1 /* SPDX-License-Identifier: GPL-2.0 */
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>
13 #include <linux/atomic.h>
14 #include <linux/page-flags.h>
17 struct notifier_block;
23 #define SWAP_FLAG_PREFER 0x8000 /* set if swap priority specified */
24 #define SWAP_FLAG_PRIO_MASK 0x7fff
25 #define SWAP_FLAG_PRIO_SHIFT 0
26 #define SWAP_FLAG_DISCARD 0x10000 /* enable discard for swap */
27 #define SWAP_FLAG_DISCARD_ONCE 0x20000 /* discard swap area at swapon-time */
28 #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */
30 #define SWAP_FLAGS_VALID (SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \
31 SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \
32 SWAP_FLAG_DISCARD_PAGES)
35 static inline int current_is_kswapd(void)
37 return current->flags & PF_KSWAPD;
41 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
42 * be swapped to. The swap type and the offset into that swap type are
43 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
44 * for the type means that the maximum number of swapcache pages is 27 bits
45 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
46 * the type/offset into the pte as 5/27 as well.
48 #define MAX_SWAPFILES_SHIFT 5
51 * Use some of the swap files numbers for other purposes. This
52 * is a convenient way to hook into the VM to trigger special
57 * Unaddressable device memory support. See include/linux/hmm.h and
58 * Documentation/vm/hmm.rst. Short description is we need struct pages for
59 * device memory that is unaddressable (inaccessible) by CPU, so that we can
60 * migrate part of a process memory to device memory.
62 * When a page is migrated from CPU to device, we set the CPU page table entry
63 * to a special SWP_DEVICE_* entry.
65 #ifdef CONFIG_DEVICE_PRIVATE
66 #define SWP_DEVICE_NUM 2
67 #define SWP_DEVICE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM)
68 #define SWP_DEVICE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+1)
70 #define SWP_DEVICE_NUM 0
74 * NUMA node memory migration support
76 #ifdef CONFIG_MIGRATION
77 #define SWP_MIGRATION_NUM 2
78 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
79 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
81 #define SWP_MIGRATION_NUM 0
85 * Handling of hardware poisoned pages with memory corruption.
87 #ifdef CONFIG_MEMORY_FAILURE
88 #define SWP_HWPOISON_NUM 1
89 #define SWP_HWPOISON MAX_SWAPFILES
91 #define SWP_HWPOISON_NUM 0
94 #define MAX_SWAPFILES \
95 ((1 << MAX_SWAPFILES_SHIFT) - SWP_DEVICE_NUM - \
96 SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
99 * Magic header for a swap area. The first part of the union is
100 * what the swap magic looks like for the old (limited to 128MB)
101 * swap area format, the second part of the union adds - in the
102 * old reserved area - some extra information. Note that the first
103 * kilobyte is reserved for boot loader or disk label stuff...
105 * Having the magic at the end of the PAGE_SIZE makes detecting swap
106 * areas somewhat tricky on machines that support multiple page sizes.
107 * For 2.5 we'll probably want to move the magic to just beyond the
112 char reserved[PAGE_SIZE - 10];
113 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
116 char bootbits[1024]; /* Space for disklabel etc. */
120 unsigned char sws_uuid[16];
121 unsigned char sws_volume[16];
128 * current->reclaim_state points to one of these when a task is running
131 struct reclaim_state {
132 unsigned long reclaimed_slab;
137 struct address_space;
139 struct writeback_control;
143 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
144 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
145 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
146 * from setup, they're handled identically.
148 * We always assume that blocks are of size PAGE_SIZE.
151 struct rb_node rb_node;
154 sector_t start_block;
158 * Max bad pages in the new format..
160 #define MAX_SWAP_BADPAGES \
161 ((offsetof(union swap_header, magic.magic) - \
162 offsetof(union swap_header, info.badpages)) / sizeof(int))
165 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
166 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
167 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
168 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
169 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
170 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
171 SWP_BLKDEV = (1 << 6), /* its a block device */
172 SWP_ACTIVATED = (1 << 7), /* set after swap_activate success */
173 SWP_FS = (1 << 8), /* swap file goes through fs */
174 SWP_AREA_DISCARD = (1 << 9), /* single-time swap area discards */
175 SWP_PAGE_DISCARD = (1 << 10), /* freed swap page-cluster discards */
176 SWP_STABLE_WRITES = (1 << 11), /* no overwrite PG_writeback pages */
177 SWP_SYNCHRONOUS_IO = (1 << 12), /* synchronous IO is efficient */
178 SWP_VALID = (1 << 13), /* swap is valid to be operated on? */
179 /* add others here before... */
180 SWP_SCANNING = (1 << 14), /* refcount in scan_swap_map */
183 #define SWAP_CLUSTER_MAX 32UL
184 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
186 /* Bit flag in swap_map */
187 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
188 #define COUNT_CONTINUED 0x80 /* Flag swap_map continuation for full count */
190 /* Special value in first swap_map */
191 #define SWAP_MAP_MAX 0x3e /* Max count */
192 #define SWAP_MAP_BAD 0x3f /* Note page is bad */
193 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs */
195 /* Special value in each swap_map continuation */
196 #define SWAP_CONT_MAX 0x7f /* Max count */
199 * We use this to track usage of a cluster. A cluster is a block of swap disk
200 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
201 * free clusters are organized into a list. We fetch an entry from the list to
202 * get a free cluster.
204 * The data field stores next cluster if the cluster is free or cluster usage
205 * counter otherwise. The flags field determines if a cluster is free. This is
206 * protected by swap_info_struct.lock.
208 struct swap_cluster_info {
210 * Protect swap_cluster_info fields
211 * and swap_info_struct->swap_map
212 * elements correspond to the swap
215 unsigned int data:24;
216 unsigned int flags:8;
218 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
219 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
220 #define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */
223 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
224 * its own cluster and swapout sequentially. The purpose is to optimize swapout
227 struct percpu_cluster {
228 struct swap_cluster_info index; /* Current cluster index */
229 unsigned int next; /* Likely next allocation offset */
232 struct swap_cluster_list {
233 struct swap_cluster_info head;
234 struct swap_cluster_info tail;
238 * The in-memory structure used to track swap areas.
240 struct swap_info_struct {
241 unsigned long flags; /* SWP_USED etc: see above */
242 signed short prio; /* swap priority of this type */
243 struct plist_node list; /* entry in swap_active_head */
244 signed char type; /* strange name for an index */
245 unsigned int max; /* extent of the swap_map */
246 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
247 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
248 struct swap_cluster_list free_clusters; /* free clusters list */
249 unsigned int lowest_bit; /* index of first free in swap_map */
250 unsigned int highest_bit; /* index of last free in swap_map */
251 unsigned int pages; /* total of usable pages of swap */
252 unsigned int inuse_pages; /* number of those currently in use */
253 unsigned int cluster_next; /* likely index for next allocation */
254 unsigned int cluster_nr; /* countdown to next cluster search */
255 unsigned int __percpu *cluster_next_cpu; /*percpu index for next allocation */
256 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
257 struct rb_root swap_extent_root;/* root of the swap extent rbtree */
258 struct block_device *bdev; /* swap device or bdev of swap file */
259 struct file *swap_file; /* seldom referenced */
260 unsigned int old_block_size; /* seldom referenced */
261 #ifdef CONFIG_FRONTSWAP
262 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
263 atomic_t frontswap_pages; /* frontswap pages in-use counter */
266 * protect map scan related fields like
267 * swap_map, lowest_bit, highest_bit,
268 * inuse_pages, cluster_next,
269 * cluster_nr, lowest_alloc,
270 * highest_alloc, free/discard cluster
271 * list. other fields are only changed
272 * at swapon/swapoff, so are protected
273 * by swap_lock. changing flags need
274 * hold this lock and swap_lock. If
275 * both locks need hold, hold swap_lock
278 spinlock_t cont_lock; /*
279 * protect swap count continuation page
282 struct work_struct discard_work; /* discard worker */
283 struct swap_cluster_list discard_clusters; /* discard clusters list */
284 struct plist_node avail_lists[]; /*
285 * entries in swap_avail_heads, one
287 * Must be last as the number of the
288 * array is nr_node_ids, which is not
289 * a fixed value so have to allocate
291 * And it has to be an array so that
292 * plist_for_each_* can work.
297 #define SWAP_RA_ORDER_CEILING 5
299 /* Avoid stack overflow, because we need to save part of page table */
300 #define SWAP_RA_ORDER_CEILING 3
301 #define SWAP_RA_PTE_CACHE_SIZE (1 << SWAP_RA_ORDER_CEILING)
304 struct vma_swap_readahead {
306 unsigned short offset;
307 unsigned short nr_pte;
311 pte_t ptes[SWAP_RA_PTE_CACHE_SIZE];
315 /* linux/mm/workingset.c */
316 void workingset_age_nonresident(struct lruvec *lruvec, unsigned long nr_pages);
317 void *workingset_eviction(struct page *page, struct mem_cgroup *target_memcg);
318 void workingset_refault(struct page *page, void *shadow);
319 void workingset_activation(struct page *page);
321 /* Only track the nodes of mappings with shadow entries */
322 void workingset_update_node(struct xa_node *node);
323 #define mapping_set_update(xas, mapping) do { \
324 if (!dax_mapping(mapping) && !shmem_mapping(mapping)) \
325 xas_set_update(xas, workingset_update_node); \
328 /* linux/mm/page_alloc.c */
329 extern unsigned long totalreserve_pages;
330 extern unsigned long nr_free_buffer_pages(void);
331 extern unsigned long nr_free_pagecache_pages(void);
333 /* Definition of global_zone_page_state not available yet */
334 #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES)
337 /* linux/mm/swap.c */
338 extern void lru_note_cost(struct lruvec *lruvec, bool file,
339 unsigned int nr_pages);
340 extern void lru_note_cost_page(struct page *);
341 extern void lru_cache_add(struct page *);
342 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
343 struct lruvec *lruvec, struct list_head *head);
344 extern void activate_page(struct page *);
345 extern void mark_page_accessed(struct page *);
346 extern void lru_add_drain(void);
347 extern void lru_add_drain_cpu(int cpu);
348 extern void lru_add_drain_cpu_zone(struct zone *zone);
349 extern void lru_add_drain_all(void);
350 extern void rotate_reclaimable_page(struct page *page);
351 extern void deactivate_file_page(struct page *page);
352 extern void deactivate_page(struct page *page);
353 extern void mark_page_lazyfree(struct page *page);
354 extern void swap_setup(void);
356 extern void lru_cache_add_active_or_unevictable(struct page *page,
357 struct vm_area_struct *vma);
359 /* linux/mm/vmscan.c */
360 extern unsigned long zone_reclaimable_pages(struct zone *zone);
361 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
362 gfp_t gfp_mask, nodemask_t *mask);
363 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
364 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
365 unsigned long nr_pages,
368 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
369 gfp_t gfp_mask, bool noswap,
371 unsigned long *nr_scanned);
372 extern unsigned long shrink_all_memory(unsigned long nr_pages);
373 extern int vm_swappiness;
374 extern int remove_mapping(struct address_space *mapping, struct page *page);
375 extern unsigned long vm_total_pages;
377 extern unsigned long reclaim_pages(struct list_head *page_list);
379 extern int node_reclaim_mode;
380 extern int sysctl_min_unmapped_ratio;
381 extern int sysctl_min_slab_ratio;
383 #define node_reclaim_mode 0
386 extern void check_move_unevictable_pages(struct pagevec *pvec);
388 extern int kswapd_run(int nid);
389 extern void kswapd_stop(int nid);
393 #include <linux/blk_types.h> /* for bio_end_io_t */
395 /* linux/mm/page_io.c */
396 extern int swap_readpage(struct page *page, bool do_poll);
397 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
398 extern void end_swap_bio_write(struct bio *bio);
399 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
400 bio_end_io_t end_write_func);
401 extern int swap_set_page_dirty(struct page *page);
403 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
404 unsigned long nr_pages, sector_t start_block);
405 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
408 /* linux/mm/swap_state.c */
409 /* One swap address space for each 64M swap space */
410 #define SWAP_ADDRESS_SPACE_SHIFT 14
411 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
412 extern struct address_space *swapper_spaces[];
413 #define swap_address_space(entry) \
414 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
415 >> SWAP_ADDRESS_SPACE_SHIFT])
416 extern unsigned long total_swapcache_pages(void);
417 extern void show_swap_cache_info(void);
418 extern int add_to_swap(struct page *page);
419 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
420 extern void __delete_from_swap_cache(struct page *, swp_entry_t entry);
421 extern void delete_from_swap_cache(struct page *);
422 extern void free_page_and_swap_cache(struct page *);
423 extern void free_pages_and_swap_cache(struct page **, int);
424 extern struct page *lookup_swap_cache(swp_entry_t entry,
425 struct vm_area_struct *vma,
427 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
428 struct vm_area_struct *vma, unsigned long addr,
430 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
431 struct vm_area_struct *vma, unsigned long addr,
432 bool *new_page_allocated);
433 extern struct page *swap_cluster_readahead(swp_entry_t entry, gfp_t flag,
434 struct vm_fault *vmf);
435 extern struct page *swapin_readahead(swp_entry_t entry, gfp_t flag,
436 struct vm_fault *vmf);
438 /* linux/mm/swapfile.c */
439 extern atomic_long_t nr_swap_pages;
440 extern long total_swap_pages;
441 extern atomic_t nr_rotate_swap;
442 extern bool has_usable_swap(void);
444 /* Swap 50% full? Release swapcache more aggressively.. */
445 static inline bool vm_swap_full(void)
447 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
450 static inline long get_nr_swap_pages(void)
452 return atomic_long_read(&nr_swap_pages);
455 extern void si_swapinfo(struct sysinfo *);
456 extern swp_entry_t get_swap_page(struct page *page);
457 extern void put_swap_page(struct page *page, swp_entry_t entry);
458 extern swp_entry_t get_swap_page_of_type(int);
459 extern int get_swap_pages(int n, swp_entry_t swp_entries[], int entry_size);
460 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
461 extern void swap_shmem_alloc(swp_entry_t);
462 extern int swap_duplicate(swp_entry_t);
463 extern int swapcache_prepare(swp_entry_t);
464 extern void swap_free(swp_entry_t);
465 extern void swapcache_free_entries(swp_entry_t *entries, int n);
466 extern int free_swap_and_cache(swp_entry_t);
467 extern int swap_type_of(dev_t, sector_t, struct block_device **);
468 extern unsigned int count_swap_pages(int, int);
469 extern sector_t map_swap_page(struct page *, struct block_device **);
470 extern sector_t swapdev_block(int, pgoff_t);
471 extern int page_swapcount(struct page *);
472 extern int __swap_count(swp_entry_t entry);
473 extern int __swp_swapcount(swp_entry_t entry);
474 extern int swp_swapcount(swp_entry_t entry);
475 extern struct swap_info_struct *page_swap_info(struct page *);
476 extern struct swap_info_struct *swp_swap_info(swp_entry_t entry);
477 extern bool reuse_swap_page(struct page *, int *);
478 extern int try_to_free_swap(struct page *);
479 struct backing_dev_info;
480 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
481 extern void exit_swap_address_space(unsigned int type);
482 extern struct swap_info_struct *get_swap_device(swp_entry_t entry);
484 static inline void put_swap_device(struct swap_info_struct *si)
489 #else /* CONFIG_SWAP */
491 static inline int swap_readpage(struct page *page, bool do_poll)
496 static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry)
501 #define swap_address_space(entry) (NULL)
502 #define get_nr_swap_pages() 0L
503 #define total_swap_pages 0L
504 #define total_swapcache_pages() 0UL
505 #define vm_swap_full() 0
507 #define si_swapinfo(val) \
508 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
509 /* only sparc can not include linux/pagemap.h in this file
510 * so leave put_page and release_pages undeclared... */
511 #define free_page_and_swap_cache(page) \
513 #define free_pages_and_swap_cache(pages, nr) \
514 release_pages((pages), (nr));
516 static inline void show_swap_cache_info(void)
520 #define free_swap_and_cache(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
521 #define swapcache_prepare(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
523 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
528 static inline void swap_shmem_alloc(swp_entry_t swp)
532 static inline int swap_duplicate(swp_entry_t swp)
537 static inline void swap_free(swp_entry_t swp)
541 static inline void put_swap_page(struct page *page, swp_entry_t swp)
545 static inline struct page *swap_cluster_readahead(swp_entry_t entry,
546 gfp_t gfp_mask, struct vm_fault *vmf)
551 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
552 struct vm_fault *vmf)
557 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
562 static inline struct page *lookup_swap_cache(swp_entry_t swp,
563 struct vm_area_struct *vma,
569 static inline int add_to_swap(struct page *page)
574 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
580 static inline void __delete_from_swap_cache(struct page *page,
585 static inline void delete_from_swap_cache(struct page *page)
589 static inline int page_swapcount(struct page *page)
594 static inline int __swap_count(swp_entry_t entry)
599 static inline int __swp_swapcount(swp_entry_t entry)
604 static inline int swp_swapcount(swp_entry_t entry)
609 #define reuse_swap_page(page, total_map_swapcount) \
610 (page_trans_huge_mapcount(page, total_map_swapcount) == 1)
612 static inline int try_to_free_swap(struct page *page)
617 static inline swp_entry_t get_swap_page(struct page *page)
624 #endif /* CONFIG_SWAP */
626 #ifdef CONFIG_THP_SWAP
627 extern int split_swap_cluster(swp_entry_t entry);
629 static inline int split_swap_cluster(swp_entry_t entry)
636 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
638 /* Cgroup2 doesn't have per-cgroup swappiness */
639 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
640 return vm_swappiness;
643 if (mem_cgroup_disabled() || mem_cgroup_is_root(memcg))
644 return vm_swappiness;
646 return memcg->swappiness;
649 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
651 return vm_swappiness;
655 #if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
656 extern void cgroup_throttle_swaprate(struct page *page, gfp_t gfp_mask);
658 static inline void cgroup_throttle_swaprate(struct page *page, gfp_t gfp_mask)
663 #ifdef CONFIG_MEMCG_SWAP
664 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
665 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
666 extern void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
667 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
668 extern bool mem_cgroup_swap_full(struct page *page);
670 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
674 static inline int mem_cgroup_try_charge_swap(struct page *page,
680 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
681 unsigned int nr_pages)
685 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
687 return get_nr_swap_pages();
690 static inline bool mem_cgroup_swap_full(struct page *page)
692 return vm_swap_full();
696 #endif /* __KERNEL__*/
697 #endif /* _LINUX_SWAP_H */