1 /*- 2 * modified for Lites 1.1 3 * 4 * Aug 1995, Godmar Back (gback@cs.utah.edu) 5 * University of Utah, Department of Computer Science 6 */ 7 /*- 8 * Copyright (c) 1982, 1986, 1989, 1993 9 * The Regents of the University of California. All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)ffs_inode.c 8.5 (Berkeley) 12/30/93 36 * $FreeBSD$ 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/mount.h> 42 #include <sys/bio.h> 43 #include <sys/buf.h> 44 #include <sys/vnode.h> 45 #include <sys/malloc.h> 46 #include <sys/rwlock.h> 47 48 #include <vm/vm.h> 49 #include <vm/vm_extern.h> 50 51 #include <fs/ext2fs/inode.h> 52 #include <fs/ext2fs/ext2_mount.h> 53 #include <fs/ext2fs/ext2fs.h> 54 #include <fs/ext2fs/fs.h> 55 #include <fs/ext2fs/ext2_extern.h> 56 #include <fs/ext2fs/ext2_extattr.h> 57 58 static int ext2_indirtrunc(struct inode *, daddr_t, daddr_t, 59 daddr_t, int, e4fs_daddr_t *); 60 61 /* 62 * Update the access, modified, and inode change times as specified by the 63 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode 64 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by 65 * the timestamp update). The IN_LAZYMOD flag is set to force a write 66 * later if not now. If we write now, then clear both IN_MODIFIED and 67 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is 68 * set, then wait for the write to complete. 69 */ 70 int 71 ext2_update(struct vnode *vp, int waitfor) 72 { 73 struct m_ext2fs *fs; 74 struct buf *bp; 75 struct inode *ip; 76 int error; 77 78 ASSERT_VOP_ELOCKED(vp, "ext2_update"); 79 ext2_itimes(vp); 80 ip = VTOI(vp); 81 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0) 82 return (0); 83 ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED); 84 fs = ip->i_e2fs; 85 if (fs->e2fs_ronly) 86 return (0); 87 if ((error = bread(ip->i_devvp, 88 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 89 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) { 90 brelse(bp); 91 return (error); 92 } 93 ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data + 94 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number))); 95 if (waitfor && !DOINGASYNC(vp)) 96 return (bwrite(bp)); 97 else { 98 bdwrite(bp); 99 return (0); 100 } 101 } 102 103 #define SINGLE 0 /* index of single indirect block */ 104 #define DOUBLE 1 /* index of double indirect block */ 105 #define TRIPLE 2 /* index of triple indirect block */ 106 /* 107 * Truncate the inode oip to at most length size, freeing the 108 * disk blocks. 109 */ 110 int 111 ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred, 112 struct thread *td) 113 { 114 struct vnode *ovp = vp; 115 int32_t lastblock; 116 struct inode *oip; 117 int32_t bn, lbn, lastiblock[EXT2_NIADDR], indir_lbn[EXT2_NIADDR]; 118 uint32_t oldblks[EXT2_NDADDR + EXT2_NIADDR]; 119 uint32_t newblks[EXT2_NDADDR + EXT2_NIADDR]; 120 struct m_ext2fs *fs; 121 struct buf *bp; 122 int offset, size, level; 123 e4fs_daddr_t count, nblocks, blocksreleased = 0; 124 int error, i, allerror; 125 off_t osize; 126 #ifdef INVARIANTS 127 struct bufobj *bo; 128 #endif 129 130 oip = VTOI(ovp); 131 #ifdef INVARIANTS 132 bo = &ovp->v_bufobj; 133 #endif 134 135 ASSERT_VOP_LOCKED(vp, "ext2_truncate"); 136 137 if (length < 0) 138 return (EINVAL); 139 140 if (ovp->v_type == VLNK && 141 oip->i_size < ovp->v_mount->mnt_maxsymlinklen) { 142 #ifdef INVARIANTS 143 if (length != 0) 144 panic("ext2_truncate: partial truncate of symlink"); 145 #endif 146 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size); 147 oip->i_size = 0; 148 oip->i_flag |= IN_CHANGE | IN_UPDATE; 149 return (ext2_update(ovp, 1)); 150 } 151 if (oip->i_size == length) { 152 oip->i_flag |= IN_CHANGE | IN_UPDATE; 153 return (ext2_update(ovp, 0)); 154 } 155 fs = oip->i_e2fs; 156 osize = oip->i_size; 157 /* 158 * Lengthen the size of the file. We must ensure that the 159 * last byte of the file is allocated. Since the smallest 160 * value of osize is 0, length will be at least 1. 161 */ 162 if (osize < length) { 163 if (length > oip->i_e2fs->e2fs_maxfilesize) 164 return (EFBIG); 165 vnode_pager_setsize(ovp, length); 166 offset = blkoff(fs, length - 1); 167 lbn = lblkno(fs, length - 1); 168 flags |= BA_CLRBUF; 169 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags); 170 if (error) { 171 vnode_pager_setsize(vp, osize); 172 return (error); 173 } 174 oip->i_size = length; 175 if (bp->b_bufsize == fs->e2fs_bsize) 176 bp->b_flags |= B_CLUSTEROK; 177 if (flags & IO_SYNC) 178 bwrite(bp); 179 else if (DOINGASYNC(ovp)) 180 bdwrite(bp); 181 else 182 bawrite(bp); 183 oip->i_flag |= IN_CHANGE | IN_UPDATE; 184 return (ext2_update(ovp, !DOINGASYNC(ovp))); 185 } 186 /* 187 * Shorten the size of the file. If the file is not being 188 * truncated to a block boundary, the contents of the 189 * partial block following the end of the file must be 190 * zero'ed in case it ever become accessible again because 191 * of subsequent file growth. 192 */ 193 /* I don't understand the comment above */ 194 offset = blkoff(fs, length); 195 if (offset == 0) { 196 oip->i_size = length; 197 } else { 198 lbn = lblkno(fs, length); 199 flags |= BA_CLRBUF; 200 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags); 201 if (error) 202 return (error); 203 oip->i_size = length; 204 size = blksize(fs, oip, lbn); 205 bzero((char *)bp->b_data + offset, (u_int)(size - offset)); 206 allocbuf(bp, size); 207 if (bp->b_bufsize == fs->e2fs_bsize) 208 bp->b_flags |= B_CLUSTEROK; 209 if (flags & IO_SYNC) 210 bwrite(bp); 211 else if (DOINGASYNC(ovp)) 212 bdwrite(bp); 213 else 214 bawrite(bp); 215 } 216 /* 217 * Calculate index into inode's block list of 218 * last direct and indirect blocks (if any) 219 * which we want to keep. Lastblock is -1 when 220 * the file is truncated to 0. 221 */ 222 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 223 lastiblock[SINGLE] = lastblock - EXT2_NDADDR; 224 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 225 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 226 nblocks = btodb(fs->e2fs_bsize); 227 /* 228 * Update file and block pointers on disk before we start freeing 229 * blocks. If we crash before free'ing blocks below, the blocks 230 * will be returned to the free list. lastiblock values are also 231 * normalized to -1 for calls to ext2_indirtrunc below. 232 */ 233 for (level = TRIPLE; level >= SINGLE; level--) { 234 oldblks[EXT2_NDADDR + level] = oip->i_ib[level]; 235 if (lastiblock[level] < 0) { 236 oip->i_ib[level] = 0; 237 lastiblock[level] = -1; 238 } 239 } 240 for (i = 0; i < EXT2_NDADDR; i++) { 241 oldblks[i] = oip->i_db[i]; 242 if (i > lastblock) 243 oip->i_db[i] = 0; 244 } 245 oip->i_flag |= IN_CHANGE | IN_UPDATE; 246 allerror = ext2_update(ovp, !DOINGASYNC(ovp)); 247 248 /* 249 * Having written the new inode to disk, save its new configuration 250 * and put back the old block pointers long enough to process them. 251 * Note that we save the new block configuration so we can check it 252 * when we are done. 253 */ 254 for (i = 0; i < EXT2_NDADDR; i++) { 255 newblks[i] = oip->i_db[i]; 256 oip->i_db[i] = oldblks[i]; 257 } 258 for (i = 0; i < EXT2_NIADDR; i++) { 259 newblks[EXT2_NDADDR + i] = oip->i_ib[i]; 260 oip->i_ib[i] = oldblks[EXT2_NDADDR + i]; 261 } 262 oip->i_size = osize; 263 error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize); 264 if (error && (allerror == 0)) 265 allerror = error; 266 vnode_pager_setsize(ovp, length); 267 268 /* 269 * Indirect blocks first. 270 */ 271 indir_lbn[SINGLE] = -EXT2_NDADDR; 272 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 273 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 274 for (level = TRIPLE; level >= SINGLE; level--) { 275 bn = oip->i_ib[level]; 276 if (bn != 0) { 277 error = ext2_indirtrunc(oip, indir_lbn[level], 278 fsbtodb(fs, bn), lastiblock[level], level, &count); 279 if (error) 280 allerror = error; 281 blocksreleased += count; 282 if (lastiblock[level] < 0) { 283 oip->i_ib[level] = 0; 284 ext2_blkfree(oip, bn, fs->e2fs_fsize); 285 blocksreleased += nblocks; 286 } 287 } 288 if (lastiblock[level] >= 0) 289 goto done; 290 } 291 292 /* 293 * All whole direct blocks or frags. 294 */ 295 for (i = EXT2_NDADDR - 1; i > lastblock; i--) { 296 long bsize; 297 298 bn = oip->i_db[i]; 299 if (bn == 0) 300 continue; 301 oip->i_db[i] = 0; 302 bsize = blksize(fs, oip, i); 303 ext2_blkfree(oip, bn, bsize); 304 blocksreleased += btodb(bsize); 305 } 306 if (lastblock < 0) 307 goto done; 308 309 /* 310 * Finally, look for a change in size of the 311 * last direct block; release any frags. 312 */ 313 bn = oip->i_db[lastblock]; 314 if (bn != 0) { 315 long oldspace, newspace; 316 317 /* 318 * Calculate amount of space we're giving 319 * back as old block size minus new block size. 320 */ 321 oldspace = blksize(fs, oip, lastblock); 322 oip->i_size = length; 323 newspace = blksize(fs, oip, lastblock); 324 if (newspace == 0) 325 panic("ext2_truncate: newspace"); 326 if (oldspace - newspace > 0) { 327 /* 328 * Block number of space to be free'd is 329 * the old block # plus the number of frags 330 * required for the storage we're keeping. 331 */ 332 bn += numfrags(fs, newspace); 333 ext2_blkfree(oip, bn, oldspace - newspace); 334 blocksreleased += btodb(oldspace - newspace); 335 } 336 } 337 done: 338 #ifdef INVARIANTS 339 for (level = SINGLE; level <= TRIPLE; level++) 340 if (newblks[EXT2_NDADDR + level] != oip->i_ib[level]) 341 panic("itrunc1"); 342 for (i = 0; i < EXT2_NDADDR; i++) 343 if (newblks[i] != oip->i_db[i]) 344 panic("itrunc2"); 345 BO_LOCK(bo); 346 if (length == 0 && (bo->bo_dirty.bv_cnt != 0 || 347 bo->bo_clean.bv_cnt != 0)) 348 panic("itrunc3"); 349 BO_UNLOCK(bo); 350 #endif /* INVARIANTS */ 351 /* 352 * Put back the real size. 353 */ 354 oip->i_size = length; 355 if (oip->i_blocks >= blocksreleased) 356 oip->i_blocks -= blocksreleased; 357 else /* sanity */ 358 oip->i_blocks = 0; 359 oip->i_flag |= IN_CHANGE; 360 vnode_pager_setsize(ovp, length); 361 return (allerror); 362 } 363 364 /* 365 * Release blocks associated with the inode ip and stored in the indirect 366 * block bn. Blocks are free'd in LIFO order up to (but not including) 367 * lastbn. If level is greater than SINGLE, the block is an indirect block 368 * and recursive calls to indirtrunc must be used to cleanse other indirect 369 * blocks. 370 * 371 * NB: triple indirect blocks are untested. 372 */ 373 374 static int 375 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, 376 daddr_t lastbn, int level, e4fs_daddr_t *countp) 377 { 378 struct buf *bp; 379 struct m_ext2fs *fs = ip->i_e2fs; 380 struct vnode *vp; 381 e2fs_daddr_t *bap, *copy; 382 int i, nblocks, error = 0, allerror = 0; 383 e2fs_lbn_t nb, nlbn, last; 384 e4fs_daddr_t blkcount, factor, blocksreleased = 0; 385 386 /* 387 * Calculate index in current block of last 388 * block to be kept. -1 indicates the entire 389 * block so we need not calculate the index. 390 */ 391 factor = 1; 392 for (i = SINGLE; i < level; i++) 393 factor *= NINDIR(fs); 394 last = lastbn; 395 if (lastbn > 0) 396 last /= factor; 397 nblocks = btodb(fs->e2fs_bsize); 398 /* 399 * Get buffer of block pointers, zero those entries corresponding 400 * to blocks to be free'd, and update on disk copy first. Since 401 * double(triple) indirect before single(double) indirect, calls 402 * to bmap on these blocks will fail. However, we already have 403 * the on disk address, so we have to set the b_blkno field 404 * explicitly instead of letting bread do everything for us. 405 */ 406 vp = ITOV(ip); 407 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0); 408 if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) { 409 bp->b_iocmd = BIO_READ; 410 if (bp->b_bcount > bp->b_bufsize) 411 panic("ext2_indirtrunc: bad buffer size"); 412 bp->b_blkno = dbn; 413 vfs_busy_pages(bp, 0); 414 bp->b_iooffset = dbtob(bp->b_blkno); 415 bstrategy(bp); 416 error = bufwait(bp); 417 } 418 if (error) { 419 brelse(bp); 420 *countp = 0; 421 return (error); 422 } 423 bap = (e2fs_daddr_t *)bp->b_data; 424 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK); 425 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize); 426 bzero((caddr_t)&bap[last + 1], 427 (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t)); 428 if (last == -1) 429 bp->b_flags |= B_INVAL; 430 if (DOINGASYNC(vp)) { 431 bdwrite(bp); 432 } else { 433 error = bwrite(bp); 434 if (error) 435 allerror = error; 436 } 437 bap = copy; 438 439 /* 440 * Recursively free totally unused blocks. 441 */ 442 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 443 i--, nlbn += factor) { 444 nb = bap[i]; 445 if (nb == 0) 446 continue; 447 if (level > SINGLE) { 448 if ((error = ext2_indirtrunc(ip, nlbn, 449 fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0) 450 allerror = error; 451 blocksreleased += blkcount; 452 } 453 ext2_blkfree(ip, nb, fs->e2fs_bsize); 454 blocksreleased += nblocks; 455 } 456 457 /* 458 * Recursively free last partial block. 459 */ 460 if (level > SINGLE && lastbn >= 0) { 461 last = lastbn % factor; 462 nb = bap[i]; 463 if (nb != 0) { 464 if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 465 last, level - 1, &blkcount)) != 0) 466 allerror = error; 467 blocksreleased += blkcount; 468 } 469 } 470 free(copy, M_TEMP); 471 *countp = blocksreleased; 472 return (allerror); 473 } 474 475 /* 476 * discard preallocated blocks 477 */ 478 int 479 ext2_inactive(struct vop_inactive_args *ap) 480 { 481 struct vnode *vp = ap->a_vp; 482 struct inode *ip = VTOI(vp); 483 struct thread *td = ap->a_td; 484 int mode, error = 0; 485 486 /* 487 * Ignore inodes related to stale file handles. 488 */ 489 if (ip->i_mode == 0) 490 goto out; 491 if (ip->i_nlink <= 0) { 492 ext2_extattr_free(ip); 493 error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td); 494 ip->i_rdev = 0; 495 mode = ip->i_mode; 496 ip->i_mode = 0; 497 ip->i_flag |= IN_CHANGE | IN_UPDATE; 498 ext2_vfree(vp, ip->i_number, mode); 499 } 500 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) 501 ext2_update(vp, 0); 502 out: 503 /* 504 * If we are done with the inode, reclaim it 505 * so that it can be reused immediately. 506 */ 507 if (ip->i_mode == 0) 508 vrecycle(vp); 509 return (error); 510 } 511 512 /* 513 * Reclaim an inode so that it can be used for other purposes. 514 */ 515 int 516 ext2_reclaim(struct vop_reclaim_args *ap) 517 { 518 struct inode *ip; 519 struct vnode *vp = ap->a_vp; 520 521 ip = VTOI(vp); 522 if (ip->i_flag & IN_LAZYMOD) { 523 ip->i_flag |= IN_MODIFIED; 524 ext2_update(vp, 0); 525 } 526 vfs_hash_remove(vp); 527 free(vp->v_data, M_EXT2NODE); 528 vp->v_data = 0; 529 vnode_destroy_vobject(vp); 530 return (0); 531 } 532