PageRenderTime 49ms CodeModel.GetById 23ms RepoModel.GetById 0ms app.codeStats 0ms

/fs/xfs/xfs_buf.h

https://gitlab.com/freesoftware/linux
C Header | 396 lines | 264 code | 53 blank | 79 comment | 2 complexity | d71fded8608960aed98c74e4826a2bbd MD5 | raw file
  1. // SPDX-License-Identifier: GPL-2.0
  2. /*
  3. * Copyright (c) 2000-2005 Silicon Graphics, Inc.
  4. * All Rights Reserved.
  5. */
  6. #ifndef __XFS_BUF_H__
  7. #define __XFS_BUF_H__
  8. #include <linux/list.h>
  9. #include <linux/types.h>
  10. #include <linux/spinlock.h>
  11. #include <linux/mm.h>
  12. #include <linux/fs.h>
  13. #include <linux/dax.h>
  14. #include <linux/uio.h>
  15. #include <linux/list_lru.h>
  16. /*
  17. * Base types
  18. */
  19. #define XFS_BUF_DADDR_NULL ((xfs_daddr_t) (-1LL))
  20. typedef enum {
  21. XBRW_READ = 1, /* transfer into target memory */
  22. XBRW_WRITE = 2, /* transfer from target memory */
  23. XBRW_ZERO = 3, /* Zero target memory */
  24. } xfs_buf_rw_t;
  25. #define XBF_READ (1 << 0) /* buffer intended for reading from device */
  26. #define XBF_WRITE (1 << 1) /* buffer intended for writing to device */
  27. #define XBF_READ_AHEAD (1 << 2) /* asynchronous read-ahead */
  28. #define XBF_NO_IOACCT (1 << 3) /* bypass I/O accounting (non-LRU bufs) */
  29. #define XBF_ASYNC (1 << 4) /* initiator will not wait for completion */
  30. #define XBF_DONE (1 << 5) /* all pages in the buffer uptodate */
  31. #define XBF_STALE (1 << 6) /* buffer has been staled, do not find it */
  32. #define XBF_WRITE_FAIL (1 << 24)/* async writes have failed on this buffer */
  33. /* I/O hints for the BIO layer */
  34. #define XBF_SYNCIO (1 << 10)/* treat this buffer as synchronous I/O */
  35. #define XBF_FUA (1 << 11)/* force cache write through mode */
  36. #define XBF_FLUSH (1 << 12)/* flush the disk cache before a write */
  37. /* flags used only as arguments to access routines */
  38. #define XBF_TRYLOCK (1 << 16)/* lock requested, but do not wait */
  39. #define XBF_UNMAPPED (1 << 17)/* do not map the buffer */
  40. /* flags used only internally */
  41. #define _XBF_PAGES (1 << 20)/* backed by refcounted pages */
  42. #define _XBF_KMEM (1 << 21)/* backed by heap memory */
  43. #define _XBF_DELWRI_Q (1 << 22)/* buffer on a delwri queue */
  44. #define _XBF_COMPOUND (1 << 23)/* compound buffer */
  45. typedef unsigned int xfs_buf_flags_t;
  46. #define XFS_BUF_FLAGS \
  47. { XBF_READ, "READ" }, \
  48. { XBF_WRITE, "WRITE" }, \
  49. { XBF_READ_AHEAD, "READ_AHEAD" }, \
  50. { XBF_NO_IOACCT, "NO_IOACCT" }, \
  51. { XBF_ASYNC, "ASYNC" }, \
  52. { XBF_DONE, "DONE" }, \
  53. { XBF_STALE, "STALE" }, \
  54. { XBF_WRITE_FAIL, "WRITE_FAIL" }, \
  55. { XBF_SYNCIO, "SYNCIO" }, \
  56. { XBF_FUA, "FUA" }, \
  57. { XBF_FLUSH, "FLUSH" }, \
  58. { XBF_TRYLOCK, "TRYLOCK" }, /* should never be set */\
  59. { XBF_UNMAPPED, "UNMAPPED" }, /* ditto */\
  60. { _XBF_PAGES, "PAGES" }, \
  61. { _XBF_KMEM, "KMEM" }, \
  62. { _XBF_DELWRI_Q, "DELWRI_Q" }, \
  63. { _XBF_COMPOUND, "COMPOUND" }
  64. /*
  65. * Internal state flags.
  66. */
  67. #define XFS_BSTATE_DISPOSE (1 << 0) /* buffer being discarded */
  68. #define XFS_BSTATE_IN_FLIGHT (1 << 1) /* I/O in flight */
  69. /*
  70. * The xfs_buftarg contains 2 notions of "sector size" -
  71. *
  72. * 1) The metadata sector size, which is the minimum unit and
  73. * alignment of IO which will be performed by metadata operations.
  74. * 2) The device logical sector size
  75. *
  76. * The first is specified at mkfs time, and is stored on-disk in the
  77. * superblock's sb_sectsize.
  78. *
  79. * The latter is derived from the underlying device, and controls direct IO
  80. * alignment constraints.
  81. */
  82. typedef struct xfs_buftarg {
  83. dev_t bt_dev;
  84. struct block_device *bt_bdev;
  85. struct dax_device *bt_daxdev;
  86. struct xfs_mount *bt_mount;
  87. unsigned int bt_meta_sectorsize;
  88. size_t bt_meta_sectormask;
  89. size_t bt_logical_sectorsize;
  90. size_t bt_logical_sectormask;
  91. /* LRU control structures */
  92. struct shrinker bt_shrinker;
  93. struct list_lru bt_lru;
  94. struct percpu_counter bt_io_count;
  95. } xfs_buftarg_t;
  96. struct xfs_buf;
  97. typedef void (*xfs_buf_iodone_t)(struct xfs_buf *);
  98. #define XB_PAGES 2
  99. struct xfs_buf_map {
  100. xfs_daddr_t bm_bn; /* block number for I/O */
  101. int bm_len; /* size of I/O */
  102. };
  103. #define DEFINE_SINGLE_BUF_MAP(map, blkno, numblk) \
  104. struct xfs_buf_map (map) = { .bm_bn = (blkno), .bm_len = (numblk) };
  105. struct xfs_buf_ops {
  106. char *name;
  107. union {
  108. __be32 magic[2]; /* v4 and v5 on disk magic values */
  109. __be16 magic16[2]; /* v4 and v5 on disk magic values */
  110. };
  111. void (*verify_read)(struct xfs_buf *);
  112. void (*verify_write)(struct xfs_buf *);
  113. xfs_failaddr_t (*verify_struct)(struct xfs_buf *bp);
  114. };
  115. typedef struct xfs_buf {
  116. /*
  117. * first cacheline holds all the fields needed for an uncontended cache
  118. * hit to be fully processed. The semaphore straddles the cacheline
  119. * boundary, but the counter and lock sits on the first cacheline,
  120. * which is the only bit that is touched if we hit the semaphore
  121. * fast-path on locking.
  122. */
  123. struct rhash_head b_rhash_head; /* pag buffer hash node */
  124. xfs_daddr_t b_bn; /* block number of buffer */
  125. int b_length; /* size of buffer in BBs */
  126. atomic_t b_hold; /* reference count */
  127. atomic_t b_lru_ref; /* lru reclaim ref count */
  128. xfs_buf_flags_t b_flags; /* status flags */
  129. struct semaphore b_sema; /* semaphore for lockables */
  130. /*
  131. * concurrent access to b_lru and b_lru_flags are protected by
  132. * bt_lru_lock and not by b_sema
  133. */
  134. struct list_head b_lru; /* lru list */
  135. spinlock_t b_lock; /* internal state lock */
  136. unsigned int b_state; /* internal state flags */
  137. int b_io_error; /* internal IO error state */
  138. wait_queue_head_t b_waiters; /* unpin waiters */
  139. struct list_head b_list;
  140. struct xfs_perag *b_pag; /* contains rbtree root */
  141. xfs_buftarg_t *b_target; /* buffer target (device) */
  142. void *b_addr; /* virtual address of buffer */
  143. struct work_struct b_ioend_work;
  144. struct workqueue_struct *b_ioend_wq; /* I/O completion wq */
  145. xfs_buf_iodone_t b_iodone; /* I/O completion function */
  146. struct completion b_iowait; /* queue for I/O waiters */
  147. void *b_log_item;
  148. struct list_head b_li_list; /* Log items list head */
  149. struct xfs_trans *b_transp;
  150. struct page **b_pages; /* array of page pointers */
  151. struct page *b_page_array[XB_PAGES]; /* inline pages */
  152. struct xfs_buf_map *b_maps; /* compound buffer map */
  153. struct xfs_buf_map __b_map; /* inline compound buffer map */
  154. int b_map_count;
  155. int b_io_length; /* IO size in BBs */
  156. atomic_t b_pin_count; /* pin count */
  157. atomic_t b_io_remaining; /* #outstanding I/O requests */
  158. unsigned int b_page_count; /* size of page array */
  159. unsigned int b_offset; /* page offset in first page */
  160. int b_error; /* error code on I/O */
  161. /*
  162. * async write failure retry count. Initialised to zero on the first
  163. * failure, then when it exceeds the maximum configured without a
  164. * success the write is considered to be failed permanently and the
  165. * iodone handler will take appropriate action.
  166. *
  167. * For retry timeouts, we record the jiffie of the first failure. This
  168. * means that we can change the retry timeout for buffers already under
  169. * I/O and thus avoid getting stuck in a retry loop with a long timeout.
  170. *
  171. * last_error is used to ensure that we are getting repeated errors, not
  172. * different errors. e.g. a block device might change ENOSPC to EIO when
  173. * a failure timeout occurs, so we want to re-initialise the error
  174. * retry behaviour appropriately when that happens.
  175. */
  176. int b_retries;
  177. unsigned long b_first_retry_time; /* in jiffies */
  178. int b_last_error;
  179. const struct xfs_buf_ops *b_ops;
  180. } xfs_buf_t;
  181. /* Finding and Reading Buffers */
  182. struct xfs_buf *xfs_buf_incore(struct xfs_buftarg *target,
  183. xfs_daddr_t blkno, size_t numblks,
  184. xfs_buf_flags_t flags);
  185. struct xfs_buf *_xfs_buf_alloc(struct xfs_buftarg *target,
  186. struct xfs_buf_map *map, int nmaps,
  187. xfs_buf_flags_t flags);
  188. static inline struct xfs_buf *
  189. xfs_buf_alloc(
  190. struct xfs_buftarg *target,
  191. xfs_daddr_t blkno,
  192. size_t numblks,
  193. xfs_buf_flags_t flags)
  194. {
  195. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  196. return _xfs_buf_alloc(target, &map, 1, flags);
  197. }
  198. struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
  199. struct xfs_buf_map *map, int nmaps,
  200. xfs_buf_flags_t flags);
  201. struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
  202. struct xfs_buf_map *map, int nmaps,
  203. xfs_buf_flags_t flags,
  204. const struct xfs_buf_ops *ops);
  205. void xfs_buf_readahead_map(struct xfs_buftarg *target,
  206. struct xfs_buf_map *map, int nmaps,
  207. const struct xfs_buf_ops *ops);
  208. static inline struct xfs_buf *
  209. xfs_buf_get(
  210. struct xfs_buftarg *target,
  211. xfs_daddr_t blkno,
  212. size_t numblks,
  213. xfs_buf_flags_t flags)
  214. {
  215. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  216. return xfs_buf_get_map(target, &map, 1, flags);
  217. }
  218. static inline struct xfs_buf *
  219. xfs_buf_read(
  220. struct xfs_buftarg *target,
  221. xfs_daddr_t blkno,
  222. size_t numblks,
  223. xfs_buf_flags_t flags,
  224. const struct xfs_buf_ops *ops)
  225. {
  226. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  227. return xfs_buf_read_map(target, &map, 1, flags, ops);
  228. }
  229. static inline void
  230. xfs_buf_readahead(
  231. struct xfs_buftarg *target,
  232. xfs_daddr_t blkno,
  233. size_t numblks,
  234. const struct xfs_buf_ops *ops)
  235. {
  236. DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
  237. return xfs_buf_readahead_map(target, &map, 1, ops);
  238. }
  239. void xfs_buf_set_empty(struct xfs_buf *bp, size_t numblks);
  240. int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length);
  241. struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
  242. int flags);
  243. int xfs_buf_read_uncached(struct xfs_buftarg *target, xfs_daddr_t daddr,
  244. size_t numblks, int flags, struct xfs_buf **bpp,
  245. const struct xfs_buf_ops *ops);
  246. void xfs_buf_hold(struct xfs_buf *bp);
  247. /* Releasing Buffers */
  248. extern void xfs_buf_free(xfs_buf_t *);
  249. extern void xfs_buf_rele(xfs_buf_t *);
  250. /* Locking and Unlocking Buffers */
  251. extern int xfs_buf_trylock(xfs_buf_t *);
  252. extern void xfs_buf_lock(xfs_buf_t *);
  253. extern void xfs_buf_unlock(xfs_buf_t *);
  254. #define xfs_buf_islocked(bp) \
  255. ((bp)->b_sema.count <= 0)
  256. /* Buffer Read and Write Routines */
  257. extern int xfs_bwrite(struct xfs_buf *bp);
  258. extern void xfs_buf_ioend(struct xfs_buf *bp);
  259. extern void __xfs_buf_ioerror(struct xfs_buf *bp, int error,
  260. xfs_failaddr_t failaddr);
  261. #define xfs_buf_ioerror(bp, err) __xfs_buf_ioerror((bp), (err), __this_address)
  262. extern void xfs_buf_ioerror_alert(struct xfs_buf *, const char *func);
  263. extern int __xfs_buf_submit(struct xfs_buf *bp, bool);
  264. static inline int xfs_buf_submit(struct xfs_buf *bp)
  265. {
  266. bool wait = bp->b_flags & XBF_ASYNC ? false : true;
  267. return __xfs_buf_submit(bp, wait);
  268. }
  269. extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *,
  270. xfs_buf_rw_t);
  271. #define xfs_buf_zero(bp, off, len) \
  272. xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO)
  273. /* Buffer Utility Routines */
  274. extern void *xfs_buf_offset(struct xfs_buf *, size_t);
  275. extern void xfs_buf_stale(struct xfs_buf *bp);
  276. /* Delayed Write Buffer Routines */
  277. extern void xfs_buf_delwri_cancel(struct list_head *);
  278. extern bool xfs_buf_delwri_queue(struct xfs_buf *, struct list_head *);
  279. extern int xfs_buf_delwri_submit(struct list_head *);
  280. extern int xfs_buf_delwri_submit_nowait(struct list_head *);
  281. extern int xfs_buf_delwri_pushbuf(struct xfs_buf *, struct list_head *);
  282. /* Buffer Daemon Setup Routines */
  283. extern int xfs_buf_init(void);
  284. extern void xfs_buf_terminate(void);
  285. /*
  286. * These macros use the IO block map rather than b_bn. b_bn is now really
  287. * just for the buffer cache index for cached buffers. As IO does not use b_bn
  288. * anymore, uncached buffers do not use b_bn at all and hence must modify the IO
  289. * map directly. Uncached buffers are not allowed to be discontiguous, so this
  290. * is safe to do.
  291. *
  292. * In future, uncached buffers will pass the block number directly to the io
  293. * request function and hence these macros will go away at that point.
  294. */
  295. #define XFS_BUF_ADDR(bp) ((bp)->b_maps[0].bm_bn)
  296. #define XFS_BUF_SET_ADDR(bp, bno) ((bp)->b_maps[0].bm_bn = (xfs_daddr_t)(bno))
  297. void xfs_buf_set_ref(struct xfs_buf *bp, int lru_ref);
  298. /*
  299. * If the buffer is already on the LRU, do nothing. Otherwise set the buffer
  300. * up with a reference count of 0 so it will be tossed from the cache when
  301. * released.
  302. */
  303. static inline void xfs_buf_oneshot(struct xfs_buf *bp)
  304. {
  305. if (!list_empty(&bp->b_lru) || atomic_read(&bp->b_lru_ref) > 1)
  306. return;
  307. atomic_set(&bp->b_lru_ref, 0);
  308. }
  309. static inline int xfs_buf_ispinned(struct xfs_buf *bp)
  310. {
  311. return atomic_read(&bp->b_pin_count);
  312. }
  313. static inline void xfs_buf_relse(xfs_buf_t *bp)
  314. {
  315. xfs_buf_unlock(bp);
  316. xfs_buf_rele(bp);
  317. }
  318. static inline int
  319. xfs_buf_verify_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
  320. {
  321. return xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length),
  322. cksum_offset);
  323. }
  324. static inline void
  325. xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
  326. {
  327. xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length),
  328. cksum_offset);
  329. }
  330. /*
  331. * Handling of buftargs.
  332. */
  333. extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *,
  334. struct block_device *, struct dax_device *);
  335. extern void xfs_free_buftarg(struct xfs_buftarg *);
  336. extern void xfs_wait_buftarg(xfs_buftarg_t *);
  337. extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int);
  338. #define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev)
  339. #define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev)
  340. int xfs_buf_reverify(struct xfs_buf *bp, const struct xfs_buf_ops *ops);
  341. bool xfs_verify_magic(struct xfs_buf *bp, __be32 dmagic);
  342. bool xfs_verify_magic16(struct xfs_buf *bp, __be16 dmagic);
  343. #endif /* __XFS_BUF_H__ */