include/linux/swap.h

   1 #ifndef _LINUX_SWAP_H
   2 #define _LINUX_SWAP_H
   3
   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>
  11 #include <linux/fs.h>
  12 #include <linux/atomic.h>
  13 #include <linux/page-flags.h>
  14 #include <asm/page.h>
  15
  16 struct notifier_block;
  17
  18 struct bio;
  19
  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 */
  26
  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)
  30
  31 static inline int current_is_kswapd(void)
  32 {
  33         return current->flags & PF_KSWAPD;
  34 }
  35
  36 /*
  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.
  43  */
  44 #define MAX_SWAPFILES_SHIFT     5
  45
  46 /*
  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
  49  * actions on faults.
  50  */
  51
  52 /*
  53  * NUMA node memory migration support
  54  */
  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)
  59 #else
  60 #define SWP_MIGRATION_NUM 0
  61 #endif
  62
  63 /*
  64  * Handling of hardware poisoned pages with memory corruption.
  65  */
  66 #ifdef CONFIG_MEMORY_FAILURE
  67 #define SWP_HWPOISON_NUM 1
  68 #define SWP_HWPOISON            MAX_SWAPFILES
  69 #else
  70 #define SWP_HWPOISON_NUM 0
  71 #endif
  72
  73 #define MAX_SWAPFILES \
  74         ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
  75
  76 /*
  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...
  82  *
  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
  86  * bootbits...
  87  */
  88 union swap_header {
  89         struct {
  90                 char reserved[PAGE_SIZE - 10];
  91                 char magic[10];                 /* SWAP-SPACE or SWAPSPACE2 */
  92         } magic;
  93         struct {
  94                 char            bootbits[1024]; /* Space for disklabel etc. */
  95                 __u32           version;
  96                 __u32           last_page;
  97                 __u32           nr_badpages;
  98                 unsigned char   sws_uuid[16];
  99                 unsigned char   sws_volume[16];
 100                 __u32           padding[117];
 101                 __u32           badpages[1];
 102         } info;
 103 };
 104
 105 /*
 106  * current->reclaim_state points to one of these when a task is running
 107  * memory reclaim
 108  */
 109 struct reclaim_state {
 110         unsigned long reclaimed_slab;
 111 };
 112
 113 #ifdef __KERNEL__
 114
 115 struct address_space;
 116 struct sysinfo;
 117 struct writeback_control;
 118 struct zone;
 119
 120 /*
 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.
 125  *
 126  * We always assume that blocks are of size PAGE_SIZE.
 127  */
 128 struct swap_extent {
 129         struct list_head list;
 130         pgoff_t start_page;
 131         pgoff_t nr_pages;
 132         sector_t start_block;
 133 };
 134
 135 /*
 136  * Max bad pages in the new format..
 137  */
 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))
 141
 142 enum {
 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         SWP_STABLE_WRITES = (1 << 10),  /* no overwrite PG_writeback pages */
 154                                         /* add others here before... */
 155         SWP_SCANNING    = (1 << 11),    /* refcount in scan_swap_map */
 156 };
 157
 158 #define SWAP_CLUSTER_MAX 32UL
 159 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
 160
 161 #define SWAP_MAP_MAX    0x3e    /* Max duplication count, in first swap_map */
 162 #define SWAP_MAP_BAD    0x3f    /* Note pageblock is bad, in first swap_map */
 163 #define SWAP_HAS_CACHE  0x40    /* Flag page is cached, in first swap_map */
 164 #define SWAP_CONT_MAX   0x7f    /* Max count, in each swap_map continuation */
 165 #define COUNT_CONTINUED 0x80    /* See swap_map continuation for full count */
 166 #define SWAP_MAP_SHMEM  0xbf    /* Owned by shmem/tmpfs, in first swap_map */
 167
 168 /*
 169  * We use this to track usage of a cluster. A cluster is a block of swap disk
 170  * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
 171  * free clusters are organized into a list. We fetch an entry from the list to
 172  * get a free cluster.
 173  *
 174  * The data field stores next cluster if the cluster is free or cluster usage
 175  * counter otherwise. The flags field determines if a cluster is free. This is
 176  * protected by swap_info_struct.lock.
 177  */
 178 struct swap_cluster_info {
 179         unsigned int data:24;
 180         unsigned int flags:8;
 181 };
 182 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
 183 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
 184
 185 /*
 186  * We assign a cluster to each CPU, so each CPU can allocate swap entry from
 187  * its own cluster and swapout sequentially. The purpose is to optimize swapout
 188  * throughput.
 189  */
 190 struct percpu_cluster {
 191         struct swap_cluster_info index; /* Current cluster index */
 192         unsigned int next; /* Likely next allocation offset */
 193 };
 194
 195 struct swap_cluster_list {
 196         struct swap_cluster_info head;
 197         struct swap_cluster_info tail;
 198 };
 199
 200 /*
 201  * The in-memory structure used to track swap areas.
 202  */
 203 struct swap_info_struct {
 204         unsigned long   flags;          /* SWP_USED etc: see above */
 205         signed short    prio;           /* swap priority of this type */
 206         struct plist_node list;         /* entry in swap_active_head */
 207         struct plist_node avail_list;   /* entry in swap_avail_head */
 208         signed char     type;           /* strange name for an index */
 209         unsigned int    max;            /* extent of the swap_map */
 210         unsigned char *swap_map;        /* vmalloc'ed array of usage counts */
 211         struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
 212         struct swap_cluster_list free_clusters; /* free clusters list */
 213         unsigned int lowest_bit;        /* index of first free in swap_map */
 214         unsigned int highest_bit;       /* index of last free in swap_map */
 215         unsigned int pages;             /* total of usable pages of swap */
 216         unsigned int inuse_pages;       /* number of those currently in use */
 217         unsigned int cluster_next;      /* likely index for next allocation */
 218         unsigned int cluster_nr;        /* countdown to next cluster search */
 219         struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
 220         struct swap_extent *curr_swap_extent;
 221         struct swap_extent first_swap_extent;
 222         struct block_device *bdev;      /* swap device or bdev of swap file */
 223         struct file *swap_file;         /* seldom referenced */
 224         unsigned int old_block_size;    /* seldom referenced */
 225 #ifdef CONFIG_FRONTSWAP
 226         unsigned long *frontswap_map;   /* frontswap in-use, one bit per page */
 227         atomic_t frontswap_pages;       /* frontswap pages in-use counter */
 228 #endif
 229         spinlock_t lock;                /*
 230                                          * protect map scan related fields like
 231                                          * swap_map, lowest_bit, highest_bit,
 232                                          * inuse_pages, cluster_next,
 233                                          * cluster_nr, lowest_alloc,
 234                                          * highest_alloc, free/discard cluster
 235                                          * list. other fields are only changed
 236                                          * at swapon/swapoff, so are protected
 237                                          * by swap_lock. changing flags need
 238                                          * hold this lock and swap_lock. If
 239                                          * both locks need hold, hold swap_lock
 240                                          * first.
 241                                          */
 242         struct work_struct discard_work; /* discard worker */
 243         struct swap_cluster_list discard_clusters; /* discard clusters list */
 244 };
 245
 246 /* linux/mm/workingset.c */
 247 void *workingset_eviction(struct address_space *mapping, struct page *page);
 248 bool workingset_refault(void *shadow);
 249 void workingset_activation(struct page *page);
 250 void workingset_update_node(struct radix_tree_node *node, void *private);
 251
 252 /* linux/mm/page_alloc.c */
 253 extern unsigned long totalram_pages;
 254 extern unsigned long totalreserve_pages;
 255 extern unsigned long nr_free_buffer_pages(void);
 256 extern unsigned long nr_free_pagecache_pages(void);
 257
 258 /* Definition of global_page_state not available yet */
 259 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
 260
 261
 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
 278 extern void add_page_to_unevictable_list(struct page *page);
 279
 280 extern void lru_cache_add_active_or_unevictable(struct page *page,
 281                                                 struct vm_area_struct *vma);
 282
 283 /* linux/mm/vmscan.c */
 284 extern unsigned long zone_reclaimable_pages(struct zone *zone);
 285 extern unsigned long pgdat_reclaimable_pages(struct pglist_data *pgdat);
 286 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
 287                                         gfp_t gfp_mask, nodemask_t *mask);
 288 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
 289 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
 290                                                   unsigned long nr_pages,
 291                                                   gfp_t gfp_mask,
 292                                                   bool may_swap);
 293 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
 294                                                 gfp_t gfp_mask, bool noswap,
 295                                                 pg_data_t *pgdat,
 296                                                 unsigned long *nr_scanned);
 297 extern unsigned long shrink_all_memory(unsigned long nr_pages);
 298 extern int vm_swappiness;
 299 extern int remove_mapping(struct address_space *mapping, struct page *page);
 300 extern unsigned long vm_total_pages;
 301
 302 #ifdef CONFIG_NUMA
 303 extern int node_reclaim_mode;
 304 extern int sysctl_min_unmapped_ratio;
 305 extern int sysctl_min_slab_ratio;
 306 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
 307 #else
 308 #define node_reclaim_mode 0
 309 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
 310                                 unsigned int order)
 311 {
 312         return 0;
 313 }
 314 #endif
 315
 316 extern int page_evictable(struct page *page);
 317 extern void check_move_unevictable_pages(struct page **, int nr_pages);
 318
 319 extern int kswapd_run(int nid);
 320 extern void kswapd_stop(int nid);
 321
 322 #ifdef CONFIG_SWAP
 323
 324 #include <linux/blk_types.h> /* for bio_end_io_t */
 325
 326 /* linux/mm/page_io.c */
 327 extern int swap_readpage(struct page *);
 328 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
 329 extern void end_swap_bio_write(struct bio *bio);
 330 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
 331         bio_end_io_t end_write_func);
 332 extern int swap_set_page_dirty(struct page *page);
 333
 334 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
 335                 unsigned long nr_pages, sector_t start_block);
 336 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
 337                 sector_t *);
 338
 339 /* linux/mm/swap_state.c */
 340 extern struct address_space swapper_spaces[];
 341 #define swap_address_space(entry) (&swapper_spaces[swp_type(entry)])
 342 extern unsigned long total_swapcache_pages(void);
 343 extern void show_swap_cache_info(void);
 344 extern int add_to_swap(struct page *, struct list_head *list);
 345 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
 346 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
 347 extern void __delete_from_swap_cache(struct page *);
 348 extern void delete_from_swap_cache(struct page *);
 349 extern void free_page_and_swap_cache(struct page *);
 350 extern void free_pages_and_swap_cache(struct page **, int);
 351 extern struct page *lookup_swap_cache(swp_entry_t);
 352 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
 353                         struct vm_area_struct *vma, unsigned long addr);
 354 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
 355                         struct vm_area_struct *vma, unsigned long addr,
 356                         bool *new_page_allocated);
 357 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
 358                         struct vm_area_struct *vma, unsigned long addr);
 359
 360 /* linux/mm/swapfile.c */
 361 extern atomic_long_t nr_swap_pages;
 362 extern long total_swap_pages;
 363
 364 /* Swap 50% full? Release swapcache more aggressively.. */
 365 static inline bool vm_swap_full(void)
 366 {
 367         return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
 368 }
 369
 370 static inline long get_nr_swap_pages(void)
 371 {
 372         return atomic_long_read(&nr_swap_pages);
 373 }
 374
 375 extern void si_swapinfo(struct sysinfo *);
 376 extern swp_entry_t get_swap_page(void);
 377 extern swp_entry_t get_swap_page_of_type(int);
 378 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
 379 extern void swap_shmem_alloc(swp_entry_t);
 380 extern int swap_duplicate(swp_entry_t);
 381 extern int swapcache_prepare(swp_entry_t);
 382 extern void swap_free(swp_entry_t);
 383 extern void swapcache_free(swp_entry_t);
 384 extern int free_swap_and_cache(swp_entry_t);
 385 extern int swap_type_of(dev_t, sector_t, struct block_device **);
 386 extern unsigned int count_swap_pages(int, int);
 387 extern sector_t map_swap_page(struct page *, struct block_device **);
 388 extern sector_t swapdev_block(int, pgoff_t);
 389 extern int page_swapcount(struct page *);
 390 extern int swp_swapcount(swp_entry_t entry);
 391 extern struct swap_info_struct *page_swap_info(struct page *);
 392 extern bool reuse_swap_page(struct page *, int *);
 393 extern int try_to_free_swap(struct page *);
 394 struct backing_dev_info;
 395
 396 #else /* CONFIG_SWAP */
 397
 398 #define swap_address_space(entry)               (NULL)
 399 #define get_nr_swap_pages()                     0L
 400 #define total_swap_pages                        0L
 401 #define total_swapcache_pages()                 0UL
 402 #define vm_swap_full()                          0
 403
 404 #define si_swapinfo(val) \
 405         do { (val)->freeswap = (val)->totalswap = 0; } while (0)
 406 /* only sparc can not include linux/pagemap.h in this file
 407  * so leave put_page and release_pages undeclared... */
 408 #define free_page_and_swap_cache(page) \
 409         put_page(page)
 410 #define free_pages_and_swap_cache(pages, nr) \
 411         release_pages((pages), (nr), false);
 412
 413 static inline void show_swap_cache_info(void)
 414 {
 415 }
 416
 417 #define free_swap_and_cache(swp)        is_migration_entry(swp)
 418 #define swapcache_prepare(swp)          is_migration_entry(swp)
 419
 420 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
 421 {
 422         return 0;
 423 }
 424
 425 static inline void swap_shmem_alloc(swp_entry_t swp)
 426 {
 427 }
 428
 429 static inline int swap_duplicate(swp_entry_t swp)
 430 {
 431         return 0;
 432 }
 433
 434 static inline void swap_free(swp_entry_t swp)
 435 {
 436 }
 437
 438 static inline void swapcache_free(swp_entry_t swp)
 439 {
 440 }
 441
 442 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
 443                         struct vm_area_struct *vma, unsigned long addr)
 444 {
 445         return NULL;
 446 }
 447
 448 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
 449 {
 450         return 0;
 451 }
 452
 453 static inline struct page *lookup_swap_cache(swp_entry_t swp)
 454 {
 455         return NULL;
 456 }
 457
 458 static inline int add_to_swap(struct page *page, struct list_head *list)
 459 {
 460         return 0;
 461 }
 462
 463 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
 464                                                         gfp_t gfp_mask)
 465 {
 466         return -1;
 467 }
 468
 469 static inline void __delete_from_swap_cache(struct page *page)
 470 {
 471 }
 472
 473 static inline void delete_from_swap_cache(struct page *page)
 474 {
 475 }
 476
 477 static inline int page_swapcount(struct page *page)
 478 {
 479         return 0;
 480 }
 481
 482 static inline int swp_swapcount(swp_entry_t entry)
 483 {
 484         return 0;
 485 }
 486
 487 #define reuse_swap_page(page, total_mapcount) \
 488         (page_trans_huge_mapcount(page, total_mapcount) == 1)
 489
 490 static inline int try_to_free_swap(struct page *page)
 491 {
 492         return 0;
 493 }
 494
 495 static inline swp_entry_t get_swap_page(void)
 496 {
 497         swp_entry_t entry;
 498         entry.val = 0;
 499         return entry;
 500 }
 501
 502 #endif /* CONFIG_SWAP */
 503
 504 #ifdef CONFIG_MEMCG
 505 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
 506 {
 507         /* Cgroup2 doesn't have per-cgroup swappiness */
 508         if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
 509                 return vm_swappiness;
 510
 511         /* root ? */
 512         if (mem_cgroup_disabled() || !memcg->css.parent)
 513                 return vm_swappiness;
 514
 515         return memcg->swappiness;
 516 }
 517
 518 #else
 519 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
 520 {
 521         return vm_swappiness;
 522 }
 523 #endif
 524
 525 #ifdef CONFIG_MEMCG_SWAP
 526 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
 527 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
 528 extern void mem_cgroup_uncharge_swap(swp_entry_t entry);
 529 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
 530 extern bool mem_cgroup_swap_full(struct page *page);
 531 #else
 532 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
 533 {
 534 }
 535
 536 static inline int mem_cgroup_try_charge_swap(struct page *page,
 537                                              swp_entry_t entry)
 538 {
 539         return 0;
 540 }
 541
 542 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry)
 543 {
 544 }
 545
 546 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
 547 {
 548         return get_nr_swap_pages();
 549 }
 550
 551 static inline bool mem_cgroup_swap_full(struct page *page)
 552 {
 553         return vm_swap_full();
 554 }
 555 #endif
 556
 557 #endif /* __KERNEL__*/
 558 #endif /* _LINUX_SWAP_H */