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 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/bio.h> 40 #include <sys/buf.h> 41 #include <sys/proc.h> 42 #include <sys/rwlock.h> 43 #include <sys/vnode.h> 44 #include <sys/mount.h> 45 #include <sys/racct.h> 46 #include <sys/resourcevar.h> 47 #include <sys/stat.h> 48 49 #include <vm/vm.h> 50 #include <vm/vm_object.h> 51 52 #include <ufs/ufs/extattr.h> 53 #include <ufs/ufs/quota.h> 54 #include <ufs/ufs/inode.h> 55 #include <ufs/ufs/ufsmount.h> 56 #include <ufs/ufs/ufs_extern.h> 57 58 static ufs_lbn_t lbn_count(struct ufsmount *, int); 59 static int readindir(struct vnode *, ufs_lbn_t, ufs2_daddr_t, struct buf **); 60 61 /* 62 * Bmap converts the logical block number of a file to its physical block 63 * number on the disk. The conversion is done by using the logical block 64 * number to index into the array of block pointers described by the dinode. 65 */ 66 int 67 ufs_bmap( 68 struct vop_bmap_args /* { 69 struct vnode *a_vp; 70 daddr_t a_bn; 71 struct bufobj **a_bop; 72 daddr_t *a_bnp; 73 int *a_runp; 74 int *a_runb; 75 } */ *ap) 76 { 77 ufs2_daddr_t blkno; 78 int error; 79 80 /* 81 * Check for underlying vnode requests and ensure that logical 82 * to physical mapping is requested. 83 */ 84 if (ap->a_bop != NULL) 85 *ap->a_bop = &VFSTOUFS(ap->a_vp->v_mount)->um_devvp->v_bufobj; 86 if (ap->a_bnp == NULL) 87 return (0); 88 89 error = ufs_bmaparray(ap->a_vp, ap->a_bn, &blkno, NULL, 90 ap->a_runp, ap->a_runb); 91 *ap->a_bnp = blkno; 92 return (error); 93 } 94 95 static int 96 readindir(struct vnode *vp, 97 ufs_lbn_t lbn, 98 ufs2_daddr_t daddr, 99 struct buf **bpp) 100 { 101 struct buf *bp; 102 struct mount *mp; 103 struct ufsmount *ump; 104 int error; 105 106 mp = vp->v_mount; 107 ump = VFSTOUFS(mp); 108 109 bp = getblk(vp, lbn, mp->mnt_stat.f_iosize, 0, 0, 0); 110 if ((bp->b_flags & B_CACHE) == 0) { 111 KASSERT(daddr != 0, 112 ("readindir: indirect block not in cache")); 113 114 bp->b_blkno = blkptrtodb(ump, daddr); 115 bp->b_iocmd = BIO_READ; 116 bp->b_flags &= ~B_INVAL; 117 bp->b_ioflags &= ~BIO_ERROR; 118 vfs_busy_pages(bp, 0); 119 bp->b_iooffset = dbtob(bp->b_blkno); 120 bstrategy(bp); 121 #ifdef RACCT 122 if (racct_enable) { 123 PROC_LOCK(curproc); 124 racct_add_buf(curproc, bp, 0); 125 PROC_UNLOCK(curproc); 126 } 127 #endif 128 curthread->td_ru.ru_inblock++; 129 error = bufwait(bp); 130 if (error != 0) { 131 brelse(bp); 132 return (error); 133 } 134 } 135 *bpp = bp; 136 return (0); 137 } 138 139 /* 140 * Indirect blocks are now on the vnode for the file. They are given negative 141 * logical block numbers. Indirect blocks are addressed by the negative 142 * address of the first data block to which they point. Double indirect blocks 143 * are addressed by one less than the address of the first indirect block to 144 * which they point. Triple indirect blocks are addressed by one less than 145 * the address of the first double indirect block to which they point. 146 * 147 * ufs_bmaparray does the bmap conversion, and if requested returns the 148 * array of logical blocks which must be traversed to get to a block. 149 * Each entry contains the offset into that block that gets you to the 150 * next block and the disk address of the block (if it is assigned). 151 */ 152 153 int 154 ufs_bmaparray(struct vnode *vp, 155 ufs2_daddr_t bn, 156 ufs2_daddr_t *bnp, 157 struct buf *nbp, 158 int *runp, 159 int *runb) 160 { 161 struct inode *ip; 162 struct buf *bp; 163 struct ufsmount *ump; 164 struct mount *mp; 165 struct indir a[UFS_NIADDR+1], *ap; 166 ufs2_daddr_t daddr; 167 ufs_lbn_t metalbn; 168 int error, num, maxrun = 0; 169 int *nump; 170 171 ap = NULL; 172 ip = VTOI(vp); 173 mp = vp->v_mount; 174 ump = VFSTOUFS(mp); 175 176 if (runp) { 177 maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1; 178 *runp = 0; 179 } 180 181 if (runb) { 182 *runb = 0; 183 } 184 185 ap = a; 186 nump = # 187 error = ufs_getlbns(vp, bn, ap, nump); 188 if (error) 189 return (error); 190 191 num = *nump; 192 if (num == 0) { 193 if (bn >= 0 && bn < UFS_NDADDR) { 194 *bnp = blkptrtodb(ump, DIP(ip, i_db[bn])); 195 } else if (bn < 0 && bn >= -UFS_NXADDR) { 196 *bnp = blkptrtodb(ump, ip->i_din2->di_extb[-1 - bn]); 197 if (*bnp == 0) 198 *bnp = -1; 199 if (nbp == NULL) { 200 /* indirect block not found */ 201 return (EINVAL); 202 } 203 nbp->b_xflags |= BX_ALTDATA; 204 return (0); 205 } else { 206 /* blkno out of range */ 207 return (EINVAL); 208 } 209 /* 210 * Since this is FFS independent code, we are out of 211 * scope for the definitions of BLK_NOCOPY and 212 * BLK_SNAP, but we do know that they will fall in 213 * the range 1..um_seqinc, so we use that test and 214 * return a request for a zeroed out buffer if attempts 215 * are made to read a BLK_NOCOPY or BLK_SNAP block. 216 */ 217 if (IS_SNAPSHOT(ip) && DIP(ip, i_db[bn]) > 0 && 218 DIP(ip, i_db[bn]) < ump->um_seqinc) { 219 *bnp = -1; 220 } else if (*bnp == 0) { 221 *bnp = IS_SNAPSHOT(ip) ? blkptrtodb(ump, 222 bn * ump->um_seqinc) : -1; 223 } else if (runp) { 224 ufs2_daddr_t bnb = bn; 225 for (++bn; bn < UFS_NDADDR && *runp < maxrun && 226 is_sequential(ump, DIP(ip, i_db[bn - 1]), 227 DIP(ip, i_db[bn])); 228 ++bn, ++*runp); 229 bn = bnb; 230 if (runb && (bn > 0)) { 231 for (--bn; (bn >= 0) && (*runb < maxrun) && 232 is_sequential(ump, DIP(ip, i_db[bn]), 233 DIP(ip, i_db[bn+1])); 234 --bn, ++*runb); 235 } 236 } 237 return (0); 238 } 239 240 /* Get disk address out of indirect block array */ 241 daddr = DIP(ip, i_ib[ap->in_off]); 242 243 for (bp = NULL, ++ap; --num; ++ap) { 244 /* 245 * Exit the loop if there is no disk address assigned yet and 246 * the indirect block isn't in the cache, or if we were 247 * looking for an indirect block and we've found it. 248 */ 249 250 metalbn = ap->in_lbn; 251 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 252 break; 253 /* 254 * If we get here, we've either got the block in the cache 255 * or we have a disk address for it, go fetch it. 256 */ 257 if (bp) 258 bqrelse(bp); 259 error = readindir(vp, metalbn, daddr, &bp); 260 if (error != 0) 261 return (error); 262 263 if (I_IS_UFS1(ip)) 264 daddr = ((ufs1_daddr_t *)bp->b_data)[ap->in_off]; 265 else 266 daddr = ((ufs2_daddr_t *)bp->b_data)[ap->in_off]; 267 if ((error = UFS_CHECK_BLKNO(mp, ip->i_number, daddr, 268 mp->mnt_stat.f_iosize)) != 0) { 269 bqrelse(bp); 270 return (error); 271 } 272 if (I_IS_UFS1(ip)) { 273 if (num == 1 && daddr && runp) { 274 for (bn = ap->in_off + 1; 275 bn < MNINDIR(ump) && *runp < maxrun && 276 is_sequential(ump, 277 ((ufs1_daddr_t *)bp->b_data)[bn - 1], 278 ((ufs1_daddr_t *)bp->b_data)[bn]); 279 ++bn, ++*runp); 280 bn = ap->in_off; 281 if (runb && bn) { 282 for (--bn; bn >= 0 && *runb < maxrun && 283 is_sequential(ump, 284 ((ufs1_daddr_t *)bp->b_data)[bn], 285 ((ufs1_daddr_t *)bp->b_data)[bn+1]); 286 --bn, ++*runb); 287 } 288 } 289 continue; 290 } 291 if (num == 1 && daddr && runp) { 292 for (bn = ap->in_off + 1; 293 bn < MNINDIR(ump) && *runp < maxrun && 294 is_sequential(ump, 295 ((ufs2_daddr_t *)bp->b_data)[bn - 1], 296 ((ufs2_daddr_t *)bp->b_data)[bn]); 297 ++bn, ++*runp); 298 bn = ap->in_off; 299 if (runb && bn) { 300 for (--bn; bn >= 0 && *runb < maxrun && 301 is_sequential(ump, 302 ((ufs2_daddr_t *)bp->b_data)[bn], 303 ((ufs2_daddr_t *)bp->b_data)[bn + 1]); 304 --bn, ++*runb); 305 } 306 } 307 } 308 if (bp) 309 bqrelse(bp); 310 311 /* 312 * Since this is FFS independent code, we are out of scope for the 313 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 314 * will fall in the range 1..um_seqinc, so we use that test and 315 * return a request for a zeroed out buffer if attempts are made 316 * to read a BLK_NOCOPY or BLK_SNAP block. 317 */ 318 if (IS_SNAPSHOT(ip) && daddr > 0 && daddr < ump->um_seqinc){ 319 *bnp = -1; 320 return (0); 321 } 322 *bnp = blkptrtodb(ump, daddr); 323 if (*bnp == 0) { 324 if (IS_SNAPSHOT(ip)) 325 *bnp = blkptrtodb(ump, bn * ump->um_seqinc); 326 else 327 *bnp = -1; 328 } 329 return (0); 330 } 331 332 static ufs_lbn_t 333 lbn_count(struct ufsmount *ump, int level) 334 { 335 ufs_lbn_t blockcnt; 336 337 for (blockcnt = 1; level > 0; level--) 338 blockcnt *= MNINDIR(ump); 339 return (blockcnt); 340 } 341 342 int 343 ufs_bmap_seekdata(struct vnode *vp, off_t *offp) 344 { 345 struct buf *bp; 346 struct indir a[UFS_NIADDR + 1], *ap; 347 struct inode *ip; 348 struct mount *mp; 349 struct ufsmount *ump; 350 vm_object_t obj; 351 ufs2_daddr_t bn, daddr, nextbn; 352 uint64_t bsize; 353 off_t numblks; 354 int error, num, num1, off; 355 356 bp = NULL; 357 error = 0; 358 ip = VTOI(vp); 359 mp = vp->v_mount; 360 ump = VFSTOUFS(mp); 361 362 if (vp->v_type != VREG || IS_SNAPSHOT(ip)) 363 return (EINVAL); 364 if (*offp < 0 || *offp >= ip->i_size) 365 return (ENXIO); 366 367 /* 368 * We could have pages on the vnode' object queue which still 369 * do not have the data blocks allocated. Convert all dirty 370 * pages into buffer writes to ensure that we see all 371 * allocated data. 372 */ 373 obj = vp->v_object; 374 if (obj != NULL) { 375 VM_OBJECT_WLOCK(obj); 376 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC); 377 VM_OBJECT_WUNLOCK(obj); 378 } 379 380 bsize = mp->mnt_stat.f_iosize; 381 for (bn = *offp / bsize, numblks = howmany(ip->i_size, bsize); 382 bn < numblks; bn = nextbn) { 383 if (bn < UFS_NDADDR) { 384 daddr = DIP(ip, i_db[bn]); 385 if (daddr != 0) 386 break; 387 nextbn = bn + 1; 388 continue; 389 } 390 391 ap = a; 392 error = ufs_getlbns(vp, bn, ap, &num); 393 if (error != 0) 394 break; 395 MPASS(num >= 2); 396 daddr = DIP(ip, i_ib[ap->in_off]); 397 ap++, num--; 398 for (nextbn = UFS_NDADDR, num1 = num - 1; num1 > 0; num1--) 399 nextbn += lbn_count(ump, num1); 400 if (daddr == 0) { 401 nextbn += lbn_count(ump, num); 402 continue; 403 } 404 405 for (; daddr != 0 && num > 0; ap++, num--) { 406 if (bp != NULL) 407 bqrelse(bp); 408 error = readindir(vp, ap->in_lbn, daddr, &bp); 409 if (error != 0) 410 return (error); 411 412 /* 413 * Scan the indirect block until we find a non-zero 414 * pointer. 415 */ 416 off = ap->in_off; 417 do { 418 daddr = I_IS_UFS1(ip) ? 419 ((ufs1_daddr_t *)bp->b_data)[off] : 420 ((ufs2_daddr_t *)bp->b_data)[off]; 421 } while (daddr == 0 && ++off < MNINDIR(ump)); 422 nextbn += off * lbn_count(ump, num - 1); 423 424 /* 425 * We need to recompute the LBNs of indirect 426 * blocks, so restart with the updated block offset. 427 */ 428 if (off != ap->in_off) 429 break; 430 } 431 if (num == 0) { 432 /* 433 * We found a data block. 434 */ 435 bn = nextbn; 436 break; 437 } 438 } 439 if (bp != NULL) 440 bqrelse(bp); 441 if (bn >= numblks) 442 error = ENXIO; 443 if (error == 0 && *offp < bn * bsize) 444 *offp = bn * bsize; 445 return (error); 446 } 447 448 /* 449 * Create an array of logical block number/offset pairs which represent the 450 * path of indirect blocks required to access a data block. The first "pair" 451 * contains the logical block number of the appropriate single, double or 452 * triple indirect block and the offset into the inode indirect block array. 453 * Note, the logical block number of the inode single/double/triple indirect 454 * block appears twice in the array, once with the offset into the i_ib and 455 * once with the offset into the page itself. 456 */ 457 int 458 ufs_getlbns(struct vnode *vp, 459 ufs2_daddr_t bn, 460 struct indir *ap, 461 int *nump) 462 { 463 ufs2_daddr_t blockcnt; 464 ufs_lbn_t metalbn, realbn; 465 struct ufsmount *ump; 466 int i, numlevels, off; 467 468 ump = VFSTOUFS(vp->v_mount); 469 if (nump) 470 *nump = 0; 471 numlevels = 0; 472 realbn = bn; 473 if (bn < 0) 474 bn = -bn; 475 476 /* The first UFS_NDADDR blocks are direct blocks. */ 477 if (bn < UFS_NDADDR) 478 return (0); 479 480 /* 481 * Determine the number of levels of indirection. After this loop 482 * is done, blockcnt indicates the number of data blocks possible 483 * at the previous level of indirection, and UFS_NIADDR - i is the 484 * number of levels of indirection needed to locate the requested block. 485 */ 486 for (blockcnt = 1, i = UFS_NIADDR, bn -= UFS_NDADDR; ; 487 i--, bn -= blockcnt) { 488 if (i == 0) 489 return (EFBIG); 490 blockcnt *= MNINDIR(ump); 491 if (bn < blockcnt) 492 break; 493 } 494 495 /* Calculate the address of the first meta-block. */ 496 if (realbn >= 0) 497 metalbn = -(realbn - bn + UFS_NIADDR - i); 498 else 499 metalbn = -(-realbn - bn + UFS_NIADDR - i); 500 501 /* 502 * At each iteration, off is the offset into the bap array which is 503 * an array of disk addresses at the current level of indirection. 504 * The logical block number and the offset in that block are stored 505 * into the argument array. 506 */ 507 ap->in_lbn = metalbn; 508 ap->in_off = off = UFS_NIADDR - i; 509 ap++; 510 for (++numlevels; i <= UFS_NIADDR; i++) { 511 /* If searching for a meta-data block, quit when found. */ 512 if (metalbn == realbn) 513 break; 514 515 blockcnt /= MNINDIR(ump); 516 off = (bn / blockcnt) % MNINDIR(ump); 517 518 ++numlevels; 519 ap->in_lbn = metalbn; 520 ap->in_off = off; 521 ++ap; 522 523 metalbn -= -1 + off * blockcnt; 524 } 525 if (nump) 526 *nump = numlevels; 527 return (0); 528 } 529