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  /* A swap entry has to fit into a "unsigned long", as
 106   * the entry is hidden in the "index" field of the
 107   * swapper address space.
 108   */
 109 typedef struct {
 110         unsigned long val;
 111 } swp_entry_t;
 112
 113 /*
 114  * current->reclaim_state points to one of these when a task is running
 115  * memory reclaim
 116  */
 117 struct reclaim_state {
 118         unsigned long reclaimed_slab;
 119 };
 120
 121 #ifdef __KERNEL__
 122
 123 struct address_space;
 124 struct sysinfo;
 125 struct writeback_control;
 126 struct zone;
 127
 128 /*
 129  * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
 130  * disk blocks.  A list of swap extents maps the entire swapfile.  (Where the
 131  * term `swapfile' refers to either a blockdevice or an IS_REG file.  Apart
 132  * from setup, they're handled identically.
 133  *
 134  * We always assume that blocks are of size PAGE_SIZE.
 135  */
 136 struct swap_extent {
 137         struct list_head list;
 138         pgoff_t start_page;
 139         pgoff_t nr_pages;
 140         sector_t start_block;
 141 };
 142
 143 /*
 144  * Max bad pages in the new format..
 145  */
 146 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
 147 #define MAX_SWAP_BADPAGES \
 148         ((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
 149
 150 enum {
 151         SWP_USED        = (1 << 0),     /* is slot in swap_info[] used? */
 152         SWP_WRITEOK     = (1 << 1),     /* ok to write to this swap?    */
 153         SWP_DISCARDABLE = (1 << 2),     /* blkdev support discard */
 154         SWP_DISCARDING  = (1 << 3),     /* now discarding a free cluster */
 155         SWP_SOLIDSTATE  = (1 << 4),     /* blkdev seeks are cheap */
 156         SWP_CONTINUED   = (1 << 5),     /* swap_map has count continuation */
 157         SWP_BLKDEV      = (1 << 6),     /* its a block device */
 158         SWP_FILE        = (1 << 7),     /* set after swap_activate success */
 159         SWP_AREA_DISCARD = (1 << 8),    /* single-time swap area discards */
 160         SWP_PAGE_DISCARD = (1 << 9),    /* freed swap page-cluster discards */
 161                                         /* add others here before... */
 162         SWP_SCANNING    = (1 << 10),    /* refcount in scan_swap_map */
 163 };
 164
 165 #define SWAP_CLUSTER_MAX 32UL
 166 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
 167
 168 /*
 169  * Ratio between the present memory in the zone and the "gap" that
 170  * we're allowing kswapd to shrink in addition to the per-zone high
 171  * wmark, even for zones that already have the high wmark satisfied,
 172  * in order to provide better per-zone lru behavior. We are ok to
 173  * spend not more than 1% of the memory for this zone balancing "gap".
 174  */
 175 #define KSWAPD_ZONE_BALANCE_GAP_RATIO 100
 176
 177 #define SWAP_MAP_MAX    0x3e    /* Max duplication count, in first swap_map */
 178 #define SWAP_MAP_BAD    0x3f    /* Note pageblock is bad, in first swap_map */
 179 #define SWAP_HAS_CACHE  0x40    /* Flag page is cached, in first swap_map */
 180 #define SWAP_CONT_MAX   0x7f    /* Max count, in each swap_map continuation */
 181 #define COUNT_CONTINUED 0x80    /* See swap_map continuation for full count */
 182 #define SWAP_MAP_SHMEM  0xbf    /* Owned by shmem/tmpfs, in first swap_map */
 183
 184 /*
 185  * We use this to track usage of a cluster. A cluster is a block of swap disk
 186  * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
 187  * free clusters are organized into a list. We fetch an entry from the list to
 188  * get a free cluster.
 189  *
 190  * The data field stores next cluster if the cluster is free or cluster usage
 191  * counter otherwise. The flags field determines if a cluster is free. This is
 192  * protected by swap_info_struct.lock.
 193  */
 194 struct swap_cluster_info {
 195         unsigned int data:24;
 196         unsigned int flags:8;
 197 };
 198 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
 199 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
 200
 201 /*
 202  * We assign a cluster to each CPU, so each CPU can allocate swap entry from
 203  * its own cluster and swapout sequentially. The purpose is to optimize swapout
 204  * throughput.
 205  */
 206 struct percpu_cluster {
 207         struct swap_cluster_info index; /* Current cluster index */
 208         unsigned int next; /* Likely next allocation offset */
 209 };
 210
 211 /*
 212  * The in-memory structure used to track swap areas.
 213  */
 214 struct swap_info_struct {
 215         unsigned long   flags;          /* SWP_USED etc: see above */
 216         signed short    prio;           /* swap priority of this type */
 217         signed char     type;           /* strange name for an index */
 218         signed char     next;           /* next type on the swap list */
 219         unsigned int    max;            /* extent of the swap_map */
 220         unsigned char *swap_map;        /* vmalloc'ed array of usage counts */
 221         struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
 222         struct swap_cluster_info free_cluster_head; /* free cluster list head */
 223         struct swap_cluster_info free_cluster_tail; /* free cluster list tail */
 224         unsigned int lowest_bit;        /* index of first free in swap_map */
 225         unsigned int highest_bit;       /* index of last free in swap_map */
 226         unsigned int pages;             /* total of usable pages of swap */
 227         unsigned int inuse_pages;       /* number of those currently in use */
 228         unsigned int cluster_next;      /* likely index for next allocation */
 229         unsigned int cluster_nr;        /* countdown to next cluster search */
 230         struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
 231         struct swap_extent *curr_swap_extent;
 232         struct swap_extent first_swap_extent;
 233         struct block_device *bdev;      /* swap device or bdev of swap file */
 234         struct file *swap_file;         /* seldom referenced */
 235         unsigned int old_block_size;    /* seldom referenced */
 236 #ifdef CONFIG_FRONTSWAP
 237         unsigned long *frontswap_map;   /* frontswap in-use, one bit per page */
 238         atomic_t frontswap_pages;       /* frontswap pages in-use counter */
 239 #endif
 240         spinlock_t lock;                /*
 241                                          * protect map scan related fields like
 242                                          * swap_map, lowest_bit, highest_bit,
 243                                          * inuse_pages, cluster_next,
 244                                          * cluster_nr, lowest_alloc,
 245                                          * highest_alloc, free/discard cluster
 246                                          * list. other fields are only changed
 247                                          * at swapon/swapoff, so are protected
 248                                          * by swap_lock. changing flags need
 249                                          * hold this lock and swap_lock. If
 250                                          * both locks need hold, hold swap_lock
 251                                          * first.
 252                                          */
 253         struct work_struct discard_work; /* discard worker */
 254         struct swap_cluster_info discard_cluster_head; /* list head of discard clusters */
 255         struct swap_cluster_info discard_cluster_tail; /* list tail of discard clusters */
 256 };
 257
 258 struct swap_list_t {
 259         int head;       /* head of priority-ordered swapfile list */
 260         int next;       /* swapfile to be used next */
 261 };
 262
 263 /* linux/mm/workingset.c */
 264 void *workingset_eviction(struct address_space *mapping, struct page *page);
 265 bool workingset_refault(void *shadow);
 266 void workingset_activation(struct page *page);
 267 extern struct list_lru workingset_shadow_nodes;
 268
 269 static inline unsigned int workingset_node_pages(struct radix_tree_node *node)
 270 {
 271         return node->count & RADIX_TREE_COUNT_MASK;
 272 }
 273
 274 static inline void workingset_node_pages_inc(struct radix_tree_node *node)
 275 {
 276         node->count++;
 277 }
 278
 279 static inline void workingset_node_pages_dec(struct radix_tree_node *node)
 280 {
 281         node->count--;
 282 }
 283
 284 static inline unsigned int workingset_node_shadows(struct radix_tree_node *node)
 285 {
 286         return node->count >> RADIX_TREE_COUNT_SHIFT;
 287 }
 288
 289 static inline void workingset_node_shadows_inc(struct radix_tree_node *node)
 290 {
 291         node->count += 1U << RADIX_TREE_COUNT_SHIFT;
 292 }
 293
 294 static inline void workingset_node_shadows_dec(struct radix_tree_node *node)
 295 {
 296         node->count -= 1U << RADIX_TREE_COUNT_SHIFT;
 297 }
 298
 299 /* linux/mm/page_alloc.c */
 300 extern unsigned long totalram_pages;
 301 extern unsigned long totalreserve_pages;
 302 extern unsigned long dirty_balance_reserve;
 303 extern unsigned long nr_free_buffer_pages(void);
 304 extern unsigned long nr_free_pagecache_pages(void);
 305
 306 /* Definition of global_page_state not available yet */
 307 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
 308
 309
 310 /* linux/mm/swap.c */
 311 extern void lru_cache_add(struct page *);
 312 extern void lru_cache_add_anon(struct page *page);
 313 extern void lru_cache_add_file(struct page *page);
 314 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
 315                          struct lruvec *lruvec, struct list_head *head);
 316 extern void activate_page(struct page *);
 317 extern void mark_page_accessed(struct page *);
 318 extern void lru_add_drain(void);
 319 extern void lru_add_drain_cpu(int cpu);
 320 extern void lru_add_drain_all(void);
 321 extern void rotate_reclaimable_page(struct page *page);
 322 extern void deactivate_page(struct page *page);
 323 extern void swap_setup(void);
 324
 325 extern void add_page_to_unevictable_list(struct page *page);
 326
 327 /* linux/mm/vmscan.c */
 328 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
 329                                         gfp_t gfp_mask, nodemask_t *mask);
 330 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
 331 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem,
 332                                                   gfp_t gfp_mask, bool noswap);
 333 extern unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
 334                                                 gfp_t gfp_mask, bool noswap,
 335                                                 struct zone *zone,
 336                                                 unsigned long *nr_scanned);
 337 extern unsigned long shrink_all_memory(unsigned long nr_pages);
 338 extern int vm_swappiness;
 339 extern int remove_mapping(struct address_space *mapping, struct page *page);
 340 extern unsigned long vm_total_pages;
 341
 342 #ifdef CONFIG_NUMA
 343 extern int zone_reclaim_mode;
 344 extern int sysctl_min_unmapped_ratio;
 345 extern int sysctl_min_slab_ratio;
 346 extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
 347 #else
 348 #define zone_reclaim_mode 0
 349 static inline int zone_reclaim(struct zone *z, gfp_t mask, unsigned int order)
 350 {
 351         return 0;
 352 }
 353 #endif
 354
 355 extern int page_evictable(struct page *page);
 356 extern void check_move_unevictable_pages(struct page **, int nr_pages);
 357
 358 extern unsigned long scan_unevictable_pages;
 359 extern int scan_unevictable_handler(struct ctl_table *, int,
 360                                         void __user *, size_t *, loff_t *);
 361 #ifdef CONFIG_NUMA
 362 extern int scan_unevictable_register_node(struct node *node);
 363 extern void scan_unevictable_unregister_node(struct node *node);
 364 #else
 365 static inline int scan_unevictable_register_node(struct node *node)
 366 {
 367         return 0;
 368 }
 369 static inline void scan_unevictable_unregister_node(struct node *node)
 370 {
 371 }
 372 #endif
 373
 374 extern int kswapd_run(int nid);
 375 extern void kswapd_stop(int nid);
 376 #ifdef CONFIG_MEMCG
 377 extern int mem_cgroup_swappiness(struct mem_cgroup *mem);
 378 #else
 379 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
 380 {
 381         return vm_swappiness;
 382 }
 383 #endif
 384 #ifdef CONFIG_MEMCG_SWAP
 385 extern void mem_cgroup_uncharge_swap(swp_entry_t ent);
 386 #else
 387 static inline void mem_cgroup_uncharge_swap(swp_entry_t ent)
 388 {
 389 }
 390 #endif
 391 #ifdef CONFIG_SWAP
 392 /* linux/mm/page_io.c */
 393 extern int swap_readpage(struct page *);
 394 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
 395 extern void end_swap_bio_write(struct bio *bio, int err);
 396 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
 397         void (*end_write_func)(struct bio *, int));
 398 extern int swap_set_page_dirty(struct page *page);
 399 extern void end_swap_bio_read(struct bio *bio, int err);
 400
 401 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
 402                 unsigned long nr_pages, sector_t start_block);
 403 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
 404                 sector_t *);
 405
 406 /* linux/mm/swap_state.c */
 407 extern struct address_space swapper_spaces[];
 408 #define swap_address_space(entry) (&swapper_spaces[swp_type(entry)])
 409 extern unsigned long total_swapcache_pages(void);
 410 extern void show_swap_cache_info(void);
 411 extern int add_to_swap(struct page *, struct list_head *list);
 412 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
 413 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
 414 extern void __delete_from_swap_cache(struct page *);
 415 extern void delete_from_swap_cache(struct page *);
 416 extern void free_page_and_swap_cache(struct page *);
 417 extern void free_pages_and_swap_cache(struct page **, int);
 418 extern struct page *lookup_swap_cache(swp_entry_t);
 419 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
 420                         struct vm_area_struct *vma, unsigned long addr);
 421 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
 422                         struct vm_area_struct *vma, unsigned long addr);
 423
 424 /* linux/mm/swapfile.c */
 425 extern atomic_long_t nr_swap_pages;
 426 extern long total_swap_pages;
 427
 428 /* Swap 50% full? Release swapcache more aggressively.. */
 429 static inline bool vm_swap_full(void)
 430 {
 431         return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
 432 }
 433
 434 static inline long get_nr_swap_pages(void)
 435 {
 436         return atomic_long_read(&nr_swap_pages);
 437 }
 438
 439 extern void si_swapinfo(struct sysinfo *);
 440 extern swp_entry_t get_swap_page(void);
 441 extern swp_entry_t get_swap_page_of_type(int);
 442 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
 443 extern void swap_shmem_alloc(swp_entry_t);
 444 extern int swap_duplicate(swp_entry_t);
 445 extern int swapcache_prepare(swp_entry_t);
 446 extern void swap_free(swp_entry_t);
 447 extern void swapcache_free(swp_entry_t, struct page *page);
 448 extern int free_swap_and_cache(swp_entry_t);
 449 extern int swap_type_of(dev_t, sector_t, struct block_device **);
 450 extern unsigned int count_swap_pages(int, int);
 451 extern sector_t map_swap_page(struct page *, struct block_device **);
 452 extern sector_t swapdev_block(int, pgoff_t);
 453 extern int page_swapcount(struct page *);
 454 extern struct swap_info_struct *page_swap_info(struct page *);
 455 extern int reuse_swap_page(struct page *);
 456 extern int try_to_free_swap(struct page *);
 457 struct backing_dev_info;
 458
 459 #ifdef CONFIG_MEMCG
 460 extern void
 461 mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout);
 462 #else
 463 static inline void
 464 mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
 465 {
 466 }
 467 #endif
 468
 469 #else /* CONFIG_SWAP */
 470
 471 #define swap_address_space(entry)               (NULL)
 472 #define get_nr_swap_pages()                     0L
 473 #define total_swap_pages                        0L
 474 #define total_swapcache_pages()                 0UL
 475 #define vm_swap_full()                          0
 476
 477 #define si_swapinfo(val) \
 478         do { (val)->freeswap = (val)->totalswap = 0; } while (0)
 479 /* only sparc can not include linux/pagemap.h in this file
 480  * so leave page_cache_release and release_pages undeclared... */
 481 #define free_page_and_swap_cache(page) \
 482         page_cache_release(page)
 483 #define free_pages_and_swap_cache(pages, nr) \
 484         release_pages((pages), (nr), 0);
 485
 486 static inline void show_swap_cache_info(void)
 487 {
 488 }
 489
 490 #define free_swap_and_cache(swp)        is_migration_entry(swp)
 491 #define swapcache_prepare(swp)          is_migration_entry(swp)
 492
 493 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
 494 {
 495         return 0;
 496 }
 497
 498 static inline void swap_shmem_alloc(swp_entry_t swp)
 499 {
 500 }
 501
 502 static inline int swap_duplicate(swp_entry_t swp)
 503 {
 504         return 0;
 505 }
 506
 507 static inline void swap_free(swp_entry_t swp)
 508 {
 509 }
 510
 511 static inline void swapcache_free(swp_entry_t swp, struct page *page)
 512 {
 513 }
 514
 515 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
 516                         struct vm_area_struct *vma, unsigned long addr)
 517 {
 518         return NULL;
 519 }
 520
 521 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
 522 {
 523         return 0;
 524 }
 525
 526 static inline struct page *lookup_swap_cache(swp_entry_t swp)
 527 {
 528         return NULL;
 529 }
 530
 531 static inline int add_to_swap(struct page *page, struct list_head *list)
 532 {
 533         return 0;
 534 }
 535
 536 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
 537                                                         gfp_t gfp_mask)
 538 {
 539         return -1;
 540 }
 541
 542 static inline void __delete_from_swap_cache(struct page *page)
 543 {
 544 }
 545
 546 static inline void delete_from_swap_cache(struct page *page)
 547 {
 548 }
 549
 550 static inline int page_swapcount(struct page *page)
 551 {
 552         return 0;
 553 }
 554
 555 #define reuse_swap_page(page)   (page_mapcount(page) == 1)
 556
 557 static inline int try_to_free_swap(struct page *page)
 558 {
 559         return 0;
 560 }
 561
 562 static inline swp_entry_t get_swap_page(void)
 563 {
 564         swp_entry_t entry;
 565         entry.val = 0;
 566         return entry;
 567 }
 568
 569 static inline void
 570 mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
 571 {
 572 }
 573
 574 #endif /* CONFIG_SWAP */
 575 #endif /* __KERNEL__*/
 576 #endif /* _LINUX_SWAP_H */