1 /* $NetBSD: ext2fs_inode.c,v 1.71 2010/02/07 17:12:40 bouyer 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 1.71 2010/02/07 17:12:40 bouyer 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 mutex_enter(&vp->v_interlock); 160 vp->v_iflag |= VI_FREEING; 161 mutex_exit(&vp->v_interlock); 162 ext2fs_vfree(vp, ip->i_number, ip->i_e2fs_mode); 163 } 164 if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) { 165 ext2fs_update(vp, NULL, NULL, 0); 166 } 167 out: 168 /* 169 * If we are done with the inode, reclaim it 170 * so that it can be reused immediately. 171 */ 172 *ap->a_recycle = (ip->i_e2fs_dtime != 0); 173 VOP_UNLOCK(vp, 0); 174 return (error); 175 } 176 177 178 /* 179 * Update the access, modified, and inode change times as specified by the 180 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 181 * used to specify that the inode needs to be updated but that the times have 182 * already been set. The access and modified times are taken from the second 183 * and third parameters; the inode change time is always taken from the current 184 * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk 185 * write of the inode to complete. 186 */ 187 int 188 ext2fs_update(struct vnode *vp, const struct timespec *acc, 189 const struct timespec *mod, int updflags) 190 { 191 struct m_ext2fs *fs; 192 struct buf *bp; 193 struct inode *ip; 194 int error; 195 void *cp; 196 int flags; 197 198 if (vp->v_mount->mnt_flag & MNT_RDONLY) 199 return (0); 200 ip = VTOI(vp); 201 EXT2FS_ITIMES(ip, acc, mod, NULL); 202 if (updflags & UPDATE_CLOSE) 203 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED); 204 else 205 flags = ip->i_flag & IN_MODIFIED; 206 if (flags == 0) 207 return (0); 208 fs = ip->i_e2fs; 209 210 error = bread(ip->i_devvp, 211 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 212 (int)fs->e2fs_bsize, NOCRED, B_MODIFY, &bp); 213 if (error) { 214 brelse(bp, 0); 215 return (error); 216 } 217 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED); 218 cp = (char *)bp->b_data + 219 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs)); 220 e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp); 221 if ((updflags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 && 222 (flags & IN_MODIFIED) != 0 && 223 (vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 224 return (bwrite(bp)); 225 else { 226 bdwrite(bp); 227 return (0); 228 } 229 } 230 231 #define SINGLE 0 /* index of single indirect block */ 232 #define DOUBLE 1 /* index of double indirect block */ 233 #define TRIPLE 2 /* index of triple indirect block */ 234 /* 235 * Truncate the inode oip to at most length size, freeing the 236 * disk blocks. 237 */ 238 int 239 ext2fs_truncate(struct vnode *ovp, off_t length, int ioflag, 240 kauth_cred_t cred) 241 { 242 daddr_t lastblock; 243 struct inode *oip = VTOI(ovp); 244 daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR]; 245 /* XXX ondisk32 */ 246 int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 247 struct m_ext2fs *fs; 248 int offset, size, level; 249 long count, blocksreleased = 0; 250 int i, nblocks; 251 int error, allerror = 0; 252 off_t osize; 253 int sync; 254 struct ufsmount *ump = oip->i_ump; 255 256 if (ovp->v_type == VCHR || ovp->v_type == VBLK || 257 ovp->v_type == VFIFO || ovp->v_type == VSOCK) { 258 return 0; 259 } 260 261 if (length < 0) 262 return (EINVAL); 263 264 if (ovp->v_type == VLNK && 265 (ext2fs_size(oip) < ump->um_maxsymlinklen || 266 (ump->um_maxsymlinklen == 0 && oip->i_e2fs_nblock == 0))) { 267 KDASSERT(length == 0); 268 memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0, 269 (u_int)ext2fs_size(oip)); 270 (void)ext2fs_setsize(oip, 0); 271 oip->i_flag |= IN_CHANGE | IN_UPDATE; 272 return (ext2fs_update(ovp, NULL, NULL, 0)); 273 } 274 if (ext2fs_size(oip) == length) { 275 /* still do a uvm_vnp_setsize() as writesize may be larger */ 276 uvm_vnp_setsize(ovp, length); 277 oip->i_flag |= IN_CHANGE | IN_UPDATE; 278 return (ext2fs_update(ovp, NULL, NULL, 0)); 279 } 280 fs = oip->i_e2fs; 281 if (length > ump->um_maxfilesize) 282 return (EFBIG); 283 284 osize = ext2fs_size(oip); 285 286 /* 287 * Lengthen the size of the file. We must ensure that the 288 * last byte of the file is allocated. Since the smallest 289 * value of osize is 0, length will be at least 1. 290 */ 291 if (osize < length) { 292 uvm_vnp_setwritesize(ovp, length); 293 error = ufs_balloc_range(ovp, length - 1, 1, cred, 294 ioflag & IO_SYNC ? B_SYNC : 0); 295 if (error) { 296 (void) ext2fs_truncate(ovp, osize, ioflag & IO_SYNC, 297 cred); 298 return (error); 299 } 300 uvm_vnp_setsize(ovp, length); 301 oip->i_flag |= IN_CHANGE | IN_UPDATE; 302 KASSERT(error || ovp->v_size == ext2fs_size(oip)); 303 return (ext2fs_update(ovp, NULL, NULL, 0)); 304 } 305 /* 306 * Shorten the size of the file. If the file is not being 307 * truncated to a block boundry, the contents of the 308 * partial block following the end of the file must be 309 * zero'ed in case it ever become accessible again because 310 * of subsequent file growth. 311 */ 312 offset = blkoff(fs, length); 313 if (offset != 0) { 314 size = fs->e2fs_bsize; 315 316 /* XXXUBC we should handle more than just VREG */ 317 uvm_vnp_zerorange(ovp, length, size - offset); 318 } 319 (void)ext2fs_setsize(oip, length); 320 uvm_vnp_setsize(ovp, length); 321 /* 322 * Calculate index into inode's block list of 323 * last direct and indirect blocks (if any) 324 * which we want to keep. Lastblock is -1 when 325 * the file is truncated to 0. 326 */ 327 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 328 lastiblock[SINGLE] = lastblock - NDADDR; 329 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 330 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 331 nblocks = btodb(fs->e2fs_bsize); 332 /* 333 * Update file and block pointers on disk before we start freeing 334 * blocks. If we crash before free'ing blocks below, the blocks 335 * will be returned to the free list. lastiblock values are also 336 * normalized to -1 for calls to ext2fs_indirtrunc below. 337 */ 338 memcpy((void *)oldblks, (void *)&oip->i_e2fs_blocks[0], sizeof oldblks); 339 sync = 0; 340 for (level = TRIPLE; level >= SINGLE; level--) { 341 if (lastiblock[level] < 0 && oldblks[NDADDR + level] != 0) { 342 sync = 1; 343 oip->i_e2fs_blocks[NDADDR + level] = 0; 344 lastiblock[level] = -1; 345 } 346 } 347 for (i = 0; i < NDADDR; i++) { 348 if (i > lastblock && oldblks[i] != 0) { 349 sync = 1; 350 oip->i_e2fs_blocks[i] = 0; 351 } 352 } 353 oip->i_flag |= IN_CHANGE | IN_UPDATE; 354 if (sync) { 355 error = ext2fs_update(ovp, NULL, NULL, UPDATE_WAIT); 356 if (error && !allerror) 357 allerror = error; 358 } 359 360 /* 361 * Having written the new inode to disk, save its new configuration 362 * and put back the old block pointers long enough to process them. 363 * Note that we save the new block configuration so we can check it 364 * when we are done. 365 */ 366 memcpy((void *)newblks, (void *)&oip->i_e2fs_blocks[0], sizeof newblks); 367 memcpy((void *)&oip->i_e2fs_blocks[0], (void *)oldblks, sizeof oldblks); 368 369 (void)ext2fs_setsize(oip, osize); 370 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 371 if (error && !allerror) 372 allerror = error; 373 374 /* 375 * Indirect blocks first. 376 */ 377 indir_lbn[SINGLE] = -NDADDR; 378 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 379 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 380 for (level = TRIPLE; level >= SINGLE; level--) { 381 /* XXX ondisk32 */ 382 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 383 if (bn != 0) { 384 error = ext2fs_indirtrunc(oip, indir_lbn[level], 385 fsbtodb(fs, bn), lastiblock[level], level, &count); 386 if (error) 387 allerror = error; 388 blocksreleased += count; 389 if (lastiblock[level] < 0) { 390 oip->i_e2fs_blocks[NDADDR + level] = 0; 391 ext2fs_blkfree(oip, bn); 392 blocksreleased += nblocks; 393 } 394 } 395 if (lastiblock[level] >= 0) 396 goto done; 397 } 398 399 /* 400 * All whole direct blocks or frags. 401 */ 402 for (i = NDADDR - 1; i > lastblock; i--) { 403 /* XXX ondisk32 */ 404 bn = fs2h32(oip->i_e2fs_blocks[i]); 405 if (bn == 0) 406 continue; 407 oip->i_e2fs_blocks[i] = 0; 408 ext2fs_blkfree(oip, bn); 409 blocksreleased += btodb(fs->e2fs_bsize); 410 } 411 412 done: 413 #ifdef DIAGNOSTIC 414 for (level = SINGLE; level <= TRIPLE; level++) 415 if (newblks[NDADDR + level] != 416 oip->i_e2fs_blocks[NDADDR + level]) 417 panic("ext2fs_truncate1"); 418 for (i = 0; i < NDADDR; i++) 419 if (newblks[i] != oip->i_e2fs_blocks[i]) 420 panic("ext2fs_truncate2"); 421 if (length == 0 && 422 (!LIST_EMPTY(&ovp->v_cleanblkhd) || 423 !LIST_EMPTY(&ovp->v_dirtyblkhd))) 424 panic("ext2fs_truncate3"); 425 #endif /* DIAGNOSTIC */ 426 /* 427 * Put back the real size. 428 */ 429 (void)ext2fs_setsize(oip, length); 430 oip->i_e2fs_nblock -= blocksreleased; 431 oip->i_flag |= IN_CHANGE; 432 KASSERT(ovp->v_type != VREG || ovp->v_size == ext2fs_size(oip)); 433 return (allerror); 434 } 435 436 /* 437 * Release blocks associated with the inode ip and stored in the indirect 438 * block bn. Blocks are free'd in LIFO order up to (but not including) 439 * lastbn. If level is greater than SINGLE, the block is an indirect block 440 * and recursive calls to indirtrunc must be used to cleanse other indirect 441 * blocks. 442 * 443 * NB: triple indirect blocks are untested. 444 */ 445 static int 446 ext2fs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn, 447 int level, long *countp) 448 { 449 int i; 450 struct buf *bp; 451 struct m_ext2fs *fs = ip->i_e2fs; 452 int32_t *bap; /* XXX ondisk32 */ 453 struct vnode *vp; 454 daddr_t nb, nlbn, last; 455 int32_t *copy = NULL; /* XXX ondisk32 */ 456 long blkcount, factor; 457 int nblocks, blocksreleased = 0; 458 int error = 0, allerror = 0; 459 460 /* 461 * Calculate index in current block of last 462 * block to be kept. -1 indicates the entire 463 * block so we need not calculate the index. 464 */ 465 factor = 1; 466 for (i = SINGLE; i < level; i++) 467 factor *= NINDIR(fs); 468 last = lastbn; 469 if (lastbn > 0) 470 last /= factor; 471 nblocks = btodb(fs->e2fs_bsize); 472 /* 473 * Get buffer of block pointers, zero those entries corresponding 474 * to blocks to be free'd, and update on disk copy first. Since 475 * double(triple) indirect before single(double) indirect, calls 476 * to bmap on these blocks will fail. However, we already have 477 * the on disk address, so we have to set the b_blkno field 478 * explicitly instead of letting bread do everything for us. 479 */ 480 vp = ITOV(ip); 481 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 482 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) { 483 /* Braces must be here in case trace evaluates to nothing. */ 484 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 485 } else { 486 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 487 curlwp->l_ru.ru_inblock++; /* pay for read */ 488 bp->b_flags |= B_READ; 489 if (bp->b_bcount > bp->b_bufsize) 490 panic("ext2fs_indirtrunc: bad buffer size"); 491 bp->b_blkno = dbn; 492 VOP_STRATEGY(vp, bp); 493 error = biowait(bp); 494 } 495 if (error) { 496 brelse(bp, 0); 497 *countp = 0; 498 return (error); 499 } 500 501 bap = (int32_t *)bp->b_data; /* XXX ondisk32 */ 502 if (lastbn >= 0) { 503 /* XXX ondisk32 */ 504 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK); 505 memcpy((void *)copy, (void *)bap, (u_int)fs->e2fs_bsize); 506 memset((void *)&bap[last + 1], 0, 507 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (uint32_t)); 508 error = bwrite(bp); 509 if (error) 510 allerror = error; 511 bap = copy; 512 } 513 514 /* 515 * Recursively free totally unused blocks. 516 */ 517 for (i = NINDIR(fs) - 1, 518 nlbn = lbn + 1 - i * factor; i > last; 519 i--, nlbn += factor) { 520 /* XXX ondisk32 */ 521 nb = fs2h32(bap[i]); 522 if (nb == 0) 523 continue; 524 if (level > SINGLE) { 525 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 526 (daddr_t)-1, level - 1, 527 &blkcount); 528 if (error) 529 allerror = error; 530 blocksreleased += blkcount; 531 } 532 ext2fs_blkfree(ip, nb); 533 blocksreleased += nblocks; 534 } 535 536 /* 537 * Recursively free last partial block. 538 */ 539 if (level > SINGLE && lastbn >= 0) { 540 last = lastbn % factor; 541 /* XXX ondisk32 */ 542 nb = fs2h32(bap[i]); 543 if (nb != 0) { 544 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 545 last, level - 1, &blkcount); 546 if (error) 547 allerror = error; 548 blocksreleased += blkcount; 549 } 550 } 551 552 if (copy != NULL) { 553 free(copy, M_TEMP); 554 } else { 555 brelse(bp, BC_INVAL); 556 } 557 558 *countp = blocksreleased; 559 return (allerror); 560 } 561