bitmap.h 9.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279
  1. /* SPDX-License-Identifier: GPL-2.0 */
  2. /*
  3. * bitmap.h: Copyright (C) Peter T. Breuer (ptb@ot.uc3m.es) 2003
  4. *
  5. * additions: Copyright (C) 2003-2004, Paul Clements, SteelEye Technology, Inc.
  6. */
  7. #ifndef BITMAP_H
  8. #define BITMAP_H 1
  9. #define BITMAP_MAJOR_LO 3
  10. /* version 4 insists the bitmap is in little-endian order
  11. * with version 3, it is host-endian which is non-portable
  12. * Version 5 is currently set only for clustered devices
  13. */
  14. #define BITMAP_MAJOR_HI 4
  15. #define BITMAP_MAJOR_CLUSTERED 5
  16. #define BITMAP_MAJOR_HOSTENDIAN 3
  17. /*
  18. * in-memory bitmap:
  19. *
  20. * Use 16 bit block counters to track pending writes to each "chunk".
  21. * The 2 high order bits are special-purpose, the first is a flag indicating
  22. * whether a resync is needed. The second is a flag indicating whether a
  23. * resync is active.
  24. * This means that the counter is actually 14 bits:
  25. *
  26. * +--------+--------+------------------------------------------------+
  27. * | resync | resync | counter |
  28. * | needed | active | |
  29. * | (0-1) | (0-1) | (0-16383) |
  30. * +--------+--------+------------------------------------------------+
  31. *
  32. * The "resync needed" bit is set when:
  33. * a '1' bit is read from storage at startup.
  34. * a write request fails on some drives
  35. * a resync is aborted on a chunk with 'resync active' set
  36. * It is cleared (and resync-active set) when a resync starts across all drives
  37. * of the chunk.
  38. *
  39. *
  40. * The "resync active" bit is set when:
  41. * a resync is started on all drives, and resync_needed is set.
  42. * resync_needed will be cleared (as long as resync_active wasn't already set).
  43. * It is cleared when a resync completes.
  44. *
  45. * The counter counts pending write requests, plus the on-disk bit.
  46. * When the counter is '1' and the resync bits are clear, the on-disk
  47. * bit can be cleared as well, thus setting the counter to 0.
  48. * When we set a bit, or in the counter (to start a write), if the fields is
  49. * 0, we first set the disk bit and set the counter to 1.
  50. *
  51. * If the counter is 0, the on-disk bit is clear and the stripe is clean
  52. * Anything that dirties the stripe pushes the counter to 2 (at least)
  53. * and sets the on-disk bit (lazily).
  54. * If a periodic sweep find the counter at 2, it is decremented to 1.
  55. * If the sweep find the counter at 1, the on-disk bit is cleared and the
  56. * counter goes to zero.
  57. *
  58. * Also, we'll hijack the "map" pointer itself and use it as two 16 bit block
  59. * counters as a fallback when "page" memory cannot be allocated:
  60. *
  61. * Normal case (page memory allocated):
  62. *
  63. * page pointer (32-bit)
  64. *
  65. * [ ] ------+
  66. * |
  67. * +-------> [ ][ ]..[ ] (4096 byte page == 2048 counters)
  68. * c1 c2 c2048
  69. *
  70. * Hijacked case (page memory allocation failed):
  71. *
  72. * hijacked page pointer (32-bit)
  73. *
  74. * [ ][ ] (no page memory allocated)
  75. * counter #1 (16-bit) counter #2 (16-bit)
  76. *
  77. */
  78. #ifdef __KERNEL__
  79. #define PAGE_BITS (PAGE_SIZE << 3)
  80. #define PAGE_BIT_SHIFT (PAGE_SHIFT + 3)
  81. typedef __u16 bitmap_counter_t;
  82. #define COUNTER_BITS 16
  83. #define COUNTER_BIT_SHIFT 4
  84. #define COUNTER_BYTE_SHIFT (COUNTER_BIT_SHIFT - 3)
  85. #define NEEDED_MASK ((bitmap_counter_t) (1 << (COUNTER_BITS - 1)))
  86. #define RESYNC_MASK ((bitmap_counter_t) (1 << (COUNTER_BITS - 2)))
  87. #define COUNTER_MAX ((bitmap_counter_t) RESYNC_MASK - 1)
  88. #define NEEDED(x) (((bitmap_counter_t) x) & NEEDED_MASK)
  89. #define RESYNC(x) (((bitmap_counter_t) x) & RESYNC_MASK)
  90. #define COUNTER(x) (((bitmap_counter_t) x) & COUNTER_MAX)
  91. /* how many counters per page? */
  92. #define PAGE_COUNTER_RATIO (PAGE_BITS / COUNTER_BITS)
  93. /* same, except a shift value for more efficient bitops */
  94. #define PAGE_COUNTER_SHIFT (PAGE_BIT_SHIFT - COUNTER_BIT_SHIFT)
  95. /* same, except a mask value for more efficient bitops */
  96. #define PAGE_COUNTER_MASK (PAGE_COUNTER_RATIO - 1)
  97. #define BITMAP_BLOCK_SHIFT 9
  98. #endif
  99. /*
  100. * bitmap structures:
  101. */
  102. #define BITMAP_MAGIC 0x6d746962
  103. /* use these for bitmap->flags and bitmap->sb->state bit-fields */
  104. enum bitmap_state {
  105. BITMAP_STALE = 1, /* the bitmap file is out of date or had -EIO */
  106. BITMAP_WRITE_ERROR = 2, /* A write error has occurred */
  107. BITMAP_HOSTENDIAN =15,
  108. };
  109. /* the superblock at the front of the bitmap file -- little endian */
  110. typedef struct bitmap_super_s {
  111. __le32 magic; /* 0 BITMAP_MAGIC */
  112. __le32 version; /* 4 the bitmap major for now, could change... */
  113. __u8 uuid[16]; /* 8 128 bit uuid - must match md device uuid */
  114. __le64 events; /* 24 event counter for the bitmap (1)*/
  115. __le64 events_cleared;/*32 event counter when last bit cleared (2) */
  116. __le64 sync_size; /* 40 the size of the md device's sync range(3) */
  117. __le32 state; /* 48 bitmap state information */
  118. __le32 chunksize; /* 52 the bitmap chunk size in bytes */
  119. __le32 daemon_sleep; /* 56 seconds between disk flushes */
  120. __le32 write_behind; /* 60 number of outstanding write-behind writes */
  121. __le32 sectors_reserved; /* 64 number of 512-byte sectors that are
  122. * reserved for the bitmap. */
  123. __le32 nodes; /* 68 the maximum number of nodes in cluster. */
  124. __u8 cluster_name[64]; /* 72 cluster name to which this md belongs */
  125. __u8 pad[256 - 136]; /* set to zero */
  126. } bitmap_super_t;
  127. /* notes:
  128. * (1) This event counter is updated before the eventcounter in the md superblock
  129. * When a bitmap is loaded, it is only accepted if this event counter is equal
  130. * to, or one greater than, the event counter in the superblock.
  131. * (2) This event counter is updated when the other one is *if*and*only*if* the
  132. * array is not degraded. As bits are not cleared when the array is degraded,
  133. * this represents the last time that any bits were cleared.
  134. * If a device is being added that has an event count with this value or
  135. * higher, it is accepted as conforming to the bitmap.
  136. * (3)This is the number of sectors represented by the bitmap, and is the range that
  137. * resync happens across. For raid1 and raid5/6 it is the size of individual
  138. * devices. For raid10 it is the size of the array.
  139. */
  140. #ifdef __KERNEL__
  141. /* the in-memory bitmap is represented by bitmap_pages */
  142. struct bitmap_page {
  143. /*
  144. * map points to the actual memory page
  145. */
  146. char *map;
  147. /*
  148. * in emergencies (when map cannot be alloced), hijack the map
  149. * pointer and use it as two counters itself
  150. */
  151. unsigned int hijacked:1;
  152. /*
  153. * If any counter in this page is '1' or '2' - and so could be
  154. * cleared then that page is marked as 'pending'
  155. */
  156. unsigned int pending:1;
  157. /*
  158. * count of dirty bits on the page
  159. */
  160. unsigned int count:30;
  161. };
  162. /* the main bitmap structure - one per mddev */
  163. struct bitmap {
  164. struct bitmap_counts {
  165. spinlock_t lock;
  166. struct bitmap_page *bp;
  167. unsigned long pages; /* total number of pages
  168. * in the bitmap */
  169. unsigned long missing_pages; /* number of pages
  170. * not yet allocated */
  171. unsigned long chunkshift; /* chunksize = 2^chunkshift
  172. * (for bitops) */
  173. unsigned long chunks; /* Total number of data
  174. * chunks for the array */
  175. } counts;
  176. struct mddev *mddev; /* the md device that the bitmap is for */
  177. __u64 events_cleared;
  178. int need_sync;
  179. struct bitmap_storage {
  180. struct file *file; /* backing disk file */
  181. struct page *sb_page; /* cached copy of the bitmap
  182. * file superblock */
  183. struct page **filemap; /* list of cache pages for
  184. * the file */
  185. unsigned long *filemap_attr; /* attributes associated
  186. * w/ filemap pages */
  187. unsigned long file_pages; /* number of pages in the file*/
  188. unsigned long bytes; /* total bytes in the bitmap */
  189. } storage;
  190. unsigned long flags;
  191. int allclean;
  192. atomic_t behind_writes;
  193. unsigned long behind_writes_used; /* highest actual value at runtime */
  194. /*
  195. * the bitmap daemon - periodically wakes up and sweeps the bitmap
  196. * file, cleaning up bits and flushing out pages to disk as necessary
  197. */
  198. unsigned long daemon_lastrun; /* jiffies of last run */
  199. unsigned long last_end_sync; /* when we lasted called end_sync to
  200. * update bitmap with resync progress */
  201. atomic_t pending_writes; /* pending writes to the bitmap file */
  202. wait_queue_head_t write_wait;
  203. wait_queue_head_t overflow_wait;
  204. wait_queue_head_t behind_wait;
  205. struct kernfs_node *sysfs_can_clear;
  206. int cluster_slot; /* Slot offset for clustered env */
  207. };
  208. /* the bitmap API */
  209. /* these are used only by md/bitmap */
  210. struct bitmap *bitmap_create(struct mddev *mddev, int slot);
  211. int bitmap_load(struct mddev *mddev);
  212. void bitmap_flush(struct mddev *mddev);
  213. void bitmap_destroy(struct mddev *mddev);
  214. void bitmap_print_sb(struct bitmap *bitmap);
  215. void bitmap_update_sb(struct bitmap *bitmap);
  216. void bitmap_status(struct seq_file *seq, struct bitmap *bitmap);
  217. int bitmap_setallbits(struct bitmap *bitmap);
  218. void bitmap_write_all(struct bitmap *bitmap);
  219. void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e);
  220. /* these are exported */
  221. int bitmap_startwrite(struct bitmap *bitmap, sector_t offset,
  222. unsigned long sectors, int behind);
  223. void bitmap_endwrite(struct bitmap *bitmap, sector_t offset,
  224. unsigned long sectors, int success, int behind);
  225. int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks, int degraded);
  226. void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks, int aborted);
  227. void bitmap_close_sync(struct bitmap *bitmap);
  228. void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector, bool force);
  229. void bitmap_sync_with_cluster(struct mddev *mddev,
  230. sector_t old_lo, sector_t old_hi,
  231. sector_t new_lo, sector_t new_hi);
  232. void bitmap_unplug(struct bitmap *bitmap);
  233. void bitmap_daemon_work(struct mddev *mddev);
  234. int bitmap_resize(struct bitmap *bitmap, sector_t blocks,
  235. int chunksize, int init);
  236. struct bitmap *get_bitmap_from_slot(struct mddev *mddev, int slot);
  237. int bitmap_copy_from_slot(struct mddev *mddev, int slot,
  238. sector_t *lo, sector_t *hi, bool clear_bits);
  239. void md_bitmap_free(struct bitmap *bitmap);
  240. void bitmap_wait_behind_writes(struct mddev *mddev);
  241. #endif
  242. #endif