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 47 #include <vm/vm.h> 48 #include <vm/vm_extern.h> 49 50 #include <fs/ext2fs/inode.h> 51 #include <fs/ext2fs/ext2_mount.h> 52 #include <fs/ext2fs/ext2fs.h> 53 #include <fs/ext2fs/fs.h> 54 #include <fs/ext2fs/ext2_extern.h> 55 56 static int ext2_indirtrunc(struct inode *, int32_t, int32_t, int32_t, int, 57 long *); 58 59 /* 60 * Update the access, modified, and inode change times as specified by the 61 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode 62 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by 63 * the timestamp update). The IN_LAZYMOD flag is set to force a write 64 * later if not now. If we write now, then clear both IN_MODIFIED and 65 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is 66 * set, then wait for the write to complete. 67 */ 68 int 69 ext2_update(vp, waitfor) 70 struct vnode *vp; 71 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(vp, length, flags, cred, td) 112 struct vnode *vp; 113 off_t length; 114 int flags; 115 struct ucred *cred; 116 struct thread *td; 117 { 118 struct vnode *ovp = vp; 119 int32_t lastblock; 120 struct inode *oip; 121 int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 122 uint32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 123 struct bufobj *bo; 124 struct m_ext2fs *fs; 125 struct buf *bp; 126 int offset, size, level; 127 long count, nblocks, blocksreleased = 0; 128 int error, i, allerror; 129 off_t osize; 130 131 oip = VTOI(ovp); 132 bo = &ovp->v_bufobj; 133 134 ASSERT_VOP_LOCKED(vp, "ext2_truncate"); 135 136 if (length < 0) 137 return (EINVAL); 138 139 if (ovp->v_type == VLNK && 140 oip->i_size < ovp->v_mount->mnt_maxsymlinklen) { 141 #ifdef DIAGNOSTIC 142 if (length != 0) 143 panic("ext2_truncate: partial truncate of symlink"); 144 #endif 145 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size); 146 oip->i_size = 0; 147 oip->i_flag |= IN_CHANGE | IN_UPDATE; 148 return (ext2_update(ovp, 1)); 149 } 150 if (oip->i_size == length) { 151 oip->i_flag |= IN_CHANGE | IN_UPDATE; 152 return (ext2_update(ovp, 0)); 153 } 154 fs = oip->i_e2fs; 155 osize = oip->i_size; 156 /* 157 * Lengthen the size of the file. We must ensure that the 158 * last byte of the file is allocated. Since the smallest 159 * value of osize is 0, length will be at least 1. 160 */ 161 if (osize < length) { 162 if (length > oip->i_e2fs->e2fs_maxfilesize) 163 return (EFBIG); 164 vnode_pager_setsize(ovp, length); 165 offset = blkoff(fs, length - 1); 166 lbn = lblkno(fs, length - 1); 167 flags |= BA_CLRBUF; 168 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags); 169 if (error) { 170 vnode_pager_setsize(vp, osize); 171 return (error); 172 } 173 oip->i_size = length; 174 if (bp->b_bufsize == fs->e2fs_bsize) 175 bp->b_flags |= B_CLUSTEROK; 176 if (flags & IO_SYNC) 177 bwrite(bp); 178 else if (DOINGASYNC(ovp)) 179 bdwrite(bp); 180 else 181 bawrite(bp); 182 oip->i_flag |= IN_CHANGE | IN_UPDATE; 183 return (ext2_update(ovp, !DOINGASYNC(ovp))); 184 } 185 /* 186 * Shorten the size of the file. If the file is not being 187 * truncated to a block boundry, the contents of the 188 * partial block following the end of the file must be 189 * zero'ed in case it ever become accessible again because 190 * of subsequent file growth. 191 */ 192 /* I don't understand the comment above */ 193 offset = blkoff(fs, length); 194 if (offset == 0) { 195 oip->i_size = length; 196 } else { 197 lbn = lblkno(fs, length); 198 flags |= BA_CLRBUF; 199 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags); 200 if (error) 201 return (error); 202 oip->i_size = length; 203 size = blksize(fs, oip, lbn); 204 bzero((char *)bp->b_data + offset, (u_int)(size - offset)); 205 allocbuf(bp, size); 206 if (bp->b_bufsize == fs->e2fs_bsize) 207 bp->b_flags |= B_CLUSTEROK; 208 if (flags & IO_SYNC) 209 bwrite(bp); 210 else if (DOINGASYNC(ovp)) 211 bdwrite(bp); 212 else 213 bawrite(bp); 214 } 215 /* 216 * Calculate index into inode's block list of 217 * last direct and indirect blocks (if any) 218 * which we want to keep. Lastblock is -1 when 219 * the file is truncated to 0. 220 */ 221 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 222 lastiblock[SINGLE] = lastblock - NDADDR; 223 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 224 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 225 nblocks = btodb(fs->e2fs_bsize); 226 /* 227 * Update file and block pointers on disk before we start freeing 228 * blocks. If we crash before free'ing blocks below, the blocks 229 * will be returned to the free list. lastiblock values are also 230 * normalized to -1 for calls to ext2_indirtrunc below. 231 */ 232 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof(oldblks)); 233 for (level = TRIPLE; level >= SINGLE; level--) 234 if (lastiblock[level] < 0) { 235 oip->i_ib[level] = 0; 236 lastiblock[level] = -1; 237 } 238 for (i = NDADDR - 1; i > lastblock; i--) 239 oip->i_db[i] = 0; 240 oip->i_flag |= IN_CHANGE | IN_UPDATE; 241 allerror = ext2_update(ovp, !DOINGASYNC(ovp)); 242 243 /* 244 * Having written the new inode to disk, save its new configuration 245 * and put back the old block pointers long enough to process them. 246 * Note that we save the new block configuration so we can check it 247 * when we are done. 248 */ 249 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof(newblks)); 250 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof(oldblks)); 251 oip->i_size = osize; 252 error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize); 253 if (error && (allerror == 0)) 254 allerror = error; 255 vnode_pager_setsize(ovp, length); 256 257 /* 258 * Indirect blocks first. 259 */ 260 indir_lbn[SINGLE] = -NDADDR; 261 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 262 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 263 for (level = TRIPLE; level >= SINGLE; level--) { 264 bn = oip->i_ib[level]; 265 if (bn != 0) { 266 error = ext2_indirtrunc(oip, indir_lbn[level], 267 fsbtodb(fs, bn), lastiblock[level], level, &count); 268 if (error) 269 allerror = error; 270 blocksreleased += count; 271 if (lastiblock[level] < 0) { 272 oip->i_ib[level] = 0; 273 ext2_blkfree(oip, bn, fs->e2fs_fsize); 274 blocksreleased += nblocks; 275 } 276 } 277 if (lastiblock[level] >= 0) 278 goto done; 279 } 280 281 /* 282 * All whole direct blocks or frags. 283 */ 284 for (i = NDADDR - 1; i > lastblock; i--) { 285 long bsize; 286 287 bn = oip->i_db[i]; 288 if (bn == 0) 289 continue; 290 oip->i_db[i] = 0; 291 bsize = blksize(fs, oip, i); 292 ext2_blkfree(oip, bn, bsize); 293 blocksreleased += btodb(bsize); 294 } 295 if (lastblock < 0) 296 goto done; 297 298 /* 299 * Finally, look for a change in size of the 300 * last direct block; release any frags. 301 */ 302 bn = oip->i_db[lastblock]; 303 if (bn != 0) { 304 long oldspace, newspace; 305 306 /* 307 * Calculate amount of space we're giving 308 * back as old block size minus new block size. 309 */ 310 oldspace = blksize(fs, oip, lastblock); 311 oip->i_size = length; 312 newspace = blksize(fs, oip, lastblock); 313 if (newspace == 0) 314 panic("itrunc: newspace"); 315 if (oldspace - newspace > 0) { 316 /* 317 * Block number of space to be free'd is 318 * the old block # plus the number of frags 319 * required for the storage we're keeping. 320 */ 321 bn += numfrags(fs, newspace); 322 ext2_blkfree(oip, bn, oldspace - newspace); 323 blocksreleased += btodb(oldspace - newspace); 324 } 325 } 326 done: 327 #ifdef DIAGNOSTIC 328 for (level = SINGLE; level <= TRIPLE; level++) 329 if (newblks[NDADDR + level] != oip->i_ib[level]) 330 panic("itrunc1"); 331 for (i = 0; i < NDADDR; i++) 332 if (newblks[i] != oip->i_db[i]) 333 panic("itrunc2"); 334 BO_LOCK(bo); 335 if (length == 0 && (bo->bo_dirty.bv_cnt != 0 || 336 bo->bo_clean.bv_cnt != 0)) 337 panic("itrunc3"); 338 BO_UNLOCK(bo); 339 #endif /* DIAGNOSTIC */ 340 /* 341 * Put back the real size. 342 */ 343 oip->i_size = length; 344 oip->i_blocks -= blocksreleased; 345 if (oip->i_blocks > blocksreleased) 346 oip->i_blocks -= blocksreleased; 347 else /* sanity */ 348 oip->i_blocks = 0; 349 oip->i_flag |= IN_CHANGE; 350 vnode_pager_setsize(ovp, length); 351 return (allerror); 352 } 353 354 /* 355 * Release blocks associated with the inode ip and stored in the indirect 356 * block bn. Blocks are free'd in LIFO order up to (but not including) 357 * lastbn. If level is greater than SINGLE, the block is an indirect block 358 * and recursive calls to indirtrunc must be used to cleanse other indirect 359 * blocks. 360 * 361 * NB: triple indirect blocks are untested. 362 */ 363 364 static int 365 ext2_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 366 struct inode *ip; 367 int32_t lbn, lastbn; 368 int32_t dbn; 369 int level; 370 long *countp; 371 { 372 struct buf *bp; 373 struct m_ext2fs *fs = ip->i_e2fs; 374 struct vnode *vp; 375 int32_t *bap, *copy, nb, nlbn, last; 376 long blkcount, factor; 377 int i, nblocks, blocksreleased = 0; 378 int error = 0, allerror = 0; 379 380 /* 381 * Calculate index in current block of last 382 * block to be kept. -1 indicates the entire 383 * block so we need not calculate the index. 384 */ 385 factor = 1; 386 for (i = SINGLE; i < level; i++) 387 factor *= NINDIR(fs); 388 last = lastbn; 389 if (lastbn > 0) 390 last /= factor; 391 nblocks = btodb(fs->e2fs_bsize); 392 /* 393 * Get buffer of block pointers, zero those entries corresponding 394 * to blocks to be free'd, and update on disk copy first. Since 395 * double(triple) indirect before single(double) indirect, calls 396 * to bmap on these blocks will fail. However, we already have 397 * the on disk address, so we have to set the b_blkno field 398 * explicitly instead of letting bread do everything for us. 399 */ 400 vp = ITOV(ip); 401 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0); 402 if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) { 403 bp->b_iocmd = BIO_READ; 404 if (bp->b_bcount > bp->b_bufsize) 405 panic("ext2_indirtrunc: bad buffer size"); 406 bp->b_blkno = dbn; 407 vfs_busy_pages(bp, 0); 408 bp->b_iooffset = dbtob(bp->b_blkno); 409 bstrategy(bp); 410 error = bufwait(bp); 411 } 412 if (error) { 413 brelse(bp); 414 *countp = 0; 415 return (error); 416 } 417 418 bap = (int32_t *)bp->b_data; 419 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK); 420 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize); 421 bzero((caddr_t)&bap[last + 1], 422 (u_int)(NINDIR(fs) - (last + 1)) * sizeof(int32_t)); 423 if (last == -1) 424 bp->b_flags |= B_INVAL; 425 if (DOINGASYNC(vp)) { 426 bdwrite(bp); 427 } else { 428 error = bwrite(bp); 429 if (error) 430 allerror = error; 431 } 432 bap = copy; 433 434 /* 435 * Recursively free totally unused blocks. 436 */ 437 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 438 i--, nlbn += factor) { 439 nb = bap[i]; 440 if (nb == 0) 441 continue; 442 if (level > SINGLE) { 443 if ((error = ext2_indirtrunc(ip, nlbn, 444 fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0) 445 allerror = error; 446 blocksreleased += blkcount; 447 } 448 ext2_blkfree(ip, nb, fs->e2fs_bsize); 449 blocksreleased += nblocks; 450 } 451 452 /* 453 * Recursively free last partial block. 454 */ 455 if (level > SINGLE && lastbn >= 0) { 456 last = lastbn % factor; 457 nb = bap[i]; 458 if (nb != 0) { 459 if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 460 last, level - 1, &blkcount)) != 0) 461 allerror = error; 462 blocksreleased += blkcount; 463 } 464 } 465 free(copy, M_TEMP); 466 *countp = blocksreleased; 467 return (allerror); 468 } 469 470 /* 471 * discard preallocated blocks 472 */ 473 int 474 ext2_inactive(ap) 475 struct vop_inactive_args /* { 476 struct vnode *a_vp; 477 struct thread *a_td; 478 } */ *ap; 479 { 480 struct vnode *vp = ap->a_vp; 481 struct inode *ip = VTOI(vp); 482 struct thread *td = ap->a_td; 483 int mode, error = 0; 484 485 /* 486 * Ignore inodes related to stale file handles. 487 */ 488 if (ip->i_mode == 0) 489 goto out; 490 if (ip->i_nlink <= 0) { 491 error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td); 492 ip->i_rdev = 0; 493 mode = ip->i_mode; 494 ip->i_mode = 0; 495 ip->i_flag |= IN_CHANGE | IN_UPDATE; 496 ext2_vfree(vp, ip->i_number, mode); 497 } 498 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) 499 ext2_update(vp, 0); 500 out: 501 /* 502 * If we are done with the inode, reclaim it 503 * so that it can be reused immediately. 504 */ 505 if (ip->i_mode == 0) 506 vrecycle(vp); 507 return (error); 508 } 509 510 /* 511 * Reclaim an inode so that it can be used for other purposes. 512 */ 513 int 514 ext2_reclaim(ap) 515 struct vop_reclaim_args /* { 516 struct vnode *a_vp; 517 struct thread *a_td; 518 } */ *ap; 519 { 520 struct inode *ip; 521 struct vnode *vp = ap->a_vp; 522 523 ip = VTOI(vp); 524 if (ip->i_flag & IN_LAZYMOD) { 525 ip->i_flag |= IN_MODIFIED; 526 ext2_update(vp, 0); 527 } 528 vfs_hash_remove(vp); 529 free(vp->v_data, M_EXT2NODE); 530 vp->v_data = 0; 531 vnode_destroy_vobject(vp); 532 return (0); 533 } 534