1 /* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. 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 8.13 (Berkeley) 4/21/95 34 * $FreeBSD$ 35 */ 36 37 #include "opt_quota.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/mount.h> 42 #include <sys/proc.h> 43 #include <sys/bio.h> 44 #include <sys/buf.h> 45 #include <sys/vnode.h> 46 #include <sys/malloc.h> 47 #include <sys/resourcevar.h> 48 #include <sys/vmmeter.h> 49 #include <sys/stat.h> 50 51 #include <vm/vm.h> 52 #include <vm/vm_extern.h> 53 54 #include <ufs/ufs/extattr.h> 55 #include <ufs/ufs/quota.h> 56 #include <ufs/ufs/ufsmount.h> 57 #include <ufs/ufs/inode.h> 58 #include <ufs/ufs/ufs_extern.h> 59 60 #include <ufs/ffs/fs.h> 61 #include <ufs/ffs/ffs_extern.h> 62 63 static int ffs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t, 64 ufs_daddr_t, int, long *)); 65 66 /* 67 * Update the access, modified, and inode change times as specified by the 68 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode 69 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by 70 * the timestamp update). The IN_LAZYMOD flag is set to force a write 71 * later if not now. If we write now, then clear both IN_MODIFIED and 72 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is 73 * set, then wait for the write to complete. 74 */ 75 int 76 ffs_update(vp, waitfor) 77 struct vnode *vp; 78 int waitfor; 79 { 80 register struct fs *fs; 81 struct buf *bp; 82 struct inode *ip; 83 int error; 84 85 ufs_itimes(vp); 86 ip = VTOI(vp); 87 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0) 88 return (0); 89 ip->i_flag &= ~(IN_LAZYMOD | IN_MODIFIED); 90 if (vp->v_mount->mnt_flag & MNT_RDONLY) 91 return (0); 92 fs = ip->i_fs; 93 /* 94 * Ensure that uid and gid are correct. This is a temporary 95 * fix until fsck has been changed to do the update. 96 */ 97 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 98 ip->i_din.di_ouid = ip->i_uid; /* XXX */ 99 ip->i_din.di_ogid = ip->i_gid; /* XXX */ 100 } /* XXX */ 101 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 102 (int)fs->fs_bsize, NOCRED, &bp); 103 if (error) { 104 brelse(bp); 105 return (error); 106 } 107 if (DOINGSOFTDEP(vp)) 108 softdep_update_inodeblock(ip, bp, waitfor); 109 else if (ip->i_effnlink != ip->i_nlink) 110 panic("ffs_update: bad link cnt"); 111 *((struct dinode *)bp->b_data + 112 ino_to_fsbo(fs, ip->i_number)) = ip->i_din; 113 if (waitfor && !DOINGASYNC(vp)) { 114 return (bwrite(bp)); 115 } else if (vm_page_count_severe() || buf_dirty_count_severe()) { 116 return (bwrite(bp)); 117 } else { 118 if (bp->b_bufsize == fs->fs_bsize) 119 bp->b_flags |= B_CLUSTEROK; 120 bdwrite(bp); 121 return (0); 122 } 123 } 124 125 #define SINGLE 0 /* index of single indirect block */ 126 #define DOUBLE 1 /* index of double indirect block */ 127 #define TRIPLE 2 /* index of triple indirect block */ 128 /* 129 * Truncate the inode oip to at most length size, freeing the 130 * disk blocks. 131 */ 132 int 133 ffs_truncate(vp, length, flags, cred, p) 134 struct vnode *vp; 135 off_t length; 136 int flags; 137 struct ucred *cred; 138 struct proc *p; 139 { 140 register struct vnode *ovp = vp; 141 ufs_daddr_t lastblock; 142 register struct inode *oip; 143 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 144 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 145 register struct fs *fs; 146 struct buf *bp; 147 int offset, size, level; 148 long count, nblocks, blocksreleased = 0; 149 register int i; 150 int aflags, error, allerror; 151 off_t osize; 152 153 oip = VTOI(ovp); 154 fs = oip->i_fs; 155 if (length < 0) 156 return (EINVAL); 157 if (length > fs->fs_maxfilesize) 158 return (EFBIG); 159 if (ovp->v_type == VLNK && 160 (oip->i_size < ovp->v_mount->mnt_maxsymlinklen || oip->i_din.di_blocks == 0)) { 161 #ifdef DIAGNOSTIC 162 if (length != 0) 163 panic("ffs_truncate: partial truncate of symlink"); 164 #endif 165 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size); 166 oip->i_size = 0; 167 oip->i_flag |= IN_CHANGE | IN_UPDATE; 168 return (UFS_UPDATE(ovp, 1)); 169 } 170 if (oip->i_size == length) { 171 oip->i_flag |= IN_CHANGE | IN_UPDATE; 172 return (UFS_UPDATE(ovp, 0)); 173 } 174 #ifdef QUOTA 175 error = getinoquota(oip); 176 if (error) 177 return (error); 178 #endif 179 if ((oip->i_flags & SF_SNAPSHOT) != 0) 180 ffs_snapremove(ovp); 181 ovp->v_lasta = ovp->v_clen = ovp->v_cstart = ovp->v_lastw = 0; 182 if (DOINGSOFTDEP(ovp)) { 183 if (length > 0 || softdep_slowdown(ovp)) { 184 /* 185 * If a file is only partially truncated, then 186 * we have to clean up the data structures 187 * describing the allocation past the truncation 188 * point. Finding and deallocating those structures 189 * is a lot of work. Since partial truncation occurs 190 * rarely, we solve the problem by syncing the file 191 * so that it will have no data structures left. 192 */ 193 if ((error = VOP_FSYNC(ovp, cred, MNT_WAIT, 194 p)) != 0) 195 return (error); 196 } else { 197 #ifdef QUOTA 198 (void) chkdq(oip, -oip->i_blocks, NOCRED, 0); 199 #endif 200 softdep_setup_freeblocks(oip, length); 201 vinvalbuf(ovp, 0, cred, p, 0, 0); 202 oip->i_flag |= IN_CHANGE | IN_UPDATE; 203 return (ffs_update(ovp, 0)); 204 } 205 } 206 osize = oip->i_size; 207 /* 208 * Lengthen the size of the file. We must ensure that the 209 * last byte of the file is allocated. Since the smallest 210 * value of osize is 0, length will be at least 1. 211 */ 212 if (osize < length) { 213 vnode_pager_setsize(ovp, length); 214 aflags = B_CLRBUF; 215 if (flags & IO_SYNC) 216 aflags |= B_SYNC; 217 error = VOP_BALLOC(ovp, length - 1, 1, 218 cred, aflags, &bp); 219 if (error) 220 return (error); 221 oip->i_size = length; 222 if (bp->b_bufsize == fs->fs_bsize) 223 bp->b_flags |= B_CLUSTEROK; 224 if (aflags & B_SYNC) 225 bwrite(bp); 226 else 227 bawrite(bp); 228 oip->i_flag |= IN_CHANGE | IN_UPDATE; 229 return (UFS_UPDATE(ovp, 1)); 230 } 231 /* 232 * Shorten the size of the file. If the file is not being 233 * truncated to a block boundary, the contents of the 234 * partial block following the end of the file must be 235 * zero'ed in case it ever becomes accessible again because 236 * of subsequent file growth. Directories however are not 237 * zero'ed as they should grow back initialized to empty. 238 */ 239 offset = blkoff(fs, length); 240 if (offset == 0) { 241 oip->i_size = length; 242 } else { 243 lbn = lblkno(fs, length); 244 aflags = B_CLRBUF; 245 if (flags & IO_SYNC) 246 aflags |= B_SYNC; 247 error = VOP_BALLOC(ovp, length - 1, 1, cred, aflags, &bp); 248 if (error) { 249 return (error); 250 } 251 oip->i_size = length; 252 size = blksize(fs, oip, lbn); 253 if (ovp->v_type != VDIR) 254 bzero((char *)bp->b_data + offset, 255 (u_int)(size - offset)); 256 /* Kirk's code has reallocbuf(bp, size, 1) here */ 257 allocbuf(bp, size); 258 if (bp->b_bufsize == fs->fs_bsize) 259 bp->b_flags |= B_CLUSTEROK; 260 if (aflags & B_SYNC) 261 bwrite(bp); 262 else 263 bawrite(bp); 264 } 265 /* 266 * Calculate index into inode's block list of 267 * last direct and indirect blocks (if any) 268 * which we want to keep. Lastblock is -1 when 269 * the file is truncated to 0. 270 */ 271 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1; 272 lastiblock[SINGLE] = lastblock - NDADDR; 273 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 274 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 275 nblocks = btodb(fs->fs_bsize); 276 /* 277 * Update file and block pointers on disk before we start freeing 278 * blocks. If we crash before free'ing blocks below, the blocks 279 * will be returned to the free list. lastiblock values are also 280 * normalized to -1 for calls to ffs_indirtrunc below. 281 */ 282 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks); 283 for (level = TRIPLE; level >= SINGLE; level--) 284 if (lastiblock[level] < 0) { 285 oip->i_ib[level] = 0; 286 lastiblock[level] = -1; 287 } 288 for (i = NDADDR - 1; i > lastblock; i--) 289 oip->i_db[i] = 0; 290 oip->i_flag |= IN_CHANGE | IN_UPDATE; 291 allerror = UFS_UPDATE(ovp, 1); 292 293 /* 294 * Having written the new inode to disk, save its new configuration 295 * and put back the old block pointers long enough to process them. 296 * Note that we save the new block configuration so we can check it 297 * when we are done. 298 */ 299 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks); 300 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks); 301 oip->i_size = osize; 302 303 error = vtruncbuf(ovp, cred, p, length, fs->fs_bsize); 304 if (error && (allerror == 0)) 305 allerror = error; 306 307 /* 308 * Indirect blocks first. 309 */ 310 indir_lbn[SINGLE] = -NDADDR; 311 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 312 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 313 for (level = TRIPLE; level >= SINGLE; level--) { 314 bn = oip->i_ib[level]; 315 if (bn != 0) { 316 error = ffs_indirtrunc(oip, indir_lbn[level], 317 fsbtodb(fs, bn), lastiblock[level], level, &count); 318 if (error) 319 allerror = error; 320 blocksreleased += count; 321 if (lastiblock[level] < 0) { 322 oip->i_ib[level] = 0; 323 ffs_blkfree(oip, bn, fs->fs_bsize); 324 blocksreleased += nblocks; 325 } 326 } 327 if (lastiblock[level] >= 0) 328 goto done; 329 } 330 331 /* 332 * All whole direct blocks or frags. 333 */ 334 for (i = NDADDR - 1; i > lastblock; i--) { 335 register long bsize; 336 337 bn = oip->i_db[i]; 338 if (bn == 0) 339 continue; 340 oip->i_db[i] = 0; 341 bsize = blksize(fs, oip, i); 342 ffs_blkfree(oip, bn, bsize); 343 blocksreleased += btodb(bsize); 344 } 345 if (lastblock < 0) 346 goto done; 347 348 /* 349 * Finally, look for a change in size of the 350 * last direct block; release any frags. 351 */ 352 bn = oip->i_db[lastblock]; 353 if (bn != 0) { 354 long oldspace, newspace; 355 356 /* 357 * Calculate amount of space we're giving 358 * back as old block size minus new block size. 359 */ 360 oldspace = blksize(fs, oip, lastblock); 361 oip->i_size = length; 362 newspace = blksize(fs, oip, lastblock); 363 if (newspace == 0) 364 panic("ffs_truncate: newspace"); 365 if (oldspace - newspace > 0) { 366 /* 367 * Block number of space to be free'd is 368 * the old block # plus the number of frags 369 * required for the storage we're keeping. 370 */ 371 bn += numfrags(fs, newspace); 372 ffs_blkfree(oip, bn, oldspace - newspace); 373 blocksreleased += btodb(oldspace - newspace); 374 } 375 } 376 done: 377 #ifdef DIAGNOSTIC 378 for (level = SINGLE; level <= TRIPLE; level++) 379 if (newblks[NDADDR + level] != oip->i_ib[level]) 380 panic("ffs_truncate1"); 381 for (i = 0; i < NDADDR; i++) 382 if (newblks[i] != oip->i_db[i]) 383 panic("ffs_truncate2"); 384 if (length == 0 && 385 (!TAILQ_EMPTY(&ovp->v_dirtyblkhd) || 386 !TAILQ_EMPTY(&ovp->v_cleanblkhd))) 387 panic("ffs_truncate3"); 388 #endif /* DIAGNOSTIC */ 389 /* 390 * Put back the real size. 391 */ 392 oip->i_size = length; 393 oip->i_blocks -= blocksreleased; 394 395 if (oip->i_blocks < 0) /* sanity */ 396 oip->i_blocks = 0; 397 oip->i_flag |= IN_CHANGE; 398 #ifdef QUOTA 399 (void) chkdq(oip, -blocksreleased, NOCRED, 0); 400 #endif 401 return (allerror); 402 } 403 404 /* 405 * Release blocks associated with the inode ip and stored in the indirect 406 * block bn. Blocks are free'd in LIFO order up to (but not including) 407 * lastbn. If level is greater than SINGLE, the block is an indirect block 408 * and recursive calls to indirtrunc must be used to cleanse other indirect 409 * blocks. 410 * 411 * NB: triple indirect blocks are untested. 412 */ 413 static int 414 ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 415 register struct inode *ip; 416 ufs_daddr_t lbn, lastbn; 417 ufs_daddr_t dbn; 418 int level; 419 long *countp; 420 { 421 register int i; 422 struct buf *bp; 423 register struct fs *fs = ip->i_fs; 424 register ufs_daddr_t *bap; 425 struct vnode *vp; 426 ufs_daddr_t *copy = NULL, nb, nlbn, last; 427 long blkcount, factor; 428 int nblocks, blocksreleased = 0; 429 int error = 0, allerror = 0; 430 431 /* 432 * Calculate index in current block of last 433 * block to be kept. -1 indicates the entire 434 * block so we need not calculate the index. 435 */ 436 factor = 1; 437 for (i = SINGLE; i < level; i++) 438 factor *= NINDIR(fs); 439 last = lastbn; 440 if (lastbn > 0) 441 last /= factor; 442 nblocks = btodb(fs->fs_bsize); 443 /* 444 * Get buffer of block pointers, zero those entries corresponding 445 * to blocks to be free'd, and update on disk copy first. Since 446 * double(triple) indirect before single(double) indirect, calls 447 * to bmap on these blocks will fail. However, we already have 448 * the on disk address, so we have to set the b_blkno field 449 * explicitly instead of letting bread do everything for us. 450 */ 451 vp = ITOV(ip); 452 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0); 453 if ((bp->b_flags & B_CACHE) == 0) { 454 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 455 bp->b_iocmd = BIO_READ; 456 bp->b_flags &= ~B_INVAL; 457 bp->b_ioflags &= ~BIO_ERROR; 458 if (bp->b_bcount > bp->b_bufsize) 459 panic("ffs_indirtrunc: bad buffer size"); 460 bp->b_blkno = dbn; 461 vfs_busy_pages(bp, 0); 462 BUF_STRATEGY(bp); 463 error = bufwait(bp); 464 } 465 if (error) { 466 brelse(bp); 467 *countp = 0; 468 return (error); 469 } 470 471 bap = (ufs_daddr_t *)bp->b_data; 472 if (lastbn != -1) { 473 MALLOC(copy, ufs_daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK); 474 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize); 475 bzero((caddr_t)&bap[last + 1], 476 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t)); 477 if (DOINGASYNC(vp)) { 478 bawrite(bp); 479 } else { 480 error = bwrite(bp); 481 if (error) 482 allerror = error; 483 } 484 bap = copy; 485 } 486 487 /* 488 * Recursively free totally unused blocks. 489 */ 490 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 491 i--, nlbn += factor) { 492 nb = bap[i]; 493 if (nb == 0) 494 continue; 495 if (level > SINGLE) { 496 if ((error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 497 (ufs_daddr_t)-1, level - 1, &blkcount)) != 0) 498 allerror = error; 499 blocksreleased += blkcount; 500 } 501 ffs_blkfree(ip, nb, fs->fs_bsize); 502 blocksreleased += nblocks; 503 } 504 505 /* 506 * Recursively free last partial block. 507 */ 508 if (level > SINGLE && lastbn >= 0) { 509 last = lastbn % factor; 510 nb = bap[i]; 511 if (nb != 0) { 512 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 513 last, level - 1, &blkcount); 514 if (error) 515 allerror = error; 516 blocksreleased += blkcount; 517 } 518 } 519 if (copy != NULL) { 520 FREE(copy, M_TEMP); 521 } else { 522 bp->b_flags |= B_INVAL | B_NOCACHE; 523 brelse(bp); 524 } 525 526 *countp = blocksreleased; 527 return (allerror); 528 } 529