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 1.7 2003/08/20 07:31:21 rob 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 struct buf; 57 struct mount; 58 struct vnode; 59 60 /* 61 * To avoid including <ufs/ffs/softdep.h> 62 */ 63 LIST_HEAD(workhead, worklist); 64 /* 65 * These are currently used only by the soft dependency code, hence 66 * are stored once in a global variable. If other subsystems wanted 67 * to use these hooks, a pointer to a set of bio_ops could be added 68 * to each buffer. 69 */ 70 extern struct bio_ops { 71 void (*io_start) (struct buf *); 72 void (*io_complete) (struct buf *); 73 void (*io_deallocate) (struct buf *); 74 int (*io_fsync) (struct vnode *); 75 int (*io_sync) (struct mount *); 76 void (*io_movedeps) (struct buf *, struct buf *); 77 int (*io_countdeps) (struct buf *, int); 78 } bioops; 79 80 struct iodone_chain { 81 long ic_prev_flags; 82 void (*ic_prev_iodone) (struct buf *); 83 void *ic_prev_iodone_chain; 84 struct { 85 long ia_long; 86 void *ia_ptr; 87 } ic_args[5]; 88 }; 89 90 /* 91 * The buffer header describes an I/O operation in the kernel. 92 * 93 * NOTES: 94 * b_bufsize, b_bcount. b_bufsize is the allocation size of the 95 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the 96 * originally requested buffer size and can serve as a bounds check 97 * against EOF. For most, but not all uses, b_bcount == b_bufsize. 98 * 99 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned 100 * ranges of dirty data that need to be written to backing store. 101 * The range is typically clipped at b_bcount ( not b_bufsize ). 102 * 103 * b_resid. Number of bytes remaining in I/O. After an I/O operation 104 * completes, b_resid is usually 0 indicating 100% success. 105 */ 106 struct buf { 107 LIST_ENTRY(buf) b_hash; /* Hash chain. */ 108 TAILQ_ENTRY(buf) b_vnbufs; /* Buffer's associated vnode. */ 109 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */ 110 TAILQ_ENTRY(buf) b_act; /* Device driver queue when active. *new* */ 111 long b_flags; /* B_* flags. */ 112 unsigned short b_qindex; /* buffer queue index */ 113 unsigned char b_xflags; /* extra flags */ 114 struct lock b_lock; /* Buffer lock */ 115 int b_error; /* Errno value. */ 116 long b_bufsize; /* Allocated buffer size. */ 117 long b_runningbufspace; /* when I/O is running, pipelining */ 118 long b_bcount; /* Valid bytes in buffer. */ 119 long b_resid; /* Remaining I/O. */ 120 dev_t b_dev; /* Device associated with buffer. */ 121 caddr_t b_data; /* Memory, superblocks, indirect etc. */ 122 caddr_t b_kvabase; /* base kva for buffer */ 123 int b_kvasize; /* size of kva for buffer */ 124 daddr_t b_lblkno; /* Logical block number. */ 125 daddr_t b_blkno; /* Underlying physical block number. */ 126 off_t b_offset; /* Offset into file */ 127 /* Function to call upon completion. */ 128 void (*b_iodone) (struct buf *); 129 /* For nested b_iodone's. */ 130 struct iodone_chain *b_iodone_chain; 131 struct vnode *b_vp; /* Device vnode. */ 132 int b_dirtyoff; /* Offset in buffer of dirty region. */ 133 int b_dirtyend; /* Offset of end of dirty region. */ 134 daddr_t b_pblkno; /* physical block number */ 135 void *b_saveaddr; /* Original b_addr for physio. */ 136 void *b_driver1; /* for private use by the driver */ 137 void *b_driver2; /* for private use by the driver */ 138 void *b_caller1; /* for private use by the caller */ 139 void *b_caller2; /* for private use by the caller */ 140 union pager_info { 141 void *pg_spc; 142 int pg_reqpage; 143 } b_pager; 144 union cluster_info { 145 TAILQ_HEAD(cluster_list_head, buf) cluster_head; 146 TAILQ_ENTRY(buf) cluster_entry; 147 } b_cluster; 148 struct vm_page *b_pages[btoc(MAXPHYS)]; 149 int b_npages; 150 struct workhead b_dep; /* List of filesystem dependencies. */ 151 struct chain_info { /* buffer chaining */ 152 struct buf *parent; 153 int count; 154 } b_chain; 155 }; 156 157 #define b_spc b_pager.pg_spc 158 159 /* 160 * These flags are kept in b_flags. 161 * 162 * Notes: 163 * 164 * B_ASYNC VOP calls on bp's are usually async whether or not 165 * B_ASYNC is set, but some subsystems, such as NFS, like 166 * to know what is best for the caller so they can 167 * optimize the I/O. 168 * 169 * B_PAGING Indicates that bp is being used by the paging system or 170 * some paging system and that the bp is not linked into 171 * the b_vp's clean/dirty linked lists or ref counts. 172 * Buffer vp reassignments are illegal in this case. 173 * 174 * B_CACHE This may only be set if the buffer is entirely valid. 175 * The situation where B_DELWRI is set and B_CACHE is 176 * clear MUST be committed to disk by getblk() so 177 * B_DELWRI can also be cleared. See the comments for 178 * getblk() in kern/vfs_bio.c. If B_CACHE is clear, 179 * the caller is expected to clear B_ERROR|B_INVAL, 180 * set B_READ, and initiate an I/O. 181 * 182 * The 'entire buffer' is defined to be the range from 183 * 0 through b_bcount. 184 * 185 * B_MALLOC Request that the buffer be allocated from the malloc 186 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned. 187 * 188 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers 189 * by filesystems that allow clustering when the buffer 190 * is fully dirty and indicates that it may be clustered 191 * with other adjacent dirty buffers. Note the clustering 192 * may not be used with the stage 1 data write under NFS 193 * but may be used for the commit rpc portion. 194 * 195 * B_VMIO Indicates that the buffer is tied into an VM object. 196 * The buffer's data is always PAGE_SIZE aligned even 197 * if b_bufsize and b_bcount are not. ( b_bufsize is 198 * always at least DEV_BSIZE aligned, though ). 199 * 200 * B_DIRECT Hint that we should attempt to completely free 201 * the pages underlying the buffer. B_DIRECT is 202 * sticky until the buffer is released and typically 203 * only has an effect when B_RELBUF is also set. 204 * 205 * B_NOWDRAIN This flag should be set when a device (like VN) 206 * does a turn-around VOP_WRITE from its strategy 207 * routine. This flag prevents bwrite() from blocking 208 * in wdrain, avoiding a deadlock situation. 209 */ 210 211 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */ 212 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */ 213 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */ 214 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */ 215 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */ 216 #define B_CACHE 0x00000020 /* Bread found us in the cache. */ 217 #define B_CALL 0x00000040 /* Call b_iodone from biodone. */ 218 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */ 219 #define B_FREEBUF 0x00000100 /* Instruct driver: free blocks */ 220 #define B_DONE 0x00000200 /* I/O completed. */ 221 #define B_EINTR 0x00000400 /* I/O was interrupted */ 222 #define B_ERROR 0x00000800 /* I/O error occurred. */ 223 #define B_SCANNED 0x00001000 /* VOP_FSYNC funcs mark written bufs */ 224 #define B_INVAL 0x00002000 /* Does not contain valid info. */ 225 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */ 226 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */ 227 #define B_MALLOC 0x00010000 /* malloced b_data */ 228 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */ 229 #define B_PHYS 0x00040000 /* I/O to user memory. */ 230 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */ 231 #define B_READ 0x00100000 /* Read buffer. */ 232 #define B_DIRTY 0x00200000 /* Needs writing later. */ 233 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */ 234 #define B_WANT 0x00800000 /* Used by vm_pager.c */ 235 #define B_WRITE 0x00000000 /* Write buffer (pseudo flag). */ 236 #define B_WRITEINPROG 0x01000000 /* Write in progress. */ 237 #define B_XXX 0x02000000 /* Debugging flag. */ 238 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */ 239 #define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */ 240 #define B_RAM 0x10000000 /* Read ahead mark (flag) */ 241 #define B_VMIO 0x20000000 /* VMIO flag */ 242 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */ 243 #define B_NOWDRAIN 0x80000000 /* Avoid wdrain deadlock */ 244 245 #define PRINT_BUF_FLAGS "\20\40nowdrain\37cluster\36vmio\35ram\34ordered" \ 246 "\33paging\32xxx\31writeinprog\30want\27relbuf\26dirty" \ 247 "\25read\24raw\23phys\22clusterok\21malloc\20nocache" \ 248 "\17locked\16inval\15scanned\14error\13eintr\12done\11freebuf" \ 249 "\10delwri\7call\6cache\4direct\3async\2needcommit\1age" 250 251 /* 252 * These flags are kept in b_xflags. 253 */ 254 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */ 255 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */ 256 #define BX_BKGRDWRITE 0x00000004 /* Do writes in background */ 257 #define BX_BKGRDINPROG 0x00000008 /* Background write in progress */ 258 #define BX_BKGRDWAIT 0x00000010 /* Background write waiting */ 259 #define BX_AUTOCHAINDONE 0x00000020 /* pager I/O chain auto mode */ 260 261 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */ 262 263 #ifdef _KERNEL 264 /* 265 * Buffer locking. See sys/buf2.h for inline functions. 266 */ 267 struct lwkt_token buftimetoken; /* Interlock on setting prio and timo */ 268 extern char *buf_wmesg; /* Default buffer lock message */ 269 #define BUF_WMESG "bufwait" 270 271 #endif /* _KERNEL */ 272 273 struct buf_queue_head { 274 TAILQ_HEAD(buf_queue, buf) queue; 275 daddr_t last_pblkno; 276 struct buf *insert_point; 277 struct buf *switch_point; 278 }; 279 280 /* 281 * This structure describes a clustered I/O. It is stored in the b_saveaddr 282 * field of the buffer on which I/O is done. At I/O completion, cluster 283 * callback uses the structure to parcel I/O's to individual buffers, and 284 * then free's this structure. 285 */ 286 struct cluster_save { 287 long bs_bcount; /* Saved b_bcount. */ 288 long bs_bufsize; /* Saved b_bufsize. */ 289 void *bs_saveaddr; /* Saved b_addr. */ 290 int bs_nchildren; /* Number of associated buffers. */ 291 struct buf **bs_children; /* List of associated buffers. */ 292 }; 293 294 /* 295 * Definitions for the buffer free lists. 296 */ 297 #define BUFFER_QUEUES 6 /* number of free buffer queues */ 298 299 #define QUEUE_NONE 0 /* on no queue */ 300 #define QUEUE_LOCKED 1 /* locked buffers */ 301 #define QUEUE_CLEAN 2 /* non-B_DELWRI buffers */ 302 #define QUEUE_DIRTY 3 /* B_DELWRI buffers */ 303 #define QUEUE_EMPTYKVA 4 /* empty buffer headers w/KVA assignment */ 304 #define QUEUE_EMPTY 5 /* empty buffer headers */ 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 extern TAILQ_HEAD(swqueue, buf) bswlist; 338 extern TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES]; 339 340 struct uio; 341 342 caddr_t bufhashinit (caddr_t); 343 void bufinit (void); 344 void bwillwrite (void); 345 int buf_dirty_count_severe (void); 346 void bremfree (struct buf *); 347 int bread (struct vnode *, daddr_t, int, struct buf **); 348 int breadn (struct vnode *, daddr_t, int, daddr_t *, int *, int, 349 struct buf **); 350 int bwrite (struct buf *); 351 void bdwrite (struct buf *); 352 void bawrite (struct buf *); 353 void bdirty (struct buf *); 354 void bundirty (struct buf *); 355 int bowrite (struct buf *); 356 void brelse (struct buf *); 357 void bqrelse (struct buf *); 358 int vfs_bio_awrite (struct buf *); 359 struct buf * getpbuf (int *); 360 struct buf *incore (struct vnode *, daddr_t); 361 struct buf *gbincore (struct vnode *, daddr_t); 362 int inmem (struct vnode *, daddr_t); 363 struct buf *getblk (struct vnode *, daddr_t, int, int, int); 364 struct buf *geteblk (int); 365 int biowait (struct buf *); 366 void biodone (struct buf *); 367 368 void cluster_callback (struct buf *); 369 int cluster_read (struct vnode *, u_quad_t, daddr_t, long, 370 long, int, struct buf **); 371 int cluster_wbuild (struct vnode *, long, daddr_t, int); 372 void cluster_write (struct buf *, u_quad_t, int); 373 int physio (dev_t dev, struct uio *uio, int ioflag); 374 #define physread physio 375 #define physwrite physio 376 void vfs_bio_set_validclean (struct buf *, int base, int size); 377 void vfs_bio_clrbuf (struct buf *); 378 void vfs_busy_pages (struct buf *, int clear_modify); 379 void vfs_unbusy_pages (struct buf *); 380 void vwakeup (struct buf *); 381 int vmapbuf (struct buf *); 382 void vunmapbuf (struct buf *); 383 void relpbuf (struct buf *, int *); 384 void brelvp (struct buf *); 385 void bgetvp (struct vnode *, struct buf *); 386 void pbgetvp (struct vnode *, struct buf *); 387 void pbrelvp (struct buf *); 388 int allocbuf (struct buf *bp, int size); 389 void reassignbuf (struct buf *, struct vnode *); 390 void pbreassignbuf (struct buf *, struct vnode *); 391 struct buf *trypbuf (int *); 392 393 #endif /* _KERNEL */ 394 395 #endif /* !_SYS_BUF_H_ */ 396