ffs_inode.c revision 1.45
1 /* $NetBSD: ffs_inode.c,v 1.45 2001/09/28 11:43:23 chs 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_OPT) 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 <ufs/ufs/quota.h> 56 #include <ufs/ufs/inode.h> 57 #include <ufs/ufs/ufsmount.h> 58 #include <ufs/ufs/ufs_extern.h> 59 #include <ufs/ufs/ufs_bswap.h> 60 61 #include <ufs/ffs/fs.h> 62 #include <ufs/ffs/ffs_extern.h> 63 64 static int ffs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t, 65 ufs_daddr_t, int, long *)); 66 67 /* 68 * Update the access, modified, and inode change times as specified 69 * by the IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. 70 * The IN_MODIFIED flag is used to specify that the inode needs to be 71 * updated but that the times have already been set. The access 72 * and modified times are taken from the second and third parameters; 73 * the inode change time is always taken from the current time. If 74 * UPDATE_WAIT flag is set, or UPDATE_DIROP is set and we are not doing 75 * softupdates, then wait for the disk write of the inode to complete. 76 */ 77 78 int 79 ffs_update(v) 80 void *v; 81 { 82 struct vop_update_args /* { 83 struct vnode *a_vp; 84 struct timespec *a_access; 85 struct timespec *a_modify; 86 int a_flags; 87 } */ *ap = v; 88 struct fs *fs; 89 struct buf *bp; 90 struct inode *ip; 91 int error; 92 struct timespec ts; 93 caddr_t cp; 94 int waitfor, flags; 95 96 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) 97 return (0); 98 ip = VTOI(ap->a_vp); 99 TIMEVAL_TO_TIMESPEC(&time, &ts); 100 FFS_ITIMES(ip, 101 ap->a_access ? ap->a_access : &ts, 102 ap->a_modify ? ap->a_modify : &ts, &ts); 103 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED); 104 if (flags == 0) 105 return (0); 106 fs = ip->i_fs; 107 108 if ((flags & IN_MODIFIED) != 0 && 109 (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) { 110 waitfor = ap->a_flags & UPDATE_WAIT; 111 if ((ap->a_flags & UPDATE_DIROP) && !DOINGSOFTDEP(ap->a_vp)) 112 waitfor |= UPDATE_WAIT; 113 } else 114 waitfor = 0; 115 116 /* 117 * Ensure that uid and gid are correct. This is a temporary 118 * fix until fsck has been changed to do the update. 119 */ 120 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 121 ip->i_din.ffs_din.di_ouid = ip->i_ffs_uid; /* XXX */ 122 ip->i_din.ffs_din.di_ogid = ip->i_ffs_gid; /* XXX */ 123 } /* XXX */ 124 error = bread(ip->i_devvp, 125 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 126 (int)fs->fs_bsize, NOCRED, &bp); 127 if (error) { 128 brelse(bp); 129 return (error); 130 } 131 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED); 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) { 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 struct genfs_node *gp = VTOG(ovp); 172 ufs_daddr_t lastblock; 173 struct inode *oip; 174 ufs_daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR]; 175 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 176 off_t length = ap->a_length; 177 struct fs *fs; 178 int offset, size, level; 179 long count, nblocks, blocksreleased = 0; 180 int i; 181 int error, allerror = 0; 182 off_t osize; 183 184 if (length < 0) 185 return (EINVAL); 186 oip = VTOI(ovp); 187 if (ovp->v_type == VLNK && 188 (oip->i_ffs_size < ovp->v_mount->mnt_maxsymlinklen || 189 (ovp->v_mount->mnt_maxsymlinklen == 0 && 190 oip->i_din.ffs_din.di_blocks == 0))) { 191 KDASSERT(length == 0); 192 memset(&oip->i_ffs_shortlink, 0, (size_t)oip->i_ffs_size); 193 oip->i_ffs_size = 0; 194 oip->i_flag |= IN_CHANGE | IN_UPDATE; 195 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)); 196 } 197 if (oip->i_ffs_size == length) { 198 oip->i_flag |= IN_CHANGE | IN_UPDATE; 199 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 200 } 201 #ifdef QUOTA 202 if ((error = getinoquota(oip)) != 0) 203 return (error); 204 #endif 205 fs = oip->i_fs; 206 if (length > fs->fs_maxfilesize) 207 return (EFBIG); 208 209 osize = oip->i_ffs_size; 210 211 /* 212 * Lengthen the size of the file. We must ensure that the 213 * last byte of the file is allocated. Since the smallest 214 * value of osize is 0, length will be at least 1. 215 */ 216 217 if (osize < length) { 218 error = ufs_balloc_range(ovp, length - 1, 1, ap->a_cred, 219 ap->a_flags & IO_SYNC ? B_SYNC : 0); 220 if (error) { 221 return error; 222 } 223 uvm_vnp_setsize(ovp, length); 224 oip->i_flag |= IN_CHANGE | IN_UPDATE; 225 KASSERT(ovp->v_size == oip->i_ffs_size); 226 return (VOP_UPDATE(ovp, NULL, NULL, 1)); 227 } 228 229 /* 230 * When truncating a regular file down to a non-block-aligned size, 231 * we must zero the part of last block which is past the new EOF. 232 * We must synchronously flush the zeroed pages to disk 233 * since the new pages will be invalidated as soon as we 234 * inform the VM system of the new, smaller size. 235 * We must to this before acquiring the GLOCK, since fetching 236 * the pages will acquire the GLOCK internally. 237 * So there is a window where another thread could see a whole 238 * zeroed page past EOF, but that's life. 239 */ 240 241 offset = blkoff(fs, length); 242 if (ovp->v_type == VREG && length < osize && offset != 0) { 243 struct uvm_object *uobj; 244 voff_t eoz; 245 246 size = blksize(fs, oip, lblkno(fs, length)); 247 eoz = MIN(lblktosize(fs, lblkno(fs, length)) + size, osize); 248 uvm_vnp_zerorange(ovp, length, eoz - length); 249 uobj = &ovp->v_uobj; 250 simple_lock(&uobj->vmobjlock); 251 error = (uobj->pgops->pgo_put)(uobj, trunc_page(length), 252 round_page(eoz), PGO_CLEANIT|PGO_DEACTIVATE|PGO_SYNCIO); 253 if (error) { 254 return error; 255 } 256 } 257 258 lockmgr(&gp->g_glock, LK_EXCLUSIVE, NULL); 259 260 if (DOINGSOFTDEP(ovp)) { 261 if (length > 0) { 262 /* 263 * If a file is only partially truncated, then 264 * we have to clean up the data structures 265 * describing the allocation past the truncation 266 * point. Finding and deallocating those structures 267 * is a lot of work. Since partial truncation occurs 268 * rarely, we solve the problem by syncing the file 269 * so that it will have no data structures left. 270 */ 271 if ((error = VOP_FSYNC(ovp, ap->a_cred, FSYNC_WAIT, 272 0, 0, ap->a_p)) != 0) { 273 lockmgr(&gp->g_glock, LK_RELEASE, NULL); 274 return (error); 275 } 276 } else { 277 uvm_vnp_setsize(ovp, length); 278 #ifdef QUOTA 279 (void) chkdq(oip, -oip->i_ffs_blocks, NOCRED, 0); 280 #endif 281 softdep_setup_freeblocks(oip, length); 282 (void) vinvalbuf(ovp, 0, ap->a_cred, ap->a_p, 0, 0); 283 lockmgr(&gp->g_glock, LK_RELEASE, NULL); 284 oip->i_flag |= IN_CHANGE | IN_UPDATE; 285 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 286 } 287 } 288 289 /* 290 * Reduce the size of the file. 291 */ 292 oip->i_ffs_size = length; 293 uvm_vnp_setsize(ovp, length); 294 /* 295 * Calculate index into inode's block list of 296 * last direct and indirect blocks (if any) 297 * which we want to keep. Lastblock is -1 when 298 * the file is truncated to 0. 299 */ 300 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1; 301 lastiblock[SINGLE] = lastblock - NDADDR; 302 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 303 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 304 nblocks = btodb(fs->fs_bsize); 305 /* 306 * Update file and block pointers on disk before we start freeing 307 * blocks. If we crash before free'ing blocks below, the blocks 308 * will be returned to the free list. lastiblock values are also 309 * normalized to -1 for calls to ffs_indirtrunc below. 310 */ 311 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_ffs_db[0], sizeof oldblks); 312 for (level = TRIPLE; level >= SINGLE; level--) 313 if (lastiblock[level] < 0) { 314 oip->i_ffs_ib[level] = 0; 315 lastiblock[level] = -1; 316 } 317 for (i = NDADDR - 1; i > lastblock; i--) 318 oip->i_ffs_db[i] = 0; 319 oip->i_flag |= IN_CHANGE | IN_UPDATE; 320 error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT); 321 if (error && !allerror) 322 allerror = error; 323 324 /* 325 * Having written the new inode to disk, save its new configuration 326 * and put back the old block pointers long enough to process them. 327 * Note that we save the new block configuration so we can check it 328 * when we are done. 329 */ 330 memcpy((caddr_t)newblks, (caddr_t)&oip->i_ffs_db[0], sizeof newblks); 331 memcpy((caddr_t)&oip->i_ffs_db[0], (caddr_t)oldblks, sizeof oldblks); 332 oip->i_ffs_size = osize; 333 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 334 if (error && !allerror) 335 allerror = error; 336 337 /* 338 * Indirect blocks first. 339 */ 340 indir_lbn[SINGLE] = -NDADDR; 341 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 342 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 343 for (level = TRIPLE; level >= SINGLE; level--) { 344 bn = ufs_rw32(oip->i_ffs_ib[level], UFS_FSNEEDSWAP(fs)); 345 if (bn != 0) { 346 error = ffs_indirtrunc(oip, indir_lbn[level], 347 fsbtodb(fs, bn), lastiblock[level], level, &count); 348 if (error) 349 allerror = error; 350 blocksreleased += count; 351 if (lastiblock[level] < 0) { 352 oip->i_ffs_ib[level] = 0; 353 ffs_blkfree(oip, bn, fs->fs_bsize); 354 blocksreleased += nblocks; 355 } 356 } 357 if (lastiblock[level] >= 0) 358 goto done; 359 } 360 361 /* 362 * All whole direct blocks or frags. 363 */ 364 for (i = NDADDR - 1; i > lastblock; i--) { 365 long bsize; 366 367 bn = ufs_rw32(oip->i_ffs_db[i], UFS_FSNEEDSWAP(fs)); 368 if (bn == 0) 369 continue; 370 oip->i_ffs_db[i] = 0; 371 bsize = blksize(fs, oip, i); 372 ffs_blkfree(oip, bn, bsize); 373 blocksreleased += btodb(bsize); 374 } 375 if (lastblock < 0) 376 goto done; 377 378 /* 379 * Finally, look for a change in size of the 380 * last direct block; release any frags. 381 */ 382 bn = ufs_rw32(oip->i_ffs_db[lastblock], UFS_FSNEEDSWAP(fs)); 383 if (bn != 0) { 384 long oldspace, newspace; 385 386 /* 387 * Calculate amount of space we're giving 388 * back as old block size minus new block size. 389 */ 390 oldspace = blksize(fs, oip, lastblock); 391 oip->i_ffs_size = length; 392 newspace = blksize(fs, oip, lastblock); 393 if (newspace == 0) 394 panic("itrunc: newspace"); 395 if (oldspace - newspace > 0) { 396 /* 397 * Block number of space to be free'd is 398 * the old block # plus the number of frags 399 * required for the storage we're keeping. 400 */ 401 bn += numfrags(fs, newspace); 402 ffs_blkfree(oip, bn, oldspace - newspace); 403 blocksreleased += btodb(oldspace - newspace); 404 } 405 } 406 407 done: 408 #ifdef DIAGNOSTIC 409 for (level = SINGLE; level <= TRIPLE; level++) 410 if (newblks[NDADDR + level] != oip->i_ffs_ib[level]) 411 panic("itrunc1"); 412 for (i = 0; i < NDADDR; i++) 413 if (newblks[i] != oip->i_ffs_db[i]) 414 panic("itrunc2"); 415 if (length == 0 && 416 (!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd))) 417 panic("itrunc3"); 418 #endif /* DIAGNOSTIC */ 419 /* 420 * Put back the real size. 421 */ 422 oip->i_ffs_size = length; 423 oip->i_ffs_blocks -= blocksreleased; 424 if (oip->i_ffs_blocks < 0) /* sanity */ 425 oip->i_ffs_blocks = 0; 426 lockmgr(&gp->g_glock, LK_RELEASE, NULL); 427 oip->i_flag |= IN_CHANGE; 428 #ifdef QUOTA 429 (void) chkdq(oip, -blocksreleased, NOCRED, 0); 430 #endif 431 KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_ffs_size); 432 return (allerror); 433 } 434 435 /* 436 * Release blocks associated with the inode ip and stored in the indirect 437 * block bn. Blocks are free'd in LIFO order up to (but not including) 438 * lastbn. If level is greater than SINGLE, the block is an indirect block 439 * and recursive calls to indirtrunc must be used to cleanse other indirect 440 * blocks. 441 * 442 * NB: triple indirect blocks are untested. 443 */ 444 static int 445 ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 446 struct inode *ip; 447 ufs_daddr_t lbn, lastbn; 448 ufs_daddr_t dbn; 449 int level; 450 long *countp; 451 { 452 int i; 453 struct buf *bp; 454 struct fs *fs = ip->i_fs; 455 ufs_daddr_t *bap; 456 struct vnode *vp; 457 ufs_daddr_t *copy = NULL, nb, nlbn, last; 458 long blkcount, factor; 459 int nblocks, blocksreleased = 0; 460 int error = 0, allerror = 0; 461 462 /* 463 * Calculate index in current block of last 464 * block to be kept. -1 indicates the entire 465 * block so we need not calculate the index. 466 */ 467 factor = 1; 468 for (i = SINGLE; i < level; i++) 469 factor *= NINDIR(fs); 470 last = lastbn; 471 if (lastbn > 0) 472 last /= factor; 473 nblocks = btodb(fs->fs_bsize); 474 /* 475 * Get buffer of block pointers, zero those entries corresponding 476 * to blocks to be free'd, and update on disk copy first. Since 477 * double(triple) indirect before single(double) indirect, calls 478 * to bmap on these blocks will fail. However, we already have 479 * the on disk address, so we have to set the b_blkno field 480 * explicitly instead of letting bread do everything for us. 481 */ 482 vp = ITOV(ip); 483 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0); 484 if (bp->b_flags & (B_DONE | B_DELWRI)) { 485 /* Braces must be here in case trace evaluates to nothing. */ 486 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn); 487 } else { 488 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn); 489 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 490 bp->b_flags |= B_READ; 491 if (bp->b_bcount > bp->b_bufsize) 492 panic("ffs_indirtrunc: bad buffer size"); 493 bp->b_blkno = dbn; 494 VOP_STRATEGY(bp); 495 error = biowait(bp); 496 } 497 if (error) { 498 brelse(bp); 499 *countp = 0; 500 return (error); 501 } 502 503 bap = (ufs_daddr_t *)bp->b_data; 504 if (lastbn >= 0) { 505 copy = (ufs_daddr_t *) malloc(fs->fs_bsize, M_TEMP, M_WAITOK); 506 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->fs_bsize); 507 memset((caddr_t)&bap[last + 1], 0, 508 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t)); 509 error = bwrite(bp); 510 if (error) 511 allerror = error; 512 bap = copy; 513 } 514 515 /* 516 * Recursively free totally unused blocks. 517 */ 518 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 519 i--, nlbn += factor) { 520 nb = ufs_rw32(bap[i], UFS_FSNEEDSWAP(fs)); 521 if (nb == 0) 522 continue; 523 if (level > SINGLE) { 524 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 525 (ufs_daddr_t)-1, level - 1, 526 &blkcount); 527 if (error) 528 allerror = error; 529 blocksreleased += blkcount; 530 } 531 ffs_blkfree(ip, nb, fs->fs_bsize); 532 blocksreleased += nblocks; 533 } 534 535 /* 536 * Recursively free last partial block. 537 */ 538 if (level > SINGLE && lastbn >= 0) { 539 last = lastbn % factor; 540 nb = ufs_rw32(bap[i], UFS_FSNEEDSWAP(fs)); 541 if (nb != 0) { 542 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 543 last, level - 1, &blkcount); 544 if (error) 545 allerror = error; 546 blocksreleased += blkcount; 547 } 548 } 549 550 if (copy != NULL) { 551 FREE(copy, M_TEMP); 552 } else { 553 bp->b_flags |= B_INVAL; 554 brelse(bp); 555 } 556 557 *countp = blocksreleased; 558 return (allerror); 559 } 560
Indexes created Tue Sep 30 07:10:03 GMT 2025