ffs_inode.c revision 1.31
1 /* $NetBSD: ffs_inode.c,v 1.31 2000/05/13 23:43:13 perseant Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 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 8.13 (Berkeley) 4/21/95 36 */ 37 38 #if defined(_KERNEL) && !defined(_LKM) 39 #include "opt_ffs.h" 40 #include "opt_quota.h" 41 #endif 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/mount.h> 46 #include <sys/proc.h> 47 #include <sys/file.h> 48 #include <sys/buf.h> 49 #include <sys/vnode.h> 50 #include <sys/kernel.h> 51 #include <sys/malloc.h> 52 #include <sys/trace.h> 53 #include <sys/resourcevar.h> 54 55 #include <vm/vm.h> 56 57 #include <uvm/uvm_extern.h> 58 59 #include <ufs/ufs/quota.h> 60 #include <ufs/ufs/inode.h> 61 #include <ufs/ufs/ufsmount.h> 62 #include <ufs/ufs/ufs_extern.h> 63 #include <ufs/ufs/ufs_bswap.h> 64 65 #include <ufs/ffs/fs.h> 66 #include <ufs/ffs/ffs_extern.h> 67 68 static int ffs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t, 69 ufs_daddr_t, int, long *)); 70 71 /* 72 * Update the access, modified, and inode change times as specified 73 * by the IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. 74 * The IN_MODIFIED flag is used to specify that the inode needs to be 75 * updated but that the times have already been set. The access 76 * and modified times are taken from the second and third parameters; 77 * the inode change time is always taken from the current time. If 78 * UPDATE_WAIT flag is set, or UPDATE_DIROP is set and we are not doing 79 * softupdates, then wait for the disk write of the inode to complete. 80 */ 81 82 int 83 ffs_update(v) 84 void *v; 85 { 86 struct vop_update_args /* { 87 struct vnode *a_vp; 88 struct timespec *a_access; 89 struct timespec *a_modify; 90 int a_flags; 91 } */ *ap = v; 92 struct fs *fs; 93 struct buf *bp; 94 struct inode *ip; 95 int error; 96 struct timespec ts; 97 caddr_t cp; 98 int waitfor; 99 100 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) 101 return (0); 102 ip = VTOI(ap->a_vp); 103 TIMEVAL_TO_TIMESPEC(&time, &ts); 104 FFS_ITIMES(ip, 105 ap->a_access ? ap->a_access : &ts, 106 ap->a_modify ? ap->a_modify : &ts, &ts); 107 if ((ip->i_flag & IN_MODIFIED) == 0 && (ap->a_flags & UPDATE_WAIT) == 0) 108 return (0); 109 ip->i_flag &= ~IN_MODIFIED; 110 fs = ip->i_fs; 111 112 waitfor=0; 113 if ((ap->a_flags & UPDATE_DIROP) && !DOINGSOFTDEP(ap->a_vp)) 114 waitfor = UPDATE_WAIT; 115 waitfor |= ap->a_flags & UPDATE_WAIT; 116 117 /* 118 * Ensure that uid and gid are correct. This is a temporary 119 * fix until fsck has been changed to do the update. 120 */ 121 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 122 ip->i_din.ffs_din.di_ouid = ip->i_ffs_uid; /* XXX */ 123 ip->i_din.ffs_din.di_ogid = ip->i_ffs_gid; /* XXX */ 124 } /* XXX */ 125 error = bread(ip->i_devvp, 126 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 127 (int)fs->fs_bsize, NOCRED, &bp); 128 if (error) { 129 brelse(bp); 130 return (error); 131 } 132 if (DOINGSOFTDEP(ap->a_vp)) 133 softdep_update_inodeblock(ip, bp, waitfor); 134 else if (ip->i_ffs_effnlink != ip->i_ffs_nlink) 135 panic("ffs_update: bad link cnt"); 136 cp = (caddr_t)bp->b_data + 137 (ino_to_fsbo(fs, ip->i_number) * DINODE_SIZE); 138 #ifdef FFS_EI 139 if (UFS_FSNEEDSWAP(fs)) 140 ffs_dinode_swap(&ip->i_din.ffs_din, (struct dinode *)cp); 141 else 142 #endif 143 memcpy(cp, &ip->i_din.ffs_din, DINODE_SIZE); 144 if (waitfor && (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) { 145 return (bwrite(bp)); 146 } else { 147 bdwrite(bp); 148 return (0); 149 } 150 } 151 152 #define SINGLE 0 /* index of single indirect block */ 153 #define DOUBLE 1 /* index of double indirect block */ 154 #define TRIPLE 2 /* index of triple indirect block */ 155 /* 156 * Truncate the inode oip to at most length size, freeing the 157 * disk blocks. 158 */ 159 int 160 ffs_truncate(v) 161 void *v; 162 { 163 struct vop_truncate_args /* { 164 struct vnode *a_vp; 165 off_t a_length; 166 int a_flags; 167 struct ucred *a_cred; 168 struct proc *a_p; 169 } */ *ap = v; 170 struct vnode *ovp = ap->a_vp; 171 ufs_daddr_t lastblock; 172 struct inode *oip; 173 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 174 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 175 off_t length = ap->a_length; 176 struct fs *fs; 177 struct buf *bp; 178 int offset, size, level; 179 long count, nblocks, vflags, blocksreleased = 0; 180 int i; 181 int aflags, error, allerror; 182 off_t osize; 183 184 if (length < 0) 185 return (EINVAL); 186 oip = VTOI(ovp); 187 #if 1 188 /* 189 * XXX. Was in Kirk's patches. Is it good behavior to just 190 * return and not update modification times? 191 */ 192 if (oip->i_ffs_size == length) 193 return (0); 194 #endif 195 if (ovp->v_type == VLNK && 196 (oip->i_ffs_size < ovp->v_mount->mnt_maxsymlinklen || 197 (ovp->v_mount->mnt_maxsymlinklen == 0 && 198 oip->i_din.ffs_din.di_blocks == 0))) { 199 #ifdef DIAGNOSTIC 200 if (length != 0) 201 panic("ffs_truncate: partial truncate of symlink"); 202 #endif 203 memset((char *)&oip->i_ffs_shortlink, 0, (u_int)oip->i_ffs_size); 204 oip->i_ffs_size = 0; 205 oip->i_flag |= IN_CHANGE | IN_UPDATE; 206 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)); 207 } 208 if (oip->i_ffs_size == length) { 209 oip->i_flag |= IN_CHANGE | IN_UPDATE; 210 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 211 } 212 #ifdef QUOTA 213 if ((error = getinoquota(oip)) != 0) 214 return (error); 215 #endif 216 fs = oip->i_fs; 217 osize = oip->i_ffs_size; 218 ovp->v_lasta = ovp->v_clen = ovp->v_cstart = ovp->v_lastw = 0; 219 220 if (DOINGSOFTDEP(ovp)) { 221 uvm_vnp_setsize(ovp, length); 222 (void) uvm_vnp_uncache(ovp); 223 if (length > 0) { 224 /* 225 * If a file is only partially truncated, then 226 * we have to clean up the data structures 227 * describing the allocation past the truncation 228 * point. Finding and deallocating those structures 229 * is a lot of work. Since partial truncation occurs 230 * rarely, we solve the problem by syncing the file 231 * so that it will have no data structures left. 232 */ 233 if ((error = VOP_FSYNC(ovp, ap->a_cred, FSYNC_WAIT, 234 ap->a_p)) != 0) 235 return (error); 236 } else { 237 #ifdef QUOTA 238 (void) chkdq(oip, -oip->i_ffs_blocks, NOCRED, 0); 239 #endif 240 softdep_setup_freeblocks(oip, length); 241 (void) vinvalbuf(ovp, 0, ap->a_cred, ap->a_p, 0, 0); 242 oip->i_flag |= IN_CHANGE | IN_UPDATE; 243 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 244 } 245 } 246 /* 247 * Lengthen the size of the file. We must ensure that the 248 * last byte of the file is allocated. Since the smallest 249 * value of osize is 0, length will be at least 1. 250 */ 251 if (osize < length) { 252 if (length > fs->fs_maxfilesize) 253 return (EFBIG); 254 aflags = B_CLRBUF; 255 if (ap->a_flags & IO_SYNC) 256 aflags |= B_SYNC; 257 error = VOP_BALLOC(ovp, length - 1, 1, ap->a_cred, aflags, &bp); 258 if (error) 259 return (error); 260 oip->i_ffs_size = length; 261 uvm_vnp_setsize(ovp, length); 262 (void) uvm_vnp_uncache(ovp); 263 if (aflags & B_SYNC) 264 bwrite(bp); 265 else 266 bawrite(bp); 267 oip->i_flag |= IN_CHANGE | IN_UPDATE; 268 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)); 269 } 270 /* 271 * Shorten the size of the file. If the file is not being 272 * truncated to a block boundary, the contents of the 273 * partial block following the end of the file must be 274 * zero'ed in case it ever becomes accessible again because 275 * of subsequent file growth. Directories however are not 276 * zero'ed as they should grow back initialized to empty. 277 */ 278 offset = blkoff(fs, length); 279 if (offset == 0) { 280 oip->i_ffs_size = length; 281 } else { 282 lbn = lblkno(fs, length); 283 aflags = B_CLRBUF; 284 if (ap->a_flags & IO_SYNC) 285 aflags |= B_SYNC; 286 error = VOP_BALLOC(ovp, length - 1, 1, ap->a_cred, aflags, &bp); 287 if (error) 288 return (error); 289 oip->i_ffs_size = length; 290 size = blksize(fs, oip, lbn); 291 (void) uvm_vnp_uncache(ovp); 292 if (ovp->v_type != VDIR) 293 memset((char *)bp->b_data + offset, 0, 294 (u_int)(size - offset)); 295 allocbuf(bp, size); 296 if (aflags & B_SYNC) 297 bwrite(bp); 298 else 299 bawrite(bp); 300 } 301 uvm_vnp_setsize(ovp, length); 302 /* 303 * Calculate index into inode's block list of 304 * last direct and indirect blocks (if any) 305 * which we want to keep. Lastblock is -1 when 306 * the file is truncated to 0. 307 */ 308 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1; 309 lastiblock[SINGLE] = lastblock - NDADDR; 310 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 311 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 312 nblocks = btodb(fs->fs_bsize); 313 /* 314 * Update file and block pointers on disk before we start freeing 315 * blocks. If we crash before free'ing blocks below, the blocks 316 * will be returned to the free list. lastiblock values are also 317 * normalized to -1 for calls to ffs_indirtrunc below. 318 */ 319 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_ffs_db[0], sizeof oldblks); 320 for (level = TRIPLE; level >= SINGLE; level--) 321 if (lastiblock[level] < 0) { 322 oip->i_ffs_ib[level] = 0; 323 lastiblock[level] = -1; 324 } 325 for (i = NDADDR - 1; i > lastblock; i--) 326 oip->i_ffs_db[i] = 0; 327 oip->i_flag |= IN_CHANGE | IN_UPDATE; 328 if ((error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)) != 0) 329 allerror = error; 330 /* 331 * Having written the new inode to disk, save its new configuration 332 * and put back the old block pointers long enough to process them. 333 * Note that we save the new block configuration so we can check it 334 * when we are done. 335 */ 336 memcpy((caddr_t)newblks, (caddr_t)&oip->i_ffs_db[0], sizeof newblks); 337 memcpy((caddr_t)&oip->i_ffs_db[0], (caddr_t)oldblks, sizeof oldblks); 338 oip->i_ffs_size = osize; 339 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA; 340 allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0); 341 342 /* 343 * Indirect blocks first. 344 */ 345 indir_lbn[SINGLE] = -NDADDR; 346 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 347 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 348 for (level = TRIPLE; level >= SINGLE; level--) { 349 bn = ufs_rw32(oip->i_ffs_ib[level], UFS_FSNEEDSWAP(fs)); 350 if (bn != 0) { 351 error = ffs_indirtrunc(oip, indir_lbn[level], 352 fsbtodb(fs, bn), lastiblock[level], level, &count); 353 if (error) 354 allerror = error; 355 blocksreleased += count; 356 if (lastiblock[level] < 0) { 357 oip->i_ffs_ib[level] = 0; 358 ffs_blkfree(oip, bn, fs->fs_bsize); 359 blocksreleased += nblocks; 360 } 361 } 362 if (lastiblock[level] >= 0) 363 goto done; 364 } 365 366 /* 367 * All whole direct blocks or frags. 368 */ 369 for (i = NDADDR - 1; i > lastblock; i--) { 370 long bsize; 371 372 bn = ufs_rw32(oip->i_ffs_db[i], UFS_FSNEEDSWAP(fs)); 373 if (bn == 0) 374 continue; 375 oip->i_ffs_db[i] = 0; 376 bsize = blksize(fs, oip, i); 377 ffs_blkfree(oip, bn, bsize); 378 blocksreleased += btodb(bsize); 379 } 380 if (lastblock < 0) 381 goto done; 382 383 /* 384 * Finally, look for a change in size of the 385 * last direct block; release any frags. 386 */ 387 bn = ufs_rw32(oip->i_ffs_db[lastblock], UFS_FSNEEDSWAP(fs)); 388 if (bn != 0) { 389 long oldspace, newspace; 390 391 /* 392 * Calculate amount of space we're giving 393 * back as old block size minus new block size. 394 */ 395 oldspace = blksize(fs, oip, lastblock); 396 oip->i_ffs_size = length; 397 newspace = blksize(fs, oip, lastblock); 398 if (newspace == 0) 399 panic("itrunc: newspace"); 400 if (oldspace - newspace > 0) { 401 /* 402 * Block number of space to be free'd is 403 * the old block # plus the number of frags 404 * required for the storage we're keeping. 405 */ 406 bn += numfrags(fs, newspace); 407 ffs_blkfree(oip, bn, oldspace - newspace); 408 blocksreleased += btodb(oldspace - newspace); 409 } 410 } 411 done: 412 #ifdef DIAGNOSTIC 413 for (level = SINGLE; level <= TRIPLE; level++) 414 if (newblks[NDADDR + level] != oip->i_ffs_ib[level]) 415 panic("itrunc1"); 416 for (i = 0; i < NDADDR; i++) 417 if (newblks[i] != oip->i_ffs_db[i]) 418 panic("itrunc2"); 419 if (length == 0 && 420 (ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first)) 421 panic("itrunc3"); 422 #endif /* DIAGNOSTIC */ 423 /* 424 * Put back the real size. 425 */ 426 oip->i_ffs_size = length; 427 oip->i_ffs_blocks -= blocksreleased; 428 if (oip->i_ffs_blocks < 0) /* sanity */ 429 oip->i_ffs_blocks = 0; 430 oip->i_flag |= IN_CHANGE; 431 #ifdef QUOTA 432 (void) chkdq(oip, -blocksreleased, NOCRED, 0); 433 #endif 434 return (allerror); 435 } 436 437 /* 438 * Release blocks associated with the inode ip and stored in the indirect 439 * block bn. Blocks are free'd in LIFO order up to (but not including) 440 * lastbn. If level is greater than SINGLE, the block is an indirect block 441 * and recursive calls to indirtrunc must be used to cleanse other indirect 442 * blocks. 443 * 444 * NB: triple indirect blocks are untested. 445 */ 446 static int 447 ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 448 struct inode *ip; 449 ufs_daddr_t lbn, lastbn; 450 ufs_daddr_t dbn; 451 int level; 452 long *countp; 453 { 454 int i; 455 struct buf *bp; 456 struct fs *fs = ip->i_fs; 457 ufs_daddr_t *bap; 458 struct vnode *vp; 459 ufs_daddr_t *copy = NULL, nb, nlbn, last; 460 long blkcount, factor; 461 int nblocks, blocksreleased = 0; 462 int error = 0, allerror = 0; 463 464 /* 465 * Calculate index in current block of last 466 * block to be kept. -1 indicates the entire 467 * block so we need not calculate the index. 468 */ 469 factor = 1; 470 for (i = SINGLE; i < level; i++) 471 factor *= NINDIR(fs); 472 last = lastbn; 473 if (lastbn > 0) 474 last /= factor; 475 nblocks = btodb(fs->fs_bsize); 476 /* 477 * Get buffer of block pointers, zero those entries corresponding 478 * to blocks to be free'd, and update on disk copy first. Since 479 * double(triple) indirect before single(double) indirect, calls 480 * to bmap on these blocks will fail. However, we already have 481 * the on disk address, so we have to set the b_blkno field 482 * explicitly instead of letting bread do everything for us. 483 */ 484 vp = ITOV(ip); 485 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0); 486 if (bp->b_flags & (B_DONE | B_DELWRI)) { 487 /* Braces must be here in case trace evaluates to nothing. */ 488 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn); 489 } else { 490 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn); 491 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 492 bp->b_flags |= B_READ; 493 if (bp->b_bcount > bp->b_bufsize) 494 panic("ffs_indirtrunc: bad buffer size"); 495 bp->b_blkno = dbn; 496 VOP_STRATEGY(bp); 497 error = biowait(bp); 498 } 499 if (error) { 500 brelse(bp); 501 *countp = 0; 502 return (error); 503 } 504 505 bap = (ufs_daddr_t *)bp->b_data; 506 if (lastbn != -1) { 507 MALLOC(copy, ufs_daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK); 508 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->fs_bsize); 509 memset((caddr_t)&bap[last + 1], 0, 510 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t)); 511 error = bwrite(bp); 512 if (error) 513 allerror = error; 514 bap = copy; 515 } 516 517 /* 518 * Recursively free totally unused blocks. 519 */ 520 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 521 i--, nlbn += factor) { 522 nb = ufs_rw32(bap[i], UFS_FSNEEDSWAP(fs)); 523 if (nb == 0) 524 continue; 525 if (level > SINGLE) { 526 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 527 (ufs_daddr_t)-1, level - 1, 528 &blkcount); 529 if (error) 530 allerror = error; 531 blocksreleased += blkcount; 532 } 533 ffs_blkfree(ip, nb, fs->fs_bsize); 534 blocksreleased += nblocks; 535 } 536 537 /* 538 * Recursively free last partial block. 539 */ 540 if (level > SINGLE && lastbn >= 0) { 541 last = lastbn % factor; 542 nb = ufs_rw32(bap[i], UFS_FSNEEDSWAP(fs)); 543 if (nb != 0) { 544 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 545 last, level - 1, &blkcount); 546 if (error) 547 allerror = error; 548 blocksreleased += blkcount; 549 } 550 } 551 552 if (copy != NULL) { 553 FREE(copy, M_TEMP); 554 } else { 555 bp->b_flags |= B_INVAL; 556 brelse(bp); 557 } 558 559 *countp = blocksreleased; 560 return (allerror); 561 } 562
Indexes created Tue Sep 30 17:09:57 GMT 2025