1 /* $NetBSD: ext2fs_inode.c,v 1.19 2000/05/30 19:09:44 mycroft 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/param.h> 41 #include <sys/systm.h> 42 #include <sys/mount.h> 43 #include <sys/proc.h> 44 #include <sys/file.h> 45 #include <sys/buf.h> 46 #include <sys/vnode.h> 47 #include <sys/kernel.h> 48 #include <sys/malloc.h> 49 #include <sys/trace.h> 50 #include <sys/resourcevar.h> 51 52 #include <vm/vm.h> 53 54 #include <uvm/uvm_extern.h> 55 56 #include <ufs/ufs/quota.h> 57 #include <ufs/ufs/inode.h> 58 #include <ufs/ufs/ufsmount.h> 59 #include <ufs/ufs/ufs_extern.h> 60 61 #include <ufs/ext2fs/ext2fs.h> 62 #include <ufs/ext2fs/ext2fs_extern.h> 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 extern int prtactive; 84 85 if (prtactive && vp->v_usecount != 0) 86 vprint("ext2fs_inactive: pushing active", vp); 87 /* Get rid of inodes related to stale file handles. */ 88 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0) 89 goto out; 90 91 error = 0; 92 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 93 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL); 94 TIMEVAL_TO_TIMESPEC(&time, &ts); 95 ip->i_e2fs_dtime = ts.tv_sec; 96 ip->i_flag |= IN_CHANGE | IN_UPDATE; 97 VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode); 98 } 99 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFIED | IN_ACCESSED)) 100 VOP_UPDATE(vp, NULL, NULL, 0); 101 out: 102 VOP_UNLOCK(vp, 0); 103 /* 104 * If we are done with the inode, reclaim it 105 * so that it can be reused immediately. 106 */ 107 if (ip->i_e2fs_dtime != 0) 108 vrecycle(vp, (struct simplelock *)0, p); 109 return (error); 110 } 111 112 113 /* 114 * Update the access, modified, and inode change times as specified by the 115 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 116 * used to specify that the inode needs to be updated but that the times have 117 * already been set. The access and modified times are taken from the second 118 * and third parameters; the inode change time is always taken from the current 119 * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk 120 * write of the inode to complete. 121 */ 122 int 123 ext2fs_update(v) 124 void *v; 125 { 126 struct vop_update_args /* { 127 struct vnode *a_vp; 128 struct timespec *a_access; 129 struct timespec *a_modify; 130 int a_flags; 131 } */ *ap = v; 132 struct m_ext2fs *fs; 133 struct buf *bp; 134 struct inode *ip; 135 int error; 136 struct timespec ts; 137 caddr_t cp; 138 int flags; 139 140 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) 141 return (0); 142 ip = VTOI(ap->a_vp); 143 TIMEVAL_TO_TIMESPEC(&time, &ts); 144 EXT2FS_ITIMES(ip, 145 ap->a_access ? ap->a_access : &ts, 146 ap->a_modify ? ap->a_modify : &ts, &ts); 147 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED); 148 if (flags == 0) 149 return (0); 150 fs = ip->i_e2fs; 151 152 error = bread(ip->i_devvp, 153 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 154 (int)fs->e2fs_bsize, NOCRED, &bp); 155 if (error) { 156 brelse(bp); 157 return (error); 158 } 159 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED); 160 cp = (caddr_t)bp->b_data + 161 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE); 162 e2fs_isave(&ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp); 163 if ((ap->a_flags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 && 164 (flags & IN_MODIFIED) != 0 && 165 (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) 166 return (bwrite(bp)); 167 else { 168 bdwrite(bp); 169 return (0); 170 } 171 } 172 173 #define SINGLE 0 /* index of single indirect block */ 174 #define DOUBLE 1 /* index of double indirect block */ 175 #define TRIPLE 2 /* index of triple indirect block */ 176 /* 177 * Truncate the inode oip to at most length size, freeing the 178 * disk blocks. 179 */ 180 int 181 ext2fs_truncate(v) 182 void *v; 183 { 184 struct vop_truncate_args /* { 185 struct vnode *a_vp; 186 off_t a_length; 187 int a_flags; 188 struct ucred *a_cred; 189 struct proc *a_p; 190 } */ *ap = v; 191 struct vnode *ovp = ap->a_vp; 192 ufs_daddr_t lastblock; 193 struct inode *oip; 194 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 195 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 196 off_t length = ap->a_length; 197 struct m_ext2fs *fs; 198 struct buf *bp; 199 int offset, size, level; 200 long count, nblocks, blocksreleased = 0; 201 int i; 202 int aflags, 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 offset = blkoff(fs, length - 1); 240 lbn = lblkno(fs, length - 1); 241 aflags = B_CLRBUF; 242 if (ap->a_flags & IO_SYNC) 243 aflags |= B_SYNC; 244 error = ext2fs_balloc(oip, lbn, offset + 1, ap->a_cred, &bp, 245 aflags); 246 if (error) 247 return (error); 248 oip->i_e2fs_size = length; 249 uvm_vnp_setsize(ovp, length); 250 (void) uvm_vnp_uncache(ovp); 251 if (aflags & B_SYNC) 252 bwrite(bp); 253 else 254 bawrite(bp); 255 oip->i_flag |= IN_CHANGE | IN_UPDATE; 256 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT)); 257 } 258 /* 259 * Shorten the size of the file. If the file is not being 260 * truncated to a block boundry, the contents of the 261 * partial block following the end of the file must be 262 * zero'ed in case it ever become accessable again because 263 * of subsequent file growth. 264 */ 265 offset = blkoff(fs, length); 266 if (offset == 0) { 267 oip->i_e2fs_size = length; 268 } else { 269 lbn = lblkno(fs, length); 270 aflags = B_CLRBUF; 271 if (ap->a_flags & IO_SYNC) 272 aflags |= B_SYNC; 273 error = ext2fs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags); 274 if (error) 275 return (error); 276 oip->i_e2fs_size = length; 277 size = fs->e2fs_bsize; 278 (void) uvm_vnp_uncache(ovp); 279 memset((char *)bp->b_data + offset, 0, (u_int)(size - offset)); 280 allocbuf(bp, size); 281 if (aflags & B_SYNC) 282 bwrite(bp); 283 else 284 bawrite(bp); 285 } 286 uvm_vnp_setsize(ovp, length); 287 288 /* 289 * Calculate index into inode's block list of 290 * last direct and indirect blocks (if any) 291 * which we want to keep. Lastblock is -1 when 292 * the file is truncated to 0. 293 */ 294 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 295 lastiblock[SINGLE] = lastblock - NDADDR; 296 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 297 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 298 nblocks = btodb(fs->e2fs_bsize); 299 /* 300 * Update file and block pointers on disk before we start freeing 301 * blocks. If we crash before free'ing blocks below, the blocks 302 * will be returned to the free list. lastiblock values are also 303 * normalized to -1 for calls to ext2fs_indirtrunc below. 304 */ 305 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks); 306 for (level = TRIPLE; level >= SINGLE; level--) 307 if (lastiblock[level] < 0) { 308 oip->i_e2fs_blocks[NDADDR + level] = 0; 309 lastiblock[level] = -1; 310 } 311 for (i = NDADDR - 1; i > lastblock; i--) 312 oip->i_e2fs_blocks[i] = 0; 313 oip->i_flag |= IN_CHANGE | IN_UPDATE; 314 error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT); 315 if (error && !allerror) 316 allerror = error; 317 318 /* 319 * Having written the new inode to disk, save its new configuration 320 * and put back the old block pointers long enough to process them. 321 * Note that we save the new block configuration so we can check it 322 * when we are done. 323 */ 324 memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks); 325 memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks); 326 oip->i_e2fs_size = osize; 327 error = vtruncbuf(ovp, lastblock + 1, 0, 0); 328 if (error && !allerror) 329 allerror = error; 330 331 /* 332 * Indirect blocks first. 333 */ 334 indir_lbn[SINGLE] = -NDADDR; 335 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 336 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 337 for (level = TRIPLE; level >= SINGLE; level--) { 338 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 339 if (bn != 0) { 340 error = ext2fs_indirtrunc(oip, indir_lbn[level], 341 fsbtodb(fs, bn), lastiblock[level], level, &count); 342 if (error) 343 allerror = error; 344 blocksreleased += count; 345 if (lastiblock[level] < 0) { 346 oip->i_e2fs_blocks[NDADDR + level] = 0; 347 ext2fs_blkfree(oip, bn); 348 blocksreleased += nblocks; 349 } 350 } 351 if (lastiblock[level] >= 0) 352 goto done; 353 } 354 355 /* 356 * All whole direct blocks or frags. 357 */ 358 for (i = NDADDR - 1; i > lastblock; i--) { 359 bn = fs2h32(oip->i_e2fs_blocks[i]); 360 if (bn == 0) 361 continue; 362 oip->i_e2fs_blocks[i] = 0; 363 ext2fs_blkfree(oip, bn); 364 blocksreleased += btodb(fs->e2fs_bsize); 365 } 366 if (lastblock < 0) 367 goto done; 368 369 done: 370 #ifdef DIAGNOSTIC 371 for (level = SINGLE; level <= TRIPLE; level++) 372 if (newblks[NDADDR + level] != oip->i_e2fs_blocks[NDADDR + level]) 373 panic("itrunc1"); 374 for (i = 0; i < NDADDR; i++) 375 if (newblks[i] != oip->i_e2fs_blocks[i]) 376 panic("itrunc2"); 377 if (length == 0 && 378 (!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd))) 379 panic("itrunc3"); 380 #endif /* DIAGNOSTIC */ 381 /* 382 * Put back the real size. 383 */ 384 oip->i_e2fs_size = length; 385 oip->i_e2fs_nblock -= blocksreleased; 386 if (oip->i_e2fs_nblock < 0) /* sanity */ 387 oip->i_e2fs_nblock = 0; 388 oip->i_flag |= IN_CHANGE; 389 return (allerror); 390 } 391 392 /* 393 * Release blocks associated with the inode ip and stored in the indirect 394 * block bn. Blocks are free'd in LIFO order up to (but not including) 395 * lastbn. If level is greater than SINGLE, the block is an indirect block 396 * and recursive calls to indirtrunc must be used to cleanse other indirect 397 * blocks. 398 * 399 * NB: triple indirect blocks are untested. 400 */ 401 static int 402 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 403 struct inode *ip; 404 ufs_daddr_t lbn, lastbn; 405 ufs_daddr_t dbn; 406 int level; 407 long *countp; 408 { 409 int i; 410 struct buf *bp; 411 struct m_ext2fs *fs = ip->i_e2fs; 412 ufs_daddr_t *bap; 413 struct vnode *vp; 414 ufs_daddr_t *copy = NULL, nb, nlbn, last; 415 long blkcount, factor; 416 int nblocks, blocksreleased = 0; 417 int error = 0, allerror = 0; 418 419 /* 420 * Calculate index in current block of last 421 * block to be kept. -1 indicates the entire 422 * block so we need not calculate the index. 423 */ 424 factor = 1; 425 for (i = SINGLE; i < level; i++) 426 factor *= NINDIR(fs); 427 last = lastbn; 428 if (lastbn > 0) 429 last /= factor; 430 nblocks = btodb(fs->e2fs_bsize); 431 /* 432 * Get buffer of block pointers, zero those entries corresponding 433 * to blocks to be free'd, and update on disk copy first. Since 434 * double(triple) indirect before single(double) indirect, calls 435 * to bmap on these blocks will fail. However, we already have 436 * the on disk address, so we have to set the b_blkno field 437 * explicitly instead of letting bread do everything for us. 438 */ 439 vp = ITOV(ip); 440 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 441 if (bp->b_flags & (B_DONE | B_DELWRI)) { 442 /* Braces must be here in case trace evaluates to nothing. */ 443 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 444 } else { 445 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 446 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 447 bp->b_flags |= B_READ; 448 if (bp->b_bcount > bp->b_bufsize) 449 panic("ext2fs_indirtrunc: bad buffer size"); 450 bp->b_blkno = dbn; 451 VOP_STRATEGY(bp); 452 error = biowait(bp); 453 } 454 if (error) { 455 brelse(bp); 456 *countp = 0; 457 return (error); 458 } 459 460 bap = (ufs_daddr_t *)bp->b_data; 461 if (lastbn >= 0) { 462 MALLOC(copy, ufs_daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK); 463 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize); 464 memset((caddr_t)&bap[last + 1], 0, 465 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t)); 466 error = bwrite(bp); 467 if (error) 468 allerror = error; 469 bap = copy; 470 } 471 472 /* 473 * Recursively free totally unused blocks. 474 */ 475 for (i = NINDIR(fs) - 1, 476 nlbn = lbn + 1 - i * factor; i > last; 477 i--, nlbn += factor) { 478 nb = fs2h32(bap[i]); 479 if (nb == 0) 480 continue; 481 if (level > SINGLE) { 482 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 483 (ufs_daddr_t)-1, level - 1, 484 &blkcount); 485 if (error) 486 allerror = error; 487 blocksreleased += blkcount; 488 } 489 ext2fs_blkfree(ip, nb); 490 blocksreleased += nblocks; 491 } 492 493 /* 494 * Recursively free last partial block. 495 */ 496 if (level > SINGLE && lastbn >= 0) { 497 last = lastbn % factor; 498 nb = fs2h32(bap[i]); 499 if (nb != 0) { 500 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 501 last, level - 1, &blkcount); 502 if (error) 503 allerror = error; 504 blocksreleased += blkcount; 505 } 506 } 507 508 if (copy != NULL) { 509 FREE(copy, M_TEMP); 510 } else { 511 bp->b_flags |= B_INVAL; 512 brelse(bp); 513 } 514 515 *countp = blocksreleased; 516 return (allerror); 517 } 518