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;
21 #define SWAP_FLAG_PREFER 0x8000 /* set if swap priority specified */
22 #define SWAP_FLAG_PRIO_MASK 0x7fff
23 #define SWAP_FLAG_PRIO_SHIFT 0
24 #define SWAP_FLAG_DISCARD 0x10000 /* enable discard for swap */
25 #define SWAP_FLAG_DISCARD_ONCE 0x20000 /* discard swap area at swapon-time */
26 #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */
28 #define SWAP_FLAGS_VALID (SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \
29 SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \
30 SWAP_FLAG_DISCARD_PAGES)
33 static inline int current_is_kswapd(void)
35 return current->flags & PF_KSWAPD;
39 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
40 * be swapped to. The swap type and the offset into that swap type are
41 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
42 * for the type means that the maximum number of swapcache pages is 27 bits
43 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
44 * the type/offset into the pte as 5/27 as well.
46 #define MAX_SWAPFILES_SHIFT 5
49 * Use some of the swap files numbers for other purposes. This
50 * is a convenient way to hook into the VM to trigger special
55 * Unaddressable device memory support. See include/linux/hmm.h and
56 * Documentation/vm/hmm.rst. Short description is we need struct pages for
57 * device memory that is unaddressable (inaccessible) by CPU, so that we can
58 * migrate part of a process memory to device memory.
60 * When a page is migrated from CPU to device, we set the CPU page table entry
61 * to a special SWP_DEVICE_* entry.
63 #ifdef CONFIG_DEVICE_PRIVATE
64 #define SWP_DEVICE_NUM 2
65 #define SWP_DEVICE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM)
66 #define SWP_DEVICE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+1)
68 #define SWP_DEVICE_NUM 0
72 * NUMA node memory migration support
74 #ifdef CONFIG_MIGRATION
75 #define SWP_MIGRATION_NUM 2
76 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
77 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
79 #define SWP_MIGRATION_NUM 0
83 * Handling of hardware poisoned pages with memory corruption.
85 #ifdef CONFIG_MEMORY_FAILURE
86 #define SWP_HWPOISON_NUM 1
87 #define SWP_HWPOISON MAX_SWAPFILES
89 #define SWP_HWPOISON_NUM 0
92 #define MAX_SWAPFILES \
93 ((1 << MAX_SWAPFILES_SHIFT) - SWP_DEVICE_NUM - \
94 SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
97 * Magic header for a swap area. The first part of the union is
98 * what the swap magic looks like for the old (limited to 128MB)
99 * swap area format, the second part of the union adds - in the
100 * old reserved area - some extra information. Note that the first
101 * kilobyte is reserved for boot loader or disk label stuff...
103 * Having the magic at the end of the PAGE_SIZE makes detecting swap
104 * areas somewhat tricky on machines that support multiple page sizes.
105 * For 2.5 we'll probably want to move the magic to just beyond the
110 char reserved[PAGE_SIZE - 10];
111 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
114 char bootbits[1024]; /* Space for disklabel etc. */
118 unsigned char sws_uuid[16];
119 unsigned char sws_volume[16];
126 * current->reclaim_state points to one of these when a task is running
129 struct reclaim_state {
130 unsigned long reclaimed_slab;
135 struct address_space;
137 struct writeback_control;
141 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
142 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
143 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
144 * from setup, they're handled identically.
146 * We always assume that blocks are of size PAGE_SIZE.
149 struct list_head list;
152 sector_t start_block;
156 * Max bad pages in the new format..
158 #define MAX_SWAP_BADPAGES \
159 ((offsetof(union swap_header, magic.magic) - \
160 offsetof(union swap_header, info.badpages)) / sizeof(int))
163 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
164 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
165 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
166 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
167 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
168 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
169 SWP_BLKDEV = (1 << 6), /* its a block device */
170 SWP_FILE = (1 << 7), /* set after swap_activate success */
171 SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
172 SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
173 SWP_STABLE_WRITES = (1 << 10), /* no overwrite PG_writeback pages */
174 SWP_SYNCHRONOUS_IO = (1 << 11), /* synchronous IO is efficient */
175 /* add others here before... */
176 SWP_SCANNING = (1 << 12), /* refcount in scan_swap_map */
179 #define SWAP_CLUSTER_MAX 32UL
180 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
182 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
183 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
184 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
185 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
186 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
187 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
190 * We use this to track usage of a cluster. A cluster is a block of swap disk
191 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
192 * free clusters are organized into a list. We fetch an entry from the list to
193 * get a free cluster.
195 * The data field stores next cluster if the cluster is free or cluster usage
196 * counter otherwise. The flags field determines if a cluster is free. This is
197 * protected by swap_info_struct.lock.
199 struct swap_cluster_info {
201 * Protect swap_cluster_info fields
202 * and swap_info_struct->swap_map
203 * elements correspond to the swap
206 unsigned int data:24;
207 unsigned int flags:8;
209 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
210 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
211 #define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */
214 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
215 * its own cluster and swapout sequentially. The purpose is to optimize swapout
218 struct percpu_cluster {
219 struct swap_cluster_info index; /* Current cluster index */
220 unsigned int next; /* Likely next allocation offset */
223 struct swap_cluster_list {
224 struct swap_cluster_info head;
225 struct swap_cluster_info tail;
229 * The in-memory structure used to track swap areas.
231 struct swap_info_struct {
232 unsigned long flags; /* SWP_USED etc: see above */
233 signed short prio; /* swap priority of this type */
234 struct plist_node list; /* entry in swap_active_head */
235 signed char type; /* strange name for an index */
236 unsigned int max; /* extent of the swap_map */
237 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
238 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
239 struct swap_cluster_list free_clusters; /* free clusters list */
240 unsigned int lowest_bit; /* index of first free in swap_map */
241 unsigned int highest_bit; /* index of last free in swap_map */
242 unsigned int pages; /* total of usable pages of swap */
243 unsigned int inuse_pages; /* number of those currently in use */
244 unsigned int cluster_next; /* likely index for next allocation */
245 unsigned int cluster_nr; /* countdown to next cluster search */
246 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
247 struct swap_extent *curr_swap_extent;
248 struct swap_extent first_swap_extent;
249 struct block_device *bdev; /* swap device or bdev of swap file */
250 struct file *swap_file; /* seldom referenced */
251 unsigned int old_block_size; /* seldom referenced */
252 #ifdef CONFIG_FRONTSWAP
253 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
254 atomic_t frontswap_pages; /* frontswap pages in-use counter */
257 * protect map scan related fields like
258 * swap_map, lowest_bit, highest_bit,
259 * inuse_pages, cluster_next,
260 * cluster_nr, lowest_alloc,
261 * highest_alloc, free/discard cluster
262 * list. other fields are only changed
263 * at swapon/swapoff, so are protected
264 * by swap_lock. changing flags need
265 * hold this lock and swap_lock. If
266 * both locks need hold, hold swap_lock
269 spinlock_t cont_lock; /*
270 * protect swap count continuation page
273 struct work_struct discard_work; /* discard worker */
274 struct swap_cluster_list discard_clusters; /* discard clusters list */
275 struct plist_node avail_lists[0]; /*
276 * entries in swap_avail_heads, one
278 * Must be last as the number of the
279 * array is nr_node_ids, which is not
280 * a fixed value so have to allocate
282 * And it has to be an array so that
283 * plist_for_each_* can work.
288 #define SWAP_RA_ORDER_CEILING 5
290 /* Avoid stack overflow, because we need to save part of page table */
291 #define SWAP_RA_ORDER_CEILING 3
292 #define SWAP_RA_PTE_CACHE_SIZE (1 << SWAP_RA_ORDER_CEILING)
295 struct vma_swap_readahead {
297 unsigned short offset;
298 unsigned short nr_pte;
302 pte_t ptes[SWAP_RA_PTE_CACHE_SIZE];
306 /* linux/mm/workingset.c */
307 void *workingset_eviction(struct address_space *mapping, struct page *page);
308 bool workingset_refault(void *shadow);
309 void workingset_activation(struct page *page);
311 /* Do not use directly, use workingset_lookup_update */
312 void workingset_update_node(struct radix_tree_node *node);
314 /* Returns workingset_update_node() if the mapping has shadow entries. */
315 #define workingset_lookup_update(mapping) \
317 radix_tree_update_node_t __helper = workingset_update_node; \
318 if (dax_mapping(mapping) || shmem_mapping(mapping)) \
323 /* linux/mm/page_alloc.c */
324 extern unsigned long totalram_pages;
325 extern unsigned long totalreserve_pages;
326 extern unsigned long nr_free_buffer_pages(void);
327 extern unsigned long nr_free_pagecache_pages(void);
329 /* Definition of global_zone_page_state not available yet */
330 #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES)
333 /* linux/mm/swap.c */
334 extern void lru_cache_add(struct page *);
335 extern void lru_cache_add_anon(struct page *page);
336 extern void lru_cache_add_file(struct page *page);
337 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
338 struct lruvec *lruvec, struct list_head *head);
339 extern void activate_page(struct page *);
340 extern void mark_page_accessed(struct page *);
341 extern void lru_add_drain(void);
342 extern void lru_add_drain_cpu(int cpu);
343 extern void lru_add_drain_all(void);
344 extern void rotate_reclaimable_page(struct page *page);
345 extern void deactivate_file_page(struct page *page);
346 extern void mark_page_lazyfree(struct page *page);
347 extern void swap_setup(void);
349 extern void lru_cache_add_active_or_unevictable(struct page *page,
350 struct vm_area_struct *vma);
352 /* linux/mm/vmscan.c */
353 extern unsigned long zone_reclaimable_pages(struct zone *zone);
354 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
355 gfp_t gfp_mask, nodemask_t *mask);
356 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
357 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
358 unsigned long nr_pages,
361 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
362 gfp_t gfp_mask, bool noswap,
364 unsigned long *nr_scanned);
365 extern unsigned long shrink_all_memory(unsigned long nr_pages);
366 extern int vm_swappiness;
367 extern int remove_mapping(struct address_space *mapping, struct page *page);
368 extern unsigned long vm_total_pages;
371 extern int node_reclaim_mode;
372 extern int sysctl_min_unmapped_ratio;
373 extern int sysctl_min_slab_ratio;
375 #define node_reclaim_mode 0
378 extern int page_evictable(struct page *page);
379 extern void check_move_unevictable_pages(struct page **, int nr_pages);
381 extern int kswapd_run(int nid);
382 extern void kswapd_stop(int nid);
386 #include <linux/blk_types.h> /* for bio_end_io_t */
388 /* linux/mm/page_io.c */
389 extern int swap_readpage(struct page *page, bool do_poll);
390 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
391 extern void end_swap_bio_write(struct bio *bio);
392 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
393 bio_end_io_t end_write_func);
394 extern int swap_set_page_dirty(struct page *page);
396 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
397 unsigned long nr_pages, sector_t start_block);
398 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
401 /* linux/mm/swap_state.c */
402 /* One swap address space for each 64M swap space */
403 #define SWAP_ADDRESS_SPACE_SHIFT 14
404 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
405 extern struct address_space *swapper_spaces[];
406 #define swap_address_space(entry) \
407 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
408 >> SWAP_ADDRESS_SPACE_SHIFT])
409 extern unsigned long total_swapcache_pages(void);
410 extern void show_swap_cache_info(void);
411 extern int add_to_swap(struct page *page);
412 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
413 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
414 extern void __delete_from_swap_cache(struct page *);
415 extern void delete_from_swap_cache(struct page *);
416 extern void free_page_and_swap_cache(struct page *);
417 extern void free_pages_and_swap_cache(struct page **, int);
418 extern struct page *lookup_swap_cache(swp_entry_t entry,
419 struct vm_area_struct *vma,
421 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
422 struct vm_area_struct *vma, unsigned long addr,
424 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
425 struct vm_area_struct *vma, unsigned long addr,
426 bool *new_page_allocated);
427 extern struct page *swap_cluster_readahead(swp_entry_t entry, gfp_t flag,
428 struct vm_fault *vmf);
429 extern struct page *swapin_readahead(swp_entry_t entry, gfp_t flag,
430 struct vm_fault *vmf);
432 /* linux/mm/swapfile.c */
433 extern atomic_long_t nr_swap_pages;
434 extern long total_swap_pages;
435 extern atomic_t nr_rotate_swap;
436 extern bool has_usable_swap(void);
438 /* Swap 50% full? Release swapcache more aggressively.. */
439 static inline bool vm_swap_full(void)
441 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
444 static inline long get_nr_swap_pages(void)
446 return atomic_long_read(&nr_swap_pages);
449 extern void si_swapinfo(struct sysinfo *);
450 extern swp_entry_t get_swap_page(struct page *page);
451 extern void put_swap_page(struct page *page, swp_entry_t entry);
452 extern swp_entry_t get_swap_page_of_type(int);
453 extern int get_swap_pages(int n, swp_entry_t swp_entries[], int entry_size);
454 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
455 extern void swap_shmem_alloc(swp_entry_t);
456 extern int swap_duplicate(swp_entry_t);
457 extern int swapcache_prepare(swp_entry_t);
458 extern void swap_free(swp_entry_t);
459 extern void swapcache_free_entries(swp_entry_t *entries, int n);
460 extern int free_swap_and_cache(swp_entry_t);
461 extern int swap_type_of(dev_t, sector_t, struct block_device **);
462 extern unsigned int count_swap_pages(int, int);
463 extern sector_t map_swap_page(struct page *, struct block_device **);
464 extern sector_t swapdev_block(int, pgoff_t);
465 extern int page_swapcount(struct page *);
466 extern int __swap_count(struct swap_info_struct *si, swp_entry_t entry);
467 extern int __swp_swapcount(swp_entry_t entry);
468 extern int swp_swapcount(swp_entry_t entry);
469 extern struct swap_info_struct *page_swap_info(struct page *);
470 extern struct swap_info_struct *swp_swap_info(swp_entry_t entry);
471 extern bool reuse_swap_page(struct page *, int *);
472 extern int try_to_free_swap(struct page *);
473 struct backing_dev_info;
474 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
475 extern void exit_swap_address_space(unsigned int type);
477 #else /* CONFIG_SWAP */
479 static inline int swap_readpage(struct page *page, bool do_poll)
484 static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry)
489 #define swap_address_space(entry) (NULL)
490 #define get_nr_swap_pages() 0L
491 #define total_swap_pages 0L
492 #define total_swapcache_pages() 0UL
493 #define vm_swap_full() 0
495 #define si_swapinfo(val) \
496 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
497 /* only sparc can not include linux/pagemap.h in this file
498 * so leave put_page and release_pages undeclared... */
499 #define free_page_and_swap_cache(page) \
501 #define free_pages_and_swap_cache(pages, nr) \
502 release_pages((pages), (nr));
504 static inline void show_swap_cache_info(void)
508 #define free_swap_and_cache(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
509 #define swapcache_prepare(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
511 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
516 static inline void swap_shmem_alloc(swp_entry_t swp)
520 static inline int swap_duplicate(swp_entry_t swp)
525 static inline void swap_free(swp_entry_t swp)
529 static inline void put_swap_page(struct page *page, swp_entry_t swp)
533 static inline struct page *swap_cluster_readahead(swp_entry_t entry,
534 gfp_t gfp_mask, struct vm_fault *vmf)
539 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
540 struct vm_fault *vmf)
545 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
550 static inline struct page *lookup_swap_cache(swp_entry_t swp,
551 struct vm_area_struct *vma,
557 static inline int add_to_swap(struct page *page)
562 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
568 static inline void __delete_from_swap_cache(struct page *page)
572 static inline void delete_from_swap_cache(struct page *page)
576 static inline int page_swapcount(struct page *page)
581 static inline int __swap_count(struct swap_info_struct *si, swp_entry_t entry)
586 static inline int __swp_swapcount(swp_entry_t entry)
591 static inline int swp_swapcount(swp_entry_t entry)
596 #define reuse_swap_page(page, total_map_swapcount) \
597 (page_trans_huge_mapcount(page, total_map_swapcount) == 1)
599 static inline int try_to_free_swap(struct page *page)
604 static inline swp_entry_t get_swap_page(struct page *page)
611 #endif /* CONFIG_SWAP */
613 #ifdef CONFIG_THP_SWAP
614 extern int split_swap_cluster(swp_entry_t entry);
616 static inline int split_swap_cluster(swp_entry_t entry)
623 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
625 /* Cgroup2 doesn't have per-cgroup swappiness */
626 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
627 return vm_swappiness;
630 if (mem_cgroup_disabled() || !memcg->css.parent)
631 return vm_swappiness;
633 return memcg->swappiness;
636 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
638 return vm_swappiness;
642 #if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
643 extern void mem_cgroup_throttle_swaprate(struct mem_cgroup *memcg, int node,
646 static inline void mem_cgroup_throttle_swaprate(struct mem_cgroup *memcg,
647 int node, gfp_t gfp_mask)
652 #ifdef CONFIG_MEMCG_SWAP
653 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
654 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
655 extern void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
656 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
657 extern bool mem_cgroup_swap_full(struct page *page);
659 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
663 static inline int mem_cgroup_try_charge_swap(struct page *page,
669 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
670 unsigned int nr_pages)
674 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
676 return get_nr_swap_pages();
679 static inline bool mem_cgroup_swap_full(struct page *page)
681 return vm_swap_full();
685 #endif /* __KERNEL__*/
686 #endif /* _LINUX_SWAP_H */