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 1.41 2007/11/07 00:46:38 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 typedef enum buf_cmd { 90 BUF_CMD_DONE = 0, 91 BUF_CMD_READ, 92 BUF_CMD_WRITE, 93 BUF_CMD_FREEBLKS, 94 BUF_CMD_FORMAT 95 } buf_cmd_t; 96 97 /* 98 * The buffer header describes an I/O operation in the kernel. 99 * 100 * NOTES: 101 * b_bufsize represents the filesystem block size (for this particular 102 * block) and/or the allocation size or original request size. This 103 * field is NOT USED by lower device layers. VNode and device 104 * strategy routines WILL NEVER ACCESS THIS FIELD. 105 * 106 * b_bcount represents the I/O request size. Unless B_NOBCLIP is set, 107 * the device chain is allowed to clip b_bcount to accomodate the device 108 * EOF. Note that this is different from the byte oriented file EOF. 109 * If B_NOBCLIP is set, the device chain is required to generate an 110 * error if it would othrewise have to clip the request. Buffers 111 * obtained via getblk() automatically set B_NOBCLIP. It is important 112 * to note that EOF clipping via b_bcount is different from EOF clipping 113 * via returning a b_actual < b_bcount. B_NOBCLIP only effects block 114 * oriented EOF clipping (b_bcount modifications). 115 * 116 * b_actual represents the number of bytes of I/O that actually occured, 117 * whether an error occured or not. b_actual must be initialized to 0 118 * prior to initiating I/O as the device drivers will assume it to 119 * start at 0. 120 * 121 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned 122 * ranges of dirty data that need to be written to backing store. 123 * The range is typically clipped at b_bcount (not b_bufsize). 124 * 125 * b_bio1 and b_bio2 represent the two primary I/O layers. Additional 126 * I/O layers are allocated out of the object cache and may also exist. 127 * 128 * b_bio1 is the logical layer and contains offset or block number 129 * data for the primary vnode, b_vp. I/O operations are almost 130 * universally initiated from the logical layer, so you will often 131 * see things like: vn_strategy(bp->b_vp, &bp->b_bio1). 132 * 133 * b_bio2 is the first physical layer (typically the slice-relative 134 * layer) and contains the translated offset or block number for 135 * the block device underlying a filesystem. Filesystems such as UFS 136 * will maintain cached translations and you may see them initiate 137 * a 'physical' I/O using vn_strategy(devvp, &bp->b_bio2). BUT, 138 * remember that the layering is relative to bp->b_vp, so the 139 * device-relative block numbers for buffer cache operations that occur 140 * directly on a block device will be in the first BIO layer. 141 * 142 * b_ops - initialized if a buffer has a bio_ops 143 * 144 * NOTE!!! Only the BIO subsystem accesses b_bio1 and b_bio2 directly. 145 * ALL STRATEGY LAYERS FOR BOTH VNODES AND DEVICES ONLY ACCESS THE BIO 146 * PASSED TO THEM, AND WILL PUSH ANOTHER BIO LAYER IF FORWARDING THE 147 * I/O DEEPER. In particular, a vn_strategy() or dev_dstrategy() 148 * call should not ever access buf->b_vp as this vnode may be totally 149 * unrelated to the vnode/device whos strategy routine was called. 150 */ 151 struct buf { 152 RB_ENTRY(buf) b_rbnode; /* RB node in vnode clean/dirty tree */ 153 RB_ENTRY(buf) b_rbhash; /* RB node in vnode hash tree */ 154 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */ 155 struct buf *b_cluster_next; /* Next buffer (cluster code) */ 156 struct vnode *b_vp; /* (vp, loffset) index */ 157 struct bio b_bio_array[NBUF_BIO]; /* BIO translation layers */ 158 u_int32_t b_flags; /* B_* flags. */ 159 unsigned short b_qindex; /* buffer queue index */ 160 unsigned short b_unused01; 161 struct lock b_lock; /* Buffer lock */ 162 buf_cmd_t b_cmd; /* I/O command */ 163 int b_bufsize; /* Allocated buffer size. */ 164 int b_runningbufspace; /* when I/O is running, pipelining */ 165 int b_bcount; /* Valid bytes in buffer. */ 166 int b_resid; /* Remaining I/O */ 167 int b_error; /* Error return */ 168 caddr_t b_data; /* Memory, superblocks, indirect etc. */ 169 caddr_t b_kvabase; /* base kva for buffer */ 170 int b_kvasize; /* size of kva for buffer */ 171 int b_dirtyoff; /* Offset in buffer of dirty region. */ 172 int b_dirtyend; /* Offset of end of dirty region. */ 173 struct xio b_xio; /* data buffer page list management */ 174 struct bio_ops *b_ops; /* bio_ops used w/ b_dep */ 175 struct workhead b_dep; /* List of filesystem dependencies. */ 176 }; 177 178 /* 179 * XXX temporary 180 */ 181 #define b_bio1 b_bio_array[0] /* logical layer */ 182 #define b_bio2 b_bio_array[1] /* (typically) the disk layer */ 183 #define b_loffset b_bio1.bio_offset 184 185 /* 186 * These flags are kept in b_flags. 187 * 188 * Notes: 189 * 190 * B_ASYNC VOP calls on bp's are usually async whether or not 191 * B_ASYNC is set, but some subsystems, such as NFS, like 192 * to know what is best for the caller so they can 193 * optimize the I/O. 194 * 195 * B_PAGING Indicates that bp is being used by the paging system or 196 * some paging system and that the bp is not linked into 197 * the b_vp's clean/dirty linked lists or ref counts. 198 * Buffer vp reassignments are illegal in this case. 199 * 200 * B_CACHE This may only be set if the buffer is entirely valid. 201 * The situation where B_DELWRI is set and B_CACHE is 202 * clear MUST be committed to disk by getblk() so 203 * B_DELWRI can also be cleared. See the comments for 204 * getblk() in kern/vfs_bio.c. If B_CACHE is clear, 205 * the caller is expected to clear B_ERROR|B_INVAL, 206 * set BUF_CMD_READ, and initiate an I/O. 207 * 208 * The 'entire buffer' is defined to be the range from 209 * 0 through b_bcount. 210 * 211 * B_MALLOC Request that the buffer be allocated from the malloc 212 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned. 213 * 214 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers 215 * by filesystems that allow clustering when the buffer 216 * is fully dirty and indicates that it may be clustered 217 * with other adjacent dirty buffers. Note the clustering 218 * may not be used with the stage 1 data write under NFS 219 * but may be used for the commit rpc portion. 220 * 221 * B_VMIO Indicates that the buffer is tied into an VM object. 222 * The buffer's data is always PAGE_SIZE aligned even 223 * if b_bufsize and b_bcount are not. ( b_bufsize is 224 * always at least DEV_BSIZE aligned, though ). 225 * 226 * B_DIRECT Hint that we should attempt to completely free 227 * the pages underlying the buffer. B_DIRECT is 228 * sticky until the buffer is released and typically 229 * only has an effect when B_RELBUF is also set. 230 * 231 * B_NOWDRAIN This flag should be set when a device (like VN) 232 * does a turn-around VOP_WRITE from its strategy 233 * routine. This flag prevents bwrite() from blocking 234 * in wdrain, avoiding a deadlock situation. 235 * 236 * B_LOCKED The buffer will be released to the locked queue 237 * regardless of its current state. Note that 238 * if B_DELWRI is set, no I/O occurs until the caller 239 * acquires the buffer, clears B_LOCKED, then releases 240 * it again. 241 */ 242 243 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */ 244 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */ 245 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */ 246 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */ 247 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */ 248 #define B_CACHE 0x00000020 /* Bread found us in the cache. */ 249 #define B_HASHED 0x00000040 /* Indexed via v_rbhash_tree */ 250 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */ 251 #define B_BNOCLIP 0x00000100 /* EOF clipping b_bcount not allowed */ 252 #define B_UNUSED0200 0x00000200 253 #define B_EINTR 0x00000400 /* I/O was interrupted */ 254 #define B_ERROR 0x00000800 /* I/O error occurred. */ 255 #define B_UNUSED1000 0x00001000 /* Unused */ 256 #define B_INVAL 0x00002000 /* Does not contain valid info. */ 257 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */ 258 #define B_NOCACHE 0x00008000 /* Destroy buffer AND backing store */ 259 #define B_MALLOC 0x00010000 /* malloced b_data */ 260 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */ 261 #define B_UNUSED40000 0x00040000 262 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */ 263 #define B_UNUSED100000 0x00100000 264 #define B_DIRTY 0x00200000 /* Needs writing later. */ 265 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */ 266 #define B_WANT 0x00800000 /* Used by vm_pager.c */ 267 #define B_VNCLEAN 0x01000000 /* On vnode clean list */ 268 #define B_VNDIRTY 0x02000000 /* On vnode dirty list */ 269 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */ 270 #define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */ 271 #define B_RAM 0x10000000 /* Read ahead mark (flag) */ 272 #define B_VMIO 0x20000000 /* VMIO flag */ 273 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */ 274 #define B_NOWDRAIN 0x80000000 /* Avoid wdrain deadlock */ 275 276 #define PRINT_BUF_FLAGS "\20" \ 277 "\40nowdrain\37cluster\36vmio\35ram\34ordered" \ 278 "\33paging\32vndirty\31vnclean\30want\27relbuf\26dirty" \ 279 "\25unused20\24raw\23unused18\22clusterok\21malloc\20nocache" \ 280 "\17locked\16inval\15unused12\14error\13eintr\12unused9\11unused8" \ 281 "\10delwri\7hashed\6cache\5deferred\4direct\3async\2needcommit\1age" 282 283 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */ 284 285 #ifdef _KERNEL 286 /* 287 * Buffer locking. See sys/buf2.h for inline functions. 288 */ 289 extern char *buf_wmesg; /* Default buffer lock message */ 290 #define BUF_WMESG "bufwait" 291 292 #endif /* _KERNEL */ 293 294 struct bio_queue_head { 295 TAILQ_HEAD(bio_queue, bio) queue; 296 off_t last_offset; 297 struct bio *insert_point; 298 struct bio *switch_point; 299 }; 300 301 /* 302 * This structure describes a clustered I/O. 303 */ 304 struct cluster_save { 305 int bs_nchildren; /* Number of associated buffers. */ 306 struct buf **bs_children; /* List of associated buffers. */ 307 }; 308 309 /* 310 * Zero out the buffer's data area. 311 */ 312 #define clrbuf(bp) { \ 313 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \ 314 (bp)->b_resid = 0; \ 315 } 316 317 /* 318 * Flags to low-level bitmap allocation routines (balloc). 319 * 320 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count 321 * to 127. 322 */ 323 #define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */ 324 #define B_SEQSHIFT 24 /* Sequential heuristic shift. */ 325 #define B_SEQMAX 0x7F 326 #define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */ 327 #define B_SYNC 0x02 /* Do all allocations synchronously. */ 328 329 #ifdef _KERNEL 330 extern int nbuf; /* The number of buffer headers */ 331 extern int maxswzone; /* Max KVA for swap structures */ 332 extern int maxbcache; /* Max KVA for buffer cache */ 333 extern int runningbufspace; 334 extern int buf_maxio; /* nominal maximum I/O for buffer */ 335 extern struct buf *buf; /* The buffer headers. */ 336 extern char *buffers; /* The buffer contents. */ 337 extern int bufpages; /* Number of memory pages in the buffer pool. */ 338 extern struct buf *swbuf; /* Swap I/O buffer headers. */ 339 extern int nswbuf; /* Number of swap I/O buffer headers. */ 340 341 struct uio; 342 343 void bufinit (void); 344 void bwillwrite (void); 345 int buf_dirty_count_severe (void); 346 void initbufbio(struct buf *); 347 void reinitbufbio(struct buf *); 348 void clearbiocache(struct bio *); 349 void bremfree (struct buf *); 350 int bread (struct vnode *, off_t, int, struct buf **); 351 int breadn (struct vnode *, off_t, int, off_t *, int *, int, 352 struct buf **); 353 int bwrite (struct buf *); 354 void bdwrite (struct buf *); 355 void bawrite (struct buf *); 356 void bdirty (struct buf *); 357 void bundirty (struct buf *); 358 int bowrite (struct buf *); 359 void brelse (struct buf *); 360 void bqrelse (struct buf *); 361 int vfs_bio_awrite (struct buf *); 362 struct buf *getpbuf (int *); 363 int inmem (struct vnode *, off_t); 364 struct buf *findblk (struct vnode *, off_t); 365 struct buf *getblk (struct vnode *, off_t, int, int, int); 366 struct buf *geteblk (int); 367 void regetblk(struct buf *bp); 368 struct bio *push_bio(struct bio *); 369 void pop_bio(struct bio *); 370 int biowait (struct buf *); 371 void biodone (struct bio *); 372 373 void cluster_append(struct bio *, struct buf *); 374 int cluster_read (struct vnode *, off_t, off_t, int, 375 int, int, struct buf **); 376 int cluster_wbuild (struct vnode *, int, off_t, int); 377 void cluster_write (struct buf *, off_t, int); 378 int physread (struct dev_read_args *); 379 int physwrite (struct dev_write_args *); 380 void vfs_bio_set_validclean (struct buf *, int base, int size); 381 void vfs_bio_clrbuf (struct buf *); 382 void vfs_busy_pages (struct vnode *, struct buf *); 383 void vfs_unbusy_pages (struct buf *); 384 int vmapbuf (struct buf *, caddr_t, int); 385 void vunmapbuf (struct buf *); 386 void relpbuf (struct buf *, int *); 387 void brelvp (struct buf *); 388 void bgetvp (struct vnode *, struct buf *); 389 int allocbuf (struct buf *bp, int size); 390 int scan_all_buffers (int (*)(struct buf *, void *), void *); 391 void reassignbuf (struct buf *); 392 struct buf *trypbuf (int *); 393 void bio_ops_sync(struct mount *mp); 394 395 #endif /* _KERNEL */ 396 #endif /* _KERNEL || _KERNEL_STRUCTURES */ 397 #endif /* !_SYS_BUF_H_ */ 398