net: tcp_memcontrol: sanitize tcp memory accounting callbacks
[deliverable/linux.git] / include / linux / memcontrol.h
1 /* memcontrol.h - Memory Controller
2 *
3 * Copyright IBM Corporation, 2007
4 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
5 *
6 * Copyright 2007 OpenVZ SWsoft Inc
7 * Author: Pavel Emelianov <xemul@openvz.org>
8 *
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
18 */
19
20 #ifndef _LINUX_MEMCONTROL_H
21 #define _LINUX_MEMCONTROL_H
22 #include <linux/cgroup.h>
23 #include <linux/vm_event_item.h>
24 #include <linux/hardirq.h>
25 #include <linux/jump_label.h>
26 #include <linux/page_counter.h>
27 #include <linux/vmpressure.h>
28 #include <linux/eventfd.h>
29 #include <linux/mmzone.h>
30 #include <linux/writeback.h>
31
32 struct mem_cgroup;
33 struct page;
34 struct mm_struct;
35 struct kmem_cache;
36
37 /*
38 * The corresponding mem_cgroup_stat_names is defined in mm/memcontrol.c,
39 * These two lists should keep in accord with each other.
40 */
41 enum mem_cgroup_stat_index {
42 /*
43 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
44 */
45 MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
46 MEM_CGROUP_STAT_RSS, /* # of pages charged as anon rss */
47 MEM_CGROUP_STAT_RSS_HUGE, /* # of pages charged as anon huge */
48 MEM_CGROUP_STAT_FILE_MAPPED, /* # of pages charged as file rss */
49 MEM_CGROUP_STAT_DIRTY, /* # of dirty pages in page cache */
50 MEM_CGROUP_STAT_WRITEBACK, /* # of pages under writeback */
51 MEM_CGROUP_STAT_SWAP, /* # of pages, swapped out */
52 MEM_CGROUP_STAT_NSTATS,
53 };
54
55 struct mem_cgroup_reclaim_cookie {
56 struct zone *zone;
57 int priority;
58 unsigned int generation;
59 };
60
61 enum mem_cgroup_events_index {
62 MEM_CGROUP_EVENTS_PGPGIN, /* # of pages paged in */
63 MEM_CGROUP_EVENTS_PGPGOUT, /* # of pages paged out */
64 MEM_CGROUP_EVENTS_PGFAULT, /* # of page-faults */
65 MEM_CGROUP_EVENTS_PGMAJFAULT, /* # of major page-faults */
66 MEM_CGROUP_EVENTS_NSTATS,
67 /* default hierarchy events */
68 MEMCG_LOW = MEM_CGROUP_EVENTS_NSTATS,
69 MEMCG_HIGH,
70 MEMCG_MAX,
71 MEMCG_OOM,
72 MEMCG_NR_EVENTS,
73 };
74
75 /*
76 * Per memcg event counter is incremented at every pagein/pageout. With THP,
77 * it will be incremated by the number of pages. This counter is used for
78 * for trigger some periodic events. This is straightforward and better
79 * than using jiffies etc. to handle periodic memcg event.
80 */
81 enum mem_cgroup_events_target {
82 MEM_CGROUP_TARGET_THRESH,
83 MEM_CGROUP_TARGET_SOFTLIMIT,
84 MEM_CGROUP_TARGET_NUMAINFO,
85 MEM_CGROUP_NTARGETS,
86 };
87
88 struct cg_proto {
89 struct page_counter memory_allocated; /* Current allocated memory. */
90 int memory_pressure;
91 bool active;
92 /*
93 * memcg field is used to find which memcg we belong directly
94 * Each memcg struct can hold more than one cg_proto, so container_of
95 * won't really cut.
96 *
97 * The elegant solution would be having an inverse function to
98 * proto_cgroup in struct proto, but that means polluting the structure
99 * for everybody, instead of just for memcg users.
100 */
101 struct mem_cgroup *memcg;
102 };
103
104 #ifdef CONFIG_MEMCG
105 struct mem_cgroup_stat_cpu {
106 long count[MEM_CGROUP_STAT_NSTATS];
107 unsigned long events[MEMCG_NR_EVENTS];
108 unsigned long nr_page_events;
109 unsigned long targets[MEM_CGROUP_NTARGETS];
110 };
111
112 struct mem_cgroup_reclaim_iter {
113 struct mem_cgroup *position;
114 /* scan generation, increased every round-trip */
115 unsigned int generation;
116 };
117
118 /*
119 * per-zone information in memory controller.
120 */
121 struct mem_cgroup_per_zone {
122 struct lruvec lruvec;
123 unsigned long lru_size[NR_LRU_LISTS];
124
125 struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
126
127 struct rb_node tree_node; /* RB tree node */
128 unsigned long usage_in_excess;/* Set to the value by which */
129 /* the soft limit is exceeded*/
130 bool on_tree;
131 struct mem_cgroup *memcg; /* Back pointer, we cannot */
132 /* use container_of */
133 };
134
135 struct mem_cgroup_per_node {
136 struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
137 };
138
139 struct mem_cgroup_threshold {
140 struct eventfd_ctx *eventfd;
141 unsigned long threshold;
142 };
143
144 /* For threshold */
145 struct mem_cgroup_threshold_ary {
146 /* An array index points to threshold just below or equal to usage. */
147 int current_threshold;
148 /* Size of entries[] */
149 unsigned int size;
150 /* Array of thresholds */
151 struct mem_cgroup_threshold entries[0];
152 };
153
154 struct mem_cgroup_thresholds {
155 /* Primary thresholds array */
156 struct mem_cgroup_threshold_ary *primary;
157 /*
158 * Spare threshold array.
159 * This is needed to make mem_cgroup_unregister_event() "never fail".
160 * It must be able to store at least primary->size - 1 entries.
161 */
162 struct mem_cgroup_threshold_ary *spare;
163 };
164
165 /*
166 * The memory controller data structure. The memory controller controls both
167 * page cache and RSS per cgroup. We would eventually like to provide
168 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
169 * to help the administrator determine what knobs to tune.
170 */
171 struct mem_cgroup {
172 struct cgroup_subsys_state css;
173
174 /* Accounted resources */
175 struct page_counter memory;
176 struct page_counter memsw;
177 struct page_counter kmem;
178
179 /* Normal memory consumption range */
180 unsigned long low;
181 unsigned long high;
182
183 unsigned long soft_limit;
184
185 /* vmpressure notifications */
186 struct vmpressure vmpressure;
187
188 /* css_online() has been completed */
189 int initialized;
190
191 /*
192 * Should the accounting and control be hierarchical, per subtree?
193 */
194 bool use_hierarchy;
195
196 /* protected by memcg_oom_lock */
197 bool oom_lock;
198 int under_oom;
199
200 int swappiness;
201 /* OOM-Killer disable */
202 int oom_kill_disable;
203
204 /* handle for "memory.events" */
205 struct cgroup_file events_file;
206
207 /* protect arrays of thresholds */
208 struct mutex thresholds_lock;
209
210 /* thresholds for memory usage. RCU-protected */
211 struct mem_cgroup_thresholds thresholds;
212
213 /* thresholds for mem+swap usage. RCU-protected */
214 struct mem_cgroup_thresholds memsw_thresholds;
215
216 /* For oom notifier event fd */
217 struct list_head oom_notify;
218
219 /*
220 * Should we move charges of a task when a task is moved into this
221 * mem_cgroup ? And what type of charges should we move ?
222 */
223 unsigned long move_charge_at_immigrate;
224 /*
225 * set > 0 if pages under this cgroup are moving to other cgroup.
226 */
227 atomic_t moving_account;
228 /* taken only while moving_account > 0 */
229 spinlock_t move_lock;
230 struct task_struct *move_lock_task;
231 unsigned long move_lock_flags;
232 /*
233 * percpu counter.
234 */
235 struct mem_cgroup_stat_cpu __percpu *stat;
236
237 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
238 struct cg_proto tcp_mem;
239 #endif
240 #if defined(CONFIG_MEMCG_KMEM)
241 /* Index in the kmem_cache->memcg_params.memcg_caches array */
242 int kmemcg_id;
243 bool kmem_acct_activated;
244 bool kmem_acct_active;
245 #endif
246
247 int last_scanned_node;
248 #if MAX_NUMNODES > 1
249 nodemask_t scan_nodes;
250 atomic_t numainfo_events;
251 atomic_t numainfo_updating;
252 #endif
253
254 #ifdef CONFIG_CGROUP_WRITEBACK
255 struct list_head cgwb_list;
256 struct wb_domain cgwb_domain;
257 #endif
258
259 /* List of events which userspace want to receive */
260 struct list_head event_list;
261 spinlock_t event_list_lock;
262
263 struct mem_cgroup_per_node *nodeinfo[0];
264 /* WARNING: nodeinfo must be the last member here */
265 };
266
267 extern struct mem_cgroup *root_mem_cgroup;
268
269 /**
270 * mem_cgroup_events - count memory events against a cgroup
271 * @memcg: the memory cgroup
272 * @idx: the event index
273 * @nr: the number of events to account for
274 */
275 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
276 enum mem_cgroup_events_index idx,
277 unsigned int nr)
278 {
279 this_cpu_add(memcg->stat->events[idx], nr);
280 cgroup_file_notify(&memcg->events_file);
281 }
282
283 bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg);
284
285 int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
286 gfp_t gfp_mask, struct mem_cgroup **memcgp);
287 void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
288 bool lrucare);
289 void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg);
290 void mem_cgroup_uncharge(struct page *page);
291 void mem_cgroup_uncharge_list(struct list_head *page_list);
292
293 void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage);
294
295 struct lruvec *mem_cgroup_zone_lruvec(struct zone *, struct mem_cgroup *);
296 struct lruvec *mem_cgroup_page_lruvec(struct page *, struct zone *);
297
298 bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg);
299 struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p);
300 struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg);
301
302 static inline
303 struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *css){
304 return css ? container_of(css, struct mem_cgroup, css) : NULL;
305 }
306
307 struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
308 struct mem_cgroup *,
309 struct mem_cgroup_reclaim_cookie *);
310 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
311
312 static inline bool mem_cgroup_is_descendant(struct mem_cgroup *memcg,
313 struct mem_cgroup *root)
314 {
315 if (root == memcg)
316 return true;
317 if (!root->use_hierarchy)
318 return false;
319 return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
320 }
321
322 static inline bool mm_match_cgroup(struct mm_struct *mm,
323 struct mem_cgroup *memcg)
324 {
325 struct mem_cgroup *task_memcg;
326 bool match = false;
327
328 rcu_read_lock();
329 task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
330 if (task_memcg)
331 match = mem_cgroup_is_descendant(task_memcg, memcg);
332 rcu_read_unlock();
333 return match;
334 }
335
336 struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page);
337 ino_t page_cgroup_ino(struct page *page);
338
339 static inline bool mem_cgroup_disabled(void)
340 {
341 return !cgroup_subsys_enabled(memory_cgrp_subsys);
342 }
343
344 /*
345 * For memory reclaim.
346 */
347 int mem_cgroup_select_victim_node(struct mem_cgroup *memcg);
348
349 void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
350 int nr_pages);
351
352 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
353 {
354 struct mem_cgroup_per_zone *mz;
355 struct mem_cgroup *memcg;
356
357 if (mem_cgroup_disabled())
358 return true;
359
360 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
361 memcg = mz->memcg;
362
363 return !!(memcg->css.flags & CSS_ONLINE);
364 }
365
366 static inline
367 unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
368 {
369 struct mem_cgroup_per_zone *mz;
370
371 mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
372 return mz->lru_size[lru];
373 }
374
375 static inline bool mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
376 {
377 unsigned long inactive_ratio;
378 unsigned long inactive;
379 unsigned long active;
380 unsigned long gb;
381
382 inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
383 active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
384
385 gb = (inactive + active) >> (30 - PAGE_SHIFT);
386 if (gb)
387 inactive_ratio = int_sqrt(10 * gb);
388 else
389 inactive_ratio = 1;
390
391 return inactive * inactive_ratio < active;
392 }
393
394 void mem_cgroup_handle_over_high(void);
395
396 void mem_cgroup_print_oom_info(struct mem_cgroup *memcg,
397 struct task_struct *p);
398
399 static inline void mem_cgroup_oom_enable(void)
400 {
401 WARN_ON(current->memcg_may_oom);
402 current->memcg_may_oom = 1;
403 }
404
405 static inline void mem_cgroup_oom_disable(void)
406 {
407 WARN_ON(!current->memcg_may_oom);
408 current->memcg_may_oom = 0;
409 }
410
411 static inline bool task_in_memcg_oom(struct task_struct *p)
412 {
413 return p->memcg_in_oom;
414 }
415
416 bool mem_cgroup_oom_synchronize(bool wait);
417
418 #ifdef CONFIG_MEMCG_SWAP
419 extern int do_swap_account;
420 #endif
421
422 struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page);
423 void mem_cgroup_end_page_stat(struct mem_cgroup *memcg);
424
425 /**
426 * mem_cgroup_update_page_stat - update page state statistics
427 * @memcg: memcg to account against
428 * @idx: page state item to account
429 * @val: number of pages (positive or negative)
430 *
431 * See mem_cgroup_begin_page_stat() for locking requirements.
432 */
433 static inline void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
434 enum mem_cgroup_stat_index idx, int val)
435 {
436 VM_BUG_ON(!rcu_read_lock_held());
437
438 if (memcg)
439 this_cpu_add(memcg->stat->count[idx], val);
440 }
441
442 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
443 enum mem_cgroup_stat_index idx)
444 {
445 mem_cgroup_update_page_stat(memcg, idx, 1);
446 }
447
448 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
449 enum mem_cgroup_stat_index idx)
450 {
451 mem_cgroup_update_page_stat(memcg, idx, -1);
452 }
453
454 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
455 gfp_t gfp_mask,
456 unsigned long *total_scanned);
457
458 static inline void mem_cgroup_count_vm_event(struct mm_struct *mm,
459 enum vm_event_item idx)
460 {
461 struct mem_cgroup *memcg;
462
463 if (mem_cgroup_disabled())
464 return;
465
466 rcu_read_lock();
467 memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
468 if (unlikely(!memcg))
469 goto out;
470
471 switch (idx) {
472 case PGFAULT:
473 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
474 break;
475 case PGMAJFAULT:
476 this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
477 break;
478 default:
479 BUG();
480 }
481 out:
482 rcu_read_unlock();
483 }
484 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
485 void mem_cgroup_split_huge_fixup(struct page *head);
486 #endif
487
488 #else /* CONFIG_MEMCG */
489 struct mem_cgroup;
490
491 static inline void mem_cgroup_events(struct mem_cgroup *memcg,
492 enum mem_cgroup_events_index idx,
493 unsigned int nr)
494 {
495 }
496
497 static inline bool mem_cgroup_low(struct mem_cgroup *root,
498 struct mem_cgroup *memcg)
499 {
500 return false;
501 }
502
503 static inline int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
504 gfp_t gfp_mask,
505 struct mem_cgroup **memcgp)
506 {
507 *memcgp = NULL;
508 return 0;
509 }
510
511 static inline void mem_cgroup_commit_charge(struct page *page,
512 struct mem_cgroup *memcg,
513 bool lrucare)
514 {
515 }
516
517 static inline void mem_cgroup_cancel_charge(struct page *page,
518 struct mem_cgroup *memcg)
519 {
520 }
521
522 static inline void mem_cgroup_uncharge(struct page *page)
523 {
524 }
525
526 static inline void mem_cgroup_uncharge_list(struct list_head *page_list)
527 {
528 }
529
530 static inline void mem_cgroup_replace_page(struct page *old, struct page *new)
531 {
532 }
533
534 static inline struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
535 struct mem_cgroup *memcg)
536 {
537 return &zone->lruvec;
538 }
539
540 static inline struct lruvec *mem_cgroup_page_lruvec(struct page *page,
541 struct zone *zone)
542 {
543 return &zone->lruvec;
544 }
545
546 static inline bool mm_match_cgroup(struct mm_struct *mm,
547 struct mem_cgroup *memcg)
548 {
549 return true;
550 }
551
552 static inline bool task_in_mem_cgroup(struct task_struct *task,
553 const struct mem_cgroup *memcg)
554 {
555 return true;
556 }
557
558 static inline struct mem_cgroup *
559 mem_cgroup_iter(struct mem_cgroup *root,
560 struct mem_cgroup *prev,
561 struct mem_cgroup_reclaim_cookie *reclaim)
562 {
563 return NULL;
564 }
565
566 static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
567 struct mem_cgroup *prev)
568 {
569 }
570
571 static inline bool mem_cgroup_disabled(void)
572 {
573 return true;
574 }
575
576 static inline bool
577 mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
578 {
579 return true;
580 }
581
582 static inline bool mem_cgroup_lruvec_online(struct lruvec *lruvec)
583 {
584 return true;
585 }
586
587 static inline unsigned long
588 mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
589 {
590 return 0;
591 }
592
593 static inline void
594 mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
595 int increment)
596 {
597 }
598
599 static inline void
600 mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
601 {
602 }
603
604 static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
605 {
606 return NULL;
607 }
608
609 static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
610 {
611 }
612
613 static inline void mem_cgroup_handle_over_high(void)
614 {
615 }
616
617 static inline void mem_cgroup_oom_enable(void)
618 {
619 }
620
621 static inline void mem_cgroup_oom_disable(void)
622 {
623 }
624
625 static inline bool task_in_memcg_oom(struct task_struct *p)
626 {
627 return false;
628 }
629
630 static inline bool mem_cgroup_oom_synchronize(bool wait)
631 {
632 return false;
633 }
634
635 static inline void mem_cgroup_inc_page_stat(struct mem_cgroup *memcg,
636 enum mem_cgroup_stat_index idx)
637 {
638 }
639
640 static inline void mem_cgroup_dec_page_stat(struct mem_cgroup *memcg,
641 enum mem_cgroup_stat_index idx)
642 {
643 }
644
645 static inline
646 unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
647 gfp_t gfp_mask,
648 unsigned long *total_scanned)
649 {
650 return 0;
651 }
652
653 static inline void mem_cgroup_split_huge_fixup(struct page *head)
654 {
655 }
656
657 static inline
658 void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
659 {
660 }
661 #endif /* CONFIG_MEMCG */
662
663 #ifdef CONFIG_CGROUP_WRITEBACK
664
665 struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg);
666 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
667 void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
668 unsigned long *pheadroom, unsigned long *pdirty,
669 unsigned long *pwriteback);
670
671 #else /* CONFIG_CGROUP_WRITEBACK */
672
673 static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
674 {
675 return NULL;
676 }
677
678 static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb,
679 unsigned long *pfilepages,
680 unsigned long *pheadroom,
681 unsigned long *pdirty,
682 unsigned long *pwriteback)
683 {
684 }
685
686 #endif /* CONFIG_CGROUP_WRITEBACK */
687
688 struct sock;
689 void sock_update_memcg(struct sock *sk);
690 void sock_release_memcg(struct sock *sk);
691 bool mem_cgroup_charge_skmem(struct cg_proto *proto, unsigned int nr_pages);
692 void mem_cgroup_uncharge_skmem(struct cg_proto *proto, unsigned int nr_pages);
693 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
694 static inline bool mem_cgroup_under_socket_pressure(struct cg_proto *proto)
695 {
696 return proto->memory_pressure;
697 }
698 #else
699 static inline bool mem_cgroup_under_pressure(struct cg_proto *proto)
700 {
701 return false;
702 }
703 #endif
704
705 #ifdef CONFIG_MEMCG_KMEM
706 extern struct static_key memcg_kmem_enabled_key;
707
708 extern int memcg_nr_cache_ids;
709 void memcg_get_cache_ids(void);
710 void memcg_put_cache_ids(void);
711
712 /*
713 * Helper macro to loop through all memcg-specific caches. Callers must still
714 * check if the cache is valid (it is either valid or NULL).
715 * the slab_mutex must be held when looping through those caches
716 */
717 #define for_each_memcg_cache_index(_idx) \
718 for ((_idx) = 0; (_idx) < memcg_nr_cache_ids; (_idx)++)
719
720 static inline bool memcg_kmem_enabled(void)
721 {
722 return static_key_false(&memcg_kmem_enabled_key);
723 }
724
725 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
726 {
727 return memcg->kmem_acct_active;
728 }
729
730 /*
731 * In general, we'll do everything in our power to not incur in any overhead
732 * for non-memcg users for the kmem functions. Not even a function call, if we
733 * can avoid it.
734 *
735 * Therefore, we'll inline all those functions so that in the best case, we'll
736 * see that kmemcg is off for everybody and proceed quickly. If it is on,
737 * we'll still do most of the flag checking inline. We check a lot of
738 * conditions, but because they are pretty simple, they are expected to be
739 * fast.
740 */
741 int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
742 struct mem_cgroup *memcg);
743 int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order);
744 void __memcg_kmem_uncharge(struct page *page, int order);
745
746 /*
747 * helper for acessing a memcg's index. It will be used as an index in the
748 * child cache array in kmem_cache, and also to derive its name. This function
749 * will return -1 when this is not a kmem-limited memcg.
750 */
751 static inline int memcg_cache_id(struct mem_cgroup *memcg)
752 {
753 return memcg ? memcg->kmemcg_id : -1;
754 }
755
756 struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp);
757 void __memcg_kmem_put_cache(struct kmem_cache *cachep);
758
759 static inline bool __memcg_kmem_bypass(void)
760 {
761 if (!memcg_kmem_enabled())
762 return true;
763 if (in_interrupt() || (!current->mm) || (current->flags & PF_KTHREAD))
764 return true;
765 return false;
766 }
767
768 /**
769 * memcg_kmem_charge: charge a kmem page
770 * @page: page to charge
771 * @gfp: reclaim mode
772 * @order: allocation order
773 *
774 * Returns 0 on success, an error code on failure.
775 */
776 static __always_inline int memcg_kmem_charge(struct page *page,
777 gfp_t gfp, int order)
778 {
779 if (__memcg_kmem_bypass())
780 return 0;
781 if (!(gfp & __GFP_ACCOUNT))
782 return 0;
783 return __memcg_kmem_charge(page, gfp, order);
784 }
785
786 /**
787 * memcg_kmem_uncharge: uncharge a kmem page
788 * @page: page to uncharge
789 * @order: allocation order
790 */
791 static __always_inline void memcg_kmem_uncharge(struct page *page, int order)
792 {
793 if (memcg_kmem_enabled())
794 __memcg_kmem_uncharge(page, order);
795 }
796
797 /**
798 * memcg_kmem_get_cache: selects the correct per-memcg cache for allocation
799 * @cachep: the original global kmem cache
800 *
801 * All memory allocated from a per-memcg cache is charged to the owner memcg.
802 */
803 static __always_inline struct kmem_cache *
804 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
805 {
806 if (__memcg_kmem_bypass())
807 return cachep;
808 return __memcg_kmem_get_cache(cachep, gfp);
809 }
810
811 static __always_inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
812 {
813 if (memcg_kmem_enabled())
814 __memcg_kmem_put_cache(cachep);
815 }
816 #else
817 #define for_each_memcg_cache_index(_idx) \
818 for (; NULL; )
819
820 static inline bool memcg_kmem_enabled(void)
821 {
822 return false;
823 }
824
825 static inline bool memcg_kmem_is_active(struct mem_cgroup *memcg)
826 {
827 return false;
828 }
829
830 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
831 {
832 return 0;
833 }
834
835 static inline void memcg_kmem_uncharge(struct page *page, int order)
836 {
837 }
838
839 static inline int memcg_cache_id(struct mem_cgroup *memcg)
840 {
841 return -1;
842 }
843
844 static inline void memcg_get_cache_ids(void)
845 {
846 }
847
848 static inline void memcg_put_cache_ids(void)
849 {
850 }
851
852 static inline struct kmem_cache *
853 memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
854 {
855 return cachep;
856 }
857
858 static inline void memcg_kmem_put_cache(struct kmem_cache *cachep)
859 {
860 }
861 #endif /* CONFIG_MEMCG_KMEM */
862 #endif /* _LINUX_MEMCONTROL_H */
This page took 0.09543 seconds and 6 git commands to generate.