1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SWAP_H
3 #define _LINUX_SWAP_H
4
5 #include <linux/spinlock.h>
6 #include <linux/linkage.h>
7 #include <linux/mmzone.h>
8 #include <linux/list.h>
9 #include <linux/memcontrol.h>
10 #include <linux/sched.h>
11 #include <linux/node.h>
12 #include <linux/fs.h>
13 #include <linux/pagemap.h>
14 #include <linux/atomic.h>
15 #include <linux/page-flags.h>
16 #include <uapi/linux/mempolicy.h>
17 #include <asm/page.h>
18
19 struct notifier_block;
20
21 struct bio;
22
23 struct pagevec;
24
25 #define SWAP_FLAG_PREFER 0x8000 /* set if swap priority specified */
26 #define SWAP_FLAG_PRIO_MASK 0x7fff
27 #define SWAP_FLAG_PRIO_SHIFT 0
28 #define SWAP_FLAG_DISCARD 0x10000 /* enable discard for swap */
29 #define SWAP_FLAG_DISCARD_ONCE 0x20000 /* discard swap area at swapon-time */
30 #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */
31
32 #define SWAP_FLAGS_VALID (SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \
33 SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \
34 SWAP_FLAG_DISCARD_PAGES)
35 #define SWAP_BATCH 64
36
37 static inline int current_is_kswapd(void)
38 {
39 return current->flags & PF_KSWAPD;
40 }
41
42 /*
43 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
44 * be swapped to. The swap type and the offset into that swap type are
45 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
46 * for the type means that the maximum number of swapcache pages is 27 bits
47 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
48 * the type/offset into the pte as 5/27 as well.
49 */
50 #define MAX_SWAPFILES_SHIFT 5
51
52 /*
53 * Use some of the swap files numbers for other purposes. This
54 * is a convenient way to hook into the VM to trigger special
55 * actions on faults.
56 */
57
58 /*
59 * PTE markers are used to persist information onto PTEs that otherwise
60 * should be a none pte. As its name "PTE" hints, it should only be
61 * applied to the leaves of pgtables.
62 */
63 #define SWP_PTE_MARKER_NUM 1
64 #define SWP_PTE_MARKER (MAX_SWAPFILES + SWP_HWPOISON_NUM + \
65 SWP_MIGRATION_NUM + SWP_DEVICE_NUM)
66
67 /*
68 * Unaddressable device memory support. See include/linux/hmm.h and
69 * Documentation/mm/hmm.rst. Short description is we need struct pages for
70 * device memory that is unaddressable (inaccessible) by CPU, so that we can
71 * migrate part of a process memory to device memory.
72 *
73 * When a page is migrated from CPU to device, we set the CPU page table entry
74 * to a special SWP_DEVICE_{READ|WRITE} entry.
75 *
76 * When a page is mapped by the device for exclusive access we set the CPU page
77 * table entries to special SWP_DEVICE_EXCLUSIVE_* entries.
78 */
79 #ifdef CONFIG_DEVICE_PRIVATE
80 #define SWP_DEVICE_NUM 4
81 #define SWP_DEVICE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM)
82 #define SWP_DEVICE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+1)
83 #define SWP_DEVICE_EXCLUSIVE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+2)
84 #define SWP_DEVICE_EXCLUSIVE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+3)
85 #else
86 #define SWP_DEVICE_NUM 0
87 #endif
88
89 /*
90 * Page migration support.
91 *
92 * SWP_MIGRATION_READ_EXCLUSIVE is only applicable to anonymous pages and
93 * indicates that the referenced (part of) an anonymous page is exclusive to
94 * a single process. For SWP_MIGRATION_WRITE, that information is implicit:
95 * (part of) an anonymous page that are mapped writable are exclusive to a
96 * single process.
97 */
98 #ifdef CONFIG_MIGRATION
99 #define SWP_MIGRATION_NUM 3
100 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
101 #define SWP_MIGRATION_READ_EXCLUSIVE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
102 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 2)
103 #else
104 #define SWP_MIGRATION_NUM 0
105 #endif
106
107 /*
108 * Handling of hardware poisoned pages with memory corruption.
109 */
110 #ifdef CONFIG_MEMORY_FAILURE
111 #define SWP_HWPOISON_NUM 1
112 #define SWP_HWPOISON MAX_SWAPFILES
113 #else
114 #define SWP_HWPOISON_NUM 0
115 #endif
116
117 #define MAX_SWAPFILES \
118 ((1 << MAX_SWAPFILES_SHIFT) - SWP_DEVICE_NUM - \
119 SWP_MIGRATION_NUM - SWP_HWPOISON_NUM - \
120 SWP_PTE_MARKER_NUM)
121
122 /*
123 * Magic header for a swap area. The first part of the union is
124 * what the swap magic looks like for the old (limited to 128MB)
125 * swap area format, the second part of the union adds - in the
126 * old reserved area - some extra information. Note that the first
127 * kilobyte is reserved for boot loader or disk label stuff...
128 *
129 * Having the magic at the end of the PAGE_SIZE makes detecting swap
130 * areas somewhat tricky on machines that support multiple page sizes.
131 * For 2.5 we'll probably want to move the magic to just beyond the
132 * bootbits...
133 */
134 union swap_header {
135 struct {
136 char reserved[PAGE_SIZE - 10];
137 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
138 } magic;
139 struct {
140 char bootbits[1024]; /* Space for disklabel etc. */
141 __u32 version;
142 __u32 last_page;
143 __u32 nr_badpages;
144 unsigned char sws_uuid[16];
145 unsigned char sws_volume[16];
146 __u32 padding[117];
147 __u32 badpages[1];
148 } info;
149 };
150
151 /*
152 * current->reclaim_state points to one of these when a task is running
153 * memory reclaim
154 */
155 struct reclaim_state {
156 /* pages reclaimed outside of LRU-based reclaim */
157 unsigned long reclaimed;
158 #ifdef CONFIG_LRU_GEN
159 /* per-thread mm walk data */
160 struct lru_gen_mm_walk *mm_walk;
161 #endif
162 };
163
164 /*
165 * mm_account_reclaimed_pages(): account reclaimed pages outside of LRU-based
166 * reclaim
167 * @pages: number of pages reclaimed
168 *
169 * If the current process is undergoing a reclaim operation, increment the
170 * number of reclaimed pages by @pages.
171 */
172 static inline void mm_account_reclaimed_pages(unsigned long pages)
173 {
174 if (current->reclaim_state)
175 current->reclaim_state->reclaimed += pages;
176 }
177
178 #ifdef __KERNEL__
179
180 struct address_space;
181 struct sysinfo;
182 struct writeback_control;
183 struct zone;
184
185 /*
186 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
187 * disk blocks. A rbtree of swap extents maps the entire swapfile (Where the
188 * term `swapfile' refers to either a blockdevice or an IS_REG file). Apart
189 * from setup, they're handled identically.
190 *
191 * We always assume that blocks are of size PAGE_SIZE.
192 */
193 struct swap_extent {
194 struct rb_node rb_node;
195 pgoff_t start_page;
196 pgoff_t nr_pages;
197 sector_t start_block;
198 };
199
200 /*
201 * Max bad pages in the new format..
202 */
203 #define MAX_SWAP_BADPAGES \
204 ((offsetof(union swap_header, magic.magic) - \
205 offsetof(union swap_header, info.badpages)) / sizeof(int))
206
207 enum {
208 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
209 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
210 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
211 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
212 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
213 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
214 SWP_BLKDEV = (1 << 6), /* its a block device */
215 SWP_ACTIVATED = (1 << 7), /* set after swap_activate success */
216 SWP_FS_OPS = (1 << 8), /* swapfile operations go through fs */
217 SWP_AREA_DISCARD = (1 << 9), /* single-time swap area discards */
218 SWP_PAGE_DISCARD = (1 << 10), /* freed swap page-cluster discards */
219 SWP_STABLE_WRITES = (1 << 11), /* no overwrite PG_writeback pages */
220 SWP_SYNCHRONOUS_IO = (1 << 12), /* synchronous IO is efficient */
221 /* add others here before... */
222 SWP_SCANNING = (1 << 14), /* refcount in scan_swap_map */
223 };
224
225 #define SWAP_CLUSTER_MAX 32UL
226 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
227
228 /* Bit flag in swap_map */
229 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
230 #define COUNT_CONTINUED 0x80 /* Flag swap_map continuation for full count */
231
232 /* Special value in first swap_map */
233 #define SWAP_MAP_MAX 0x3e /* Max count */
234 #define SWAP_MAP_BAD 0x3f /* Note page is bad */
235 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs */
236
237 /* Special value in each swap_map continuation */
238 #define SWAP_CONT_MAX 0x7f /* Max count */
239
240 /*
241 * We use this to track usage of a cluster. A cluster is a block of swap disk
242 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
243 * free clusters are organized into a list. We fetch an entry from the list to
244 * get a free cluster.
245 *
246 * The data field stores next cluster if the cluster is free or cluster usage
247 * counter otherwise. The flags field determines if a cluster is free. This is
248 * protected by swap_info_struct.lock.
249 */
250 struct swap_cluster_info {
251 spinlock_t lock; /*
252 * Protect swap_cluster_info fields
253 * and swap_info_struct->swap_map
254 * elements correspond to the swap
255 * cluster
256 */
257 unsigned int data:24;
258 unsigned int flags:8;
259 };
260 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
261 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
262
263 /*
264 * The first page in the swap file is the swap header, which is always marked
265 * bad to prevent it from being allocated as an entry. This also prevents the
266 * cluster to which it belongs being marked free. Therefore 0 is safe to use as
267 * a sentinel to indicate next is not valid in percpu_cluster.
268 */
269 #define SWAP_NEXT_INVALID 0
270
271 #ifdef CONFIG_THP_SWAP
272 #define SWAP_NR_ORDERS (PMD_ORDER + 1)
273 #else
274 #define SWAP_NR_ORDERS 1
275 #endif
276
277 /*
278 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
279 * its own cluster and swapout sequentially. The purpose is to optimize swapout
280 * throughput.
281 */
282 struct percpu_cluster {
283 unsigned int next[SWAP_NR_ORDERS]; /* Likely next allocation offset */
284 };
285
286 struct swap_cluster_list {
287 struct swap_cluster_info head;
288 struct swap_cluster_info tail;
289 };
290
291 /*
292 * The in-memory structure used to track swap areas.
293 */
294 struct swap_info_struct {
295 struct percpu_ref users; /* indicate and keep swap device valid. */
296 unsigned long flags; /* SWP_USED etc: see above */
297 signed short prio; /* swap priority of this type */
298 struct plist_node list; /* entry in swap_active_head */
299 signed char type; /* strange name for an index */
300 unsigned int max; /* extent of the swap_map */
301 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
302 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
303 struct swap_cluster_list free_clusters; /* free clusters list */
304 unsigned int lowest_bit; /* index of first free in swap_map */
305 unsigned int highest_bit; /* index of last free in swap_map */
306 unsigned int pages; /* total of usable pages of swap */
307 unsigned int inuse_pages; /* number of those currently in use */
308 unsigned int cluster_next; /* likely index for next allocation */
309 unsigned int cluster_nr; /* countdown to next cluster search */
310 unsigned int __percpu *cluster_next_cpu; /*percpu index for next allocation */
311 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
312 struct rb_root swap_extent_root;/* root of the swap extent rbtree */
313 struct block_device *bdev; /* swap device or bdev of swap file */
314 struct file *swap_file; /* seldom referenced */
315 struct completion comp; /* seldom referenced */
316 spinlock_t lock; /*
317 * protect map scan related fields like
318 * swap_map, lowest_bit, highest_bit,
319 * inuse_pages, cluster_next,
320 * cluster_nr, lowest_alloc,
321 * highest_alloc, free/discard cluster
322 * list. other fields are only changed
323 * at swapon/swapoff, so are protected
324 * by swap_lock. changing flags need
325 * hold this lock and swap_lock. If
326 * both locks need hold, hold swap_lock
327 * first.
328 */
329 spinlock_t cont_lock; /*
330 * protect swap count continuation page
331 * list.
332 */
333 struct work_struct discard_work; /* discard worker */
334 struct swap_cluster_list discard_clusters; /* discard clusters list */
335 struct plist_node avail_lists[]; /*
336 * entries in swap_avail_heads, one
337 * entry per node.
338 * Must be last as the number of the
339 * array is nr_node_ids, which is not
340 * a fixed value so have to allocate
341 * dynamically.
342 * And it has to be an array so that
343 * plist_for_each_* can work.
344 */
345 };
346
347 static inline swp_entry_t page_swap_entry(struct page *page)
348 {
349 struct folio *folio = page_folio(page);
350 swp_entry_t entry = folio->swap;
351
352 entry.val += folio_page_idx(folio, page);
353 return entry;
354 }
355
356 /* linux/mm/workingset.c */
357 bool workingset_test_recent(void *shadow, bool file, bool *workingset,
358 bool flush);
359 void workingset_age_nonresident(struct lruvec *lruvec, unsigned long nr_pages);
360 void *workingset_eviction(struct folio *folio, struct mem_cgroup *target_memcg);
361 void workingset_refault(struct folio *folio, void *shadow);
362 void workingset_activation(struct folio *folio);
363
364 /* linux/mm/page_alloc.c */
365 extern unsigned long totalreserve_pages;
366
367 /* Definition of global_zone_page_state not available yet */
368 #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES)
369
370
371 /* linux/mm/swap.c */
372 void lru_note_cost(struct lruvec *lruvec, bool file,
373 unsigned int nr_io, unsigned int nr_rotated);
374 void lru_note_cost_refault(struct folio *);
375 void folio_add_lru(struct folio *);
376 void folio_add_lru_vma(struct folio *, struct vm_area_struct *);
377 void mark_page_accessed(struct page *);
378 void folio_mark_accessed(struct folio *);
379
380 extern atomic_t lru_disable_count;
381
382 static inline bool lru_cache_disabled(void)
383 {
384 return atomic_read(&lru_disable_count);
385 }
386
387 static inline void lru_cache_enable(void)
388 {
389 atomic_dec(&lru_disable_count);
390 }
391
392 extern void lru_cache_disable(void);
393 extern void lru_add_drain(void);
394 extern void lru_add_drain_cpu(int cpu);
395 extern void lru_add_drain_cpu_zone(struct zone *zone);
396 extern void lru_add_drain_all(void);
397 void folio_deactivate(struct folio *folio);
398 void folio_mark_lazyfree(struct folio *folio);
399 extern void swap_setup(void);
400
401 /* linux/mm/vmscan.c */
402 extern unsigned long zone_reclaimable_pages(struct zone *zone);
403 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
404 gfp_t gfp_mask, nodemask_t *mask);
405
406 #define MEMCG_RECLAIM_MAY_SWAP (1 << 1)
407 #define MEMCG_RECLAIM_PROACTIVE (1 << 2)
408 #define MIN_SWAPPINESS 0
409 #define MAX_SWAPPINESS 200
410 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
411 unsigned long nr_pages,
412 gfp_t gfp_mask,
413 unsigned int reclaim_options,
414 int *swappiness);
415 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
416 gfp_t gfp_mask, bool noswap,
417 pg_data_t *pgdat,
418 unsigned long *nr_scanned);
419 extern unsigned long shrink_all_memory(unsigned long nr_pages);
420 extern int vm_swappiness;
421 long remove_mapping(struct address_space *mapping, struct folio *folio);
422
423 #ifdef CONFIG_NUMA
424 extern int node_reclaim_mode;
425 extern int sysctl_min_unmapped_ratio;
426 extern int sysctl_min_slab_ratio;
427 #else
428 #define node_reclaim_mode 0
429 #endif
430
431 static inline bool node_reclaim_enabled(void)
432 {
433 /* Is any node_reclaim_mode bit set? */
434 return node_reclaim_mode & (RECLAIM_ZONE|RECLAIM_WRITE|RECLAIM_UNMAP);
435 }
436
437 void check_move_unevictable_folios(struct folio_batch *fbatch);
438
439 extern void __meminit kswapd_run(int nid);
440 extern void __meminit kswapd_stop(int nid);
441
442 #ifdef CONFIG_SWAP
443
444 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
445 unsigned long nr_pages, sector_t start_block);
446 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
447 sector_t *);
448
449 static inline unsigned long total_swapcache_pages(void)
450 {
451 return global_node_page_state(NR_SWAPCACHE);
452 }
453
454 void free_swap_cache(struct folio *folio);
455 void free_page_and_swap_cache(struct page *);
456 void free_pages_and_swap_cache(struct encoded_page **, int);
457 /* linux/mm/swapfile.c */
458 extern atomic_long_t nr_swap_pages;
459 extern long total_swap_pages;
460 extern atomic_t nr_rotate_swap;
461 extern bool has_usable_swap(void);
462
463 /* Swap 50% full? Release swapcache more aggressively.. */
464 static inline bool vm_swap_full(void)
465 {
466 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
467 }
468
469 static inline long get_nr_swap_pages(void)
470 {
471 return atomic_long_read(&nr_swap_pages);
472 }
473
474 extern void si_swapinfo(struct sysinfo *);
475 swp_entry_t folio_alloc_swap(struct folio *folio);
476 bool folio_free_swap(struct folio *folio);
477 void put_swap_folio(struct folio *folio, swp_entry_t entry);
478 extern swp_entry_t get_swap_page_of_type(int);
479 extern int get_swap_pages(int n, swp_entry_t swp_entries[], int order);
480 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
481 extern void swap_shmem_alloc(swp_entry_t);
482 extern int swap_duplicate(swp_entry_t);
483 extern int swapcache_prepare(swp_entry_t);
484 extern void swap_free_nr(swp_entry_t entry, int nr_pages);
485 extern void swapcache_free_entries(swp_entry_t *entries, int n);
486 extern void free_swap_and_cache_nr(swp_entry_t entry, int nr);
487 int swap_type_of(dev_t device, sector_t offset);
488 int find_first_swap(dev_t *device);
489 extern unsigned int count_swap_pages(int, int);
490 extern sector_t swapdev_block(int, pgoff_t);
491 extern int __swap_count(swp_entry_t entry);
492 extern int swap_swapcount(struct swap_info_struct *si, swp_entry_t entry);
493 extern int swp_swapcount(swp_entry_t entry);
494 struct swap_info_struct *swp_swap_info(swp_entry_t entry);
495 struct backing_dev_info;
496 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
497 extern void exit_swap_address_space(unsigned int type);
498 extern struct swap_info_struct *get_swap_device(swp_entry_t entry);
499 sector_t swap_folio_sector(struct folio *folio);
500
501 static inline void put_swap_device(struct swap_info_struct *si)
502 {
503 percpu_ref_put(&si->users);
504 }
505
506 #else /* CONFIG_SWAP */
507 static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry)
508 {
509 return NULL;
510 }
511
512 static inline struct swap_info_struct *get_swap_device(swp_entry_t entry)
513 {
514 return NULL;
515 }
516
517 static inline void put_swap_device(struct swap_info_struct *si)
518 {
519 }
520
521 #define get_nr_swap_pages() 0L
522 #define total_swap_pages 0L
523 #define total_swapcache_pages() 0UL
524 #define vm_swap_full() 0
525
526 #define si_swapinfo(val) \
527 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
528 /* only sparc can not include linux/pagemap.h in this file
529 * so leave put_page and release_pages undeclared... */
530 #define free_page_and_swap_cache(page) \
531 put_page(page)
532 #define free_pages_and_swap_cache(pages, nr) \
533 release_pages((pages), (nr));
534
535 static inline void free_swap_and_cache_nr(swp_entry_t entry, int nr)
536 {
537 }
538
539 static inline void free_swap_cache(struct folio *folio)
540 {
541 }
542
543 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
544 {
545 return 0;
546 }
547
548 static inline void swap_shmem_alloc(swp_entry_t swp)
549 {
550 }
551
552 static inline int swap_duplicate(swp_entry_t swp)
553 {
554 return 0;
555 }
556
557 static inline int swapcache_prepare(swp_entry_t swp)
558 {
559 return 0;
560 }
561
562 static inline void swap_free_nr(swp_entry_t entry, int nr_pages)
563 {
564 }
565
566 static inline void put_swap_folio(struct folio *folio, swp_entry_t swp)
567 {
568 }
569
570 static inline int __swap_count(swp_entry_t entry)
571 {
572 return 0;
573 }
574
575 static inline int swap_swapcount(struct swap_info_struct *si, swp_entry_t entry)
576 {
577 return 0;
578 }
579
580 static inline int swp_swapcount(swp_entry_t entry)
581 {
582 return 0;
583 }
584
585 static inline swp_entry_t folio_alloc_swap(struct folio *folio)
586 {
587 swp_entry_t entry;
588 entry.val = 0;
589 return entry;
590 }
591
592 static inline bool folio_free_swap(struct folio *folio)
593 {
594 return false;
595 }
596
597 static inline int add_swap_extent(struct swap_info_struct *sis,
598 unsigned long start_page,
599 unsigned long nr_pages, sector_t start_block)
600 {
601 return -EINVAL;
602 }
603 #endif /* CONFIG_SWAP */
604
605 static inline void free_swap_and_cache(swp_entry_t entry)
606 {
607 free_swap_and_cache_nr(entry, 1);
608 }
609
610 static inline void swap_free(swp_entry_t entry)
611 {
612 swap_free_nr(entry, 1);
613 }
614
615 #ifdef CONFIG_MEMCG
616 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
617 {
618 /* Cgroup2 doesn't have per-cgroup swappiness */
619 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
620 return READ_ONCE(vm_swappiness);
621
622 /* root ? */
623 if (mem_cgroup_disabled() || mem_cgroup_is_root(memcg))
624 return READ_ONCE(vm_swappiness);
625
626 return READ_ONCE(memcg->swappiness);
627 }
628 #else
629 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
630 {
631 return READ_ONCE(vm_swappiness);
632 }
633 #endif
634
635 #if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
636 void __folio_throttle_swaprate(struct folio *folio, gfp_t gfp);
637 static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp)
638 {
639 if (mem_cgroup_disabled())
640 return;
641 __folio_throttle_swaprate(folio, gfp);
642 }
643 #else
644 static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp)
645 {
646 }
647 #endif
648
649 #if defined(CONFIG_MEMCG) && defined(CONFIG_SWAP)
650 void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry);
651 int __mem_cgroup_try_charge_swap(struct folio *folio, swp_entry_t entry);
652 static inline int mem_cgroup_try_charge_swap(struct folio *folio,
653 swp_entry_t entry)
654 {
655 if (mem_cgroup_disabled())
656 return 0;
657 return __mem_cgroup_try_charge_swap(folio, entry);
658 }
659
660 extern void __mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
661 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages)
662 {
663 if (mem_cgroup_disabled())
664 return;
665 __mem_cgroup_uncharge_swap(entry, nr_pages);
666 }
667
668 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
669 extern bool mem_cgroup_swap_full(struct folio *folio);
670 #else
671 static inline void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry)
672 {
673 }
674
675 static inline int mem_cgroup_try_charge_swap(struct folio *folio,
676 swp_entry_t entry)
677 {
678 return 0;
679 }
680
681 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
682 unsigned int nr_pages)
683 {
684 }
685
686 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
687 {
688 return get_nr_swap_pages();
689 }
690
691 static inline bool mem_cgroup_swap_full(struct folio *folio)
692 {
693 return vm_swap_full();
694 }
695 #endif
696
697 #endif /* __KERNEL__*/
698 #endif /* _LINUX_SWAP_H */
699
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.