swap.h 18 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559
  1. #ifndef _LINUX_SWAP_H
  2. #define _LINUX_SWAP_H
  3. #include <linux/spinlock.h>
  4. #include <linux/linkage.h>
  5. #include <linux/mmzone.h>
  6. #include <linux/list.h>
  7. #include <linux/memcontrol.h>
  8. #include <linux/sched.h>
  9. #include <linux/node.h>
  10. #include <linux/fs.h>
  11. #include <linux/atomic.h>
  12. #include <linux/page-flags.h>
  13. #include <asm/page.h>
  14. struct notifier_block;
  15. struct bio;
  16. #define SWAP_FLAG_PREFER 0x8000 /* set if swap priority specified */
  17. #define SWAP_FLAG_PRIO_MASK 0x7fff
  18. #define SWAP_FLAG_PRIO_SHIFT 0
  19. #define SWAP_FLAG_DISCARD 0x10000 /* enable discard for swap */
  20. #define SWAP_FLAG_DISCARD_ONCE 0x20000 /* discard swap area at swapon-time */
  21. #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */
  22. #define SWAP_FLAGS_VALID (SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \
  23. SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \
  24. SWAP_FLAG_DISCARD_PAGES)
  25. static inline int current_is_kswapd(void)
  26. {
  27. return current->flags & PF_KSWAPD;
  28. }
  29. /*
  30. * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
  31. * be swapped to. The swap type and the offset into that swap type are
  32. * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
  33. * for the type means that the maximum number of swapcache pages is 27 bits
  34. * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
  35. * the type/offset into the pte as 5/27 as well.
  36. */
  37. #define MAX_SWAPFILES_SHIFT 5
  38. /*
  39. * Use some of the swap files numbers for other purposes. This
  40. * is a convenient way to hook into the VM to trigger special
  41. * actions on faults.
  42. */
  43. /*
  44. * NUMA node memory migration support
  45. */
  46. #ifdef CONFIG_MIGRATION
  47. #define SWP_MIGRATION_NUM 2
  48. #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
  49. #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
  50. #else
  51. #define SWP_MIGRATION_NUM 0
  52. #endif
  53. /*
  54. * Handling of hardware poisoned pages with memory corruption.
  55. */
  56. #ifdef CONFIG_MEMORY_FAILURE
  57. #define SWP_HWPOISON_NUM 1
  58. #define SWP_HWPOISON MAX_SWAPFILES
  59. #else
  60. #define SWP_HWPOISON_NUM 0
  61. #endif
  62. #define MAX_SWAPFILES \
  63. ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
  64. /*
  65. * Magic header for a swap area. The first part of the union is
  66. * what the swap magic looks like for the old (limited to 128MB)
  67. * swap area format, the second part of the union adds - in the
  68. * old reserved area - some extra information. Note that the first
  69. * kilobyte is reserved for boot loader or disk label stuff...
  70. *
  71. * Having the magic at the end of the PAGE_SIZE makes detecting swap
  72. * areas somewhat tricky on machines that support multiple page sizes.
  73. * For 2.5 we'll probably want to move the magic to just beyond the
  74. * bootbits...
  75. */
  76. union swap_header {
  77. struct {
  78. char reserved[PAGE_SIZE - 10];
  79. char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
  80. } magic;
  81. struct {
  82. char bootbits[1024]; /* Space for disklabel etc. */
  83. __u32 version;
  84. __u32 last_page;
  85. __u32 nr_badpages;
  86. unsigned char sws_uuid[16];
  87. unsigned char sws_volume[16];
  88. __u32 padding[117];
  89. __u32 badpages[1];
  90. } info;
  91. };
  92. /*
  93. * current->reclaim_state points to one of these when a task is running
  94. * memory reclaim
  95. */
  96. struct reclaim_state {
  97. unsigned long reclaimed_slab;
  98. };
  99. #ifdef __KERNEL__
  100. struct address_space;
  101. struct sysinfo;
  102. struct writeback_control;
  103. struct zone;
  104. /*
  105. * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
  106. * disk blocks. A list of swap extents maps the entire swapfile. (Where the
  107. * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
  108. * from setup, they're handled identically.
  109. *
  110. * We always assume that blocks are of size PAGE_SIZE.
  111. */
  112. struct swap_extent {
  113. struct list_head list;
  114. pgoff_t start_page;
  115. pgoff_t nr_pages;
  116. sector_t start_block;
  117. };
  118. /*
  119. * Max bad pages in the new format..
  120. */
  121. #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
  122. #define MAX_SWAP_BADPAGES \
  123. ((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
  124. enum {
  125. SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
  126. SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
  127. SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
  128. SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
  129. SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
  130. SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
  131. SWP_BLKDEV = (1 << 6), /* its a block device */
  132. SWP_FILE = (1 << 7), /* set after swap_activate success */
  133. SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
  134. SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
  135. /* add others here before... */
  136. SWP_SCANNING = (1 << 10), /* refcount in scan_swap_map */
  137. };
  138. #define SWAP_CLUSTER_MAX 32UL
  139. #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
  140. /*
  141. * Ratio between zone->managed_pages and the "gap" that above the per-zone
  142. * "high_wmark". While balancing nodes, We allow kswapd to shrink zones that
  143. * do not meet the (high_wmark + gap) watermark, even which already met the
  144. * high_wmark, in order to provide better per-zone lru behavior. We are ok to
  145. * spend not more than 1% of the memory for this zone balancing "gap".
  146. */
  147. #define KSWAPD_ZONE_BALANCE_GAP_RATIO 100
  148. #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
  149. #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
  150. #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
  151. #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
  152. #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
  153. #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
  154. /*
  155. * We use this to track usage of a cluster. A cluster is a block of swap disk
  156. * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
  157. * free clusters are organized into a list. We fetch an entry from the list to
  158. * get a free cluster.
  159. *
  160. * The data field stores next cluster if the cluster is free or cluster usage
  161. * counter otherwise. The flags field determines if a cluster is free. This is
  162. * protected by swap_info_struct.lock.
  163. */
  164. struct swap_cluster_info {
  165. unsigned int data:24;
  166. unsigned int flags:8;
  167. };
  168. #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
  169. #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
  170. /*
  171. * We assign a cluster to each CPU, so each CPU can allocate swap entry from
  172. * its own cluster and swapout sequentially. The purpose is to optimize swapout
  173. * throughput.
  174. */
  175. struct percpu_cluster {
  176. struct swap_cluster_info index; /* Current cluster index */
  177. unsigned int next; /* Likely next allocation offset */
  178. };
  179. /*
  180. * The in-memory structure used to track swap areas.
  181. */
  182. struct swap_info_struct {
  183. unsigned long flags; /* SWP_USED etc: see above */
  184. signed short prio; /* swap priority of this type */
  185. struct plist_node list; /* entry in swap_active_head */
  186. struct plist_node avail_list; /* entry in swap_avail_head */
  187. signed char type; /* strange name for an index */
  188. unsigned int max; /* extent of the swap_map */
  189. unsigned char *swap_map; /* vmalloc'ed array of usage counts */
  190. struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
  191. struct swap_cluster_info free_cluster_head; /* free cluster list head */
  192. struct swap_cluster_info free_cluster_tail; /* free cluster list tail */
  193. unsigned int lowest_bit; /* index of first free in swap_map */
  194. unsigned int highest_bit; /* index of last free in swap_map */
  195. unsigned int pages; /* total of usable pages of swap */
  196. unsigned int inuse_pages; /* number of those currently in use */
  197. unsigned int cluster_next; /* likely index for next allocation */
  198. unsigned int cluster_nr; /* countdown to next cluster search */
  199. struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
  200. struct swap_extent *curr_swap_extent;
  201. struct swap_extent first_swap_extent;
  202. struct block_device *bdev; /* swap device or bdev of swap file */
  203. struct file *swap_file; /* seldom referenced */
  204. unsigned int old_block_size; /* seldom referenced */
  205. #ifdef CONFIG_FRONTSWAP
  206. unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
  207. atomic_t frontswap_pages; /* frontswap pages in-use counter */
  208. #endif
  209. spinlock_t lock; /*
  210. * protect map scan related fields like
  211. * swap_map, lowest_bit, highest_bit,
  212. * inuse_pages, cluster_next,
  213. * cluster_nr, lowest_alloc,
  214. * highest_alloc, free/discard cluster
  215. * list. other fields are only changed
  216. * at swapon/swapoff, so are protected
  217. * by swap_lock. changing flags need
  218. * hold this lock and swap_lock. If
  219. * both locks need hold, hold swap_lock
  220. * first.
  221. */
  222. struct work_struct discard_work; /* discard worker */
  223. struct swap_cluster_info discard_cluster_head; /* list head of discard clusters */
  224. struct swap_cluster_info discard_cluster_tail; /* list tail of discard clusters */
  225. };
  226. /* linux/mm/workingset.c */
  227. void *workingset_eviction(struct address_space *mapping, struct page *page);
  228. bool workingset_refault(void *shadow);
  229. void workingset_activation(struct page *page);
  230. extern struct list_lru workingset_shadow_nodes;
  231. static inline unsigned int workingset_node_pages(struct radix_tree_node *node)
  232. {
  233. return node->count & RADIX_TREE_COUNT_MASK;
  234. }
  235. static inline void workingset_node_pages_inc(struct radix_tree_node *node)
  236. {
  237. node->count++;
  238. }
  239. static inline void workingset_node_pages_dec(struct radix_tree_node *node)
  240. {
  241. node->count--;
  242. }
  243. static inline unsigned int workingset_node_shadows(struct radix_tree_node *node)
  244. {
  245. return node->count >> RADIX_TREE_COUNT_SHIFT;
  246. }
  247. static inline void workingset_node_shadows_inc(struct radix_tree_node *node)
  248. {
  249. node->count += 1U << RADIX_TREE_COUNT_SHIFT;
  250. }
  251. static inline void workingset_node_shadows_dec(struct radix_tree_node *node)
  252. {
  253. node->count -= 1U << RADIX_TREE_COUNT_SHIFT;
  254. }
  255. /* linux/mm/page_alloc.c */
  256. extern unsigned long totalram_pages;
  257. extern unsigned long totalreserve_pages;
  258. extern unsigned long nr_free_buffer_pages(void);
  259. extern unsigned long nr_free_pagecache_pages(void);
  260. /* Definition of global_page_state not available yet */
  261. #define nr_free_pages() global_page_state(NR_FREE_PAGES)
  262. /* linux/mm/swap.c */
  263. extern void lru_cache_add(struct page *);
  264. extern void lru_cache_add_anon(struct page *page);
  265. extern void lru_cache_add_file(struct page *page);
  266. extern void lru_add_page_tail(struct page *page, struct page *page_tail,
  267. struct lruvec *lruvec, struct list_head *head);
  268. extern void activate_page(struct page *);
  269. extern void mark_page_accessed(struct page *);
  270. extern void lru_add_drain(void);
  271. extern void lru_add_drain_cpu(int cpu);
  272. extern void lru_add_drain_all(void);
  273. extern void rotate_reclaimable_page(struct page *page);
  274. extern void deactivate_file_page(struct page *page);
  275. extern void deactivate_page(struct page *page);
  276. extern void swap_setup(void);
  277. extern void add_page_to_unevictable_list(struct page *page);
  278. extern void lru_cache_add_active_or_unevictable(struct page *page,
  279. struct vm_area_struct *vma);
  280. /* linux/mm/vmscan.c */
  281. extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
  282. gfp_t gfp_mask, nodemask_t *mask);
  283. extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
  284. extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
  285. unsigned long nr_pages,
  286. gfp_t gfp_mask,
  287. bool may_swap);
  288. extern unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
  289. gfp_t gfp_mask, bool noswap,
  290. struct zone *zone,
  291. unsigned long *nr_scanned);
  292. extern unsigned long shrink_all_memory(unsigned long nr_pages);
  293. extern int vm_swappiness;
  294. extern int remove_mapping(struct address_space *mapping, struct page *page);
  295. extern unsigned long vm_total_pages;
  296. #ifdef CONFIG_NUMA
  297. extern int zone_reclaim_mode;
  298. extern int sysctl_min_unmapped_ratio;
  299. extern int sysctl_min_slab_ratio;
  300. extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
  301. #else
  302. #define zone_reclaim_mode 0
  303. static inline int zone_reclaim(struct zone *z, gfp_t mask, unsigned int order)
  304. {
  305. return 0;
  306. }
  307. #endif
  308. extern int page_evictable(struct page *page);
  309. extern void check_move_unevictable_pages(struct page **, int nr_pages);
  310. extern int kswapd_run(int nid);
  311. extern void kswapd_stop(int nid);
  312. #ifdef CONFIG_MEMCG
  313. static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
  314. {
  315. /* root ? */
  316. if (mem_cgroup_disabled() || !memcg->css.parent)
  317. return vm_swappiness;
  318. return memcg->swappiness;
  319. }
  320. #else
  321. static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
  322. {
  323. return vm_swappiness;
  324. }
  325. #endif
  326. #ifdef CONFIG_MEMCG_SWAP
  327. extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
  328. extern void mem_cgroup_uncharge_swap(swp_entry_t entry);
  329. #else
  330. static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
  331. {
  332. }
  333. static inline void mem_cgroup_uncharge_swap(swp_entry_t entry)
  334. {
  335. }
  336. #endif
  337. #ifdef CONFIG_SWAP
  338. /* linux/mm/page_io.c */
  339. extern int swap_readpage(struct page *);
  340. extern int swap_writepage(struct page *page, struct writeback_control *wbc);
  341. extern void end_swap_bio_write(struct bio *bio);
  342. extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
  343. bio_end_io_t end_write_func);
  344. extern int swap_set_page_dirty(struct page *page);
  345. int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
  346. unsigned long nr_pages, sector_t start_block);
  347. int generic_swapfile_activate(struct swap_info_struct *, struct file *,
  348. sector_t *);
  349. /* linux/mm/swap_state.c */
  350. extern struct address_space swapper_spaces[];
  351. #define swap_address_space(entry) (&swapper_spaces[swp_type(entry)])
  352. extern unsigned long total_swapcache_pages(void);
  353. extern void show_swap_cache_info(void);
  354. extern int add_to_swap(struct page *, struct list_head *list);
  355. extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
  356. extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
  357. extern void __delete_from_swap_cache(struct page *);
  358. extern void delete_from_swap_cache(struct page *);
  359. extern void free_page_and_swap_cache(struct page *);
  360. extern void free_pages_and_swap_cache(struct page **, int);
  361. extern struct page *lookup_swap_cache(swp_entry_t);
  362. extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
  363. struct vm_area_struct *vma, unsigned long addr);
  364. extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
  365. struct vm_area_struct *vma, unsigned long addr,
  366. bool *new_page_allocated);
  367. extern struct page *swapin_readahead(swp_entry_t, gfp_t,
  368. struct vm_area_struct *vma, unsigned long addr);
  369. /* linux/mm/swapfile.c */
  370. extern atomic_long_t nr_swap_pages;
  371. extern long total_swap_pages;
  372. /* Swap 50% full? Release swapcache more aggressively.. */
  373. static inline bool vm_swap_full(void)
  374. {
  375. return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
  376. }
  377. static inline long get_nr_swap_pages(void)
  378. {
  379. return atomic_long_read(&nr_swap_pages);
  380. }
  381. extern void si_swapinfo(struct sysinfo *);
  382. extern swp_entry_t get_swap_page(void);
  383. extern swp_entry_t get_swap_page_of_type(int);
  384. extern int add_swap_count_continuation(swp_entry_t, gfp_t);
  385. extern void swap_shmem_alloc(swp_entry_t);
  386. extern int swap_duplicate(swp_entry_t);
  387. extern int swapcache_prepare(swp_entry_t);
  388. extern void swap_free(swp_entry_t);
  389. extern void swapcache_free(swp_entry_t);
  390. extern int free_swap_and_cache(swp_entry_t);
  391. extern int swap_type_of(dev_t, sector_t, struct block_device **);
  392. extern unsigned int count_swap_pages(int, int);
  393. extern sector_t map_swap_page(struct page *, struct block_device **);
  394. extern sector_t swapdev_block(int, pgoff_t);
  395. extern int page_swapcount(struct page *);
  396. extern int swp_swapcount(swp_entry_t entry);
  397. extern struct swap_info_struct *page_swap_info(struct page *);
  398. extern int reuse_swap_page(struct page *);
  399. extern int try_to_free_swap(struct page *);
  400. struct backing_dev_info;
  401. #else /* CONFIG_SWAP */
  402. #define swap_address_space(entry) (NULL)
  403. #define get_nr_swap_pages() 0L
  404. #define total_swap_pages 0L
  405. #define total_swapcache_pages() 0UL
  406. #define vm_swap_full() 0
  407. #define si_swapinfo(val) \
  408. do { (val)->freeswap = (val)->totalswap = 0; } while (0)
  409. /* only sparc can not include linux/pagemap.h in this file
  410. * so leave page_cache_release and release_pages undeclared... */
  411. #define free_page_and_swap_cache(page) \
  412. page_cache_release(page)
  413. #define free_pages_and_swap_cache(pages, nr) \
  414. release_pages((pages), (nr), false);
  415. static inline void show_swap_cache_info(void)
  416. {
  417. }
  418. #define free_swap_and_cache(swp) is_migration_entry(swp)
  419. #define swapcache_prepare(swp) is_migration_entry(swp)
  420. static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
  421. {
  422. return 0;
  423. }
  424. static inline void swap_shmem_alloc(swp_entry_t swp)
  425. {
  426. }
  427. static inline int swap_duplicate(swp_entry_t swp)
  428. {
  429. return 0;
  430. }
  431. static inline void swap_free(swp_entry_t swp)
  432. {
  433. }
  434. static inline void swapcache_free(swp_entry_t swp)
  435. {
  436. }
  437. static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
  438. struct vm_area_struct *vma, unsigned long addr)
  439. {
  440. return NULL;
  441. }
  442. static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
  443. {
  444. return 0;
  445. }
  446. static inline struct page *lookup_swap_cache(swp_entry_t swp)
  447. {
  448. return NULL;
  449. }
  450. static inline int add_to_swap(struct page *page, struct list_head *list)
  451. {
  452. return 0;
  453. }
  454. static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
  455. gfp_t gfp_mask)
  456. {
  457. return -1;
  458. }
  459. static inline void __delete_from_swap_cache(struct page *page)
  460. {
  461. }
  462. static inline void delete_from_swap_cache(struct page *page)
  463. {
  464. }
  465. static inline int page_swapcount(struct page *page)
  466. {
  467. return 0;
  468. }
  469. static inline int swp_swapcount(swp_entry_t entry)
  470. {
  471. return 0;
  472. }
  473. #define reuse_swap_page(page) \
  474. (!PageTransCompound(page) && page_mapcount(page) == 1)
  475. static inline int try_to_free_swap(struct page *page)
  476. {
  477. return 0;
  478. }
  479. static inline swp_entry_t get_swap_page(void)
  480. {
  481. swp_entry_t entry;
  482. entry.val = 0;
  483. return entry;
  484. }
  485. #endif /* CONFIG_SWAP */
  486. #endif /* __KERNEL__*/
  487. #endif /* _LINUX_SWAP_H */