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 list_head list;
154 sector_t start_block;
158 * Max bad pages in the new format..
160 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
161 #define MAX_SWAP_BADPAGES \
162 ((__swapoffset(magic.magic) - __swapoffset(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 /* add others here before... */
179 SWP_SCANNING = (1 << 13), /* refcount in scan_swap_map */
182 #define SWAP_CLUSTER_MAX 32UL
183 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
185 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
186 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
187 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
188 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
189 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
190 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
193 * We use this to track usage of a cluster. A cluster is a block of swap disk
194 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
195 * free clusters are organized into a list. We fetch an entry from the list to
196 * get a free cluster.
198 * The data field stores next cluster if the cluster is free or cluster usage
199 * counter otherwise. The flags field determines if a cluster is free. This is
200 * protected by swap_info_struct.lock.
202 struct swap_cluster_info {
204 * Protect swap_cluster_info fields
205 * and swap_info_struct->swap_map
206 * elements correspond to the swap
209 unsigned int data:24;
210 unsigned int flags:8;
212 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
213 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
214 #define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */
217 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
218 * its own cluster and swapout sequentially. The purpose is to optimize swapout
221 struct percpu_cluster {
222 struct swap_cluster_info index; /* Current cluster index */
223 unsigned int next; /* Likely next allocation offset */
226 struct swap_cluster_list {
227 struct swap_cluster_info head;
228 struct swap_cluster_info tail;
232 * The in-memory structure used to track swap areas.
234 struct swap_info_struct {
235 unsigned long flags; /* SWP_USED etc: see above */
236 signed short prio; /* swap priority of this type */
237 struct plist_node list; /* entry in swap_active_head */
238 struct plist_node avail_lists[MAX_NUMNODES];/* entry in swap_avail_heads */
239 signed char type; /* strange name for an index */
240 unsigned int max; /* extent of the swap_map */
241 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
242 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
243 struct swap_cluster_list free_clusters; /* free clusters list */
244 unsigned int lowest_bit; /* index of first free in swap_map */
245 unsigned int highest_bit; /* index of last free in swap_map */
246 unsigned int pages; /* total of usable pages of swap */
247 unsigned int inuse_pages; /* number of those currently in use */
248 unsigned int cluster_next; /* likely index for next allocation */
249 unsigned int cluster_nr; /* countdown to next cluster search */
250 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
251 struct swap_extent *curr_swap_extent;
252 struct swap_extent first_swap_extent;
253 struct block_device *bdev; /* swap device or bdev of swap file */
254 struct file *swap_file; /* seldom referenced */
255 unsigned int old_block_size; /* seldom referenced */
256 #ifdef CONFIG_FRONTSWAP
257 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
258 atomic_t frontswap_pages; /* frontswap pages in-use counter */
261 * protect map scan related fields like
262 * swap_map, lowest_bit, highest_bit,
263 * inuse_pages, cluster_next,
264 * cluster_nr, lowest_alloc,
265 * highest_alloc, free/discard cluster
266 * list. other fields are only changed
267 * at swapon/swapoff, so are protected
268 * by swap_lock. changing flags need
269 * hold this lock and swap_lock. If
270 * both locks need hold, hold swap_lock
273 spinlock_t cont_lock; /*
274 * protect swap count continuation page
277 struct work_struct discard_work; /* discard worker */
278 struct swap_cluster_list discard_clusters; /* discard clusters list */
282 #define SWAP_RA_ORDER_CEILING 5
284 /* Avoid stack overflow, because we need to save part of page table */
285 #define SWAP_RA_ORDER_CEILING 3
286 #define SWAP_RA_PTE_CACHE_SIZE (1 << SWAP_RA_ORDER_CEILING)
289 struct vma_swap_readahead {
291 unsigned short offset;
292 unsigned short nr_pte;
296 pte_t ptes[SWAP_RA_PTE_CACHE_SIZE];
300 /* linux/mm/workingset.c */
301 void *workingset_eviction(struct address_space *mapping, struct page *page);
302 void workingset_refault(struct page *page, void *shadow);
303 void workingset_activation(struct page *page);
305 /* Only track the nodes of mappings with shadow entries */
306 void workingset_update_node(struct xa_node *node);
307 #define mapping_set_update(xas, mapping) do { \
308 if (!dax_mapping(mapping) && !shmem_mapping(mapping)) \
309 xas_set_update(xas, workingset_update_node); \
312 /* linux/mm/page_alloc.c */
313 extern unsigned long totalram_pages;
314 extern unsigned long totalreserve_pages;
315 extern unsigned long nr_free_buffer_pages(void);
316 extern unsigned long nr_free_pagecache_pages(void);
318 /* Definition of global_zone_page_state not available yet */
319 #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES)
322 /* linux/mm/swap.c */
323 extern void lru_cache_add(struct page *);
324 extern void lru_cache_add_anon(struct page *page);
325 extern void lru_cache_add_file(struct page *page);
326 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
327 struct lruvec *lruvec, struct list_head *head);
328 extern void activate_page(struct page *);
329 extern void mark_page_accessed(struct page *);
330 extern void lru_add_drain(void);
331 extern void lru_add_drain_cpu(int cpu);
332 extern void lru_add_drain_all(void);
333 extern void rotate_reclaimable_page(struct page *page);
334 extern void deactivate_file_page(struct page *page);
335 extern void mark_page_lazyfree(struct page *page);
336 extern void swap_setup(void);
338 extern void lru_cache_add_active_or_unevictable(struct page *page,
339 struct vm_area_struct *vma);
341 /* linux/mm/vmscan.c */
342 extern unsigned long zone_reclaimable_pages(struct zone *zone);
343 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
344 gfp_t gfp_mask, nodemask_t *mask);
345 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
346 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
347 unsigned long nr_pages,
350 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
351 gfp_t gfp_mask, bool noswap,
353 unsigned long *nr_scanned);
354 extern unsigned long shrink_all_memory(unsigned long nr_pages);
355 extern int vm_swappiness;
356 extern int remove_mapping(struct address_space *mapping, struct page *page);
357 extern unsigned long vm_total_pages;
360 extern int node_reclaim_mode;
361 extern int sysctl_min_unmapped_ratio;
362 extern int sysctl_min_slab_ratio;
363 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
365 #define node_reclaim_mode 0
366 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
373 extern int page_evictable(struct page *page);
374 extern void check_move_unevictable_pages(struct pagevec *pvec);
376 extern int kswapd_run(int nid);
377 extern void kswapd_stop(int nid);
381 #include <linux/blk_types.h> /* for bio_end_io_t */
383 /* linux/mm/page_io.c */
384 extern int swap_readpage(struct page *page, bool do_poll);
385 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
386 extern void end_swap_bio_write(struct bio *bio);
387 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
388 bio_end_io_t end_write_func);
389 extern int swap_set_page_dirty(struct page *page);
391 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
392 unsigned long nr_pages, sector_t start_block);
393 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
396 /* linux/mm/swap_state.c */
397 /* One swap address space for each 64M swap space */
398 #define SWAP_ADDRESS_SPACE_SHIFT 14
399 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
400 extern struct address_space *swapper_spaces[];
401 #define swap_address_space(entry) \
402 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
403 >> SWAP_ADDRESS_SPACE_SHIFT])
404 extern unsigned long total_swapcache_pages(void);
405 extern void show_swap_cache_info(void);
406 extern int add_to_swap(struct page *page);
407 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
408 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
409 extern void __delete_from_swap_cache(struct page *, swp_entry_t entry);
410 extern void delete_from_swap_cache(struct page *);
411 extern void free_page_and_swap_cache(struct page *);
412 extern void free_pages_and_swap_cache(struct page **, int);
413 extern struct page *lookup_swap_cache(swp_entry_t entry,
414 struct vm_area_struct *vma,
416 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
417 struct vm_area_struct *vma, unsigned long addr,
419 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
420 struct vm_area_struct *vma, unsigned long addr,
421 bool *new_page_allocated);
422 extern struct page *swap_cluster_readahead(swp_entry_t entry, gfp_t flag,
423 struct vm_fault *vmf);
424 extern struct page *swapin_readahead(swp_entry_t entry, gfp_t flag,
425 struct vm_fault *vmf);
427 /* linux/mm/swapfile.c */
428 extern atomic_long_t nr_swap_pages;
429 extern long total_swap_pages;
430 extern atomic_t nr_rotate_swap;
431 extern bool has_usable_swap(void);
433 /* Swap 50% full? Release swapcache more aggressively.. */
434 static inline bool vm_swap_full(void)
436 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
439 static inline long get_nr_swap_pages(void)
441 return atomic_long_read(&nr_swap_pages);
444 extern void si_swapinfo(struct sysinfo *);
445 extern swp_entry_t get_swap_page(struct page *page);
446 extern void put_swap_page(struct page *page, swp_entry_t entry);
447 extern swp_entry_t get_swap_page_of_type(int);
448 extern int get_swap_pages(int n, swp_entry_t swp_entries[], int entry_size);
449 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
450 extern void swap_shmem_alloc(swp_entry_t);
451 extern int swap_duplicate(swp_entry_t);
452 extern int swapcache_prepare(swp_entry_t);
453 extern void swap_free(swp_entry_t);
454 extern void swapcache_free_entries(swp_entry_t *entries, int n);
455 extern int free_swap_and_cache(swp_entry_t);
456 extern int swap_type_of(dev_t, sector_t, struct block_device **);
457 extern unsigned int count_swap_pages(int, int);
458 extern sector_t map_swap_page(struct page *, struct block_device **);
459 extern sector_t swapdev_block(int, pgoff_t);
460 extern int page_swapcount(struct page *);
461 extern int __swap_count(struct swap_info_struct *si, swp_entry_t entry);
462 extern int __swp_swapcount(swp_entry_t entry);
463 extern int swp_swapcount(swp_entry_t entry);
464 extern struct swap_info_struct *page_swap_info(struct page *);
465 extern struct swap_info_struct *swp_swap_info(swp_entry_t entry);
466 extern bool reuse_swap_page(struct page *, int *);
467 extern int try_to_free_swap(struct page *);
468 struct backing_dev_info;
469 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
470 extern void exit_swap_address_space(unsigned int type);
472 #else /* CONFIG_SWAP */
474 static inline int swap_readpage(struct page *page, bool do_poll)
479 static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry)
484 #define swap_address_space(entry) (NULL)
485 #define get_nr_swap_pages() 0L
486 #define total_swap_pages 0L
487 #define total_swapcache_pages() 0UL
488 #define vm_swap_full() 0
490 #define si_swapinfo(val) \
491 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
492 /* only sparc can not include linux/pagemap.h in this file
493 * so leave put_page and release_pages undeclared... */
494 #define free_page_and_swap_cache(page) \
496 #define free_pages_and_swap_cache(pages, nr) \
497 release_pages((pages), (nr));
499 static inline void show_swap_cache_info(void)
503 #define free_swap_and_cache(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
504 #define swapcache_prepare(e) ({(is_migration_entry(e) || is_device_private_entry(e));})
506 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
511 static inline void swap_shmem_alloc(swp_entry_t swp)
515 static inline int swap_duplicate(swp_entry_t swp)
520 static inline void swap_free(swp_entry_t swp)
524 static inline void put_swap_page(struct page *page, swp_entry_t swp)
528 static inline struct page *swap_cluster_readahead(swp_entry_t entry,
529 gfp_t gfp_mask, struct vm_fault *vmf)
534 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
535 struct vm_fault *vmf)
540 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
545 static inline struct page *lookup_swap_cache(swp_entry_t swp,
546 struct vm_area_struct *vma,
552 static inline int add_to_swap(struct page *page)
557 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
563 static inline void __delete_from_swap_cache(struct page *page,
568 static inline void delete_from_swap_cache(struct page *page)
572 static inline int page_swapcount(struct page *page)
577 static inline int __swap_count(struct swap_info_struct *si, swp_entry_t entry)
582 static inline int __swp_swapcount(swp_entry_t entry)
587 static inline int swp_swapcount(swp_entry_t entry)
592 #define reuse_swap_page(page, total_map_swapcount) \
593 (page_trans_huge_mapcount(page, total_map_swapcount) == 1)
595 static inline int try_to_free_swap(struct page *page)
600 static inline swp_entry_t get_swap_page(struct page *page)
607 #endif /* CONFIG_SWAP */
609 #ifdef CONFIG_THP_SWAP
610 extern int split_swap_cluster(swp_entry_t entry);
612 static inline int split_swap_cluster(swp_entry_t entry)
619 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
621 /* Cgroup2 doesn't have per-cgroup swappiness */
622 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
623 return vm_swappiness;
626 if (mem_cgroup_disabled() || !memcg->css.parent)
627 return vm_swappiness;
629 return memcg->swappiness;
632 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
634 return vm_swappiness;
638 #if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
639 extern void mem_cgroup_throttle_swaprate(struct mem_cgroup *memcg, int node,
642 static inline void mem_cgroup_throttle_swaprate(struct mem_cgroup *memcg,
643 int node, gfp_t gfp_mask)
648 #ifdef CONFIG_MEMCG_SWAP
649 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
650 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
651 extern void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
652 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
653 extern bool mem_cgroup_swap_full(struct page *page);
655 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
659 static inline int mem_cgroup_try_charge_swap(struct page *page,
665 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
666 unsigned int nr_pages)
670 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
672 return get_nr_swap_pages();
675 static inline bool mem_cgroup_swap_full(struct page *page)
677 return vm_swap_full();
681 #endif /* __KERNEL__*/
682 #endif /* _LINUX_SWAP_H */