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 1.19 2005/08/09 05:11:42 hmp Exp $ 41 */ 42 43 #ifndef _SYS_BUF_H_ 44 #define _SYS_BUF_H_ 45 46 #ifndef _SYS_QUEUE_H_ 47 #include <sys/queue.h> 48 #endif 49 #ifndef _SYS_LOCK_H_ 50 #include <sys/lock.h> 51 #endif 52 #ifndef _SYS_DEVICE_H_ 53 #include <sys/device.h> 54 #endif 55 56 #ifndef _SYS_XIO_H_ 57 #include <sys/xio.h> 58 #endif 59 #ifndef _SYS_TREE_H_ 60 #include <sys/tree.h> 61 #endif 62 63 #ifndef _SYS_BIO_H_ 64 #include <sys/bio.h> 65 #endif 66 67 struct buf; 68 struct mount; 69 struct vnode; 70 struct xio; 71 72 struct buf_rb_tree; 73 RB_PROTOTYPE(buf_rb_tree, buf, b_rbnode, rb_buf_compare); 74 75 /* 76 * To avoid including <ufs/ffs/softdep.h> 77 */ 78 LIST_HEAD(workhead, worklist); 79 /* 80 * These are currently used only by the soft dependency code, hence 81 * are stored once in a global variable. If other subsystems wanted 82 * to use these hooks, a pointer to a set of bio_ops could be added 83 * to each buffer. 84 */ 85 extern struct bio_ops { 86 void (*io_start) (struct buf *); 87 void (*io_complete) (struct buf *); 88 void (*io_deallocate) (struct buf *); 89 int (*io_fsync) (struct vnode *); 90 int (*io_sync) (struct mount *); 91 void (*io_movedeps) (struct buf *, struct buf *); 92 int (*io_countdeps) (struct buf *, int); 93 } bioops; 94 95 struct iodone_chain { 96 long ic_prev_flags; 97 void (*ic_prev_iodone) (struct buf *); 98 void *ic_prev_iodone_chain; 99 struct { 100 long ia_long; 101 void *ia_ptr; 102 } ic_args[5]; 103 }; 104 105 /* 106 * The buffer header describes an I/O operation in the kernel. 107 * 108 * NOTES: 109 * b_bufsize, b_bcount. b_bufsize is the allocation size of the 110 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the 111 * originally requested buffer size and can serve as a bounds check 112 * against EOF. For most, but not all uses, b_bcount == b_bufsize. 113 * 114 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned 115 * ranges of dirty data that need to be written to backing store. 116 * The range is typically clipped at b_bcount ( not b_bufsize ). 117 * 118 * b_resid. Number of bytes remaining in I/O. After an I/O operation 119 * completes, b_resid is usually 0 indicating 100% success. 120 */ 121 struct buf { 122 LIST_ENTRY(buf) b_hash; /* Hash chain. */ 123 RB_ENTRY(buf) b_rbnode; /* Red-Black node in vnode RB tree */ 124 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */ 125 TAILQ_ENTRY(buf) b_act; /* driver queue when active. *new* */ 126 struct bio b_bio; /* Underlying I/O */ 127 long b_flags; /* B_* flags. */ 128 unsigned short b_qindex; /* buffer queue index */ 129 unsigned char b_xflags; /* extra flags */ 130 struct lock b_lock; /* Buffer lock */ 131 long b_bufsize; /* Allocated buffer size. */ 132 long b_runningbufspace; /* when I/O is running, pipelining */ 133 long b_bcount; /* Valid bytes in buffer. */ 134 caddr_t b_data; /* Memory, superblocks, indirect etc. */ 135 caddr_t b_kvabase; /* base kva for buffer */ 136 int b_kvasize; /* size of kva for buffer */ 137 off_t b_offset; /* Offset into file */ 138 /* For nested b_iodone's. */ 139 struct iodone_chain *b_iodone_chain; 140 struct vnode *b_vp; /* Device vnode. */ 141 int b_dirtyoff; /* Offset in buffer of dirty region. */ 142 int b_dirtyend; /* Offset of end of dirty region. */ 143 void *b_saveaddr; /* Original b_addr for physio. */ 144 union pager_info { 145 void *pg_spc; 146 int pg_reqpage; 147 } b_pager; 148 union cluster_info { 149 TAILQ_HEAD(cluster_list_head, buf) cluster_head; 150 TAILQ_ENTRY(buf) cluster_entry; 151 } b_cluster; 152 struct xio b_xio; /* page list management for buffer head. */ 153 struct workhead b_dep; /* List of filesystem dependencies. */ 154 struct chain_info { /* buffer chaining */ 155 struct buf *parent; 156 int count; 157 } b_chain; 158 }; 159 160 #define b_dev b_bio.bio_dev 161 #define b_resid b_bio.bio_resid 162 #define b_error b_bio.bio_error 163 #define b_lblkno b_bio.bio_lblkno 164 #define b_blkno b_bio.bio_blkno 165 #define b_pblkno b_bio.bio_pblkno 166 #define b_driver1 b_bio.bio_driver_ctx 167 #define b_caller1 b_bio.bio_caller_ctx 168 #define b_iodone b_bio.bio_done 169 #define b_spc b_pager.pg_spc 170 171 /* 172 * These flags are kept in b_flags. 173 * 174 * Notes: 175 * 176 * B_ASYNC VOP calls on bp's are usually async whether or not 177 * B_ASYNC is set, but some subsystems, such as NFS, like 178 * to know what is best for the caller so they can 179 * optimize the I/O. 180 * 181 * B_PAGING Indicates that bp is being used by the paging system or 182 * some paging system and that the bp is not linked into 183 * the b_vp's clean/dirty linked lists or ref counts. 184 * Buffer vp reassignments are illegal in this case. 185 * 186 * B_CACHE This may only be set if the buffer is entirely valid. 187 * The situation where B_DELWRI is set and B_CACHE is 188 * clear MUST be committed to disk by getblk() so 189 * B_DELWRI can also be cleared. See the comments for 190 * getblk() in kern/vfs_bio.c. If B_CACHE is clear, 191 * the caller is expected to clear B_ERROR|B_INVAL, 192 * set B_READ, and initiate an I/O. 193 * 194 * The 'entire buffer' is defined to be the range from 195 * 0 through b_bcount. 196 * 197 * B_MALLOC Request that the buffer be allocated from the malloc 198 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned. 199 * 200 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers 201 * by filesystems that allow clustering when the buffer 202 * is fully dirty and indicates that it may be clustered 203 * with other adjacent dirty buffers. Note the clustering 204 * may not be used with the stage 1 data write under NFS 205 * but may be used for the commit rpc portion. 206 * 207 * B_VMIO Indicates that the buffer is tied into an VM object. 208 * The buffer's data is always PAGE_SIZE aligned even 209 * if b_bufsize and b_bcount are not. ( b_bufsize is 210 * always at least DEV_BSIZE aligned, though ). 211 * 212 * B_DIRECT Hint that we should attempt to completely free 213 * the pages underlying the buffer. B_DIRECT is 214 * sticky until the buffer is released and typically 215 * only has an effect when B_RELBUF is also set. 216 * 217 * B_NOWDRAIN This flag should be set when a device (like VN) 218 * does a turn-around VOP_WRITE from its strategy 219 * routine. This flag prevents bwrite() from blocking 220 * in wdrain, avoiding a deadlock situation. 221 */ 222 223 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */ 224 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */ 225 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */ 226 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */ 227 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */ 228 #define B_CACHE 0x00000020 /* Bread found us in the cache. */ 229 #define B_UNUSED40 0x00000040 /* Unused */ 230 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */ 231 #define B_FREEBUF 0x00000100 /* Instruct driver: free blocks */ 232 #define B_DONE 0x00000200 /* I/O completed. */ 233 #define B_EINTR 0x00000400 /* I/O was interrupted */ 234 #define B_ERROR 0x00000800 /* I/O error occurred. */ 235 #define B_UNUSED1000 0x00001000 /* Unused */ 236 #define B_INVAL 0x00002000 /* Does not contain valid info. */ 237 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */ 238 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */ 239 #define B_MALLOC 0x00010000 /* malloced b_data */ 240 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */ 241 #define B_PHYS 0x00040000 /* I/O to user memory. */ 242 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */ 243 #define B_READ 0x00100000 /* Read buffer. */ 244 #define B_DIRTY 0x00200000 /* Needs writing later. */ 245 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */ 246 #define B_WANT 0x00800000 /* Used by vm_pager.c */ 247 #define B_WRITE 0x00000000 /* Write buffer (pseudo flag). */ 248 #define B_UNUSED1000000 0x01000000 /* Unused */ 249 #define B_XXX 0x02000000 /* Debugging flag. */ 250 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */ 251 #define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */ 252 #define B_RAM 0x10000000 /* Read ahead mark (flag) */ 253 #define B_VMIO 0x20000000 /* VMIO flag */ 254 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */ 255 #define B_NOWDRAIN 0x80000000 /* Avoid wdrain deadlock */ 256 257 #define PRINT_BUF_FLAGS "\20\40nowdrain\37cluster\36vmio\35ram\34ordered" \ 258 "\33paging\32xxx\31writeinprog\30want\27relbuf\26dirty" \ 259 "\25read\24raw\23phys\22clusterok\21malloc\20nocache" \ 260 "\17locked\16inval\15scanned\14error\13eintr\12done\11freebuf" \ 261 "\10delwri\7call\6cache\4direct\3async\2needcommit\1age" 262 263 /* 264 * These flags are kept in b_xflags. 265 */ 266 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */ 267 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */ 268 #define BX_BKGRDWRITE 0x00000004 /* Do writes in background */ 269 #define BX_BKGRDINPROG 0x00000008 /* Background write in progress */ 270 #define BX_BKGRDWAIT 0x00000010 /* Background write waiting */ 271 #define BX_AUTOCHAINDONE 0x00000020 /* pager I/O chain auto mode */ 272 273 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */ 274 275 #ifdef _KERNEL 276 /* 277 * Buffer locking. See sys/buf2.h for inline functions. 278 */ 279 extern struct lwkt_token buftimetoken; /* Interlock on setting prio and timo */ 280 extern char *buf_wmesg; /* Default buffer lock message */ 281 #define BUF_WMESG "bufwait" 282 283 #endif /* _KERNEL */ 284 285 struct buf_queue_head { 286 TAILQ_HEAD(buf_queue, buf) queue; 287 daddr_t last_pblkno; 288 struct buf *insert_point; 289 struct buf *switch_point; 290 }; 291 292 /* 293 * This structure describes a clustered I/O. It is stored in the b_saveaddr 294 * field of the buffer on which I/O is done. At I/O completion, cluster 295 * callback uses the structure to parcel I/O's to individual buffers, and 296 * then free's this structure. 297 */ 298 struct cluster_save { 299 long bs_bcount; /* Saved b_bcount. */ 300 long bs_bufsize; /* Saved b_bufsize. */ 301 void *bs_saveaddr; /* Saved b_addr. */ 302 int bs_nchildren; /* Number of associated buffers. */ 303 struct buf **bs_children; /* List of associated buffers. */ 304 }; 305 306 /* 307 * Zero out the buffer's data area. 308 */ 309 #define clrbuf(bp) { \ 310 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \ 311 (bp)->b_resid = 0; \ 312 } 313 314 /* 315 * Flags to low-level bitmap allocation routines (balloc). 316 * 317 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count 318 * to 127. 319 */ 320 #define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */ 321 #define B_SEQSHIFT 24 /* Sequential heuristic shift. */ 322 #define B_SEQMAX 0x7F 323 #define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */ 324 #define B_SYNC 0x02 /* Do all allocations synchronously. */ 325 326 #ifdef _KERNEL 327 extern int nbuf; /* The number of buffer headers */ 328 extern int maxswzone; /* Max KVA for swap structures */ 329 extern int maxbcache; /* Max KVA for buffer cache */ 330 extern int runningbufspace; 331 extern int buf_maxio; /* nominal maximum I/O for buffer */ 332 extern struct buf *buf; /* The buffer headers. */ 333 extern char *buffers; /* The buffer contents. */ 334 extern int bufpages; /* Number of memory pages in the buffer pool. */ 335 extern struct buf *swbuf; /* Swap I/O buffer headers. */ 336 extern int nswbuf; /* Number of swap I/O buffer headers. */ 337 338 struct uio; 339 340 caddr_t bufhashinit (caddr_t); 341 void bufinit (void); 342 void bwillwrite (void); 343 int buf_dirty_count_severe (void); 344 void bremfree (struct buf *); 345 int bread (struct vnode *, daddr_t, int, struct buf **); 346 int breadn (struct vnode *, daddr_t, int, daddr_t *, int *, int, 347 struct buf **); 348 int bwrite (struct buf *); 349 void bdwrite (struct buf *); 350 void bawrite (struct buf *); 351 void bdirty (struct buf *); 352 void bundirty (struct buf *); 353 int bowrite (struct buf *); 354 void brelse (struct buf *); 355 void bqrelse (struct buf *); 356 int vfs_bio_awrite (struct buf *); 357 struct buf * getpbuf (int *); 358 struct buf *incore (struct vnode *, daddr_t); 359 struct buf *gbincore (struct vnode *, daddr_t); 360 int inmem (struct vnode *, daddr_t); 361 struct buf *getblk (struct vnode *, daddr_t, int, int, int); 362 struct buf *geteblk (int); 363 int biowait (struct buf *); 364 void biodone (struct buf *); 365 366 void cluster_callback (struct buf *); 367 int cluster_read (struct vnode *, u_quad_t, daddr_t, long, 368 long, int, struct buf **); 369 int cluster_wbuild (struct vnode *, long, daddr_t, int); 370 void cluster_write (struct buf *, u_quad_t, int); 371 int physio (dev_t dev, struct uio *uio, int ioflag); 372 #define physread physio 373 #define physwrite physio 374 void vfs_bio_set_validclean (struct buf *, int base, int size); 375 void vfs_bio_clrbuf (struct buf *); 376 void vfs_busy_pages (struct buf *, int clear_modify); 377 void vfs_unbusy_pages (struct buf *); 378 void vwakeup (struct buf *); 379 int vmapbuf (struct buf *); 380 void vunmapbuf (struct buf *); 381 void relpbuf (struct buf *, int *); 382 void brelvp (struct buf *); 383 void bgetvp (struct vnode *, struct buf *); 384 void pbgetvp (struct vnode *, struct buf *); 385 void pbrelvp (struct buf *); 386 int allocbuf (struct buf *bp, int size); 387 void reassignbuf (struct buf *, struct vnode *); 388 void pbreassignbuf (struct buf *, struct vnode *); 389 struct buf *trypbuf (int *); 390 391 #endif /* _KERNEL */ 392 393 #endif /* !_SYS_BUF_H_ */ 394