xref: /linux-6.15/include/linux/swap.h (revision 2ec8107d)
1 #ifndef _LINUX_SWAP_H
2 #define _LINUX_SWAP_H
3 
4 #include <linux/spinlock.h>
5 #include <linux/linkage.h>
6 #include <linux/mmzone.h>
7 #include <linux/list.h>
8 #include <linux/memcontrol.h>
9 #include <linux/sched.h>
10 #include <linux/node.h>
11 #include <linux/fs.h>
12 #include <linux/atomic.h>
13 #include <linux/page-flags.h>
14 #include <asm/page.h>
15 
16 struct notifier_block;
17 
18 struct bio;
19 
20 #define SWAP_FLAG_PREFER	0x8000	/* set if swap priority specified */
21 #define SWAP_FLAG_PRIO_MASK	0x7fff
22 #define SWAP_FLAG_PRIO_SHIFT	0
23 #define SWAP_FLAG_DISCARD	0x10000 /* enable discard for swap */
24 #define SWAP_FLAG_DISCARD_ONCE	0x20000 /* discard swap area at swapon-time */
25 #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */
26 
27 #define SWAP_FLAGS_VALID	(SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \
28 				 SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \
29 				 SWAP_FLAG_DISCARD_PAGES)
30 
31 static inline int current_is_kswapd(void)
32 {
33 	return current->flags & PF_KSWAPD;
34 }
35 
36 /*
37  * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
38  * be swapped to.  The swap type and the offset into that swap type are
39  * encoded into pte's and into pgoff_t's in the swapcache.  Using five bits
40  * for the type means that the maximum number of swapcache pages is 27 bits
41  * on 32-bit-pgoff_t architectures.  And that assumes that the architecture packs
42  * the type/offset into the pte as 5/27 as well.
43  */
44 #define MAX_SWAPFILES_SHIFT	5
45 
46 /*
47  * Use some of the swap files numbers for other purposes. This
48  * is a convenient way to hook into the VM to trigger special
49  * actions on faults.
50  */
51 
52 /*
53  * NUMA node memory migration support
54  */
55 #ifdef CONFIG_MIGRATION
56 #define SWP_MIGRATION_NUM 2
57 #define SWP_MIGRATION_READ	(MAX_SWAPFILES + SWP_HWPOISON_NUM)
58 #define SWP_MIGRATION_WRITE	(MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
59 #else
60 #define SWP_MIGRATION_NUM 0
61 #endif
62 
63 /*
64  * Handling of hardware poisoned pages with memory corruption.
65  */
66 #ifdef CONFIG_MEMORY_FAILURE
67 #define SWP_HWPOISON_NUM 1
68 #define SWP_HWPOISON		MAX_SWAPFILES
69 #else
70 #define SWP_HWPOISON_NUM 0
71 #endif
72 
73 #define MAX_SWAPFILES \
74 	((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
75 
76 /*
77  * Magic header for a swap area. The first part of the union is
78  * what the swap magic looks like for the old (limited to 128MB)
79  * swap area format, the second part of the union adds - in the
80  * old reserved area - some extra information. Note that the first
81  * kilobyte is reserved for boot loader or disk label stuff...
82  *
83  * Having the magic at the end of the PAGE_SIZE makes detecting swap
84  * areas somewhat tricky on machines that support multiple page sizes.
85  * For 2.5 we'll probably want to move the magic to just beyond the
86  * bootbits...
87  */
88 union swap_header {
89 	struct {
90 		char reserved[PAGE_SIZE - 10];
91 		char magic[10];			/* SWAP-SPACE or SWAPSPACE2 */
92 	} magic;
93 	struct {
94 		char		bootbits[1024];	/* Space for disklabel etc. */
95 		__u32		version;
96 		__u32		last_page;
97 		__u32		nr_badpages;
98 		unsigned char	sws_uuid[16];
99 		unsigned char	sws_volume[16];
100 		__u32		padding[117];
101 		__u32		badpages[1];
102 	} info;
103 };
104 
105 /*
106  * current->reclaim_state points to one of these when a task is running
107  * memory reclaim
108  */
109 struct reclaim_state {
110 	unsigned long reclaimed_slab;
111 };
112 
113 #ifdef __KERNEL__
114 
115 struct address_space;
116 struct sysinfo;
117 struct writeback_control;
118 struct zone;
119 
120 /*
121  * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
122  * disk blocks.  A list of swap extents maps the entire swapfile.  (Where the
123  * term `swapfile' refers to either a blockdevice or an IS_REG file.  Apart
124  * from setup, they're handled identically.
125  *
126  * We always assume that blocks are of size PAGE_SIZE.
127  */
128 struct swap_extent {
129 	struct list_head list;
130 	pgoff_t start_page;
131 	pgoff_t nr_pages;
132 	sector_t start_block;
133 };
134 
135 /*
136  * Max bad pages in the new format..
137  */
138 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
139 #define MAX_SWAP_BADPAGES \
140 	((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
141 
142 enum {
143 	SWP_USED	= (1 << 0),	/* is slot in swap_info[] used? */
144 	SWP_WRITEOK	= (1 << 1),	/* ok to write to this swap?	*/
145 	SWP_DISCARDABLE = (1 << 2),	/* blkdev support discard */
146 	SWP_DISCARDING	= (1 << 3),	/* now discarding a free cluster */
147 	SWP_SOLIDSTATE	= (1 << 4),	/* blkdev seeks are cheap */
148 	SWP_CONTINUED	= (1 << 5),	/* swap_map has count continuation */
149 	SWP_BLKDEV	= (1 << 6),	/* its a block device */
150 	SWP_FILE	= (1 << 7),	/* set after swap_activate success */
151 	SWP_AREA_DISCARD = (1 << 8),	/* single-time swap area discards */
152 	SWP_PAGE_DISCARD = (1 << 9),	/* freed swap page-cluster discards */
153 					/* add others here before... */
154 	SWP_SCANNING	= (1 << 10),	/* refcount in scan_swap_map */
155 };
156 
157 #define SWAP_CLUSTER_MAX 32UL
158 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
159 
160 #define SWAP_MAP_MAX	0x3e	/* Max duplication count, in first swap_map */
161 #define SWAP_MAP_BAD	0x3f	/* Note pageblock is bad, in first swap_map */
162 #define SWAP_HAS_CACHE	0x40	/* Flag page is cached, in first swap_map */
163 #define SWAP_CONT_MAX	0x7f	/* Max count, in each swap_map continuation */
164 #define COUNT_CONTINUED	0x80	/* See swap_map continuation for full count */
165 #define SWAP_MAP_SHMEM	0xbf	/* Owned by shmem/tmpfs, in first swap_map */
166 
167 /*
168  * We use this to track usage of a cluster. A cluster is a block of swap disk
169  * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
170  * free clusters are organized into a list. We fetch an entry from the list to
171  * get a free cluster.
172  *
173  * The data field stores next cluster if the cluster is free or cluster usage
174  * counter otherwise. The flags field determines if a cluster is free. This is
175  * protected by swap_info_struct.lock.
176  */
177 struct swap_cluster_info {
178 	unsigned int data:24;
179 	unsigned int flags:8;
180 };
181 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
182 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
183 
184 /*
185  * We assign a cluster to each CPU, so each CPU can allocate swap entry from
186  * its own cluster and swapout sequentially. The purpose is to optimize swapout
187  * throughput.
188  */
189 struct percpu_cluster {
190 	struct swap_cluster_info index; /* Current cluster index */
191 	unsigned int next; /* Likely next allocation offset */
192 };
193 
194 struct swap_cluster_list {
195 	struct swap_cluster_info head;
196 	struct swap_cluster_info tail;
197 };
198 
199 /*
200  * The in-memory structure used to track swap areas.
201  */
202 struct swap_info_struct {
203 	unsigned long	flags;		/* SWP_USED etc: see above */
204 	signed short	prio;		/* swap priority of this type */
205 	struct plist_node list;		/* entry in swap_active_head */
206 	struct plist_node avail_list;	/* entry in swap_avail_head */
207 	signed char	type;		/* strange name for an index */
208 	unsigned int	max;		/* extent of the swap_map */
209 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
210 	struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
211 	struct swap_cluster_list free_clusters; /* free clusters list */
212 	unsigned int lowest_bit;	/* index of first free in swap_map */
213 	unsigned int highest_bit;	/* index of last free in swap_map */
214 	unsigned int pages;		/* total of usable pages of swap */
215 	unsigned int inuse_pages;	/* number of those currently in use */
216 	unsigned int cluster_next;	/* likely index for next allocation */
217 	unsigned int cluster_nr;	/* countdown to next cluster search */
218 	struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
219 	struct swap_extent *curr_swap_extent;
220 	struct swap_extent first_swap_extent;
221 	struct block_device *bdev;	/* swap device or bdev of swap file */
222 	struct file *swap_file;		/* seldom referenced */
223 	unsigned int old_block_size;	/* seldom referenced */
224 #ifdef CONFIG_FRONTSWAP
225 	unsigned long *frontswap_map;	/* frontswap in-use, one bit per page */
226 	atomic_t frontswap_pages;	/* frontswap pages in-use counter */
227 #endif
228 	spinlock_t lock;		/*
229 					 * protect map scan related fields like
230 					 * swap_map, lowest_bit, highest_bit,
231 					 * inuse_pages, cluster_next,
232 					 * cluster_nr, lowest_alloc,
233 					 * highest_alloc, free/discard cluster
234 					 * list. other fields are only changed
235 					 * at swapon/swapoff, so are protected
236 					 * by swap_lock. changing flags need
237 					 * hold this lock and swap_lock. If
238 					 * both locks need hold, hold swap_lock
239 					 * first.
240 					 */
241 	struct work_struct discard_work; /* discard worker */
242 	struct swap_cluster_list discard_clusters; /* discard clusters list */
243 };
244 
245 /* linux/mm/workingset.c */
246 void *workingset_eviction(struct address_space *mapping, struct page *page);
247 bool workingset_refault(void *shadow);
248 void workingset_activation(struct page *page);
249 extern struct list_lru workingset_shadow_nodes;
250 
251 static inline unsigned int workingset_node_pages(struct radix_tree_node *node)
252 {
253 	return node->count & RADIX_TREE_COUNT_MASK;
254 }
255 
256 static inline void workingset_node_pages_inc(struct radix_tree_node *node)
257 {
258 	node->count++;
259 }
260 
261 static inline void workingset_node_pages_dec(struct radix_tree_node *node)
262 {
263 	VM_WARN_ON_ONCE(!workingset_node_pages(node));
264 	node->count--;
265 }
266 
267 static inline unsigned int workingset_node_shadows(struct radix_tree_node *node)
268 {
269 	return node->count >> RADIX_TREE_COUNT_SHIFT;
270 }
271 
272 static inline void workingset_node_shadows_inc(struct radix_tree_node *node)
273 {
274 	node->count += 1U << RADIX_TREE_COUNT_SHIFT;
275 }
276 
277 static inline void workingset_node_shadows_dec(struct radix_tree_node *node)
278 {
279 	VM_WARN_ON_ONCE(!workingset_node_shadows(node));
280 	node->count -= 1U << RADIX_TREE_COUNT_SHIFT;
281 }
282 
283 /* linux/mm/page_alloc.c */
284 extern unsigned long totalram_pages;
285 extern unsigned long totalreserve_pages;
286 extern unsigned long nr_free_buffer_pages(void);
287 extern unsigned long nr_free_pagecache_pages(void);
288 
289 /* Definition of global_page_state not available yet */
290 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
291 
292 
293 /* linux/mm/swap.c */
294 extern void lru_cache_add(struct page *);
295 extern void lru_cache_add_anon(struct page *page);
296 extern void lru_cache_add_file(struct page *page);
297 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
298 			 struct lruvec *lruvec, struct list_head *head);
299 extern void activate_page(struct page *);
300 extern void mark_page_accessed(struct page *);
301 extern void lru_add_drain(void);
302 extern void lru_add_drain_cpu(int cpu);
303 extern void lru_add_drain_all(void);
304 extern void rotate_reclaimable_page(struct page *page);
305 extern void deactivate_file_page(struct page *page);
306 extern void deactivate_page(struct page *page);
307 extern void swap_setup(void);
308 
309 extern void add_page_to_unevictable_list(struct page *page);
310 
311 extern void lru_cache_add_active_or_unevictable(struct page *page,
312 						struct vm_area_struct *vma);
313 
314 /* linux/mm/vmscan.c */
315 extern unsigned long zone_reclaimable_pages(struct zone *zone);
316 extern unsigned long pgdat_reclaimable_pages(struct pglist_data *pgdat);
317 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
318 					gfp_t gfp_mask, nodemask_t *mask);
319 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
320 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
321 						  unsigned long nr_pages,
322 						  gfp_t gfp_mask,
323 						  bool may_swap);
324 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
325 						gfp_t gfp_mask, bool noswap,
326 						pg_data_t *pgdat,
327 						unsigned long *nr_scanned);
328 extern unsigned long shrink_all_memory(unsigned long nr_pages);
329 extern int vm_swappiness;
330 extern int remove_mapping(struct address_space *mapping, struct page *page);
331 extern unsigned long vm_total_pages;
332 
333 #ifdef CONFIG_NUMA
334 extern int node_reclaim_mode;
335 extern int sysctl_min_unmapped_ratio;
336 extern int sysctl_min_slab_ratio;
337 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
338 #else
339 #define node_reclaim_mode 0
340 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
341 				unsigned int order)
342 {
343 	return 0;
344 }
345 #endif
346 
347 extern int page_evictable(struct page *page);
348 extern void check_move_unevictable_pages(struct page **, int nr_pages);
349 
350 extern int kswapd_run(int nid);
351 extern void kswapd_stop(int nid);
352 
353 #ifdef CONFIG_SWAP
354 /* linux/mm/page_io.c */
355 extern int swap_readpage(struct page *);
356 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
357 extern void end_swap_bio_write(struct bio *bio);
358 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
359 	bio_end_io_t end_write_func);
360 extern int swap_set_page_dirty(struct page *page);
361 
362 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
363 		unsigned long nr_pages, sector_t start_block);
364 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
365 		sector_t *);
366 
367 /* linux/mm/swap_state.c */
368 extern struct address_space swapper_spaces[];
369 #define swap_address_space(entry) (&swapper_spaces[swp_type(entry)])
370 extern unsigned long total_swapcache_pages(void);
371 extern void show_swap_cache_info(void);
372 extern int add_to_swap(struct page *, struct list_head *list);
373 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
374 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
375 extern void __delete_from_swap_cache(struct page *);
376 extern void delete_from_swap_cache(struct page *);
377 extern void free_page_and_swap_cache(struct page *);
378 extern void free_pages_and_swap_cache(struct page **, int);
379 extern struct page *lookup_swap_cache(swp_entry_t);
380 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
381 			struct vm_area_struct *vma, unsigned long addr);
382 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
383 			struct vm_area_struct *vma, unsigned long addr,
384 			bool *new_page_allocated);
385 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
386 			struct vm_area_struct *vma, unsigned long addr);
387 
388 /* linux/mm/swapfile.c */
389 extern atomic_long_t nr_swap_pages;
390 extern long total_swap_pages;
391 
392 /* Swap 50% full? Release swapcache more aggressively.. */
393 static inline bool vm_swap_full(void)
394 {
395 	return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
396 }
397 
398 static inline long get_nr_swap_pages(void)
399 {
400 	return atomic_long_read(&nr_swap_pages);
401 }
402 
403 extern void si_swapinfo(struct sysinfo *);
404 extern swp_entry_t get_swap_page(void);
405 extern swp_entry_t get_swap_page_of_type(int);
406 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
407 extern void swap_shmem_alloc(swp_entry_t);
408 extern int swap_duplicate(swp_entry_t);
409 extern int swapcache_prepare(swp_entry_t);
410 extern void swap_free(swp_entry_t);
411 extern void swapcache_free(swp_entry_t);
412 extern int free_swap_and_cache(swp_entry_t);
413 extern int swap_type_of(dev_t, sector_t, struct block_device **);
414 extern unsigned int count_swap_pages(int, int);
415 extern sector_t map_swap_page(struct page *, struct block_device **);
416 extern sector_t swapdev_block(int, pgoff_t);
417 extern int page_swapcount(struct page *);
418 extern int swp_swapcount(swp_entry_t entry);
419 extern struct swap_info_struct *page_swap_info(struct page *);
420 extern bool reuse_swap_page(struct page *, int *);
421 extern int try_to_free_swap(struct page *);
422 struct backing_dev_info;
423 
424 #else /* CONFIG_SWAP */
425 
426 #define swap_address_space(entry)		(NULL)
427 #define get_nr_swap_pages()			0L
428 #define total_swap_pages			0L
429 #define total_swapcache_pages()			0UL
430 #define vm_swap_full()				0
431 
432 #define si_swapinfo(val) \
433 	do { (val)->freeswap = (val)->totalswap = 0; } while (0)
434 /* only sparc can not include linux/pagemap.h in this file
435  * so leave put_page and release_pages undeclared... */
436 #define free_page_and_swap_cache(page) \
437 	put_page(page)
438 #define free_pages_and_swap_cache(pages, nr) \
439 	release_pages((pages), (nr), false);
440 
441 static inline void show_swap_cache_info(void)
442 {
443 }
444 
445 #define free_swap_and_cache(swp)	is_migration_entry(swp)
446 #define swapcache_prepare(swp)		is_migration_entry(swp)
447 
448 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
449 {
450 	return 0;
451 }
452 
453 static inline void swap_shmem_alloc(swp_entry_t swp)
454 {
455 }
456 
457 static inline int swap_duplicate(swp_entry_t swp)
458 {
459 	return 0;
460 }
461 
462 static inline void swap_free(swp_entry_t swp)
463 {
464 }
465 
466 static inline void swapcache_free(swp_entry_t swp)
467 {
468 }
469 
470 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
471 			struct vm_area_struct *vma, unsigned long addr)
472 {
473 	return NULL;
474 }
475 
476 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
477 {
478 	return 0;
479 }
480 
481 static inline struct page *lookup_swap_cache(swp_entry_t swp)
482 {
483 	return NULL;
484 }
485 
486 static inline int add_to_swap(struct page *page, struct list_head *list)
487 {
488 	return 0;
489 }
490 
491 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
492 							gfp_t gfp_mask)
493 {
494 	return -1;
495 }
496 
497 static inline void __delete_from_swap_cache(struct page *page)
498 {
499 }
500 
501 static inline void delete_from_swap_cache(struct page *page)
502 {
503 }
504 
505 static inline int page_swapcount(struct page *page)
506 {
507 	return 0;
508 }
509 
510 static inline int swp_swapcount(swp_entry_t entry)
511 {
512 	return 0;
513 }
514 
515 #define reuse_swap_page(page, total_mapcount) \
516 	(page_trans_huge_mapcount(page, total_mapcount) == 1)
517 
518 static inline int try_to_free_swap(struct page *page)
519 {
520 	return 0;
521 }
522 
523 static inline swp_entry_t get_swap_page(void)
524 {
525 	swp_entry_t entry;
526 	entry.val = 0;
527 	return entry;
528 }
529 
530 #endif /* CONFIG_SWAP */
531 
532 #ifdef CONFIG_MEMCG
533 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
534 {
535 	/* Cgroup2 doesn't have per-cgroup swappiness */
536 	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
537 		return vm_swappiness;
538 
539 	/* root ? */
540 	if (mem_cgroup_disabled() || !memcg->css.parent)
541 		return vm_swappiness;
542 
543 	return memcg->swappiness;
544 }
545 
546 #else
547 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
548 {
549 	return vm_swappiness;
550 }
551 #endif
552 
553 #ifdef CONFIG_MEMCG_SWAP
554 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
555 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
556 extern void mem_cgroup_uncharge_swap(swp_entry_t entry);
557 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
558 extern bool mem_cgroup_swap_full(struct page *page);
559 #else
560 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
561 {
562 }
563 
564 static inline int mem_cgroup_try_charge_swap(struct page *page,
565 					     swp_entry_t entry)
566 {
567 	return 0;
568 }
569 
570 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry)
571 {
572 }
573 
574 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
575 {
576 	return get_nr_swap_pages();
577 }
578 
579 static inline bool mem_cgroup_swap_full(struct page *page)
580 {
581 	return vm_swap_full();
582 }
583 #endif
584 
585 #endif /* __KERNEL__*/
586 #endif /* _LINUX_SWAP_H */
587