1 /* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)buf.h 8.9 (Berkeley) 3/30/95 39 * $FreeBSD: src/sys/sys/buf.h,v 1.88.2.10 2003/01/25 19:02:23 dillon Exp $ 40 * $DragonFly: src/sys/sys/buf.h,v 1.38 2006/07/28 02:17:41 dillon Exp $ 41 */ 42 43 #ifndef _SYS_BUF_H_ 44 #define _SYS_BUF_H_ 45 46 #if defined(_KERNEL) || defined(_KERNEL_STRUCTURES) 47 48 #ifndef _SYS_QUEUE_H_ 49 #include <sys/queue.h> 50 #endif 51 #ifndef _SYS_LOCK_H_ 52 #include <sys/lock.h> 53 #endif 54 #ifndef _SYS_DEVICE_H_ 55 #include <sys/device.h> 56 #endif 57 58 #ifndef _SYS_XIO_H_ 59 #include <sys/xio.h> 60 #endif 61 #ifndef _SYS_TREE_H_ 62 #include <sys/tree.h> 63 #endif 64 #ifndef _SYS_BIO_H_ 65 #include <sys/bio.h> 66 #endif 67 #ifndef _SYS_SPINLOCK_H_ 68 #include <sys/spinlock.h> 69 #endif 70 71 struct buf; 72 struct bio; 73 struct mount; 74 struct vnode; 75 struct xio; 76 77 #define NBUF_BIO 4 78 79 struct buf_rb_tree; 80 struct buf_rb_hash; 81 RB_PROTOTYPE2(buf_rb_tree, buf, b_rbnode, rb_buf_compare, off_t); 82 RB_PROTOTYPE2(buf_rb_hash, buf, b_rbhash, rb_buf_compare, off_t); 83 84 /* 85 * To avoid including <ufs/ffs/softdep.h> 86 */ 87 LIST_HEAD(workhead, worklist); 88 /* 89 * These are currently used only by the soft dependency code, hence 90 * are stored once in a global variable. If other subsystems wanted 91 * to use these hooks, a pointer to a set of bio_ops could be added 92 * to each buffer. 93 */ 94 extern struct bio_ops { 95 void (*io_start) (struct buf *); 96 void (*io_complete) (struct buf *); 97 void (*io_deallocate) (struct buf *); 98 int (*io_fsync) (struct vnode *); 99 int (*io_sync) (struct mount *); 100 void (*io_movedeps) (struct buf *, struct buf *); 101 int (*io_countdeps) (struct buf *, int); 102 } bioops; 103 104 typedef enum buf_cmd { 105 BUF_CMD_DONE = 0, 106 BUF_CMD_READ, 107 BUF_CMD_WRITE, 108 BUF_CMD_FREEBLKS, 109 BUF_CMD_FORMAT 110 } buf_cmd_t; 111 112 /* 113 * The buffer header describes an I/O operation in the kernel. 114 * 115 * NOTES: 116 * b_bufsize represents the filesystem block size (for this particular 117 * block) and/or the allocation size or original request size. This 118 * field is NOT USED by lower device layers. VNode and device 119 * strategy routines WILL NEVER ACCESS THIS FIELD. 120 * 121 * b_bcount represents the I/O request size. Unless B_NOBCLIP is set, 122 * the device chain is allowed to clip b_bcount to accomodate the device 123 * EOF. Note that this is different from the byte oriented file EOF. 124 * If B_NOBCLIP is set, the device chain is required to generate an 125 * error if it would othrewise have to clip the request. Buffers 126 * obtained via getblk() automatically set B_NOBCLIP. It is important 127 * to note that EOF clipping via b_bcount is different from EOF clipping 128 * via returning a b_actual < b_bcount. B_NOBCLIP only effects block 129 * oriented EOF clipping (b_bcount modifications). 130 * 131 * b_actual represents the number of bytes of I/O that actually occured, 132 * whether an error occured or not. b_actual must be initialized to 0 133 * prior to initiating I/O as the device drivers will assume it to 134 * start at 0. 135 * 136 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned 137 * ranges of dirty data that need to be written to backing store. 138 * The range is typically clipped at b_bcount (not b_bufsize). 139 * 140 * b_bio1 and b_bio2 represent the two primary I/O layers. Additional 141 * I/O layers are allocated out of the object cache and may also exist. 142 * 143 * b_bio1 is the logical layer and contains offset or block number 144 * data for the primary vnode, b_vp. I/O operations are almost 145 * universally initiated from the logical layer, so you will often 146 * see things like: vn_strategy(bp->b_vp, &bp->b_bio1). 147 * 148 * b_bio2 is the first physical layer (typically the slice-relative 149 * layer) and contains the translated offset or block number for 150 * the block device underlying a filesystem. Filesystems such as UFS 151 * will maintain cached translations and you may see them initiate 152 * a 'physical' I/O using vn_strategy(devvp, &bp->b_bio2). BUT, 153 * remember that the layering is relative to bp->b_vp, so the 154 * device-relative block numbers for buffer cache operations that occur 155 * directly on a block device will be in the first BIO layer. 156 * 157 * NOTE!!! Only the BIO subsystem accesses b_bio1 and b_bio2 directly. 158 * ALL STRATEGY LAYERS FOR BOTH VNODES AND DEVICES ONLY ACCESS THE BIO 159 * PASSED TO THEM, AND WILL PUSH ANOTHER BIO LAYER IF FORWARDING THE 160 * I/O DEEPER. In particular, a vn_strategy() or dev_dstrategy() 161 * call should not ever access buf->b_vp as this vnode may be totally 162 * unrelated to the vnode/device whos strategy routine was called. 163 */ 164 struct buf { 165 RB_ENTRY(buf) b_rbnode; /* RB node in vnode clean/dirty tree */ 166 RB_ENTRY(buf) b_rbhash; /* RB node in vnode hash tree */ 167 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */ 168 struct buf *b_cluster_next; /* Next buffer (cluster code) */ 169 struct vnode *b_vp; /* (vp, loffset) index */ 170 struct bio b_bio_array[NBUF_BIO]; /* BIO translation layers */ 171 u_int32_t b_flags; /* B_* flags. */ 172 unsigned short b_qindex; /* buffer queue index */ 173 unsigned short b_unused01; 174 struct lock b_lock; /* Buffer lock */ 175 buf_cmd_t b_cmd; /* I/O command */ 176 int b_bufsize; /* Allocated buffer size. */ 177 int b_runningbufspace; /* when I/O is running, pipelining */ 178 int b_bcount; /* Valid bytes in buffer. */ 179 int b_resid; /* Remaining I/O */ 180 int b_error; /* Error return */ 181 caddr_t b_data; /* Memory, superblocks, indirect etc. */ 182 caddr_t b_kvabase; /* base kva for buffer */ 183 int b_kvasize; /* size of kva for buffer */ 184 int b_dirtyoff; /* Offset in buffer of dirty region. */ 185 int b_dirtyend; /* Offset of end of dirty region. */ 186 struct xio b_xio; /* data buffer page list management */ 187 struct workhead b_dep; /* List of filesystem dependencies. */ 188 }; 189 190 /* 191 * XXX temporary 192 */ 193 #define b_bio1 b_bio_array[0] /* logical layer */ 194 #define b_bio2 b_bio_array[1] /* (typically) the disk layer */ 195 #define b_loffset b_bio1.bio_offset 196 197 /* 198 * These flags are kept in b_flags. 199 * 200 * Notes: 201 * 202 * B_ASYNC VOP calls on bp's are usually async whether or not 203 * B_ASYNC is set, but some subsystems, such as NFS, like 204 * to know what is best for the caller so they can 205 * optimize the I/O. 206 * 207 * B_PAGING Indicates that bp is being used by the paging system or 208 * some paging system and that the bp is not linked into 209 * the b_vp's clean/dirty linked lists or ref counts. 210 * Buffer vp reassignments are illegal in this case. 211 * 212 * B_CACHE This may only be set if the buffer is entirely valid. 213 * The situation where B_DELWRI is set and B_CACHE is 214 * clear MUST be committed to disk by getblk() so 215 * B_DELWRI can also be cleared. See the comments for 216 * getblk() in kern/vfs_bio.c. If B_CACHE is clear, 217 * the caller is expected to clear B_ERROR|B_INVAL, 218 * set BUF_CMD_READ, and initiate an I/O. 219 * 220 * The 'entire buffer' is defined to be the range from 221 * 0 through b_bcount. 222 * 223 * B_MALLOC Request that the buffer be allocated from the malloc 224 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned. 225 * 226 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers 227 * by filesystems that allow clustering when the buffer 228 * is fully dirty and indicates that it may be clustered 229 * with other adjacent dirty buffers. Note the clustering 230 * may not be used with the stage 1 data write under NFS 231 * but may be used for the commit rpc portion. 232 * 233 * B_VMIO Indicates that the buffer is tied into an VM object. 234 * The buffer's data is always PAGE_SIZE aligned even 235 * if b_bufsize and b_bcount are not. ( b_bufsize is 236 * always at least DEV_BSIZE aligned, though ). 237 * 238 * B_DIRECT Hint that we should attempt to completely free 239 * the pages underlying the buffer. B_DIRECT is 240 * sticky until the buffer is released and typically 241 * only has an effect when B_RELBUF is also set. 242 * 243 * B_NOWDRAIN This flag should be set when a device (like VN) 244 * does a turn-around VOP_WRITE from its strategy 245 * routine. This flag prevents bwrite() from blocking 246 * in wdrain, avoiding a deadlock situation. 247 */ 248 249 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */ 250 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */ 251 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */ 252 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */ 253 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */ 254 #define B_CACHE 0x00000020 /* Bread found us in the cache. */ 255 #define B_HASHED 0x00000040 /* Indexed via v_rbhash_tree */ 256 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */ 257 #define B_BNOCLIP 0x00000100 /* EOF clipping b_bcount not allowed */ 258 #define B_UNUSED0200 0x00000200 259 #define B_EINTR 0x00000400 /* I/O was interrupted */ 260 #define B_ERROR 0x00000800 /* I/O error occurred. */ 261 #define B_UNUSED1000 0x00001000 /* Unused */ 262 #define B_INVAL 0x00002000 /* Does not contain valid info. */ 263 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */ 264 #define B_NOCACHE 0x00008000 /* Destroy buffer AND backing store */ 265 #define B_MALLOC 0x00010000 /* malloced b_data */ 266 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */ 267 #define B_UNUSED40000 0x00040000 268 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */ 269 #define B_UNUSED100000 0x00100000 270 #define B_DIRTY 0x00200000 /* Needs writing later. */ 271 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */ 272 #define B_WANT 0x00800000 /* Used by vm_pager.c */ 273 #define B_VNCLEAN 0x01000000 /* On vnode clean list */ 274 #define B_VNDIRTY 0x02000000 /* On vnode dirty list */ 275 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */ 276 #define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */ 277 #define B_RAM 0x10000000 /* Read ahead mark (flag) */ 278 #define B_VMIO 0x20000000 /* VMIO flag */ 279 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */ 280 #define B_NOWDRAIN 0x80000000 /* Avoid wdrain deadlock */ 281 282 #define PRINT_BUF_FLAGS "\20" \ 283 "\40nowdrain\37cluster\36vmio\35ram\34ordered" \ 284 "\33paging\32vndirty\31vnclean\30want\27relbuf\26dirty" \ 285 "\25unused20\24raw\23unused18\22clusterok\21malloc\20nocache" \ 286 "\17locked\16inval\15unused12\14error\13eintr\12unused9\11unused8" \ 287 "\10delwri\7hashed\6cache\5deferred\4direct\3async\2needcommit\1age" 288 289 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */ 290 291 #ifdef _KERNEL 292 /* 293 * Buffer locking. See sys/buf2.h for inline functions. 294 */ 295 extern char *buf_wmesg; /* Default buffer lock message */ 296 #define BUF_WMESG "bufwait" 297 298 #endif /* _KERNEL */ 299 300 struct bio_queue_head { 301 TAILQ_HEAD(bio_queue, bio) queue; 302 off_t last_offset; 303 struct bio *insert_point; 304 struct bio *switch_point; 305 }; 306 307 /* 308 * This structure describes a clustered I/O. 309 */ 310 struct cluster_save { 311 int bs_nchildren; /* Number of associated buffers. */ 312 struct buf **bs_children; /* List of associated buffers. */ 313 }; 314 315 /* 316 * Zero out the buffer's data area. 317 */ 318 #define clrbuf(bp) { \ 319 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \ 320 (bp)->b_resid = 0; \ 321 } 322 323 /* 324 * Flags to low-level bitmap allocation routines (balloc). 325 * 326 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count 327 * to 127. 328 */ 329 #define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */ 330 #define B_SEQSHIFT 24 /* Sequential heuristic shift. */ 331 #define B_SEQMAX 0x7F 332 #define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */ 333 #define B_SYNC 0x02 /* Do all allocations synchronously. */ 334 335 #ifdef _KERNEL 336 extern int nbuf; /* The number of buffer headers */ 337 extern int maxswzone; /* Max KVA for swap structures */ 338 extern int maxbcache; /* Max KVA for buffer cache */ 339 extern int runningbufspace; 340 extern int buf_maxio; /* nominal maximum I/O for buffer */ 341 extern struct buf *buf; /* The buffer headers. */ 342 extern char *buffers; /* The buffer contents. */ 343 extern int bufpages; /* Number of memory pages in the buffer pool. */ 344 extern struct buf *swbuf; /* Swap I/O buffer headers. */ 345 extern int nswbuf; /* Number of swap I/O buffer headers. */ 346 347 struct uio; 348 349 void bufinit (void); 350 void bwillwrite (void); 351 int buf_dirty_count_severe (void); 352 void initbufbio(struct buf *); 353 void reinitbufbio(struct buf *); 354 void clearbiocache(struct bio *); 355 void bremfree (struct buf *); 356 int bread (struct vnode *, off_t, int, struct buf **); 357 int breadn (struct vnode *, off_t, int, off_t *, int *, int, 358 struct buf **); 359 int bwrite (struct buf *); 360 void bdwrite (struct buf *); 361 void bawrite (struct buf *); 362 void bdirty (struct buf *); 363 void bundirty (struct buf *); 364 int bowrite (struct buf *); 365 void brelse (struct buf *); 366 void bqrelse (struct buf *); 367 int vfs_bio_awrite (struct buf *); 368 struct buf *getpbuf (int *); 369 int inmem (struct vnode *, off_t); 370 struct buf *findblk (struct vnode *, off_t); 371 struct buf *getblk (struct vnode *, off_t, int, int, int); 372 struct buf *geteblk (int); 373 struct bio *push_bio(struct bio *); 374 void pop_bio(struct bio *); 375 int biowait (struct buf *); 376 void biodone (struct bio *); 377 378 void cluster_append(struct bio *, struct buf *); 379 int cluster_read (struct vnode *, off_t, off_t, int, 380 int, int, struct buf **); 381 int cluster_wbuild (struct vnode *, int, off_t, int); 382 void cluster_write (struct buf *, off_t, int); 383 int physread (struct dev_read_args *); 384 int physwrite (struct dev_write_args *); 385 void vfs_bio_set_validclean (struct buf *, int base, int size); 386 void vfs_bio_clrbuf (struct buf *); 387 void vfs_busy_pages (struct vnode *, struct buf *); 388 void vfs_unbusy_pages (struct buf *); 389 int vmapbuf (struct buf *, caddr_t, int); 390 void vunmapbuf (struct buf *); 391 void relpbuf (struct buf *, int *); 392 void brelvp (struct buf *); 393 void bgetvp (struct vnode *, struct buf *); 394 int allocbuf (struct buf *bp, int size); 395 int scan_all_buffers (int (*)(struct buf *, void *), void *); 396 void reassignbuf (struct buf *); 397 struct buf *trypbuf (int *); 398 399 #endif /* _KERNEL */ 400 #endif /* _KERNEL || _KERNEL_STRUCTURES */ 401 #endif /* !_SYS_BUF_H_ */ 402