1 /* $OpenBSD: ext2fs_inode.c,v 1.58 2016/03/19 12:04:16 natano Exp $ */ 2 /* $NetBSD: ext2fs_inode.c,v 1.24 2001/06/19 12:59:18 wiz Exp $ */ 3 4 /* 5 * Copyright (c) 1997 Manuel Bouyer. 6 * Copyright (c) 1982, 1986, 1989, 1993 7 * The Regents of the University of California. All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94 34 * Modified for ext2fs by Manuel Bouyer. 35 */ 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/mount.h> 40 #include <sys/proc.h> 41 #include <sys/buf.h> 42 #include <sys/vnode.h> 43 #include <sys/kernel.h> 44 #include <sys/malloc.h> 45 #include <sys/resourcevar.h> 46 47 #include <ufs/ufs/quota.h> 48 #include <ufs/ufs/inode.h> 49 #include <ufs/ufs/ufsmount.h> 50 #include <ufs/ufs/ufs_extern.h> 51 52 #include <ufs/ext2fs/ext2fs.h> 53 #include <ufs/ext2fs/ext2fs_extern.h> 54 55 static int ext2fs_indirtrunc(struct inode *, int32_t, int32_t, 56 int32_t, int, long *); 57 58 /* 59 * Get the size of an inode. 60 */ 61 u_int64_t 62 ext2fs_size(struct inode *ip) 63 { 64 u_int64_t size = ip->i_e2fs_size; 65 66 if ((ip->i_e2fs_mode & IFMT) == IFREG) 67 size |= (u_int64_t)ip->i_e2fs_size_hi << 32; 68 69 return (size); 70 } 71 72 int 73 ext2fs_setsize(struct inode *ip, u_int64_t size) 74 { 75 struct m_ext2fs *fs = ip->i_e2fs; 76 77 if (size <= fs->e2fs_maxfilesize) { 78 /* If HUGE_FILEs are off, e2fs_maxfilesize will protect us. */ 79 if ((ip->i_e2fs_mode & IFMT) == IFREG || ip->i_e2fs_mode == 0) 80 ip->i_e2fs_size_hi = size >> 32; 81 82 ip->i_e2fs_size = size; 83 return (0); 84 } 85 86 /* Linux automagically upgrades to REV1 here! */ 87 if (fs->e2fs.e2fs_rev <= E2FS_REV0) 88 return (EFBIG); 89 90 if ((fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_LARGEFILE) == 0) { 91 fs->e2fs.e2fs_features_rocompat |= EXT2F_ROCOMPAT_LARGEFILE; 92 fs->e2fs_fmod = 1; 93 } 94 return (EFBIG); 95 } 96 97 98 /* 99 * Last reference to an inode. If necessary, write or delete it. 100 */ 101 int 102 ext2fs_inactive(void *v) 103 { 104 struct vop_inactive_args *ap = v; 105 struct vnode *vp = ap->a_vp; 106 struct inode *ip = VTOI(vp); 107 struct proc *p = ap->a_p; 108 struct timespec ts; 109 int error = 0; 110 #ifdef DIAGNOSTIC 111 extern int prtactive; 112 113 if (prtactive && vp->v_usecount != 0) 114 vprint("ext2fs_inactive: pushing active", vp); 115 #endif 116 117 /* Get rid of inodes related to stale file handles. */ 118 if (ip->i_e2din == NULL || ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime) 119 goto out; 120 121 error = 0; 122 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 123 if (ext2fs_size(ip) != 0) { 124 error = ext2fs_truncate(ip, (off_t)0, 0, NOCRED); 125 } 126 getnanotime(&ts); 127 ip->i_e2fs_dtime = ts.tv_sec; 128 ip->i_flag |= IN_CHANGE | IN_UPDATE; 129 ext2fs_inode_free(ip, ip->i_number, ip->i_e2fs_mode); 130 } 131 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) { 132 ext2fs_update(ip, 0); 133 } 134 out: 135 VOP_UNLOCK(vp, p); 136 /* 137 * If we are done with the inode, reclaim it 138 * so that it can be reused immediately. 139 */ 140 if (ip->i_e2din == NULL || ip->i_e2fs_dtime != 0) 141 vrecycle(vp, p); 142 return (error); 143 } 144 145 146 /* 147 * Update the access, modified, and inode change times as specified by the 148 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 149 * used to specify that the inode needs to be updated but that the times have 150 * already been set. The access and modified times are taken from the second 151 * and third parameters; the inode change time is always taken from the current 152 * time. If waitfor is set, then wait for the disk write of the inode to 153 * complete. 154 */ 155 int 156 ext2fs_update(struct inode *ip, int waitfor) 157 { 158 struct m_ext2fs *fs; 159 struct buf *bp; 160 int error; 161 caddr_t cp; 162 163 if (ITOV(ip)->v_mount->mnt_flag & MNT_RDONLY) 164 return (0); 165 EXT2FS_ITIMES(ip); 166 if ((ip->i_flag & IN_MODIFIED) == 0) 167 return (0); 168 ip->i_flag &= ~IN_MODIFIED; 169 fs = ip->i_e2fs; 170 error = bread(ip->i_devvp, 171 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 172 (int)fs->e2fs_bsize, &bp); 173 if (error) { 174 brelse(bp); 175 return (error); 176 } 177 ip->i_flag &= ~(IN_MODIFIED); 178 cp = (caddr_t)bp->b_data + 179 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs)); 180 181 /* 182 * See note about 16-bit UID/GID limitation in ext2fs_vget(). Now 183 * that we are about to write the inode, construct the split UID and 184 * GID fields out of the two 32-bit fields we kept in memory. 185 */ 186 ip->i_e2fs_uid_low = (u_int16_t)ip->i_e2fs_uid; 187 ip->i_e2fs_gid_low = (u_int16_t)ip->i_e2fs_gid; 188 ip->i_e2fs_uid_high = ip->i_e2fs_uid >> 16; 189 ip->i_e2fs_gid_high = ip->i_e2fs_gid >> 16; 190 191 e2fs_isave(fs, ip->i_e2din, (struct ext2fs_dinode *)cp); 192 if (waitfor) 193 return (bwrite(bp)); 194 else { 195 bdwrite(bp); 196 return (0); 197 } 198 } 199 200 #define SINGLE 0 /* index of single indirect block */ 201 #define DOUBLE 1 /* index of double indirect block */ 202 #define TRIPLE 2 /* index of triple indirect block */ 203 /* 204 * Truncate the inode oip to at most length size, freeing the 205 * disk blocks. 206 */ 207 int 208 ext2fs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred) 209 { 210 struct vnode *ovp = ITOV(oip); 211 int32_t lastblock; 212 int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 213 int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 214 struct m_ext2fs *fs; 215 struct buf *bp; 216 int offset, size, level; 217 long count, nblocks, vflags, blocksreleased = 0; 218 int i; 219 int aflags, error, allerror; 220 off_t osize; 221 222 if (length < 0) 223 return (EINVAL); 224 225 if (ovp->v_type != VREG && 226 ovp->v_type != VDIR && 227 ovp->v_type != VLNK) 228 return (0); 229 230 if (ovp->v_type == VLNK && ext2fs_size(oip) < EXT2_MAXSYMLINKLEN) { 231 #ifdef DIAGNOSTIC 232 if (length != 0) 233 panic("ext2fs_truncate: partial truncate of symlink"); 234 #endif 235 memset(&oip->i_e2din->e2di_shortlink, 0, ext2fs_size(oip)); 236 (void)ext2fs_setsize(oip, 0); 237 oip->i_flag |= IN_CHANGE | IN_UPDATE; 238 return (ext2fs_update(oip, 1)); 239 } 240 241 if (ext2fs_size(oip) == length) { 242 oip->i_flag |= IN_CHANGE | IN_UPDATE; 243 return (ext2fs_update(oip, 0)); 244 } 245 fs = oip->i_e2fs; 246 osize = ext2fs_size(oip); 247 /* 248 * Lengthen the size of the file. We must ensure that the 249 * last byte of the file is allocated. Since the smallest 250 * value of osize is 0, length will be at least 1. 251 */ 252 if (osize < length) { 253 #if 0 /* XXX */ 254 if (length > fs->fs_maxfilesize) 255 return (EFBIG); 256 #endif 257 offset = blkoff(fs, length - 1); 258 lbn = lblkno(fs, length - 1); 259 aflags = B_CLRBUF; 260 if (flags & IO_SYNC) 261 aflags |= B_SYNC; 262 error = ext2fs_buf_alloc(oip, lbn, offset + 1, cred, &bp, 263 aflags); 264 if (error) 265 return (error); 266 (void)ext2fs_setsize(oip, length); 267 uvm_vnp_setsize(ovp, length); 268 uvm_vnp_uncache(ovp); 269 if (aflags & B_SYNC) 270 bwrite(bp); 271 else 272 bawrite(bp); 273 oip->i_flag |= IN_CHANGE | IN_UPDATE; 274 return (ext2fs_update(oip, 1)); 275 } 276 /* 277 * Shorten the size of the file. If the file is not being 278 * truncated to a block boundry, the contents of the 279 * partial block following the end of the file must be 280 * zero'ed in case it ever become accessible again because 281 * of subsequent file growth. 282 */ 283 offset = blkoff(fs, length); 284 if (offset == 0) { 285 (void)ext2fs_setsize(oip, length); 286 } else { 287 lbn = lblkno(fs, length); 288 aflags = B_CLRBUF; 289 if (flags & IO_SYNC) 290 aflags |= B_SYNC; 291 error = ext2fs_buf_alloc(oip, lbn, offset, cred, &bp, 292 aflags); 293 if (error) 294 return (error); 295 (void)ext2fs_setsize(oip, length); 296 size = fs->e2fs_bsize; 297 uvm_vnp_setsize(ovp, length); 298 uvm_vnp_uncache(ovp); 299 memset(bp->b_data + offset, 0, size - offset); 300 bp->b_bcount = size; 301 if (aflags & B_SYNC) 302 bwrite(bp); 303 else 304 bawrite(bp); 305 } 306 /* 307 * Calculate index into inode's block list of 308 * last direct and indirect blocks (if any) 309 * which we want to keep. Lastblock is -1 when 310 * the file is truncated to 0. 311 */ 312 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 313 lastiblock[SINGLE] = lastblock - NDADDR; 314 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 315 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 316 nblocks = btodb(fs->e2fs_bsize); 317 /* 318 * Update file and block pointers on disk before we start freeing 319 * blocks. If we crash before free'ing blocks below, the blocks 320 * will be returned to the free list. lastiblock values are also 321 * normalized to -1 for calls to ext2fs_indirtrunc below. 322 */ 323 memcpy(oldblks, &oip->i_e2fs_blocks[0], sizeof(oldblks)); 324 for (level = TRIPLE; level >= SINGLE; level--) 325 if (lastiblock[level] < 0) { 326 oip->i_e2fs_blocks[NDADDR + level] = 0; 327 lastiblock[level] = -1; 328 } 329 for (i = NDADDR - 1; i > lastblock; i--) 330 oip->i_e2fs_blocks[i] = 0; 331 oip->i_flag |= IN_CHANGE | IN_UPDATE; 332 if ((error = ext2fs_update(oip, 1)) != 0) 333 allerror = error; 334 /* 335 * Having written the new inode to disk, save its new configuration 336 * and put back the old block pointers long enough to process them. 337 * Note that we save the new block configuration so we can check it 338 * when we are done. 339 */ 340 memcpy(newblks, &oip->i_e2fs_blocks[0], sizeof(newblks)); 341 memcpy(&oip->i_e2fs_blocks[0], oldblks, sizeof(oldblks)); 342 (void)ext2fs_setsize(oip, osize); 343 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA; 344 allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0); 345 346 /* 347 * Indirect blocks first. 348 */ 349 indir_lbn[SINGLE] = -NDADDR; 350 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 351 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 352 for (level = TRIPLE; level >= SINGLE; level--) { 353 bn = letoh32(oip->i_e2fs_blocks[NDADDR + level]); 354 if (bn != 0) { 355 error = ext2fs_indirtrunc(oip, indir_lbn[level], 356 fsbtodb(fs, bn), lastiblock[level], level, &count); 357 if (error) 358 allerror = error; 359 blocksreleased += count; 360 if (lastiblock[level] < 0) { 361 oip->i_e2fs_blocks[NDADDR + level] = 0; 362 ext2fs_blkfree(oip, bn); 363 blocksreleased += nblocks; 364 } 365 } 366 if (lastiblock[level] >= 0) 367 goto done; 368 } 369 370 /* 371 * All whole direct blocks or frags. 372 */ 373 for (i = NDADDR - 1; i > lastblock; i--) { 374 bn = letoh32(oip->i_e2fs_blocks[i]); 375 if (bn == 0) 376 continue; 377 oip->i_e2fs_blocks[i] = 0; 378 ext2fs_blkfree(oip, bn); 379 blocksreleased += btodb(fs->e2fs_bsize); 380 } 381 382 done: 383 #ifdef DIAGNOSTIC 384 for (level = SINGLE; level <= TRIPLE; level++) 385 if (newblks[NDADDR + level] != 386 oip->i_e2fs_blocks[NDADDR + level]) 387 panic("ext2fs_truncate1"); 388 for (i = 0; i < NDADDR; i++) 389 if (newblks[i] != oip->i_e2fs_blocks[i]) 390 panic("ext2fs_truncate2"); 391 if (length == 0 && 392 (!LIST_EMPTY(&ovp->v_cleanblkhd) || 393 !LIST_EMPTY(&ovp->v_dirtyblkhd))) 394 panic("ext2fs_truncate3"); 395 #endif /* DIAGNOSTIC */ 396 /* 397 * Put back the real size. 398 */ 399 (void)ext2fs_setsize(oip, length); 400 if (blocksreleased >= oip->i_e2fs_nblock) 401 oip->i_e2fs_nblock = 0; 402 else 403 oip->i_e2fs_nblock -= blocksreleased; 404 oip->i_flag |= IN_CHANGE; 405 return (allerror); 406 } 407 408 /* 409 * Release blocks associated with the inode ip and stored in the indirect 410 * block bn. Blocks are free'd in LIFO order up to (but not including) 411 * lastbn. If level is greater than SINGLE, the block is an indirect block 412 * and recursive calls to indirtrunc must be used to cleanse other indirect 413 * blocks. 414 * 415 * NB: triple indirect blocks are untested. 416 */ 417 static int 418 ext2fs_indirtrunc(struct inode *ip, int32_t lbn, int32_t dbn, int32_t lastbn, int level, long *countp) 419 { 420 int i; 421 struct buf *bp; 422 struct m_ext2fs *fs = ip->i_e2fs; 423 int32_t *bap; 424 struct vnode *vp; 425 int32_t *copy = NULL, nb, nlbn, last; 426 long blkcount, factor; 427 int nblocks, blocksreleased = 0; 428 int error = 0, allerror = 0; 429 430 /* 431 * Calculate index in current block of last 432 * block to be kept. -1 indicates the entire 433 * block so we need not calculate the index. 434 */ 435 factor = 1; 436 for (i = SINGLE; i < level; i++) 437 factor *= NINDIR(fs); 438 last = lastbn; 439 if (lastbn > 0) 440 last /= factor; 441 nblocks = btodb(fs->e2fs_bsize); 442 /* 443 * Get buffer of block pointers, zero those entries corresponding 444 * to blocks to be free'd, and update on disk copy first. Since 445 * double(triple) indirect before single(double) indirect, calls 446 * to bmap on these blocks will fail. However, we already have 447 * the on disk address, so we have to set the b_blkno field 448 * explicitly instead of letting bread do everything for us. 449 */ 450 vp = ITOV(ip); 451 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 452 if (!(bp->b_flags & (B_DONE | B_DELWRI))) { 453 curproc->p_ru.ru_inblock++; /* pay for read */ 454 bcstats.pendingreads++; 455 bcstats.numreads++; 456 bp->b_flags |= B_READ; 457 if (bp->b_bcount > bp->b_bufsize) 458 panic("ext2fs_indirtrunc: bad buffer size"); 459 bp->b_blkno = dbn; 460 VOP_STRATEGY(bp); 461 error = biowait(bp); 462 } 463 if (error) { 464 brelse(bp); 465 *countp = 0; 466 return (error); 467 } 468 469 bap = (int32_t *)bp->b_data; 470 if (lastbn >= 0) { 471 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK); 472 memcpy(copy, bap, fs->e2fs_bsize); 473 memset(&bap[last + 1], 0, 474 (NINDIR(fs) - (last + 1)) * sizeof(u_int32_t)); 475 error = bwrite(bp); 476 if (error) 477 allerror = error; 478 bap = copy; 479 } 480 481 /* 482 * Recursively free totally unused blocks. 483 */ 484 for (i = NINDIR(fs) - 1, 485 nlbn = lbn + 1 - i * factor; i > last; 486 i--, nlbn += factor) { 487 nb = letoh32(bap[i]); 488 if (nb == 0) 489 continue; 490 if (level > SINGLE) { 491 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 492 (int32_t)-1, level - 1, 493 &blkcount); 494 if (error) 495 allerror = error; 496 blocksreleased += blkcount; 497 } 498 ext2fs_blkfree(ip, nb); 499 blocksreleased += nblocks; 500 } 501 502 /* 503 * Recursively free last partial block. 504 */ 505 if (level > SINGLE && lastbn >= 0) { 506 last = lastbn % factor; 507 nb = letoh32(bap[i]); 508 if (nb != 0) { 509 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 510 last, level - 1, &blkcount); 511 if (error) 512 allerror = error; 513 blocksreleased += blkcount; 514 } 515 } 516 517 if (copy != NULL) { 518 free(copy, M_TEMP, fs->e2fs_bsize); 519 } else { 520 bp->b_flags |= B_INVAL; 521 brelse(bp); 522 } 523 524 *countp = blocksreleased; 525 return (allerror); 526 } 527