1 /* $NetBSD: ufs_bmap.c,v 1.17 2002/05/11 12:23:53 enami Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. All advertising materials mentioning features or use of this software 21 * must display the following acknowledgement: 22 * This product includes software developed by the University of 23 * California, Berkeley and its contributors. 24 * 4. Neither the name of the University nor the names of its contributors 25 * may be used to endorse or promote products derived from this software 26 * without specific prior written permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 38 * SUCH DAMAGE. 39 * 40 * @(#)ufs_bmap.c 8.8 (Berkeley) 8/11/95 41 */ 42 43 #include <sys/cdefs.h> 44 __KERNEL_RCSID(0, "$NetBSD: ufs_bmap.c,v 1.17 2002/05/11 12:23:53 enami Exp $"); 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/buf.h> 49 #include <sys/proc.h> 50 #include <sys/vnode.h> 51 #include <sys/mount.h> 52 #include <sys/resourcevar.h> 53 #include <sys/trace.h> 54 55 #include <miscfs/specfs/specdev.h> 56 57 #include <ufs/ufs/inode.h> 58 #include <ufs/ufs/ufsmount.h> 59 #include <ufs/ufs/ufs_extern.h> 60 #include <ufs/ufs/ufs_bswap.h> 61 62 /* 63 * Bmap converts a the logical block number of a file to its physical block 64 * number on the disk. The conversion is done by using the logical block 65 * number to index into the array of block pointers described by the dinode. 66 */ 67 int 68 ufs_bmap(v) 69 void *v; 70 { 71 struct vop_bmap_args /* { 72 struct vnode *a_vp; 73 daddr_t a_bn; 74 struct vnode **a_vpp; 75 daddr_t *a_bnp; 76 int *a_runp; 77 } */ *ap = v; 78 /* 79 * Check for underlying vnode requests and ensure that logical 80 * to physical mapping is requested. 81 */ 82 if (ap->a_vpp != NULL) 83 *ap->a_vpp = VTOI(ap->a_vp)->i_devvp; 84 if (ap->a_bnp == NULL) 85 return (0); 86 87 return (ufs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL, 88 ap->a_runp)); 89 } 90 91 /* 92 * Indirect blocks are now on the vnode for the file. They are given negative 93 * logical block numbers. Indirect blocks are addressed by the negative 94 * address of the first data block to which they point. Double indirect blocks 95 * are addressed by one less than the address of the first indirect block to 96 * which they point. Triple indirect blocks are addressed by one less than 97 * the address of the first double indirect block to which they point. 98 * 99 * ufs_bmaparray does the bmap conversion, and if requested returns the 100 * array of logical blocks which must be traversed to get to a block. 101 * Each entry contains the offset into that block that gets you to the 102 * next block and the disk address of the block (if it is assigned). 103 */ 104 105 int 106 ufs_bmaparray(vp, bn, bnp, ap, nump, runp) 107 struct vnode *vp; 108 ufs_daddr_t bn; 109 ufs_daddr_t *bnp; 110 struct indir *ap; 111 int *nump; 112 int *runp; 113 { 114 struct inode *ip; 115 struct buf *bp; 116 struct ufsmount *ump; 117 struct mount *mp; 118 struct indir a[NIADDR + 1], *xap; 119 ufs_daddr_t daddr; 120 long metalbn; 121 int error, maxrun = 0, num; 122 123 ip = VTOI(vp); 124 mp = vp->v_mount; 125 ump = VFSTOUFS(mp); 126 #ifdef DIAGNOSTIC 127 if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL)) 128 panic("ufs_bmaparray: invalid arguments"); 129 #endif 130 131 if (runp) { 132 /* 133 * XXX 134 * If MAXBSIZE is the largest transfer the disks can handle, 135 * we probably want maxrun to be 1 block less so that we 136 * don't create a block larger than the device can handle. 137 */ 138 *runp = 0; 139 maxrun = MAXBSIZE / mp->mnt_stat.f_iosize - 1; 140 } 141 142 if (bn >= 0 && bn < NDADDR) { 143 if (nump != NULL) 144 *nump = 0; 145 *bnp = blkptrtodb(ump, ufs_rw32(ip->i_ffs_db[bn], 146 UFS_MPNEEDSWAP(vp->v_mount))); 147 if (*bnp == 0) 148 *bnp = -1; 149 else if (runp) 150 for (++bn; bn < NDADDR && *runp < maxrun && 151 is_sequential(ump, 152 ufs_rw32(ip->i_ffs_db[bn - 1], 153 UFS_MPNEEDSWAP(vp->v_mount)), 154 ufs_rw32(ip->i_ffs_db[bn], 155 UFS_MPNEEDSWAP(vp->v_mount))); 156 ++bn, ++*runp); 157 return (0); 158 } 159 160 xap = ap == NULL ? a : ap; 161 if (!nump) 162 nump = # 163 if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0) 164 return (error); 165 166 num = *nump; 167 168 /* Get disk address out of indirect block array */ 169 daddr = ufs_rw32(ip->i_ffs_ib[xap->in_off], 170 UFS_MPNEEDSWAP(vp->v_mount)); 171 172 for (bp = NULL, ++xap; --num; ++xap) { 173 /* 174 * Exit the loop if there is no disk address assigned yet and 175 * the indirect block isn't in the cache, or if we were 176 * looking for an indirect block and we've found it. 177 */ 178 179 metalbn = xap->in_lbn; 180 if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn) 181 break; 182 /* 183 * If we get here, we've either got the block in the cache 184 * or we have a disk address for it, go fetch it. 185 */ 186 if (bp) 187 brelse(bp); 188 189 xap->in_exists = 1; 190 bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0); 191 if (bp == NULL) { 192 193 /* 194 * getblk() above returns NULL only iff we are 195 * pagedaemon. See the implementation of getblk 196 * for detail. 197 */ 198 199 return (ENOMEM); 200 } 201 if (bp->b_flags & (B_DONE | B_DELWRI)) { 202 trace(TR_BREADHIT, pack(vp, size), metalbn); 203 } 204 #ifdef DIAGNOSTIC 205 else if (!daddr) 206 panic("ufs_bmaparry: indirect block not in cache"); 207 #endif 208 else { 209 trace(TR_BREADMISS, pack(vp, size), metalbn); 210 bp->b_blkno = blkptrtodb(ump, daddr); 211 bp->b_flags |= B_READ; 212 VOP_STRATEGY(bp); 213 curproc->p_stats->p_ru.ru_inblock++; /* XXX */ 214 if ((error = biowait(bp)) != 0) { 215 brelse(bp); 216 return (error); 217 } 218 } 219 daddr = ufs_rw32(((ufs_daddr_t *)bp->b_data)[xap->in_off], 220 UFS_MPNEEDSWAP(mp)); 221 if (num == 1 && daddr && runp) 222 for (bn = xap->in_off + 1; 223 bn < MNINDIR(ump) && *runp < maxrun && 224 is_sequential(ump, 225 ufs_rw32(((ufs_daddr_t *)bp->b_data)[bn - 1], 226 UFS_MPNEEDSWAP(mp)), 227 ufs_rw32(((ufs_daddr_t *)bp->b_data)[bn], 228 UFS_MPNEEDSWAP(mp))); 229 ++bn, ++*runp); 230 } 231 if (bp) 232 brelse(bp); 233 234 daddr = blkptrtodb(ump, daddr); 235 *bnp = daddr == 0 ? -1 : daddr; 236 return (0); 237 } 238 239 /* 240 * Create an array of logical block number/offset pairs which represent the 241 * path of indirect blocks required to access a data block. The first "pair" 242 * contains the logical block number of the appropriate single, double or 243 * triple indirect block and the offset into the inode indirect block array. 244 * Note, the logical block number of the inode single/double/triple indirect 245 * block appears twice in the array, once with the offset into the i_ffs_ib and 246 * once with the offset into the page itself. 247 */ 248 int 249 ufs_getlbns(vp, bn, ap, nump) 250 struct vnode *vp; 251 ufs_daddr_t bn; 252 struct indir *ap; 253 int *nump; 254 { 255 long metalbn, realbn; 256 struct ufsmount *ump; 257 int64_t blockcnt; 258 int lbc; 259 int i, numlevels, off; 260 261 ump = VFSTOUFS(vp->v_mount); 262 if (nump) 263 *nump = 0; 264 numlevels = 0; 265 realbn = bn; 266 if ((long)bn < 0) 267 bn = -(long)bn; 268 KASSERT(bn >= NDADDR); 269 270 /* 271 * Determine the number of levels of indirection. After this loop 272 * is done, blockcnt indicates the number of data blocks possible 273 * at the given level of indirection, and NIADDR - i is the number 274 * of levels of indirection needed to locate the requested block. 275 */ 276 277 bn -= NDADDR; 278 for (lbc = 0, i = NIADDR;; i--, bn -= blockcnt) { 279 if (i == 0) 280 return (EFBIG); 281 282 lbc += ump->um_lognindir; 283 blockcnt = (int64_t)1 << lbc; 284 285 if (bn < blockcnt) 286 break; 287 } 288 289 /* Calculate the address of the first meta-block. */ 290 if (realbn >= 0) 291 metalbn = -(realbn - bn + NIADDR - i); 292 else 293 metalbn = -(-realbn - bn + NIADDR - i); 294 295 /* 296 * At each iteration, off is the offset into the bap array which is 297 * an array of disk addresses at the current level of indirection. 298 * The logical block number and the offset in that block are stored 299 * into the argument array. 300 */ 301 ap->in_lbn = metalbn; 302 ap->in_off = off = NIADDR - i; 303 ap->in_exists = 0; 304 ap++; 305 for (++numlevels; i <= NIADDR; i++) { 306 /* If searching for a meta-data block, quit when found. */ 307 if (metalbn == realbn) 308 break; 309 310 lbc -= ump->um_lognindir; 311 blockcnt = (int64_t)1 << lbc; 312 off = (bn >> lbc) & (MNINDIR(ump) - 1); 313 314 ++numlevels; 315 ap->in_lbn = metalbn; 316 ap->in_off = off; 317 ap->in_exists = 0; 318 ++ap; 319 320 metalbn -= -1 + (off << lbc); 321 } 322 if (nump) 323 *nump = numlevels; 324 return (0); 325 } 326