1 /* $NetBSD: ext2fs_inode.c,v 1.34 2003/06/29 22:32:32 fvdl Exp $ */ 2 3 /* 4 * Copyright (c) 1997 Manuel Bouyer. 5 * Copyright (c) 1982, 1986, 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94 37 * Modified for ext2fs by Manuel Bouyer. 38 */ 39 40 #include <sys/cdefs.h> 41 __KERNEL_RCSID(0, "$NetBSD: ext2fs_inode.c,v 1.34 2003/06/29 22:32:32 fvdl Exp $"); 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/inode.h> 56 #include <ufs/ufs/ufsmount.h> 57 #include <ufs/ufs/ufs_extern.h> 58 59 #include <ufs/ext2fs/ext2fs.h> 60 #include <ufs/ext2fs/ext2fs_extern.h> 61 62 extern int prtactive; 63 64 static int ext2fs_indirtrunc __P((struct inode *, daddr_t, daddr_t, 65 daddr_t, int, long *)); 66 67 /* 68 * Last reference to an inode. If necessary, write or delete it. 69 */ 70 int 71 ext2fs_inactive(v) 72 void *v; 73 { 74 struct vop_inactive_args /* { 75 struct vnode *a_vp; 76 struct proc *a_p; 77 } */ *ap = v; 78 struct vnode *vp = ap->a_vp; 79 struct inode *ip = VTOI(vp); 80 struct proc *p = ap->a_p; 81 struct timespec ts; 82 int error = 0; 83 84 if (prtactive && vp->v_usecount != 0) 85 vprint("ext2fs_inactive: pushing active", vp); 86 /* Get rid of inodes related to stale file handles. */ 87 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0) 88 goto out; 89 90 error = 0; 91 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 92 if (ip->i_e2fs_size != 0) { 93 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL); 94 } 95 TIMEVAL_TO_TIMESPEC(&time, &ts); 96 ip->i_e2fs_dtime = ts.tv_sec; 97 ip->i_flag |= IN_CHANGE | IN_UPDATE; 98 VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode); 99 } 100 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFIED | IN_ACCESSED)) 101 VOP_UPDATE(vp, NULL, NULL, 0); 102 out: 103 VOP_UNLOCK(vp, 0); 104 /* 105 * If we are done with the inode, reclaim it 106 * so that it can be reused immediately. 107 */ 108 if (ip->i_e2fs_dtime != 0) 109 vrecycle(vp, NULL, p); 110 return (error); 111 } 112 113 114 /* 115 * Update the access, modified, and inode change times as specified by the 116 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 117 * used to specify that the inode needs to be updated but that the times have 118 * already been set. The access and modified times are taken from the second 119 * and third parameters; the inode change time is always taken from the current 120 * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk 121 * write of the inode to complete. 122 */ 123 int 124 ext2fs_update(v) 125 void *v; 126 { 127 struct vop_update_args /* { 128 struct vnode *a_vp; 129 struct timespec *a_access; 130 struct timespec *a_modify; 131 int a_flags; 132 } */ *ap = v; 133 struct m_ext2fs *fs; 134 struct buf *bp; 135 struct inode *ip; 136 int error; 137 struct timespec ts; 138 caddr_t cp; 139 int flags; 140 141 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) 142 return (0); 143 ip = VTOI(ap->a_vp); 144 TIMEVAL_TO_TIMESPEC(&time, &ts); 145 EXT2FS_ITIMES(ip, 146 ap->a_access ? ap->a_access : &ts, 147 ap->a_modify ? ap->a_modify : &ts, &ts); 148 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED); 149 if (flags == 0) 150 return (0); 151 fs = ip->i_e2fs; 152 153 error = bread(ip->i_devvp, 154 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 155 (int)fs->e2fs_bsize, NOCRED, &bp); 156 if (error) { 157 brelse(bp); 158 return (error); 159 } 160 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED); 161 cp = (caddr_t)bp->b_data + 162 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE); 163 e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp); 164 if ((ap->a_flags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 && 165 (flags & IN_MODIFIED) != 0 && 166 (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 167 return (bwrite(bp)); 168 else { 169 bdwrite(bp); 170 return (0); 171 } 172 } 173 174 #define SINGLE 0 /* index of single indirect block */ 175 #define DOUBLE 1 /* index of double indirect block */ 176 #define TRIPLE 2 /* index of triple indirect block */ 177 /* 178 * Truncate the inode oip to at most length size, freeing the 179 * disk blocks. 180 */ 181 int 182 ext2fs_truncate(v) 183 void *v; 184 { 185 struct vop_truncate_args /* { 186 struct vnode *a_vp; 187 off_t a_length; 188 int a_flags; 189 struct ucred *a_cred; 190 struct proc *a_p; 191 } */ *ap = v; 192 struct vnode *ovp = ap->a_vp; 193 daddr_t lastblock; 194 struct inode *oip; 195 daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR]; 196 /* XXX ondisk32 */ 197 int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 198 off_t length = ap->a_length; 199 struct m_ext2fs *fs; 200 int offset, size, level; 201 long count, nblocks, blocksreleased = 0; 202 int i; 203 int error, allerror = 0; 204 off_t osize; 205 206 if (length < 0) 207 return (EINVAL); 208 209 oip = VTOI(ovp); 210 if (ovp->v_type == VLNK && 211 (oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen || 212 (ovp->v_mount->mnt_maxsymlinklen == 0 && 213 oip->i_e2fs_nblock == 0))) { 214 #ifdef DIAGNOSTIC 215 if (length != 0) 216 panic("ext2fs_truncate: partial truncate of symlink"); 217 #endif 218 memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0, 219 (u_int)oip->i_e2fs_size); 220 oip->i_e2fs_size = 0; 221 oip->i_flag |= IN_CHANGE | IN_UPDATE; 222 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)); 223 } 224 if (oip->i_e2fs_size == length) { 225 oip->i_flag |= IN_CHANGE | IN_UPDATE; 226 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 227 } 228 fs = oip->i_e2fs; 229 osize = oip->i_e2fs_size; 230 /* 231 * Lengthen the size of the file. We must ensure that the 232 * last byte of the file is allocated. Since the smallest 233 * value of osize is 0, length will be at least 1. 234 */ 235 if (osize < length) { 236 #if 0 /* XXX */ 237 if (length > fs->fs_maxfilesize) 238 return (EFBIG); 239 #endif 240 ufs_balloc_range(ovp, length - 1, 1, ap->a_cred, 241 ap->a_flags & IO_SYNC ? B_SYNC : 0); 242 oip->i_flag |= IN_CHANGE | IN_UPDATE; 243 return (VOP_UPDATE(ovp, NULL, NULL, 1)); 244 } 245 /* 246 * Shorten the size of the file. If the file is not being 247 * truncated to a block boundry, the contents of the 248 * partial block following the end of the file must be 249 * zero'ed in case it ever become accessible again because 250 * of subsequent file growth. 251 */ 252 offset = blkoff(fs, length); 253 if (offset != 0) { 254 size = fs->e2fs_bsize; 255 256 /* XXXUBC we should handle more than just VREG */ 257 uvm_vnp_zerorange(ovp, length, size - offset); 258 } 259 oip->i_e2fs_size = length; 260 uvm_vnp_setsize(ovp, length); 261 262 /* 263 * Calculate index into inode's block list of 264 * last direct and indirect blocks (if any) 265 * which we want to keep. Lastblock is -1 when 266 * the file is truncated to 0. 267 */ 268 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 269 lastiblock[SINGLE] = lastblock - NDADDR; 270 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 271 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 272 nblocks = btodb(fs->e2fs_bsize); 273 /* 274 * Update file and block pointers on disk before we start freeing 275 * blocks. If we crash before free'ing blocks below, the blocks 276 * will be returned to the free list. lastiblock values are also 277 * normalized to -1 for calls to ext2fs_indirtrunc below. 278 */ 279 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks); 280 for (level = TRIPLE; level >= SINGLE; level--) 281 if (lastiblock[level] < 0) { 282 oip->i_e2fs_blocks[NDADDR + level] = 0; 283 lastiblock[level] = -1; 284 } 285 for (i = NDADDR - 1; i > lastblock; i--) 286 oip->i_e2fs_blocks[i] = 0; 287 oip->i_flag |= IN_CHANGE | IN_UPDATE; 288 error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT); 289 if (error && !allerror) 290 allerror = error; 291 292 /* 293 * Having written the new inode to disk, save its new configuration 294 * and put back the old block pointers long enough to process them. 295 * Note that we save the new block configuration so we can check it 296 * when we are done. 297 */ 298 299 memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks); 300 memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks); 301 oip->i_e2fs_size = osize; 302 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 303 if (error && !allerror) 304 allerror = error; 305 306 /* 307 * Indirect blocks first. 308 */ 309 indir_lbn[SINGLE] = -NDADDR; 310 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 311 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 312 for (level = TRIPLE; level >= SINGLE; level--) { 313 /* XXX ondisk32 */ 314 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 315 if (bn != 0) { 316 error = ext2fs_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_e2fs_blocks[NDADDR + level] = 0; 323 ext2fs_blkfree(oip, bn); 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 /* XXX ondisk32 */ 336 bn = fs2h32(oip->i_e2fs_blocks[i]); 337 if (bn == 0) 338 continue; 339 oip->i_e2fs_blocks[i] = 0; 340 ext2fs_blkfree(oip, bn); 341 blocksreleased += btodb(fs->e2fs_bsize); 342 } 343 344 done: 345 #ifdef DIAGNOSTIC 346 for (level = SINGLE; level <= TRIPLE; level++) 347 if (newblks[NDADDR + level] != 348 oip->i_e2fs_blocks[NDADDR + level]) 349 panic("ext2fs_truncate1"); 350 for (i = 0; i < NDADDR; i++) 351 if (newblks[i] != oip->i_e2fs_blocks[i]) 352 panic("ext2fs_truncate2"); 353 if (length == 0 && 354 (!LIST_EMPTY(&ovp->v_cleanblkhd) || 355 !LIST_EMPTY(&ovp->v_dirtyblkhd))) 356 panic("ext2fs_truncate3"); 357 #endif /* DIAGNOSTIC */ 358 /* 359 * Put back the real size. 360 */ 361 oip->i_e2fs_size = length; 362 oip->i_e2fs_nblock -= blocksreleased; 363 oip->i_flag |= IN_CHANGE; 364 return (allerror); 365 } 366 367 /* 368 * Release blocks associated with the inode ip and stored in the indirect 369 * block bn. Blocks are free'd in LIFO order up to (but not including) 370 * lastbn. If level is greater than SINGLE, the block is an indirect block 371 * and recursive calls to indirtrunc must be used to cleanse other indirect 372 * blocks. 373 * 374 * NB: triple indirect blocks are untested. 375 */ 376 static int 377 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 378 struct inode *ip; 379 daddr_t lbn, lastbn; 380 daddr_t dbn; 381 int level; 382 long *countp; 383 { 384 int i; 385 struct buf *bp; 386 struct m_ext2fs *fs = ip->i_e2fs; 387 int32_t *bap; /* XXX ondisk32 */ 388 struct vnode *vp; 389 daddr_t nb, nlbn, last; 390 int32_t *copy = NULL; /* XXX ondisk32 */ 391 long blkcount, factor; 392 int nblocks, blocksreleased = 0; 393 int error = 0, allerror = 0; 394 395 /* 396 * Calculate index in current block of last 397 * block to be kept. -1 indicates the entire 398 * block so we need not calculate the index. 399 */ 400 factor = 1; 401 for (i = SINGLE; i < level; i++) 402 factor *= NINDIR(fs); 403 last = lastbn; 404 if (lastbn > 0) 405 last /= factor; 406 nblocks = btodb(fs->e2fs_bsize); 407 /* 408 * Get buffer of block pointers, zero those entries corresponding 409 * to blocks to be free'd, and update on disk copy first. Since 410 * double(triple) indirect before single(double) indirect, calls 411 * to bmap on these blocks will fail. However, we already have 412 * the on disk address, so we have to set the b_blkno field 413 * explicitly instead of letting bread do everything for us. 414 */ 415 vp = ITOV(ip); 416 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 417 if (bp->b_flags & (B_DONE | B_DELWRI)) { 418 /* Braces must be here in case trace evaluates to nothing. */ 419 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 420 } else { 421 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 422 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 423 bp->b_flags |= B_READ; 424 if (bp->b_bcount > bp->b_bufsize) 425 panic("ext2fs_indirtrunc: bad buffer size"); 426 bp->b_blkno = dbn; 427 VOP_STRATEGY(bp); 428 error = biowait(bp); 429 } 430 if (error) { 431 brelse(bp); 432 *countp = 0; 433 return (error); 434 } 435 436 bap = (int32_t *)bp->b_data; /* XXX ondisk32 */ 437 if (lastbn >= 0) { 438 /* XXX ondisk32 */ 439 MALLOC(copy, int32_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK); 440 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize); 441 memset((caddr_t)&bap[last + 1], 0, 442 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t)); 443 error = bwrite(bp); 444 if (error) 445 allerror = error; 446 bap = copy; 447 } 448 449 /* 450 * Recursively free totally unused blocks. 451 */ 452 for (i = NINDIR(fs) - 1, 453 nlbn = lbn + 1 - i * factor; i > last; 454 i--, nlbn += factor) { 455 /* XXX ondisk32 */ 456 nb = fs2h32(bap[i]); 457 if (nb == 0) 458 continue; 459 if (level > SINGLE) { 460 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 461 (daddr_t)-1, level - 1, 462 &blkcount); 463 if (error) 464 allerror = error; 465 blocksreleased += blkcount; 466 } 467 ext2fs_blkfree(ip, nb); 468 blocksreleased += nblocks; 469 } 470 471 /* 472 * Recursively free last partial block. 473 */ 474 if (level > SINGLE && lastbn >= 0) { 475 last = lastbn % factor; 476 /* XXX ondisk32 */ 477 nb = fs2h32(bap[i]); 478 if (nb != 0) { 479 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 480 last, level - 1, &blkcount); 481 if (error) 482 allerror = error; 483 blocksreleased += blkcount; 484 } 485 } 486 487 if (copy != NULL) { 488 FREE(copy, M_TEMP); 489 } else { 490 bp->b_flags |= B_INVAL; 491 brelse(bp); 492 } 493 494 *countp = blocksreleased; 495 return (allerror); 496 } 497