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 __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
159 #define MAX_SWAP_BADPAGES \
160 ((__swapoffset(magic.magic) - __swapoffset(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_ACTIVATED = (1 << 7), /* set after swap_activate success */
171 SWP_FS = (1 << 8), /* swap file goes through fs */
172 SWP_AREA_DISCARD = (1 << 9), /* single-time swap area discards */
173 SWP_PAGE_DISCARD = (1 << 10), /* freed swap page-cluster discards */
174 SWP_STABLE_WRITES = (1 << 11), /* no overwrite PG_writeback pages */
175 SWP_SYNCHRONOUS_IO = (1 << 12), /* synchronous IO is efficient */
176 /* add others here before... */
177 SWP_SCANNING = (1 << 13), /* refcount in scan_swap_map */
180 #define SWAP_CLUSTER_MAX 32UL
181 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
183 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
184 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
185 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
186 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
187 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
188 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
191 * We use this to track usage of a cluster. A cluster is a block of swap disk
192 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
193 * free clusters are organized into a list. We fetch an entry from the list to
194 * get a free cluster.
196 * The data field stores next cluster if the cluster is free or cluster usage
197 * counter otherwise. The flags field determines if a cluster is free. This is
198 * protected by swap_info_struct.lock.
200 struct swap_cluster_info {
202 * Protect swap_cluster_info fields
203 * and swap_info_struct->swap_map
204 * elements correspond to the swap
207 unsigned int data:24;
208 unsigned int flags:8;
210 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
211 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
212 #define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */
215 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
216 * its own cluster and swapout sequentially. The purpose is to optimize swapout
219 struct percpu_cluster {
220 struct swap_cluster_info index; /* Current cluster index */
221 unsigned int next; /* Likely next allocation offset */
224 struct swap_cluster_list {
225 struct swap_cluster_info head;
226 struct swap_cluster_info tail;
230 * The in-memory structure used to track swap areas.
232 struct swap_info_struct {
233 unsigned long flags; /* SWP_USED etc: see above */
234 signed short prio; /* swap priority of this type */
235 struct plist_node list; /* entry in swap_active_head */
236 struct plist_node avail_lists[MAX_NUMNODES];/* entry in swap_avail_heads */
237 signed char type; /* strange name for an index */
238 unsigned int max; /* extent of the swap_map */
239 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
240 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
241 struct swap_cluster_list free_clusters; /* free clusters list */
242 unsigned int lowest_bit; /* index of first free in swap_map */
243 unsigned int highest_bit; /* index of last free in swap_map */
244 unsigned int pages; /* total of usable pages of swap */
245 unsigned int inuse_pages; /* number of those currently in use */
246 unsigned int cluster_next; /* likely index for next allocation */
247 unsigned int cluster_nr; /* countdown to next cluster search */
248 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
249 struct swap_extent *curr_swap_extent;
250 struct swap_extent first_swap_extent;
251 struct block_device *bdev; /* swap device or bdev of swap file */
252 struct file *swap_file; /* seldom referenced */
253 unsigned int old_block_size; /* seldom referenced */
254 #ifdef CONFIG_FRONTSWAP
255 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
256 atomic_t frontswap_pages; /* frontswap pages in-use counter */
259 * protect map scan related fields like
260 * swap_map, lowest_bit, highest_bit,
261 * inuse_pages, cluster_next,
262 * cluster_nr, lowest_alloc,
263 * highest_alloc, free/discard cluster
264 * list. other fields are only changed
265 * at swapon/swapoff, so are protected
266 * by swap_lock. changing flags need
267 * hold this lock and swap_lock. If
268 * both locks need hold, hold swap_lock
271 spinlock_t cont_lock; /*
272 * protect swap count continuation page
275 struct work_struct discard_work; /* discard worker */
276 struct swap_cluster_list discard_clusters; /* discard clusters list */
280 #define SWAP_RA_ORDER_CEILING 5
282 /* Avoid stack overflow, because we need to save part of page table */
283 #define SWAP_RA_ORDER_CEILING 3
284 #define SWAP_RA_PTE_CACHE_SIZE (1 << SWAP_RA_ORDER_CEILING)
287 struct vma_swap_readahead {
289 unsigned short offset;
290 unsigned short nr_pte;
294 pte_t ptes[SWAP_RA_PTE_CACHE_SIZE];
298 /* linux/mm/workingset.c */
299 void *workingset_eviction(struct address_space *mapping, struct page *page);
300 void workingset_refault(struct page *page, void *shadow);
301 void workingset_activation(struct page *page);
303 /* Only track the nodes of mappings with shadow entries */
304 void workingset_update_node(struct xa_node *node);
305 #define mapping_set_update(xas, mapping) do { \
306 if (!dax_mapping(mapping) && !shmem_mapping(mapping)) \
307 xas_set_update(xas, workingset_update_node); \
310 /* linux/mm/page_alloc.c */
311 extern unsigned long totalram_pages;
312 extern unsigned long totalreserve_pages;
313 extern unsigned long nr_free_buffer_pages(void);
314 extern unsigned long nr_free_pagecache_pages(void);
316 /* Definition of global_zone_page_state not available yet */
317 #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES)
320 /* linux/mm/swap.c */
321 extern void lru_cache_add(struct page *);
322 extern void lru_cache_add_anon(struct page *page);
323 extern void lru_cache_add_file(struct page *page);
324 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
325 struct lruvec *lruvec, struct list_head *head);
326 extern void activate_page(struct page *);
327 extern void mark_page_accessed(struct page *);
328 extern void lru_add_drain(void);
329 extern void lru_add_drain_cpu(int cpu);
330 extern void lru_add_drain_all(void);
331 extern void rotate_reclaimable_page(struct page *page);
332 extern void deactivate_file_page(struct page *page);
333 extern void mark_page_lazyfree(struct page *page);
334 extern void swap_setup(void);
336 extern void lru_cache_add_active_or_unevictable(struct page *page,
337 struct vm_area_struct *vma);
339 /* linux/mm/vmscan.c */
340 extern unsigned long zone_reclaimable_pages(struct zone *zone);
341 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
342 gfp_t gfp_mask, nodemask_t *mask);
343 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
344 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
345 unsigned long nr_pages,
348 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
349 gfp_t gfp_mask, bool noswap,
351 unsigned long *nr_scanned);
352 extern unsigned long shrink_all_memory(unsigned long nr_pages);
353 extern int vm_swappiness;
354 extern int remove_mapping(struct address_space *mapping, struct page *page);
355 extern unsigned long vm_total_pages;
358 extern int node_reclaim_mode;
359 extern int sysctl_min_unmapped_ratio;
360 extern int sysctl_min_slab_ratio;
361 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
363 #define node_reclaim_mode 0
364 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
371 extern int page_evictable(struct page *page);
372 extern void check_move_unevictable_pages(struct page **, int nr_pages);
374 extern int kswapd_run(int nid);
375 extern void kswapd_stop(int nid);
379 #include <linux/blk_types.h> /* for bio_end_io_t */
381 /* linux/mm/page_io.c */
382 extern int swap_readpage(struct page *page, bool do_poll);
383 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
384 extern void end_swap_bio_write(struct bio *bio);
385 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
386 bio_end_io_t end_write_func);
387 extern int swap_set_page_dirty(struct page *page);
389 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
390 unsigned long nr_pages, sector_t start_block);
391 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
394 /* linux/mm/swap_state.c */
395 /* One swap address space for each 64M swap space */
396 #define SWAP_ADDRESS_SPACE_SHIFT 14
397 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
398 extern struct address_space *swapper_spaces[];
399 #define swap_address_space(entry) \
400 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
401 >> SWAP_ADDRESS_SPACE_SHIFT])
402 extern unsigned long total_swapcache_pages(void);
403 extern void show_swap_cache_info(void);
404 extern int add_to_swap(struct page *page);
405 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
406 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
407 extern void __delete_from_swap_cache(struct page *, swp_entry_t entry);
408 extern void delete_from_swap_cache(struct page *);
409 extern void free_page_and_swap_cache(struct page *);
410 extern void free_pages_and_swap_cache(struct page **, int);
411 extern struct page *lookup_swap_cache(swp_entry_t entry,
412 struct vm_area_struct *vma,
414 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
415 struct vm_area_struct *vma, unsigned long addr,
417 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
418 struct vm_area_struct *vma, unsigned long addr,
419 bool *new_page_allocated);
420 extern struct page *swap_cluster_readahead(swp_entry_t entry, gfp_t flag,
421 struct vm_fault *vmf);
422 extern struct page *swapin_readahead(swp_entry_t entry, gfp_t flag,
423 struct vm_fault *vmf);
425 /* linux/mm/swapfile.c */
426 extern atomic_long_t nr_swap_pages;
427 extern long total_swap_pages;
428 extern atomic_t nr_rotate_swap;
429 extern bool has_usable_swap(void);
431 /* Swap 50% full? Release swapcache more aggressively.. */
432 static inline bool vm_swap_full(void)
434 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
437 static inline long get_nr_swap_pages(void)
439 return atomic_long_read(&nr_swap_pages);
442 extern void si_swapinfo(struct sysinfo *);
443 extern swp_entry_t get_swap_page(struct page *page);
444 extern void put_swap_page(struct page *page, swp_entry_t entry);
445 extern swp_entry_t get_swap_page_of_type(int);
446 extern int get_swap_pages(int n, swp_entry_t swp_entries[], int entry_size);
447 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
448 extern void swap_shmem_alloc(swp_entry_t);
449 extern int swap_duplicate(swp_entry_t);
450 extern int swapcache_prepare(swp_entry_t);
451 extern void swap_free(swp_entry_t);
452 extern void swapcache_free_entries(swp_entry_t *entries, int n);
453 extern int free_swap_and_cache(swp_entry_t);
454 extern int swap_type_of(dev_t, sector_t, struct block_device **);
455 extern unsigned int count_swap_pages(int, int);
456 extern sector_t map_swap_page(struct page *, struct block_device **);
457 extern sector_t swapdev_block(int, pgoff_t);
458 extern int page_swapcount(struct page *);
459 extern int __swap_count(struct swap_info_struct *si, swp_entry_t entry);
460 extern int __swp_swapcount(swp_entry_t entry);
461 extern int swp_swapcount(swp_entry_t entry);
462 extern struct swap_info_struct *page_swap_info(struct page *);
463 extern struct swap_info_struct *swp_swap_info(swp_entry_t entry);
464 extern bool reuse_swap_page(struct page *, int *);
465 extern int try_to_free_swap(struct page *);
466 struct backing_dev_info;
467 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
468 extern void exit_swap_address_space(unsigned int type);
470 #else /* CONFIG_SWAP */
472 static inline int swap_readpage(struct page *page, bool do_poll)
477 static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry)
482 #define swap_address_space(entry) (NULL)
483 #define get_nr_swap_pages() 0L
484 #define total_swap_pages 0L
485 #define total_swapcache_pages() 0UL
486 #define vm_swap_full() 0
488 #define si_swapinfo(val) \
489 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
490 /* only sparc can not include linux/pagemap.h in this file
491 * so leave put_page and release_pages undeclared... */
492 #define free_page_and_swap_cache(page) \
494 #define free_pages_and_swap_cache(pages, nr) \
495 release_pages((pages), (nr));
497 static inline void show_swap_cache_info(void)
501 #define free_swap_and_cache(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
502 #define swapcache_prepare(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
504 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
509 static inline void swap_shmem_alloc(swp_entry_t swp)
513 static inline int swap_duplicate(swp_entry_t swp)
518 static inline void swap_free(swp_entry_t swp)
522 static inline void put_swap_page(struct page *page, swp_entry_t swp)
526 static inline struct page *swap_cluster_readahead(swp_entry_t entry,
527 gfp_t gfp_mask, struct vm_fault *vmf)
532 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
533 struct vm_fault *vmf)
538 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
543 static inline struct page *lookup_swap_cache(swp_entry_t swp,
544 struct vm_area_struct *vma,
550 static inline int add_to_swap(struct page *page)
555 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
561 static inline void __delete_from_swap_cache(struct page *page,
566 static inline void delete_from_swap_cache(struct page *page)
570 static inline int page_swapcount(struct page *page)
575 static inline int __swap_count(struct swap_info_struct *si, swp_entry_t entry)
580 static inline int __swp_swapcount(swp_entry_t entry)
585 static inline int swp_swapcount(swp_entry_t entry)
590 #define reuse_swap_page(page, total_map_swapcount) \
591 (page_trans_huge_mapcount(page, total_map_swapcount) == 1)
593 static inline int try_to_free_swap(struct page *page)
598 static inline swp_entry_t get_swap_page(struct page *page)
605 #endif /* CONFIG_SWAP */
607 #ifdef CONFIG_THP_SWAP
608 extern int split_swap_cluster(swp_entry_t entry);
610 static inline int split_swap_cluster(swp_entry_t entry)
617 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
619 /* Cgroup2 doesn't have per-cgroup swappiness */
620 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
621 return vm_swappiness;
624 if (mem_cgroup_disabled() || !memcg->css.parent)
625 return vm_swappiness;
627 return memcg->swappiness;
630 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
632 return vm_swappiness;
636 #if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
637 extern void mem_cgroup_throttle_swaprate(struct mem_cgroup *memcg, int node,
640 static inline void mem_cgroup_throttle_swaprate(struct mem_cgroup *memcg,
641 int node, gfp_t gfp_mask)
646 #ifdef CONFIG_MEMCG_SWAP
647 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
648 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
649 extern void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
650 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
651 extern bool mem_cgroup_swap_full(struct page *page);
653 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
657 static inline int mem_cgroup_try_charge_swap(struct page *page,
663 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
664 unsigned int nr_pages)
668 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
670 return get_nr_swap_pages();
673 static inline bool mem_cgroup_swap_full(struct page *page)
675 return vm_swap_full();
679 #endif /* __KERNEL__*/
680 #endif /* _LINUX_SWAP_H */