2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
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.
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.
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
21 #include <linux/list.h>
22 #include <linux/types.h>
23 #include <linux/spinlock.h>
26 #include <linux/buffer_head.h>
27 #include <linux/uio.h>
28 #include <linux/list_lru.h>
34 #define XFS_BUF_DADDR_NULL ((xfs_daddr_t) (-1LL))
37 XBRW_READ = 1, /* transfer into target memory */
38 XBRW_WRITE = 2, /* transfer from target memory */
39 XBRW_ZERO = 3, /* Zero target memory */
42 #define XBF_READ (1 << 0) /* buffer intended for reading from device */
43 #define XBF_WRITE (1 << 1) /* buffer intended for writing to device */
44 #define XBF_READ_AHEAD (1 << 2) /* asynchronous read-ahead */
45 #define XBF_ASYNC (1 << 4) /* initiator will not wait for completion */
46 #define XBF_DONE (1 << 5) /* all pages in the buffer uptodate */
47 #define XBF_STALE (1 << 6) /* buffer has been staled, do not find it */
48 #define XBF_WRITE_FAIL (1 << 24)/* async writes have failed on this buffer */
50 /* I/O hints for the BIO layer */
51 #define XBF_SYNCIO (1 << 10)/* treat this buffer as synchronous I/O */
52 #define XBF_FUA (1 << 11)/* force cache write through mode */
53 #define XBF_FLUSH (1 << 12)/* flush the disk cache before a write */
55 /* flags used only as arguments to access routines */
56 #define XBF_TRYLOCK (1 << 16)/* lock requested, but do not wait */
57 #define XBF_UNMAPPED (1 << 17)/* do not map the buffer */
59 /* flags used only internally */
60 #define _XBF_PAGES (1 << 20)/* backed by refcounted pages */
61 #define _XBF_KMEM (1 << 21)/* backed by heap memory */
62 #define _XBF_DELWRI_Q (1 << 22)/* buffer on a delwri queue */
63 #define _XBF_COMPOUND (1 << 23)/* compound buffer */
65 typedef unsigned int xfs_buf_flags_t;
67 #define XFS_BUF_FLAGS \
68 { XBF_READ, "READ" }, \
69 { XBF_WRITE, "WRITE" }, \
70 { XBF_READ_AHEAD, "READ_AHEAD" }, \
71 { XBF_ASYNC, "ASYNC" }, \
72 { XBF_DONE, "DONE" }, \
73 { XBF_STALE, "STALE" }, \
74 { XBF_WRITE_FAIL, "WRITE_FAIL" }, \
75 { XBF_SYNCIO, "SYNCIO" }, \
77 { XBF_FLUSH, "FLUSH" }, \
78 { XBF_TRYLOCK, "TRYLOCK" }, /* should never be set */\
79 { XBF_UNMAPPED, "UNMAPPED" }, /* ditto */\
80 { _XBF_PAGES, "PAGES" }, \
81 { _XBF_KMEM, "KMEM" }, \
82 { _XBF_DELWRI_Q, "DELWRI_Q" }, \
83 { _XBF_COMPOUND, "COMPOUND" }
87 * Internal state flags.
89 #define XFS_BSTATE_DISPOSE (1 << 0) /* buffer being discarded */
92 * The xfs_buftarg contains 2 notions of "sector size" -
94 * 1) The metadata sector size, which is the minimum unit and
95 * alignment of IO which will be performed by metadata operations.
96 * 2) The device logical sector size
98 * The first is specified at mkfs time, and is stored on-disk in the
99 * superblock's sb_sectsize.
101 * The latter is derived from the underlying device, and controls direct IO
102 * alignment constraints.
104 typedef struct xfs_buftarg {
106 struct block_device *bt_bdev;
107 struct backing_dev_info *bt_bdi;
108 struct xfs_mount *bt_mount;
109 unsigned int bt_meta_sectorsize;
110 size_t bt_meta_sectormask;
111 size_t bt_logical_sectorsize;
112 size_t bt_logical_sectormask;
114 /* LRU control structures */
115 struct shrinker bt_shrinker;
116 struct list_lru bt_lru;
120 typedef void (*xfs_buf_iodone_t)(struct xfs_buf *);
126 xfs_daddr_t bm_bn; /* block number for I/O */
127 int bm_len; /* size of I/O */
130 #define DEFINE_SINGLE_BUF_MAP(map, blkno, numblk) \
131 struct xfs_buf_map (map) = { .bm_bn = (blkno), .bm_len = (numblk) };
134 void (*verify_read)(struct xfs_buf *);
135 void (*verify_write)(struct xfs_buf *);
138 typedef struct xfs_buf {
140 * first cacheline holds all the fields needed for an uncontended cache
141 * hit to be fully processed. The semaphore straddles the cacheline
142 * boundary, but the counter and lock sits on the first cacheline,
143 * which is the only bit that is touched if we hit the semaphore
144 * fast-path on locking.
146 struct rb_node b_rbnode; /* rbtree node */
147 xfs_daddr_t b_bn; /* block number of buffer */
148 int b_length; /* size of buffer in BBs */
149 atomic_t b_hold; /* reference count */
150 atomic_t b_lru_ref; /* lru reclaim ref count */
151 xfs_buf_flags_t b_flags; /* status flags */
152 struct semaphore b_sema; /* semaphore for lockables */
155 * concurrent access to b_lru and b_lru_flags are protected by
156 * bt_lru_lock and not by b_sema
158 struct list_head b_lru; /* lru list */
159 spinlock_t b_lock; /* internal state lock */
160 unsigned int b_state; /* internal state flags */
161 int b_io_error; /* internal IO error state */
162 wait_queue_head_t b_waiters; /* unpin waiters */
163 struct list_head b_list;
164 struct xfs_perag *b_pag; /* contains rbtree root */
165 xfs_buftarg_t *b_target; /* buffer target (device) */
166 void *b_addr; /* virtual address of buffer */
167 struct work_struct b_ioend_work;
168 struct workqueue_struct *b_ioend_wq; /* I/O completion wq */
169 xfs_buf_iodone_t b_iodone; /* I/O completion function */
170 struct completion b_iowait; /* queue for I/O waiters */
172 struct xfs_trans *b_transp;
173 struct page **b_pages; /* array of page pointers */
174 struct page *b_page_array[XB_PAGES]; /* inline pages */
175 struct xfs_buf_map *b_maps; /* compound buffer map */
176 struct xfs_buf_map __b_map; /* inline compound buffer map */
178 int b_io_length; /* IO size in BBs */
179 atomic_t b_pin_count; /* pin count */
180 atomic_t b_io_remaining; /* #outstanding I/O requests */
181 unsigned int b_page_count; /* size of page array */
182 unsigned int b_offset; /* page offset in first page */
183 int b_error; /* error code on I/O */
184 const struct xfs_buf_ops *b_ops;
186 #ifdef XFS_BUF_LOCK_TRACKING
191 /* Finding and Reading Buffers */
192 struct xfs_buf *_xfs_buf_find(struct xfs_buftarg *target,
193 struct xfs_buf_map *map, int nmaps,
194 xfs_buf_flags_t flags, struct xfs_buf *new_bp);
196 static inline struct xfs_buf *
198 struct xfs_buftarg *target,
201 xfs_buf_flags_t flags)
203 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
204 return _xfs_buf_find(target, &map, 1, flags, NULL);
207 struct xfs_buf *_xfs_buf_alloc(struct xfs_buftarg *target,
208 struct xfs_buf_map *map, int nmaps,
209 xfs_buf_flags_t flags);
211 static inline struct xfs_buf *
213 struct xfs_buftarg *target,
216 xfs_buf_flags_t flags)
218 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
219 return _xfs_buf_alloc(target, &map, 1, flags);
222 struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
223 struct xfs_buf_map *map, int nmaps,
224 xfs_buf_flags_t flags);
225 struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
226 struct xfs_buf_map *map, int nmaps,
227 xfs_buf_flags_t flags,
228 const struct xfs_buf_ops *ops);
229 void xfs_buf_readahead_map(struct xfs_buftarg *target,
230 struct xfs_buf_map *map, int nmaps,
231 const struct xfs_buf_ops *ops);
233 static inline struct xfs_buf *
235 struct xfs_buftarg *target,
238 xfs_buf_flags_t flags)
240 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
241 return xfs_buf_get_map(target, &map, 1, flags);
244 static inline struct xfs_buf *
246 struct xfs_buftarg *target,
249 xfs_buf_flags_t flags,
250 const struct xfs_buf_ops *ops)
252 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
253 return xfs_buf_read_map(target, &map, 1, flags, ops);
258 struct xfs_buftarg *target,
261 const struct xfs_buf_ops *ops)
263 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
264 return xfs_buf_readahead_map(target, &map, 1, ops);
267 struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks);
268 void xfs_buf_set_empty(struct xfs_buf *bp, size_t numblks);
269 int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length);
271 struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
273 int xfs_buf_read_uncached(struct xfs_buftarg *target, xfs_daddr_t daddr,
274 size_t numblks, int flags, struct xfs_buf **bpp,
275 const struct xfs_buf_ops *ops);
276 void xfs_buf_hold(struct xfs_buf *bp);
278 /* Releasing Buffers */
279 extern void xfs_buf_free(xfs_buf_t *);
280 extern void xfs_buf_rele(xfs_buf_t *);
282 /* Locking and Unlocking Buffers */
283 extern int xfs_buf_trylock(xfs_buf_t *);
284 extern void xfs_buf_lock(xfs_buf_t *);
285 extern void xfs_buf_unlock(xfs_buf_t *);
286 #define xfs_buf_islocked(bp) \
287 ((bp)->b_sema.count <= 0)
289 /* Buffer Read and Write Routines */
290 extern int xfs_bwrite(struct xfs_buf *bp);
291 extern void xfs_buf_ioend(struct xfs_buf *bp);
292 extern void xfs_buf_ioerror(xfs_buf_t *, int);
293 extern void xfs_buf_ioerror_alert(struct xfs_buf *, const char *func);
294 extern void xfs_buf_submit(struct xfs_buf *bp);
295 extern int xfs_buf_submit_wait(struct xfs_buf *bp);
296 extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *,
298 #define xfs_buf_zero(bp, off, len) \
299 xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO)
301 /* Buffer Utility Routines */
302 extern xfs_caddr_t xfs_buf_offset(xfs_buf_t *, size_t);
304 /* Delayed Write Buffer Routines */
305 extern bool xfs_buf_delwri_queue(struct xfs_buf *, struct list_head *);
306 extern int xfs_buf_delwri_submit(struct list_head *);
307 extern int xfs_buf_delwri_submit_nowait(struct list_head *);
309 /* Buffer Daemon Setup Routines */
310 extern int xfs_buf_init(void);
311 extern void xfs_buf_terminate(void);
313 #define XFS_BUF_ZEROFLAGS(bp) \
314 ((bp)->b_flags &= ~(XBF_READ|XBF_WRITE|XBF_ASYNC| \
315 XBF_SYNCIO|XBF_FUA|XBF_FLUSH| \
318 void xfs_buf_stale(struct xfs_buf *bp);
319 #define XFS_BUF_UNSTALE(bp) ((bp)->b_flags &= ~XBF_STALE)
320 #define XFS_BUF_ISSTALE(bp) ((bp)->b_flags & XBF_STALE)
322 #define XFS_BUF_DONE(bp) ((bp)->b_flags |= XBF_DONE)
323 #define XFS_BUF_UNDONE(bp) ((bp)->b_flags &= ~XBF_DONE)
324 #define XFS_BUF_ISDONE(bp) ((bp)->b_flags & XBF_DONE)
326 #define XFS_BUF_ASYNC(bp) ((bp)->b_flags |= XBF_ASYNC)
327 #define XFS_BUF_UNASYNC(bp) ((bp)->b_flags &= ~XBF_ASYNC)
328 #define XFS_BUF_ISASYNC(bp) ((bp)->b_flags & XBF_ASYNC)
330 #define XFS_BUF_READ(bp) ((bp)->b_flags |= XBF_READ)
331 #define XFS_BUF_UNREAD(bp) ((bp)->b_flags &= ~XBF_READ)
332 #define XFS_BUF_ISREAD(bp) ((bp)->b_flags & XBF_READ)
334 #define XFS_BUF_WRITE(bp) ((bp)->b_flags |= XBF_WRITE)
335 #define XFS_BUF_UNWRITE(bp) ((bp)->b_flags &= ~XBF_WRITE)
336 #define XFS_BUF_ISWRITE(bp) ((bp)->b_flags & XBF_WRITE)
339 * These macros use the IO block map rather than b_bn. b_bn is now really
340 * just for the buffer cache index for cached buffers. As IO does not use b_bn
341 * anymore, uncached buffers do not use b_bn at all and hence must modify the IO
342 * map directly. Uncached buffers are not allowed to be discontiguous, so this
345 * In future, uncached buffers will pass the block number directly to the io
346 * request function and hence these macros will go away at that point.
348 #define XFS_BUF_ADDR(bp) ((bp)->b_maps[0].bm_bn)
349 #define XFS_BUF_SET_ADDR(bp, bno) ((bp)->b_maps[0].bm_bn = (xfs_daddr_t)(bno))
351 static inline void xfs_buf_set_ref(struct xfs_buf *bp, int lru_ref)
353 atomic_set(&bp->b_lru_ref, lru_ref);
356 static inline int xfs_buf_ispinned(struct xfs_buf *bp)
358 return atomic_read(&bp->b_pin_count);
361 static inline void xfs_buf_relse(xfs_buf_t *bp)
368 xfs_buf_verify_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
370 return xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length),
375 xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
377 xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length),
382 * Handling of buftargs.
384 extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *,
385 struct block_device *);
386 extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *);
387 extern void xfs_wait_buftarg(xfs_buftarg_t *);
388 extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int);
390 #define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev)
391 #define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev)
393 #endif /* __XFS_BUF_H__ */