1 /* $NetBSD: ext2fs_inode.c,v 1.3 1997/10/09 15:42:50 bouyer 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 <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/ext2fs/ext2fs.h> 60 #include <ufs/ext2fs/ext2fs_extern.h> 61 62 static int ext2fs_indirtrunc __P((struct inode *, daddr_t, daddr_t, 63 daddr_t, int, long *)); 64 65 void 66 ext2fs_init() 67 { 68 static int done = 0; 69 70 if (done) 71 return; 72 done = 1; 73 ufs_ihashinit(); 74 return; 75 } 76 77 /* 78 * Last reference to an inode. If necessary, write or delete it. 79 */ 80 int 81 ext2fs_inactive(v) 82 void *v; 83 { 84 struct vop_inactive_args /* { 85 struct vnode *a_vp; 86 } */ *ap = v; 87 register struct vnode *vp = ap->a_vp; 88 register struct inode *ip = VTOI(vp); 89 struct timespec ts; 90 int error; 91 extern int prtactive; 92 93 if (prtactive && vp->v_usecount != 0) 94 vprint("ffs_inactive: pushing active", vp); 95 /* Get rid of inodes related to stale file handles. */ 96 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0) { 97 if ((vp->v_flag & VXLOCK) == 0) 98 vgone(vp); 99 return (0); 100 } 101 102 error = 0; 103 #ifdef DIAGNOSTIC 104 if (VOP_ISLOCKED(vp)) 105 panic("ffs_inactive: locked inode"); 106 if (curproc) 107 ip->i_lockholder = curproc->p_pid; 108 else 109 ip->i_lockholder = -1; 110 #endif 111 ip->i_flag |= IN_LOCKED; 112 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 113 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL); 114 TIMEVAL_TO_TIMESPEC(&time, &ts); 115 ip->i_e2fs_dtime = ts.tv_sec; 116 ip->i_flag |= IN_CHANGE | IN_UPDATE; 117 VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode); 118 } 119 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) { 120 TIMEVAL_TO_TIMESPEC(&time, &ts); 121 VOP_UPDATE(vp, &ts, &ts, 0); 122 } 123 VOP_UNLOCK(vp); 124 /* 125 * If we are done with the inode, reclaim it 126 * so that it can be reused immediately. 127 */ 128 if (vp->v_usecount == 0 && ip->i_e2fs_dtime != 0) 129 vgone(vp); 130 return (error); 131 } 132 133 134 /* 135 * Update the access, modified, and inode change times as specified by the 136 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is 137 * used to specify that the inode needs to be updated but that the times have 138 * already been set. The access and modified times are taken from the second 139 * and third parameters; the inode change time is always taken from the current 140 * time. If waitfor is set, then wait for the disk write of the inode to 141 * complete. 142 */ 143 int 144 ext2fs_update(v) 145 void *v; 146 { 147 struct vop_update_args /* { 148 struct vnode *a_vp; 149 struct timespec *a_access; 150 struct timespec *a_modify; 151 int a_waitfor; 152 } */ *ap = v; 153 register struct m_ext2fs *fs; 154 struct buf *bp; 155 struct inode *ip; 156 int error; 157 struct timespec ts; 158 159 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) 160 return (0); 161 ip = VTOI(ap->a_vp); 162 TIMEVAL_TO_TIMESPEC(&time, &ts); 163 EXT2FS_ITIMES(ip, ap->a_access, ap->a_modify, &ts); 164 if ((ip->i_flag & IN_MODIFIED) == 0) 165 return (0); 166 ip->i_flag &= ~IN_MODIFIED; 167 fs = ip->i_e2fs; 168 error = bread(ip->i_devvp, 169 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 170 (int)fs->e2fs_bsize, NOCRED, &bp); 171 if (error) { 172 brelse(bp); 173 return (error); 174 } 175 e2fs_isave(&ip->i_din.e2fs_din, 176 (struct ext2fs_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)); 177 if (ap->a_waitfor) 178 return (bwrite(bp)); 179 else { 180 bdwrite(bp); 181 return (0); 182 } 183 } 184 185 #define SINGLE 0 /* index of single indirect block */ 186 #define DOUBLE 1 /* index of double indirect block */ 187 #define TRIPLE 2 /* index of triple indirect block */ 188 /* 189 * Truncate the inode oip to at most length size, freeing the 190 * disk blocks. 191 */ 192 int 193 ext2fs_truncate(v) 194 void *v; 195 { 196 struct vop_truncate_args /* { 197 struct vnode *a_vp; 198 off_t a_length; 199 int a_flags; 200 struct ucred *a_cred; 201 struct proc *a_p; 202 } */ *ap = v; 203 register struct vnode *ovp = ap->a_vp; 204 register daddr_t lastblock; 205 register struct inode *oip; 206 daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR]; 207 daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR]; 208 off_t length = ap->a_length; 209 register struct m_ext2fs *fs; 210 struct buf *bp; 211 int offset, size, level; 212 long count, nblocks, vflags, blocksreleased = 0; 213 struct timespec ts; 214 register int i; 215 int aflags, error, allerror; 216 off_t osize; 217 218 if (length < 0) 219 return (EINVAL); 220 221 oip = VTOI(ovp); 222 TIMEVAL_TO_TIMESPEC(&time, &ts); 223 if (ovp->v_type == VLNK && 224 (oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen || 225 (ovp->v_mount->mnt_maxsymlinklen == 0 && 226 oip->i_e2fs_nblock == 0))) { 227 #ifdef DIAGNOSTIC 228 if (length != 0) 229 panic("ext2fs_truncate: partial truncate of symlink"); 230 #endif 231 bzero((char *)&oip->i_din.e2fs_din.e2di_shortlink, 232 (u_int)oip->i_e2fs_size); 233 oip->i_e2fs_size = 0; 234 oip->i_flag |= IN_CHANGE | IN_UPDATE; 235 return (VOP_UPDATE(ovp, &ts, &ts, 1)); 236 } 237 if (oip->i_e2fs_size == length) { 238 oip->i_flag |= IN_CHANGE | IN_UPDATE; 239 return (VOP_UPDATE(ovp, &ts, &ts, 0)); 240 } 241 fs = oip->i_e2fs; 242 osize = oip->i_e2fs_size; 243 /* 244 * Lengthen the size of the file. We must ensure that the 245 * last byte of the file is allocated. Since the smallest 246 * value of osize is 0, length will be at least 1. 247 */ 248 if (osize < length) { 249 #if 0 /* XXX */ 250 if (length > fs->fs_maxfilesize) 251 return (EFBIG); 252 #endif 253 offset = blkoff(fs, length - 1); 254 lbn = lblkno(fs, length - 1); 255 aflags = B_CLRBUF; 256 if (ap->a_flags & IO_SYNC) 257 aflags |= B_SYNC; 258 error = ext2fs_balloc(oip, lbn, offset + 1, ap->a_cred, &bp, 259 aflags); 260 if (error) 261 return (error); 262 oip->i_e2fs_size = length; 263 vnode_pager_setsize(ovp, length); 264 (void) vnode_pager_uncache(ovp); 265 if (aflags & B_SYNC) 266 bwrite(bp); 267 else 268 bawrite(bp); 269 oip->i_flag |= IN_CHANGE | IN_UPDATE; 270 return (VOP_UPDATE(ovp, &ts, &ts, 1)); 271 } 272 /* 273 * Shorten the size of the file. If the file is not being 274 * truncated to a block boundry, the contents of the 275 * partial block following the end of the file must be 276 * zero'ed in case it ever become accessable again because 277 * of subsequent file growth. 278 */ 279 offset = blkoff(fs, length); 280 if (offset == 0) { 281 oip->i_e2fs_size = length; 282 } else { 283 lbn = lblkno(fs, length); 284 aflags = B_CLRBUF; 285 if (ap->a_flags & IO_SYNC) 286 aflags |= B_SYNC; 287 error = ext2fs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags); 288 if (error) 289 return (error); 290 oip->i_e2fs_size = length; 291 size = fs->e2fs_bsize; 292 vnode_pager_setsize(ovp, length); 293 (void) vnode_pager_uncache(ovp); 294 bzero((char *)bp->b_data + offset, (u_int)(size - offset)); 295 allocbuf(bp, size); 296 if (aflags & B_SYNC) 297 bwrite(bp); 298 else 299 bawrite(bp); 300 } 301 /* 302 * Calculate index into inode's block list of 303 * last direct and indirect blocks (if any) 304 * which we want to keep. Lastblock is -1 when 305 * the file is truncated to 0. 306 */ 307 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1; 308 lastiblock[SINGLE] = lastblock - NDADDR; 309 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); 310 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); 311 nblocks = btodb(fs->e2fs_bsize); 312 /* 313 * Update file and block pointers on disk before we start freeing 314 * blocks. If we crash before free'ing blocks below, the blocks 315 * will be returned to the free list. lastiblock values are also 316 * normalized to -1 for calls to ext2fs_indirtrunc below. 317 */ 318 bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks); 319 for (level = TRIPLE; level >= SINGLE; level--) 320 if (lastiblock[level] < 0) { 321 oip->i_e2fs_blocks[NDADDR + level] = 0; 322 lastiblock[level] = -1; 323 } 324 for (i = NDADDR - 1; i > lastblock; i--) 325 oip->i_e2fs_blocks[i] = 0; 326 oip->i_flag |= IN_CHANGE | IN_UPDATE; 327 if ((error = VOP_UPDATE(ovp, &ts, &ts, 1)) != 0) 328 allerror = error; 329 /* 330 * Having written the new inode to disk, save its new configuration 331 * and put back the old block pointers long enough to process them. 332 * Note that we save the new block configuration so we can check it 333 * when we are done. 334 */ 335 bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)newblks, sizeof newblks); 336 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks); 337 oip->i_e2fs_size = osize; 338 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA; 339 allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0); 340 341 /* 342 * Indirect blocks first. 343 */ 344 indir_lbn[SINGLE] = -NDADDR; 345 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1; 346 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1; 347 for (level = TRIPLE; level >= SINGLE; level--) { 348 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]); 349 if (bn != 0) { 350 error = ext2fs_indirtrunc(oip, indir_lbn[level], 351 fsbtodb(fs, bn), lastiblock[level], level, &count); 352 if (error) 353 allerror = error; 354 blocksreleased += count; 355 if (lastiblock[level] < 0) { 356 oip->i_e2fs_blocks[NDADDR + level] = 0; 357 ext2fs_blkfree(oip, bn); 358 blocksreleased += nblocks; 359 } 360 } 361 if (lastiblock[level] >= 0) 362 goto done; 363 } 364 365 /* 366 * All whole direct blocks or frags. 367 */ 368 for (i = NDADDR - 1; i > lastblock; i--) { 369 bn = fs2h32(oip->i_e2fs_blocks[i]); 370 if (bn == 0) 371 continue; 372 oip->i_e2fs_blocks[i] = 0; 373 ext2fs_blkfree(oip, bn); 374 blocksreleased += btodb(fs->e2fs_bsize); 375 } 376 if (lastblock < 0) 377 goto done; 378 379 done: 380 #ifdef DIAGNOSTIC 381 for (level = SINGLE; level <= TRIPLE; level++) 382 if (newblks[NDADDR + level] != 383 fs2h32(oip->i_e2fs_blocks[NDADDR + level])) 384 panic("itrunc1"); 385 for (i = 0; i < NDADDR; i++) 386 if (newblks[i] != fs2h32(oip->i_e2fs_blocks[i])) 387 panic("itrunc2"); 388 if (length == 0 && 389 (ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first)) 390 panic("itrunc3"); 391 #endif /* DIAGNOSTIC */ 392 /* 393 * Put back the real size. 394 */ 395 oip->i_e2fs_size = length; 396 oip->i_e2fs_nblock -= blocksreleased; 397 if (oip->i_e2fs_nblock < 0) /* sanity */ 398 oip->i_e2fs_nblock = 0; 399 oip->i_flag |= IN_CHANGE; 400 return (allerror); 401 } 402 403 /* 404 * Release blocks associated with the inode ip and stored in the indirect 405 * block bn. Blocks are free'd in LIFO order up to (but not including) 406 * lastbn. If level is greater than SINGLE, the block is an indirect block 407 * and recursive calls to indirtrunc must be used to cleanse other indirect 408 * blocks. 409 * 410 * NB: triple indirect blocks are untested. 411 */ 412 static int 413 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp) 414 register struct inode *ip; 415 daddr_t lbn, lastbn; 416 daddr_t dbn; 417 int level; 418 long *countp; 419 { 420 register int i; 421 struct buf *bp; 422 register struct m_ext2fs *fs = ip->i_e2fs; 423 register daddr_t *bap; 424 struct vnode *vp; 425 daddr_t *copy = NULL, nb, nlbn, last; 426 long blkcount, factor; 427 int nblocks, blocksreleased = 0; 428 int error = 0, allerror = 0; 429 430 /* 431 * Calculate index in current block of last 432 * block to be kept. -1 indicates the entire 433 * block so we need not calculate the index. 434 */ 435 factor = 1; 436 for (i = SINGLE; i < level; i++) 437 factor *= NINDIR(fs); 438 last = lastbn; 439 if (lastbn > 0) 440 last /= factor; 441 nblocks = btodb(fs->e2fs_bsize); 442 /* 443 * Get buffer of block pointers, zero those entries corresponding 444 * to blocks to be free'd, and update on disk copy first. Since 445 * double(triple) indirect before single(double) indirect, calls 446 * to bmap on these blocks will fail. However, we already have 447 * the on disk address, so we have to set the b_blkno field 448 * explicitly instead of letting bread do everything for us. 449 */ 450 vp = ITOV(ip); 451 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0); 452 if (bp->b_flags & (B_DONE | B_DELWRI)) { 453 /* Braces must be here in case trace evaluates to nothing. */ 454 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn); 455 } else { 456 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn); 457 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */ 458 bp->b_flags |= B_READ; 459 if (bp->b_bcount > bp->b_bufsize) 460 panic("ext2fs_indirtrunc: bad buffer size"); 461 bp->b_blkno = dbn; 462 VOP_STRATEGY(bp); 463 error = biowait(bp); 464 } 465 if (error) { 466 brelse(bp); 467 *countp = 0; 468 return (error); 469 } 470 471 bap = (daddr_t *)bp->b_data; 472 if (lastbn != -1) { 473 MALLOC(copy, daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK); 474 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize); 475 bzero((caddr_t)&bap[last + 1], 476 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t)); 477 error = bwrite(bp); 478 if (error) 479 allerror = error; 480 bap = copy; 481 } 482 483 /* 484 * Recursively free totally unused blocks. 485 */ 486 for (i = NINDIR(fs) - 1, 487 nlbn = lbn + 1 - i * factor; i > last; 488 i--, nlbn += factor) { 489 nb = fs2h32(bap[i]); 490 if (nb == 0) 491 continue; 492 if (level > SINGLE) { 493 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 494 (daddr_t)-1, level - 1, 495 &blkcount); 496 if (error) 497 allerror = error; 498 blocksreleased += blkcount; 499 } 500 ext2fs_blkfree(ip, nb); 501 blocksreleased += nblocks; 502 } 503 504 /* 505 * Recursively free last partial block. 506 */ 507 if (level > SINGLE && lastbn >= 0) { 508 last = lastbn % factor; 509 nb = fs2h32(bap[i]); 510 if (nb != 0) { 511 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb), 512 last, level - 1, &blkcount); 513 if (error) 514 allerror = error; 515 blocksreleased += blkcount; 516 } 517 } 518 519 if (copy != NULL) { 520 FREE(copy, M_TEMP); 521 } else { 522 bp->b_flags |= B_INVAL; 523 brelse(bp); 524 } 525 526 *countp = blocksreleased; 527 return (allerror); 528 } 529