | ||
33398cf2 MH |
241 | /* protected by memcg_oom_lock */ |
242 | bool oom_lock; | |
243 | int under_oom; | |
244 | ||
245 | int swappiness; | |
246 | /* OOM-Killer disable */ | |
247 | int oom_kill_disable; | |
248 | ||
1e577f97 | 249 | /* memory.events and memory.events.local */ |
472912a2 | 250 | struct cgroup_file events_file; |
1e577f97 | 251 | struct cgroup_file events_local_file; |
472912a2 | 252 | |
f3a53a3a TH |
253 | /* handle for "memory.swap.events" */ |
254 | struct cgroup_file swap_events_file; | |
255 | ||
33398cf2 MH |
256 | /* protect arrays of thresholds */ |
257 | struct mutex thresholds_lock; | |
258 | ||
259 | /* thresholds for memory usage. RCU-protected */ | |
260 | struct mem_cgroup_thresholds thresholds; | |
261 | ||
262 | /* thresholds for mem+swap usage. RCU-protected */ | |
263 | struct mem_cgroup_thresholds memsw_thresholds; | |
264 | ||
265 | /* For oom notifier event fd */ | |
266 | struct list_head oom_notify; | |
267 | ||
268 | /* | |
269 | * Should we move charges of a task when a task is moved into this | |
270 | * mem_cgroup ? And what type of charges should we move ? | |
271 | */ | |
272 | unsigned long move_charge_at_immigrate; | |
e81bf979 AL |
273 | /* taken only while moving_account > 0 */ |
274 | spinlock_t move_lock; | |
275 | unsigned long move_lock_flags; | |
276 | ||
e6ad640b | 277 | CACHELINE_PADDING(_pad1_); |
e81bf979 | 278 | |
2d146aa3 | 279 | /* memory.stat */ |
410f8e82 | 280 | struct memcg_vmstats *vmstats; |
42a30035 | 281 | |
815744d7 | 282 | /* memory.events */ |
42a30035 | 283 | atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS]; |
1e577f97 | 284 | atomic_long_t memory_events_local[MEMCG_NR_MEMORY_EVENTS]; |
33398cf2 | 285 | |
d886f4e4 JW |
286 | unsigned long socket_pressure; |
287 | ||
288 | /* Legacy tcp memory accounting */ | |
0db15298 JW |
289 | bool tcpmem_active; |
290 | int tcpmem_pressure; | |
d886f4e4 | 291 | |
84c07d11 | 292 | #ifdef CONFIG_MEMCG_KMEM |
33398cf2 | 293 | int kmemcg_id; |
bf4f0599 | 294 | struct obj_cgroup __rcu *objcg; |
0764db9b RG |
295 | /* list of inherited objcgs, protected by objcg_lock */ |
296 | struct list_head objcg_list; | |
33398cf2 MH |
297 | #endif |
298 | ||
e6ad640b | 299 | CACHELINE_PADDING(_pad2_); |
4df91062 FT |
300 | |
301 | /* | |
302 | * set > 0 if pages under this cgroup are moving to other cgroup. | |
303 | */ | |
304 | atomic_t moving_account; | |
305 | struct task_struct *move_lock_task; | |
306 | ||
4df91062 FT |
307 | struct memcg_vmstats_percpu __percpu *vmstats_percpu; |
308 | ||
33398cf2 MH |
309 | #ifdef CONFIG_CGROUP_WRITEBACK |
310 | struct list_head cgwb_list; | |
311 | struct wb_domain cgwb_domain; | |
97b27821 | 312 | struct memcg_cgwb_frn cgwb_frn[MEMCG_CGWB_FRN_CNT]; |
33398cf2 MH |
313 | #endif |
314 | ||
315 | /* List of events which userspace want to receive */ | |
316 | struct list_head event_list; | |
317 | spinlock_t event_list_lock; | |
318 | ||
87eaceb3 YS |
319 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
320 | struct deferred_split deferred_split_queue; | |
321 | #endif | |
322 | ||
bd74fdae YZ |
323 | #ifdef CONFIG_LRU_GEN |
324 | /* per-memcg mm_struct list */ | |
325 | struct lru_gen_mm_list mm_list; | |
326 | #endif | |
327 | ||
b51478a0 | 328 | struct mem_cgroup_per_node *nodeinfo[]; |
33398cf2 | 329 | }; |
7d828602 | 330 | |
a983b5eb | 331 | /* |
1813e51e SB |
332 | * size of first charge trial. |
333 | * TODO: maybe necessary to use big numbers in big irons or dynamic based of the | |
334 | * workload. | |
a983b5eb | 335 | */ |
1813e51e | 336 | #define MEMCG_CHARGE_BATCH 64U |
a983b5eb | 337 | |
7d828602 | 338 | extern struct mem_cgroup *root_mem_cgroup; |
56161634 | 339 | |
87944e29 RG |
340 | enum page_memcg_data_flags { |
341 | /* page->memcg_data is a pointer to an objcgs vector */ | |
342 | MEMCG_DATA_OBJCGS = (1UL << 0), | |
18b2db3b RG |
343 | /* page has been accounted as a non-slab kernel page */ |
344 | MEMCG_DATA_KMEM = (1UL << 1), | |
87944e29 | 345 | /* the next bit after the last actual flag */ |
18b2db3b | 346 | __NR_MEMCG_DATA_FLAGS = (1UL << 2), |
87944e29 RG |
347 | }; |
348 | ||
349 | #define MEMCG_DATA_FLAGS_MASK (__NR_MEMCG_DATA_FLAGS - 1) | |
350 | ||
1b7e4464 | 351 | static inline bool folio_memcg_kmem(struct folio *folio); |
b4e0b68f MS |
352 | |
353 | /* | |
354 | * After the initialization objcg->memcg is always pointing at | |
355 | * a valid memcg, but can be atomically swapped to the parent memcg. | |
356 | * | |
357 | * The caller must ensure that the returned memcg won't be released: | |
358 | * e.g. acquire the rcu_read_lock or css_set_lock. | |
359 | */ | |
360 | static inline struct mem_cgroup *obj_cgroup_memcg(struct obj_cgroup *objcg) | |
361 | { | |
362 | return READ_ONCE(objcg->memcg); | |
363 | } | |
364 | ||
365 | /* | |
1b7e4464 MWO |
366 | * __folio_memcg - Get the memory cgroup associated with a non-kmem folio |
367 | * @folio: Pointer to the folio. | |
b4e0b68f | 368 | * |
1b7e4464 MWO |
369 | * Returns a pointer to the memory cgroup associated with the folio, |
370 | * or NULL. This function assumes that the folio is known to have a | |
b4e0b68f | 371 | * proper memory cgroup pointer. It's not safe to call this function |
1b7e4464 MWO |
372 | * against some type of folios, e.g. slab folios or ex-slab folios or |
373 | * kmem folios. | |
b4e0b68f | 374 | */ |
1b7e4464 | 375 | static inline struct mem_cgroup *__folio_memcg(struct folio *folio) |
b4e0b68f | 376 | { |
1b7e4464 | 377 | unsigned long memcg_data = folio->memcg_data; |
b4e0b68f | 378 | |
1b7e4464 MWO |
379 | VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); |
380 | VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio); | |
381 | VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_KMEM, folio); | |
b4e0b68f MS |
382 | |
383 | return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); | |
384 | } | |
385 | ||
386 | /* | |
1b7e4464 MWO |
387 | * __folio_objcg - get the object cgroup associated with a kmem folio. |
388 | * @folio: Pointer to the folio. | |
b4e0b68f | 389 | * |
1b7e4464 MWO |
390 | * Returns a pointer to the object cgroup associated with the folio, |
391 | * or NULL. This function assumes that the folio is known to have a | |
b4e0b68f | 392 | * proper object cgroup pointer. It's not safe to call this function |
1b7e4464 MWO |
393 | * against some type of folios, e.g. slab folios or ex-slab folios or |
394 | * LRU folios. | |
b4e0b68f | 395 | */ |
1b7e4464 | 396 | static inline struct obj_cgroup *__folio_objcg(struct folio *folio) |
b4e0b68f | 397 | { |
1b7e4464 | 398 | unsigned long memcg_data = folio->memcg_data; |
b4e0b68f | 399 | |
1b7e4464 MWO |
400 | VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); |
401 | VM_BUG_ON_FOLIO(memcg_data & MEMCG_DATA_OBJCGS, folio); | |
402 | VM_BUG_ON_FOLIO(!(memcg_data & MEMCG_DATA_KMEM), folio); | |
b4e0b68f MS |
403 | |
404 | return (struct obj_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); | |
405 | } | |
406 | ||
bcfe06bf | 407 | /* |
1b7e4464 MWO |
408 | * folio_memcg - Get the memory cgroup associated with a folio. |
409 | * @folio: Pointer to the folio. | |
bcfe06bf | 410 | * |
1b7e4464 MWO |
411 | * Returns a pointer to the memory cgroup associated with the folio, |
412 | * or NULL. This function assumes that the folio is known to have a | |
bcfe06bf | 413 | * proper memory cgroup pointer. It's not safe to call this function |
1b7e4464 | 414 | * against some type of folios, e.g. slab folios or ex-slab folios. |
bcfe06bf | 415 | * |
1b7e4464 | 416 | * For a non-kmem folio any of the following ensures folio and memcg binding |
b4e0b68f MS |
417 | * stability: |
418 | * | |
1b7e4464 | 419 | * - the folio lock |
bcfe06bf RG |
420 | * - LRU isolation |
421 | * - lock_page_memcg() | |
422 | * - exclusive reference | |
018ee47f | 423 | * - mem_cgroup_trylock_pages() |
b4e0b68f | 424 | * |
1b7e4464 MWO |
425 | * For a kmem folio a caller should hold an rcu read lock to protect memcg |
426 | * associated with a kmem folio from being released. | |
bcfe06bf | 427 | */ |
1b7e4464 MWO |
428 | static inline struct mem_cgroup *folio_memcg(struct folio *folio) |
429 | { | |
430 | if (folio_memcg_kmem(folio)) | |
431 | return obj_cgroup_memcg(__folio_objcg(folio)); | |
432 | return __folio_memcg(folio); | |
433 | } | |
434 | ||
bcfe06bf RG |
435 | static inline struct mem_cgroup *page_memcg(struct page *page) |
436 | { | |
1b7e4464 | 437 | return folio_memcg(page_folio(page)); |
bcfe06bf RG |
438 | } |
439 | ||
c5ce619a MWO |
440 | /** |
441 | * folio_memcg_rcu - Locklessly get the memory cgroup associated with a folio. | |
442 | * @folio: Pointer to the folio. | |
bcfe06bf | 443 | * |
c5ce619a | 444 | * This function assumes that the folio is known to have a |
bcfe06bf | 445 | * proper memory cgroup pointer. It's not safe to call this function |
c5ce619a MWO |
446 | * against some type of folios, e.g. slab folios or ex-slab folios. |
447 | * | |
448 | * Return: A pointer to the memory cgroup associated with the folio, | |
449 | * or NULL. | |
bcfe06bf | 450 | */ |
c5ce619a | 451 | static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) |
bcfe06bf | 452 | { |
c5ce619a | 453 | unsigned long memcg_data = READ_ONCE(folio->memcg_data); |
b4e0b68f | 454 | |
c5ce619a | 455 | VM_BUG_ON_FOLIO(folio_test_slab(folio), folio); |
bcfe06bf RG |
456 | WARN_ON_ONCE(!rcu_read_lock_held()); |
457 | ||
b4e0b68f MS |
458 | if (memcg_data & MEMCG_DATA_KMEM) { |
459 | struct obj_cgroup *objcg; | |
460 | ||
461 | objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); | |
462 | return obj_cgroup_memcg(objcg); | |
463 | } | |
464 | ||
465 | return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); | |
bcfe06bf RG |
466 | } |
467 | ||
468 | /* | |
469 | * page_memcg_check - get the memory cgroup associated with a page | |
470 | * @page: a pointer to the page struct | |
471 | * | |
472 | * Returns a pointer to the memory cgroup associated with the page, | |
b4e0b68f | 473 | * or NULL. This function unlike page_memcg() can take any page |
bcfe06bf | 474 | * as an argument. It has to be used in cases when it's not known if a page |
b4e0b68f MS |
475 | * has an associated memory cgroup pointer or an object cgroups vector or |
476 | * an object cgroup. | |
477 | * | |
478 | * For a non-kmem page any of the following ensures page and memcg binding | |
479 | * stability: | |
bcfe06bf | 480 | * |
bcfe06bf RG |
481 | * - the page lock |
482 | * - LRU isolation | |
483 | * - lock_page_memcg() | |
484 | * - exclusive reference | |
018ee47f | 485 | * - mem_cgroup_trylock_pages() |
b4e0b68f MS |
486 | * |
487 | * For a kmem page a caller should hold an rcu read lock to protect memcg | |
488 | * associated with a kmem page from being released. | |
bcfe06bf RG |
489 | */ |
490 | static inline struct mem_cgroup *page_memcg_check(struct page *page) | |
491 | { | |
492 | /* | |
493 | * Because page->memcg_data might be changed asynchronously | |
494 | * for slab pages, READ_ONCE() should be used here. | |
495 | */ | |
496 | unsigned long memcg_data = READ_ONCE(page->memcg_data); | |
497 | ||
87944e29 | 498 | if (memcg_data & MEMCG_DATA_OBJCGS) |
bcfe06bf RG |
499 | return NULL; |
500 | ||
b4e0b68f MS |
501 | if (memcg_data & MEMCG_DATA_KMEM) { |
502 | struct obj_cgroup *objcg; | |
503 | ||
504 | objcg = (void *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); | |
505 | return obj_cgroup_memcg(objcg); | |
506 | } | |
507 | ||
18b2db3b RG |
508 | return (struct mem_cgroup *)(memcg_data & ~MEMCG_DATA_FLAGS_MASK); |
509 | } | |
510 | ||
88f2ef73 MS |
511 | static inline struct mem_cgroup *get_mem_cgroup_from_objcg(struct obj_cgroup *objcg) |
512 | { | |
513 | struct mem_cgroup *memcg; | |
514 | ||
515 | rcu_read_lock(); | |
516 | retry: | |
517 | memcg = obj_cgroup_memcg(objcg); | |
518 | if (unlikely(!css_tryget(&memcg->css))) | |
519 | goto retry; | |
520 | rcu_read_unlock(); | |
521 | ||
522 | return memcg; | |
523 | } | |
524 | ||
bd290e1e | 525 | #ifdef CONFIG_MEMCG_KMEM |
18b2db3b | 526 | /* |
1b7e4464 MWO |
527 | * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set. |
528 | * @folio: Pointer to the folio. | |
18b2db3b | 529 | * |
1b7e4464 MWO |
530 | * Checks if the folio has MemcgKmem flag set. The caller must ensure |
531 | * that the folio has an associated memory cgroup. It's not safe to call | |
532 | * this function against some types of folios, e.g. slab folios. | |
18b2db3b | 533 | */ |
1b7e4464 | 534 | static inline bool folio_memcg_kmem(struct folio *folio) |
18b2db3b | 535 | { |
1b7e4464 MWO |
536 | VM_BUG_ON_PGFLAGS(PageTail(&folio->page), &folio->page); |
537 | VM_BUG_ON_FOLIO(folio->memcg_data & MEMCG_DATA_OBJCGS, folio); | |
538 | return folio->memcg_data & MEMCG_DATA_KMEM; | |
bcfe06bf RG |
539 | } |
540 | ||
270c6a71 | 541 | |
270c6a71 | 542 | #else |
1b7e4464 | 543 | static inline bool folio_memcg_kmem(struct folio *folio) |
bd290e1e MS |
544 | { |
545 | return false; | |
546 | } | |
547 | ||
270c6a71 RG |
548 | #endif |
549 | ||
1b7e4464 MWO |
550 | static inline bool PageMemcgKmem(struct page *page) |
551 | { | |
552 | return folio_memcg_kmem(page_folio(page)); | |
553 | } | |
554 | ||
dfd2f10c KT |
555 | static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) |
556 | { | |
557 | return (memcg == root_mem_cgroup); | |
558 | } | |
559 | ||
23047a96 JW |
560 | static inline bool mem_cgroup_disabled(void) |
561 | { | |
562 | return !cgroup_subsys_enabled(memory_cgrp_subsys); | |
563 | } | |
564 | ||
f56ce412 JW |
565 | static inline void mem_cgroup_protection(struct mem_cgroup *root, |
566 | struct mem_cgroup *memcg, | |
567 | unsigned long *min, | |
568 | unsigned long *low) | |
9783aa99 | 569 | { |
f56ce412 JW |
570 | *min = *low = 0; |
571 | ||
1bc63fb1 | 572 | if (mem_cgroup_disabled()) |
f56ce412 | 573 | return; |
1bc63fb1 | 574 | |
22f7496f YS |
575 | /* |
576 | * There is no reclaim protection applied to a targeted reclaim. | |
577 | * We are special casing this specific case here because | |
578 | * mem_cgroup_protected calculation is not robust enough to keep | |
579 | * the protection invariant for calculated effective values for | |
580 | * parallel reclaimers with different reclaim target. This is | |
581 | * especially a problem for tail memcgs (as they have pages on LRU) | |
582 | * which would want to have effective values 0 for targeted reclaim | |
583 | * but a different value for external reclaim. | |
584 | * | |
585 | * Example | |
586 | * Let's have global and A's reclaim in parallel: | |
587 | * | | |
588 | * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G) | |
589 | * |\ | |
590 | * | C (low = 1G, usage = 2.5G) | |
591 | * B (low = 1G, usage = 0.5G) | |
592 | * | |
593 | * For the global reclaim | |
594 | * A.elow = A.low | |
595 | * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow | |
596 | * C.elow = min(C.usage, C.low) | |
597 | * | |
598 | * With the effective values resetting we have A reclaim | |
599 | * A.elow = 0 | |
600 | * B.elow = B.low | |
601 | * C.elow = C.low | |
602 | * | |
603 | * If the global reclaim races with A's reclaim then | |
604 | * B.elow = C.elow = 0 because children_low_usage > A.elow) | |
605 | * is possible and reclaiming B would be violating the protection. | |
606 | * | |
607 | */ | |
608 | if (root == memcg) | |
f56ce412 | 609 | return; |
9783aa99 | 610 | |
f56ce412 JW |
611 | *min = READ_ONCE(memcg->memory.emin); |
612 | *low = READ_ONCE(memcg->memory.elow); | |
9783aa99 CD |
613 | } |
614 | ||
45c7f7e1 CD |
615 | void mem_cgroup_calculate_protection(struct mem_cgroup *root, |
616 | struct mem_cgroup *memcg); | |
617 | ||
adb82130 YA |
618 | static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, |
619 | struct mem_cgroup *memcg) | |
45c7f7e1 CD |
620 | { |
621 | /* | |
622 | * The root memcg doesn't account charges, and doesn't support | |
adb82130 YA |
623 | * protection. The target memcg's protection is ignored, see |
624 | * mem_cgroup_calculate_protection() and mem_cgroup_protection() | |
45c7f7e1 | 625 | */ |
adb82130 YA |
626 | return mem_cgroup_disabled() || mem_cgroup_is_root(memcg) || |
627 | memcg == target; | |
45c7f7e1 CD |
628 | } |
629 | ||
adb82130 YA |
630 | static inline bool mem_cgroup_below_low(struct mem_cgroup *target, |
631 | struct mem_cgroup *memcg) | |
45c7f7e1 | 632 | { |
adb82130 | 633 | if (mem_cgroup_unprotected(target, memcg)) |
45c7f7e1 CD |
634 | return false; |
635 | ||
636 | return READ_ONCE(memcg->memory.elow) >= | |
637 | page_counter_read(&memcg->memory); | |
638 | } | |
639 | ||
adb82130 YA |
640 | static inline bool mem_cgroup_below_min(struct mem_cgroup *target, |
641 | struct mem_cgroup *memcg) | |
45c7f7e1 | 642 | { |
adb82130 | 643 | if (mem_cgroup_unprotected(target, memcg)) |
45c7f7e1 CD |
644 | return false; |
645 | ||
646 | return READ_ONCE(memcg->memory.emin) >= | |
647 | page_counter_read(&memcg->memory); | |
648 | } | |
241994ed | 649 | |
8f425e4e MWO |
650 | int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp); |
651 | ||
652 | /** | |
653 | * mem_cgroup_charge - Charge a newly allocated folio to a cgroup. | |
654 | * @folio: Folio to charge. | |
655 | * @mm: mm context of the allocating task. | |
656 | * @gfp: Reclaim mode. | |
657 | * | |
658 | * Try to charge @folio to the memcg that @mm belongs to, reclaiming | |
659 | * pages according to @gfp if necessary. If @mm is NULL, try to | |
660 | * charge to the active memcg. | |
661 | * | |
662 | * Do not use this for folios allocated for swapin. | |
663 | * | |
664 | * Return: 0 on success. Otherwise, an error code is returned. | |
665 | */ | |
666 | static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, | |
667 | gfp_t gfp) | |
2c8d8f97 SB |
668 | { |
669 | if (mem_cgroup_disabled()) | |
670 | return 0; | |
8f425e4e | 671 | return __mem_cgroup_charge(folio, mm, gfp); |
2c8d8f97 SB |
672 | } |
673 | ||
65995918 | 674 | int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm, |
0add0c77 SB |
675 | gfp_t gfp, swp_entry_t entry); |
676 | void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry); | |
3fea5a49 | 677 | |
bbc6b703 MWO |
678 | void __mem_cgroup_uncharge(struct folio *folio); |
679 | ||
680 | /** | |
681 | * mem_cgroup_uncharge - Uncharge a folio. | |
682 | * @folio: Folio to uncharge. | |
683 | * | |
684 | * Uncharge a folio previously charged with mem_cgroup_charge(). | |
685 | */ | |
686 | static inline void mem_cgroup_uncharge(struct folio *folio) | |
2c8d8f97 SB |
687 | { |
688 | if (mem_cgroup_disabled()) | |
689 | return; | |
bbc6b703 | 690 | __mem_cgroup_uncharge(folio); |
2c8d8f97 SB |
691 | } |
692 | ||
693 | void __mem_cgroup_uncharge_list(struct list_head *page_list); | |
694 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) | |
695 | { | |
696 | if (mem_cgroup_disabled()) | |
697 | return; | |
698 | __mem_cgroup_uncharge_list(page_list); | |
699 | } | |
569b846d | 700 | |
d21bba2b | 701 | void mem_cgroup_migrate(struct folio *old, struct folio *new); |
569b846d | 702 | |
55779ec7 | 703 | /** |
867e5e1d | 704 | * mem_cgroup_lruvec - get the lru list vector for a memcg & node |
55779ec7 | 705 | * @memcg: memcg of the wanted lruvec |
9a1ac228 | 706 | * @pgdat: pglist_data |
55779ec7 | 707 | * |
867e5e1d | 708 | * Returns the lru list vector holding pages for a given @memcg & |
9a1ac228 | 709 | * @pgdat combination. This can be the node lruvec, if the memory |
867e5e1d | 710 | * controller is disabled. |
55779ec7 | 711 | */ |
867e5e1d JW |
712 | static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, |
713 | struct pglist_data *pgdat) | |
55779ec7 | 714 | { |
ef8f2327 | 715 | struct mem_cgroup_per_node *mz; |
55779ec7 JW |
716 | struct lruvec *lruvec; |
717 | ||
718 | if (mem_cgroup_disabled()) { | |
867e5e1d | 719 | lruvec = &pgdat->__lruvec; |
55779ec7 JW |
720 | goto out; |
721 | } | |
722 | ||
1b05117d JW |
723 | if (!memcg) |
724 | memcg = root_mem_cgroup; | |
725 | ||
a3747b53 | 726 | mz = memcg->nodeinfo[pgdat->node_id]; |
55779ec7 JW |
727 | lruvec = &mz->lruvec; |
728 | out: | |
729 | /* | |
730 | * Since a node can be onlined after the mem_cgroup was created, | |
599d0c95 | 731 | * we have to be prepared to initialize lruvec->pgdat here; |
55779ec7 JW |
732 | * and if offlined then reonlined, we need to reinitialize it. |
733 | */ | |
ef8f2327 MG |
734 | if (unlikely(lruvec->pgdat != pgdat)) |
735 | lruvec->pgdat = pgdat; | |
55779ec7 JW |
736 | return lruvec; |
737 | } | |
738 | ||
9a1ac228 | 739 | /** |
b1baabd9 MWO |
740 | * folio_lruvec - return lruvec for isolating/putting an LRU folio |
741 | * @folio: Pointer to the folio. | |
9a1ac228 | 742 | * |
b1baabd9 | 743 | * This function relies on folio->mem_cgroup being stable. |
9a1ac228 | 744 | */ |
b1baabd9 | 745 | static inline struct lruvec *folio_lruvec(struct folio *folio) |
9a1ac228 | 746 | { |
b1baabd9 | 747 | struct mem_cgroup *memcg = folio_memcg(folio); |
9a1ac228 | 748 | |
b1baabd9 MWO |
749 | VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio); |
750 | return mem_cgroup_lruvec(memcg, folio_pgdat(folio)); | |
9a1ac228 | 751 | } |
c9b0ed51 | 752 | |
64219994 | 753 | struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p); |
e993d905 | 754 | |
d46eb14b SB |
755 | struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm); |
756 | ||
e809c3fe MWO |
757 | struct lruvec *folio_lruvec_lock(struct folio *folio); |
758 | struct lruvec *folio_lruvec_lock_irq(struct folio *folio); | |
759 | struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, | |
6168d0da AS |
760 | unsigned long *flags); |
761 | ||
762 | #ifdef CONFIG_DEBUG_VM | |
e809c3fe | 763 | void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio); |
6168d0da | 764 | #else |
e809c3fe MWO |
765 | static inline |
766 | void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) | |
6168d0da AS |
767 | { |
768 | } | |
769 | #endif | |
770 | ||
33398cf2 MH |
771 | static inline |
772 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){ | |
773 | return css ? container_of(css, struct mem_cgroup, css) : NULL; | |
774 | } | |
775 | ||
bf4f0599 RG |
776 | static inline bool obj_cgroup_tryget(struct obj_cgroup *objcg) |
777 | { | |
778 | return percpu_ref_tryget(&objcg->refcnt); | |
779 | } | |
780 | ||
781 | static inline void obj_cgroup_get(struct obj_cgroup *objcg) | |
782 | { | |
783 | percpu_ref_get(&objcg->refcnt); | |
784 | } | |
785 | ||
b4e0b68f MS |
786 | static inline void obj_cgroup_get_many(struct obj_cgroup *objcg, |
787 | unsigned long nr) | |
bf4f0599 | 788 | { |
b4e0b68f | 789 | percpu_ref_get_many(&objcg->refcnt, nr); |
bf4f0599 RG |
790 | } |
791 | ||
b4e0b68f | 792 | static inline void obj_cgroup_put(struct obj_cgroup *objcg) |
bf4f0599 | 793 | { |
b4e0b68f | 794 | percpu_ref_put(&objcg->refcnt); |
bf4f0599 RG |
795 | } |
796 | ||
dc0b5864 RG |
797 | static inline void mem_cgroup_put(struct mem_cgroup *memcg) |
798 | { | |
d46eb14b SB |
799 | if (memcg) |
800 | css_put(&memcg->css); | |
dc0b5864 RG |
801 | } |
802 | ||
8e8ae645 JW |
803 | #define mem_cgroup_from_counter(counter, member) \ |
804 | container_of(counter, struct mem_cgroup, member) | |
805 | ||
33398cf2 MH |
806 | struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, |
807 | struct mem_cgroup *, | |
808 | struct mem_cgroup_reclaim_cookie *); | |
809 | void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *); | |
7c5f64f8 VD |
810 | int mem_cgroup_scan_tasks(struct mem_cgroup *, |
811 | int (*)(struct task_struct *, void *), void *); | |
33398cf2 | 812 | |
23047a96 JW |
813 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
814 | { | |
815 | if (mem_cgroup_disabled()) | |
816 | return 0; | |
817 | ||
73f576c0 | 818 | return memcg->id.id; |
23047a96 | 819 | } |
73f576c0 | 820 | struct mem_cgroup *mem_cgroup_from_id(unsigned short id); |
23047a96 | 821 | |
c15187a4 RG |
822 | #ifdef CONFIG_SHRINKER_DEBUG |
823 | static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) | |
824 | { | |
825 | return memcg ? cgroup_ino(memcg->css.cgroup) : 0; | |
826 | } | |
827 | ||
828 | struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino); | |
829 | #endif | |
830 | ||
aa9694bb CD |
831 | static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) |
832 | { | |
833 | return mem_cgroup_from_css(seq_css(m)); | |
834 | } | |
835 | ||
2262185c RG |
836 | static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) |
837 | { | |
838 | struct mem_cgroup_per_node *mz; | |
839 | ||
840 | if (mem_cgroup_disabled()) | |
841 | return NULL; | |
842 | ||
843 | mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); | |
844 | return mz->memcg; | |
845 | } | |
846 | ||
8e8ae645 JW |
847 | /** |
848 | * parent_mem_cgroup - find the accounting parent of a memcg | |
849 | * @memcg: memcg whose parent to find | |
850 | * | |
851 | * Returns the parent memcg, or NULL if this is the root or the memory | |
852 | * controller is in legacy no-hierarchy mode. | |
853 | */ | |
854 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) | |
855 | { | |
486bc706 | 856 | return mem_cgroup_from_css(memcg->css.parent); |
8e8ae645 JW |
857 | } |
858 | ||
33398cf2 MH |
859 | static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, |
860 | struct mem_cgroup *root) | |
861 | { | |
862 | if (root == memcg) | |
863 | return true; | |
33398cf2 MH |
864 | return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup); |
865 | } | |
e1aab161 | 866 | |
2314b42d JW |
867 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
868 | struct mem_cgroup *memcg) | |
2e4d4091 | 869 | { |
587af308 | 870 | struct mem_cgroup *task_memcg; |
413918bb | 871 | bool match = false; |
c3ac9a8a | 872 | |
2e4d4091 | 873 | rcu_read_lock(); |
587af308 | 874 | task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); |
413918bb | 875 | if (task_memcg) |
2314b42d | 876 | match = mem_cgroup_is_descendant(task_memcg, memcg); |
2e4d4091 | 877 | rcu_read_unlock(); |
c3ac9a8a | 878 | return match; |
2e4d4091 | 879 | } |
8a9f3ccd | 880 | |
64219994 | 881 | struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); |
2fc04524 | 882 | ino_t page_cgroup_ino(struct page *page); |
d324236b | 883 | |
eb01aaab VD |
884 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
885 | { | |
886 | if (mem_cgroup_disabled()) | |
887 | return true; | |
888 | return !!(memcg->css.flags & CSS_ONLINE); | |
889 | } | |
890 | ||
33398cf2 | 891 | void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru, |
b4536f0c | 892 | int zid, int nr_pages); |
33398cf2 | 893 | |
b4536f0c MH |
894 | static inline |
895 | unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, | |
896 | enum lru_list lru, int zone_idx) | |
897 | { | |
898 | struct mem_cgroup_per_node *mz; | |
899 | ||
900 | mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec); | |
e0e3f42f | 901 | return READ_ONCE(mz->lru_zone_size[zone_idx][lru]); |
33398cf2 MH |
902 | } |
903 | ||
b23afb93 TH |
904 | void mem_cgroup_handle_over_high(void); |
905 | ||
bbec2e15 | 906 | unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg); |
7c5f64f8 | 907 | |
9783aa99 CD |
908 | unsigned long mem_cgroup_size(struct mem_cgroup *memcg); |
909 | ||
f0c867d9 | 910 | void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, |
64219994 | 911 | struct task_struct *p); |
58ae83db | 912 | |
f0c867d9 | 913 | void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg); |
914 | ||
29ef680a | 915 | static inline void mem_cgroup_enter_user_fault(void) |
519e5247 | 916 | { |
29ef680a MH |
917 | WARN_ON(current->in_user_fault); |
918 | current->in_user_fault = 1; | |
519e5247 JW |
919 | } |
920 | ||
29ef680a | 921 | static inline void mem_cgroup_exit_user_fault(void) |
519e5247 | 922 | { |
29ef680a MH |
923 | WARN_ON(!current->in_user_fault); |
924 | current->in_user_fault = 0; | |
519e5247 JW |
925 | } |
926 | ||
3812c8c8 JW |
927 | static inline bool task_in_memcg_oom(struct task_struct *p) |
928 | { | |
626ebc41 | 929 | return p->memcg_in_oom; |
3812c8c8 JW |
930 | } |
931 | ||
49426420 | 932 | bool mem_cgroup_oom_synchronize(bool wait); |
3d8b38eb RG |
933 | struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim, |
934 | struct mem_cgroup *oom_domain); | |
935 | void mem_cgroup_print_oom_group(struct mem_cgroup *memcg); | |
3812c8c8 | 936 | |
f70ad448 MWO |
937 | void folio_memcg_lock(struct folio *folio); |
938 | void folio_memcg_unlock(struct folio *folio); | |
1c824a68 | 939 | void lock_page_memcg(struct page *page); |
62cccb8c | 940 | void unlock_page_memcg(struct page *page); |
d7365e78 | 941 | |
db9adbcb | 942 | void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val); |
2a2e4885 | 943 | |
018ee47f YZ |
944 | /* try to stablize folio_memcg() for all the pages in a memcg */ |
945 | static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg) | |
946 | { | |
947 | rcu_read_lock(); | |
948 | ||
949 | if (mem_cgroup_disabled() || !atomic_read(&memcg->moving_account)) | |
950 | return true; | |
951 | ||
952 | rcu_read_unlock(); | |
953 | return false; | |
954 | } | |
955 | ||
956 | static inline void mem_cgroup_unlock_pages(void) | |
957 | { | |
958 | rcu_read_unlock(); | |
959 | } | |
960 | ||
04fecbf5 | 961 | /* idx can be of type enum memcg_stat_item or node_stat_item */ |
00f3ca2c | 962 | static inline void mod_memcg_state(struct mem_cgroup *memcg, |
04fecbf5 | 963 | int idx, int val) |
2a2e4885 | 964 | { |
c3cc3911 JW |
965 | unsigned long flags; |
966 | ||
967 | local_irq_save(flags); | |
a983b5eb | 968 | __mod_memcg_state(memcg, idx, val); |
c3cc3911 | 969 | local_irq_restore(flags); |
2a2e4885 JW |
970 | } |
971 | ||
4e5aa1f4 SB |
972 | static inline void mod_memcg_page_state(struct page *page, |
973 | int idx, int val) | |
974 | { | |
975 | struct mem_cgroup *memcg; | |
976 | ||
977 | if (mem_cgroup_disabled()) | |
978 | return; | |
979 | ||
980 | rcu_read_lock(); | |
981 | memcg = page_memcg(page); | |
982 | if (memcg) | |
983 | mod_memcg_state(memcg, idx, val); | |
984 | rcu_read_unlock(); | |
985 | } | |
986 | ||
410f8e82 | 987 | unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx); |
7490a2d2 | 988 | |
42a30035 JW |
989 | static inline unsigned long lruvec_page_state(struct lruvec *lruvec, |
990 | enum node_stat_item idx) | |
991 | { | |
992 | struct mem_cgroup_per_node *pn; | |
dbb16df6 | 993 | long x; |
42a30035 JW |
994 | |
995 | if (mem_cgroup_disabled()) | |
996 | return node_page_state(lruvec_pgdat(lruvec), idx); | |
997 | ||
998 | pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); | |
dbb16df6 SB |
999 | x = READ_ONCE(pn->lruvec_stats.state[idx]); |
1000 | #ifdef CONFIG_SMP | |
1001 | if (x < 0) | |
1002 | x = 0; | |
1003 | #endif | |
1004 | return x; | |
42a30035 JW |
1005 | } |
1006 | ||
205b20cc JW |
1007 | static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, |
1008 | enum node_stat_item idx) | |
2a7106f2 | 1009 | { |
00f3ca2c | 1010 | struct mem_cgroup_per_node *pn; |
815744d7 JW |
1011 | long x = 0; |
1012 | int cpu; | |
00f3ca2c JW |
1013 | |
1014 | if (mem_cgroup_disabled()) | |
1015 | return node_page_state(lruvec_pgdat(lruvec), idx); | |
1016 | ||
1017 | pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec); | |
815744d7 | 1018 | for_each_possible_cpu(cpu) |
7e1c0d6f | 1019 | x += per_cpu(pn->lruvec_stats_percpu->state[idx], cpu); |
a983b5eb JW |
1020 | #ifdef CONFIG_SMP |
1021 | if (x < 0) | |
1022 | x = 0; | |
1023 | #endif | |
1024 | return x; | |
2a7106f2 GT |
1025 | } |
1026 | ||
aa48e47e | 1027 | void mem_cgroup_flush_stats(void); |
9b301615 | 1028 | void mem_cgroup_flush_stats_delayed(void); |
aa48e47e | 1029 | |
eedc4e5a RG |
1030 | void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx, |
1031 | int val); | |
da3ceeff | 1032 | void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val); |
991e7673 | 1033 | |
da3ceeff | 1034 | static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, |
991e7673 SB |
1035 | int val) |
1036 | { | |
1037 | unsigned long flags; | |
1038 | ||
1039 | local_irq_save(flags); | |
da3ceeff | 1040 | __mod_lruvec_kmem_state(p, idx, val); |
991e7673 SB |
1041 | local_irq_restore(flags); |
1042 | } | |
1043 | ||
eedc4e5a RG |
1044 | static inline void mod_memcg_lruvec_state(struct lruvec *lruvec, |
1045 | enum node_stat_item idx, int val) | |
1046 | { | |
1047 | unsigned long flags; | |
1048 | ||
1049 | local_irq_save(flags); | |
1050 | __mod_memcg_lruvec_state(lruvec, idx, val); | |
1051 | local_irq_restore(flags); | |
1052 | } | |
1053 | ||
db9adbcb JW |
1054 | void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx, |
1055 | unsigned long count); | |
c9019e9b | 1056 | |
2262185c | 1057 | static inline void count_memcg_events(struct mem_cgroup *memcg, |
e27be240 JW |
1058 | enum vm_event_item idx, |
1059 | unsigned long count) | |
2262185c | 1060 | { |
c3cc3911 JW |
1061 | unsigned long flags; |
1062 | ||
1063 | local_irq_save(flags); | |
a983b5eb | 1064 | __count_memcg_events(memcg, idx, count); |
c3cc3911 | 1065 | local_irq_restore(flags); |
2262185c RG |
1066 | } |
1067 | ||
1068 | static inline void count_memcg_page_event(struct page *page, | |
e27be240 | 1069 | enum vm_event_item idx) |
2262185c | 1070 | { |
bcfe06bf RG |
1071 | struct mem_cgroup *memcg = page_memcg(page); |
1072 | ||
1073 | if (memcg) | |
1074 | count_memcg_events(memcg, idx, 1); | |
2262185c RG |
1075 | } |
1076 | ||
64daa5d8 MWO |
1077 | static inline void count_memcg_folio_events(struct folio *folio, |
1078 | enum vm_event_item idx, unsigned long nr) | |
1079 | { | |
1080 | struct mem_cgroup *memcg = folio_memcg(folio); | |
1081 | ||
1082 | if (memcg) | |
1083 | count_memcg_events(memcg, idx, nr); | |
1084 | } | |
1085 | ||
2262185c RG |
1086 | static inline void count_memcg_event_mm(struct mm_struct *mm, |
1087 | enum vm_event_item idx) | |
68ae564b | 1088 | { |
33398cf2 MH |
1089 | struct mem_cgroup *memcg; |
1090 | ||
68ae564b DR |
1091 | if (mem_cgroup_disabled()) |
1092 | return; | |
33398cf2 MH |
1093 | |
1094 | rcu_read_lock(); | |
1095 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
fe6bdfc8 | 1096 | if (likely(memcg)) |
c9019e9b | 1097 | count_memcg_events(memcg, idx, 1); |
33398cf2 | 1098 | rcu_read_unlock(); |
68ae564b | 1099 | } |
c9019e9b | 1100 | |
e27be240 JW |
1101 | static inline void memcg_memory_event(struct mem_cgroup *memcg, |
1102 | enum memcg_memory_event event) | |
c9019e9b | 1103 | { |
8b21ca02 MS |
1104 | bool swap_event = event == MEMCG_SWAP_HIGH || event == MEMCG_SWAP_MAX || |
1105 | event == MEMCG_SWAP_FAIL; | |
1106 | ||
1e577f97 | 1107 | atomic_long_inc(&memcg->memory_events_local[event]); |
8b21ca02 MS |
1108 | if (!swap_event) |
1109 | cgroup_file_notify(&memcg->events_local_file); | |
1e577f97 | 1110 | |
9852ae3f CD |
1111 | do { |
1112 | atomic_long_inc(&memcg->memory_events[event]); | |
8b21ca02 MS |
1113 | if (swap_event) |
1114 | cgroup_file_notify(&memcg->swap_events_file); | |
1115 | else | |
1116 | cgroup_file_notify(&memcg->events_file); | |
9852ae3f | 1117 | |
04fd61a4 YS |
1118 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) |
1119 | break; | |
9852ae3f CD |
1120 | if (cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_LOCAL_EVENTS) |
1121 | break; | |
1122 | } while ((memcg = parent_mem_cgroup(memcg)) && | |
1123 | !mem_cgroup_is_root(memcg)); | |
c9019e9b JW |
1124 | } |
1125 | ||
fe6bdfc8 RG |
1126 | static inline void memcg_memory_event_mm(struct mm_struct *mm, |
1127 | enum memcg_memory_event event) | |
1128 | { | |
1129 | struct mem_cgroup *memcg; | |
1130 | ||
1131 | if (mem_cgroup_disabled()) | |
1132 | return; | |
1133 | ||
1134 | rcu_read_lock(); | |
1135 | memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); | |
1136 | if (likely(memcg)) | |
1137 | memcg_memory_event(memcg, event); | |
1138 | rcu_read_unlock(); | |
1139 | } | |
1140 | ||
be6c8982 | 1141 | void split_page_memcg(struct page *head, unsigned int nr); |
ca3e0214 | 1142 | |
2d146aa3 JW |
1143 | unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, |
1144 | gfp_t gfp_mask, | |
1145 | unsigned long *total_scanned); | |
1146 | ||
c255a458 | 1147 | #else /* CONFIG_MEMCG */ |
23047a96 JW |
1148 | |
1149 | #define MEM_CGROUP_ID_SHIFT 0 | |
1150 | #define MEM_CGROUP_ID_MAX 0 | |
1151 | ||
1b7e4464 MWO |
1152 | static inline struct mem_cgroup *folio_memcg(struct folio *folio) |
1153 | { | |
1154 | return NULL; | |
1155 | } | |
1156 | ||
bcfe06bf RG |
1157 | static inline struct mem_cgroup *page_memcg(struct page *page) |
1158 | { | |
1159 | return NULL; | |
1160 | } | |
1161 | ||
c5ce619a | 1162 | static inline struct mem_cgroup *folio_memcg_rcu(struct folio *folio) |
bcfe06bf RG |
1163 | { |
1164 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
1165 | return NULL; | |
1166 | } | |
1167 | ||
1168 | static inline struct mem_cgroup *page_memcg_check(struct page *page) | |
1169 | { | |
1170 | return NULL; | |
1171 | } | |
1172 | ||
1b7e4464 MWO |
1173 | static inline bool folio_memcg_kmem(struct folio *folio) |
1174 | { | |
1175 | return false; | |
1176 | } | |
1177 | ||
18b2db3b RG |
1178 | static inline bool PageMemcgKmem(struct page *page) |
1179 | { | |
1180 | return false; | |
1181 | } | |
1182 | ||
dfd2f10c KT |
1183 | static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg) |
1184 | { | |
1185 | return true; | |
1186 | } | |
1187 | ||
23047a96 JW |
1188 | static inline bool mem_cgroup_disabled(void) |
1189 | { | |
1190 | return true; | |
1191 | } | |
1192 | ||
e27be240 JW |
1193 | static inline void memcg_memory_event(struct mem_cgroup *memcg, |
1194 | enum memcg_memory_event event) | |
241994ed JW |
1195 | { |
1196 | } | |
1197 | ||
fe6bdfc8 RG |
1198 | static inline void memcg_memory_event_mm(struct mm_struct *mm, |
1199 | enum memcg_memory_event event) | |
1200 | { | |
1201 | } | |
1202 | ||
f56ce412 JW |
1203 | static inline void mem_cgroup_protection(struct mem_cgroup *root, |
1204 | struct mem_cgroup *memcg, | |
1205 | unsigned long *min, | |
1206 | unsigned long *low) | |
9783aa99 | 1207 | { |
f56ce412 | 1208 | *min = *low = 0; |
9783aa99 CD |
1209 | } |
1210 | ||
45c7f7e1 CD |
1211 | static inline void mem_cgroup_calculate_protection(struct mem_cgroup *root, |
1212 | struct mem_cgroup *memcg) | |
1213 | { | |
1214 | } | |
1215 | ||
adb82130 YA |
1216 | static inline bool mem_cgroup_unprotected(struct mem_cgroup *target, |
1217 | struct mem_cgroup *memcg) | |
1218 | { | |
1219 | return true; | |
1220 | } | |
1221 | static inline bool mem_cgroup_below_low(struct mem_cgroup *target, | |
1222 | struct mem_cgroup *memcg) | |
45c7f7e1 CD |
1223 | { |
1224 | return false; | |
1225 | } | |
1226 | ||
adb82130 YA |
1227 | static inline bool mem_cgroup_below_min(struct mem_cgroup *target, |
1228 | struct mem_cgroup *memcg) | |
241994ed | 1229 | { |
45c7f7e1 | 1230 | return false; |
241994ed JW |
1231 | } |
1232 | ||
8f425e4e MWO |
1233 | static inline int mem_cgroup_charge(struct folio *folio, |
1234 | struct mm_struct *mm, gfp_t gfp) | |
3fea5a49 JW |
1235 | { |
1236 | return 0; | |
1237 | } | |
1238 | ||
65995918 | 1239 | static inline int mem_cgroup_swapin_charge_folio(struct folio *folio, |
0add0c77 SB |
1240 | struct mm_struct *mm, gfp_t gfp, swp_entry_t entry) |
1241 | { | |
1242 | return 0; | |
1243 | } | |
1244 | ||
1245 | static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry) | |
1246 | { | |
1247 | } | |
1248 | ||
bbc6b703 | 1249 | static inline void mem_cgroup_uncharge(struct folio *folio) |
569b846d KH |
1250 | { |
1251 | } | |
1252 | ||
747db954 | 1253 | static inline void mem_cgroup_uncharge_list(struct list_head *page_list) |
8a9f3ccd BS |
1254 | { |
1255 | } | |
1256 | ||
d21bba2b | 1257 | static inline void mem_cgroup_migrate(struct folio *old, struct folio *new) |
69029cd5 KH |
1258 | { |
1259 | } | |
1260 | ||
867e5e1d JW |
1261 | static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg, |
1262 | struct pglist_data *pgdat) | |
08e552c6 | 1263 | { |
867e5e1d | 1264 | return &pgdat->__lruvec; |
08e552c6 KH |
1265 | } |
1266 | ||
b1baabd9 | 1267 | static inline struct lruvec *folio_lruvec(struct folio *folio) |
66e1707b | 1268 | { |
b1baabd9 | 1269 | struct pglist_data *pgdat = folio_pgdat(folio); |
867e5e1d | 1270 | return &pgdat->__lruvec; |
66e1707b BS |
1271 | } |
1272 | ||
e809c3fe MWO |
1273 | static inline |
1274 | void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio) | |
2d146aa3 JW |
1275 | { |
1276 | } | |
1277 | ||
b910718a JW |
1278 | static inline struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg) |
1279 | { | |
1280 | return NULL; | |
1281 | } | |
1282 | ||
587af308 | 1283 | static inline bool mm_match_cgroup(struct mm_struct *mm, |
c0ff4b85 | 1284 | struct mem_cgroup *memcg) |
bed7161a | 1285 | { |
587af308 | 1286 | return true; |
bed7161a BS |
1287 | } |
1288 | ||
d46eb14b SB |
1289 | static inline struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm) |
1290 | { | |
1291 | return NULL; | |
1292 | } | |
1293 | ||
c74d40e8 DS |
1294 | static inline |
1295 | struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css) | |
1296 | { | |
1297 | return NULL; | |
1298 | } | |
1299 | ||
f4840ccf JW |
1300 | static inline void obj_cgroup_put(struct obj_cgroup *objcg) |
1301 | { | |
1302 | } | |
1303 | ||
dc0b5864 RG |
1304 | static inline void mem_cgroup_put(struct mem_cgroup *memcg) |
1305 | { | |
1306 | } | |
1307 | ||
e809c3fe | 1308 | static inline struct lruvec *folio_lruvec_lock(struct folio *folio) |
6168d0da | 1309 | { |
e809c3fe | 1310 | struct pglist_data *pgdat = folio_pgdat(folio); |
6168d0da AS |
1311 | |
1312 | spin_lock(&pgdat->__lruvec.lru_lock); | |
1313 | return &pgdat->__lruvec; | |
1314 | } | |
1315 | ||
e809c3fe | 1316 | static inline struct lruvec *folio_lruvec_lock_irq(struct folio *folio) |
6168d0da | 1317 | { |
e809c3fe | 1318 | struct pglist_data *pgdat = folio_pgdat(folio); |
6168d0da AS |
1319 | |
1320 | spin_lock_irq(&pgdat->__lruvec.lru_lock); | |
1321 | return &pgdat->__lruvec; | |
1322 | } | |
1323 | ||
e809c3fe | 1324 | static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio, |
6168d0da AS |
1325 | unsigned long *flagsp) |
1326 | { | |
e809c3fe | 1327 | struct pglist_data *pgdat = folio_pgdat(folio); |
6168d0da AS |
1328 | |
1329 | spin_lock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp); | |
1330 | return &pgdat->__lruvec; | |
1331 | } | |
1332 | ||
5660048c JW |
1333 | static inline struct mem_cgroup * |
1334 | mem_cgroup_iter(struct mem_cgroup *root, | |
1335 | struct mem_cgroup *prev, | |
1336 | struct mem_cgroup_reclaim_cookie *reclaim) | |
1337 | { | |
1338 | return NULL; | |
1339 | } | |
1340 | ||
1341 | static inline void mem_cgroup_iter_break(struct mem_cgroup *root, | |
1342 | struct mem_cgroup *prev) | |
1343 | { | |
1344 | } | |
1345 | ||
7c5f64f8 VD |
1346 | static inline int mem_cgroup_scan_tasks(struct mem_cgroup *memcg, |
1347 | int (*fn)(struct task_struct *, void *), void *arg) | |
1348 | { | |
1349 | return 0; | |
1350 | } | |
1351 | ||
23047a96 | 1352 | static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg) |
f8d66542 | 1353 | { |
23047a96 JW |
1354 | return 0; |
1355 | } | |
1356 | ||
1357 | static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id) | |
1358 | { | |
1359 | WARN_ON_ONCE(id); | |
1360 | /* XXX: This should always return root_mem_cgroup */ | |
1361 | return NULL; | |
f8d66542 | 1362 | } |
a636b327 | 1363 | |
c15187a4 RG |
1364 | #ifdef CONFIG_SHRINKER_DEBUG |
1365 | static inline unsigned long mem_cgroup_ino(struct mem_cgroup *memcg) | |
1366 | { | |
1367 | return 0; | |
1368 | } | |
1369 | ||
1370 | static inline struct mem_cgroup *mem_cgroup_get_from_ino(unsigned long ino) | |
1371 | { | |
1372 | return NULL; | |
1373 | } | |
1374 | #endif | |
1375 | ||
aa9694bb CD |
1376 | static inline struct mem_cgroup *mem_cgroup_from_seq(struct seq_file *m) |
1377 | { | |
1378 | return NULL; | |
1379 | } | |
1380 | ||
2262185c RG |
1381 | static inline struct mem_cgroup *lruvec_memcg(struct lruvec *lruvec) |
1382 | { | |
1383 | return NULL; | |
1384 | } | |
1385 | ||
eb01aaab | 1386 | static inline bool mem_cgroup_online(struct mem_cgroup *memcg) |
14797e23 | 1387 | { |
13308ca9 | 1388 | return true; |
14797e23 KM |
1389 | } |
1390 | ||
b4536f0c MH |
1391 | static inline |
1392 | unsigned long mem_cgroup_get_zone_lru_size(struct lruvec *lruvec, | |
1393 | enum lru_list lru, int zone_idx) | |
1394 | { | |
1395 | return 0; | |
1396 | } | |
a3d8e054 | 1397 | |
bbec2e15 | 1398 | static inline unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg) |
7c5f64f8 VD |
1399 | { |
1400 | return 0; | |
1401 | } | |
1402 | ||
9783aa99 CD |
1403 | static inline unsigned long mem_cgroup_size(struct mem_cgroup *memcg) |
1404 | { | |
1405 | return 0; | |
1406 | } | |
1407 | ||
e222432b | 1408 | static inline void |
f0c867d9 | 1409 | mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p) |
1410 | { | |
1411 | } | |
1412 | ||
1413 | static inline void | |
1414 | mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg) | |
e222432b BS |
1415 | { |
1416 | } | |
1417 | ||
1c824a68 | 1418 | static inline void lock_page_memcg(struct page *page) |
89c06bd5 KH |
1419 | { |
1420 | } | |
1421 | ||
62cccb8c | 1422 | static inline void unlock_page_memcg(struct page *page) |
89c06bd5 KH |
1423 | { |
1424 | } | |
1425 | ||
f70ad448 MWO |
1426 | static inline void folio_memcg_lock(struct folio *folio) |
1427 | { | |
1428 | } | |
1429 | ||
1430 | static inline void folio_memcg_unlock(struct folio *folio) | |
1431 | { | |
1432 | } | |
1433 | ||
018ee47f YZ |
1434 | static inline bool mem_cgroup_trylock_pages(struct mem_cgroup *memcg) |
1435 | { | |
1436 | /* to match folio_memcg_rcu() */ | |
1437 | rcu_read_lock(); | |
1438 | return true; | |
1439 | } | |
1440 | ||
1441 | static inline void mem_cgroup_unlock_pages(void) | |
1442 | { | |
1443 | rcu_read_unlock(); | |
1444 | } | |
1445 | ||
b23afb93 TH |
1446 | static inline void mem_cgroup_handle_over_high(void) |
1447 | { | |
1448 | } | |
1449 | ||
29ef680a | 1450 | static inline void mem_cgroup_enter_user_fault(void) |
519e5247 JW |
1451 | { |
1452 | } | |
1453 | ||
29ef680a | 1454 | static inline void mem_cgroup_exit_user_fault(void) |
519e5247 JW |
1455 | { |
1456 | } | |
1457 | ||
3812c8c8 JW |
1458 | static inline bool task_in_memcg_oom(struct task_struct *p) |
1459 | { | |
1460 | return false; | |
1461 | } | |
1462 | ||
49426420 | 1463 | static inline bool mem_cgroup_oom_synchronize(bool wait) |
3812c8c8 JW |
1464 | { |
1465 | return false; | |
1466 | } | |
1467 | ||
3d8b38eb RG |
1468 | static inline struct mem_cgroup *mem_cgroup_get_oom_group( |
1469 | struct task_struct *victim, struct mem_cgroup *oom_domain) | |
1470 | { | |
1471 | return NULL; | |
1472 | } | |
1473 | ||
1474 | static inline void mem_cgroup_print_oom_group(struct mem_cgroup *memcg) | |
1475 | { | |
1476 | } | |
1477 | ||
00f3ca2c | 1478 | static inline void __mod_memcg_state(struct mem_cgroup *memcg, |
04fecbf5 | 1479 | int idx, |
00f3ca2c | 1480 | int nr) |
2a2e4885 JW |
1481 | { |
1482 | } | |
1483 | ||
00f3ca2c | 1484 | static inline void mod_memcg_state(struct mem_cgroup *memcg, |
04fecbf5 | 1485 | int idx, |
00f3ca2c | 1486 | int nr) |
2a2e4885 JW |
1487 | { |
1488 | } | |
1489 | ||
4e5aa1f4 SB |
1490 | static inline void mod_memcg_page_state(struct page *page, |
1491 | int idx, int val) | |
1492 | { | |
1493 | } | |
1494 | ||
7490a2d2 SB |
1495 | static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx) |
1496 | { | |
1497 | return 0; | |
1498 | } | |
1499 | ||
42a30035 JW |
1500 | static inline unsigned long lruvec_page_state(struct lruvec *lruvec, |
1501 | enum node_stat_item idx) | |
1502 | { | |
1503 | return node_page_state(lruvec_pgdat(lruvec), idx); | |
1504 | } | |
1505 | ||
205b20cc JW |
1506 | static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec, |
1507 | enum node_stat_item idx) | |
2a7106f2 | 1508 | { |
00f3ca2c | 1509 | return node_page_state(lruvec_pgdat(lruvec), idx); |
2a7106f2 GT |
1510 | } |
1511 | ||
aa48e47e SB |
1512 | static inline void mem_cgroup_flush_stats(void) |
1513 | { | |
1514 | } | |
1515 | ||
9b301615 SB |
1516 | static inline void mem_cgroup_flush_stats_delayed(void) |
1517 | { | |
1518 | } | |
1519 | ||
eedc4e5a RG |
1520 | static inline void __mod_memcg_lruvec_state(struct lruvec *lruvec, |
1521 | enum node_stat_item idx, int val) | |
1522 | { | |
1523 | } | |
1524 | ||
da3ceeff | 1525 | static inline void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, |
ec9f0238 RG |
1526 | int val) |
1527 | { | |
1528 | struct page *page = virt_to_head_page(p); | |
1529 | ||
1530 | __mod_node_page_state(page_pgdat(page), idx, val); | |
1531 | } | |
1532 | ||
da3ceeff | 1533 | static inline void mod_lruvec_kmem_state(void *p, enum node_stat_item idx, |
991e7673 SB |
1534 | int val) |
1535 | { | |
1536 | struct page *page = virt_to_head_page(p); | |
1537 | ||
1538 | mod_node_page_state(page_pgdat(page), idx, val); | |
1539 | } | |
1540 | ||
2262185c RG |
1541 | static inline void count_memcg_events(struct mem_cgroup *memcg, |
1542 | enum vm_event_item idx, | |
1543 | unsigned long count) | |
1544 | { | |
1545 | } | |
1546 | ||
9851ac13 KT |
1547 | static inline void __count_memcg_events(struct mem_cgroup *memcg, |
1548 | enum vm_event_item idx, | |
1549 | unsigned long count) | |
1550 | { | |
1551 | } | |
1552 | ||
2262185c | 1553 | static inline void count_memcg_page_event(struct page *page, |
04fecbf5 | 1554 | int idx) |
2262185c RG |
1555 | { |
1556 | } | |
1557 | ||
64daa5d8 MWO |
1558 | static inline void count_memcg_folio_events(struct folio *folio, |
1559 | enum vm_event_item idx, unsigned long nr) | |
1560 | { | |
1561 | } | |
1562 | ||
456f998e | 1563 | static inline |
2262185c | 1564 | void count_memcg_event_mm(struct mm_struct *mm, enum vm_event_item idx) |
456f998e YH |
1565 | { |
1566 | } | |
6168d0da | 1567 | |
2d146aa3 JW |
1568 | static inline void split_page_memcg(struct page *head, unsigned int nr) |
1569 | { | |
1570 | } | |
1571 | ||
1572 | static inline | |
1573 | unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order, | |
1574 | gfp_t gfp_mask, | |
1575 | unsigned long *total_scanned) | |
6168d0da | 1576 | { |
2d146aa3 | 1577 | return 0; |
6168d0da | 1578 | } |
c255a458 | 1579 | #endif /* CONFIG_MEMCG */ |
78fb7466 | 1580 | |
da3ceeff | 1581 | static inline void __inc_lruvec_kmem_state(void *p, enum node_stat_item idx) |
ec9f0238 | 1582 | { |
da3ceeff | 1583 | __mod_lruvec_kmem_state(p, idx, 1); |
ec9f0238 RG |
1584 | } |
1585 | ||
da3ceeff | 1586 | static inline void __dec_lruvec_kmem_state(void *p, enum node_stat_item idx) |
ec9f0238 | 1587 | { |
da3ceeff | 1588 | __mod_lruvec_kmem_state(p, idx, -1); |
ec9f0238 RG |
1589 | } |
1590 | ||
7cf111bc JW |
1591 | static inline struct lruvec *parent_lruvec(struct lruvec *lruvec) |
1592 | { | |
1593 | struct mem_cgroup *memcg; | |
1594 | ||
1595 | memcg = lruvec_memcg(lruvec); | |
1596 | if (!memcg) | |
1597 | return NULL; | |
1598 | memcg = parent_mem_cgroup(memcg); | |
1599 | if (!memcg) | |
1600 | return NULL; | |
1601 | return mem_cgroup_lruvec(memcg, lruvec_pgdat(lruvec)); | |
1602 | } | |
1603 | ||
6168d0da AS |
1604 | static inline void unlock_page_lruvec(struct lruvec *lruvec) |
1605 | { | |
1606 | spin_unlock(&lruvec->lru_lock); | |
1607 | } | |
1608 | ||
1609 | static inline void unlock_page_lruvec_irq(struct lruvec *lruvec) | |
1610 | { | |
1611 | spin_unlock_irq(&lruvec->lru_lock); | |
1612 | } | |
1613 | ||
1614 | static inline void unlock_page_lruvec_irqrestore(struct lruvec *lruvec, | |
1615 | unsigned long flags) | |
1616 | { | |
1617 | spin_unlock_irqrestore(&lruvec->lru_lock, flags); | |
1618 | } | |
1619 | ||
7467c391 | 1620 | /* Test requires a stable page->memcg binding, see page_memcg() */ |
0de340cb MWO |
1621 | static inline bool folio_matches_lruvec(struct folio *folio, |
1622 | struct lruvec *lruvec) | |
f2e4d28d | 1623 | { |
0de340cb MWO |
1624 | return lruvec_pgdat(lruvec) == folio_pgdat(folio) && |
1625 | lruvec_memcg(lruvec) == folio_memcg(folio); | |
f2e4d28d MS |
1626 | } |
1627 | ||
2a5e4e34 | 1628 | /* Don't lock again iff page's lruvec locked */ |
0de340cb | 1629 | static inline struct lruvec *folio_lruvec_relock_irq(struct folio *folio, |
2a5e4e34 AD |
1630 | struct lruvec *locked_lruvec) |
1631 | { | |
1632 | if (locked_lruvec) { | |
0de340cb | 1633 | if (folio_matches_lruvec(folio, locked_lruvec)) |
2a5e4e34 AD |
1634 | return locked_lruvec; |
1635 | ||
1636 | unlock_page_lruvec_irq(locked_lruvec); | |
1637 | } | |
1638 | ||
e809c3fe | 1639 | return folio_lruvec_lock_irq(folio); |
2a5e4e34 AD |
1640 | } |
1641 | ||
1642 | /* Don't lock again iff page's lruvec locked */ | |
0de340cb | 1643 | static inline struct lruvec *folio_lruvec_relock_irqsave(struct folio *folio, |
2a5e4e34 AD |
1644 | struct lruvec *locked_lruvec, unsigned long *flags) |
1645 | { | |
1646 | if (locked_lruvec) { | |
0de340cb | 1647 | if (folio_matches_lruvec(folio, locked_lruvec)) |
2a5e4e34 AD |
1648 | return locked_lruvec; |
1649 | ||
1650 | unlock_page_lruvec_irqrestore(locked_lruvec, *flags); | |
1651 | } | |
1652 | ||
e809c3fe | 1653 | return folio_lruvec_lock_irqsave(folio, flags); |
2a5e4e34 AD |
1654 | } |
1655 | ||
52ebea74 | 1656 | #ifdef CONFIG_CGROUP_WRITEBACK |
841710aa | 1657 | |
841710aa | 1658 | struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); |
c5edf9cd TH |
1659 | void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages, |
1660 | unsigned long *pheadroom, unsigned long *pdirty, | |
1661 | unsigned long *pwriteback); | |
841710aa | 1662 | |
9d8053fc | 1663 | void mem_cgroup_track_foreign_dirty_slowpath(struct folio *folio, |
97b27821 TH |
1664 | struct bdi_writeback *wb); |
1665 | ||
203a3151 | 1666 | static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, |
97b27821 TH |
1667 | struct bdi_writeback *wb) |
1668 | { | |
ac86f547 KW |
1669 | struct mem_cgroup *memcg; |
1670 | ||
08d1d0e6 BH |
1671 | if (mem_cgroup_disabled()) |
1672 | return; | |
1673 | ||
ac86f547 KW |
1674 | memcg = folio_memcg(folio); |
1675 | if (unlikely(memcg && &memcg->css != wb->memcg_css)) | |
9d8053fc | 1676 | mem_cgroup_track_foreign_dirty_slowpath(folio, wb); |
97b27821 TH |
1677 | } |
1678 | ||
1679 | void mem_cgroup_flush_foreign(struct bdi_writeback *wb); | |
1680 | ||
841710aa TH |
1681 | #else /* CONFIG_CGROUP_WRITEBACK */ |
1682 | ||
1683 | static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) | |
1684 | { | |
1685 | return NULL; | |
1686 | } | |
1687 | ||
c2aa723a | 1688 | static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, |
c5edf9cd TH |
1689 | unsigned long *pfilepages, |
1690 | unsigned long *pheadroom, | |
c2aa723a TH |
1691 | unsigned long *pdirty, |
1692 | unsigned long *pwriteback) | |
1693 | { | |
1694 | } | |
1695 | ||
203a3151 | 1696 | static inline void mem_cgroup_track_foreign_dirty(struct folio *folio, |
97b27821 TH |
1697 | struct bdi_writeback *wb) |
1698 | { | |
1699 | } | |
1700 | ||
1701 | static inline void mem_cgroup_flush_foreign(struct bdi_writeback *wb) | |
1702 | { | |
1703 | } | |
1704 | ||
841710aa | 1705 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
52ebea74 | 1706 | |
e1aab161 | 1707 | struct sock; |
4b1327be WW |
1708 | bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages, |
1709 | gfp_t gfp_mask); | |
baac50bb | 1710 | void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages); |
d886f4e4 | 1711 | #ifdef CONFIG_MEMCG |
ef12947c JW |
1712 | extern struct static_key_false memcg_sockets_enabled_key; |
1713 | #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key) | |
2d758073 JW |
1714 | void mem_cgroup_sk_alloc(struct sock *sk); |
1715 | void mem_cgroup_sk_free(struct sock *sk); | |
baac50bb | 1716 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c | 1717 | { |
0db15298 | 1718 | if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure) |
8e8ae645 | 1719 | return true; |
8e8ae645 | 1720 | do { |
7e6ec49c | 1721 | if (time_before(jiffies, READ_ONCE(memcg->socket_pressure))) |
8e8ae645 JW |
1722 | return true; |
1723 | } while ((memcg = parent_mem_cgroup(memcg))); | |
1724 | return false; | |
e805605c | 1725 | } |
0a432dcb | 1726 | |
e4262c4f YS |
1727 | int alloc_shrinker_info(struct mem_cgroup *memcg); |
1728 | void free_shrinker_info(struct mem_cgroup *memcg); | |
2bfd3637 | 1729 | void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id); |
a178015c | 1730 | void reparent_shrinker_deferred(struct mem_cgroup *memcg); |
e805605c | 1731 | #else |
80e95fe0 | 1732 | #define mem_cgroup_sockets_enabled 0 |
2d758073 JW |
1733 | static inline void mem_cgroup_sk_alloc(struct sock *sk) { }; |
1734 | static inline void mem_cgroup_sk_free(struct sock *sk) { }; | |
baac50bb | 1735 | static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg) |
e805605c JW |
1736 | { |
1737 | return false; | |
1738 | } | |
0a432dcb | 1739 | |
2bfd3637 YS |
1740 | static inline void set_shrinker_bit(struct mem_cgroup *memcg, |
1741 | int nid, int shrinker_id) | |
0a432dcb YS |
1742 | { |
1743 | } | |
e805605c | 1744 | #endif |
7ae1e1d0 | 1745 | |
9b6f7e16 | 1746 | #ifdef CONFIG_MEMCG_KMEM |
4d5c8aed | 1747 | bool mem_cgroup_kmem_disabled(void); |
f4b00eab RG |
1748 | int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order); |
1749 | void __memcg_kmem_uncharge_page(struct page *page, int order); | |
45264778 | 1750 | |
bf4f0599 | 1751 | struct obj_cgroup *get_obj_cgroup_from_current(void); |
f4840ccf | 1752 | struct obj_cgroup *get_obj_cgroup_from_page(struct page *page); |
bf4f0599 RG |
1753 | |
1754 | int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size); | |
1755 | void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size); | |
1756 | ||
ef12947c | 1757 | extern struct static_key_false memcg_kmem_enabled_key; |
749c5415 | 1758 | |
7ae1e1d0 GC |
1759 | static inline bool memcg_kmem_enabled(void) |
1760 | { | |
eda330e5 | 1761 | return static_branch_likely(&memcg_kmem_enabled_key); |
7ae1e1d0 GC |
1762 | } |
1763 | ||
f4b00eab RG |
1764 | static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, |
1765 | int order) | |
60cd4bcd SB |
1766 | { |
1767 | if (memcg_kmem_enabled()) | |
f4b00eab | 1768 | return __memcg_kmem_charge_page(page, gfp, order); |
60cd4bcd SB |
1769 | return 0; |
1770 | } | |
1771 | ||
f4b00eab | 1772 | static inline void memcg_kmem_uncharge_page(struct page *page, int order) |
60cd4bcd SB |
1773 | { |
1774 | if (memcg_kmem_enabled()) | |
f4b00eab | 1775 | __memcg_kmem_uncharge_page(page, order); |
60cd4bcd SB |
1776 | } |
1777 | ||
33398cf2 | 1778 | /* |
a7cb874b RG |
1779 | * A helper for accessing memcg's kmem_id, used for getting |
1780 | * corresponding LRU lists. | |
33398cf2 | 1781 | */ |
7c52f65d | 1782 | static inline int memcg_kmem_id(struct mem_cgroup *memcg) |
33398cf2 MH |
1783 | { |
1784 | return memcg ? memcg->kmemcg_id : -1; | |
1785 | } | |
5722d094 | 1786 | |
8380ce47 | 1787 | struct mem_cgroup *mem_cgroup_from_obj(void *p); |
fc4db90f | 1788 | struct mem_cgroup *mem_cgroup_from_slab_obj(void *p); |
8380ce47 | 1789 | |
f4840ccf JW |
1790 | static inline void count_objcg_event(struct obj_cgroup *objcg, |
1791 | enum vm_event_item idx) | |
1792 | { | |
1793 | struct mem_cgroup *memcg; | |
1794 | ||
2eb98919 | 1795 | if (!memcg_kmem_enabled()) |
f4840ccf JW |
1796 | return; |
1797 | ||
1798 | rcu_read_lock(); | |
1799 | memcg = obj_cgroup_memcg(objcg); | |
1800 | count_memcg_events(memcg, idx, 1); | |
1801 | rcu_read_unlock(); | |
1802 | } | |
1803 | ||
7ae1e1d0 | 1804 | #else |
4d5c8aed RG |
1805 | static inline bool mem_cgroup_kmem_disabled(void) |
1806 | { | |
1807 | return true; | |
1808 | } | |
9b6f7e16 | 1809 | |
f4b00eab RG |
1810 | static inline int memcg_kmem_charge_page(struct page *page, gfp_t gfp, |
1811 | int order) | |
9b6f7e16 RG |
1812 | { |
1813 | return 0; | |
1814 | } | |
1815 | ||
f4b00eab | 1816 | static inline void memcg_kmem_uncharge_page(struct page *page, int order) |
9b6f7e16 RG |
1817 | { |
1818 | } | |
1819 | ||
f4b00eab RG |
1820 | static inline int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, |
1821 | int order) | |
60cd4bcd SB |
1822 | { |
1823 | return 0; | |
1824 | } | |
1825 | ||
f4b00eab | 1826 | static inline void __memcg_kmem_uncharge_page(struct page *page, int order) |
60cd4bcd SB |
1827 | { |
1828 | } | |
1829 | ||
f4840ccf JW |
1830 | static inline struct obj_cgroup *get_obj_cgroup_from_page(struct page *page) |
1831 | { | |
1832 | return NULL; | |
1833 | } | |
1834 | ||
b9ce5ef4 GC |
1835 | static inline bool memcg_kmem_enabled(void) |
1836 | { | |
1837 | return false; | |
1838 | } | |
1839 | ||
7c52f65d | 1840 | static inline int memcg_kmem_id(struct mem_cgroup *memcg) |
2633d7a0 GC |
1841 | { |
1842 | return -1; | |
1843 | } | |
1844 | ||
8380ce47 RG |
1845 | static inline struct mem_cgroup *mem_cgroup_from_obj(void *p) |
1846 | { | |
1d0403d2 | 1847 | return NULL; |
8380ce47 RG |
1848 | } |
1849 | ||
fc4db90f RG |
1850 | static inline struct mem_cgroup *mem_cgroup_from_slab_obj(void *p) |
1851 | { | |
1852 | return NULL; | |
1853 | } | |
1854 | ||
f4840ccf JW |
1855 | static inline void count_objcg_event(struct obj_cgroup *objcg, |
1856 | enum vm_event_item idx) | |
1857 | { | |
1858 | } | |
1859 | ||
84c07d11 | 1860 | #endif /* CONFIG_MEMCG_KMEM */ |
127424c8 | 1861 | |
f4840ccf JW |
1862 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_ZSWAP) |
1863 | bool obj_cgroup_may_zswap(struct obj_cgroup *objcg); | |
1864 | void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, size_t size); | |
1865 | void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, size_t size); | |
1866 | #else | |
1867 | static inline bool obj_cgroup_may_zswap(struct obj_cgroup *objcg) | |
1868 | { | |
1869 | return true; | |
1870 | } | |
1871 | static inline void obj_cgroup_charge_zswap(struct obj_cgroup *objcg, | |
1872 | size_t size) | |
1873 | { | |
1874 | } | |
1875 | static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup *objcg, | |
1876 | size_t size) | |
1877 | { | |
1878 | } | |
1879 | #endif | |
1880 | ||
8cdea7c0 | 1881 | #endif /* _LINUX_MEMCONTROL_H */ |