1 /* $NetBSD: ext2fs_inode.c,v 1.48 2005/09/12 16:24:41 christos Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94 32 * Modified for ext2fs by Manuel Bouyer. 33 */ 34 35 /* 36 * Copyright (c) 1997 Manuel Bouyer. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. All advertising materials mentioning features or use of this software 47 * must display the following acknowledgement: 48 * This product includes software developed by Manuel Bouyer. 49 * 4. The name of the author may not be used to endorse or promote products 50 * derived from this software without specific prior written permission. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 53 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 54 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 55 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 56 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 57 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 58 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 59 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 60 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 61 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 62 * 63 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94 64 * Modified for ext2fs by Manuel Bouyer. 65 */ 66 67 #include <sys/cdefs.h> 68 __KERNEL_RCSID(0, "$NetBSD: ext2fs_inode.c,v 1.48 2005/09/12 16:24:41 christos Exp $"); 69 70 #include <sys/param.h> 71 #include <sys/systm.h> 72 #include <sys/mount.h> 73 #include <sys/proc.h> 74 #include <sys/file.h> 75 #include <sys/buf.h> 76 #include <sys/vnode.h> 77 #include <sys/kernel.h> 78 #include <sys/malloc.h> 79 #include <sys/trace.h> 80 #include <sys/resourcevar.h> 81 82 #include <ufs/ufs/inode.h> 83 #include <ufs/ufs/ufsmount.h> 84 #include <ufs/ufs/ufs_extern.h> 85 86 #include <ufs/ext2fs/ext2fs.h> 87 #include <ufs/ext2fs/ext2fs_extern.h> 88 89 extern int prtactive; 90 91 static int ext2fs_indirtrunc(struct inode *, daddr_t, daddr_t, 92 daddr_t, int, long *); 93 94 /* 95 * Get the size of an inode. 96 */ 97 u_int64_t 98 ext2fs_size(struct inode *ip) 99 { 100 u_int64_t size = ip->i_e2fs_size; 101 102 if ((ip->i_e2fs_mode & IFMT) == IFREG) 103 size |= (u_int64_t)ip->i_e2fs_dacl << 32; 104 return size; 105 } 106 107 int 108 ext2fs_setsize(struct inode *ip, u_int64_t size) 109 { 110 if ((ip->i_e2fs_mode & IFMT) == IFREG || 111 ip->i_e2fs_mode == 0) { 112 ip->i_e2fs_dacl = size >> 32; 113 if (size >= 0x80000000U) { 114 struct m_ext2fs *fs = ip->i_e2fs; 115 116 if (fs->e2fs.e2fs_rev <= E2FS_REV0) { 117 /* Linux automagically upgrades to REV1 here! */ 118 return EFBIG; 119 } 120 if (!(fs->e2fs.e2fs_features_rocompat 121 & EXT2F_ROCOMPAT_LARGEFILE)) { 122 fs->e2fs.e2fs_features_rocompat |= 123 EXT2F_ROCOMPAT_LARGEFILE; 124 fs->e2fs_fmod = 1; 125 } 126 } 127 } else if (size >= 0x80000000U) 128 return EFBIG; 129 130 ip->i_e2fs_size = size; 131 132 return 0; 133 } 134 135 /* 136 * Last reference to an inode. If necessary, write or delete it. 137 */ 138 int 139 ext2fs_inactive(void *v) 140 { 141 struct vop_inactive_args /* { 142 struct vnode *a_vp; 143 struct proc *a_p; 144 } */ *ap = v; 145 struct vnode *vp = ap->a_vp; 146 struct inode *ip = VTOI(vp); 147 struct mount *mp; 148 struct proc *p = ap->a_p; 149 struct timespec ts; 150 int error = 0; 151 152 if (prtactive && vp->v_usecount != 0) 153 vprint("ext2fs_inactive: pushing active", vp); 154 /* Get rid of inodes related to stale file handles. */ 155 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0) 156 goto out; 157 158 error = 0; 159 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 160 vn_start_write(vp, &mp, V_WAIT | V_LOWER); 161 if (ext2fs_size(ip) != 0) { 162 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL); 163 } 164 TIMEVAL_TO_TIMESPEC(&time, &ts); 165 ip->i_e2fs_dtime = ts.tv_sec; 166 ip->i_flag |= IN_CHANGE | IN_UPDATE; 167 VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode); 168 vn_finished_write(mp, V_LOWER); 169 } 170 if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) { 171 vn_start_write(vp, &mp, V_WAIT | V_LOWER); 172 VOP_UPDATE(vp, NULL, NULL, 0); 173 vn_finished_write(mp, V_LOWER); 174 } 175 out: 176 VOP_UNLOCK(vp, 0); 177 /* 178 * If we are done with the inode, reclaim it 179 * so that it can be reused immediately. 180 */ 181 if (ip->i_e2fs_dtime != 0) 182 vrecycle(vp, NULL, p); 183 return (error); 184 } 185 186 187 /* 188 * Update the access, modified, and inode change times as specified by the 189 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 190 * used to specify that the inode needs to be updated but that the times have 191 * already been set. The access and modified times are taken from the second 192 * and third parameters; the inode change time is always taken from the current 193 * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk 194 * write of the inode to complete. 195 */ 196 int 197 ext2fs_update(void *v) 198 { 199 struct vop_update_args /* { 200 struct vnode *a_vp; 201 struct timespec *a_access; 202 struct timespec *a_modify; 203 int a_flags; 204 } */ *ap = v; 205 struct m_ext2fs *fs; 206 struct buf *bp; 207 struct inode *ip; 208 int error; 209 caddr_t cp; 210 int flags; 211 212 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) 213 return (0); 214 ip = VTOI(ap->a_vp); 215 EXT2FS_ITIMES(ip, ap->a_access, ap->a_modify, NULL); 216 if (ap->a_flags & UPDATE_CLOSE) 217 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED); 218 else 219 flags = ip->i_flag & IN_MODIFIED; 220 if (flags == 0) 221 return (0); 222 fs = ip->i_e2fs; 223 224 error = bread(ip->i_devvp, 225 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 226 (int)fs->e2fs_bsize, NOCRED, &bp); 227 if (error) { 228 brelse(bp); 229 return (error); 230 } 231 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED); 232 cp = (caddr_t)bp->b_data + 233 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE); 234 e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp); 235 if ((ap->a_flags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 && 236 (flags & IN_MODIFIED) != 0 && 237 (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 238 return (bwrite(bp)); 239 else { 240 bdwrite(bp); 241 return (0); 242 } 243 } 244 245 #define SINGLE 0 /* index of single indirect block */ 246 #define DOUBLE 1 /* index of double indirect block */ 247 #define TRIPLE 2 /* index of triple indirect block */ 248 /* 249 * Truncate the inode oip to at most length size, freeing the 250 * disk blocks. 251 */ 252 int 253 ext2fs_truncate(void *v) 254 { 255 struct vop_truncate_args /* { 256 struct vnode *a_vp; 257 off_t a_length; 258 int a_flags; 259 struct ucred *a_cred; 260 struct proc *a_p; 261 } */ *ap = v; 262 struct vnode *ovp = ap->a_vp; 263 daddr_t lastblock; 264 struct inode *oip = VTOI(ovp); 265 daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR]; 266 /* XXX ondisk32 */ 267 int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 268 off_t length = ap->a_length; 269 struct m_ext2fs *fs; 270 int offset, size, level; 271 long count, blocksreleased = 0; 272 int i, ioflag, nblocks; 273 int error, allerror = 0; 274 off_t osize; 275 int sync; 276 struct ufsmount *ump = oip->i_ump; 277 278 if (length < 0) 279 return (EINVAL); 280 281 if (ovp->v_type == VLNK && 282 (ext2fs_size(oip) < ump->um_maxsymlinklen || 283 (ump->um_maxsymlinklen == 0 && oip->i_e2fs_nblock == 0))) { 284 KDASSERT(length == 0); 285 memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0, 286 (u_int)ext2fs_size(oip)); 287 (void)ext2fs_setsize(oip, 0); 288 oip->i_flag |= IN_CHANGE | IN_UPDATE; 289 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 290 } 291 if (ext2fs_size(oip) == length) { 292 oip->i_flag |= IN_CHANGE | IN_UPDATE; 293 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 294 } 295 fs = oip->i_e2fs; 296 if (length > ump->um_maxfilesize) 297 return (EFBIG); 298 299 osize = ext2fs_size(oip); 300 ioflag = ap->a_flags; 301 302 /* 303 * Lengthen the size of the file. We must ensure that the 304 * last byte of the file is allocated. Since the smallest 305 * value of osize is 0, length will be at least 1. 306 */ 307 if (osize < length) { 308 error = ufs_balloc_range(ovp, length - 1, 1, ap->a_cred, 309 ioflag & IO_SYNC ? B_SYNC : 0); 310 if (error) { 311 (void) VOP_TRUNCATE(ovp, osize, ioflag & IO_SYNC, 312 ap->a_cred, ap->a_p); 313 return (error); 314 } 315 uvm_vnp_setsize(ovp, length); 316 oip->i_flag |= IN_CHANGE | IN_UPDATE; 317 KASSERT(error || ovp->v_size == ext2fs_size(oip)); 318 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 319 } 320 /* 321 * Shorten the size of the file. If the file is not being 322 * truncated to a block boundry, the contents of the 323 * partial block following the end of the file must be 324 * zero'ed in case it ever become accessible again because 325 * of subsequent file growth. 326 */ 327 offset = blkoff(fs, length); 328 if (offset != 0) { 329 size = fs->e2fs_bsize; 330 331 /* XXXUBC we should handle more than just VREG */ 332 uvm_vnp_zerorange(ovp, length, size - offset); 333 } 334 (void)ext2fs_setsize(oip, length); 335 uvm_vnp_setsize(ovp, length); 336 /* 337 * Calculate index into inode's block list of 338 * last direct and indirect blocks (if any) 339 * which we want to keep. Lastblock is -1 when 340 * the file is truncated to 0. 341 */ 342 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 343 lastiblock[SINGLE] = lastblock - NDADDR; 344 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 345 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 346 nblocks = btodb(fs->e2fs_bsize); 347 /* 348 * Update file and block pointers on disk before we start freeing 349 * blocks. If we crash before free'ing blocks below, the blocks 350 * will be returned to the free list. lastiblock values are also 351 * normalized to -1 for calls to ext2fs_indirtrunc below. 352 */ 353 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks); 354 sync = 0; 355 for (level = TRIPLE; level >= SINGLE; level--) { 356 if (lastiblock[level] < 0 && oldblks[NDADDR + level] != 0) { 357 sync = 1; 358 oip->i_e2fs_blocks[NDADDR + level] = 0; 359 lastiblock[level] = -1; 360 } 361 } 362 for (i = 0; i < NDADDR; i++) { 363 if (i > lastblock && oldblks[i] != 0) { 364 sync = 1; 365 oip->i_e2fs_blocks[i] = 0; 366 } 367 } 368 oip->i_flag |= IN_CHANGE | IN_UPDATE; 369 if (sync) { 370 error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT); 371 if (error && !allerror) 372 allerror = error; 373 } 374 375 /* 376 * Having written the new inode to disk, save its new configuration 377 * and put back the old block pointers long enough to process them. 378 * Note that we save the new block configuration so we can check it 379 * when we are done. 380 */ 381 memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks); 382 memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks); 383 384 (void)ext2fs_setsize(oip, osize); 385 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 386 if (error && !allerror) 387 allerror = error; 388 389 /* 390 * Indirect blocks first. 391 */ 392 indir_lbn[SINGLE] = -NDADDR; 393 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 394 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 395 for (level = TRIPLE; level >= SINGLE; level--) { 396 /* XXX ondisk32 */ 397 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 398 if (bn != 0) { 399 error = ext2fs_indirtrunc(oip, indir_lbn[level], 400 fsbtodb(fs, bn), lastiblock[level], level, &count); 401 if (error) 402 allerror = error; 403 blocksreleased += count; 404 if (lastiblock[level] < 0) { 405 oip->i_e2fs_blocks[NDADDR + level] = 0; 406 ext2fs_blkfree(oip, bn); 407 blocksreleased += nblocks; 408 } 409 } 410 if (lastiblock[level] >= 0) 411 goto done; 412 } 413 414 /* 415 * All whole direct blocks or frags. 416 */ 417 for (i = NDADDR - 1; i > lastblock; i--) { 418 /* XXX ondisk32 */ 419 bn = fs2h32(oip->i_e2fs_blocks[i]); 420 if (bn == 0) 421 continue; 422 oip->i_e2fs_blocks[i] = 0; 423 ext2fs_blkfree(oip, bn); 424 blocksreleased += btodb(fs->e2fs_bsize); 425 } 426 427 done: 428 #ifdef DIAGNOSTIC 429 for (level = SINGLE; level <= TRIPLE; level++) 430 if (newblks[NDADDR + level] != 431 oip->i_e2fs_blocks[NDADDR + level]) 432 panic("ext2fs_truncate1"); 433 for (i = 0; i < NDADDR; i++) 434 if (newblks[i] != oip->i_e2fs_blocks[i]) 435 panic("ext2fs_truncate2"); 436 if (length == 0 && 437 (!LIST_EMPTY(&ovp->v_cleanblkhd) || 438 !LIST_EMPTY(&ovp->v_dirtyblkhd))) 439 panic("ext2fs_truncate3"); 440 #endif /* DIAGNOSTIC */ 441 /* 442 * Put back the real size. 443 */ 444 (void)ext2fs_setsize(oip, length); 445 oip->i_e2fs_nblock -= blocksreleased; 446 oip->i_flag |= IN_CHANGE; 447 KASSERT(ovp->v_type != VREG || ovp->v_size == ext2fs_size(oip)); 448 return (allerror); 449 } 450 451 /* 452 * Release blocks associated with the inode ip and stored in the indirect 453 * block bn. Blocks are free'd in LIFO order up to (but not including) 454 * lastbn. If level is greater than SINGLE, the block is an indirect block 455 * and recursive calls to indirtrunc must be used to cleanse other indirect 456 * blocks. 457 * 458 * NB: triple indirect blocks are untested. 459 */ 460 static int 461 ext2fs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn, 462 int level, long *countp) 463 { 464 int i; 465 struct buf *bp; 466 struct m_ext2fs *fs = ip->i_e2fs; 467 int32_t *bap; /* XXX ondisk32 */ 468 struct vnode *vp; 469 daddr_t nb, nlbn, last; 470 int32_t *copy = NULL; /* XXX ondisk32 */ 471 long blkcount, factor; 472 int nblocks, blocksreleased = 0; 473 int error = 0, allerror = 0; 474 475 /* 476 * Calculate index in current block of last 477 * block to be kept. -1 indicates the entire 478 * block so we need not calculate the index. 479 */ 480 factor = 1; 481 for (i = SINGLE; i < level; i++) 482 factor *= NINDIR(fs); 483 last = lastbn; 484 if (lastbn > 0) 485 last /= factor; 486 nblocks = btodb(fs->e2fs_bsize); 487 /* 488 * Get buffer of block pointers, zero those entries corresponding 489 * to blocks to be free'd, and update on disk copy first. Since 490 * double(triple) indirect before single(double) indirect, calls 491 * to bmap on these blocks will fail. However, we already have 492 * the on disk address, so we have to set the b_blkno field 493 * explicitly instead of letting bread do everything for us. 494 */ 495 vp = ITOV(ip); 496 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 497 if (bp->b_flags & (B_DONE | B_DELWRI)) { 498 /* Braces must be here in case trace evaluates to nothing. */ 499 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 500 } else { 501 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 502 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 503 bp->b_flags |= B_READ; 504 if (bp->b_bcount > bp->b_bufsize) 505 panic("ext2fs_indirtrunc: bad buffer size"); 506 bp->b_blkno = dbn; 507 VOP_STRATEGY(vp, bp); 508 error = biowait(bp); 509 } 510 if (error) { 511 brelse(bp); 512 *countp = 0; 513 return (error); 514 } 515 516 bap = (int32_t *)bp->b_data; /* XXX ondisk32 */ 517 if (lastbn >= 0) { 518 /* XXX ondisk32 */ 519 MALLOC(copy, int32_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK); 520 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize); 521 memset((caddr_t)&bap[last + 1], 0, 522 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t)); 523 error = bwrite(bp); 524 if (error) 525 allerror = error; 526 bap = copy; 527 } 528 529 /* 530 * Recursively free totally unused blocks. 531 */ 532 for (i = NINDIR(fs) - 1, 533 nlbn = lbn + 1 - i * factor; i > last; 534 i--, nlbn += factor) { 535 /* XXX ondisk32 */ 536 nb = fs2h32(bap[i]); 537 if (nb == 0) 538 continue; 539 if (level > SINGLE) { 540 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 541 (daddr_t)-1, level - 1, 542 &blkcount); 543 if (error) 544 allerror = error; 545 blocksreleased += blkcount; 546 } 547 ext2fs_blkfree(ip, nb); 548 blocksreleased += nblocks; 549 } 550 551 /* 552 * Recursively free last partial block. 553 */ 554 if (level > SINGLE && lastbn >= 0) { 555 last = lastbn % factor; 556 /* XXX ondisk32 */ 557 nb = fs2h32(bap[i]); 558 if (nb != 0) { 559 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 560 last, level - 1, &blkcount); 561 if (error) 562 allerror = error; 563 blocksreleased += blkcount; 564 } 565 } 566 567 if (copy != NULL) { 568 FREE(copy, M_TEMP); 569 } else { 570 bp->b_flags |= B_INVAL; 571 brelse(bp); 572 } 573 574 *countp = blocksreleased; 575 return (allerror); 576 } 577