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. 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 * @(#)ufs_bmap.c 8.7 (Berkeley) 3/21/95 39 * $FreeBSD: src/sys/ufs/ufs/ufs_bmap.c,v 1.34.2.1 2000/03/17 10:12:14 ps Exp $ 40 */ 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/buf.h> 45 #include <sys/proc.h> 46 #include <sys/vnode.h> 47 #include <sys/mount.h> 48 #include <sys/resourcevar.h> 49 #include <sys/conf.h> 50 51 #include "quota.h" 52 #include "inode.h" 53 #include "ufsmount.h" 54 #include "ufs_extern.h" 55 #include "fs.h" 56 57 /* 58 * Bmap converts the logical block number of a file to its physical block 59 * number on the disk. The conversion is done by using the logical block 60 * number to index into the array of block pointers described by the dinode. 61 * 62 * BMAP must return the contiguous before and after run in bytes, inclusive 63 * of the returned block. 64 * 65 * ufs_bmap(struct vnode *a_vp, off_t a_loffset, 66 * off_t *a_doffsetp, int *a_runp, int *a_runb) 67 */ 68 int 69 ufs_bmap(struct vop_bmap_args *ap) 70 { 71 struct fs *fs; 72 ufs_daddr_t lbn; 73 ufs_daddr_t dbn; 74 int error; 75 76 /* 77 * Check for underlying vnode requests and ensure that logical 78 * to physical mapping is requested. 79 */ 80 if (ap->a_doffsetp == NULL) 81 return (0); 82 83 fs = VTOI(ap->a_vp)->i_fs; 84 KKASSERT(((int)ap->a_loffset & ((1 << fs->fs_bshift) - 1)) == 0); 85 lbn = ap->a_loffset >> fs->fs_bshift; 86 87 error = ufs_bmaparray(ap->a_vp, lbn, &dbn, NULL, NULL, 88 ap->a_runp, ap->a_runb); 89 90 if (error || dbn == (ufs_daddr_t)-1) { 91 *ap->a_doffsetp = NOOFFSET; 92 } else { 93 *ap->a_doffsetp = dbtodoff(fs, dbn); 94 if (ap->a_runp) 95 *ap->a_runp = (*ap->a_runp + 1) << fs->fs_bshift; 96 if (ap->a_runb) 97 *ap->a_runb = *ap->a_runb << fs->fs_bshift; 98 } 99 return (error); 100 } 101 102 /* 103 * Indirect blocks are now on the vnode for the file. They are given negative 104 * logical block numbers. Indirect blocks are addressed by the negative 105 * address of the first data block to which they point. Double indirect blocks 106 * are addressed by one less than the address of the first indirect block to 107 * which they point. Triple indirect blocks are addressed by one less than 108 * the address of the first double indirect block to which they point. 109 * 110 * ufs_bmaparray does the bmap conversion, and if requested returns the 111 * array of logical blocks which must be traversed to get to a block. 112 * Each entry contains the offset into that block that gets you to the 113 * next block and the disk address of the block (if it is assigned). 114 */ 115 int 116 ufs_bmaparray(struct vnode *vp, ufs_daddr_t bn, ufs_daddr_t *bnp, 117 struct indir *ap, int *nump, int *runp, int *runb) 118 { 119 struct inode *ip; 120 struct buf *bp; 121 struct ufsmount *ump; 122 struct mount *mp; 123 struct fs *fs; 124 struct indir a[NIADDR+1], *xap; 125 ufs_daddr_t daddr; 126 long metalbn; 127 int error, maxrun, num; 128 129 ip = VTOI(vp); 130 mp = vp->v_mount; 131 ump = VFSTOUFS(mp); 132 fs = ip->i_fs; 133 #ifdef DIAGNOSTIC 134 if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL)) 135 panic("ufs_bmaparray: invalid arguments"); 136 #endif 137 138 if (runp) { 139 *runp = 0; 140 } 141 142 if (runb) { 143 *runb = 0; 144 } 145 146 maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1; 147 148 xap = ap == NULL ? a : ap; 149 if (!nump) 150 nump = # 151 error = ufs_getlbns(vp, bn, xap, nump); 152 if (error) 153 return (error); 154 155 num = *nump; 156 if (num == 0) { 157 *bnp = blkptrtodb(ump, ip->i_db[bn]); 158 if (*bnp == 0) 159 *bnp = -1; 160 else if (runp) { 161 daddr_t bnb = bn; 162 for (++bn; bn < NDADDR && *runp < maxrun && 163 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]); 164 ++bn, ++*runp); 165 bn = bnb; 166 if (runb && (bn > 0)) { 167 for (--bn; (bn >= 0) && (*runb < maxrun) && 168 is_sequential(ump, ip->i_db[bn], 169 ip->i_db[bn+1]); 170 --bn, ++*runb); 171 } 172 } 173 return (0); 174 } 175 176 177 /* Get disk address out of indirect block array */ 178 daddr = ip->i_ib[xap->in_off]; 179 180 for (bp = NULL, ++xap; --num; ++xap) { 181 /* 182 * Exit the loop if there is no disk address assigned yet and 183 * the indirect block isn't in the cache, or if we were 184 * looking for an indirect block and we've found it. 185 */ 186 metalbn = xap->in_lbn; 187 if ((daddr == 0 && 188 !findblk(vp, dbtodoff(fs, metalbn), FINDBLK_TEST)) || 189 metalbn == bn) { 190 break; 191 } 192 /* 193 * If we get here, we've either got the block in the cache 194 * or we have a disk address for it, go fetch it. 195 */ 196 if (bp) 197 bqrelse(bp); 198 199 xap->in_exists = 1; 200 bp = getblk(vp, lblktodoff(fs, metalbn), 201 mp->mnt_stat.f_iosize, 0, 0); 202 if ((bp->b_flags & B_CACHE) == 0) { 203 #ifdef DIAGNOSTIC 204 if (!daddr) 205 panic("ufs_bmaparray: indirect block not in cache"); 206 #endif 207 /* 208 * cached disk addr in bio2, do I/O on bio1. It 209 * will probably hit the vfs's strategy function 210 * which will then use the cached offset in bio2. 211 */ 212 bp->b_bio1.bio_done = biodone_sync; 213 bp->b_bio1.bio_flags |= BIO_SYNC; 214 bp->b_bio2.bio_offset = fsbtodoff(fs, daddr); 215 bp->b_flags &= ~(B_INVAL|B_ERROR); 216 bp->b_cmd = BUF_CMD_READ; 217 vfs_busy_pages(bp->b_vp, bp); 218 vn_strategy(bp->b_vp, &bp->b_bio1); 219 error = biowait(&bp->b_bio1, "biord"); 220 if (error) { 221 brelse(bp); 222 return (error); 223 } 224 } 225 226 daddr = ((ufs_daddr_t *)bp->b_data)[xap->in_off]; 227 if (num == 1 && daddr && runp) { 228 for (bn = xap->in_off + 1; 229 bn < MNINDIR(ump) && *runp < maxrun && 230 is_sequential(ump, 231 ((ufs_daddr_t *)bp->b_data)[bn - 1], 232 ((ufs_daddr_t *)bp->b_data)[bn]); 233 ++bn, ++*runp); 234 bn = xap->in_off; 235 if (runb && bn) { 236 for(--bn; bn >= 0 && *runb < maxrun && 237 is_sequential(ump, ((daddr_t *)bp->b_data)[bn], 238 ((daddr_t *)bp->b_data)[bn+1]); 239 --bn, ++*runb); 240 } 241 } 242 } 243 if (bp) 244 bqrelse(bp); 245 246 daddr = blkptrtodb(ump, daddr); 247 *bnp = daddr == 0 ? -1 : daddr; 248 return (0); 249 } 250 251 /* 252 * Create an array of logical block number/offset pairs which represent the 253 * path of indirect blocks required to access a data block. The first "pair" 254 * contains the logical block number of the appropriate single, double or 255 * triple indirect block and the offset into the inode indirect block array. 256 * Note, the logical block number of the inode single/double/triple indirect 257 * block appears twice in the array, once with the offset into the i_ib and 258 * once with the offset into the page itself. 259 */ 260 int 261 ufs_getlbns(struct vnode *vp, ufs_daddr_t bn, struct indir *ap, int *nump) 262 { 263 long blockcnt, metalbn, realbn; 264 struct ufsmount *ump; 265 int i, numlevels, off; 266 int64_t qblockcnt; 267 268 ump = VFSTOUFS(vp->v_mount); 269 if (nump) 270 *nump = 0; 271 numlevels = 0; 272 realbn = bn; 273 if ((long)bn < 0) 274 bn = -(long)bn; 275 276 /* The first NDADDR blocks are direct blocks. */ 277 if (bn < NDADDR) 278 return (0); 279 280 /* 281 * Determine the number of levels of indirection. After this loop 282 * is done, blockcnt indicates the number of data blocks possible 283 * at the previous level of indirection, and NIADDR - i is the number 284 * of levels of indirection needed to locate the requested block. 285 */ 286 for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) { 287 if (i == 0) 288 return (EFBIG); 289 /* 290 * Use int64_t's here to avoid overflow for triple indirect 291 * blocks when longs have 32 bits and the block size is more 292 * than 4K. 293 */ 294 qblockcnt = (int64_t)blockcnt * MNINDIR(ump); 295 if (bn < qblockcnt) 296 break; 297 blockcnt = qblockcnt; 298 } 299 300 /* Calculate the address of the first meta-block. */ 301 if (realbn >= 0) 302 metalbn = -(realbn - bn + NIADDR - i); 303 else 304 metalbn = -(-realbn - bn + NIADDR - i); 305 306 /* 307 * At each iteration, off is the offset into the bap array which is 308 * an array of disk addresses at the current level of indirection. 309 * The logical block number and the offset in that block are stored 310 * into the argument array. 311 */ 312 ap->in_lbn = metalbn; 313 ap->in_off = off = NIADDR - i; 314 ap->in_exists = 0; 315 ap++; 316 for (++numlevels; i <= NIADDR; i++) { 317 /* If searching for a meta-data block, quit when found. */ 318 if (metalbn == realbn) 319 break; 320 321 off = (bn / blockcnt) % MNINDIR(ump); 322 323 ++numlevels; 324 ap->in_lbn = metalbn; 325 ap->in_off = off; 326 ap->in_exists = 0; 327 ++ap; 328 329 metalbn -= -1 + off * blockcnt; 330 blockcnt /= MNINDIR(ump); 331 } 332 if (nump) 333 *nump = numlevels; 334 return (0); 335 } 336