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, td) 134 struct vnode *vp; 135 off_t length; 136 int flags; 137 struct ucred *cred; 138 struct thread *td; 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 td)) != 0) 195 return (error); 196 if (oip->i_flag & IN_SPACECOUNTED) 197 fs->fs_pendingblocks -= oip->i_blocks; 198 } else { 199 #ifdef QUOTA 200 (void) chkdq(oip, -oip->i_blocks, NOCRED, 0); 201 #endif 202 softdep_setup_freeblocks(oip, length); 203 vinvalbuf(ovp, 0, cred, td, 0, 0); 204 oip->i_flag |= IN_CHANGE | IN_UPDATE; 205 return (ffs_update(ovp, 0)); 206 } 207 } 208 osize = oip->i_size; 209 /* 210 * Lengthen the size of the file. We must ensure that the 211 * last byte of the file is allocated. Since the smallest 212 * value of osize is 0, length will be at least 1. 213 */ 214 if (osize < length) { 215 vnode_pager_setsize(ovp, length); 216 aflags = B_CLRBUF; 217 if (flags & IO_SYNC) 218 aflags |= B_SYNC; 219 error = UFS_BALLOC(ovp, length - 1, 1, 220 cred, aflags, &bp); 221 if (error) 222 return (error); 223 oip->i_size = length; 224 if (bp->b_bufsize == fs->fs_bsize) 225 bp->b_flags |= B_CLUSTEROK; 226 if (aflags & B_SYNC) 227 bwrite(bp); 228 else 229 bawrite(bp); 230 oip->i_flag |= IN_CHANGE | IN_UPDATE; 231 return (UFS_UPDATE(ovp, 1)); 232 } 233 /* 234 * Shorten the size of the file. If the file is not being 235 * truncated to a block boundary, the contents of the 236 * partial block following the end of the file must be 237 * zero'ed in case it ever becomes accessible again because 238 * of subsequent file growth. Directories however are not 239 * zero'ed as they should grow back initialized to empty. 240 */ 241 offset = blkoff(fs, length); 242 if (offset == 0) { 243 oip->i_size = length; 244 } else { 245 lbn = lblkno(fs, length); 246 aflags = B_CLRBUF; 247 if (flags & IO_SYNC) 248 aflags |= B_SYNC; 249 error = UFS_BALLOC(ovp, length - 1, 1, cred, aflags, &bp); 250 if (error) { 251 return (error); 252 } 253 /* 254 * When we are doing soft updates and the UFS_BALLOC 255 * above fills in a direct block hole with a full sized 256 * block that will be truncated down to a fragment below, 257 * we must flush out the block dependency with an FSYNC 258 * so that we do not get a soft updates inconsistency 259 * when we create the fragment below. 260 */ 261 if (DOINGSOFTDEP(ovp) && lbn < NDADDR && 262 fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize && 263 (error = VOP_FSYNC(ovp, cred, MNT_WAIT, td)) != 0) 264 return (error); 265 oip->i_size = length; 266 size = blksize(fs, oip, lbn); 267 if (ovp->v_type != VDIR) 268 bzero((char *)bp->b_data + offset, 269 (u_int)(size - offset)); 270 /* Kirk's code has reallocbuf(bp, size, 1) here */ 271 allocbuf(bp, size); 272 if (bp->b_bufsize == fs->fs_bsize) 273 bp->b_flags |= B_CLUSTEROK; 274 if (aflags & B_SYNC) 275 bwrite(bp); 276 else 277 bawrite(bp); 278 } 279 /* 280 * Calculate index into inode's block list of 281 * last direct and indirect blocks (if any) 282 * which we want to keep. Lastblock is -1 when 283 * the file is truncated to 0. 284 */ 285 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1; 286 lastiblock[SINGLE] = lastblock - NDADDR; 287 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 288 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 289 nblocks = btodb(fs->fs_bsize); 290 /* 291 * Update file and block pointers on disk before we start freeing 292 * blocks. If we crash before free'ing blocks below, the blocks 293 * will be returned to the free list. lastiblock values are also 294 * normalized to -1 for calls to ffs_indirtrunc below. 295 */ 296 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks); 297 for (level = TRIPLE; level >= SINGLE; level--) 298 if (lastiblock[level] < 0) { 299 oip->i_ib[level] = 0; 300 lastiblock[level] = -1; 301 } 302 for (i = NDADDR - 1; i > lastblock; i--) 303 oip->i_db[i] = 0; 304 oip->i_flag |= IN_CHANGE | IN_UPDATE; 305 allerror = UFS_UPDATE(ovp, 1); 306 307 /* 308 * Having written the new inode to disk, save its new configuration 309 * and put back the old block pointers long enough to process them. 310 * Note that we save the new block configuration so we can check it 311 * when we are done. 312 */ 313 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks); 314 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks); 315 oip->i_size = osize; 316 317 error = vtruncbuf(ovp, cred, td, length, fs->fs_bsize); 318 if (error && (allerror == 0)) 319 allerror = error; 320 321 /* 322 * Indirect blocks first. 323 */ 324 indir_lbn[SINGLE] = -NDADDR; 325 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 326 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 327 for (level = TRIPLE; level >= SINGLE; level--) { 328 bn = oip->i_ib[level]; 329 if (bn != 0) { 330 error = ffs_indirtrunc(oip, indir_lbn[level], 331 fsbtodb(fs, bn), lastiblock[level], level, &count); 332 if (error) 333 allerror = error; 334 blocksreleased += count; 335 if (lastiblock[level] < 0) { 336 oip->i_ib[level] = 0; 337 ffs_blkfree(oip, bn, fs->fs_bsize); 338 blocksreleased += nblocks; 339 } 340 } 341 if (lastiblock[level] >= 0) 342 goto done; 343 } 344 345 /* 346 * All whole direct blocks or frags. 347 */ 348 for (i = NDADDR - 1; i > lastblock; i--) { 349 register long bsize; 350 351 bn = oip->i_db[i]; 352 if (bn == 0) 353 continue; 354 oip->i_db[i] = 0; 355 bsize = blksize(fs, oip, i); 356 ffs_blkfree(oip, bn, bsize); 357 blocksreleased += btodb(bsize); 358 } 359 if (lastblock < 0) 360 goto done; 361 362 /* 363 * Finally, look for a change in size of the 364 * last direct block; release any frags. 365 */ 366 bn = oip->i_db[lastblock]; 367 if (bn != 0) { 368 long oldspace, newspace; 369 370 /* 371 * Calculate amount of space we're giving 372 * back as old block size minus new block size. 373 */ 374 oldspace = blksize(fs, oip, lastblock); 375 oip->i_size = length; 376 newspace = blksize(fs, oip, lastblock); 377 if (newspace == 0) 378 panic("ffs_truncate: newspace"); 379 if (oldspace - newspace > 0) { 380 /* 381 * Block number of space to be free'd is 382 * the old block # plus the number of frags 383 * required for the storage we're keeping. 384 */ 385 bn += numfrags(fs, newspace); 386 ffs_blkfree(oip, bn, oldspace - newspace); 387 blocksreleased += btodb(oldspace - newspace); 388 } 389 } 390 done: 391 #ifdef DIAGNOSTIC 392 for (level = SINGLE; level <= TRIPLE; level++) 393 if (newblks[NDADDR + level] != oip->i_ib[level]) 394 panic("ffs_truncate1"); 395 for (i = 0; i < NDADDR; i++) 396 if (newblks[i] != oip->i_db[i]) 397 panic("ffs_truncate2"); 398 if (length == 0 && 399 (!TAILQ_EMPTY(&ovp->v_dirtyblkhd) || 400 !TAILQ_EMPTY(&ovp->v_cleanblkhd))) 401 panic("ffs_truncate3"); 402 #endif /* DIAGNOSTIC */ 403 /* 404 * Put back the real size. 405 */ 406 oip->i_size = length; 407 oip->i_blocks -= blocksreleased; 408 409 if (oip->i_blocks < 0) /* sanity */ 410 oip->i_blocks = 0; 411 oip->i_flag |= IN_CHANGE; 412 #ifdef QUOTA 413 (void) chkdq(oip, -blocksreleased, NOCRED, 0); 414 #endif 415 return (allerror); 416 } 417 418 /* 419 * Release blocks associated with the inode ip and stored in the indirect 420 * block bn. Blocks are free'd in LIFO order up to (but not including) 421 * lastbn. If level is greater than SINGLE, the block is an indirect block 422 * and recursive calls to indirtrunc must be used to cleanse other indirect 423 * blocks. 424 * 425 * NB: triple indirect blocks are untested. 426 */ 427 static int 428 ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 429 register struct inode *ip; 430 ufs_daddr_t lbn, lastbn; 431 ufs_daddr_t dbn; 432 int level; 433 long *countp; 434 { 435 register int i; 436 struct buf *bp; 437 register struct fs *fs = ip->i_fs; 438 register ufs_daddr_t *bap; 439 struct vnode *vp; 440 ufs_daddr_t *copy = NULL, nb, nlbn, last; 441 long blkcount, factor; 442 int nblocks, blocksreleased = 0; 443 int error = 0, allerror = 0; 444 445 /* 446 * Calculate index in current block of last 447 * block to be kept. -1 indicates the entire 448 * block so we need not calculate the index. 449 */ 450 factor = 1; 451 for (i = SINGLE; i < level; i++) 452 factor *= NINDIR(fs); 453 last = lastbn; 454 if (lastbn > 0) 455 last /= factor; 456 nblocks = btodb(fs->fs_bsize); 457 /* 458 * Get buffer of block pointers, zero those entries corresponding 459 * to blocks to be free'd, and update on disk copy first. Since 460 * double(triple) indirect before single(double) indirect, calls 461 * to bmap on these blocks will fail. However, we already have 462 * the on disk address, so we have to set the b_blkno field 463 * explicitly instead of letting bread do everything for us. 464 */ 465 vp = ITOV(ip); 466 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0); 467 if ((bp->b_flags & B_CACHE) == 0) { 468 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 469 bp->b_iocmd = BIO_READ; 470 bp->b_flags &= ~B_INVAL; 471 bp->b_ioflags &= ~BIO_ERROR; 472 if (bp->b_bcount > bp->b_bufsize) 473 panic("ffs_indirtrunc: bad buffer size"); 474 bp->b_blkno = dbn; 475 vfs_busy_pages(bp, 0); 476 BUF_STRATEGY(bp); 477 error = bufwait(bp); 478 } 479 if (error) { 480 brelse(bp); 481 *countp = 0; 482 return (error); 483 } 484 485 bap = (ufs_daddr_t *)bp->b_data; 486 if (lastbn != -1) { 487 MALLOC(copy, ufs_daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK); 488 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize); 489 bzero((caddr_t)&bap[last + 1], 490 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t)); 491 if (DOINGASYNC(vp)) { 492 bawrite(bp); 493 } else { 494 error = bwrite(bp); 495 if (error) 496 allerror = error; 497 } 498 bap = copy; 499 } 500 501 /* 502 * Recursively free totally unused blocks. 503 */ 504 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 505 i--, nlbn += factor) { 506 nb = bap[i]; 507 if (nb == 0) 508 continue; 509 if (level > SINGLE) { 510 if ((error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 511 (ufs_daddr_t)-1, level - 1, &blkcount)) != 0) 512 allerror = error; 513 blocksreleased += blkcount; 514 } 515 ffs_blkfree(ip, nb, fs->fs_bsize); 516 blocksreleased += nblocks; 517 } 518 519 /* 520 * Recursively free last partial block. 521 */ 522 if (level > SINGLE && lastbn >= 0) { 523 last = lastbn % factor; 524 nb = bap[i]; 525 if (nb != 0) { 526 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 527 last, level - 1, &blkcount); 528 if (error) 529 allerror = error; 530 blocksreleased += blkcount; 531 } 532 } 533 if (copy != NULL) { 534 FREE(copy, M_TEMP); 535 } else { 536 bp->b_flags |= B_INVAL | B_NOCACHE; 537 brelse(bp); 538 } 539 540 *countp = blocksreleased; 541 return (allerror); 542 } 543