1 /* 2 * Copyright (c) 1989, 1991, 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. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)ufs_bmap.c 8.7 (Berkeley) 3/21/95 35 * $FreeBSD: src/sys/ufs/ufs/ufs_bmap.c,v 1.34.2.1 2000/03/17 10:12:14 ps Exp $ 36 */ 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/buf.h> 41 #include <sys/proc.h> 42 #include <sys/vnode.h> 43 #include <sys/mount.h> 44 #include <sys/resourcevar.h> 45 #include <sys/conf.h> 46 47 #include "quota.h" 48 #include "inode.h" 49 #include "ufsmount.h" 50 #include "ufs_extern.h" 51 #include "fs.h" 52 53 /* 54 * Bmap converts the logical block number of a file to its physical block 55 * number on the disk. The conversion is done by using the logical block 56 * number to index into the array of block pointers described by the dinode. 57 * 58 * BMAP must return the contiguous before and after run in bytes, inclusive 59 * of the returned block. 60 * 61 * ufs_bmap(struct vnode *a_vp, off_t a_loffset, 62 * off_t *a_doffsetp, int *a_runp, int *a_runb) 63 */ 64 int 65 ufs_bmap(struct vop_bmap_args *ap) 66 { 67 struct fs *fs; 68 ufs_daddr_t lbn; 69 ufs_daddr_t dbn; 70 int error; 71 72 /* 73 * Check for underlying vnode requests and ensure that logical 74 * to physical mapping is requested. 75 */ 76 if (ap->a_doffsetp == NULL) 77 return (0); 78 79 fs = VTOI(ap->a_vp)->i_fs; 80 KKASSERT(((int)ap->a_loffset & ((1 << fs->fs_bshift) - 1)) == 0); 81 lbn = ap->a_loffset >> fs->fs_bshift; 82 83 error = ufs_bmaparray(ap->a_vp, lbn, &dbn, NULL, NULL, 84 ap->a_runp, ap->a_runb); 85 86 if (error || dbn == (ufs_daddr_t)-1) { 87 *ap->a_doffsetp = NOOFFSET; 88 } else { 89 *ap->a_doffsetp = dbtodoff(fs, dbn); 90 if (ap->a_runp) 91 *ap->a_runp = (*ap->a_runp + 1) << fs->fs_bshift; 92 if (ap->a_runb) 93 *ap->a_runb = *ap->a_runb << fs->fs_bshift; 94 } 95 return (error); 96 } 97 98 /* 99 * Indirect blocks are now on the vnode for the file. They are given negative 100 * logical block numbers. Indirect blocks are addressed by the negative 101 * address of the first data block to which they point. Double indirect blocks 102 * are addressed by one less than the address of the first indirect block to 103 * which they point. Triple indirect blocks are addressed by one less than 104 * the address of the first double indirect block to which they point. 105 * 106 * ufs_bmaparray does the bmap conversion, and if requested returns the 107 * array of logical blocks which must be traversed to get to a block. 108 * Each entry contains the offset into that block that gets you to the 109 * next block and the disk address of the block (if it is assigned). 110 */ 111 int 112 ufs_bmaparray(struct vnode *vp, ufs_daddr_t bn, ufs_daddr_t *bnp, 113 struct indir *ap, int *nump, int *runp, int *runb) 114 { 115 struct inode *ip; 116 struct buf *bp; 117 struct ufsmount *ump; 118 struct mount *mp; 119 struct fs *fs; 120 struct indir a[UFS_NIADDR+1], *xap; 121 ufs_daddr_t daddr; 122 long metalbn; 123 int error, maxrun, num; 124 125 ip = VTOI(vp); 126 mp = vp->v_mount; 127 ump = VFSTOUFS(mp); 128 fs = ip->i_fs; 129 #ifdef DIAGNOSTIC 130 if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL)) 131 panic("ufs_bmaparray: invalid arguments"); 132 #endif 133 134 if (runp) { 135 *runp = 0; 136 } 137 138 if (runb) { 139 *runb = 0; 140 } 141 142 maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1; 143 144 xap = ap == NULL ? a : ap; 145 if (!nump) 146 nump = # 147 error = ufs_getlbns(vp, bn, xap, nump); 148 if (error) 149 return (error); 150 151 num = *nump; 152 if (num == 0) { 153 *bnp = blkptrtodb(ump, ip->i_db[bn]); 154 if (*bnp == 0) 155 *bnp = -1; 156 else if (runp) { 157 daddr_t bnb = bn; 158 for (++bn; bn < UFS_NDADDR && *runp < maxrun && 159 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]); 160 ++bn, ++*runp); 161 bn = bnb; 162 if (runb && (bn > 0)) { 163 for (--bn; (bn >= 0) && (*runb < maxrun) && 164 is_sequential(ump, ip->i_db[bn], 165 ip->i_db[bn+1]); 166 --bn, ++*runb); 167 } 168 } 169 return (0); 170 } 171 172 173 /* Get disk address out of indirect block array */ 174 daddr = ip->i_ib[xap->in_off]; 175 176 for (bp = NULL, ++xap; --num; ++xap) { 177 /* 178 * Exit the loop if there is no disk address assigned yet and 179 * the indirect block isn't in the cache, or if we were 180 * looking for an indirect block and we've found it. 181 */ 182 metalbn = xap->in_lbn; 183 if ((daddr == 0 && 184 !findblk(vp, dbtodoff(fs, metalbn), FINDBLK_TEST)) || 185 metalbn == bn) { 186 break; 187 } 188 /* 189 * If we get here, we've either got the block in the cache 190 * or we have a disk address for it, go fetch it. 191 */ 192 if (bp) 193 bqrelse(bp); 194 195 bp = getblk(vp, lblktodoff(fs, metalbn), 196 mp->mnt_stat.f_iosize, 0, 0); 197 if ((bp->b_flags & B_CACHE) == 0) { 198 #ifdef DIAGNOSTIC 199 if (!daddr) 200 panic("ufs_bmaparray: indirect block not in cache"); 201 #endif 202 /* 203 * cached disk addr in bio2, do I/O on bio1. It 204 * will probably hit the vfs's strategy function 205 * which will then use the cached offset in bio2. 206 */ 207 bp->b_bio1.bio_done = biodone_sync; 208 bp->b_bio1.bio_flags |= BIO_SYNC; 209 bp->b_bio2.bio_offset = fsbtodoff(fs, daddr); 210 bp->b_flags &= ~(B_INVAL|B_ERROR); 211 bp->b_cmd = BUF_CMD_READ; 212 vfs_busy_pages(bp->b_vp, bp); 213 vn_strategy(bp->b_vp, &bp->b_bio1); 214 error = biowait(&bp->b_bio1, "biord"); 215 if (error) { 216 brelse(bp); 217 return (error); 218 } 219 } 220 221 daddr = ((ufs_daddr_t *)bp->b_data)[xap->in_off]; 222 if (num == 1 && daddr && runp) { 223 for (bn = xap->in_off + 1; 224 bn < MNINDIR(ump) && *runp < maxrun && 225 is_sequential(ump, 226 ((ufs_daddr_t *)bp->b_data)[bn - 1], 227 ((ufs_daddr_t *)bp->b_data)[bn]); 228 ++bn, ++*runp); 229 bn = xap->in_off; 230 if (runb && bn) { 231 for(--bn; bn >= 0 && *runb < maxrun && 232 is_sequential(ump, ((daddr_t *)bp->b_data)[bn], 233 ((daddr_t *)bp->b_data)[bn+1]); 234 --bn, ++*runb); 235 } 236 } 237 } 238 if (bp) 239 bqrelse(bp); 240 241 daddr = blkptrtodb(ump, daddr); 242 *bnp = daddr == 0 ? -1 : daddr; 243 return (0); 244 } 245 246 /* 247 * Create an array of logical block number/offset pairs which represent the 248 * path of indirect blocks required to access a data block. The first "pair" 249 * contains the logical block number of the appropriate single, double or 250 * triple indirect block and the offset into the inode indirect block array. 251 * Note, the logical block number of the inode single/double/triple indirect 252 * block appears twice in the array, once with the offset into the i_ib and 253 * once with the offset into the page itself. 254 */ 255 int 256 ufs_getlbns(struct vnode *vp, ufs_daddr_t bn, struct indir *ap, int *nump) 257 { 258 long blockcnt, metalbn, realbn; 259 struct ufsmount *ump; 260 int i, numlevels, off; 261 int64_t qblockcnt; 262 263 ump = VFSTOUFS(vp->v_mount); 264 if (nump) 265 *nump = 0; 266 numlevels = 0; 267 realbn = bn; 268 if ((long)bn < 0) 269 bn = -(long)bn; 270 271 /* The first UFS_NDADDR blocks are direct blocks. */ 272 if (bn < UFS_NDADDR) 273 return (0); 274 275 /* 276 * Determine the number of levels of indirection. After this loop 277 * is done, blockcnt indicates the number of data blocks possible 278 * at the previous level of indirection, and UFS_NIADDR-i is the number 279 * of levels of indirection needed to locate the requested block. 280 */ 281 for (blockcnt = 1, i = UFS_NIADDR, bn -= UFS_NDADDR;; i--, bn -= blockcnt) { 282 if (i == 0) 283 return (EFBIG); 284 /* 285 * Use int64_t's here to avoid overflow for triple indirect 286 * blocks when longs have 32 bits and the block size is more 287 * than 4K. 288 */ 289 qblockcnt = (int64_t)blockcnt * MNINDIR(ump); 290 if (bn < qblockcnt) 291 break; 292 blockcnt = qblockcnt; 293 } 294 295 /* Calculate the address of the first meta-block. */ 296 if (realbn >= 0) 297 metalbn = -(realbn - bn + UFS_NIADDR - i); 298 else 299 metalbn = -(-realbn - bn + UFS_NIADDR - i); 300 301 /* 302 * At each iteration, off is the offset into the bap array which is 303 * an array of disk addresses at the current level of indirection. 304 * The logical block number and the offset in that block are stored 305 * into the argument array. 306 */ 307 ap->in_lbn = metalbn; 308 ap->in_off = off = UFS_NIADDR - i; 309 ap++; 310 for (++numlevels; i <= UFS_NIADDR; i++) { 311 /* If searching for a meta-data block, quit when found. */ 312 if (metalbn == realbn) 313 break; 314 315 off = (bn / blockcnt) % MNINDIR(ump); 316 317 ++numlevels; 318 ap->in_lbn = metalbn; 319 ap->in_off = off; 320 ++ap; 321 322 metalbn -= -1 + off * blockcnt; 323 blockcnt /= MNINDIR(ump); 324 } 325 if (nump) 326 *nump = numlevels; 327 return (0); 328 } 329