1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* memcontrol.h - Memory Controller
4 * Copyright IBM Corporation, 2007
5 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
7 * Copyright 2007 OpenVZ SWsoft Inc
8 * Author: Pavel Emelianov <xemul@openvz.org>
11 #ifndef _LINUX_MEMCONTROL_H
12 #define _LINUX_MEMCONTROL_H
13 #include <linux/cgroup.h>
14 #include <linux/vm_event_item.h>
15 #include <linux/hardirq.h>
16 #include <linux/jump_label.h>
17 #include <linux/kernel.h>
18 #include <linux/page_counter.h>
19 #include <linux/vmpressure.h>
20 #include <linux/eventfd.h>
22 #include <linux/vmstat.h>
23 #include <linux/writeback.h>
24 #include <linux/page-flags.h>
25 #include <linux/shrinker.h>
33 /* Cgroup-specific page state, on top of universal node page state */
34 enum memcg_stat_item
{
35 MEMCG_SWAP
= NR_VM_NODE_STAT_ITEMS
,
45 enum memcg_memory_event
{
55 MEMCG_NR_MEMORY_EVENTS
,
58 struct mem_cgroup_reclaim_cookie
{
65 #define MEM_CGROUP_ID_SHIFT 16
67 struct mem_cgroup_id
{
72 struct memcg_vmstats_percpu
;
73 struct memcg1_events_percpu
;
75 struct lruvec_stats_percpu
;
78 struct mem_cgroup_reclaim_iter
{
79 struct mem_cgroup
*position
;
80 /* scan generation, increased every round-trip */
85 * per-node information in memory controller.
87 struct mem_cgroup_per_node
{
88 /* Keep the read-only fields at the start */
89 struct mem_cgroup
*memcg
; /* Back pointer, we cannot */
90 /* use container_of */
92 struct lruvec_stats_percpu __percpu
*lruvec_stats_percpu
;
93 struct lruvec_stats
*lruvec_stats
;
94 struct shrinker_info __rcu
*shrinker_info
;
96 #ifdef CONFIG_MEMCG_V1
98 * Memcg-v1 only stuff in middle as buffer between read mostly fields
99 * and update often fields to avoid false sharing. If v1 stuff is
100 * not present, an explicit padding is needed.
103 struct rb_node tree_node
; /* RB tree node */
104 unsigned long usage_in_excess
;/* Set to the value by which */
105 /* the soft limit is exceeded*/
108 CACHELINE_PADDING(_pad1_
);
111 /* Fields which get updated often at the end. */
112 struct lruvec lruvec
;
113 CACHELINE_PADDING(_pad2_
);
114 unsigned long lru_zone_size
[MAX_NR_ZONES
][NR_LRU_LISTS
];
115 struct mem_cgroup_reclaim_iter iter
;
118 struct mem_cgroup_threshold
{
119 struct eventfd_ctx
*eventfd
;
120 unsigned long threshold
;
124 struct mem_cgroup_threshold_ary
{
125 /* An array index points to threshold just below or equal to usage. */
126 int current_threshold
;
127 /* Size of entries[] */
129 /* Array of thresholds */
130 struct mem_cgroup_threshold entries
[] __counted_by(size
);
133 struct mem_cgroup_thresholds
{
134 /* Primary thresholds array */
135 struct mem_cgroup_threshold_ary
*primary
;
137 * Spare threshold array.
138 * This is needed to make mem_cgroup_unregister_event() "never fail".
139 * It must be able to store at least primary->size - 1 entries.
141 struct mem_cgroup_threshold_ary
*spare
;
145 * Remember four most recent foreign writebacks with dirty pages in this
146 * cgroup. Inode sharing is expected to be uncommon and, even if we miss
147 * one in a given round, we're likely to catch it later if it keeps
148 * foreign-dirtying, so a fairly low count should be enough.
150 * See mem_cgroup_track_foreign_dirty_slowpath() for details.
152 #define MEMCG_CGWB_FRN_CNT 4
154 struct memcg_cgwb_frn
{
155 u64 bdi_id
; /* bdi->id of the foreign inode */
156 int memcg_id
; /* memcg->css.id of foreign inode */
157 u64 at
; /* jiffies_64 at the time of dirtying */
158 struct wb_completion done
; /* tracks in-flight foreign writebacks */
162 * Bucket for arbitrarily byte-sized objects charged to a memory
163 * cgroup. The bucket can be reparented in one piece when the cgroup
164 * is destroyed, without having to round up the individual references
165 * of all live memory objects in the wild.
168 struct percpu_ref refcnt
;
169 struct mem_cgroup
*memcg
;
170 atomic_t nr_charged_bytes
;
172 struct list_head list
; /* protected by objcg_lock */
178 * The memory controller data structure. The memory controller controls both
179 * page cache and RSS per cgroup. We would eventually like to provide
180 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
181 * to help the administrator determine what knobs to tune.
184 struct cgroup_subsys_state css
;
186 /* Private memcg ID. Used to ID objects that outlive the cgroup */
187 struct mem_cgroup_id id
;
189 /* Accounted resources */
190 struct page_counter memory
; /* Both v1 & v2 */
193 struct page_counter swap
; /* v2 only */
194 struct page_counter memsw
; /* v1 only */
197 /* registered local peak watchers */
198 struct list_head memory_peaks
;
199 struct list_head swap_peaks
;
200 spinlock_t peaks_lock
;
202 /* Range enforcement for interrupt charges */
203 struct work_struct high_work
;
206 unsigned long zswap_max
;
209 * Prevent pages from this memcg from being written back from zswap to
210 * swap, and from being swapped out on zswap store failures.
212 bool zswap_writeback
;
215 /* vmpressure notifications */
216 struct vmpressure vmpressure
;
219 * Should the OOM killer kill all belonging tasks, had it kill one?
225 /* memory.events and memory.events.local */
226 struct cgroup_file events_file
;
227 struct cgroup_file events_local_file
;
229 /* handle for "memory.swap.events" */
230 struct cgroup_file swap_events_file
;
233 struct memcg_vmstats
*vmstats
;
236 atomic_long_t memory_events
[MEMCG_NR_MEMORY_EVENTS
];
237 atomic_long_t memory_events_local
[MEMCG_NR_MEMORY_EVENTS
];
240 * Hint of reclaim pressure for socket memroy management. Note
241 * that this indicator should NOT be used in legacy cgroup mode
242 * where socket memory is accounted/charged separately.
244 unsigned long socket_pressure
;
248 * memcg->objcg is wiped out as a part of the objcg repaprenting
249 * process. memcg->orig_objcg preserves a pointer (and a reference)
250 * to the original objcg until the end of live of memcg.
252 struct obj_cgroup __rcu
*objcg
;
253 struct obj_cgroup
*orig_objcg
;
254 /* list of inherited objcgs, protected by objcg_lock */
255 struct list_head objcg_list
;
257 struct memcg_vmstats_percpu __percpu
*vmstats_percpu
;
259 #ifdef CONFIG_CGROUP_WRITEBACK
260 struct list_head cgwb_list
;
261 struct wb_domain cgwb_domain
;
262 struct memcg_cgwb_frn cgwb_frn
[MEMCG_CGWB_FRN_CNT
];
265 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
266 struct deferred_split deferred_split_queue
;
269 #ifdef CONFIG_LRU_GEN_WALKS_MMU
270 /* per-memcg mm_struct list */
271 struct lru_gen_mm_list mm_list
;
274 #ifdef CONFIG_MEMCG_V1
275 /* Legacy consumer-oriented counters */
276 struct page_counter kmem
; /* v1 only */
277 struct page_counter tcpmem
; /* v1 only */
279 struct memcg1_events_percpu __percpu
*events_percpu
;
281 unsigned long soft_limit
;
283 /* protected by memcg_oom_lock */
287 /* OOM-Killer disable */
288 int oom_kill_disable
;
290 /* protect arrays of thresholds */
291 struct mutex thresholds_lock
;
293 /* thresholds for memory usage. RCU-protected */
294 struct mem_cgroup_thresholds thresholds
;
296 /* thresholds for mem+swap usage. RCU-protected */
297 struct mem_cgroup_thresholds memsw_thresholds
;
299 /* For oom notifier event fd */
300 struct list_head oom_notify
;
302 /* Legacy tcp memory accounting */
306 /* List of events which userspace want to receive */
307 struct list_head event_list
;
308 spinlock_t event_list_lock
;
309 #endif /* CONFIG_MEMCG_V1 */
311 struct mem_cgroup_per_node
*nodeinfo
[];
315 * size of first charge trial.
316 * TODO: maybe necessary to use big numbers in big irons or dynamic based of the
319 #define MEMCG_CHARGE_BATCH 64U
321 extern struct mem_cgroup
*root_mem_cgroup
;
323 enum page_memcg_data_flags
{
324 /* page->memcg_data is a pointer to an slabobj_ext vector */
325 MEMCG_DATA_OBJEXTS
= (1UL << 0),
326 /* page has been accounted as a non-slab kernel page */
327 MEMCG_DATA_KMEM
= (1UL << 1),
328 /* the next bit after the last actual flag */
329 __NR_MEMCG_DATA_FLAGS
= (1UL << 2),
332 #define __FIRST_OBJEXT_FLAG __NR_MEMCG_DATA_FLAGS
334 #else /* CONFIG_MEMCG */
336 #define __FIRST_OBJEXT_FLAG (1UL << 0)
338 #endif /* CONFIG_MEMCG */
341 /* slabobj_ext vector failed to allocate */
342 OBJEXTS_ALLOC_FAIL
= __FIRST_OBJEXT_FLAG
,
343 /* the next bit after the last actual flag */
344 __NR_OBJEXTS_FLAGS
= (__FIRST_OBJEXT_FLAG
<< 1),
347 #define OBJEXTS_FLAGS_MASK (__NR_OBJEXTS_FLAGS - 1)
351 static inline bool folio_memcg_kmem(struct folio
*folio
);
354 * After the initialization objcg->memcg is always pointing at
355 * a valid memcg, but can be atomically swapped to the parent memcg.
357 * The caller must ensure that the returned memcg won't be released.
359 static inline struct mem_cgroup
*obj_cgroup_memcg(struct obj_cgroup
*objcg
)
361 lockdep_assert_once(rcu_read_lock_held() || lockdep_is_held(&cgroup_mutex
));
362 return READ_ONCE(objcg
->memcg
);
366 * __folio_memcg - Get the memory cgroup associated with a non-kmem folio
367 * @folio: Pointer to the folio.
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
371 * proper memory cgroup pointer. It's not safe to call this function
372 * against some type of folios, e.g. slab folios or ex-slab folios or
375 static inline struct mem_cgroup
*__folio_memcg(struct folio
*folio
)
377 unsigned long memcg_data
= folio
->memcg_data
;
379 VM_BUG_ON_FOLIO(folio_test_slab(folio
), folio
);
380 VM_BUG_ON_FOLIO(memcg_data
& MEMCG_DATA_OBJEXTS
, folio
);
381 VM_BUG_ON_FOLIO(memcg_data
& MEMCG_DATA_KMEM
, folio
);
383 return (struct mem_cgroup
*)(memcg_data
& ~OBJEXTS_FLAGS_MASK
);
387 * __folio_objcg - get the object cgroup associated with a kmem folio.
388 * @folio: Pointer to the folio.
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
392 * proper object cgroup pointer. It's not safe to call this function
393 * against some type of folios, e.g. slab folios or ex-slab folios or
396 static inline struct obj_cgroup
*__folio_objcg(struct folio
*folio
)
398 unsigned long memcg_data
= folio
->memcg_data
;
400 VM_BUG_ON_FOLIO(folio_test_slab(folio
), folio
);
401 VM_BUG_ON_FOLIO(memcg_data
& MEMCG_DATA_OBJEXTS
, folio
);
402 VM_BUG_ON_FOLIO(!(memcg_data
& MEMCG_DATA_KMEM
), folio
);
404 return (struct obj_cgroup
*)(memcg_data
& ~OBJEXTS_FLAGS_MASK
);
408 * folio_memcg - Get the memory cgroup associated with a folio.
409 * @folio: Pointer to the folio.
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
413 * proper memory cgroup pointer. It's not safe to call this function
414 * against some type of folios, e.g. slab folios or ex-slab folios.
416 * For a non-kmem folio any of the following ensures folio and memcg binding
421 * - exclusive reference
423 * For a kmem folio a caller should hold an rcu read lock to protect memcg
424 * associated with a kmem folio from being released.
426 static inline struct mem_cgroup
*folio_memcg(struct folio
*folio
)
428 if (folio_memcg_kmem(folio
))
429 return obj_cgroup_memcg(__folio_objcg(folio
));
430 return __folio_memcg(folio
);
434 * folio_memcg_charged - If a folio is charged to a memory cgroup.
435 * @folio: Pointer to the folio.
437 * Returns true if folio is charged to a memory cgroup, otherwise returns false.
439 static inline bool folio_memcg_charged(struct folio
*folio
)
441 if (folio_memcg_kmem(folio
))
442 return __folio_objcg(folio
) != NULL
;
443 return __folio_memcg(folio
) != NULL
;
447 * folio_memcg_check - Get the memory cgroup associated with a folio.
448 * @folio: Pointer to the folio.
450 * Returns a pointer to the memory cgroup associated with the folio,
451 * or NULL. This function unlike folio_memcg() can take any folio
452 * as an argument. It has to be used in cases when it's not known if a folio
453 * has an associated memory cgroup pointer or an object cgroups vector or
456 * For a non-kmem folio any of the following ensures folio and memcg binding
461 * - exclusive reference
463 * For a kmem folio a caller should hold an rcu read lock to protect memcg
464 * associated with a kmem folio from being released.
466 static inline struct mem_cgroup
*folio_memcg_check(struct folio
*folio
)
469 * Because folio->memcg_data might be changed asynchronously
470 * for slabs, READ_ONCE() should be used here.
472 unsigned long memcg_data
= READ_ONCE(folio
->memcg_data
);
474 if (memcg_data
& MEMCG_DATA_OBJEXTS
)
477 if (memcg_data
& MEMCG_DATA_KMEM
) {
478 struct obj_cgroup
*objcg
;
480 objcg
= (void *)(memcg_data
& ~OBJEXTS_FLAGS_MASK
);
481 return obj_cgroup_memcg(objcg
);
484 return (struct mem_cgroup
*)(memcg_data
& ~OBJEXTS_FLAGS_MASK
);
487 static inline struct mem_cgroup
*page_memcg_check(struct page
*page
)
491 return folio_memcg_check((struct folio
*)page
);
494 static inline struct mem_cgroup
*get_mem_cgroup_from_objcg(struct obj_cgroup
*objcg
)
496 struct mem_cgroup
*memcg
;
500 memcg
= obj_cgroup_memcg(objcg
);
501 if (unlikely(!css_tryget(&memcg
->css
)))
509 * folio_memcg_kmem - Check if the folio has the memcg_kmem flag set.
510 * @folio: Pointer to the folio.
512 * Checks if the folio has MemcgKmem flag set. The caller must ensure
513 * that the folio has an associated memory cgroup. It's not safe to call
514 * this function against some types of folios, e.g. slab folios.
516 static inline bool folio_memcg_kmem(struct folio
*folio
)
518 VM_BUG_ON_PGFLAGS(PageTail(&folio
->page
), &folio
->page
);
519 VM_BUG_ON_FOLIO(folio
->memcg_data
& MEMCG_DATA_OBJEXTS
, folio
);
520 return folio
->memcg_data
& MEMCG_DATA_KMEM
;
523 static inline bool PageMemcgKmem(struct page
*page
)
525 return folio_memcg_kmem(page_folio(page
));
528 static inline bool mem_cgroup_is_root(struct mem_cgroup
*memcg
)
530 return (memcg
== root_mem_cgroup
);
533 static inline bool mem_cgroup_disabled(void)
535 return !cgroup_subsys_enabled(memory_cgrp_subsys
);
538 static inline void mem_cgroup_protection(struct mem_cgroup
*root
,
539 struct mem_cgroup
*memcg
,
545 if (mem_cgroup_disabled())
549 * There is no reclaim protection applied to a targeted reclaim.
550 * We are special casing this specific case here because
551 * mem_cgroup_calculate_protection is not robust enough to keep
552 * the protection invariant for calculated effective values for
553 * parallel reclaimers with different reclaim target. This is
554 * especially a problem for tail memcgs (as they have pages on LRU)
555 * which would want to have effective values 0 for targeted reclaim
556 * but a different value for external reclaim.
559 * Let's have global and A's reclaim in parallel:
561 * A (low=2G, usage = 3G, max = 3G, children_low_usage = 1.5G)
563 * | C (low = 1G, usage = 2.5G)
564 * B (low = 1G, usage = 0.5G)
566 * For the global reclaim
568 * B.elow = min(B.usage, B.low) because children_low_usage <= A.elow
569 * C.elow = min(C.usage, C.low)
571 * With the effective values resetting we have A reclaim
576 * If the global reclaim races with A's reclaim then
577 * B.elow = C.elow = 0 because children_low_usage > A.elow)
578 * is possible and reclaiming B would be violating the protection.
584 *min
= READ_ONCE(memcg
->memory
.emin
);
585 *low
= READ_ONCE(memcg
->memory
.elow
);
588 void mem_cgroup_calculate_protection(struct mem_cgroup
*root
,
589 struct mem_cgroup
*memcg
);
591 static inline bool mem_cgroup_unprotected(struct mem_cgroup
*target
,
592 struct mem_cgroup
*memcg
)
595 * The root memcg doesn't account charges, and doesn't support
596 * protection. The target memcg's protection is ignored, see
597 * mem_cgroup_calculate_protection() and mem_cgroup_protection()
599 return mem_cgroup_disabled() || mem_cgroup_is_root(memcg
) ||
603 static inline bool mem_cgroup_below_low(struct mem_cgroup
*target
,
604 struct mem_cgroup
*memcg
)
606 if (mem_cgroup_unprotected(target
, memcg
))
609 return READ_ONCE(memcg
->memory
.elow
) >=
610 page_counter_read(&memcg
->memory
);
613 static inline bool mem_cgroup_below_min(struct mem_cgroup
*target
,
614 struct mem_cgroup
*memcg
)
616 if (mem_cgroup_unprotected(target
, memcg
))
619 return READ_ONCE(memcg
->memory
.emin
) >=
620 page_counter_read(&memcg
->memory
);
623 void mem_cgroup_commit_charge(struct folio
*folio
, struct mem_cgroup
*memcg
);
625 int __mem_cgroup_charge(struct folio
*folio
, struct mm_struct
*mm
, gfp_t gfp
);
628 * mem_cgroup_charge - Charge a newly allocated folio to a cgroup.
629 * @folio: Folio to charge.
630 * @mm: mm context of the allocating task.
631 * @gfp: Reclaim mode.
633 * Try to charge @folio to the memcg that @mm belongs to, reclaiming
634 * pages according to @gfp if necessary. If @mm is NULL, try to
635 * charge to the active memcg.
637 * Do not use this for folios allocated for swapin.
639 * Return: 0 on success. Otherwise, an error code is returned.
641 static inline int mem_cgroup_charge(struct folio
*folio
, struct mm_struct
*mm
,
644 if (mem_cgroup_disabled())
646 return __mem_cgroup_charge(folio
, mm
, gfp
);
649 int mem_cgroup_hugetlb_try_charge(struct mem_cgroup
*memcg
, gfp_t gfp
,
652 int mem_cgroup_swapin_charge_folio(struct folio
*folio
, struct mm_struct
*mm
,
653 gfp_t gfp
, swp_entry_t entry
);
655 void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry
, unsigned int nr_pages
);
657 void __mem_cgroup_uncharge(struct folio
*folio
);
660 * mem_cgroup_uncharge - Uncharge a folio.
661 * @folio: Folio to uncharge.
663 * Uncharge a folio previously charged with mem_cgroup_charge().
665 static inline void mem_cgroup_uncharge(struct folio
*folio
)
667 if (mem_cgroup_disabled())
669 __mem_cgroup_uncharge(folio
);
672 void __mem_cgroup_uncharge_folios(struct folio_batch
*folios
);
673 static inline void mem_cgroup_uncharge_folios(struct folio_batch
*folios
)
675 if (mem_cgroup_disabled())
677 __mem_cgroup_uncharge_folios(folios
);
680 void mem_cgroup_cancel_charge(struct mem_cgroup
*memcg
, unsigned int nr_pages
);
681 void mem_cgroup_replace_folio(struct folio
*old
, struct folio
*new);
682 void mem_cgroup_migrate(struct folio
*old
, struct folio
*new);
685 * mem_cgroup_lruvec - get the lru list vector for a memcg & node
686 * @memcg: memcg of the wanted lruvec
687 * @pgdat: pglist_data
689 * Returns the lru list vector holding pages for a given @memcg &
690 * @pgdat combination. This can be the node lruvec, if the memory
691 * controller is disabled.
693 static inline struct lruvec
*mem_cgroup_lruvec(struct mem_cgroup
*memcg
,
694 struct pglist_data
*pgdat
)
696 struct mem_cgroup_per_node
*mz
;
697 struct lruvec
*lruvec
;
699 if (mem_cgroup_disabled()) {
700 lruvec
= &pgdat
->__lruvec
;
705 memcg
= root_mem_cgroup
;
707 mz
= memcg
->nodeinfo
[pgdat
->node_id
];
708 lruvec
= &mz
->lruvec
;
711 * Since a node can be onlined after the mem_cgroup was created,
712 * we have to be prepared to initialize lruvec->pgdat here;
713 * and if offlined then reonlined, we need to reinitialize it.
715 if (unlikely(lruvec
->pgdat
!= pgdat
))
716 lruvec
->pgdat
= pgdat
;
721 * folio_lruvec - return lruvec for isolating/putting an LRU folio
722 * @folio: Pointer to the folio.
724 * This function relies on folio->mem_cgroup being stable.
726 static inline struct lruvec
*folio_lruvec(struct folio
*folio
)
728 struct mem_cgroup
*memcg
= folio_memcg(folio
);
730 VM_WARN_ON_ONCE_FOLIO(!memcg
&& !mem_cgroup_disabled(), folio
);
731 return mem_cgroup_lruvec(memcg
, folio_pgdat(folio
));
734 struct mem_cgroup
*mem_cgroup_from_task(struct task_struct
*p
);
736 struct mem_cgroup
*get_mem_cgroup_from_mm(struct mm_struct
*mm
);
738 struct mem_cgroup
*get_mem_cgroup_from_current(void);
740 struct mem_cgroup
*get_mem_cgroup_from_folio(struct folio
*folio
);
742 struct lruvec
*folio_lruvec_lock(struct folio
*folio
);
743 struct lruvec
*folio_lruvec_lock_irq(struct folio
*folio
);
744 struct lruvec
*folio_lruvec_lock_irqsave(struct folio
*folio
,
745 unsigned long *flags
);
747 #ifdef CONFIG_DEBUG_VM
748 void lruvec_memcg_debug(struct lruvec
*lruvec
, struct folio
*folio
);
751 void lruvec_memcg_debug(struct lruvec
*lruvec
, struct folio
*folio
)
757 struct mem_cgroup
*mem_cgroup_from_css(struct cgroup_subsys_state
*css
){
758 return css
? container_of(css
, struct mem_cgroup
, css
) : NULL
;
761 static inline bool obj_cgroup_tryget(struct obj_cgroup
*objcg
)
763 return percpu_ref_tryget(&objcg
->refcnt
);
766 static inline void obj_cgroup_get(struct obj_cgroup
*objcg
)
768 percpu_ref_get(&objcg
->refcnt
);
771 static inline void obj_cgroup_get_many(struct obj_cgroup
*objcg
,
774 percpu_ref_get_many(&objcg
->refcnt
, nr
);
777 static inline void obj_cgroup_put(struct obj_cgroup
*objcg
)
780 percpu_ref_put(&objcg
->refcnt
);
783 static inline bool mem_cgroup_tryget(struct mem_cgroup
*memcg
)
785 return !memcg
|| css_tryget(&memcg
->css
);
788 static inline bool mem_cgroup_tryget_online(struct mem_cgroup
*memcg
)
790 return !memcg
|| css_tryget_online(&memcg
->css
);
793 static inline void mem_cgroup_put(struct mem_cgroup
*memcg
)
796 css_put(&memcg
->css
);
799 #define mem_cgroup_from_counter(counter, member) \
800 container_of(counter, struct mem_cgroup, member)
802 struct mem_cgroup
*mem_cgroup_iter(struct mem_cgroup
*,
804 struct mem_cgroup_reclaim_cookie
*);
805 void mem_cgroup_iter_break(struct mem_cgroup
*, struct mem_cgroup
*);
806 void mem_cgroup_scan_tasks(struct mem_cgroup
*memcg
,
807 int (*)(struct task_struct
*, void *), void *arg
);
809 static inline unsigned short mem_cgroup_id(struct mem_cgroup
*memcg
)
811 if (mem_cgroup_disabled())
816 struct mem_cgroup
*mem_cgroup_from_id(unsigned short id
);
818 #ifdef CONFIG_SHRINKER_DEBUG
819 static inline unsigned long mem_cgroup_ino(struct mem_cgroup
*memcg
)
821 return memcg
? cgroup_ino(memcg
->css
.cgroup
) : 0;
824 struct mem_cgroup
*mem_cgroup_get_from_ino(unsigned long ino
);
827 static inline struct mem_cgroup
*mem_cgroup_from_seq(struct seq_file
*m
)
829 return mem_cgroup_from_css(seq_css(m
));
832 static inline struct mem_cgroup
*lruvec_memcg(struct lruvec
*lruvec
)
834 struct mem_cgroup_per_node
*mz
;
836 if (mem_cgroup_disabled())
839 mz
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
844 * parent_mem_cgroup - find the accounting parent of a memcg
845 * @memcg: memcg whose parent to find
847 * Returns the parent memcg, or NULL if this is the root.
849 static inline struct mem_cgroup
*parent_mem_cgroup(struct mem_cgroup
*memcg
)
851 return mem_cgroup_from_css(memcg
->css
.parent
);
854 static inline bool mem_cgroup_is_descendant(struct mem_cgroup
*memcg
,
855 struct mem_cgroup
*root
)
859 return cgroup_is_descendant(memcg
->css
.cgroup
, root
->css
.cgroup
);
862 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
863 struct mem_cgroup
*memcg
)
865 struct mem_cgroup
*task_memcg
;
869 task_memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
871 match
= mem_cgroup_is_descendant(task_memcg
, memcg
);
876 struct cgroup_subsys_state
*mem_cgroup_css_from_folio(struct folio
*folio
);
877 ino_t
page_cgroup_ino(struct page
*page
);
879 static inline bool mem_cgroup_online(struct mem_cgroup
*memcg
)
881 if (mem_cgroup_disabled())
883 return !!(memcg
->css
.flags
& CSS_ONLINE
);
886 void mem_cgroup_update_lru_size(struct lruvec
*lruvec
, enum lru_list lru
,
887 int zid
, int nr_pages
);
890 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec
*lruvec
,
891 enum lru_list lru
, int zone_idx
)
893 struct mem_cgroup_per_node
*mz
;
895 mz
= container_of(lruvec
, struct mem_cgroup_per_node
, lruvec
);
896 return READ_ONCE(mz
->lru_zone_size
[zone_idx
][lru
]);
899 void mem_cgroup_handle_over_high(gfp_t gfp_mask
);
901 unsigned long mem_cgroup_get_max(struct mem_cgroup
*memcg
);
903 unsigned long mem_cgroup_size(struct mem_cgroup
*memcg
);
905 void mem_cgroup_print_oom_context(struct mem_cgroup
*memcg
,
906 struct task_struct
*p
);
908 void mem_cgroup_print_oom_meminfo(struct mem_cgroup
*memcg
);
910 struct mem_cgroup
*mem_cgroup_get_oom_group(struct task_struct
*victim
,
911 struct mem_cgroup
*oom_domain
);
912 void mem_cgroup_print_oom_group(struct mem_cgroup
*memcg
);
914 void __mod_memcg_state(struct mem_cgroup
*memcg
, enum memcg_stat_item idx
,
917 /* idx can be of type enum memcg_stat_item or node_stat_item */
918 static inline void mod_memcg_state(struct mem_cgroup
*memcg
,
919 enum memcg_stat_item idx
, int val
)
923 local_irq_save(flags
);
924 __mod_memcg_state(memcg
, idx
, val
);
925 local_irq_restore(flags
);
928 static inline void mod_memcg_page_state(struct page
*page
,
929 enum memcg_stat_item idx
, int val
)
931 struct mem_cgroup
*memcg
;
933 if (mem_cgroup_disabled())
937 memcg
= folio_memcg(page_folio(page
));
939 mod_memcg_state(memcg
, idx
, val
);
943 unsigned long memcg_page_state(struct mem_cgroup
*memcg
, int idx
);
944 unsigned long lruvec_page_state(struct lruvec
*lruvec
, enum node_stat_item idx
);
945 unsigned long lruvec_page_state_local(struct lruvec
*lruvec
,
946 enum node_stat_item idx
);
948 void mem_cgroup_flush_stats(struct mem_cgroup
*memcg
);
949 void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup
*memcg
);
951 void __mod_lruvec_kmem_state(void *p
, enum node_stat_item idx
, int val
);
953 static inline void mod_lruvec_kmem_state(void *p
, enum node_stat_item idx
,
958 local_irq_save(flags
);
959 __mod_lruvec_kmem_state(p
, idx
, val
);
960 local_irq_restore(flags
);
963 void __count_memcg_events(struct mem_cgroup
*memcg
, enum vm_event_item idx
,
964 unsigned long count
);
966 static inline void count_memcg_events(struct mem_cgroup
*memcg
,
967 enum vm_event_item idx
,
972 local_irq_save(flags
);
973 __count_memcg_events(memcg
, idx
, count
);
974 local_irq_restore(flags
);
977 static inline void count_memcg_folio_events(struct folio
*folio
,
978 enum vm_event_item idx
, unsigned long nr
)
980 struct mem_cgroup
*memcg
= folio_memcg(folio
);
983 count_memcg_events(memcg
, idx
, nr
);
986 static inline void count_memcg_events_mm(struct mm_struct
*mm
,
987 enum vm_event_item idx
, unsigned long count
)
989 struct mem_cgroup
*memcg
;
991 if (mem_cgroup_disabled())
995 memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
997 count_memcg_events(memcg
, idx
, count
);
1001 static inline void count_memcg_event_mm(struct mm_struct
*mm
,
1002 enum vm_event_item idx
)
1004 count_memcg_events_mm(mm
, idx
, 1);
1007 static inline void memcg_memory_event(struct mem_cgroup
*memcg
,
1008 enum memcg_memory_event event
)
1010 bool swap_event
= event
== MEMCG_SWAP_HIGH
|| event
== MEMCG_SWAP_MAX
||
1011 event
== MEMCG_SWAP_FAIL
;
1013 atomic_long_inc(&memcg
->memory_events_local
[event
]);
1015 cgroup_file_notify(&memcg
->events_local_file
);
1018 atomic_long_inc(&memcg
->memory_events
[event
]);
1020 cgroup_file_notify(&memcg
->swap_events_file
);
1022 cgroup_file_notify(&memcg
->events_file
);
1024 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys
))
1026 if (cgrp_dfl_root
.flags
& CGRP_ROOT_MEMORY_LOCAL_EVENTS
)
1028 } while ((memcg
= parent_mem_cgroup(memcg
)) &&
1029 !mem_cgroup_is_root(memcg
));
1032 static inline void memcg_memory_event_mm(struct mm_struct
*mm
,
1033 enum memcg_memory_event event
)
1035 struct mem_cgroup
*memcg
;
1037 if (mem_cgroup_disabled())
1041 memcg
= mem_cgroup_from_task(rcu_dereference(mm
->owner
));
1043 memcg_memory_event(memcg
, event
);
1047 void split_page_memcg(struct page
*head
, int old_order
, int new_order
);
1049 #else /* CONFIG_MEMCG */
1051 #define MEM_CGROUP_ID_SHIFT 0
1053 static inline struct mem_cgroup
*folio_memcg(struct folio
*folio
)
1058 static inline bool folio_memcg_charged(struct folio
*folio
)
1063 static inline struct mem_cgroup
*folio_memcg_check(struct folio
*folio
)
1068 static inline struct mem_cgroup
*page_memcg_check(struct page
*page
)
1073 static inline struct mem_cgroup
*get_mem_cgroup_from_objcg(struct obj_cgroup
*objcg
)
1078 static inline bool folio_memcg_kmem(struct folio
*folio
)
1083 static inline bool PageMemcgKmem(struct page
*page
)
1088 static inline bool mem_cgroup_is_root(struct mem_cgroup
*memcg
)
1093 static inline bool mem_cgroup_disabled(void)
1098 static inline void memcg_memory_event(struct mem_cgroup
*memcg
,
1099 enum memcg_memory_event event
)
1103 static inline void memcg_memory_event_mm(struct mm_struct
*mm
,
1104 enum memcg_memory_event event
)
1108 static inline void mem_cgroup_protection(struct mem_cgroup
*root
,
1109 struct mem_cgroup
*memcg
,
1116 static inline void mem_cgroup_calculate_protection(struct mem_cgroup
*root
,
1117 struct mem_cgroup
*memcg
)
1121 static inline bool mem_cgroup_unprotected(struct mem_cgroup
*target
,
1122 struct mem_cgroup
*memcg
)
1126 static inline bool mem_cgroup_below_low(struct mem_cgroup
*target
,
1127 struct mem_cgroup
*memcg
)
1132 static inline bool mem_cgroup_below_min(struct mem_cgroup
*target
,
1133 struct mem_cgroup
*memcg
)
1138 static inline void mem_cgroup_commit_charge(struct folio
*folio
,
1139 struct mem_cgroup
*memcg
)
1143 static inline int mem_cgroup_charge(struct folio
*folio
,
1144 struct mm_struct
*mm
, gfp_t gfp
)
1149 static inline int mem_cgroup_hugetlb_try_charge(struct mem_cgroup
*memcg
,
1150 gfp_t gfp
, long nr_pages
)
1155 static inline int mem_cgroup_swapin_charge_folio(struct folio
*folio
,
1156 struct mm_struct
*mm
, gfp_t gfp
, swp_entry_t entry
)
1161 static inline void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry
, unsigned int nr
)
1165 static inline void mem_cgroup_uncharge(struct folio
*folio
)
1169 static inline void mem_cgroup_uncharge_folios(struct folio_batch
*folios
)
1173 static inline void mem_cgroup_cancel_charge(struct mem_cgroup
*memcg
,
1174 unsigned int nr_pages
)
1178 static inline void mem_cgroup_replace_folio(struct folio
*old
,
1183 static inline void mem_cgroup_migrate(struct folio
*old
, struct folio
*new)
1187 static inline struct lruvec
*mem_cgroup_lruvec(struct mem_cgroup
*memcg
,
1188 struct pglist_data
*pgdat
)
1190 return &pgdat
->__lruvec
;
1193 static inline struct lruvec
*folio_lruvec(struct folio
*folio
)
1195 struct pglist_data
*pgdat
= folio_pgdat(folio
);
1196 return &pgdat
->__lruvec
;
1200 void lruvec_memcg_debug(struct lruvec
*lruvec
, struct folio
*folio
)
1204 static inline struct mem_cgroup
*parent_mem_cgroup(struct mem_cgroup
*memcg
)
1209 static inline bool mm_match_cgroup(struct mm_struct
*mm
,
1210 struct mem_cgroup
*memcg
)
1215 static inline struct mem_cgroup
*get_mem_cgroup_from_mm(struct mm_struct
*mm
)
1220 static inline struct mem_cgroup
*get_mem_cgroup_from_current(void)
1225 static inline struct mem_cgroup
*get_mem_cgroup_from_folio(struct folio
*folio
)
1231 struct mem_cgroup
*mem_cgroup_from_css(struct cgroup_subsys_state
*css
)
1236 static inline void obj_cgroup_get(struct obj_cgroup
*objcg
)
1240 static inline void obj_cgroup_put(struct obj_cgroup
*objcg
)
1244 static inline bool mem_cgroup_tryget(struct mem_cgroup
*memcg
)
1249 static inline bool mem_cgroup_tryget_online(struct mem_cgroup
*memcg
)
1254 static inline void mem_cgroup_put(struct mem_cgroup
*memcg
)
1258 static inline struct lruvec
*folio_lruvec_lock(struct folio
*folio
)
1260 struct pglist_data
*pgdat
= folio_pgdat(folio
);
1262 spin_lock(&pgdat
->__lruvec
.lru_lock
);
1263 return &pgdat
->__lruvec
;
1266 static inline struct lruvec
*folio_lruvec_lock_irq(struct folio
*folio
)
1268 struct pglist_data
*pgdat
= folio_pgdat(folio
);
1270 spin_lock_irq(&pgdat
->__lruvec
.lru_lock
);
1271 return &pgdat
->__lruvec
;
1274 static inline struct lruvec
*folio_lruvec_lock_irqsave(struct folio
*folio
,
1275 unsigned long *flagsp
)
1277 struct pglist_data
*pgdat
= folio_pgdat(folio
);
1279 spin_lock_irqsave(&pgdat
->__lruvec
.lru_lock
, *flagsp
);
1280 return &pgdat
->__lruvec
;
1283 static inline struct mem_cgroup
*
1284 mem_cgroup_iter(struct mem_cgroup
*root
,
1285 struct mem_cgroup
*prev
,
1286 struct mem_cgroup_reclaim_cookie
*reclaim
)
1291 static inline void mem_cgroup_iter_break(struct mem_cgroup
*root
,
1292 struct mem_cgroup
*prev
)
1296 static inline void mem_cgroup_scan_tasks(struct mem_cgroup
*memcg
,
1297 int (*fn
)(struct task_struct
*, void *), void *arg
)
1301 static inline unsigned short mem_cgroup_id(struct mem_cgroup
*memcg
)
1306 static inline struct mem_cgroup
*mem_cgroup_from_id(unsigned short id
)
1309 /* XXX: This should always return root_mem_cgroup */
1313 #ifdef CONFIG_SHRINKER_DEBUG
1314 static inline unsigned long mem_cgroup_ino(struct mem_cgroup
*memcg
)
1319 static inline struct mem_cgroup
*mem_cgroup_get_from_ino(unsigned long ino
)
1325 static inline struct mem_cgroup
*mem_cgroup_from_seq(struct seq_file
*m
)
1330 static inline struct mem_cgroup
*lruvec_memcg(struct lruvec
*lruvec
)
1335 static inline bool mem_cgroup_online(struct mem_cgroup
*memcg
)
1341 unsigned long mem_cgroup_get_zone_lru_size(struct lruvec
*lruvec
,
1342 enum lru_list lru
, int zone_idx
)
1347 static inline unsigned long mem_cgroup_get_max(struct mem_cgroup
*memcg
)
1352 static inline unsigned long mem_cgroup_size(struct mem_cgroup
*memcg
)
1358 mem_cgroup_print_oom_context(struct mem_cgroup
*memcg
, struct task_struct
*p
)
1363 mem_cgroup_print_oom_meminfo(struct mem_cgroup
*memcg
)
1367 static inline void mem_cgroup_handle_over_high(gfp_t gfp_mask
)
1371 static inline struct mem_cgroup
*mem_cgroup_get_oom_group(
1372 struct task_struct
*victim
, struct mem_cgroup
*oom_domain
)
1377 static inline void mem_cgroup_print_oom_group(struct mem_cgroup
*memcg
)
1381 static inline void __mod_memcg_state(struct mem_cgroup
*memcg
,
1382 enum memcg_stat_item idx
,
1387 static inline void mod_memcg_state(struct mem_cgroup
*memcg
,
1388 enum memcg_stat_item idx
,
1393 static inline void mod_memcg_page_state(struct page
*page
,
1394 enum memcg_stat_item idx
, int val
)
1398 static inline unsigned long memcg_page_state(struct mem_cgroup
*memcg
, int idx
)
1403 static inline unsigned long lruvec_page_state(struct lruvec
*lruvec
,
1404 enum node_stat_item idx
)
1406 return node_page_state(lruvec_pgdat(lruvec
), idx
);
1409 static inline unsigned long lruvec_page_state_local(struct lruvec
*lruvec
,
1410 enum node_stat_item idx
)
1412 return node_page_state(lruvec_pgdat(lruvec
), idx
);
1415 static inline void mem_cgroup_flush_stats(struct mem_cgroup
*memcg
)
1419 static inline void mem_cgroup_flush_stats_ratelimited(struct mem_cgroup
*memcg
)
1423 static inline void __mod_lruvec_kmem_state(void *p
, enum node_stat_item idx
,
1426 struct page
*page
= virt_to_head_page(p
);
1428 __mod_node_page_state(page_pgdat(page
), idx
, val
);
1431 static inline void mod_lruvec_kmem_state(void *p
, enum node_stat_item idx
,
1434 struct page
*page
= virt_to_head_page(p
);
1436 mod_node_page_state(page_pgdat(page
), idx
, val
);
1439 static inline void count_memcg_events(struct mem_cgroup
*memcg
,
1440 enum vm_event_item idx
,
1441 unsigned long count
)
1445 static inline void __count_memcg_events(struct mem_cgroup
*memcg
,
1446 enum vm_event_item idx
,
1447 unsigned long count
)
1451 static inline void count_memcg_folio_events(struct folio
*folio
,
1452 enum vm_event_item idx
, unsigned long nr
)
1456 static inline void count_memcg_events_mm(struct mm_struct
*mm
,
1457 enum vm_event_item idx
, unsigned long count
)
1462 void count_memcg_event_mm(struct mm_struct
*mm
, enum vm_event_item idx
)
1466 static inline void split_page_memcg(struct page
*head
, int old_order
, int new_order
)
1469 #endif /* CONFIG_MEMCG */
1472 * Extended information for slab objects stored as an array in page->memcg_data
1473 * if MEMCG_DATA_OBJEXTS is set.
1475 struct slabobj_ext
{
1477 struct obj_cgroup
*objcg
;
1479 #ifdef CONFIG_MEM_ALLOC_PROFILING
1480 union codetag_ref ref
;
1484 static inline void __inc_lruvec_kmem_state(void *p
, enum node_stat_item idx
)
1486 __mod_lruvec_kmem_state(p
, idx
, 1);
1489 static inline void __dec_lruvec_kmem_state(void *p
, enum node_stat_item idx
)
1491 __mod_lruvec_kmem_state(p
, idx
, -1);
1494 static inline struct lruvec
*parent_lruvec(struct lruvec
*lruvec
)
1496 struct mem_cgroup
*memcg
;
1498 memcg
= lruvec_memcg(lruvec
);
1501 memcg
= parent_mem_cgroup(memcg
);
1504 return mem_cgroup_lruvec(memcg
, lruvec_pgdat(lruvec
));
1507 static inline void unlock_page_lruvec(struct lruvec
*lruvec
)
1509 spin_unlock(&lruvec
->lru_lock
);
1512 static inline void unlock_page_lruvec_irq(struct lruvec
*lruvec
)
1514 spin_unlock_irq(&lruvec
->lru_lock
);
1517 static inline void unlock_page_lruvec_irqrestore(struct lruvec
*lruvec
,
1518 unsigned long flags
)
1520 spin_unlock_irqrestore(&lruvec
->lru_lock
, flags
);
1523 /* Test requires a stable folio->memcg binding, see folio_memcg() */
1524 static inline bool folio_matches_lruvec(struct folio
*folio
,
1525 struct lruvec
*lruvec
)
1527 return lruvec_pgdat(lruvec
) == folio_pgdat(folio
) &&
1528 lruvec_memcg(lruvec
) == folio_memcg(folio
);
1531 /* Don't lock again iff page's lruvec locked */
1532 static inline struct lruvec
*folio_lruvec_relock_irq(struct folio
*folio
,
1533 struct lruvec
*locked_lruvec
)
1535 if (locked_lruvec
) {
1536 if (folio_matches_lruvec(folio
, locked_lruvec
))
1537 return locked_lruvec
;
1539 unlock_page_lruvec_irq(locked_lruvec
);
1542 return folio_lruvec_lock_irq(folio
);
1545 /* Don't lock again iff folio's lruvec locked */
1546 static inline void folio_lruvec_relock_irqsave(struct folio
*folio
,
1547 struct lruvec
**lruvecp
, unsigned long *flags
)
1550 if (folio_matches_lruvec(folio
, *lruvecp
))
1553 unlock_page_lruvec_irqrestore(*lruvecp
, *flags
);
1556 *lruvecp
= folio_lruvec_lock_irqsave(folio
, flags
);
1559 #ifdef CONFIG_CGROUP_WRITEBACK
1561 struct wb_domain
*mem_cgroup_wb_domain(struct bdi_writeback
*wb
);
1562 void mem_cgroup_wb_stats(struct bdi_writeback
*wb
, unsigned long *pfilepages
,
1563 unsigned long *pheadroom
, unsigned long *pdirty
,
1564 unsigned long *pwriteback
);
1566 void mem_cgroup_track_foreign_dirty_slowpath(struct folio
*folio
,
1567 struct bdi_writeback
*wb
);
1569 static inline void mem_cgroup_track_foreign_dirty(struct folio
*folio
,
1570 struct bdi_writeback
*wb
)
1572 struct mem_cgroup
*memcg
;
1574 if (mem_cgroup_disabled())
1577 memcg
= folio_memcg(folio
);
1578 if (unlikely(memcg
&& &memcg
->css
!= wb
->memcg_css
))
1579 mem_cgroup_track_foreign_dirty_slowpath(folio
, wb
);
1582 void mem_cgroup_flush_foreign(struct bdi_writeback
*wb
);
1584 #else /* CONFIG_CGROUP_WRITEBACK */
1586 static inline struct wb_domain
*mem_cgroup_wb_domain(struct bdi_writeback
*wb
)
1591 static inline void mem_cgroup_wb_stats(struct bdi_writeback
*wb
,
1592 unsigned long *pfilepages
,
1593 unsigned long *pheadroom
,
1594 unsigned long *pdirty
,
1595 unsigned long *pwriteback
)
1599 static inline void mem_cgroup_track_foreign_dirty(struct folio
*folio
,
1600 struct bdi_writeback
*wb
)
1604 static inline void mem_cgroup_flush_foreign(struct bdi_writeback
*wb
)
1608 #endif /* CONFIG_CGROUP_WRITEBACK */
1611 bool mem_cgroup_charge_skmem(struct mem_cgroup
*memcg
, unsigned int nr_pages
,
1613 void mem_cgroup_uncharge_skmem(struct mem_cgroup
*memcg
, unsigned int nr_pages
);
1615 extern struct static_key_false memcg_sockets_enabled_key
;
1616 #define mem_cgroup_sockets_enabled static_branch_unlikely(&memcg_sockets_enabled_key)
1617 void mem_cgroup_sk_alloc(struct sock
*sk
);
1618 void mem_cgroup_sk_free(struct sock
*sk
);
1619 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup
*memcg
)
1621 #ifdef CONFIG_MEMCG_V1
1622 if (!cgroup_subsys_on_dfl(memory_cgrp_subsys
))
1623 return !!memcg
->tcpmem_pressure
;
1624 #endif /* CONFIG_MEMCG_V1 */
1626 if (time_before(jiffies
, READ_ONCE(memcg
->socket_pressure
)))
1628 } while ((memcg
= parent_mem_cgroup(memcg
)));
1632 int alloc_shrinker_info(struct mem_cgroup
*memcg
);
1633 void free_shrinker_info(struct mem_cgroup
*memcg
);
1634 void set_shrinker_bit(struct mem_cgroup
*memcg
, int nid
, int shrinker_id
);
1635 void reparent_shrinker_deferred(struct mem_cgroup
*memcg
);
1637 #define mem_cgroup_sockets_enabled 0
1638 static inline void mem_cgroup_sk_alloc(struct sock
*sk
) { };
1639 static inline void mem_cgroup_sk_free(struct sock
*sk
) { };
1640 static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup
*memcg
)
1645 static inline void set_shrinker_bit(struct mem_cgroup
*memcg
,
1646 int nid
, int shrinker_id
)
1652 bool mem_cgroup_kmem_disabled(void);
1653 int __memcg_kmem_charge_page(struct page
*page
, gfp_t gfp
, int order
);
1654 void __memcg_kmem_uncharge_page(struct page
*page
, int order
);
1657 * The returned objcg pointer is safe to use without additional
1658 * protection within a scope. The scope is defined either by
1659 * the current task (similar to the "current" global variable)
1660 * or by set_active_memcg() pair.
1661 * Please, use obj_cgroup_get() to get a reference if the pointer
1662 * needs to be used outside of the local scope.
1664 struct obj_cgroup
*current_obj_cgroup(void);
1665 struct obj_cgroup
*get_obj_cgroup_from_folio(struct folio
*folio
);
1667 static inline struct obj_cgroup
*get_obj_cgroup_from_current(void)
1669 struct obj_cgroup
*objcg
= current_obj_cgroup();
1672 obj_cgroup_get(objcg
);
1677 int obj_cgroup_charge(struct obj_cgroup
*objcg
, gfp_t gfp
, size_t size
);
1678 void obj_cgroup_uncharge(struct obj_cgroup
*objcg
, size_t size
);
1680 extern struct static_key_false memcg_bpf_enabled_key
;
1681 static inline bool memcg_bpf_enabled(void)
1683 return static_branch_likely(&memcg_bpf_enabled_key
);
1686 extern struct static_key_false memcg_kmem_online_key
;
1688 static inline bool memcg_kmem_online(void)
1690 return static_branch_likely(&memcg_kmem_online_key
);
1693 static inline int memcg_kmem_charge_page(struct page
*page
, gfp_t gfp
,
1696 if (memcg_kmem_online())
1697 return __memcg_kmem_charge_page(page
, gfp
, order
);
1701 static inline void memcg_kmem_uncharge_page(struct page
*page
, int order
)
1703 if (memcg_kmem_online())
1704 __memcg_kmem_uncharge_page(page
, order
);
1708 * A helper for accessing memcg's kmem_id, used for getting
1709 * corresponding LRU lists.
1711 static inline int memcg_kmem_id(struct mem_cgroup
*memcg
)
1713 return memcg
? memcg
->kmemcg_id
: -1;
1716 struct mem_cgroup
*mem_cgroup_from_slab_obj(void *p
);
1718 static inline void count_objcg_events(struct obj_cgroup
*objcg
,
1719 enum vm_event_item idx
,
1720 unsigned long count
)
1722 struct mem_cgroup
*memcg
;
1724 if (!memcg_kmem_online())
1728 memcg
= obj_cgroup_memcg(objcg
);
1729 count_memcg_events(memcg
, idx
, count
);
1734 static inline bool mem_cgroup_kmem_disabled(void)
1739 static inline int memcg_kmem_charge_page(struct page
*page
, gfp_t gfp
,
1745 static inline void memcg_kmem_uncharge_page(struct page
*page
, int order
)
1749 static inline int __memcg_kmem_charge_page(struct page
*page
, gfp_t gfp
,
1755 static inline void __memcg_kmem_uncharge_page(struct page
*page
, int order
)
1759 static inline struct obj_cgroup
*get_obj_cgroup_from_folio(struct folio
*folio
)
1764 static inline bool memcg_bpf_enabled(void)
1769 static inline bool memcg_kmem_online(void)
1774 static inline int memcg_kmem_id(struct mem_cgroup
*memcg
)
1779 static inline struct mem_cgroup
*mem_cgroup_from_slab_obj(void *p
)
1784 static inline void count_objcg_events(struct obj_cgroup
*objcg
,
1785 enum vm_event_item idx
,
1786 unsigned long count
)
1790 #endif /* CONFIG_MEMCG */
1792 #if defined(CONFIG_MEMCG) && defined(CONFIG_ZSWAP)
1793 bool obj_cgroup_may_zswap(struct obj_cgroup
*objcg
);
1794 void obj_cgroup_charge_zswap(struct obj_cgroup
*objcg
, size_t size
);
1795 void obj_cgroup_uncharge_zswap(struct obj_cgroup
*objcg
, size_t size
);
1796 bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup
*memcg
);
1798 static inline bool obj_cgroup_may_zswap(struct obj_cgroup
*objcg
)
1802 static inline void obj_cgroup_charge_zswap(struct obj_cgroup
*objcg
,
1806 static inline void obj_cgroup_uncharge_zswap(struct obj_cgroup
*objcg
,
1810 static inline bool mem_cgroup_zswap_writeback_enabled(struct mem_cgroup
*memcg
)
1812 /* if zswap is disabled, do not block pages going to the swapping device */
1818 /* Cgroup v1-related declarations */
1820 #ifdef CONFIG_MEMCG_V1
1821 unsigned long memcg1_soft_limit_reclaim(pg_data_t
*pgdat
, int order
,
1823 unsigned long *total_scanned
);
1825 bool mem_cgroup_oom_synchronize(bool wait
);
1827 static inline bool task_in_memcg_oom(struct task_struct
*p
)
1829 return p
->memcg_in_oom
;
1832 static inline void mem_cgroup_enter_user_fault(void)
1834 WARN_ON(current
->in_user_fault
);
1835 current
->in_user_fault
= 1;
1838 static inline void mem_cgroup_exit_user_fault(void)
1840 WARN_ON(!current
->in_user_fault
);
1841 current
->in_user_fault
= 0;
1844 #else /* CONFIG_MEMCG_V1 */
1846 unsigned long memcg1_soft_limit_reclaim(pg_data_t
*pgdat
, int order
,
1848 unsigned long *total_scanned
)
1853 static inline bool task_in_memcg_oom(struct task_struct
*p
)
1858 static inline bool mem_cgroup_oom_synchronize(bool wait
)
1863 static inline void mem_cgroup_enter_user_fault(void)
1867 static inline void mem_cgroup_exit_user_fault(void)
1871 #endif /* CONFIG_MEMCG_V1 */
1873 #endif /* _LINUX_MEMCONTROL_H */