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 * 4. 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 * $FreeBSD$ 36 */ 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/bio.h> 41 #include <sys/buf.h> 42 #include <sys/proc.h> 43 #include <sys/vnode.h> 44 #include <sys/mount.h> 45 #include <sys/resourcevar.h> 46 #include <sys/stat.h> 47 48 #include <fs/ext2fs/inode.h> 49 #include <fs/ext2fs/fs.h> 50 #include <fs/ext2fs/ext2fs.h> 51 #include <fs/ext2fs/ext2_dinode.h> 52 #include <fs/ext2fs/ext2_extern.h> 53 #include <fs/ext2fs/ext2_mount.h> 54 55 static int ext4_bmapext(struct vnode *, int32_t, int64_t *, int *, int *); 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 int 63 ext2_bmap(struct vop_bmap_args *ap) 64 { 65 daddr_t blkno; 66 int error; 67 68 /* 69 * Check for underlying vnode requests and ensure that logical 70 * to physical mapping is requested. 71 */ 72 if (ap->a_bop != NULL) 73 *ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj; 74 if (ap->a_bnp == NULL) 75 return (0); 76 77 if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS) 78 error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno, 79 ap->a_runp, ap->a_runb); 80 else 81 error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno, 82 ap->a_runp, ap->a_runb); 83 *ap->a_bnp = blkno; 84 return (error); 85 } 86 87 /* 88 * This function converts the logical block number of a file to 89 * its physical block number on the disk within ext4 extents. 90 */ 91 static int 92 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb) 93 { 94 struct inode *ip; 95 struct m_ext2fs *fs; 96 struct ext4_extent *ep; 97 struct ext4_extent_path path = { .ep_bp = NULL }; 98 daddr_t lbn; 99 int ret = 0; 100 101 ip = VTOI(vp); 102 fs = ip->i_e2fs; 103 lbn = bn; 104 105 if (runp != NULL) 106 *runp = 0; 107 108 if (runb != NULL) 109 *runb = 0; 110 111 ext4_ext_find_extent(fs, ip, lbn, &path); 112 if (path.ep_is_sparse) { 113 *bnp = -1; 114 if (runp != NULL) 115 *runp = path.ep_sparse_ext.e_len - 116 (lbn - path.ep_sparse_ext.e_blk) - 1; 117 } else { 118 ep = path.ep_ext; 119 if (ep == NULL) 120 ret = EIO; 121 else { 122 *bnp = fsbtodb(fs, lbn - ep->e_blk + 123 (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32)); 124 125 if (*bnp == 0) 126 *bnp = -1; 127 128 if (runp != NULL) 129 *runp = ep->e_len - (lbn - ep->e_blk) - 1; 130 } 131 } 132 133 if (path.ep_bp != NULL) { 134 brelse(path.ep_bp); 135 path.ep_bp = NULL; 136 } 137 138 return (ret); 139 } 140 141 /* 142 * Indirect blocks are now on the vnode for the file. They are given negative 143 * logical block numbers. Indirect blocks are addressed by the negative 144 * address of the first data block to which they point. Double indirect blocks 145 * are addressed by one less than the address of the first indirect block to 146 * which they point. Triple indirect blocks are addressed by one less than 147 * the address of the first double indirect block to which they point. 148 * 149 * ext2_bmaparray does the bmap conversion, and if requested returns the 150 * array of logical blocks which must be traversed to get to a block. 151 * Each entry contains the offset into that block that gets you to the 152 * next block and the disk address of the block (if it is assigned). 153 */ 154 155 int 156 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb) 157 { 158 struct inode *ip; 159 struct buf *bp; 160 struct ext2mount *ump; 161 struct mount *mp; 162 struct indir a[NIADDR+1], *ap; 163 daddr_t daddr; 164 e2fs_lbn_t metalbn; 165 int error, num, maxrun = 0, bsize; 166 int *nump; 167 168 ap = NULL; 169 ip = VTOI(vp); 170 mp = vp->v_mount; 171 ump = VFSTOEXT2(mp); 172 173 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs); 174 175 if (runp) { 176 maxrun = mp->mnt_iosize_max / bsize - 1; 177 *runp = 0; 178 } 179 180 if (runb) { 181 *runb = 0; 182 } 183 184 185 ap = a; 186 nump = # 187 error = ext2_getlbns(vp, bn, ap, nump); 188 if (error) 189 return (error); 190 191 num = *nump; 192 if (num == 0) { 193 *bnp = blkptrtodb(ump, ip->i_db[bn]); 194 if (*bnp == 0) { 195 *bnp = -1; 196 } else if (runp) { 197 daddr_t bnb = bn; 198 for (++bn; bn < NDADDR && *runp < maxrun && 199 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]); 200 ++bn, ++*runp); 201 bn = bnb; 202 if (runb && (bn > 0)) { 203 for (--bn; (bn >= 0) && (*runb < maxrun) && 204 is_sequential(ump, ip->i_db[bn], 205 ip->i_db[bn + 1]); 206 --bn, ++*runb); 207 } 208 } 209 return (0); 210 } 211 212 213 /* Get disk address out of indirect block array */ 214 daddr = ip->i_ib[ap->in_off]; 215 216 for (bp = NULL, ++ap; --num; ++ap) { 217 /* 218 * Exit the loop if there is no disk address assigned yet and 219 * the indirect block isn't in the cache, or if we were 220 * looking for an indirect block and we've found it. 221 */ 222 223 metalbn = ap->in_lbn; 224 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 225 break; 226 /* 227 * If we get here, we've either got the block in the cache 228 * or we have a disk address for it, go fetch it. 229 */ 230 if (bp) 231 bqrelse(bp); 232 233 bp = getblk(vp, metalbn, bsize, 0, 0, 0); 234 if ((bp->b_flags & B_CACHE) == 0) { 235 #ifdef INVARIANTS 236 if (!daddr) 237 panic("ext2_bmaparray: indirect block not in cache"); 238 #endif 239 bp->b_blkno = blkptrtodb(ump, daddr); 240 bp->b_iocmd = BIO_READ; 241 bp->b_flags &= ~B_INVAL; 242 bp->b_ioflags &= ~BIO_ERROR; 243 vfs_busy_pages(bp, 0); 244 bp->b_iooffset = dbtob(bp->b_blkno); 245 bstrategy(bp); 246 curthread->td_ru.ru_inblock++; 247 error = bufwait(bp); 248 if (error) { 249 brelse(bp); 250 return (error); 251 } 252 } 253 254 daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off]; 255 if (num == 1 && daddr && runp) { 256 for (bn = ap->in_off + 1; 257 bn < MNINDIR(ump) && *runp < maxrun && 258 is_sequential(ump, 259 ((e2fs_daddr_t *)bp->b_data)[bn - 1], 260 ((e2fs_daddr_t *)bp->b_data)[bn]); 261 ++bn, ++*runp); 262 bn = ap->in_off; 263 if (runb && bn) { 264 for (--bn; bn >= 0 && *runb < maxrun && 265 is_sequential(ump, 266 ((e2fs_daddr_t *)bp->b_data)[bn], 267 ((e2fs_daddr_t *)bp->b_data)[bn + 1]); 268 --bn, ++*runb); 269 } 270 } 271 } 272 if (bp) 273 bqrelse(bp); 274 275 /* 276 * Since this is FFS independent code, we are out of scope for the 277 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 278 * will fall in the range 1..um_seqinc, so we use that test and 279 * return a request for a zeroed out buffer if attempts are made 280 * to read a BLK_NOCOPY or BLK_SNAP block. 281 */ 282 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){ 283 *bnp = -1; 284 return (0); 285 } 286 *bnp = blkptrtodb(ump, daddr); 287 if (*bnp == 0) { 288 *bnp = -1; 289 } 290 return (0); 291 } 292 293 /* 294 * Create an array of logical block number/offset pairs which represent the 295 * path of indirect blocks required to access a data block. The first "pair" 296 * contains the logical block number of the appropriate single, double or 297 * triple indirect block and the offset into the inode indirect block array. 298 * Note, the logical block number of the inode single/double/triple indirect 299 * block appears twice in the array, once with the offset into the i_ib and 300 * once with the offset into the page itself. 301 */ 302 int 303 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump) 304 { 305 long blockcnt; 306 e2fs_lbn_t metalbn, realbn; 307 struct ext2mount *ump; 308 int i, numlevels, off; 309 int64_t qblockcnt; 310 311 ump = VFSTOEXT2(vp->v_mount); 312 if (nump) 313 *nump = 0; 314 numlevels = 0; 315 realbn = bn; 316 if ((long)bn < 0) 317 bn = -(long)bn; 318 319 /* The first NDADDR blocks are direct blocks. */ 320 if (bn < NDADDR) 321 return (0); 322 323 /* 324 * Determine the number of levels of indirection. After this loop 325 * is done, blockcnt indicates the number of data blocks possible 326 * at the previous level of indirection, and NIADDR - i is the number 327 * of levels of indirection needed to locate the requested block. 328 */ 329 for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) { 330 if (i == 0) 331 return (EFBIG); 332 /* 333 * Use int64_t's here to avoid overflow for triple indirect 334 * blocks when longs have 32 bits and the block size is more 335 * than 4K. 336 */ 337 qblockcnt = (int64_t)blockcnt * MNINDIR(ump); 338 if (bn < qblockcnt) 339 break; 340 blockcnt = qblockcnt; 341 } 342 343 /* Calculate the address of the first meta-block. */ 344 if (realbn >= 0) 345 metalbn = -(realbn - bn + NIADDR - i); 346 else 347 metalbn = -(-realbn - bn + NIADDR - i); 348 349 /* 350 * At each iteration, off is the offset into the bap array which is 351 * an array of disk addresses at the current level of indirection. 352 * The logical block number and the offset in that block are stored 353 * into the argument array. 354 */ 355 ap->in_lbn = metalbn; 356 ap->in_off = off = NIADDR - i; 357 ap++; 358 for (++numlevels; i <= NIADDR; i++) { 359 /* If searching for a meta-data block, quit when found. */ 360 if (metalbn == realbn) 361 break; 362 363 off = (bn / blockcnt) % MNINDIR(ump); 364 365 ++numlevels; 366 ap->in_lbn = metalbn; 367 ap->in_off = off; 368 ++ap; 369 370 metalbn -= -1 + off * blockcnt; 371 blockcnt /= MNINDIR(ump); 372 } 373 if (nump) 374 *nump = numlevels; 375 return (0); 376 } 377