PageRenderTime 51ms CodeModel.GetById 24ms RepoModel.GetById 0ms app.codeStats 0ms

/fs/xfs/xfs_buf.h

https://github.com/gby/linux
C Header | 396 lines | 259 code | 51 blank | 86 comment | 0 complexity | 022f8b076cbb04bce7cb9eadd6420a5f MD5 | raw file
  1. /*
  2. * Copyright (c) 2000-2005 Silicon Graphics, Inc.
  3. * All Rights Reserved.
  4. *
  5. * This program is free software; you can redistribute it and/or
  6. * modify it under the terms of the GNU General Public License as
  7. * published by the Free Software Foundation.
  8. *
  9. * This program is distributed in the hope that it would be useful,
  10. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  12. * GNU General Public License for more details.
  13. *
  14. * You should have received a copy of the GNU General Public License
  15. * along with this program; if not, write the Free Software Foundation,
  16. * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  17. */
  18. #ifndef __XFS_BUF_H__
  19. #define __XFS_BUF_H__
  20. #include <linux/list.h>
  21. #include <linux/types.h>
  22. #include <linux/spinlock.h>
  23. #include <linux/mm.h>
  24. #include <linux/fs.h>
  25. #include <linux/dax.h>
  26. #include <linux/buffer_head.h>
  27. #include <linux/uio.h>
  28. #include <linux/list_lru.h>
  29. /*
  30. * Base types
  31. */
  32. #define XFS_BUF_DADDR_NULL ((xfs_daddr_t) (-1LL))
  33. typedef enum {
  34. XBRW_READ = 1, /* transfer into target memory */
  35. XBRW_WRITE = 2, /* transfer from target memory */
  36. XBRW_ZERO = 3, /* Zero target memory */
  37. } xfs_buf_rw_t;
  38. #define XBF_READ (1 << 0) /* buffer intended for reading from device */
  39. #define XBF_WRITE (1 << 1) /* buffer intended for writing to device */
  40. #define XBF_READ_AHEAD (1 << 2) /* asynchronous read-ahead */
  41. #define XBF_NO_IOACCT (1 << 3) /* bypass I/O accounting (non-LRU bufs) */
  42. #define XBF_ASYNC (1 << 4) /* initiator will not wait for completion */
  43. #define XBF_DONE (1 << 5) /* all pages in the buffer uptodate */
  44. #define XBF_STALE (1 << 6) /* buffer has been staled, do not find it */
  45. #define XBF_WRITE_FAIL (1 << 24)/* async writes have failed on this buffer */
  46. /* I/O hints for the BIO layer */
  47. #define XBF_SYNCIO (1 << 10)/* treat this buffer as synchronous I/O */
  48. #define XBF_FUA (1 << 11)/* force cache write through mode */
  49. #define XBF_FLUSH (1 << 12)/* flush the disk cache before a write */
  50. /* flags used only as arguments to access routines */
  51. #define XBF_TRYLOCK (1 << 16)/* lock requested, but do not wait */
  52. #define XBF_UNMAPPED (1 << 17)/* do not map the buffer */
  53. /* flags used only internally */
  54. #define _XBF_PAGES (1 << 20)/* backed by refcounted pages */
  55. #define _XBF_KMEM (1 << 21)/* backed by heap memory */
  56. #define _XBF_DELWRI_Q (1 << 22)/* buffer on a delwri queue */
  57. #define _XBF_COMPOUND (1 << 23)/* compound buffer */
  58. typedef unsigned int xfs_buf_flags_t;
  59. #define XFS_BUF_FLAGS \
  60. { XBF_READ, "READ" }, \
  61. { XBF_WRITE, "WRITE" }, \
  62. { XBF_READ_AHEAD, "READ_AHEAD" }, \
  63. { XBF_NO_IOACCT, "NO_IOACCT" }, \
  64. { XBF_ASYNC, "ASYNC" }, \
  65. { XBF_DONE, "DONE" }, \
  66. { XBF_STALE, "STALE" }, \
  67. { XBF_WRITE_FAIL, "WRITE_FAIL" }, \
  68. { XBF_SYNCIO, "SYNCIO" }, \
  69. { XBF_FUA, "FUA" }, \
  70. { XBF_FLUSH, "FLUSH" }, \
  71. { XBF_TRYLOCK, "TRYLOCK" }, /* should never be set */\
  72. { XBF_UNMAPPED, "UNMAPPED" }, /* ditto */\
  73. { _XBF_PAGES, "PAGES" }, \
  74. { _XBF_KMEM, "KMEM" }, \
  75. { _XBF_DELWRI_Q, "DELWRI_Q" }, \
  76. { _XBF_COMPOUND, "COMPOUND" }
  77. /*
  78. * Internal state flags.
  79. */
  80. #define XFS_BSTATE_DISPOSE (1 << 0) /* buffer being discarded */
  81. #define XFS_BSTATE_IN_FLIGHT (1 << 1) /* I/O in flight */
  82. /*
  83. * The xfs_buftarg contains 2 notions of "sector size" -
  84. *
  85. * 1) The metadata sector size, which is the minimum unit and
  86. * alignment of IO which will be performed by metadata operations.
  87. * 2) The device logical sector size
  88. *
  89. * The first is specified at mkfs time, and is stored on-disk in the
  90. * superblock's sb_sectsize.
  91. *
  92. * The latter is derived from the underlying device, and controls direct IO
  93. * alignment constraints.
  94. */
  95. typedef struct xfs_buftarg {
  96. dev_t bt_dev;
  97. struct block_device *bt_bdev;
  98. struct xfs_mount *bt_mount;
  99. unsigned int bt_meta_sectorsize;
  100. size_t bt_meta_sectormask;
  101. size_t bt_logical_sectorsize;
  102. size_t bt_logical_sectormask;
  103. /* LRU control structures */
  104. struct shrinker bt_shrinker;
  105. struct list_lru bt_lru;
  106. struct percpu_counter bt_io_count;
  107. } xfs_buftarg_t;
  108. struct xfs_buf;
  109. typedef void (*xfs_buf_iodone_t)(struct xfs_buf *);
  110. #define XB_PAGES 2
  111. struct xfs_buf_map {
  112. xfs_daddr_t bm_bn; /* block number for I/O */
  113. int bm_len; /* size of I/O */
  114. };
  115. #define DEFINE_SINGLE_BUF_MAP(map, blkno, numblk) \
  116. struct xfs_buf_map (map) = { .bm_bn = (blkno), .bm_len = (numblk) };
  117. struct xfs_buf_ops {
  118. char *name;
  119. void (*verify_read)(struct xfs_buf *);
  120. void (*verify_write)(struct xfs_buf *);
  121. };
  122. typedef struct xfs_buf {
  123. /*
  124. * first cacheline holds all the fields needed for an uncontended cache
  125. * hit to be fully processed. The semaphore straddles the cacheline
  126. * boundary, but the counter and lock sits on the first cacheline,
  127. * which is the only bit that is touched if we hit the semaphore
  128. * fast-path on locking.
  129. */
  130. struct rhash_head b_rhash_head; /* pag buffer hash node */
  131. xfs_daddr_t b_bn; /* block number of buffer */
  132. int b_length; /* size of buffer in BBs */
  133. atomic_t b_hold; /* reference count */
  134. atomic_t b_lru_ref; /* lru reclaim ref count */
  135. xfs_buf_flags_t b_flags; /* status flags */
  136. struct semaphore b_sema; /* semaphore for lockables */
  137. /*
  138. * concurrent access to b_lru and b_lru_flags are protected by
  139. * bt_lru_lock and not by b_sema
  140. */
  141. struct list_head b_lru; /* lru list */
  142. spinlock_t b_lock; /* internal state lock */
  143. unsigned int b_state; /* internal state flags */
  144. int b_io_error; /* internal IO error state */
  145. wait_queue_head_t b_waiters; /* unpin waiters */
  146. struct list_head b_list;
  147. struct xfs_perag *b_pag; /* contains rbtree root */
  148. xfs_buftarg_t *b_target; /* buffer target (device) */
  149. void *b_addr; /* virtual address of buffer */
  150. struct work_struct b_ioend_work;
  151. struct workqueue_struct *b_ioend_wq; /* I/O completion wq */
  152. xfs_buf_iodone_t b_iodone; /* I/O completion function */
  153. struct completion b_iowait; /* queue for I/O waiters */
  154. void *b_fspriv;
  155. struct xfs_trans *b_transp;
  156. struct page **b_pages; /* array of page pointers */
  157. struct page *b_page_array[XB_PAGES]; /* inline pages */
  158. struct xfs_buf_map *b_maps; /* compound buffer map */
  159. struct xfs_buf_map __b_map; /* inline compound buffer map */
  160. int b_map_count;
  161. int b_io_length; /* IO size in BBs */
  162. atomic_t b_pin_count; /* pin count */
  163. atomic_t b_io_remaining; /* #outstanding I/O requests */
  164. unsigned int b_page_count; /* size of page array */
  165. unsigned int b_offset; /* page offset in first page */
  166. int b_error; /* error code on I/O */
  167. /*
  168. * async write failure retry count. Initialised to zero on the first
  169. * failure, then when it exceeds the maximum configured without a
  170. * success the write is considered to be failed permanently and the
  171. * iodone handler will take appropriate action.
  172. *
  173. * For retry timeouts, we record the jiffie of the first failure. This
  174. * means that we can change the retry timeout for buffers already under
  175. * I/O and thus avoid getting stuck in a retry loop with a long timeout.
  176. *
  177. * last_error is used to ensure that we are getting repeated errors, not
  178. * different errors. e.g. a block device might change ENOSPC to EIO when
  179. * a failure timeout occurs, so we want to re-initialise the error
  180. * retry behaviour appropriately when that happens.
  181. */
  182. int b_retries;
  183. unsigned long b_first_retry_time; /* in jiffies */
  184. int b_last_error;
  185. const struct xfs_buf_ops *b_ops;
  186. #ifdef XFS_BUF_LOCK_TRACKING
  187. int b_last_holder;
  188. #endif
  189. } xfs_buf_t;
  190. /* Finding and Reading Buffers */
  191. struct xfs_buf *_xfs_buf_find(struct xfs_buftarg *target,
  192. struct xfs_buf_map *map, int nmaps,
  193. xfs_buf_flags_t flags, struct xfs_buf *new_bp);
  194. static inline struct xfs_buf *
  195. xfs_incore(
  196. struct xfs_buftarg *target,
  197. xfs_daddr_t blkno,
  198. size_t numblks,
  199. xfs_buf_flags_t flags)
  200. {
  201. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  202. return _xfs_buf_find(target, &map, 1, flags, NULL);
  203. }
  204. struct xfs_buf *_xfs_buf_alloc(struct xfs_buftarg *target,
  205. struct xfs_buf_map *map, int nmaps,
  206. xfs_buf_flags_t flags);
  207. static inline struct xfs_buf *
  208. xfs_buf_alloc(
  209. struct xfs_buftarg *target,
  210. xfs_daddr_t blkno,
  211. size_t numblks,
  212. xfs_buf_flags_t flags)
  213. {
  214. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  215. return _xfs_buf_alloc(target, &map, 1, flags);
  216. }
  217. struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
  218. struct xfs_buf_map *map, int nmaps,
  219. xfs_buf_flags_t flags);
  220. struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
  221. struct xfs_buf_map *map, int nmaps,
  222. xfs_buf_flags_t flags,
  223. const struct xfs_buf_ops *ops);
  224. void xfs_buf_readahead_map(struct xfs_buftarg *target,
  225. struct xfs_buf_map *map, int nmaps,
  226. const struct xfs_buf_ops *ops);
  227. static inline struct xfs_buf *
  228. xfs_buf_get(
  229. struct xfs_buftarg *target,
  230. xfs_daddr_t blkno,
  231. size_t numblks,
  232. xfs_buf_flags_t flags)
  233. {
  234. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  235. return xfs_buf_get_map(target, &map, 1, flags);
  236. }
  237. static inline struct xfs_buf *
  238. xfs_buf_read(
  239. struct xfs_buftarg *target,
  240. xfs_daddr_t blkno,
  241. size_t numblks,
  242. xfs_buf_flags_t flags,
  243. const struct xfs_buf_ops *ops)
  244. {
  245. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  246. return xfs_buf_read_map(target, &map, 1, flags, ops);
  247. }
  248. static inline void
  249. xfs_buf_readahead(
  250. struct xfs_buftarg *target,
  251. xfs_daddr_t blkno,
  252. size_t numblks,
  253. const struct xfs_buf_ops *ops)
  254. {
  255. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  256. return xfs_buf_readahead_map(target, &map, 1, ops);
  257. }
  258. void xfs_buf_set_empty(struct xfs_buf *bp, size_t numblks);
  259. int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length);
  260. struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
  261. int flags);
  262. int xfs_buf_read_uncached(struct xfs_buftarg *target, xfs_daddr_t daddr,
  263. size_t numblks, int flags, struct xfs_buf **bpp,
  264. const struct xfs_buf_ops *ops);
  265. void xfs_buf_hold(struct xfs_buf *bp);
  266. /* Releasing Buffers */
  267. extern void xfs_buf_free(xfs_buf_t *);
  268. extern void xfs_buf_rele(xfs_buf_t *);
  269. /* Locking and Unlocking Buffers */
  270. extern int xfs_buf_trylock(xfs_buf_t *);
  271. extern void xfs_buf_lock(xfs_buf_t *);
  272. extern void xfs_buf_unlock(xfs_buf_t *);
  273. #define xfs_buf_islocked(bp) \
  274. ((bp)->b_sema.count <= 0)
  275. /* Buffer Read and Write Routines */
  276. extern int xfs_bwrite(struct xfs_buf *bp);
  277. extern void xfs_buf_ioend(struct xfs_buf *bp);
  278. extern void xfs_buf_ioerror(xfs_buf_t *, int);
  279. extern void xfs_buf_ioerror_alert(struct xfs_buf *, const char *func);
  280. extern void xfs_buf_submit(struct xfs_buf *bp);
  281. extern int xfs_buf_submit_wait(struct xfs_buf *bp);
  282. extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *,
  283. xfs_buf_rw_t);
  284. #define xfs_buf_zero(bp, off, len) \
  285. xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO)
  286. /* Buffer Utility Routines */
  287. extern void *xfs_buf_offset(struct xfs_buf *, size_t);
  288. extern void xfs_buf_stale(struct xfs_buf *bp);
  289. /* Delayed Write Buffer Routines */
  290. extern void xfs_buf_delwri_cancel(struct list_head *);
  291. extern bool xfs_buf_delwri_queue(struct xfs_buf *, struct list_head *);
  292. extern int xfs_buf_delwri_submit(struct list_head *);
  293. extern int xfs_buf_delwri_submit_nowait(struct list_head *);
  294. extern int xfs_buf_delwri_pushbuf(struct xfs_buf *, struct list_head *);
  295. /* Buffer Daemon Setup Routines */
  296. extern int xfs_buf_init(void);
  297. extern void xfs_buf_terminate(void);
  298. /*
  299. * These macros use the IO block map rather than b_bn. b_bn is now really
  300. * just for the buffer cache index for cached buffers. As IO does not use b_bn
  301. * anymore, uncached buffers do not use b_bn at all and hence must modify the IO
  302. * map directly. Uncached buffers are not allowed to be discontiguous, so this
  303. * is safe to do.
  304. *
  305. * In future, uncached buffers will pass the block number directly to the io
  306. * request function and hence these macros will go away at that point.
  307. */
  308. #define XFS_BUF_ADDR(bp) ((bp)->b_maps[0].bm_bn)
  309. #define XFS_BUF_SET_ADDR(bp, bno) ((bp)->b_maps[0].bm_bn = (xfs_daddr_t)(bno))
  310. static inline void xfs_buf_set_ref(struct xfs_buf *bp, int lru_ref)
  311. {
  312. atomic_set(&bp->b_lru_ref, lru_ref);
  313. }
  314. static inline int xfs_buf_ispinned(struct xfs_buf *bp)
  315. {
  316. return atomic_read(&bp->b_pin_count);
  317. }
  318. static inline void xfs_buf_relse(xfs_buf_t *bp)
  319. {
  320. xfs_buf_unlock(bp);
  321. xfs_buf_rele(bp);
  322. }
  323. static inline int
  324. xfs_buf_verify_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
  325. {
  326. return xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length),
  327. cksum_offset);
  328. }
  329. static inline void
  330. xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
  331. {
  332. xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length),
  333. cksum_offset);
  334. }
  335. /*
  336. * Handling of buftargs.
  337. */
  338. extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *,
  339. struct block_device *);
  340. extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *);
  341. extern void xfs_wait_buftarg(xfs_buftarg_t *);
  342. extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int);
  343. #define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev)
  344. #define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev)
  345. #endif /* __XFS_BUF_H__ */