1 /* $NetBSD: ext2fs_inode.c,v 1.73 2010/07/28 11:03:48 hannken 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 * 47 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 48 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 49 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 50 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 51 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 52 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 53 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 54 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 55 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 56 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 57 * 58 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94 59 * Modified for ext2fs by Manuel Bouyer. 60 */ 61 62 #include <sys/cdefs.h> 63 __KERNEL_RCSID(0, "$NetBSD: ext2fs_inode.c,v 1.73 2010/07/28 11:03:48 hannken Exp $"); 64 65 #include <sys/param.h> 66 #include <sys/systm.h> 67 #include <sys/mount.h> 68 #include <sys/proc.h> 69 #include <sys/file.h> 70 #include <sys/buf.h> 71 #include <sys/vnode.h> 72 #include <sys/kernel.h> 73 #include <sys/malloc.h> 74 #include <sys/trace.h> 75 #include <sys/resourcevar.h> 76 #include <sys/kauth.h> 77 78 #include <ufs/ufs/inode.h> 79 #include <ufs/ufs/ufsmount.h> 80 #include <ufs/ufs/ufs_extern.h> 81 82 #include <ufs/ext2fs/ext2fs.h> 83 #include <ufs/ext2fs/ext2fs_extern.h> 84 85 extern int prtactive; 86 87 static int ext2fs_indirtrunc(struct inode *, daddr_t, daddr_t, 88 daddr_t, int, long *); 89 90 /* 91 * Get the size of an inode. 92 */ 93 uint64_t 94 ext2fs_size(struct inode *ip) 95 { 96 uint64_t size = ip->i_e2fs_size; 97 98 if ((ip->i_e2fs_mode & IFMT) == IFREG) 99 size |= (uint64_t)ip->i_e2fs_dacl << 32; 100 return size; 101 } 102 103 int 104 ext2fs_setsize(struct inode *ip, uint64_t size) 105 { 106 if ((ip->i_e2fs_mode & IFMT) == IFREG || 107 ip->i_e2fs_mode == 0) { 108 ip->i_e2fs_dacl = size >> 32; 109 if (size >= 0x80000000U) { 110 struct m_ext2fs *fs = ip->i_e2fs; 111 112 if (fs->e2fs.e2fs_rev <= E2FS_REV0) { 113 /* Linux automagically upgrades to REV1 here! */ 114 return EFBIG; 115 } 116 if (!(fs->e2fs.e2fs_features_rocompat 117 & EXT2F_ROCOMPAT_LARGEFILE)) { 118 fs->e2fs.e2fs_features_rocompat |= 119 EXT2F_ROCOMPAT_LARGEFILE; 120 fs->e2fs_fmod = 1; 121 } 122 } 123 } else if (size >= 0x80000000U) 124 return EFBIG; 125 126 ip->i_e2fs_size = size; 127 128 return 0; 129 } 130 131 /* 132 * Last reference to an inode. If necessary, write or delete it. 133 */ 134 int 135 ext2fs_inactive(void *v) 136 { 137 struct vop_inactive_args /* { 138 struct vnode *a_vp; 139 bool *a_recycle; 140 } */ *ap = v; 141 struct vnode *vp = ap->a_vp; 142 struct inode *ip = VTOI(vp); 143 int error = 0; 144 145 if (prtactive && vp->v_usecount != 0) 146 vprint("ext2fs_inactive: pushing active", vp); 147 /* Get rid of inodes related to stale file handles. */ 148 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0) 149 goto out; 150 151 error = 0; 152 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 153 /* Defer final inode free and update to reclaim.*/ 154 if (ext2fs_size(ip) != 0) { 155 error = ext2fs_truncate(vp, (off_t)0, 0, NOCRED); 156 } 157 ip->i_e2fs_dtime = time_second; 158 ip->i_flag |= IN_CHANGE | IN_UPDATE; 159 ip->i_omode = 1; 160 } 161 if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) { 162 ext2fs_update(vp, NULL, NULL, 0); 163 } 164 out: 165 /* 166 * If we are done with the inode, reclaim it 167 * so that it can be reused immediately. 168 */ 169 *ap->a_recycle = (ip->i_e2fs_dtime != 0); 170 VOP_UNLOCK(vp); 171 return (error); 172 } 173 174 175 /* 176 * Update the access, modified, and inode change times as specified by the 177 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 178 * used to specify that the inode needs to be updated but that the times have 179 * already been set. The access and modified times are taken from the second 180 * and third parameters; the inode change time is always taken from the current 181 * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk 182 * write of the inode to complete. 183 */ 184 int 185 ext2fs_update(struct vnode *vp, const struct timespec *acc, 186 const struct timespec *mod, int updflags) 187 { 188 struct m_ext2fs *fs; 189 struct buf *bp; 190 struct inode *ip; 191 int error; 192 void *cp; 193 int flags; 194 195 if (vp->v_mount->mnt_flag & MNT_RDONLY) 196 return (0); 197 ip = VTOI(vp); 198 EXT2FS_ITIMES(ip, acc, mod, NULL); 199 if (updflags & UPDATE_CLOSE) 200 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED); 201 else 202 flags = ip->i_flag & IN_MODIFIED; 203 if (flags == 0) 204 return (0); 205 fs = ip->i_e2fs; 206 207 error = bread(ip->i_devvp, 208 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 209 (int)fs->e2fs_bsize, NOCRED, B_MODIFY, &bp); 210 if (error) { 211 brelse(bp, 0); 212 return (error); 213 } 214 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED); 215 cp = (char *)bp->b_data + 216 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs)); 217 e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp); 218 if ((updflags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 && 219 (flags & IN_MODIFIED) != 0 && 220 (vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 221 return (bwrite(bp)); 222 else { 223 bdwrite(bp); 224 return (0); 225 } 226 } 227 228 #define SINGLE 0 /* index of single indirect block */ 229 #define DOUBLE 1 /* index of double indirect block */ 230 #define TRIPLE 2 /* index of triple indirect block */ 231 /* 232 * Truncate the inode oip to at most length size, freeing the 233 * disk blocks. 234 */ 235 int 236 ext2fs_truncate(struct vnode *ovp, off_t length, int ioflag, 237 kauth_cred_t cred) 238 { 239 daddr_t lastblock; 240 struct inode *oip = VTOI(ovp); 241 daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR]; 242 /* XXX ondisk32 */ 243 int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 244 struct m_ext2fs *fs; 245 int offset, size, level; 246 long count, blocksreleased = 0; 247 int i, nblocks; 248 int error, allerror = 0; 249 off_t osize; 250 int sync; 251 struct ufsmount *ump = oip->i_ump; 252 253 if (ovp->v_type == VCHR || ovp->v_type == VBLK || 254 ovp->v_type == VFIFO || ovp->v_type == VSOCK) { 255 return 0; 256 } 257 258 if (length < 0) 259 return (EINVAL); 260 261 if (ovp->v_type == VLNK && 262 (ext2fs_size(oip) < ump->um_maxsymlinklen || 263 (ump->um_maxsymlinklen == 0 && oip->i_e2fs_nblock == 0))) { 264 KDASSERT(length == 0); 265 memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0, 266 (u_int)ext2fs_size(oip)); 267 (void)ext2fs_setsize(oip, 0); 268 oip->i_flag |= IN_CHANGE | IN_UPDATE; 269 return (ext2fs_update(ovp, NULL, NULL, 0)); 270 } 271 if (ext2fs_size(oip) == length) { 272 /* still do a uvm_vnp_setsize() as writesize may be larger */ 273 uvm_vnp_setsize(ovp, length); 274 oip->i_flag |= IN_CHANGE | IN_UPDATE; 275 return (ext2fs_update(ovp, NULL, NULL, 0)); 276 } 277 fs = oip->i_e2fs; 278 if (length > ump->um_maxfilesize) 279 return (EFBIG); 280 281 osize = ext2fs_size(oip); 282 283 /* 284 * Lengthen the size of the file. We must ensure that the 285 * last byte of the file is allocated. Since the smallest 286 * value of osize is 0, length will be at least 1. 287 */ 288 if (osize < length) { 289 uvm_vnp_setwritesize(ovp, length); 290 error = ufs_balloc_range(ovp, length - 1, 1, cred, 291 ioflag & IO_SYNC ? B_SYNC : 0); 292 if (error) { 293 (void) ext2fs_truncate(ovp, osize, ioflag & IO_SYNC, 294 cred); 295 return (error); 296 } 297 uvm_vnp_setsize(ovp, length); 298 oip->i_flag |= IN_CHANGE | IN_UPDATE; 299 KASSERT(error || ovp->v_size == ext2fs_size(oip)); 300 return (ext2fs_update(ovp, NULL, NULL, 0)); 301 } 302 /* 303 * Shorten the size of the file. If the file is not being 304 * truncated to a block boundry, the contents of the 305 * partial block following the end of the file must be 306 * zero'ed in case it ever become accessible again because 307 * of subsequent file growth. 308 */ 309 offset = blkoff(fs, length); 310 if (offset != 0) { 311 size = fs->e2fs_bsize; 312 313 /* XXXUBC we should handle more than just VREG */ 314 uvm_vnp_zerorange(ovp, length, size - offset); 315 } 316 (void)ext2fs_setsize(oip, length); 317 uvm_vnp_setsize(ovp, length); 318 /* 319 * Calculate index into inode's block list of 320 * last direct and indirect blocks (if any) 321 * which we want to keep. Lastblock is -1 when 322 * the file is truncated to 0. 323 */ 324 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 325 lastiblock[SINGLE] = lastblock - NDADDR; 326 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 327 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 328 nblocks = btodb(fs->e2fs_bsize); 329 /* 330 * Update file and block pointers on disk before we start freeing 331 * blocks. If we crash before free'ing blocks below, the blocks 332 * will be returned to the free list. lastiblock values are also 333 * normalized to -1 for calls to ext2fs_indirtrunc below. 334 */ 335 memcpy((void *)oldblks, (void *)&oip->i_e2fs_blocks[0], sizeof oldblks); 336 sync = 0; 337 for (level = TRIPLE; level >= SINGLE; level--) { 338 if (lastiblock[level] < 0 && oldblks[NDADDR + level] != 0) { 339 sync = 1; 340 oip->i_e2fs_blocks[NDADDR + level] = 0; 341 lastiblock[level] = -1; 342 } 343 } 344 for (i = 0; i < NDADDR; i++) { 345 if (i > lastblock && oldblks[i] != 0) { 346 sync = 1; 347 oip->i_e2fs_blocks[i] = 0; 348 } 349 } 350 oip->i_flag |= IN_CHANGE | IN_UPDATE; 351 if (sync) { 352 error = ext2fs_update(ovp, NULL, NULL, UPDATE_WAIT); 353 if (error && !allerror) 354 allerror = error; 355 } 356 357 /* 358 * Having written the new inode to disk, save its new configuration 359 * and put back the old block pointers long enough to process them. 360 * Note that we save the new block configuration so we can check it 361 * when we are done. 362 */ 363 memcpy((void *)newblks, (void *)&oip->i_e2fs_blocks[0], sizeof newblks); 364 memcpy((void *)&oip->i_e2fs_blocks[0], (void *)oldblks, sizeof oldblks); 365 366 (void)ext2fs_setsize(oip, osize); 367 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 368 if (error && !allerror) 369 allerror = error; 370 371 /* 372 * Indirect blocks first. 373 */ 374 indir_lbn[SINGLE] = -NDADDR; 375 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 376 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 377 for (level = TRIPLE; level >= SINGLE; level--) { 378 /* XXX ondisk32 */ 379 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 380 if (bn != 0) { 381 error = ext2fs_indirtrunc(oip, indir_lbn[level], 382 fsbtodb(fs, bn), lastiblock[level], level, &count); 383 if (error) 384 allerror = error; 385 blocksreleased += count; 386 if (lastiblock[level] < 0) { 387 oip->i_e2fs_blocks[NDADDR + level] = 0; 388 ext2fs_blkfree(oip, bn); 389 blocksreleased += nblocks; 390 } 391 } 392 if (lastiblock[level] >= 0) 393 goto done; 394 } 395 396 /* 397 * All whole direct blocks or frags. 398 */ 399 for (i = NDADDR - 1; i > lastblock; i--) { 400 /* XXX ondisk32 */ 401 bn = fs2h32(oip->i_e2fs_blocks[i]); 402 if (bn == 0) 403 continue; 404 oip->i_e2fs_blocks[i] = 0; 405 ext2fs_blkfree(oip, bn); 406 blocksreleased += btodb(fs->e2fs_bsize); 407 } 408 409 done: 410 #ifdef DIAGNOSTIC 411 for (level = SINGLE; level <= TRIPLE; level++) 412 if (newblks[NDADDR + level] != 413 oip->i_e2fs_blocks[NDADDR + level]) 414 panic("ext2fs_truncate1"); 415 for (i = 0; i < NDADDR; i++) 416 if (newblks[i] != oip->i_e2fs_blocks[i]) 417 panic("ext2fs_truncate2"); 418 if (length == 0 && 419 (!LIST_EMPTY(&ovp->v_cleanblkhd) || 420 !LIST_EMPTY(&ovp->v_dirtyblkhd))) 421 panic("ext2fs_truncate3"); 422 #endif /* DIAGNOSTIC */ 423 /* 424 * Put back the real size. 425 */ 426 (void)ext2fs_setsize(oip, length); 427 oip->i_e2fs_nblock -= blocksreleased; 428 oip->i_flag |= IN_CHANGE; 429 KASSERT(ovp->v_type != VREG || ovp->v_size == ext2fs_size(oip)); 430 return (allerror); 431 } 432 433 /* 434 * Release blocks associated with the inode ip and stored in the indirect 435 * block bn. Blocks are free'd in LIFO order up to (but not including) 436 * lastbn. If level is greater than SINGLE, the block is an indirect block 437 * and recursive calls to indirtrunc must be used to cleanse other indirect 438 * blocks. 439 * 440 * NB: triple indirect blocks are untested. 441 */ 442 static int 443 ext2fs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn, 444 int level, long *countp) 445 { 446 int i; 447 struct buf *bp; 448 struct m_ext2fs *fs = ip->i_e2fs; 449 int32_t *bap; /* XXX ondisk32 */ 450 struct vnode *vp; 451 daddr_t nb, nlbn, last; 452 int32_t *copy = NULL; /* XXX ondisk32 */ 453 long blkcount, factor; 454 int nblocks, blocksreleased = 0; 455 int error = 0, allerror = 0; 456 457 /* 458 * Calculate index in current block of last 459 * block to be kept. -1 indicates the entire 460 * block so we need not calculate the index. 461 */ 462 factor = 1; 463 for (i = SINGLE; i < level; i++) 464 factor *= NINDIR(fs); 465 last = lastbn; 466 if (lastbn > 0) 467 last /= factor; 468 nblocks = btodb(fs->e2fs_bsize); 469 /* 470 * Get buffer of block pointers, zero those entries corresponding 471 * to blocks to be free'd, and update on disk copy first. Since 472 * double(triple) indirect before single(double) indirect, calls 473 * to bmap on these blocks will fail. However, we already have 474 * the on disk address, so we have to set the b_blkno field 475 * explicitly instead of letting bread do everything for us. 476 */ 477 vp = ITOV(ip); 478 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 479 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) { 480 /* Braces must be here in case trace evaluates to nothing. */ 481 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 482 } else { 483 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 484 curlwp->l_ru.ru_inblock++; /* pay for read */ 485 bp->b_flags |= B_READ; 486 if (bp->b_bcount > bp->b_bufsize) 487 panic("ext2fs_indirtrunc: bad buffer size"); 488 bp->b_blkno = dbn; 489 VOP_STRATEGY(vp, bp); 490 error = biowait(bp); 491 } 492 if (error) { 493 brelse(bp, 0); 494 *countp = 0; 495 return (error); 496 } 497 498 bap = (int32_t *)bp->b_data; /* XXX ondisk32 */ 499 if (lastbn >= 0) { 500 /* XXX ondisk32 */ 501 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK); 502 memcpy((void *)copy, (void *)bap, (u_int)fs->e2fs_bsize); 503 memset((void *)&bap[last + 1], 0, 504 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (uint32_t)); 505 error = bwrite(bp); 506 if (error) 507 allerror = error; 508 bap = copy; 509 } 510 511 /* 512 * Recursively free totally unused blocks. 513 */ 514 for (i = NINDIR(fs) - 1, 515 nlbn = lbn + 1 - i * factor; i > last; 516 i--, nlbn += factor) { 517 /* XXX ondisk32 */ 518 nb = fs2h32(bap[i]); 519 if (nb == 0) 520 continue; 521 if (level > SINGLE) { 522 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 523 (daddr_t)-1, level - 1, 524 &blkcount); 525 if (error) 526 allerror = error; 527 blocksreleased += blkcount; 528 } 529 ext2fs_blkfree(ip, nb); 530 blocksreleased += nblocks; 531 } 532 533 /* 534 * Recursively free last partial block. 535 */ 536 if (level > SINGLE && lastbn >= 0) { 537 last = lastbn % factor; 538 /* XXX ondisk32 */ 539 nb = fs2h32(bap[i]); 540 if (nb != 0) { 541 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 542 last, level - 1, &blkcount); 543 if (error) 544 allerror = error; 545 blocksreleased += blkcount; 546 } 547 } 548 549 if (copy != NULL) { 550 free(copy, M_TEMP); 551 } else { 552 brelse(bp, BC_INVAL); 553 } 554 555 *countp = blocksreleased; 556 return (allerror); 557 } 558