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 21 #include <linux/list.h> 22 #include <linux/types.h> 23 #include <linux/spinlock.h> 24 #include <linux/mm.h> 25 #include <linux/fs.h> 26 #include <linux/buffer_head.h> 27 #include <linux/uio.h> 28 #include <linux/list_lru.h> 29 30 /* 31 * Base types 32 */ 33 34 #define XFS_BUF_DADDR_NULL ((xfs_daddr_t) (-1LL)) 35 36 typedef enum { 37 XBRW_READ = 1, /* transfer into target memory */ 38 XBRW_WRITE = 2, /* transfer from target memory */ 39 XBRW_ZERO = 3, /* Zero target memory */ 40 } xfs_buf_rw_t; 41 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 */ 49 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 */ 54 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 */ 58 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 */ 64 65 typedef unsigned int xfs_buf_flags_t; 66 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" }, \ 76 { XBF_FUA, "FUA" }, \ 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" } 84 85 86 /* 87 * Internal state flags. 88 */ 89 #define XFS_BSTATE_DISPOSE (1 << 0) /* buffer being discarded */ 90 91 /* 92 * The xfs_buftarg contains 2 notions of "sector size" - 93 * 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 97 * 98 * The first is specified at mkfs time, and is stored on-disk in the 99 * superblock's sb_sectsize. 100 * 101 * The latter is derived from the underlying device, and controls direct IO 102 * alignment constraints. 103 */ 104 typedef struct xfs_buftarg { 105 dev_t bt_dev; 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; 113 114 /* LRU control structures */ 115 struct shrinker bt_shrinker; 116 struct list_lru bt_lru; 117 } xfs_buftarg_t; 118 119 struct xfs_buf; 120 typedef void (*xfs_buf_iodone_t)(struct xfs_buf *); 121 122 123 #define XB_PAGES 2 124 125 struct xfs_buf_map { 126 xfs_daddr_t bm_bn; /* block number for I/O */ 127 int bm_len; /* size of I/O */ 128 }; 129 130 #define DEFINE_SINGLE_BUF_MAP(map, blkno, numblk) \ 131 struct xfs_buf_map (map) = { .bm_bn = (blkno), .bm_len = (numblk) }; 132 133 struct xfs_buf_ops { 134 void (*verify_read)(struct xfs_buf *); 135 void (*verify_write)(struct xfs_buf *); 136 }; 137 138 typedef struct xfs_buf { 139 /* 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. 145 */ 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 */ 153 154 /* 155 * concurrent access to b_lru and b_lru_flags are protected by 156 * bt_lru_lock and not by b_sema 157 */ 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_iodone_work; 168 xfs_buf_iodone_t b_iodone; /* I/O completion function */ 169 struct completion b_iowait; /* queue for I/O waiters */ 170 void *b_fspriv; 171 struct xfs_trans *b_transp; 172 struct page **b_pages; /* array of page pointers */ 173 struct page *b_page_array[XB_PAGES]; /* inline pages */ 174 struct xfs_buf_map *b_maps; /* compound buffer map */ 175 struct xfs_buf_map __b_map; /* inline compound buffer map */ 176 int b_map_count; 177 int b_io_length; /* IO size in BBs */ 178 atomic_t b_pin_count; /* pin count */ 179 atomic_t b_io_remaining; /* #outstanding I/O requests */ 180 unsigned int b_page_count; /* size of page array */ 181 unsigned int b_offset; /* page offset in first page */ 182 int b_error; /* error code on I/O */ 183 const struct xfs_buf_ops *b_ops; 184 185 #ifdef XFS_BUF_LOCK_TRACKING 186 int b_last_holder; 187 #endif 188 } xfs_buf_t; 189 190 /* 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); 194 195 static inline struct xfs_buf * 196 xfs_incore( 197 struct xfs_buftarg *target, 198 xfs_daddr_t blkno, 199 size_t numblks, 200 xfs_buf_flags_t flags) 201 { 202 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 203 return _xfs_buf_find(target, &map, 1, flags, NULL); 204 } 205 206 struct xfs_buf *_xfs_buf_alloc(struct xfs_buftarg *target, 207 struct xfs_buf_map *map, int nmaps, 208 xfs_buf_flags_t flags); 209 210 static inline struct xfs_buf * 211 xfs_buf_alloc( 212 struct xfs_buftarg *target, 213 xfs_daddr_t blkno, 214 size_t numblks, 215 xfs_buf_flags_t flags) 216 { 217 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 218 return _xfs_buf_alloc(target, &map, 1, flags); 219 } 220 221 struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target, 222 struct xfs_buf_map *map, int nmaps, 223 xfs_buf_flags_t flags); 224 struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target, 225 struct xfs_buf_map *map, int nmaps, 226 xfs_buf_flags_t flags, 227 const struct xfs_buf_ops *ops); 228 void xfs_buf_readahead_map(struct xfs_buftarg *target, 229 struct xfs_buf_map *map, int nmaps, 230 const struct xfs_buf_ops *ops); 231 232 static inline struct xfs_buf * 233 xfs_buf_get( 234 struct xfs_buftarg *target, 235 xfs_daddr_t blkno, 236 size_t numblks, 237 xfs_buf_flags_t flags) 238 { 239 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 240 return xfs_buf_get_map(target, &map, 1, flags); 241 } 242 243 static inline struct xfs_buf * 244 xfs_buf_read( 245 struct xfs_buftarg *target, 246 xfs_daddr_t blkno, 247 size_t numblks, 248 xfs_buf_flags_t flags, 249 const struct xfs_buf_ops *ops) 250 { 251 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 252 return xfs_buf_read_map(target, &map, 1, flags, ops); 253 } 254 255 static inline void 256 xfs_buf_readahead( 257 struct xfs_buftarg *target, 258 xfs_daddr_t blkno, 259 size_t numblks, 260 const struct xfs_buf_ops *ops) 261 { 262 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 263 return xfs_buf_readahead_map(target, &map, 1, ops); 264 } 265 266 struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks); 267 void xfs_buf_set_empty(struct xfs_buf *bp, size_t numblks); 268 int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length); 269 270 struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks, 271 int flags); 272 int xfs_buf_read_uncached(struct xfs_buftarg *target, xfs_daddr_t daddr, 273 size_t numblks, int flags, struct xfs_buf **bpp, 274 const struct xfs_buf_ops *ops); 275 void xfs_buf_hold(struct xfs_buf *bp); 276 277 /* Releasing Buffers */ 278 extern void xfs_buf_free(xfs_buf_t *); 279 extern void xfs_buf_rele(xfs_buf_t *); 280 281 /* Locking and Unlocking Buffers */ 282 extern int xfs_buf_trylock(xfs_buf_t *); 283 extern void xfs_buf_lock(xfs_buf_t *); 284 extern void xfs_buf_unlock(xfs_buf_t *); 285 #define xfs_buf_islocked(bp) \ 286 ((bp)->b_sema.count <= 0) 287 288 /* Buffer Read and Write Routines */ 289 extern int xfs_bwrite(struct xfs_buf *bp); 290 extern void xfs_buf_ioend(struct xfs_buf *bp); 291 extern void xfs_buf_ioerror(xfs_buf_t *, int); 292 extern void xfs_buf_ioerror_alert(struct xfs_buf *, const char *func); 293 extern void xfs_buf_submit(struct xfs_buf *bp); 294 extern int xfs_buf_submit_wait(struct xfs_buf *bp); 295 extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *, 296 xfs_buf_rw_t); 297 #define xfs_buf_zero(bp, off, len) \ 298 xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO) 299 300 /* Buffer Utility Routines */ 301 extern xfs_caddr_t xfs_buf_offset(xfs_buf_t *, size_t); 302 303 /* Delayed Write Buffer Routines */ 304 extern bool xfs_buf_delwri_queue(struct xfs_buf *, struct list_head *); 305 extern int xfs_buf_delwri_submit(struct list_head *); 306 extern int xfs_buf_delwri_submit_nowait(struct list_head *); 307 308 /* Buffer Daemon Setup Routines */ 309 extern int xfs_buf_init(void); 310 extern void xfs_buf_terminate(void); 311 312 #define XFS_BUF_ZEROFLAGS(bp) \ 313 ((bp)->b_flags &= ~(XBF_READ|XBF_WRITE|XBF_ASYNC| \ 314 XBF_SYNCIO|XBF_FUA|XBF_FLUSH| \ 315 XBF_WRITE_FAIL)) 316 317 void xfs_buf_stale(struct xfs_buf *bp); 318 #define XFS_BUF_UNSTALE(bp) ((bp)->b_flags &= ~XBF_STALE) 319 #define XFS_BUF_ISSTALE(bp) ((bp)->b_flags & XBF_STALE) 320 321 #define XFS_BUF_DONE(bp) ((bp)->b_flags |= XBF_DONE) 322 #define XFS_BUF_UNDONE(bp) ((bp)->b_flags &= ~XBF_DONE) 323 #define XFS_BUF_ISDONE(bp) ((bp)->b_flags & XBF_DONE) 324 325 #define XFS_BUF_ASYNC(bp) ((bp)->b_flags |= XBF_ASYNC) 326 #define XFS_BUF_UNASYNC(bp) ((bp)->b_flags &= ~XBF_ASYNC) 327 #define XFS_BUF_ISASYNC(bp) ((bp)->b_flags & XBF_ASYNC) 328 329 #define XFS_BUF_READ(bp) ((bp)->b_flags |= XBF_READ) 330 #define XFS_BUF_UNREAD(bp) ((bp)->b_flags &= ~XBF_READ) 331 #define XFS_BUF_ISREAD(bp) ((bp)->b_flags & XBF_READ) 332 333 #define XFS_BUF_WRITE(bp) ((bp)->b_flags |= XBF_WRITE) 334 #define XFS_BUF_UNWRITE(bp) ((bp)->b_flags &= ~XBF_WRITE) 335 #define XFS_BUF_ISWRITE(bp) ((bp)->b_flags & XBF_WRITE) 336 337 /* 338 * These macros use the IO block map rather than b_bn. b_bn is now really 339 * just for the buffer cache index for cached buffers. As IO does not use b_bn 340 * anymore, uncached buffers do not use b_bn at all and hence must modify the IO 341 * map directly. Uncached buffers are not allowed to be discontiguous, so this 342 * is safe to do. 343 * 344 * In future, uncached buffers will pass the block number directly to the io 345 * request function and hence these macros will go away at that point. 346 */ 347 #define XFS_BUF_ADDR(bp) ((bp)->b_maps[0].bm_bn) 348 #define XFS_BUF_SET_ADDR(bp, bno) ((bp)->b_maps[0].bm_bn = (xfs_daddr_t)(bno)) 349 350 static inline void xfs_buf_set_ref(struct xfs_buf *bp, int lru_ref) 351 { 352 atomic_set(&bp->b_lru_ref, lru_ref); 353 } 354 355 static inline int xfs_buf_ispinned(struct xfs_buf *bp) 356 { 357 return atomic_read(&bp->b_pin_count); 358 } 359 360 static inline void xfs_buf_relse(xfs_buf_t *bp) 361 { 362 xfs_buf_unlock(bp); 363 xfs_buf_rele(bp); 364 } 365 366 static inline int 367 xfs_buf_verify_cksum(struct xfs_buf *bp, unsigned long cksum_offset) 368 { 369 return xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length), 370 cksum_offset); 371 } 372 373 static inline void 374 xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset) 375 { 376 xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length), 377 cksum_offset); 378 } 379 380 /* 381 * Handling of buftargs. 382 */ 383 extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *, 384 struct block_device *); 385 extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *); 386 extern void xfs_wait_buftarg(xfs_buftarg_t *); 387 extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int); 388 389 #define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev) 390 #define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev) 391 392 #endif /* __XFS_BUF_H__ */ 393