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 * 4. 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 57 static int ext2_indirtrunc(struct inode *, daddr_t, daddr_t, 58 daddr_t, int, e4fs_daddr_t *); 59 60 /* 61 * Update the access, modified, and inode change times as specified by the 62 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode 63 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by 64 * the timestamp update). The IN_LAZYMOD flag is set to force a write 65 * later if not now. If we write now, then clear both IN_MODIFIED and 66 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is 67 * set, then wait for the write to complete. 68 */ 69 int 70 ext2_update(struct vnode *vp, int waitfor) 71 { 72 struct m_ext2fs *fs; 73 struct buf *bp; 74 struct inode *ip; 75 int error; 76 77 ASSERT_VOP_ELOCKED(vp, "ext2_update"); 78 ext2_itimes(vp); 79 ip = VTOI(vp); 80 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0) 81 return (0); 82 ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED); 83 fs = ip->i_e2fs; 84 if(fs->e2fs_ronly) 85 return (0); 86 if ((error = bread(ip->i_devvp, 87 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 88 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) { 89 brelse(bp); 90 return (error); 91 } 92 ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data + 93 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number))); 94 if (waitfor && !DOINGASYNC(vp)) 95 return (bwrite(bp)); 96 else { 97 bdwrite(bp); 98 return (0); 99 } 100 } 101 102 #define SINGLE 0 /* index of single indirect block */ 103 #define DOUBLE 1 /* index of double indirect block */ 104 #define TRIPLE 2 /* index of triple indirect block */ 105 /* 106 * Truncate the inode oip to at most length size, freeing the 107 * disk blocks. 108 */ 109 int 110 ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred, 111 struct thread *td) 112 { 113 struct vnode *ovp = vp; 114 int32_t lastblock; 115 struct inode *oip; 116 int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 117 uint32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 118 struct m_ext2fs *fs; 119 struct buf *bp; 120 int offset, size, level; 121 e4fs_daddr_t count, nblocks, blocksreleased = 0; 122 int error, i, allerror; 123 off_t osize; 124 125 oip = VTOI(ovp); 126 127 ASSERT_VOP_LOCKED(vp, "ext2_truncate"); 128 129 if (length < 0) 130 return (EINVAL); 131 132 if (ovp->v_type == VLNK && 133 oip->i_size < ovp->v_mount->mnt_maxsymlinklen) { 134 #ifdef INVARIANTS 135 if (length != 0) 136 panic("ext2_truncate: partial truncate of symlink"); 137 #endif 138 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size); 139 oip->i_size = 0; 140 oip->i_flag |= IN_CHANGE | IN_UPDATE; 141 return (ext2_update(ovp, 1)); 142 } 143 if (oip->i_size == length) { 144 oip->i_flag |= IN_CHANGE | IN_UPDATE; 145 return (ext2_update(ovp, 0)); 146 } 147 fs = oip->i_e2fs; 148 osize = oip->i_size; 149 /* 150 * Lengthen the size of the file. We must ensure that the 151 * last byte of the file is allocated. Since the smallest 152 * value of osize is 0, length will be at least 1. 153 */ 154 if (osize < length) { 155 if (length > oip->i_e2fs->e2fs_maxfilesize) 156 return (EFBIG); 157 vnode_pager_setsize(ovp, length); 158 offset = blkoff(fs, length - 1); 159 lbn = lblkno(fs, length - 1); 160 flags |= BA_CLRBUF; 161 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags); 162 if (error) { 163 vnode_pager_setsize(vp, osize); 164 return (error); 165 } 166 oip->i_size = length; 167 if (bp->b_bufsize == fs->e2fs_bsize) 168 bp->b_flags |= B_CLUSTEROK; 169 if (flags & IO_SYNC) 170 bwrite(bp); 171 else if (DOINGASYNC(ovp)) 172 bdwrite(bp); 173 else 174 bawrite(bp); 175 oip->i_flag |= IN_CHANGE | IN_UPDATE; 176 return (ext2_update(ovp, !DOINGASYNC(ovp))); 177 } 178 /* 179 * Shorten the size of the file. If the file is not being 180 * truncated to a block boundry, the contents of the 181 * partial block following the end of the file must be 182 * zero'ed in case it ever become accessible again because 183 * of subsequent file growth. 184 */ 185 /* I don't understand the comment above */ 186 offset = blkoff(fs, length); 187 if (offset == 0) { 188 oip->i_size = length; 189 } else { 190 lbn = lblkno(fs, length); 191 flags |= BA_CLRBUF; 192 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags); 193 if (error) 194 return (error); 195 oip->i_size = length; 196 size = blksize(fs, oip, lbn); 197 bzero((char *)bp->b_data + offset, (u_int)(size - offset)); 198 allocbuf(bp, size); 199 if (bp->b_bufsize == fs->e2fs_bsize) 200 bp->b_flags |= B_CLUSTEROK; 201 if (flags & IO_SYNC) 202 bwrite(bp); 203 else if (DOINGASYNC(ovp)) 204 bdwrite(bp); 205 else 206 bawrite(bp); 207 } 208 /* 209 * Calculate index into inode's block list of 210 * last direct and indirect blocks (if any) 211 * which we want to keep. Lastblock is -1 when 212 * the file is truncated to 0. 213 */ 214 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 215 lastiblock[SINGLE] = lastblock - NDADDR; 216 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 217 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 218 nblocks = btodb(fs->e2fs_bsize); 219 /* 220 * Update file and block pointers on disk before we start freeing 221 * blocks. If we crash before free'ing blocks below, the blocks 222 * will be returned to the free list. lastiblock values are also 223 * normalized to -1 for calls to ext2_indirtrunc below. 224 */ 225 for (level = TRIPLE; level >= SINGLE; level--) { 226 oldblks[NDADDR + level] = oip->i_ib[level]; 227 if (lastiblock[level] < 0) { 228 oip->i_ib[level] = 0; 229 lastiblock[level] = -1; 230 } 231 } 232 for (i = 0; i < NDADDR; i++) { 233 oldblks[i] = oip->i_db[i]; 234 if (i > lastblock) 235 oip->i_db[i] = 0; 236 } 237 oip->i_flag |= IN_CHANGE | IN_UPDATE; 238 allerror = ext2_update(ovp, !DOINGASYNC(ovp)); 239 240 /* 241 * Having written the new inode to disk, save its new configuration 242 * and put back the old block pointers long enough to process them. 243 * Note that we save the new block configuration so we can check it 244 * when we are done. 245 */ 246 for (i = 0; i < NDADDR; i++) { 247 newblks[i] = oip->i_db[i]; 248 oip->i_db[i] = oldblks[i]; 249 } 250 for (i = 0; i < NIADDR; i++) { 251 newblks[NDADDR + i] = oip->i_ib[i]; 252 oip->i_ib[i] = oldblks[NDADDR + i]; 253 } 254 oip->i_size = osize; 255 error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize); 256 if (error && (allerror == 0)) 257 allerror = error; 258 vnode_pager_setsize(ovp, length); 259 260 /* 261 * Indirect blocks first. 262 */ 263 indir_lbn[SINGLE] = -NDADDR; 264 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 265 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 266 for (level = TRIPLE; level >= SINGLE; level--) { 267 bn = oip->i_ib[level]; 268 if (bn != 0) { 269 error = ext2_indirtrunc(oip, indir_lbn[level], 270 fsbtodb(fs, bn), lastiblock[level], level, &count); 271 if (error) 272 allerror = error; 273 blocksreleased += count; 274 if (lastiblock[level] < 0) { 275 oip->i_ib[level] = 0; 276 ext2_blkfree(oip, bn, fs->e2fs_fsize); 277 blocksreleased += nblocks; 278 } 279 } 280 if (lastiblock[level] >= 0) 281 goto done; 282 } 283 284 /* 285 * All whole direct blocks or frags. 286 */ 287 for (i = NDADDR - 1; i > lastblock; i--) { 288 long bsize; 289 290 bn = oip->i_db[i]; 291 if (bn == 0) 292 continue; 293 oip->i_db[i] = 0; 294 bsize = blksize(fs, oip, i); 295 ext2_blkfree(oip, bn, bsize); 296 blocksreleased += btodb(bsize); 297 } 298 if (lastblock < 0) 299 goto done; 300 301 /* 302 * Finally, look for a change in size of the 303 * last direct block; release any frags. 304 */ 305 bn = oip->i_db[lastblock]; 306 if (bn != 0) { 307 long oldspace, newspace; 308 309 /* 310 * Calculate amount of space we're giving 311 * back as old block size minus new block size. 312 */ 313 oldspace = blksize(fs, oip, lastblock); 314 oip->i_size = length; 315 newspace = blksize(fs, oip, lastblock); 316 if (newspace == 0) 317 panic("ext2_truncate: newspace"); 318 if (oldspace - newspace > 0) { 319 /* 320 * Block number of space to be free'd is 321 * the old block # plus the number of frags 322 * required for the storage we're keeping. 323 */ 324 bn += numfrags(fs, newspace); 325 ext2_blkfree(oip, bn, oldspace - newspace); 326 blocksreleased += btodb(oldspace - newspace); 327 } 328 } 329 done: 330 #ifdef INVARIANTS 331 for (level = SINGLE; level <= TRIPLE; level++) 332 if (newblks[NDADDR + level] != oip->i_ib[level]) 333 panic("itrunc1"); 334 for (i = 0; i < NDADDR; i++) 335 if (newblks[i] != oip->i_db[i]) 336 panic("itrunc2"); 337 BO_LOCK(bo); 338 if (length == 0 && (bo->bo_dirty.bv_cnt != 0 || 339 bo->bo_clean.bv_cnt != 0)) 340 panic("itrunc3"); 341 BO_UNLOCK(bo); 342 #endif /* INVARIANTS */ 343 /* 344 * Put back the real size. 345 */ 346 oip->i_size = length; 347 if (oip->i_blocks >= blocksreleased) 348 oip->i_blocks -= blocksreleased; 349 else /* sanity */ 350 oip->i_blocks = 0; 351 oip->i_flag |= IN_CHANGE; 352 vnode_pager_setsize(ovp, length); 353 return (allerror); 354 } 355 356 /* 357 * Release blocks associated with the inode ip and stored in the indirect 358 * block bn. Blocks are free'd in LIFO order up to (but not including) 359 * lastbn. If level is greater than SINGLE, the block is an indirect block 360 * and recursive calls to indirtrunc must be used to cleanse other indirect 361 * blocks. 362 * 363 * NB: triple indirect blocks are untested. 364 */ 365 366 static int 367 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, 368 daddr_t lastbn, int level, e4fs_daddr_t *countp) 369 { 370 struct buf *bp; 371 struct m_ext2fs *fs = ip->i_e2fs; 372 struct vnode *vp; 373 e2fs_daddr_t *bap, *copy; 374 int i, nblocks, error = 0, allerror = 0; 375 e2fs_lbn_t nb, nlbn, last; 376 e4fs_daddr_t blkcount, factor, blocksreleased = 0; 377 378 /* 379 * Calculate index in current block of last 380 * block to be kept. -1 indicates the entire 381 * block so we need not calculate the index. 382 */ 383 factor = 1; 384 for (i = SINGLE; i < level; i++) 385 factor *= NINDIR(fs); 386 last = lastbn; 387 if (lastbn > 0) 388 last /= factor; 389 nblocks = btodb(fs->e2fs_bsize); 390 /* 391 * Get buffer of block pointers, zero those entries corresponding 392 * to blocks to be free'd, and update on disk copy first. Since 393 * double(triple) indirect before single(double) indirect, calls 394 * to bmap on these blocks will fail. However, we already have 395 * the on disk address, so we have to set the b_blkno field 396 * explicitly instead of letting bread do everything for us. 397 */ 398 vp = ITOV(ip); 399 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0); 400 if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) { 401 bp->b_iocmd = BIO_READ; 402 if (bp->b_bcount > bp->b_bufsize) 403 panic("ext2_indirtrunc: bad buffer size"); 404 bp->b_blkno = dbn; 405 vfs_busy_pages(bp, 0); 406 bp->b_iooffset = dbtob(bp->b_blkno); 407 bstrategy(bp); 408 error = bufwait(bp); 409 } 410 if (error) { 411 brelse(bp); 412 *countp = 0; 413 return (error); 414 } 415 416 bap = (e2fs_daddr_t *)bp->b_data; 417 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK); 418 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize); 419 bzero((caddr_t)&bap[last + 1], 420 (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t)); 421 if (last == -1) 422 bp->b_flags |= B_INVAL; 423 if (DOINGASYNC(vp)) { 424 bdwrite(bp); 425 } else { 426 error = bwrite(bp); 427 if (error) 428 allerror = error; 429 } 430 bap = copy; 431 432 /* 433 * Recursively free totally unused blocks. 434 */ 435 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 436 i--, nlbn += factor) { 437 nb = bap[i]; 438 if (nb == 0) 439 continue; 440 if (level > SINGLE) { 441 if ((error = ext2_indirtrunc(ip, nlbn, 442 fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0) 443 allerror = error; 444 blocksreleased += blkcount; 445 } 446 ext2_blkfree(ip, nb, fs->e2fs_bsize); 447 blocksreleased += nblocks; 448 } 449 450 /* 451 * Recursively free last partial block. 452 */ 453 if (level > SINGLE && lastbn >= 0) { 454 last = lastbn % factor; 455 nb = bap[i]; 456 if (nb != 0) { 457 if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 458 last, level - 1, &blkcount)) != 0) 459 allerror = error; 460 blocksreleased += blkcount; 461 } 462 } 463 free(copy, M_TEMP); 464 *countp = blocksreleased; 465 return (allerror); 466 } 467 468 /* 469 * discard preallocated blocks 470 */ 471 int 472 ext2_inactive(struct vop_inactive_args *ap) 473 { 474 struct vnode *vp = ap->a_vp; 475 struct inode *ip = VTOI(vp); 476 struct thread *td = ap->a_td; 477 int mode, error = 0; 478 479 /* 480 * Ignore inodes related to stale file handles. 481 */ 482 if (ip->i_mode == 0) 483 goto out; 484 if (ip->i_nlink <= 0) { 485 error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td); 486 ip->i_rdev = 0; 487 mode = ip->i_mode; 488 ip->i_mode = 0; 489 ip->i_flag |= IN_CHANGE | IN_UPDATE; 490 ext2_vfree(vp, ip->i_number, mode); 491 } 492 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) 493 ext2_update(vp, 0); 494 out: 495 /* 496 * If we are done with the inode, reclaim it 497 * so that it can be reused immediately. 498 */ 499 if (ip->i_mode == 0) 500 vrecycle(vp); 501 return (error); 502 } 503 504 /* 505 * Reclaim an inode so that it can be used for other purposes. 506 */ 507 int 508 ext2_reclaim(struct vop_reclaim_args *ap) 509 { 510 struct inode *ip; 511 struct vnode *vp = ap->a_vp; 512 513 ip = VTOI(vp); 514 if (ip->i_flag & IN_LAZYMOD) { 515 ip->i_flag |= IN_MODIFIED; 516 ext2_update(vp, 0); 517 } 518 vfs_hash_remove(vp); 519 free(vp->v_data, M_EXT2NODE); 520 vp->v_data = 0; 521 vnode_destroy_vobject(vp); 522 return (0); 523 } 524