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 */ 38 39 #include <sys/cdefs.h> 40 __FBSDID("$FreeBSD$"); 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/bio.h> 45 #include <sys/buf.h> 46 #include <sys/proc.h> 47 #include <sys/vnode.h> 48 #include <sys/mount.h> 49 #include <sys/racct.h> 50 #include <sys/resourcevar.h> 51 #include <sys/stat.h> 52 53 #include <ufs/ufs/extattr.h> 54 #include <ufs/ufs/quota.h> 55 #include <ufs/ufs/inode.h> 56 #include <ufs/ufs/ufsmount.h> 57 #include <ufs/ufs/ufs_extern.h> 58 59 static ufs_lbn_t lbn_count(struct ufsmount *, int); 60 static int readindir(struct vnode *, ufs_lbn_t, ufs2_daddr_t, struct buf **); 61 62 /* 63 * Bmap converts 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(ap) 69 struct vop_bmap_args /* { 70 struct vnode *a_vp; 71 daddr_t a_bn; 72 struct bufobj **a_bop; 73 daddr_t *a_bnp; 74 int *a_runp; 75 int *a_runb; 76 } */ *ap; 77 { 78 ufs2_daddr_t blkno; 79 int error; 80 81 /* 82 * Check for underlying vnode requests and ensure that logical 83 * to physical mapping is requested. 84 */ 85 if (ap->a_bop != NULL) 86 *ap->a_bop = &VFSTOUFS(ap->a_vp->v_mount)->um_devvp->v_bufobj; 87 if (ap->a_bnp == NULL) 88 return (0); 89 90 error = ufs_bmaparray(ap->a_vp, ap->a_bn, &blkno, NULL, 91 ap->a_runp, ap->a_runb); 92 *ap->a_bnp = blkno; 93 return (error); 94 } 95 96 static int 97 readindir(vp, lbn, daddr, bpp) 98 struct vnode *vp; 99 ufs_lbn_t lbn; 100 ufs2_daddr_t daddr; 101 struct buf **bpp; 102 { 103 struct buf *bp; 104 struct mount *mp; 105 struct ufsmount *ump; 106 int error; 107 108 mp = vp->v_mount; 109 ump = VFSTOUFS(mp); 110 111 bp = getblk(vp, lbn, mp->mnt_stat.f_iosize, 0, 0, 0); 112 if ((bp->b_flags & B_CACHE) == 0) { 113 KASSERT(daddr != 0, 114 ("readindir: indirect block not in cache")); 115 116 bp->b_blkno = blkptrtodb(ump, daddr); 117 bp->b_iocmd = BIO_READ; 118 bp->b_flags &= ~B_INVAL; 119 bp->b_ioflags &= ~BIO_ERROR; 120 vfs_busy_pages(bp, 0); 121 bp->b_iooffset = dbtob(bp->b_blkno); 122 bstrategy(bp); 123 #ifdef RACCT 124 if (racct_enable) { 125 PROC_LOCK(curproc); 126 racct_add_buf(curproc, bp, 0); 127 PROC_UNLOCK(curproc); 128 } 129 #endif 130 curthread->td_ru.ru_inblock++; 131 error = bufwait(bp); 132 if (error != 0) { 133 brelse(bp); 134 return (error); 135 } 136 } 137 *bpp = bp; 138 return (0); 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 * ufs_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 ufs_bmaparray(vp, bn, bnp, nbp, runp, runb) 157 struct vnode *vp; 158 ufs2_daddr_t bn; 159 ufs2_daddr_t *bnp; 160 struct buf *nbp; 161 int *runp; 162 int *runb; 163 { 164 struct inode *ip; 165 struct buf *bp; 166 struct ufsmount *ump; 167 struct mount *mp; 168 struct indir a[UFS_NIADDR+1], *ap; 169 ufs2_daddr_t daddr; 170 ufs_lbn_t metalbn; 171 int error, num, maxrun = 0; 172 int *nump; 173 174 ap = NULL; 175 ip = VTOI(vp); 176 mp = vp->v_mount; 177 ump = VFSTOUFS(mp); 178 179 if (runp) { 180 maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1; 181 *runp = 0; 182 } 183 184 if (runb) { 185 *runb = 0; 186 } 187 188 189 ap = a; 190 nump = # 191 error = ufs_getlbns(vp, bn, ap, nump); 192 if (error) 193 return (error); 194 195 num = *nump; 196 if (num == 0) { 197 if (bn >= 0 && bn < UFS_NDADDR) { 198 *bnp = blkptrtodb(ump, DIP(ip, i_db[bn])); 199 } else if (bn < 0 && bn >= -UFS_NXADDR) { 200 *bnp = blkptrtodb(ump, ip->i_din2->di_extb[-1 - bn]); 201 if (*bnp == 0) 202 *bnp = -1; 203 if (nbp == NULL) { 204 /* indirect block not found */ 205 return (EINVAL); 206 } 207 nbp->b_xflags |= BX_ALTDATA; 208 return (0); 209 } else { 210 /* blkno out of range */ 211 return (EINVAL); 212 } 213 /* 214 * Since this is FFS independent code, we are out of 215 * scope for the definitions of BLK_NOCOPY and 216 * BLK_SNAP, but we do know that they will fall in 217 * the range 1..um_seqinc, so we use that test and 218 * return a request for a zeroed out buffer if attempts 219 * are made to read a BLK_NOCOPY or BLK_SNAP block. 220 */ 221 if ((ip->i_flags & SF_SNAPSHOT) && DIP(ip, i_db[bn]) > 0 && 222 DIP(ip, i_db[bn]) < ump->um_seqinc) { 223 *bnp = -1; 224 } else if (*bnp == 0) { 225 if (ip->i_flags & SF_SNAPSHOT) 226 *bnp = blkptrtodb(ump, bn * ump->um_seqinc); 227 else 228 *bnp = -1; 229 } else if (runp) { 230 ufs2_daddr_t bnb = bn; 231 for (++bn; bn < UFS_NDADDR && *runp < maxrun && 232 is_sequential(ump, DIP(ip, i_db[bn - 1]), 233 DIP(ip, i_db[bn])); 234 ++bn, ++*runp); 235 bn = bnb; 236 if (runb && (bn > 0)) { 237 for (--bn; (bn >= 0) && (*runb < maxrun) && 238 is_sequential(ump, DIP(ip, i_db[bn]), 239 DIP(ip, i_db[bn+1])); 240 --bn, ++*runb); 241 } 242 } 243 return (0); 244 } 245 246 247 /* Get disk address out of indirect block array */ 248 daddr = DIP(ip, i_ib[ap->in_off]); 249 250 for (bp = NULL, ++ap; --num; ++ap) { 251 /* 252 * Exit the loop if there is no disk address assigned yet and 253 * the indirect block isn't in the cache, or if we were 254 * looking for an indirect block and we've found it. 255 */ 256 257 metalbn = ap->in_lbn; 258 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 259 break; 260 /* 261 * If we get here, we've either got the block in the cache 262 * or we have a disk address for it, go fetch it. 263 */ 264 if (bp) 265 bqrelse(bp); 266 error = readindir(vp, metalbn, daddr, &bp); 267 if (error != 0) 268 return (error); 269 270 if (I_IS_UFS1(ip)) 271 daddr = ((ufs1_daddr_t *)bp->b_data)[ap->in_off]; 272 else 273 daddr = ((ufs2_daddr_t *)bp->b_data)[ap->in_off]; 274 if ((error = UFS_CHECK_BLKNO(mp, ip->i_number, daddr, 275 mp->mnt_stat.f_iosize)) != 0) { 276 bqrelse(bp); 277 return (error); 278 } 279 if (I_IS_UFS1(ip)) { 280 if (num == 1 && daddr && runp) { 281 for (bn = ap->in_off + 1; 282 bn < MNINDIR(ump) && *runp < maxrun && 283 is_sequential(ump, 284 ((ufs1_daddr_t *)bp->b_data)[bn - 1], 285 ((ufs1_daddr_t *)bp->b_data)[bn]); 286 ++bn, ++*runp); 287 bn = ap->in_off; 288 if (runb && bn) { 289 for (--bn; bn >= 0 && *runb < maxrun && 290 is_sequential(ump, 291 ((ufs1_daddr_t *)bp->b_data)[bn], 292 ((ufs1_daddr_t *)bp->b_data)[bn+1]); 293 --bn, ++*runb); 294 } 295 } 296 continue; 297 } 298 if (num == 1 && daddr && runp) { 299 for (bn = ap->in_off + 1; 300 bn < MNINDIR(ump) && *runp < maxrun && 301 is_sequential(ump, 302 ((ufs2_daddr_t *)bp->b_data)[bn - 1], 303 ((ufs2_daddr_t *)bp->b_data)[bn]); 304 ++bn, ++*runp); 305 bn = ap->in_off; 306 if (runb && bn) { 307 for (--bn; bn >= 0 && *runb < maxrun && 308 is_sequential(ump, 309 ((ufs2_daddr_t *)bp->b_data)[bn], 310 ((ufs2_daddr_t *)bp->b_data)[bn + 1]); 311 --bn, ++*runb); 312 } 313 } 314 } 315 if (bp) 316 bqrelse(bp); 317 318 /* 319 * Since this is FFS independent code, we are out of scope for the 320 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 321 * will fall in the range 1..um_seqinc, so we use that test and 322 * return a request for a zeroed out buffer if attempts are made 323 * to read a BLK_NOCOPY or BLK_SNAP block. 324 */ 325 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){ 326 *bnp = -1; 327 return (0); 328 } 329 *bnp = blkptrtodb(ump, daddr); 330 if (*bnp == 0) { 331 if (ip->i_flags & SF_SNAPSHOT) 332 *bnp = blkptrtodb(ump, bn * ump->um_seqinc); 333 else 334 *bnp = -1; 335 } 336 return (0); 337 } 338 339 static ufs_lbn_t 340 lbn_count(ump, level) 341 struct ufsmount *ump; 342 int level; 343 { 344 ufs_lbn_t blockcnt; 345 346 for (blockcnt = 1; level > 0; level--) 347 blockcnt *= MNINDIR(ump); 348 return (blockcnt); 349 } 350 351 int 352 ufs_bmap_seekdata(vp, offp) 353 struct vnode *vp; 354 off_t *offp; 355 { 356 struct buf *bp; 357 struct indir a[UFS_NIADDR + 1], *ap; 358 struct inode *ip; 359 struct mount *mp; 360 struct ufsmount *ump; 361 ufs2_daddr_t bn, daddr, nextbn; 362 uint64_t bsize; 363 off_t numblks; 364 int error, num, num1, off; 365 366 bp = NULL; 367 error = 0; 368 ip = VTOI(vp); 369 mp = vp->v_mount; 370 ump = VFSTOUFS(mp); 371 372 if (vp->v_type != VREG || (ip->i_flags & SF_SNAPSHOT) != 0) 373 return (EINVAL); 374 if (*offp < 0 || *offp >= ip->i_size) 375 return (ENXIO); 376 377 bsize = mp->mnt_stat.f_iosize; 378 for (bn = *offp / bsize, numblks = howmany(ip->i_size, bsize); 379 bn < numblks; bn = nextbn) { 380 if (bn < UFS_NDADDR) { 381 daddr = DIP(ip, i_db[bn]); 382 if (daddr != 0) 383 break; 384 nextbn = bn + 1; 385 continue; 386 } 387 388 ap = a; 389 error = ufs_getlbns(vp, bn, ap, &num); 390 if (error != 0) 391 break; 392 MPASS(num >= 2); 393 daddr = DIP(ip, i_ib[ap->in_off]); 394 ap++, num--; 395 for (nextbn = UFS_NDADDR, num1 = num - 1; num1 > 0; num1--) 396 nextbn += lbn_count(ump, num1); 397 if (daddr == 0) { 398 nextbn += lbn_count(ump, num); 399 continue; 400 } 401 402 for (; daddr != 0 && num > 0; ap++, num--) { 403 if (bp != NULL) 404 bqrelse(bp); 405 error = readindir(vp, ap->in_lbn, daddr, &bp); 406 if (error != 0) 407 return (error); 408 409 /* 410 * Scan the indirect block until we find a non-zero 411 * pointer. 412 */ 413 off = ap->in_off; 414 do { 415 daddr = I_IS_UFS1(ip) ? 416 ((ufs1_daddr_t *)bp->b_data)[off] : 417 ((ufs2_daddr_t *)bp->b_data)[off]; 418 } while (daddr == 0 && ++off < MNINDIR(ump)); 419 nextbn += off * lbn_count(ump, num - 1); 420 421 /* 422 * We need to recompute the LBNs of indirect 423 * blocks, so restart with the updated block offset. 424 */ 425 if (off != ap->in_off) 426 break; 427 } 428 if (num == 0) { 429 /* 430 * We found a data block. 431 */ 432 bn = nextbn; 433 break; 434 } 435 } 436 if (bp != NULL) 437 bqrelse(bp); 438 if (bn >= numblks) 439 error = ENXIO; 440 if (error == 0 && *offp < bn * bsize) 441 *offp = bn * bsize; 442 return (error); 443 } 444 445 /* 446 * Create an array of logical block number/offset pairs which represent the 447 * path of indirect blocks required to access a data block. The first "pair" 448 * contains the logical block number of the appropriate single, double or 449 * triple indirect block and the offset into the inode indirect block array. 450 * Note, the logical block number of the inode single/double/triple indirect 451 * block appears twice in the array, once with the offset into the i_ib and 452 * once with the offset into the page itself. 453 */ 454 int 455 ufs_getlbns(vp, bn, ap, nump) 456 struct vnode *vp; 457 ufs2_daddr_t bn; 458 struct indir *ap; 459 int *nump; 460 { 461 ufs2_daddr_t blockcnt; 462 ufs_lbn_t metalbn, realbn; 463 struct ufsmount *ump; 464 int i, numlevels, off; 465 466 ump = VFSTOUFS(vp->v_mount); 467 if (nump) 468 *nump = 0; 469 numlevels = 0; 470 realbn = bn; 471 if (bn < 0) 472 bn = -bn; 473 474 /* The first UFS_NDADDR blocks are direct blocks. */ 475 if (bn < UFS_NDADDR) 476 return (0); 477 478 /* 479 * Determine the number of levels of indirection. After this loop 480 * is done, blockcnt indicates the number of data blocks possible 481 * at the previous level of indirection, and UFS_NIADDR - i is the 482 * number of levels of indirection needed to locate the requested block. 483 */ 484 for (blockcnt = 1, i = UFS_NIADDR, bn -= UFS_NDADDR; ; 485 i--, bn -= blockcnt) { 486 if (i == 0) 487 return (EFBIG); 488 blockcnt *= MNINDIR(ump); 489 if (bn < blockcnt) 490 break; 491 } 492 493 /* Calculate the address of the first meta-block. */ 494 if (realbn >= 0) 495 metalbn = -(realbn - bn + UFS_NIADDR - i); 496 else 497 metalbn = -(-realbn - bn + UFS_NIADDR - i); 498 499 /* 500 * At each iteration, off is the offset into the bap array which is 501 * an array of disk addresses at the current level of indirection. 502 * The logical block number and the offset in that block are stored 503 * into the argument array. 504 */ 505 ap->in_lbn = metalbn; 506 ap->in_off = off = UFS_NIADDR - i; 507 ap++; 508 for (++numlevels; i <= UFS_NIADDR; i++) { 509 /* If searching for a meta-data block, quit when found. */ 510 if (metalbn == realbn) 511 break; 512 513 blockcnt /= MNINDIR(ump); 514 off = (bn / blockcnt) % MNINDIR(ump); 515 516 ++numlevels; 517 ap->in_lbn = metalbn; 518 ap->in_off = off; 519 ++ap; 520 521 metalbn -= -1 + off * blockcnt; 522 } 523 if (nump) 524 *nump = numlevels; 525 return (0); 526 } 527