1 /* $NetBSD: ext2fs_inode.c,v 1.28 2002/09/26 11:06:36 jdolecek 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 1.28 2002/09/26 11:06:36 jdolecek 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 *, ufs_daddr_t, ufs_daddr_t, 65 ufs_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 ufs_daddr_t lastblock; 194 struct inode *oip; 195 ufs_daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR]; 196 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 197 off_t length = ap->a_length; 198 struct m_ext2fs *fs; 199 int offset, size, level; 200 long count, nblocks, blocksreleased = 0; 201 int i; 202 int error, allerror = 0; 203 off_t osize; 204 205 if (length < 0) 206 return (EINVAL); 207 208 oip = VTOI(ovp); 209 if (ovp->v_type == VLNK && 210 (oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen || 211 (ovp->v_mount->mnt_maxsymlinklen == 0 && 212 oip->i_e2fs_nblock == 0))) { 213 #ifdef DIAGNOSTIC 214 if (length != 0) 215 panic("ext2fs_truncate: partial truncate of symlink"); 216 #endif 217 memset((char *)&oip->i_din.e2fs_din.e2di_shortlink, 0, 218 (u_int)oip->i_e2fs_size); 219 oip->i_e2fs_size = 0; 220 oip->i_flag |= IN_CHANGE | IN_UPDATE; 221 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)); 222 } 223 if (oip->i_e2fs_size == length) { 224 oip->i_flag |= IN_CHANGE | IN_UPDATE; 225 return (VOP_UPDATE(ovp, NULL, NULL, 0)); 226 } 227 fs = oip->i_e2fs; 228 osize = oip->i_e2fs_size; 229 /* 230 * Lengthen the size of the file. We must ensure that the 231 * last byte of the file is allocated. Since the smallest 232 * value of osize is 0, length will be at least 1. 233 */ 234 if (osize < length) { 235 #if 0 /* XXX */ 236 if (length > fs->fs_maxfilesize) 237 return (EFBIG); 238 #endif 239 ufs_balloc_range(ovp, length - 1, 1, ap->a_cred, 240 ap->a_flags & IO_SYNC ? B_SYNC : 0); 241 oip->i_flag |= IN_CHANGE | IN_UPDATE; 242 return (VOP_UPDATE(ovp, NULL, NULL, 1)); 243 } 244 /* 245 * Shorten the size of the file. If the file is not being 246 * truncated to a block boundry, the contents of the 247 * partial block following the end of the file must be 248 * zero'ed in case it ever become accessible again because 249 * of subsequent file growth. 250 */ 251 offset = blkoff(fs, length); 252 if (offset != 0) { 253 size = fs->e2fs_bsize; 254 255 /* XXXUBC we should handle more than just VREG */ 256 uvm_vnp_zerorange(ovp, length, size - offset); 257 } 258 oip->i_e2fs_size = length; 259 uvm_vnp_setsize(ovp, length); 260 261 /* 262 * Calculate index into inode's block list of 263 * last direct and indirect blocks (if any) 264 * which we want to keep. Lastblock is -1 when 265 * the file is truncated to 0. 266 */ 267 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 268 lastiblock[SINGLE] = lastblock - NDADDR; 269 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 270 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 271 nblocks = btodb(fs->e2fs_bsize); 272 /* 273 * Update file and block pointers on disk before we start freeing 274 * blocks. If we crash before free'ing blocks below, the blocks 275 * will be returned to the free list. lastiblock values are also 276 * normalized to -1 for calls to ext2fs_indirtrunc below. 277 */ 278 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks); 279 for (level = TRIPLE; level >= SINGLE; level--) 280 if (lastiblock[level] < 0) { 281 oip->i_e2fs_blocks[NDADDR + level] = 0; 282 lastiblock[level] = -1; 283 } 284 for (i = NDADDR - 1; i > lastblock; i--) 285 oip->i_e2fs_blocks[i] = 0; 286 oip->i_flag |= IN_CHANGE | IN_UPDATE; 287 error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT); 288 if (error && !allerror) 289 allerror = error; 290 291 /* 292 * Having written the new inode to disk, save its new configuration 293 * and put back the old block pointers long enough to process them. 294 * Note that we save the new block configuration so we can check it 295 * when we are done. 296 */ 297 298 memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks); 299 memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks); 300 oip->i_e2fs_size = osize; 301 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 302 if (error && !allerror) 303 allerror = error; 304 305 /* 306 * Indirect blocks first. 307 */ 308 indir_lbn[SINGLE] = -NDADDR; 309 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 310 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 311 for (level = TRIPLE; level >= SINGLE; level--) { 312 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 313 if (bn != 0) { 314 error = ext2fs_indirtrunc(oip, indir_lbn[level], 315 fsbtodb(fs, bn), lastiblock[level], level, &count); 316 if (error) 317 allerror = error; 318 blocksreleased += count; 319 if (lastiblock[level] < 0) { 320 oip->i_e2fs_blocks[NDADDR + level] = 0; 321 ext2fs_blkfree(oip, bn); 322 blocksreleased += nblocks; 323 } 324 } 325 if (lastiblock[level] >= 0) 326 goto done; 327 } 328 329 /* 330 * All whole direct blocks or frags. 331 */ 332 for (i = NDADDR - 1; i > lastblock; i--) { 333 bn = fs2h32(oip->i_e2fs_blocks[i]); 334 if (bn == 0) 335 continue; 336 oip->i_e2fs_blocks[i] = 0; 337 ext2fs_blkfree(oip, bn); 338 blocksreleased += btodb(fs->e2fs_bsize); 339 } 340 341 done: 342 #ifdef DIAGNOSTIC 343 for (level = SINGLE; level <= TRIPLE; level++) 344 if (newblks[NDADDR + level] != 345 oip->i_e2fs_blocks[NDADDR + level]) 346 panic("ext2fs_truncate1"); 347 for (i = 0; i < NDADDR; i++) 348 if (newblks[i] != oip->i_e2fs_blocks[i]) 349 panic("ext2fs_truncate2"); 350 if (length == 0 && 351 (!LIST_EMPTY(&ovp->v_cleanblkhd) || 352 !LIST_EMPTY(&ovp->v_dirtyblkhd))) 353 panic("ext2fs_truncate3"); 354 #endif /* DIAGNOSTIC */ 355 /* 356 * Put back the real size. 357 */ 358 oip->i_e2fs_size = length; 359 oip->i_e2fs_nblock -= blocksreleased; 360 oip->i_flag |= IN_CHANGE; 361 return (allerror); 362 } 363 364 /* 365 * Release blocks associated with the inode ip and stored in the indirect 366 * block bn. Blocks are free'd in LIFO order up to (but not including) 367 * lastbn. If level is greater than SINGLE, the block is an indirect block 368 * and recursive calls to indirtrunc must be used to cleanse other indirect 369 * blocks. 370 * 371 * NB: triple indirect blocks are untested. 372 */ 373 static int 374 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 375 struct inode *ip; 376 ufs_daddr_t lbn, lastbn; 377 ufs_daddr_t dbn; 378 int level; 379 long *countp; 380 { 381 int i; 382 struct buf *bp; 383 struct m_ext2fs *fs = ip->i_e2fs; 384 ufs_daddr_t *bap; 385 struct vnode *vp; 386 ufs_daddr_t *copy = NULL, nb, nlbn, last; 387 long blkcount, factor; 388 int nblocks, blocksreleased = 0; 389 int error = 0, allerror = 0; 390 391 /* 392 * Calculate index in current block of last 393 * block to be kept. -1 indicates the entire 394 * block so we need not calculate the index. 395 */ 396 factor = 1; 397 for (i = SINGLE; i < level; i++) 398 factor *= NINDIR(fs); 399 last = lastbn; 400 if (lastbn > 0) 401 last /= factor; 402 nblocks = btodb(fs->e2fs_bsize); 403 /* 404 * Get buffer of block pointers, zero those entries corresponding 405 * to blocks to be free'd, and update on disk copy first. Since 406 * double(triple) indirect before single(double) indirect, calls 407 * to bmap on these blocks will fail. However, we already have 408 * the on disk address, so we have to set the b_blkno field 409 * explicitly instead of letting bread do everything for us. 410 */ 411 vp = ITOV(ip); 412 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 413 if (bp->b_flags & (B_DONE | B_DELWRI)) { 414 /* Braces must be here in case trace evaluates to nothing. */ 415 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 416 } else { 417 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 418 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 419 bp->b_flags |= B_READ; 420 if (bp->b_bcount > bp->b_bufsize) 421 panic("ext2fs_indirtrunc: bad buffer size"); 422 bp->b_blkno = dbn; 423 VOP_STRATEGY(bp); 424 error = biowait(bp); 425 } 426 if (error) { 427 brelse(bp); 428 *countp = 0; 429 return (error); 430 } 431 432 bap = (ufs_daddr_t *)bp->b_data; 433 if (lastbn >= 0) { 434 MALLOC(copy, ufs_daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK); 435 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize); 436 memset((caddr_t)&bap[last + 1], 0, 437 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t)); 438 error = bwrite(bp); 439 if (error) 440 allerror = error; 441 bap = copy; 442 } 443 444 /* 445 * Recursively free totally unused blocks. 446 */ 447 for (i = NINDIR(fs) - 1, 448 nlbn = lbn + 1 - i * factor; i > last; 449 i--, nlbn += factor) { 450 nb = fs2h32(bap[i]); 451 if (nb == 0) 452 continue; 453 if (level > SINGLE) { 454 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 455 (ufs_daddr_t)-1, level - 1, 456 &blkcount); 457 if (error) 458 allerror = error; 459 blocksreleased += blkcount; 460 } 461 ext2fs_blkfree(ip, nb); 462 blocksreleased += nblocks; 463 } 464 465 /* 466 * Recursively free last partial block. 467 */ 468 if (level > SINGLE && lastbn >= 0) { 469 last = lastbn % factor; 470 nb = fs2h32(bap[i]); 471 if (nb != 0) { 472 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 473 last, level - 1, &blkcount); 474 if (error) 475 allerror = error; 476 blocksreleased += blkcount; 477 } 478 } 479 480 if (copy != NULL) { 481 FREE(copy, M_TEMP); 482 } else { 483 bp->b_flags |= B_INVAL; 484 brelse(bp); 485 } 486 487 *countp = blocksreleased; 488 return (allerror); 489 } 490