1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 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. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)ufs_bmap.c 8.7 (Berkeley) 3/21/95 37 * $FreeBSD$ 38 */ 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/bio.h> 43 #include <sys/buf.h> 44 #include <sys/proc.h> 45 #include <sys/vnode.h> 46 #include <sys/mount.h> 47 #include <sys/racct.h> 48 #include <sys/resourcevar.h> 49 #include <sys/stat.h> 50 51 #include <fs/ext2fs/fs.h> 52 #include <fs/ext2fs/inode.h> 53 #include <fs/ext2fs/ext2fs.h> 54 #include <fs/ext2fs/ext2_dinode.h> 55 #include <fs/ext2fs/ext2_extern.h> 56 #include <fs/ext2fs/ext2_mount.h> 57 58 /* 59 * Bmap converts the logical block number of a file to its physical block 60 * number on the disk. The conversion is done by using the logical block 61 * number to index into the array of block pointers described by the dinode. 62 */ 63 int 64 ext2_bmap(struct vop_bmap_args *ap) 65 { 66 daddr_t blkno; 67 int error; 68 69 /* 70 * Check for underlying vnode requests and ensure that logical 71 * to physical mapping is requested. 72 */ 73 if (ap->a_bop != NULL) 74 *ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj; 75 if (ap->a_bnp == NULL) 76 return (0); 77 78 if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS) 79 error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno, 80 ap->a_runp, ap->a_runb); 81 else 82 error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno, 83 ap->a_runp, ap->a_runb); 84 *ap->a_bnp = blkno; 85 return (error); 86 } 87 88 /* 89 * Convert the logical block number of a file to its physical block number 90 * on the disk within ext4 extents. 91 */ 92 int 93 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb) 94 { 95 struct inode *ip; 96 struct m_ext2fs *fs; 97 struct ext4_extent_header *ehp; 98 struct ext4_extent *ep; 99 struct ext4_extent_path *path = NULL; 100 daddr_t lbn; 101 int error, depth; 102 103 ip = VTOI(vp); 104 fs = ip->i_e2fs; 105 lbn = bn; 106 ehp = (struct ext4_extent_header *)ip->i_data; 107 depth = ehp->eh_depth; 108 109 *bnp = -1; 110 if (runp != NULL) 111 *runp = 0; 112 if (runb != NULL) 113 *runb = 0; 114 115 error = ext4_ext_find_extent(ip, lbn, &path); 116 if (error) 117 return (error); 118 119 ep = path[depth].ep_ext; 120 if(ep) { 121 if (lbn < ep->e_blk) { 122 if (runp != NULL) 123 *runp = ep->e_blk - lbn - 1; 124 } else if (ep->e_blk <= lbn && lbn < ep->e_blk + ep->e_len) { 125 *bnp = fsbtodb(fs, lbn - ep->e_blk + 126 (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32)); 127 if (runp != NULL) 128 *runp = ep->e_len - (lbn - ep->e_blk) - 1; 129 if (runb != NULL) 130 *runb = lbn - ep->e_blk; 131 } else { 132 if (runb != NULL) 133 *runb = ep->e_blk + lbn - ep->e_len; 134 } 135 } 136 137 ext4_ext_path_free(path); 138 139 return (error); 140 } 141 142 static int 143 readindir(struct vnode *vp, e2fs_lbn_t lbn, e2fs_daddr_t daddr, struct buf **bpp) 144 { 145 struct buf *bp; 146 struct mount *mp; 147 struct ext2mount *ump; 148 int error; 149 150 mp = vp->v_mount; 151 ump = VFSTOEXT2(mp); 152 153 bp = getblk(vp, lbn, mp->mnt_stat.f_iosize, 0, 0, 0); 154 if ((bp->b_flags & B_CACHE) == 0) { 155 KASSERT(daddr != 0, 156 ("readindir: indirect block not in cache")); 157 158 bp->b_blkno = blkptrtodb(ump, daddr); 159 bp->b_iocmd = BIO_READ; 160 bp->b_flags &= ~B_INVAL; 161 bp->b_ioflags &= ~BIO_ERROR; 162 vfs_busy_pages(bp, 0); 163 bp->b_iooffset = dbtob(bp->b_blkno); 164 bstrategy(bp); 165 #ifdef RACCT 166 if (racct_enable) { 167 PROC_LOCK(curproc); 168 racct_add_buf(curproc, bp, 0); 169 PROC_UNLOCK(curproc); 170 } 171 #endif 172 curthread->td_ru.ru_inblock++; 173 error = bufwait(bp); 174 if (error != 0) { 175 brelse(bp); 176 return (error); 177 } 178 } 179 *bpp = bp; 180 return (0); 181 } 182 183 /* 184 * Indirect blocks are now on the vnode for the file. They are given negative 185 * logical block numbers. Indirect blocks are addressed by the negative 186 * address of the first data block to which they point. Double indirect blocks 187 * are addressed by one less than the address of the first indirect block to 188 * which they point. Triple indirect blocks are addressed by one less than 189 * the address of the first double indirect block to which they point. 190 * 191 * ext2_bmaparray does the bmap conversion, and if requested returns the 192 * array of logical blocks which must be traversed to get to a block. 193 * Each entry contains the offset into that block that gets you to the 194 * next block and the disk address of the block (if it is assigned). 195 */ 196 197 int 198 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb) 199 { 200 struct inode *ip; 201 struct buf *bp; 202 struct ext2mount *ump; 203 struct mount *mp; 204 struct indir a[EXT2_NIADDR + 1], *ap; 205 daddr_t daddr; 206 e2fs_lbn_t metalbn; 207 int error, num, maxrun = 0, bsize; 208 int *nump; 209 210 ap = NULL; 211 ip = VTOI(vp); 212 mp = vp->v_mount; 213 ump = VFSTOEXT2(mp); 214 215 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs); 216 217 if (runp) { 218 maxrun = mp->mnt_iosize_max / bsize - 1; 219 *runp = 0; 220 } 221 if (runb) 222 *runb = 0; 223 224 225 ap = a; 226 nump = # 227 error = ext2_getlbns(vp, bn, ap, nump); 228 if (error) 229 return (error); 230 231 num = *nump; 232 if (num == 0) { 233 *bnp = blkptrtodb(ump, ip->i_db[bn]); 234 if (*bnp == 0) { 235 *bnp = -1; 236 } else if (runp) { 237 daddr_t bnb = bn; 238 239 for (++bn; bn < EXT2_NDADDR && *runp < maxrun && 240 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]); 241 ++bn, ++*runp); 242 bn = bnb; 243 if (runb && (bn > 0)) { 244 for (--bn; (bn >= 0) && (*runb < maxrun) && 245 is_sequential(ump, ip->i_db[bn], 246 ip->i_db[bn + 1]); 247 --bn, ++*runb); 248 } 249 } 250 return (0); 251 } 252 253 /* Get disk address out of indirect block array */ 254 daddr = ip->i_ib[ap->in_off]; 255 256 for (bp = NULL, ++ap; --num; ++ap) { 257 /* 258 * Exit the loop if there is no disk address assigned yet and 259 * the indirect block isn't in the cache, or if we were 260 * looking for an indirect block and we've found it. 261 */ 262 263 metalbn = ap->in_lbn; 264 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 265 break; 266 /* 267 * If we get here, we've either got the block in the cache 268 * or we have a disk address for it, go fetch it. 269 */ 270 if (bp) 271 bqrelse(bp); 272 error = readindir(vp, metalbn, daddr, &bp); 273 if (error != 0) 274 return (error); 275 276 daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off]; 277 if (num == 1 && daddr && runp) { 278 for (bn = ap->in_off + 1; 279 bn < MNINDIR(ump) && *runp < maxrun && 280 is_sequential(ump, 281 ((e2fs_daddr_t *)bp->b_data)[bn - 1], 282 ((e2fs_daddr_t *)bp->b_data)[bn]); 283 ++bn, ++*runp); 284 bn = ap->in_off; 285 if (runb && bn) { 286 for (--bn; bn >= 0 && *runb < maxrun && 287 is_sequential(ump, 288 ((e2fs_daddr_t *)bp->b_data)[bn], 289 ((e2fs_daddr_t *)bp->b_data)[bn + 1]); 290 --bn, ++*runb); 291 } 292 } 293 } 294 if (bp) 295 bqrelse(bp); 296 297 /* 298 * Since this is FFS independent code, we are out of scope for the 299 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 300 * will fall in the range 1..um_seqinc, so we use that test and 301 * return a request for a zeroed out buffer if attempts are made 302 * to read a BLK_NOCOPY or BLK_SNAP block. 303 */ 304 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc) { 305 *bnp = -1; 306 return (0); 307 } 308 *bnp = blkptrtodb(ump, daddr); 309 if (*bnp == 0) { 310 *bnp = -1; 311 } 312 return (0); 313 } 314 315 static e2fs_lbn_t 316 lbn_count(struct ext2mount *ump, int level) 317 318 { 319 e2fs_lbn_t blockcnt; 320 321 for (blockcnt = 1; level > 0; level--) 322 blockcnt *= MNINDIR(ump); 323 return (blockcnt); 324 } 325 326 int 327 ext2_bmap_seekdata(struct vnode *vp, off_t *offp) 328 { 329 struct buf *bp; 330 struct indir a[EXT2_NIADDR + 1], *ap; 331 struct inode *ip; 332 struct mount *mp; 333 struct ext2mount *ump; 334 e2fs_daddr_t bn, daddr, nextbn; 335 uint64_t bsize; 336 off_t numblks; 337 int error, num, num1, off; 338 339 bp = NULL; 340 error = 0; 341 ip = VTOI(vp); 342 mp = vp->v_mount; 343 ump = VFSTOEXT2(mp); 344 345 if (vp->v_type != VREG || (ip->i_flags & SF_SNAPSHOT) != 0) 346 return (EINVAL); 347 if (*offp < 0 || *offp >= ip->i_size) 348 return (ENXIO); 349 350 bsize = mp->mnt_stat.f_iosize; 351 for (bn = *offp / bsize, numblks = howmany(ip->i_size, bsize); 352 bn < numblks; bn = nextbn) { 353 if (bn < EXT2_NDADDR) { 354 daddr = ip->i_db[bn]; 355 if (daddr != 0) 356 break; 357 nextbn = bn + 1; 358 continue; 359 } 360 361 ap = a; 362 error = ext2_getlbns(vp, bn, ap, &num); 363 if (error != 0) 364 break; 365 MPASS(num >= 2); 366 daddr = ip->i_ib[ap->in_off]; 367 ap++, num--; 368 for (nextbn = EXT2_NDADDR, num1 = num - 1; num1 > 0; num1--) 369 nextbn += lbn_count(ump, num1); 370 if (daddr == 0) { 371 nextbn += lbn_count(ump, num); 372 continue; 373 } 374 375 for (; daddr != 0 && num > 0; ap++, num--) { 376 if (bp != NULL) 377 bqrelse(bp); 378 error = readindir(vp, ap->in_lbn, daddr, &bp); 379 if (error != 0) 380 return (error); 381 382 /* 383 * Scan the indirect block until we find a non-zero 384 * pointer. 385 */ 386 off = ap->in_off; 387 do { 388 daddr = ((e2fs_daddr_t *)bp->b_data)[off]; 389 } while (daddr == 0 && ++off < MNINDIR(ump)); 390 nextbn += off * lbn_count(ump, num - 1); 391 392 /* 393 * We need to recompute the LBNs of indirect 394 * blocks, so restart with the updated block offset. 395 */ 396 if (off != ap->in_off) 397 break; 398 } 399 if (num == 0) { 400 /* 401 * We found a data block. 402 */ 403 bn = nextbn; 404 break; 405 } 406 } 407 if (bp != NULL) 408 bqrelse(bp); 409 if (bn >= numblks) 410 error = ENXIO; 411 if (error == 0 && *offp < bn * bsize) 412 *offp = bn * bsize; 413 return (error); 414 } 415 416 /* 417 * Create an array of logical block number/offset pairs which represent the 418 * path of indirect blocks required to access a data block. The first "pair" 419 * contains the logical block number of the appropriate single, double or 420 * triple indirect block and the offset into the inode indirect block array. 421 * Note, the logical block number of the inode single/double/triple indirect 422 * block appears twice in the array, once with the offset into the i_ib and 423 * once with the offset into the page itself. 424 */ 425 int 426 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump) 427 { 428 long blockcnt; 429 e2fs_lbn_t metalbn, realbn; 430 struct ext2mount *ump; 431 int i, numlevels, off; 432 int64_t qblockcnt; 433 434 ump = VFSTOEXT2(vp->v_mount); 435 if (nump) 436 *nump = 0; 437 numlevels = 0; 438 realbn = bn; 439 if ((long)bn < 0) 440 bn = -(long)bn; 441 442 /* The first EXT2_NDADDR blocks are direct blocks. */ 443 if (bn < EXT2_NDADDR) 444 return (0); 445 446 /* 447 * Determine the number of levels of indirection. After this loop 448 * is done, blockcnt indicates the number of data blocks possible 449 * at the previous level of indirection, and EXT2_NIADDR - i is the 450 * number of levels of indirection needed to locate the requested block. 451 */ 452 for (blockcnt = 1, i = EXT2_NIADDR, bn -= EXT2_NDADDR; ; 453 i--, bn -= blockcnt) { 454 if (i == 0) 455 return (EFBIG); 456 /* 457 * Use int64_t's here to avoid overflow for triple indirect 458 * blocks when longs have 32 bits and the block size is more 459 * than 4K. 460 */ 461 qblockcnt = (int64_t)blockcnt * MNINDIR(ump); 462 if (bn < qblockcnt) 463 break; 464 blockcnt = qblockcnt; 465 } 466 467 /* Calculate the address of the first meta-block. */ 468 if (realbn >= 0) 469 metalbn = -(realbn - bn + EXT2_NIADDR - i); 470 else 471 metalbn = -(-realbn - bn + EXT2_NIADDR - i); 472 473 /* 474 * At each iteration, off is the offset into the bap array which is 475 * an array of disk addresses at the current level of indirection. 476 * The logical block number and the offset in that block are stored 477 * into the argument array. 478 */ 479 ap->in_lbn = metalbn; 480 ap->in_off = off = EXT2_NIADDR - i; 481 ap++; 482 for (++numlevels; i <= EXT2_NIADDR; i++) { 483 /* If searching for a meta-data block, quit when found. */ 484 if (metalbn == realbn) 485 break; 486 487 off = (bn / blockcnt) % MNINDIR(ump); 488 489 ++numlevels; 490 ap->in_lbn = metalbn; 491 ap->in_off = off; 492 ++ap; 493 494 metalbn -= -1 + off * blockcnt; 495 blockcnt /= MNINDIR(ump); 496 } 497 if (nump) 498 *nump = numlevels; 499 return (0); 500 } 501