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 8.5 (Berkeley) 12/30/93 34 * $Id: ffs_inode.c,v 1.19 1996/01/05 18:31:48 wollman Exp $ 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/file.h> 44 #include <sys/buf.h> 45 #include <sys/vnode.h> 46 #include <sys/kernel.h> 47 #include <sys/malloc.h> 48 #include <sys/resourcevar.h> 49 50 #include <vm/vm.h> 51 #include <vm/vm_param.h> 52 #include <vm/vm_extern.h> 53 54 #include <ufs/ufs/quota.h> 55 #include <ufs/ufs/inode.h> 56 #include <ufs/ufs/ufsmount.h> 57 #include <ufs/ufs/ufs_extern.h> 58 59 #include <ufs/ffs/fs.h> 60 #include <ufs/ffs/ffs_extern.h> 61 62 static int ffs_indirtrunc __P((struct inode *, daddr_t, daddr_t, daddr_t, int, 63 long *)); 64 65 int 66 ffs_init() 67 { 68 return (ufs_init()); 69 } 70 71 /* 72 * Update the access, modified, and inode change times as specified by the 73 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. The IN_MODIFIED 74 * flag is used to specify that the inode needs to be updated even if none 75 * of the times needs to be updated. The access and modified times are taken 76 * from the second and third parameters; the inode change time is always 77 * taken from the current time. If waitfor is set, then wait for the disk 78 * write of the inode to complete. 79 */ 80 int 81 ffs_update(ap) 82 struct vop_update_args /* { 83 struct vnode *a_vp; 84 struct timeval *a_access; 85 struct timeval *a_modify; 86 int a_waitfor; 87 } */ *ap; 88 { 89 register struct fs *fs; 90 struct buf *bp; 91 struct inode *ip; 92 int error; 93 time_t tv_sec; 94 95 ip = VTOI(ap->a_vp); 96 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) { 97 ip->i_flag &= 98 ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE); 99 return (0); 100 } 101 if ((ip->i_flag & 102 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0) 103 return (0); 104 /* 105 * Use a copy of the current time to get consistent timestamps 106 * (a_access and a_modify are sometimes aliases for &time). 107 * 108 * XXX in 2.0, a_access and a_modify are often pointers to the 109 * same copy of `time'. This is not as good. Some callers forget 110 * to make a copy; others make a copy too early (before the i/o 111 * has completed)... 112 * 113 * XXX there should be a function or macro for reading the time 114 * (e.g., some machines may require splclock()). 115 */ 116 tv_sec = time.tv_sec; 117 if (ip->i_flag & IN_ACCESS) 118 ip->i_atime.ts_sec = 119 (ap->a_access == &time ? tv_sec : ap->a_access->tv_sec); 120 if (ip->i_flag & IN_UPDATE) { 121 ip->i_mtime.ts_sec = 122 (ap->a_modify == &time ? tv_sec : ap->a_modify->tv_sec); 123 ip->i_modrev++; 124 } 125 if (ip->i_flag & IN_CHANGE) 126 ip->i_ctime.ts_sec = tv_sec; 127 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE); 128 fs = ip->i_fs; 129 /* 130 * Ensure that uid and gid are correct. This is a temporary 131 * fix until fsck has been changed to do the update. 132 */ 133 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 134 ip->i_din.di_ouid = ip->i_uid; /* XXX */ 135 ip->i_din.di_ogid = ip->i_gid; /* XXX */ 136 } /* XXX */ 137 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 138 (int)fs->fs_bsize, NOCRED, &bp); 139 if (error) { 140 brelse(bp); 141 return (error); 142 } 143 *((struct dinode *)bp->b_data + 144 ino_to_fsbo(fs, ip->i_number)) = ip->i_din; 145 if (ap->a_waitfor && (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 146 return (bwrite(bp)); 147 else { 148 bp->b_flags |= B_CLUSTEROK; 149 bdwrite(bp); 150 return (0); 151 } 152 } 153 154 #define SINGLE 0 /* index of single indirect block */ 155 #define DOUBLE 1 /* index of double indirect block */ 156 #define TRIPLE 2 /* index of triple indirect block */ 157 /* 158 * Truncate the inode oip to at most length size, freeing the 159 * disk blocks. 160 */ 161 int 162 ffs_truncate(ap) 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; 170 { 171 register struct vnode *ovp = ap->a_vp; 172 register daddr_t lastblock; 173 register struct inode *oip; 174 daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 175 daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 176 off_t length = ap->a_length; 177 register struct fs *fs; 178 struct buf *bp; 179 int offset, size, level; 180 long count, nblocks, vflags, blocksreleased = 0; 181 struct timeval tv; 182 register int i; 183 int aflags, error, allerror; 184 off_t osize; 185 186 oip = VTOI(ovp); 187 fs = oip->i_fs; 188 if (length < 0 || length > fs->fs_maxfilesize) 189 return (EINVAL); 190 tv = time; 191 if (ovp->v_type == VLNK && 192 (oip->i_size < ovp->v_mount->mnt_maxsymlinklen || oip->i_din.di_blocks == 0)) { 193 #ifdef DIAGNOSTIC 194 if (length != 0) 195 panic("ffs_truncate: partial truncate of symlink"); 196 #endif 197 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size); 198 oip->i_size = 0; 199 oip->i_flag |= IN_CHANGE | IN_UPDATE; 200 return (VOP_UPDATE(ovp, &tv, &tv, 1)); 201 } 202 if (oip->i_size == length) { 203 oip->i_flag |= IN_CHANGE | IN_UPDATE; 204 return (VOP_UPDATE(ovp, &tv, &tv, 0)); 205 } 206 #ifdef QUOTA 207 error = getinoquota(oip); 208 if (error) 209 return (error); 210 #endif 211 osize = oip->i_size; 212 /* 213 * Lengthen the size of the file. We must ensure that the 214 * last byte of the file is allocated. Since the smallest 215 * value of osize is 0, length will be at least 1. 216 */ 217 if (osize < length) { 218 offset = blkoff(fs, length - 1); 219 lbn = lblkno(fs, length - 1); 220 aflags = B_CLRBUF; 221 if (ap->a_flags & IO_SYNC) 222 aflags |= B_SYNC; 223 vnode_pager_setsize(ovp, length); 224 error = ffs_balloc(oip, lbn, offset + 1, ap->a_cred, 225 &bp, aflags); 226 if (error) 227 return (error); 228 oip->i_size = length; 229 if (aflags & B_SYNC) 230 bwrite(bp); 231 else if (ovp->v_mount->mnt_flag & MNT_ASYNC) 232 bdwrite(bp); 233 else 234 bawrite(bp); 235 oip->i_flag |= IN_CHANGE | IN_UPDATE; 236 return (VOP_UPDATE(ovp, &tv, &tv, 1)); 237 } 238 /* 239 * Shorten the size of the file. If the file is not being 240 * truncated to a block boundry, the contents of the 241 * partial block following the end of the file must be 242 * zero'ed in case it ever become accessable again because 243 * of subsequent file growth. 244 */ 245 offset = blkoff(fs, length); 246 if (offset == 0) { 247 oip->i_size = length; 248 } else { 249 lbn = lblkno(fs, length); 250 aflags = B_CLRBUF; 251 if (ap->a_flags & IO_SYNC) 252 aflags |= B_SYNC; 253 error = ffs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags); 254 if (error) 255 return (error); 256 oip->i_size = length; 257 size = blksize(fs, oip, lbn); 258 bzero((char *)bp->b_data + offset, (u_int)(size - offset)); 259 allocbuf(bp, size); 260 if (aflags & B_SYNC) 261 bwrite(bp); 262 else if (ovp->v_mount->mnt_flag & MNT_ASYNC) 263 bdwrite(bp); 264 else 265 bawrite(bp); 266 } 267 /* 268 * Calculate index into inode's block list of 269 * last direct and indirect blocks (if any) 270 * which we want to keep. Lastblock is -1 when 271 * the file is truncated to 0. 272 */ 273 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1; 274 lastiblock[SINGLE] = lastblock - NDADDR; 275 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 276 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 277 nblocks = btodb(fs->fs_bsize); 278 /* 279 * Update file and block pointers on disk before we start freeing 280 * blocks. If we crash before free'ing blocks below, the blocks 281 * will be returned to the free list. lastiblock values are also 282 * normalized to -1 for calls to ffs_indirtrunc below. 283 */ 284 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks); 285 for (level = TRIPLE; level >= SINGLE; level--) 286 if (lastiblock[level] < 0) { 287 oip->i_ib[level] = 0; 288 lastiblock[level] = -1; 289 } 290 for (i = NDADDR - 1; i > lastblock; i--) 291 oip->i_db[i] = 0; 292 oip->i_flag |= IN_CHANGE | IN_UPDATE; 293 error = VOP_UPDATE(ovp, &tv, &tv, ((length > 0) ? 0 : 1)); 294 if (error) 295 allerror = error; 296 /* 297 * Having written the new inode to disk, save its new configuration 298 * and put back the old block pointers long enough to process them. 299 * Note that we save the new block configuration so we can check it 300 * when we are done. 301 */ 302 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks); 303 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks); 304 oip->i_size = osize; 305 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA; 306 allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0); 307 308 /* 309 * Indirect blocks first. 310 */ 311 indir_lbn[SINGLE] = -NDADDR; 312 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1; 313 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 314 for (level = TRIPLE; level >= SINGLE; level--) { 315 bn = oip->i_ib[level]; 316 if (bn != 0) { 317 error = ffs_indirtrunc(oip, indir_lbn[level], 318 fsbtodb(fs, bn), lastiblock[level], level, &count); 319 if (error) 320 allerror = error; 321 blocksreleased += count; 322 if (lastiblock[level] < 0) { 323 oip->i_ib[level] = 0; 324 ffs_blkfree(oip, bn, fs->fs_bsize); 325 blocksreleased += nblocks; 326 } 327 } 328 if (lastiblock[level] >= 0) 329 goto done; 330 } 331 332 /* 333 * All whole direct blocks or frags. 334 */ 335 for (i = NDADDR - 1; i > lastblock; i--) { 336 register long bsize; 337 338 bn = oip->i_db[i]; 339 if (bn == 0) 340 continue; 341 oip->i_db[i] = 0; 342 bsize = blksize(fs, oip, i); 343 ffs_blkfree(oip, bn, bsize); 344 blocksreleased += btodb(bsize); 345 } 346 if (lastblock < 0) 347 goto done; 348 349 /* 350 * Finally, look for a change in size of the 351 * last direct block; release any frags. 352 */ 353 bn = oip->i_db[lastblock]; 354 if (bn != 0) { 355 long oldspace, newspace; 356 357 /* 358 * Calculate amount of space we're giving 359 * back as old block size minus new block size. 360 */ 361 oldspace = blksize(fs, oip, lastblock); 362 oip->i_size = length; 363 newspace = blksize(fs, oip, lastblock); 364 if (newspace == 0) 365 panic("ffs_truncate: newspace"); 366 if (oldspace - newspace > 0) { 367 /* 368 * Block number of space to be free'd is 369 * the old block # plus the number of frags 370 * required for the storage we're keeping. 371 */ 372 bn += numfrags(fs, newspace); 373 ffs_blkfree(oip, bn, oldspace - newspace); 374 blocksreleased += btodb(oldspace - newspace); 375 } 376 } 377 done: 378 #ifdef DIAGNOSTIC 379 for (level = SINGLE; level <= TRIPLE; level++) 380 if (newblks[NDADDR + level] != oip->i_ib[level]) 381 panic("ffs_truncate1"); 382 for (i = 0; i < NDADDR; i++) 383 if (newblks[i] != oip->i_db[i]) 384 panic("ffs_truncate2"); 385 if (length == 0 && 386 (ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first)) 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 if (oip->i_blocks < 0) /* sanity */ 395 oip->i_blocks = 0; 396 oip->i_flag |= IN_CHANGE; 397 vnode_pager_setsize(ovp, length); 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 daddr_t lbn, lastbn; 417 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 daddr_t *bap; 425 struct vnode *vp; 426 daddr_t *copy, 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_flags |= B_READ; 456 if (bp->b_bcount > bp->b_bufsize) 457 panic("ffs_indirtrunc: bad buffer size"); 458 bp->b_blkno = dbn; 459 vfs_busy_pages(bp, 0); 460 VOP_STRATEGY(bp); 461 error = biowait(bp); 462 } 463 if (error) { 464 brelse(bp); 465 *countp = 0; 466 return (error); 467 } 468 469 bap = (daddr_t *)bp->b_data; 470 MALLOC(copy, daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK); 471 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize); 472 bzero((caddr_t)&bap[last + 1], 473 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (daddr_t)); 474 if (last == -1) 475 bp->b_flags |= B_INVAL; 476 if ((vp->v_mount->mnt_flag & MNT_ASYNC) == 0) { 477 error = bwrite(bp); 478 } else { 479 bawrite(bp); 480 error = 0; 481 } 482 if (error) 483 allerror = error; 484 bap = copy; 485 486 /* 487 * Recursively free totally unused blocks. 488 */ 489 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last; 490 i--, nlbn += factor) { 491 nb = bap[i]; 492 if (nb == 0) 493 continue; 494 if (level > SINGLE) { 495 error = ffs_indirtrunc(ip, nlbn, 496 fsbtodb(fs, nb), (daddr_t)-1, level - 1, &blkcount); 497 if (error) 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 FREE(copy, M_TEMP); 520 *countp = blocksreleased; 521 return (allerror); 522 } 523