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         spinlock_t lock;        /*
 180                                  * Protect swap_cluster_info fields
 181                                  * and swap_info_struct->swap_map
 182                                  * elements correspond to the swap
 183                                  * cluster
 184                                  */
 185         unsigned int data:24;
 186         unsigned int flags:8;
 187 };
 188 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
 189 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
 190
 191 /*
 192  * We assign a cluster to each CPU, so each CPU can allocate swap entry from
 193  * its own cluster and swapout sequentially. The purpose is to optimize swapout
 194  * throughput.
 195  */
 196 struct percpu_cluster {
 197         struct swap_cluster_info index; /* Current cluster index */
 198         unsigned int next; /* Likely next allocation offset */
 199 };
 200
 201 struct swap_cluster_list {
 202         struct swap_cluster_info head;
 203         struct swap_cluster_info tail;
 204 };
 205
 206 /*
 207  * The in-memory structure used to track swap areas.
 208  */
 209 struct swap_info_struct {
 210         unsigned long   flags;          /* SWP_USED etc: see above */
 211         signed short    prio;           /* swap priority of this type */
 212         struct plist_node list;         /* entry in swap_active_head */
 213         struct plist_node avail_list;   /* entry in swap_avail_head */
 214         signed char     type;           /* strange name for an index */
 215         unsigned int    max;            /* extent of the swap_map */
 216         unsigned char *swap_map;        /* vmalloc'ed array of usage counts */
 217         struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
 218         struct swap_cluster_list free_clusters; /* free clusters list */
 219         unsigned int lowest_bit;        /* index of first free in swap_map */
 220         unsigned int highest_bit;       /* index of last free in swap_map */
 221         unsigned int pages;             /* total of usable pages of swap */
 222         unsigned int inuse_pages;       /* number of those currently in use */
 223         unsigned int cluster_next;      /* likely index for next allocation */
 224         unsigned int cluster_nr;        /* countdown to next cluster search */
 225         struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
 226         struct swap_extent *curr_swap_extent;
 227         struct swap_extent first_swap_extent;
 228         struct block_device *bdev;      /* swap device or bdev of swap file */
 229         struct file *swap_file;         /* seldom referenced */
 230         unsigned int old_block_size;    /* seldom referenced */
 231 #ifdef CONFIG_FRONTSWAP
 232         unsigned long *frontswap_map;   /* frontswap in-use, one bit per page */
 233         atomic_t frontswap_pages;       /* frontswap pages in-use counter */
 234 #endif
 235         spinlock_t lock;                /*
 236                                          * protect map scan related fields like
 237                                          * swap_map, lowest_bit, highest_bit,
 238                                          * inuse_pages, cluster_next,
 239                                          * cluster_nr, lowest_alloc,
 240                                          * highest_alloc, free/discard cluster
 241                                          * list. other fields are only changed
 242                                          * at swapon/swapoff, so are protected
 243                                          * by swap_lock. changing flags need
 244                                          * hold this lock and swap_lock. If
 245                                          * both locks need hold, hold swap_lock
 246                                          * first.
 247                                          */
 248         struct work_struct discard_work; /* discard worker */
 249         struct swap_cluster_list discard_clusters; /* discard clusters list */
 250 };
 251
 252 /* linux/mm/workingset.c */
 253 void *workingset_eviction(struct address_space *mapping, struct page *page);
 254 bool workingset_refault(void *shadow);
 255 void workingset_activation(struct page *page);
 256 void workingset_update_node(struct radix_tree_node *node, void *private);
 257
 258 /* linux/mm/page_alloc.c */
 259 extern unsigned long totalram_pages;
 260 extern unsigned long totalreserve_pages;
 261 extern unsigned long nr_free_buffer_pages(void);
 262 extern unsigned long nr_free_pagecache_pages(void);
 263
 264 /* Definition of global_page_state not available yet */
 265 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
 266
 267
 268 /* linux/mm/swap.c */
 269 extern void lru_cache_add(struct page *);
 270 extern void lru_cache_add_anon(struct page *page);
 271 extern void lru_cache_add_file(struct page *page);
 272 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
 273                          struct lruvec *lruvec, struct list_head *head);
 274 extern void activate_page(struct page *);
 275 extern void mark_page_accessed(struct page *);
 276 extern void lru_add_drain(void);
 277 extern void lru_add_drain_cpu(int cpu);
 278 extern void lru_add_drain_all(void);
 279 extern void rotate_reclaimable_page(struct page *page);
 280 extern void deactivate_file_page(struct page *page);
 281 extern void deactivate_page(struct page *page);
 282 extern void swap_setup(void);
 283
 284 extern void add_page_to_unevictable_list(struct page *page);
 285
 286 extern void lru_cache_add_active_or_unevictable(struct page *page,
 287                                                 struct vm_area_struct *vma);
 288
 289 /* linux/mm/vmscan.c */
 290 extern unsigned long zone_reclaimable_pages(struct zone *zone);
 291 extern unsigned long pgdat_reclaimable_pages(struct pglist_data *pgdat);
 292 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
 293                                         gfp_t gfp_mask, nodemask_t *mask);
 294 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
 295 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
 296                                                   unsigned long nr_pages,
 297                                                   gfp_t gfp_mask,
 298                                                   bool may_swap);
 299 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
 300                                                 gfp_t gfp_mask, bool noswap,
 301                                                 pg_data_t *pgdat,
 302                                                 unsigned long *nr_scanned);
 303 extern unsigned long shrink_all_memory(unsigned long nr_pages);
 304 extern int vm_swappiness;
 305 extern int remove_mapping(struct address_space *mapping, struct page *page);
 306 extern unsigned long vm_total_pages;
 307
 308 #ifdef CONFIG_NUMA
 309 extern int node_reclaim_mode;
 310 extern int sysctl_min_unmapped_ratio;
 311 extern int sysctl_min_slab_ratio;
 312 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
 313 #else
 314 #define node_reclaim_mode 0
 315 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
 316                                 unsigned int order)
 317 {
 318         return 0;
 319 }
 320 #endif
 321
 322 extern int page_evictable(struct page *page);
 323 extern void check_move_unevictable_pages(struct page **, int nr_pages);
 324
 325 extern int kswapd_run(int nid);
 326 extern void kswapd_stop(int nid);
 327
 328 #ifdef CONFIG_SWAP
 329
 330 #include <linux/blk_types.h> /* for bio_end_io_t */
 331
 332 /* linux/mm/page_io.c */
 333 extern int swap_readpage(struct page *);
 334 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
 335 extern void end_swap_bio_write(struct bio *bio);
 336 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
 337         bio_end_io_t end_write_func);
 338 extern int swap_set_page_dirty(struct page *page);
 339
 340 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
 341                 unsigned long nr_pages, sector_t start_block);
 342 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
 343                 sector_t *);
 344
 345 /* linux/mm/swap_state.c */
 346 /* One swap address space for each 64M swap space */
 347 #define SWAP_ADDRESS_SPACE_SHIFT        14
 348 #define SWAP_ADDRESS_SPACE_PAGES        (1 << SWAP_ADDRESS_SPACE_SHIFT)
 349 extern struct address_space *swapper_spaces[];
 350 #define swap_address_space(entry)                           \
 351         (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
 352                 >> SWAP_ADDRESS_SPACE_SHIFT])
 353 extern unsigned long total_swapcache_pages(void);
 354 extern void show_swap_cache_info(void);
 355 extern int add_to_swap(struct page *, struct list_head *list);
 356 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
 357 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
 358 extern void __delete_from_swap_cache(struct page *);
 359 extern void delete_from_swap_cache(struct page *);
 360 extern void free_page_and_swap_cache(struct page *);
 361 extern void free_pages_and_swap_cache(struct page **, int);
 362 extern struct page *lookup_swap_cache(swp_entry_t);
 363 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
 364                         struct vm_area_struct *vma, unsigned long addr);
 365 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
 366                         struct vm_area_struct *vma, unsigned long addr,
 367                         bool *new_page_allocated);
 368 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
 369                         struct vm_area_struct *vma, unsigned long addr);
 370
 371 /* linux/mm/swapfile.c */
 372 extern atomic_long_t nr_swap_pages;
 373 extern long total_swap_pages;
 374
 375 /* Swap 50% full? Release swapcache more aggressively.. */
 376 static inline bool vm_swap_full(void)
 377 {
 378         return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
 379 }
 380
 381 static inline long get_nr_swap_pages(void)
 382 {
 383         return atomic_long_read(&nr_swap_pages);
 384 }
 385
 386 extern void si_swapinfo(struct sysinfo *);
 387 extern swp_entry_t get_swap_page(void);
 388 extern swp_entry_t get_swap_page_of_type(int);
 389 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
 390 extern void swap_shmem_alloc(swp_entry_t);
 391 extern int swap_duplicate(swp_entry_t);
 392 extern int swapcache_prepare(swp_entry_t);
 393 extern void swap_free(swp_entry_t);
 394 extern void swapcache_free(swp_entry_t);
 395 extern int free_swap_and_cache(swp_entry_t);
 396 extern int swap_type_of(dev_t, sector_t, struct block_device **);
 397 extern unsigned int count_swap_pages(int, int);
 398 extern sector_t map_swap_page(struct page *, struct block_device **);
 399 extern sector_t swapdev_block(int, pgoff_t);
 400 extern int page_swapcount(struct page *);
 401 extern int swp_swapcount(swp_entry_t entry);
 402 extern struct swap_info_struct *page_swap_info(struct page *);
 403 extern bool reuse_swap_page(struct page *, int *);
 404 extern int try_to_free_swap(struct page *);
 405 struct backing_dev_info;
 406 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
 407 extern void exit_swap_address_space(unsigned int type);
 408
 409 #else /* CONFIG_SWAP */
 410
 411 #define swap_address_space(entry)               (NULL)
 412 #define get_nr_swap_pages()                     0L
 413 #define total_swap_pages                        0L
 414 #define total_swapcache_pages()                 0UL
 415 #define vm_swap_full()                          0
 416
 417 #define si_swapinfo(val) \
 418         do { (val)->freeswap = (val)->totalswap = 0; } while (0)
 419 /* only sparc can not include linux/pagemap.h in this file
 420  * so leave put_page and release_pages undeclared... */
 421 #define free_page_and_swap_cache(page) \
 422         put_page(page)
 423 #define free_pages_and_swap_cache(pages, nr) \
 424         release_pages((pages), (nr), false);
 425
 426 static inline void show_swap_cache_info(void)
 427 {
 428 }
 429
 430 #define free_swap_and_cache(swp)        is_migration_entry(swp)
 431 #define swapcache_prepare(swp)          is_migration_entry(swp)
 432
 433 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
 434 {
 435         return 0;
 436 }
 437
 438 static inline void swap_shmem_alloc(swp_entry_t swp)
 439 {
 440 }
 441
 442 static inline int swap_duplicate(swp_entry_t swp)
 443 {
 444         return 0;
 445 }
 446
 447 static inline void swap_free(swp_entry_t swp)
 448 {
 449 }
 450
 451 static inline void swapcache_free(swp_entry_t swp)
 452 {
 453 }
 454
 455 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
 456                         struct vm_area_struct *vma, unsigned long addr)
 457 {
 458         return NULL;
 459 }
 460
 461 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
 462 {
 463         return 0;
 464 }
 465
 466 static inline struct page *lookup_swap_cache(swp_entry_t swp)
 467 {
 468         return NULL;
 469 }
 470
 471 static inline int add_to_swap(struct page *page, struct list_head *list)
 472 {
 473         return 0;
 474 }
 475
 476 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
 477                                                         gfp_t gfp_mask)
 478 {
 479         return -1;
 480 }
 481
 482 static inline void __delete_from_swap_cache(struct page *page)
 483 {
 484 }
 485
 486 static inline void delete_from_swap_cache(struct page *page)
 487 {
 488 }
 489
 490 static inline int page_swapcount(struct page *page)
 491 {
 492         return 0;
 493 }
 494
 495 static inline int swp_swapcount(swp_entry_t entry)
 496 {
 497         return 0;
 498 }
 499
 500 #define reuse_swap_page(page, total_mapcount) \
 501         (page_trans_huge_mapcount(page, total_mapcount) == 1)
 502
 503 static inline int try_to_free_swap(struct page *page)
 504 {
 505         return 0;
 506 }
 507
 508 static inline swp_entry_t get_swap_page(void)
 509 {
 510         swp_entry_t entry;
 511         entry.val = 0;
 512         return entry;
 513 }
 514
 515 #endif /* CONFIG_SWAP */
 516
 517 #ifdef CONFIG_MEMCG
 518 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
 519 {
 520         /* Cgroup2 doesn't have per-cgroup swappiness */
 521         if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
 522                 return vm_swappiness;
 523
 524         /* root ? */
 525         if (mem_cgroup_disabled() || !memcg->css.parent)
 526                 return vm_swappiness;
 527
 528         return memcg->swappiness;
 529 }
 530
 531 #else
 532 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
 533 {
 534         return vm_swappiness;
 535 }
 536 #endif
 537
 538 #ifdef CONFIG_MEMCG_SWAP
 539 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
 540 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
 541 extern void mem_cgroup_uncharge_swap(swp_entry_t entry);
 542 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
 543 extern bool mem_cgroup_swap_full(struct page *page);
 544 #else
 545 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
 546 {
 547 }
 548
 549 static inline int mem_cgroup_try_charge_swap(struct page *page,
 550                                              swp_entry_t entry)
 551 {
 552         return 0;
 553 }
 554
 555 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry)
 556 {
 557 }
 558
 559 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
 560 {
 561         return get_nr_swap_pages();
 562 }
 563
 564 static inline bool mem_cgroup_swap_full(struct page *page)
 565 {
 566         return vm_swap_full();
 567 }
 568 #endif
 569
 570 #endif /* __KERNEL__*/
 571 #endif /* _LINUX_SWAP_H */