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 36typedef 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 65typedef 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 */ 104typedef 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 119struct xfs_buf; 120typedef void (*xfs_buf_iodone_t)(struct xfs_buf *); 121 122 123#define XB_PAGES 2 124 125struct 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 133struct xfs_buf_ops { 134 void (*verify_read)(struct xfs_buf *); 135 void (*verify_write)(struct xfs_buf *); 136}; 137 138typedef 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_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 */ 171 void *b_fspriv; 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 */ 177 int b_map_count; 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; 185 186#ifdef XFS_BUF_LOCK_TRACKING 187 int b_last_holder; 188#endif 189} xfs_buf_t; 190 191/* Finding and Reading Buffers */ 192struct 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); 195 196static inline struct xfs_buf * 197xfs_incore( 198 struct xfs_buftarg *target, 199 xfs_daddr_t blkno, 200 size_t numblks, 201 xfs_buf_flags_t flags) 202{ 203 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 204 return _xfs_buf_find(target, &map, 1, flags, NULL); 205} 206 207struct xfs_buf *_xfs_buf_alloc(struct xfs_buftarg *target, 208 struct xfs_buf_map *map, int nmaps, 209 xfs_buf_flags_t flags); 210 211static inline struct xfs_buf * 212xfs_buf_alloc( 213 struct xfs_buftarg *target, 214 xfs_daddr_t blkno, 215 size_t numblks, 216 xfs_buf_flags_t flags) 217{ 218 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 219 return _xfs_buf_alloc(target, &map, 1, flags); 220} 221 222struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target, 223 struct xfs_buf_map *map, int nmaps, 224 xfs_buf_flags_t flags); 225struct 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); 229void xfs_buf_readahead_map(struct xfs_buftarg *target, 230 struct xfs_buf_map *map, int nmaps, 231 const struct xfs_buf_ops *ops); 232 233static inline struct xfs_buf * 234xfs_buf_get( 235 struct xfs_buftarg *target, 236 xfs_daddr_t blkno, 237 size_t numblks, 238 xfs_buf_flags_t flags) 239{ 240 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 241 return xfs_buf_get_map(target, &map, 1, flags); 242} 243 244static inline struct xfs_buf * 245xfs_buf_read( 246 struct xfs_buftarg *target, 247 xfs_daddr_t blkno, 248 size_t numblks, 249 xfs_buf_flags_t flags, 250 const struct xfs_buf_ops *ops) 251{ 252 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 253 return xfs_buf_read_map(target, &map, 1, flags, ops); 254} 255 256static inline void 257xfs_buf_readahead( 258 struct xfs_buftarg *target, 259 xfs_daddr_t blkno, 260 size_t numblks, 261 const struct xfs_buf_ops *ops) 262{ 263 DEFINE_SINGLE_BUF_MAP(map, blkno, numblks); 264 return xfs_buf_readahead_map(target, &map, 1, ops); 265} 266 267struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks); 268void xfs_buf_set_empty(struct xfs_buf *bp, size_t numblks); 269int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length); 270 271struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks, 272 int flags); 273int 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); 276void xfs_buf_hold(struct xfs_buf *bp); 277 278/* Releasing Buffers */ 279extern void xfs_buf_free(xfs_buf_t *); 280extern void xfs_buf_rele(xfs_buf_t *); 281 282/* Locking and Unlocking Buffers */ 283extern int xfs_buf_trylock(xfs_buf_t *); 284extern void xfs_buf_lock(xfs_buf_t *); 285extern void xfs_buf_unlock(xfs_buf_t *); 286#define xfs_buf_islocked(bp) \ 287 ((bp)->b_sema.count <= 0) 288 289/* Buffer Read and Write Routines */ 290extern int xfs_bwrite(struct xfs_buf *bp); 291extern void xfs_buf_ioend(struct xfs_buf *bp); 292extern void xfs_buf_ioerror(xfs_buf_t *, int); 293extern void xfs_buf_ioerror_alert(struct xfs_buf *, const char *func); 294extern void xfs_buf_submit(struct xfs_buf *bp); 295extern int xfs_buf_submit_wait(struct xfs_buf *bp); 296extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *, 297 xfs_buf_rw_t); 298#define xfs_buf_zero(bp, off, len) \ 299 xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO) 300 301/* Buffer Utility Routines */ 302extern xfs_caddr_t xfs_buf_offset(xfs_buf_t *, size_t); 303 304/* Delayed Write Buffer Routines */ 305extern bool xfs_buf_delwri_queue(struct xfs_buf *, struct list_head *); 306extern int xfs_buf_delwri_submit(struct list_head *); 307extern int xfs_buf_delwri_submit_nowait(struct list_head *); 308 309/* Buffer Daemon Setup Routines */ 310extern int xfs_buf_init(void); 311extern void xfs_buf_terminate(void); 312 313#define XFS_BUF_ZEROFLAGS(bp) \ 314 ((bp)->b_flags &= ~(XBF_READ|XBF_WRITE|XBF_ASYNC| \ 315 XBF_SYNCIO|XBF_FUA|XBF_FLUSH| \ 316 XBF_WRITE_FAIL)) 317 318void 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) 321 322#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) 325 326#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) 329 330#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) 333 334#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) 337 338/* 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 343 * is safe to do. 344 * 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. 347 */ 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)) 350 351static inline void xfs_buf_set_ref(struct xfs_buf *bp, int lru_ref) 352{ 353 atomic_set(&bp->b_lru_ref, lru_ref); 354} 355 356static inline int xfs_buf_ispinned(struct xfs_buf *bp) 357{ 358 return atomic_read(&bp->b_pin_count); 359} 360 361static inline void xfs_buf_relse(xfs_buf_t *bp) 362{ 363 xfs_buf_unlock(bp); 364 xfs_buf_rele(bp); 365} 366 367static inline int 368xfs_buf_verify_cksum(struct xfs_buf *bp, unsigned long cksum_offset) 369{ 370 return xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length), 371 cksum_offset); 372} 373 374static inline void 375xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset) 376{ 377 xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length), 378 cksum_offset); 379} 380 381/* 382 * Handling of buftargs. 383 */ 384extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *, 385 struct block_device *); 386extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *); 387extern void xfs_wait_buftarg(xfs_buftarg_t *); 388extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int); 389 390#define xfs_getsize_buftarg(buftarg) block_size((buftarg)->bt_bdev) 391#define xfs_readonly_buftarg(buftarg) bdev_read_only((buftarg)->bt_bdev) 392 393#endif /* __XFS_BUF_H__ */ 394