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>
12 #include <linux/atomic.h>
13 #include <linux/page-flags.h>
16 struct notifier_block;
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 */
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)
32 static inline int current_is_kswapd(void)
34 return current->flags & PF_KSWAPD;
38 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
39 * be swapped to. The swap type and the offset into that swap type are
40 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
41 * for the type means that the maximum number of swapcache pages is 27 bits
42 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
43 * the type/offset into the pte as 5/27 as well.
45 #define MAX_SWAPFILES_SHIFT 5
48 * Use some of the swap files numbers for other purposes. This
49 * is a convenient way to hook into the VM to trigger special
54 * NUMA node memory migration support
56 #ifdef CONFIG_MIGRATION
57 #define SWP_MIGRATION_NUM 2
58 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
59 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
61 #define SWP_MIGRATION_NUM 0
65 * Handling of hardware poisoned pages with memory corruption.
67 #ifdef CONFIG_MEMORY_FAILURE
68 #define SWP_HWPOISON_NUM 1
69 #define SWP_HWPOISON MAX_SWAPFILES
71 #define SWP_HWPOISON_NUM 0
74 #define MAX_SWAPFILES \
75 ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
78 * Magic header for a swap area. The first part of the union is
79 * what the swap magic looks like for the old (limited to 128MB)
80 * swap area format, the second part of the union adds - in the
81 * old reserved area - some extra information. Note that the first
82 * kilobyte is reserved for boot loader or disk label stuff...
84 * Having the magic at the end of the PAGE_SIZE makes detecting swap
85 * areas somewhat tricky on machines that support multiple page sizes.
86 * For 2.5 we'll probably want to move the magic to just beyond the
91 char reserved[PAGE_SIZE - 10];
92 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
95 char bootbits[1024]; /* Space for disklabel etc. */
99 unsigned char sws_uuid[16];
100 unsigned char sws_volume[16];
107 * current->reclaim_state points to one of these when a task is running
110 struct reclaim_state {
111 unsigned long reclaimed_slab;
116 struct address_space;
118 struct writeback_control;
122 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
123 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
124 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
125 * from setup, they're handled identically.
127 * We always assume that blocks are of size PAGE_SIZE.
130 struct list_head list;
133 sector_t start_block;
137 * Max bad pages in the new format..
139 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
140 #define MAX_SWAP_BADPAGES \
141 ((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
144 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
145 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
146 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
147 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
148 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
149 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
150 SWP_BLKDEV = (1 << 6), /* its a block device */
151 SWP_FILE = (1 << 7), /* set after swap_activate success */
152 SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
153 SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
154 SWP_STABLE_WRITES = (1 << 10), /* no overwrite PG_writeback pages */
155 /* add others here before... */
156 SWP_SCANNING = (1 << 11), /* refcount in scan_swap_map */
159 #define SWAP_CLUSTER_MAX 32UL
160 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
162 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
163 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
164 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
165 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
166 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
167 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
170 * We use this to track usage of a cluster. A cluster is a block of swap disk
171 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
172 * free clusters are organized into a list. We fetch an entry from the list to
173 * get a free cluster.
175 * The data field stores next cluster if the cluster is free or cluster usage
176 * counter otherwise. The flags field determines if a cluster is free. This is
177 * protected by swap_info_struct.lock.
179 struct swap_cluster_info {
181 * Protect swap_cluster_info fields
182 * and swap_info_struct->swap_map
183 * elements correspond to the swap
186 unsigned int data:24;
187 unsigned int flags:8;
189 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
190 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
191 #define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */
194 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
195 * its own cluster and swapout sequentially. The purpose is to optimize swapout
198 struct percpu_cluster {
199 struct swap_cluster_info index; /* Current cluster index */
200 unsigned int next; /* Likely next allocation offset */
203 struct swap_cluster_list {
204 struct swap_cluster_info head;
205 struct swap_cluster_info tail;
209 * The in-memory structure used to track swap areas.
211 struct swap_info_struct {
212 unsigned long flags; /* SWP_USED etc: see above */
213 signed short prio; /* swap priority of this type */
214 struct plist_node list; /* entry in swap_active_head */
215 struct plist_node avail_lists[MAX_NUMNODES];/* entry in swap_avail_heads */
216 signed char type; /* strange name for an index */
217 unsigned int max; /* extent of the swap_map */
218 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
219 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
220 struct swap_cluster_list free_clusters; /* free clusters list */
221 unsigned int lowest_bit; /* index of first free in swap_map */
222 unsigned int highest_bit; /* index of last free in swap_map */
223 unsigned int pages; /* total of usable pages of swap */
224 unsigned int inuse_pages; /* number of those currently in use */
225 unsigned int cluster_next; /* likely index for next allocation */
226 unsigned int cluster_nr; /* countdown to next cluster search */
227 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
228 struct swap_extent *curr_swap_extent;
229 struct swap_extent first_swap_extent;
230 struct block_device *bdev; /* swap device or bdev of swap file */
231 struct file *swap_file; /* seldom referenced */
232 unsigned int old_block_size; /* seldom referenced */
233 #ifdef CONFIG_FRONTSWAP
234 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
235 atomic_t frontswap_pages; /* frontswap pages in-use counter */
238 * protect map scan related fields like
239 * swap_map, lowest_bit, highest_bit,
240 * inuse_pages, cluster_next,
241 * cluster_nr, lowest_alloc,
242 * highest_alloc, free/discard cluster
243 * list. other fields are only changed
244 * at swapon/swapoff, so are protected
245 * by swap_lock. changing flags need
246 * hold this lock and swap_lock. If
247 * both locks need hold, hold swap_lock
250 struct work_struct discard_work; /* discard worker */
251 struct swap_cluster_list discard_clusters; /* discard clusters list */
255 #define SWAP_RA_ORDER_CEILING 5
257 /* Avoid stack overflow, because we need to save part of page table */
258 #define SWAP_RA_ORDER_CEILING 3
259 #define SWAP_RA_PTE_CACHE_SIZE (1 << SWAP_RA_ORDER_CEILING)
262 struct vma_swap_readahead {
264 unsigned short offset;
265 unsigned short nr_pte;
269 pte_t ptes[SWAP_RA_PTE_CACHE_SIZE];
273 /* linux/mm/workingset.c */
274 void *workingset_eviction(struct address_space *mapping, struct page *page);
275 bool workingset_refault(void *shadow);
276 void workingset_activation(struct page *page);
277 void workingset_update_node(struct radix_tree_node *node, void *private);
279 /* linux/mm/page_alloc.c */
280 extern unsigned long totalram_pages;
281 extern unsigned long totalreserve_pages;
282 extern unsigned long nr_free_buffer_pages(void);
283 extern unsigned long nr_free_pagecache_pages(void);
285 /* Definition of global_zone_page_state not available yet */
286 #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES)
289 /* linux/mm/swap.c */
290 extern void lru_cache_add(struct page *);
291 extern void lru_cache_add_anon(struct page *page);
292 extern void lru_cache_add_file(struct page *page);
293 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
294 struct lruvec *lruvec, struct list_head *head);
295 extern void activate_page(struct page *);
296 extern void mark_page_accessed(struct page *);
297 extern void lru_add_drain(void);
298 extern void lru_add_drain_cpu(int cpu);
299 extern void lru_add_drain_all(void);
300 extern void lru_add_drain_all_cpuslocked(void);
301 extern void rotate_reclaimable_page(struct page *page);
302 extern void deactivate_file_page(struct page *page);
303 extern void mark_page_lazyfree(struct page *page);
304 extern void swap_setup(void);
306 extern void add_page_to_unevictable_list(struct page *page);
308 extern void lru_cache_add_active_or_unevictable(struct page *page,
309 struct vm_area_struct *vma);
311 /* linux/mm/vmscan.c */
312 extern unsigned long zone_reclaimable_pages(struct zone *zone);
313 extern unsigned long pgdat_reclaimable_pages(struct pglist_data *pgdat);
314 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
315 gfp_t gfp_mask, nodemask_t *mask);
316 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
317 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
318 unsigned long nr_pages,
321 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
322 gfp_t gfp_mask, bool noswap,
324 unsigned long *nr_scanned);
325 extern unsigned long shrink_all_memory(unsigned long nr_pages);
326 extern int vm_swappiness;
327 extern int remove_mapping(struct address_space *mapping, struct page *page);
328 extern unsigned long vm_total_pages;
331 extern int node_reclaim_mode;
332 extern int sysctl_min_unmapped_ratio;
333 extern int sysctl_min_slab_ratio;
334 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
336 #define node_reclaim_mode 0
337 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
344 extern int page_evictable(struct page *page);
345 extern void check_move_unevictable_pages(struct page **, int nr_pages);
347 extern int kswapd_run(int nid);
348 extern void kswapd_stop(int nid);
352 #include <linux/blk_types.h> /* for bio_end_io_t */
354 /* linux/mm/page_io.c */
355 extern int swap_readpage(struct page *page, bool do_poll);
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);
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 *,
367 /* linux/mm/swap_state.c */
368 /* One swap address space for each 64M swap space */
369 #define SWAP_ADDRESS_SPACE_SHIFT 14
370 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
371 extern struct address_space *swapper_spaces[];
372 extern bool swap_vma_readahead;
373 #define swap_address_space(entry) \
374 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
375 >> SWAP_ADDRESS_SPACE_SHIFT])
376 extern unsigned long total_swapcache_pages(void);
377 extern void show_swap_cache_info(void);
378 extern int add_to_swap(struct page *page);
379 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
380 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
381 extern void __delete_from_swap_cache(struct page *);
382 extern void delete_from_swap_cache(struct page *);
383 extern void free_page_and_swap_cache(struct page *);
384 extern void free_pages_and_swap_cache(struct page **, int);
385 extern struct page *lookup_swap_cache(swp_entry_t entry,
386 struct vm_area_struct *vma,
388 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
389 struct vm_area_struct *vma, unsigned long addr,
391 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
392 struct vm_area_struct *vma, unsigned long addr,
393 bool *new_page_allocated);
394 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
395 struct vm_area_struct *vma, unsigned long addr);
397 extern struct page *swap_readahead_detect(struct vm_fault *vmf,
398 struct vma_swap_readahead *swap_ra);
399 extern struct page *do_swap_page_readahead(swp_entry_t fentry, gfp_t gfp_mask,
400 struct vm_fault *vmf,
401 struct vma_swap_readahead *swap_ra);
403 /* linux/mm/swapfile.c */
404 extern atomic_long_t nr_swap_pages;
405 extern long total_swap_pages;
406 extern atomic_t nr_rotate_swap;
407 extern bool has_usable_swap(void);
409 static inline bool swap_use_vma_readahead(void)
411 return READ_ONCE(swap_vma_readahead) && !atomic_read(&nr_rotate_swap);
414 /* Swap 50% full? Release swapcache more aggressively.. */
415 static inline bool vm_swap_full(void)
417 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
420 static inline long get_nr_swap_pages(void)
422 return atomic_long_read(&nr_swap_pages);
425 extern void si_swapinfo(struct sysinfo *);
426 extern swp_entry_t get_swap_page(struct page *page);
427 extern void put_swap_page(struct page *page, swp_entry_t entry);
428 extern swp_entry_t get_swap_page_of_type(int);
429 extern int get_swap_pages(int n, bool cluster, swp_entry_t swp_entries[]);
430 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
431 extern void swap_shmem_alloc(swp_entry_t);
432 extern int swap_duplicate(swp_entry_t);
433 extern int swapcache_prepare(swp_entry_t);
434 extern void swap_free(swp_entry_t);
435 extern void swapcache_free_entries(swp_entry_t *entries, int n);
436 extern int free_swap_and_cache(swp_entry_t);
437 extern int swap_type_of(dev_t, sector_t, struct block_device **);
438 extern unsigned int count_swap_pages(int, int);
439 extern sector_t map_swap_page(struct page *, struct block_device **);
440 extern sector_t swapdev_block(int, pgoff_t);
441 extern int page_swapcount(struct page *);
442 extern int __swp_swapcount(swp_entry_t entry);
443 extern int swp_swapcount(swp_entry_t entry);
444 extern struct swap_info_struct *page_swap_info(struct page *);
445 extern bool reuse_swap_page(struct page *, int *);
446 extern int try_to_free_swap(struct page *);
447 struct backing_dev_info;
448 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
449 extern void exit_swap_address_space(unsigned int type);
451 #else /* CONFIG_SWAP */
453 #define swap_address_space(entry) (NULL)
454 #define get_nr_swap_pages() 0L
455 #define total_swap_pages 0L
456 #define total_swapcache_pages() 0UL
457 #define vm_swap_full() 0
459 #define si_swapinfo(val) \
460 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
461 /* only sparc can not include linux/pagemap.h in this file
462 * so leave put_page and release_pages undeclared... */
463 #define free_page_and_swap_cache(page) \
465 #define free_pages_and_swap_cache(pages, nr) \
466 release_pages((pages), (nr), false);
468 static inline void show_swap_cache_info(void)
472 #define free_swap_and_cache(swp) is_migration_entry(swp)
473 #define swapcache_prepare(swp) is_migration_entry(swp)
475 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
480 static inline void swap_shmem_alloc(swp_entry_t swp)
484 static inline int swap_duplicate(swp_entry_t swp)
489 static inline void swap_free(swp_entry_t swp)
493 static inline void put_swap_page(struct page *page, swp_entry_t swp)
497 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
498 struct vm_area_struct *vma, unsigned long addr)
503 static inline bool swap_use_vma_readahead(void)
508 static inline struct page *swap_readahead_detect(
509 struct vm_fault *vmf, struct vma_swap_readahead *swap_ra)
514 static inline struct page *do_swap_page_readahead(
515 swp_entry_t fentry, gfp_t gfp_mask,
516 struct vm_fault *vmf, struct vma_swap_readahead *swap_ra)
521 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
526 static inline struct page *lookup_swap_cache(swp_entry_t swp,
527 struct vm_area_struct *vma,
533 static inline int add_to_swap(struct page *page)
538 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
544 static inline void __delete_from_swap_cache(struct page *page)
548 static inline void delete_from_swap_cache(struct page *page)
552 static inline int page_swapcount(struct page *page)
557 static inline int __swp_swapcount(swp_entry_t entry)
562 static inline int swp_swapcount(swp_entry_t entry)
567 #define reuse_swap_page(page, total_map_swapcount) \
568 (page_trans_huge_mapcount(page, total_map_swapcount) == 1)
570 static inline int try_to_free_swap(struct page *page)
575 static inline swp_entry_t get_swap_page(struct page *page)
582 #endif /* CONFIG_SWAP */
584 #ifdef CONFIG_THP_SWAP
585 extern int split_swap_cluster(swp_entry_t entry);
587 static inline int split_swap_cluster(swp_entry_t entry)
594 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
596 /* Cgroup2 doesn't have per-cgroup swappiness */
597 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
598 return vm_swappiness;
601 if (mem_cgroup_disabled() || !memcg->css.parent)
602 return vm_swappiness;
604 return memcg->swappiness;
608 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
610 return vm_swappiness;
614 #ifdef CONFIG_MEMCG_SWAP
615 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
616 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
617 extern void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
618 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
619 extern bool mem_cgroup_swap_full(struct page *page);
621 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
625 static inline int mem_cgroup_try_charge_swap(struct page *page,
631 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
632 unsigned int nr_pages)
636 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
638 return get_nr_swap_pages();
641 static inline bool mem_cgroup_swap_full(struct page *page)
643 return vm_swap_full();
647 #endif /* __KERNEL__*/
648 #endif /* _LINUX_SWAP_H */