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)
31 static inline int current_is_kswapd(void)
33 return current->flags & PF_KSWAPD;
37 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
38 * be swapped to. The swap type and the offset into that swap type are
39 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
40 * for the type means that the maximum number of swapcache pages is 27 bits
41 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
42 * the type/offset into the pte as 5/27 as well.
44 #define MAX_SWAPFILES_SHIFT 5
47 * Use some of the swap files numbers for other purposes. This
48 * is a convenient way to hook into the VM to trigger special
53 * NUMA node memory migration support
55 #ifdef CONFIG_MIGRATION
56 #define SWP_MIGRATION_NUM 2
57 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
58 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
60 #define SWP_MIGRATION_NUM 0
64 * Handling of hardware poisoned pages with memory corruption.
66 #ifdef CONFIG_MEMORY_FAILURE
67 #define SWP_HWPOISON_NUM 1
68 #define SWP_HWPOISON MAX_SWAPFILES
70 #define SWP_HWPOISON_NUM 0
73 #define MAX_SWAPFILES \
74 ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
77 * Magic header for a swap area. The first part of the union is
78 * what the swap magic looks like for the old (limited to 128MB)
79 * swap area format, the second part of the union adds - in the
80 * old reserved area - some extra information. Note that the first
81 * kilobyte is reserved for boot loader or disk label stuff...
83 * Having the magic at the end of the PAGE_SIZE makes detecting swap
84 * areas somewhat tricky on machines that support multiple page sizes.
85 * For 2.5 we'll probably want to move the magic to just beyond the
90 char reserved[PAGE_SIZE - 10];
91 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
94 char bootbits[1024]; /* Space for disklabel etc. */
98 unsigned char sws_uuid[16];
99 unsigned char sws_volume[16];
106 * current->reclaim_state points to one of these when a task is running
109 struct reclaim_state {
110 unsigned long reclaimed_slab;
115 struct address_space;
117 struct writeback_control;
121 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
122 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
123 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
124 * from setup, they're handled identically.
126 * We always assume that blocks are of size PAGE_SIZE.
129 struct list_head list;
132 sector_t start_block;
136 * Max bad pages in the new format..
138 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
139 #define MAX_SWAP_BADPAGES \
140 ((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
143 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
144 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
145 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
146 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
147 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
148 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
149 SWP_BLKDEV = (1 << 6), /* its a block device */
150 SWP_FILE = (1 << 7), /* set after swap_activate success */
151 SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
152 SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
153 /* add others here before... */
154 SWP_SCANNING = (1 << 10), /* refcount in scan_swap_map */
157 #define SWAP_CLUSTER_MAX 32UL
158 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
160 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
161 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
162 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
163 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
164 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
165 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
168 * We use this to track usage of a cluster. A cluster is a block of swap disk
169 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
170 * free clusters are organized into a list. We fetch an entry from the list to
171 * get a free cluster.
173 * The data field stores next cluster if the cluster is free or cluster usage
174 * counter otherwise. The flags field determines if a cluster is free. This is
175 * protected by swap_info_struct.lock.
177 struct swap_cluster_info {
178 unsigned int data:24;
179 unsigned int flags:8;
181 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
182 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
185 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
186 * its own cluster and swapout sequentially. The purpose is to optimize swapout
189 struct percpu_cluster {
190 struct swap_cluster_info index; /* Current cluster index */
191 unsigned int next; /* Likely next allocation offset */
195 * The in-memory structure used to track swap areas.
197 struct swap_info_struct {
198 unsigned long flags; /* SWP_USED etc: see above */
199 signed short prio; /* swap priority of this type */
200 struct plist_node list; /* entry in swap_active_head */
201 struct plist_node avail_list; /* entry in swap_avail_head */
202 signed char type; /* strange name for an index */
203 unsigned int max; /* extent of the swap_map */
204 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
205 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
206 struct swap_cluster_info free_cluster_head; /* free cluster list head */
207 struct swap_cluster_info free_cluster_tail; /* free cluster list tail */
208 unsigned int lowest_bit; /* index of first free in swap_map */
209 unsigned int highest_bit; /* index of last free in swap_map */
210 unsigned int pages; /* total of usable pages of swap */
211 unsigned int inuse_pages; /* number of those currently in use */
212 unsigned int cluster_next; /* likely index for next allocation */
213 unsigned int cluster_nr; /* countdown to next cluster search */
214 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
215 struct swap_extent *curr_swap_extent;
216 struct swap_extent first_swap_extent;
217 struct block_device *bdev; /* swap device or bdev of swap file */
218 struct file *swap_file; /* seldom referenced */
219 unsigned int old_block_size; /* seldom referenced */
220 #ifdef CONFIG_FRONTSWAP
221 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
222 atomic_t frontswap_pages; /* frontswap pages in-use counter */
225 * protect map scan related fields like
226 * swap_map, lowest_bit, highest_bit,
227 * inuse_pages, cluster_next,
228 * cluster_nr, lowest_alloc,
229 * highest_alloc, free/discard cluster
230 * list. other fields are only changed
231 * at swapon/swapoff, so are protected
232 * by swap_lock. changing flags need
233 * hold this lock and swap_lock. If
234 * both locks need hold, hold swap_lock
237 struct work_struct discard_work; /* discard worker */
238 struct swap_cluster_info discard_cluster_head; /* list head of discard clusters */
239 struct swap_cluster_info discard_cluster_tail; /* list tail of discard clusters */
242 /* linux/mm/workingset.c */
243 void *workingset_eviction(struct address_space *mapping, struct page *page);
244 bool workingset_refault(void *shadow);
245 void workingset_activation(struct page *page);
246 extern struct list_lru workingset_shadow_nodes;
248 static inline unsigned int workingset_node_pages(struct radix_tree_node *node)
250 return node->count & RADIX_TREE_COUNT_MASK;
253 static inline void workingset_node_pages_inc(struct radix_tree_node *node)
258 static inline void workingset_node_pages_dec(struct radix_tree_node *node)
263 static inline unsigned int workingset_node_shadows(struct radix_tree_node *node)
265 return node->count >> RADIX_TREE_COUNT_SHIFT;
268 static inline void workingset_node_shadows_inc(struct radix_tree_node *node)
270 node->count += 1U << RADIX_TREE_COUNT_SHIFT;
273 static inline void workingset_node_shadows_dec(struct radix_tree_node *node)
275 node->count -= 1U << RADIX_TREE_COUNT_SHIFT;
278 /* linux/mm/page_alloc.c */
279 extern unsigned long totalram_pages;
280 extern unsigned long totalreserve_pages;
281 extern unsigned long nr_free_buffer_pages(void);
282 extern unsigned long nr_free_pagecache_pages(void);
284 /* Definition of global_page_state not available yet */
285 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
288 /* linux/mm/swap.c */
289 extern void lru_cache_add(struct page *);
290 extern void lru_cache_add_anon(struct page *page);
291 extern void lru_cache_add_file(struct page *page);
292 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
293 struct lruvec *lruvec, struct list_head *head);
294 extern void activate_page(struct page *);
295 extern void mark_page_accessed(struct page *);
296 extern void lru_add_drain(void);
297 extern void lru_add_drain_cpu(int cpu);
298 extern void lru_add_drain_all(void);
299 extern void rotate_reclaimable_page(struct page *page);
300 extern void deactivate_file_page(struct page *page);
301 extern void deactivate_page(struct page *page);
302 extern void swap_setup(void);
304 extern void add_page_to_unevictable_list(struct page *page);
306 extern void lru_cache_add_active_or_unevictable(struct page *page,
307 struct vm_area_struct *vma);
309 /* linux/mm/vmscan.c */
310 extern unsigned long zone_reclaimable_pages(struct zone *zone);
311 extern unsigned long pgdat_reclaimable_pages(struct pglist_data *pgdat);
312 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
313 gfp_t gfp_mask, nodemask_t *mask);
314 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
315 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
316 unsigned long nr_pages,
319 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
320 gfp_t gfp_mask, bool noswap,
322 unsigned long *nr_scanned);
323 extern unsigned long shrink_all_memory(unsigned long nr_pages);
324 extern int vm_swappiness;
325 extern int remove_mapping(struct address_space *mapping, struct page *page);
326 extern unsigned long vm_total_pages;
329 extern int zone_reclaim_mode;
330 extern int sysctl_min_unmapped_ratio;
331 extern int sysctl_min_slab_ratio;
332 extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
334 #define zone_reclaim_mode 0
335 static inline int zone_reclaim(struct zone *z, gfp_t mask, unsigned int order)
341 extern int page_evictable(struct page *page);
342 extern void check_move_unevictable_pages(struct page **, int nr_pages);
344 extern int kswapd_run(int nid);
345 extern void kswapd_stop(int nid);
348 /* linux/mm/page_io.c */
349 extern int swap_readpage(struct page *);
350 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
351 extern void end_swap_bio_write(struct bio *bio);
352 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
353 bio_end_io_t end_write_func);
354 extern int swap_set_page_dirty(struct page *page);
356 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
357 unsigned long nr_pages, sector_t start_block);
358 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
361 /* linux/mm/swap_state.c */
362 extern struct address_space swapper_spaces[];
363 #define swap_address_space(entry) (&swapper_spaces[swp_type(entry)])
364 extern unsigned long total_swapcache_pages(void);
365 extern void show_swap_cache_info(void);
366 extern int add_to_swap(struct page *, struct list_head *list);
367 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
368 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
369 extern void __delete_from_swap_cache(struct page *);
370 extern void delete_from_swap_cache(struct page *);
371 extern void free_page_and_swap_cache(struct page *);
372 extern void free_pages_and_swap_cache(struct page **, int);
373 extern struct page *lookup_swap_cache(swp_entry_t);
374 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
375 struct vm_area_struct *vma, unsigned long addr);
376 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
377 struct vm_area_struct *vma, unsigned long addr,
378 bool *new_page_allocated);
379 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
380 struct vm_area_struct *vma, unsigned long addr);
382 /* linux/mm/swapfile.c */
383 extern atomic_long_t nr_swap_pages;
384 extern long total_swap_pages;
386 /* Swap 50% full? Release swapcache more aggressively.. */
387 static inline bool vm_swap_full(void)
389 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
392 static inline long get_nr_swap_pages(void)
394 return atomic_long_read(&nr_swap_pages);
397 extern void si_swapinfo(struct sysinfo *);
398 extern swp_entry_t get_swap_page(void);
399 extern swp_entry_t get_swap_page_of_type(int);
400 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
401 extern void swap_shmem_alloc(swp_entry_t);
402 extern int swap_duplicate(swp_entry_t);
403 extern int swapcache_prepare(swp_entry_t);
404 extern void swap_free(swp_entry_t);
405 extern void swapcache_free(swp_entry_t);
406 extern int free_swap_and_cache(swp_entry_t);
407 extern int swap_type_of(dev_t, sector_t, struct block_device **);
408 extern unsigned int count_swap_pages(int, int);
409 extern sector_t map_swap_page(struct page *, struct block_device **);
410 extern sector_t swapdev_block(int, pgoff_t);
411 extern int page_swapcount(struct page *);
412 extern int swp_swapcount(swp_entry_t entry);
413 extern struct swap_info_struct *page_swap_info(struct page *);
414 extern bool reuse_swap_page(struct page *, int *);
415 extern int try_to_free_swap(struct page *);
416 struct backing_dev_info;
418 #else /* CONFIG_SWAP */
420 #define swap_address_space(entry) (NULL)
421 #define get_nr_swap_pages() 0L
422 #define total_swap_pages 0L
423 #define total_swapcache_pages() 0UL
424 #define vm_swap_full() 0
426 #define si_swapinfo(val) \
427 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
428 /* only sparc can not include linux/pagemap.h in this file
429 * so leave put_page and release_pages undeclared... */
430 #define free_page_and_swap_cache(page) \
432 #define free_pages_and_swap_cache(pages, nr) \
433 release_pages((pages), (nr), false);
435 static inline void show_swap_cache_info(void)
439 #define free_swap_and_cache(swp) is_migration_entry(swp)
440 #define swapcache_prepare(swp) is_migration_entry(swp)
442 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
447 static inline void swap_shmem_alloc(swp_entry_t swp)
451 static inline int swap_duplicate(swp_entry_t swp)
456 static inline void swap_free(swp_entry_t swp)
460 static inline void swapcache_free(swp_entry_t swp)
464 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
465 struct vm_area_struct *vma, unsigned long addr)
470 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
475 static inline struct page *lookup_swap_cache(swp_entry_t swp)
480 static inline int add_to_swap(struct page *page, struct list_head *list)
485 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
491 static inline void __delete_from_swap_cache(struct page *page)
495 static inline void delete_from_swap_cache(struct page *page)
499 static inline int page_swapcount(struct page *page)
504 static inline int swp_swapcount(swp_entry_t entry)
509 #define reuse_swap_page(page, total_mapcount) \
510 (page_trans_huge_mapcount(page, total_mapcount) == 1)
512 static inline int try_to_free_swap(struct page *page)
517 static inline swp_entry_t get_swap_page(void)
524 #endif /* CONFIG_SWAP */
527 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
529 /* Cgroup2 doesn't have per-cgroup swappiness */
530 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
531 return vm_swappiness;
534 if (mem_cgroup_disabled() || !memcg->css.parent)
535 return vm_swappiness;
537 return memcg->swappiness;
541 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
543 return vm_swappiness;
547 #ifdef CONFIG_MEMCG_SWAP
548 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
549 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
550 extern void mem_cgroup_uncharge_swap(swp_entry_t entry);
551 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
552 extern bool mem_cgroup_swap_full(struct page *page);
554 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
558 static inline int mem_cgroup_try_charge_swap(struct page *page,
564 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry)
568 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
570 return get_nr_swap_pages();
573 static inline bool mem_cgroup_swap_full(struct page *page)
575 return vm_swap_full();
579 #endif /* __KERNEL__*/
580 #endif /* _LINUX_SWAP_H */