Home | History | Annotate | Line # | Download | only in fsck_ffs
      1 /*	$NetBSD: inode.c,v 1.79 2025/02/25 15:29:00 andvar Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1980, 1986, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the University nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 #include <sys/cdefs.h>
     33 #ifndef lint
     34 #if 0
     35 static char sccsid[] = "@(#)inode.c	8.8 (Berkeley) 4/28/95";
     36 #else
     37 __RCSID("$NetBSD: inode.c,v 1.79 2025/02/25 15:29:00 andvar Exp $");
     38 #endif
     39 #endif /* not lint */
     40 
     41 #include <sys/param.h>
     42 #include <sys/time.h>
     43 #include <sys/stat.h>
     44 
     45 #include <ufs/ufs/dinode.h>
     46 #include <ufs/ufs/dir.h>
     47 #include <ufs/ffs/fs.h>
     48 #include <ufs/ffs/ffs_extern.h>
     49 #include <ufs/ufs/ufs_bswap.h>
     50 
     51 #ifndef SMALL
     52 #include <err.h>
     53 #include <pwd.h>
     54 #endif
     55 #include <stdio.h>
     56 #include <stdlib.h>
     57 #include <string.h>
     58 #include <time.h>
     59 
     60 #include "fsck.h"
     61 #include "fsutil.h"
     62 #include "extern.h"
     63 
     64 static ino_t startinum;
     65 
     66 static int iblock(struct inodesc *, long, u_int64_t);
     67 #ifndef NO_FFS_EI
     68 static void swap_dinode1(union dinode *, int);
     69 static void swap_dinode2(union dinode *, int);
     70 #endif
     71 
     72 int
     73 ckinode(union dinode *dp, struct inodesc *idesc)
     74 {
     75 	int ret, offset, i;
     76 	union dinode dino;
     77 	u_int64_t sizepb;
     78 	int64_t remsize;
     79 	daddr_t ndb;
     80 	mode_t mode;
     81 	char pathbuf[MAXPATHLEN + 1];
     82 
     83 	if (idesc->id_fix != IGNORE)
     84 		idesc->id_fix = DONTKNOW;
     85 	idesc->id_entryno = 0;
     86 	idesc->id_filesize = iswap64(DIP(dp, size));
     87 	mode = iswap16(DIP(dp, mode)) & IFMT;
     88 	if (mode == IFBLK || mode == IFCHR || (mode == IFLNK &&
     89 	    (idesc->id_filesize < sblock->fs_maxsymlinklen ||
     90 	    (isappleufs && (idesc->id_filesize < APPLEUFS_MAXSYMLINKLEN)) ||
     91 	     (sblock->fs_maxsymlinklen == 0 && DIP(dp, blocks) == 0))))
     92 		return (KEEPON);
     93 	if (is_ufs2)
     94 		dino.dp2 = dp->dp2;
     95 	else
     96 		dino.dp1 = dp->dp1;
     97 	ndb = howmany(iswap64(DIP(&dino, size)), sblock->fs_bsize);
     98 	for (i = 0; i < UFS_NDADDR; i++) {
     99 		if (--ndb == 0 &&
    100 		    (offset = ffs_blkoff(sblock, iswap64(DIP(&dino, size)))) != 0)
    101 			idesc->id_numfrags =
    102 				ffs_numfrags(sblock, ffs_fragroundup(sblock, offset));
    103 		else
    104 			idesc->id_numfrags = sblock->fs_frag;
    105 		if (DIP(&dino, db[i]) == 0) {
    106 			if (idesc->id_type == DATA && ndb >= 0) {
    107 				/* An empty block in a directory XXX */
    108 				markclean = 0;
    109 				getpathname(pathbuf, sizeof(pathbuf),
    110 				    idesc->id_number, idesc->id_number);
    111 				pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
    112 				    pathbuf);
    113 				if (reply("ADJUST LENGTH") == 1) {
    114 					dp = ginode(idesc->id_number);
    115 					DIP_SET(dp, size, iswap64(i *
    116 					    sblock->fs_bsize));
    117 					printf(
    118 					    "YOU MUST RERUN FSCK AFTERWARDS\n");
    119 					rerun = 1;
    120 					inodirty();
    121 				}
    122 			}
    123 			continue;
    124 		}
    125 		if (is_ufs2)
    126 			idesc->id_blkno = iswap64(dino.dp2.di_db[i]);
    127 		else
    128 			idesc->id_blkno = iswap32(dino.dp1.di_db[i]);
    129 		if (idesc->id_type != DATA)
    130 			ret = (*idesc->id_func)(idesc);
    131 		else
    132 			ret = dirscan(idesc);
    133 		if (ret & STOP)
    134 			return (ret);
    135 	}
    136 	idesc->id_numfrags = sblock->fs_frag;
    137 	remsize = iswap64(DIP(&dino, size)) - sblock->fs_bsize * UFS_NDADDR;
    138 	sizepb = sblock->fs_bsize;
    139 	for (i = 0; i < UFS_NIADDR; i++) {
    140 		if (DIP(&dino, ib[i])) {
    141 			if (is_ufs2)
    142 				idesc->id_blkno = iswap64(dino.dp2.di_ib[i]);
    143 			else
    144 				idesc->id_blkno = iswap32(dino.dp1.di_ib[i]);
    145 			ret = iblock(idesc, i + 1, remsize);
    146 			if (ret & STOP)
    147 				return (ret);
    148 		} else {
    149 			if (idesc->id_type == DATA && remsize > 0) {
    150 				/* An empty block in a directory XXX */
    151 				markclean = 0;
    152 				getpathname(pathbuf, sizeof(pathbuf),
    153 				    idesc->id_number, idesc->id_number);
    154 				pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
    155 				    pathbuf);
    156 				if (reply("ADJUST LENGTH") == 1) {
    157 					dp = ginode(idesc->id_number);
    158 					DIP_SET(dp, size,
    159 					    iswap64(iswap64(DIP(dp, size))
    160 						- remsize));
    161 					remsize = 0;
    162 					printf(
    163 					    "YOU MUST RERUN FSCK AFTERWARDS\n");
    164 					rerun = 1;
    165 					inodirty();
    166 					break;
    167 				}
    168 			}
    169 		}
    170 		sizepb *= FFS_NINDIR(sblock);
    171 		remsize -= sizepb;
    172 	}
    173 	return (KEEPON);
    174 }
    175 
    176 static int
    177 iblock(struct inodesc *idesc, long ilevel, u_int64_t isize)
    178 {
    179 	struct bufarea *bp;
    180 	int i, n, (*func) (struct inodesc *), nif;
    181 	u_int64_t sizepb;
    182 	char buf[BUFSIZ];
    183 	char pathbuf[MAXPATHLEN + 1];
    184 	union dinode *dp;
    185 
    186 	if (idesc->id_type != DATA) {
    187 		func = idesc->id_func;
    188 		if (((n = (*func)(idesc)) & KEEPON) == 0)
    189 			return (n);
    190 	} else
    191 		func = dirscan;
    192 	if (chkrange(idesc->id_blkno, idesc->id_numfrags))
    193 		return (SKIP);
    194 	bp = getdatablk(idesc->id_blkno, sblock->fs_bsize);
    195 	ilevel--;
    196 	for (sizepb = sblock->fs_bsize, i = 0; i < ilevel; i++)
    197 		sizepb *= FFS_NINDIR(sblock);
    198 	if (howmany(isize, sizepb) > (size_t)FFS_NINDIR(sblock))
    199 		nif = FFS_NINDIR(sblock);
    200 	else
    201 		nif = howmany(isize, sizepb);
    202 	if (do_blkswap) { /* swap byte order of the whole blk */
    203 		if (is_ufs2) {
    204 			for (i = 0; i < nif; i++)
    205 				bp->b_un.b_indir2[i] =
    206 				    bswap64(bp->b_un.b_indir2[i]);
    207 		} else {
    208 			for (i = 0; i < nif; i++)
    209 				bp->b_un.b_indir1[i] =
    210 				    bswap32(bp->b_un.b_indir1[i]);
    211 		}
    212 		dirty(bp);
    213 		flush(fswritefd, bp);
    214 	}
    215 	if (idesc->id_func == pass1check && nif < FFS_NINDIR(sblock)) {
    216 		for (i = nif; i < FFS_NINDIR(sblock); i++) {
    217 			if (IBLK(bp, i) == 0)
    218 				continue;
    219 			(void)snprintf(buf, sizeof(buf),
    220 			    "PARTIALLY TRUNCATED INODE I=%llu",
    221 			    (unsigned long long)idesc->id_number);
    222 			if (dofix(idesc, buf)) {
    223 				IBLK_SET(bp, i, 0);
    224 				dirty(bp);
    225 			} else
    226 				markclean = 0;
    227 		}
    228 		flush(fswritefd, bp);
    229 	}
    230 	for (i = 0; i < nif; i++) {
    231 		if (IBLK(bp, i)) {
    232 			if (is_ufs2)
    233 				idesc->id_blkno = iswap64(bp->b_un.b_indir2[i]);
    234 			else
    235 				idesc->id_blkno = iswap32(bp->b_un.b_indir1[i]);
    236 			if (ilevel == 0)
    237 				n = (*func)(idesc);
    238 			else
    239 				n = iblock(idesc, ilevel, isize);
    240 			if (n & STOP) {
    241 				bp->b_flags &= ~B_INUSE;
    242 				return (n);
    243 			}
    244 		} else {
    245 			if (idesc->id_type == DATA && isize > 0) {
    246 				/* An empty block in a directory XXX */
    247 				markclean = 0;
    248 				getpathname(pathbuf, sizeof(pathbuf),
    249 				    idesc->id_number, idesc->id_number);
    250 				pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
    251 				    pathbuf);
    252 				if (reply("ADJUST LENGTH") == 1) {
    253 					dp = ginode(idesc->id_number);
    254 					DIP_SET(dp, size,
    255 					    iswap64(iswap64(DIP(dp, size))
    256 						- isize));
    257 					isize = 0;
    258 					printf(
    259 					    "YOU MUST RERUN FSCK AFTERWARDS\n");
    260 					rerun = 1;
    261 					inodirty();
    262 					bp->b_flags &= ~B_INUSE;
    263 					return(STOP);
    264 				}
    265 			}
    266 		}
    267 		isize -= sizepb;
    268 	}
    269 	bp->b_flags &= ~B_INUSE;
    270 	return (KEEPON);
    271 }
    272 
    273 /*
    274  * Check that a block in a legal block number.
    275  * Return 0 if in range, 1 if out of range.
    276  */
    277 int
    278 chkrange(daddr_t blk, int cnt)
    279 {
    280 	int c;
    281 
    282 	if (cnt <= 0 || blk <= 0 || blk > maxfsblock ||
    283 	    cnt - 1 > maxfsblock - blk)
    284 		return (1);
    285 	if (cnt > sblock->fs_frag ||
    286 	    ffs_fragnum(sblock, blk) + cnt > sblock->fs_frag) {
    287 		if (debug)
    288 			printf("bad size: blk %lld, offset %d, size %d\n",
    289 			    (long long)blk, (int)ffs_fragnum(sblock, blk), cnt);
    290 	}
    291 	c = dtog(sblock, blk);
    292 	if (blk < cgdmin(sblock, c)) {
    293 		if ((blk + cnt) > cgsblock(sblock, c)) {
    294 			if (debug) {
    295 				printf("blk %lld < cgdmin %lld;",
    296 				    (long long)blk,
    297 				    (long long)cgdmin(sblock, c));
    298 				printf(" blk + cnt %lld > cgsbase %lld\n",
    299 				    (long long)(blk + cnt),
    300 				    (long long)cgsblock(sblock, c));
    301 			}
    302 			return (1);
    303 		}
    304 	} else {
    305 		if ((blk + cnt) > cgbase(sblock, c+1)) {
    306 			if (debug)  {
    307 				printf("blk %lld >= cgdmin %lld;",
    308 				    (long long)blk,
    309 				    (long long)cgdmin(sblock, c));
    310 				printf(" blk + cnt %lld > sblock->fs_fpg %d\n",
    311 				    (long long)(blk+cnt), sblock->fs_fpg);
    312 			}
    313 			return (1);
    314 		}
    315 	}
    316 	return (0);
    317 }
    318 
    319 /*
    320  * General purpose interface for reading inodes.
    321  */
    322 union dinode *
    323 ginode(ino_t inumber)
    324 {
    325 	daddr_t iblk;
    326 	int blkoff;
    327 
    328 	if (inumber < UFS_ROOTINO || inumber > maxino)
    329 		errexit("bad inode number %llu to ginode",
    330 		    (unsigned long long)inumber);
    331 	if (startinum == 0 ||
    332 	    inumber < startinum || inumber >= startinum + FFS_INOPB(sblock)) {
    333 		iblk = ino_to_fsba(sblock, inumber);
    334 		if (pbp != 0)
    335 			pbp->b_flags &= ~B_INUSE;
    336 		pbp = getdatablk(iblk, sblock->fs_bsize);
    337 		startinum = (inumber / FFS_INOPB(sblock)) * FFS_INOPB(sblock);
    338 	}
    339 	if (is_ufs2) {
    340 		blkoff = (inumber % FFS_INOPB(sblock)) * DINODE2_SIZE;
    341 		return ((union dinode *)((caddr_t)pbp->b_un.b_buf + blkoff));
    342 	}
    343 	blkoff = (inumber % FFS_INOPB(sblock)) * DINODE1_SIZE;
    344 	return ((union dinode *)((caddr_t)pbp->b_un.b_buf + blkoff));
    345 }
    346 
    347 #ifndef NO_FFS_EI
    348 static void
    349 swap_dinode1(union dinode *dp, int n)
    350 {
    351 	int i, j;
    352 	struct ufs1_dinode *dp1;
    353 	int32_t maxsymlinklen = sblock->fs_maxsymlinklen;
    354 	if (isappleufs)
    355 		maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
    356 
    357 	dp1 = (struct ufs1_dinode *)&dp->dp1;
    358 	for (i = 0; i < n; i++, dp1++) {
    359 		ffs_dinode1_swap(dp1, dp1);
    360 		if (((iswap16(dp1->di_mode) & IFMT) != IFLNK) ||
    361 		    doinglevel2 ||
    362 		    (maxsymlinklen < 0) ||
    363 		    (iswap64(dp1->di_size) > (uint64_t)maxsymlinklen)) {
    364 			for (j = 0; j < UFS_NDADDR; j++)
    365 			    dp1->di_db[j] = bswap32(dp1->di_db[j]);
    366 			for (j = 0; j < UFS_NIADDR; j++)
    367 			    dp1->di_ib[j] = bswap32(dp1->di_ib[j]);
    368 		}
    369 	}
    370 }
    371 
    372 static void
    373 swap_dinode2(union dinode *dp, int n)
    374 {
    375 	int i, j;
    376 	struct ufs2_dinode *dp2;
    377 
    378 	dp2 = (struct ufs2_dinode *)&dp->dp2;
    379 	for (i = 0; i < n; i++, dp2++) {
    380 		ffs_dinode2_swap(dp2, dp2);
    381 		if ((iswap16(dp2->di_mode) & IFMT) != IFLNK) {
    382 			for (j = 0; j < UFS_NXADDR; j++)
    383 				dp2->di_extb[j] = bswap64(dp2->di_extb[j]);
    384 			for (j = 0; j < UFS_NDADDR; j++)
    385 				dp2->di_db[j] = bswap64(dp2->di_db[j]);
    386 			for (j = 0; j < UFS_NIADDR; j++)
    387 				dp2->di_ib[j] = bswap64(dp2->di_ib[j]);
    388 		}
    389 	}
    390 }
    391 #endif /* !NO_FFS_EI */
    392 
    393 /*
    394  * Special purpose version of ginode used to optimize first pass
    395  * over all the inodes in numerical order.
    396  */
    397 ino_t nextino, lastinum, lastvalidinum;
    398 long readcnt, readpercg, fullcnt, inobufsize, partialcnt, partialsize;
    399 union dinode *inodebuf;
    400 
    401 union dinode *
    402 getnextinode(ino_t inumber)
    403 {
    404 	long size;
    405 	daddr_t dblk;
    406 	static union dinode *dp;
    407 	union dinode *ret;
    408 
    409 	if (inumber != nextino++ || inumber > lastvalidinum)
    410 		errexit("bad inode number %llu to nextinode",
    411 		    (unsigned long long)inumber);
    412 
    413 	if (inumber >= lastinum) {
    414 		readcnt++;
    415 		dblk = FFS_FSBTODB(sblock, ino_to_fsba(sblock, lastinum));
    416 		if (readcnt % readpercg == 0) {
    417 			size = partialsize;
    418 			lastinum += partialcnt;
    419 		} else {
    420 			size = inobufsize;
    421 			lastinum += fullcnt;
    422 		}
    423 		(void)bread(fsreadfd, (caddr_t)inodebuf, dblk, size);
    424 		if (doswap) {
    425 			if (is_ufs2)
    426 				swap_dinode2(inodebuf, lastinum - inumber);
    427 			else
    428 				swap_dinode1(inodebuf, lastinum - inumber);
    429 			bwrite(fswritefd, (char *)inodebuf, dblk, size);
    430 		}
    431 		dp = (union dinode *)inodebuf;
    432 	}
    433 	ret = dp;
    434 	dp = (union dinode *)
    435 	    ((char *)dp + (is_ufs2 ? DINODE2_SIZE : DINODE1_SIZE));
    436 	return ret;
    437 }
    438 
    439 void
    440 setinodebuf(ino_t inum)
    441 {
    442 
    443 	if (inum % sblock->fs_ipg != 0)
    444 		errexit("bad inode number %llu to setinodebuf",
    445 		    (unsigned long long)inum);
    446 
    447 	lastvalidinum = inum + sblock->fs_ipg - 1;
    448 	startinum = 0;
    449 	nextino = inum;
    450 	lastinum = inum;
    451 	readcnt = 0;
    452 	if (inodebuf != NULL)
    453 		return;
    454 	inobufsize = ffs_blkroundup(sblock, INOBUFSIZE);
    455 	fullcnt = inobufsize / (is_ufs2 ? DINODE2_SIZE : DINODE1_SIZE);
    456 	readpercg = sblock->fs_ipg / fullcnt;
    457 	partialcnt = sblock->fs_ipg % fullcnt;
    458 	partialsize = partialcnt * (is_ufs2 ? DINODE2_SIZE : DINODE1_SIZE);
    459 	if (partialcnt != 0) {
    460 		readpercg++;
    461 	} else {
    462 		partialcnt = fullcnt;
    463 		partialsize = inobufsize;
    464 	}
    465 	if (inodebuf == NULL &&
    466 	    (inodebuf = aligned_alloc(DEV_BSIZE, (unsigned)inobufsize)) == NULL)
    467 		errexit("Cannot allocate space for inode buffer");
    468 }
    469 
    470 void
    471 freeinodebuf(void)
    472 {
    473 
    474 	if (inodebuf != NULL)
    475 		free((char *)inodebuf);
    476 	inodebuf = NULL;
    477 }
    478 
    479 /*
    480  * Routines to maintain information about directory inodes.
    481  * This is built during the first pass and used during the
    482  * second and third passes.
    483  *
    484  * Enter inodes into the cache.
    485  */
    486 void
    487 cacheino(union dinode *dp, ino_t inumber)
    488 {
    489 	struct inoinfo *inp;
    490 	struct inoinfo **inpp, **ninpsort;
    491 	unsigned int i, blks, extra;
    492 	int64_t size;
    493 
    494 	size = iswap64(DIP(dp, size));
    495 	blks = howmany(size, sblock->fs_bsize);
    496 	if (blks > UFS_NDADDR)
    497 		blks = UFS_NDADDR + UFS_NIADDR;
    498 	if (blks > 0)
    499 		extra = (blks - 1) * sizeof (int64_t);
    500 	else
    501 		extra = 0;
    502 	inp = malloc(sizeof(*inp) + extra);
    503 	if (inp == NULL)
    504 		return;
    505 	inpp = &inphead[inumber % dirhash];
    506 	inp->i_nexthash = *inpp;
    507 	*inpp = inp;
    508 	inp->i_child = inp->i_sibling = 0;
    509 	if (inumber == UFS_ROOTINO)
    510 		inp->i_parent = UFS_ROOTINO;
    511 	else
    512 		inp->i_parent = (ino_t)0;
    513 	inp->i_dotdot = (ino_t)0;
    514 	inp->i_number = inumber;
    515 	inp->i_isize = size;
    516 	inp->i_numblks = blks;
    517 	for (i = 0; i < (blks < UFS_NDADDR ? blks : UFS_NDADDR); i++)
    518 		inp->i_blks[i] = DIP(dp, db[i]);
    519 	if (blks > UFS_NDADDR)
    520 		for (i = 0; i < UFS_NIADDR; i++)
    521 			inp->i_blks[UFS_NDADDR + i] = DIP(dp, ib[i]);
    522 	if (inplast == listmax) {
    523 		ninpsort = (struct inoinfo **)realloc((char *)inpsort,
    524 		    (unsigned)(listmax + 100) * sizeof(struct inoinfo *));
    525 		if (inpsort == NULL)
    526 			errexit("cannot increase directory list");
    527 		inpsort = ninpsort;
    528 		listmax += 100;
    529 	}
    530 	inpsort[inplast++] = inp;
    531 }
    532 
    533 /*
    534  * Look up an inode cache structure.
    535  */
    536 struct inoinfo *
    537 getinoinfo(ino_t inumber)
    538 {
    539 	struct inoinfo *inp;
    540 
    541 	for (inp = inphead[inumber % dirhash]; inp; inp = inp->i_nexthash) {
    542 		if (inp->i_number != inumber)
    543 			continue;
    544 		return (inp);
    545 	}
    546 	errexit("cannot find inode %llu", (unsigned long long)inumber);
    547 	return ((struct inoinfo *)0);
    548 }
    549 
    550 /*
    551  * Clean up all the inode cache structure.
    552  */
    553 void
    554 inocleanup(void)
    555 {
    556 	struct inoinfo **inpp;
    557 
    558 	if (inphead == NULL)
    559 		return;
    560 	for (inpp = &inpsort[inplast - 1]; inpp >= inpsort; inpp--)
    561 		free((char *)(*inpp));
    562 	free((char *)inphead);
    563 	free((char *)inpsort);
    564 	inphead = inpsort = NULL;
    565 }
    566 
    567 void
    568 inodirty(void)
    569 {
    570 
    571 	dirty(pbp);
    572 }
    573 
    574 void
    575 clri(struct inodesc *idesc, const char *type, int flag)
    576 {
    577 	union dinode *dp;
    578 
    579 	dp = ginode(idesc->id_number);
    580 	if (flag == 1) {
    581 		pwarn("%s %s", type,
    582 		    (iswap16(DIP(dp, mode)) & IFMT) == IFDIR ? "DIR" : "FILE");
    583 		pinode(idesc->id_number);
    584 	}
    585 	if (preen || reply("CLEAR") == 1) {
    586 		if (preen)
    587 			printf(" (CLEARED)\n");
    588 		n_files--;
    589 		/*
    590 		 * ckinode will call id_func (actually always pass4check)
    591 		 * which will update the block count
    592 		 */
    593 		if (idesc->id_type != SNAP)
    594 			update_uquot(idesc->id_number,
    595 			    idesc->id_uid, idesc->id_gid, 0, -1);
    596 		(void)ckinode(dp, idesc);
    597 		clearinode(dp);
    598 		inoinfo(idesc->id_number)->ino_state = USTATE;
    599 		inodirty();
    600 	} else
    601 		markclean = 0;
    602 }
    603 
    604 int
    605 findname(struct inodesc *idesc)
    606 {
    607 	struct direct *dirp = idesc->id_dirp;
    608 	size_t len;
    609 	char *buf;
    610 
    611 	if (iswap32(dirp->d_ino) != idesc->id_parent || idesc->id_entryno < 2) {
    612 		idesc->id_entryno++;
    613 		return (KEEPON);
    614 	}
    615 	if ((len = dirp->d_namlen + 1) > MAXPATHLEN) {
    616 		/* XXX: We don't fix but we ignore */
    617 		len = MAXPATHLEN;
    618 	}
    619 	/* this is namebuf from utilities.c */
    620 	buf = __UNCONST(idesc->id_name);
    621 	(void)memcpy(buf, dirp->d_name, (size_t)dirp->d_namlen + 1);
    622 	return (STOP|FOUND);
    623 }
    624 
    625 int
    626 findino(struct inodesc *idesc)
    627 {
    628 	struct direct *dirp = idesc->id_dirp;
    629 
    630 	if (dirp->d_ino == 0)
    631 		return (KEEPON);
    632 	if (strcmp(dirp->d_name, idesc->id_name) == 0 &&
    633 	    iswap32(dirp->d_ino) >= UFS_ROOTINO && iswap32(dirp->d_ino) <= maxino) {
    634 		idesc->id_parent = iswap32(dirp->d_ino);
    635 		return (STOP|FOUND);
    636 	}
    637 	return (KEEPON);
    638 }
    639 
    640 int
    641 clearentry(struct inodesc *idesc)
    642 {
    643 	struct direct *dirp = idesc->id_dirp;
    644 
    645 	if (dirp->d_ino != idesc->id_parent || idesc->id_entryno < 2) {
    646 		idesc->id_entryno++;
    647 		return (KEEPON);
    648 	}
    649 	dirp->d_ino = 0;
    650 	return (STOP|FOUND|ALTERED);
    651 }
    652 
    653 void
    654 pinode(ino_t ino)
    655 {
    656 	union dinode *dp;
    657 	struct passwd *pw;
    658 
    659 	printf(" I=%llu ", (unsigned long long)ino);
    660 	if (ino < UFS_ROOTINO || ino > maxino)
    661 		return;
    662 	dp = ginode(ino);
    663 	printf(" OWNER=");
    664 #ifndef SMALL
    665 	if (Uflag && (pw = getpwuid((int)iswap32(DIP(dp, uid)))) != 0)
    666 		printf("%s ", pw->pw_name);
    667 	else
    668 #endif
    669 		printf("%u ", (unsigned)iswap32(DIP(dp, uid)));
    670 	printf("MODE=%o\n", iswap16(DIP(dp, mode)));
    671 	if (preen)
    672 		printf("%s: ", cdevname());
    673 	printf("SIZE=%llu ", (unsigned long long)iswap64(DIP(dp, size)));
    674 	printf("MTIME=%s ", print_mtime(iswap32(DIP(dp, mtime))));
    675 }
    676 
    677 void
    678 blkerror(ino_t ino, const char *type, daddr_t blk)
    679 {
    680 	struct inostat *info;
    681 
    682 	pfatal("%lld %s I=%llu", (long long)blk, type, (unsigned long long)ino);
    683 	printf("\n");
    684 	info = inoinfo(ino);
    685 	switch (info->ino_state) {
    686 
    687 	case FSTATE:
    688 		info->ino_state = FCLEAR;
    689 		return;
    690 
    691 	case DSTATE:
    692 		info->ino_state = DCLEAR;
    693 		return;
    694 
    695 	case FCLEAR:
    696 	case DCLEAR:
    697 		return;
    698 
    699 	default:
    700 		errexit("BAD STATE %d TO BLKERR", info->ino_state);
    701 		/* NOTREACHED */
    702 	}
    703 }
    704 
    705 /*
    706  * allocate an unused inode
    707  */
    708 ino_t
    709 allocino(ino_t request, int type)
    710 {
    711 	ino_t ino;
    712 	union dinode *dp;
    713 	struct ufs1_dinode *dp1;
    714 	struct ufs2_dinode *dp2;
    715 	time_t t;
    716 	struct cg *cgp = cgrp;
    717 	int cg;
    718 	struct inostat *info = NULL;
    719 	int nfrags;
    720 
    721 	if (request == 0)
    722 		request = UFS_ROOTINO;
    723 	else if (inoinfo(request)->ino_state != USTATE)
    724 		return (0);
    725 	for (ino = request; ino < maxino; ino++) {
    726 		info = inoinfo(ino);
    727 		if (info->ino_state == USTATE)
    728 			break;
    729 	}
    730 	if (ino == maxino)
    731 		return (0);
    732 	cg = ino_to_cg(sblock, ino);
    733 	/* If necessary, extend the inoinfo array. grow exponentially */
    734 	if ((ino % sblock->fs_ipg) >= inostathead[cg].il_numalloced) {
    735 		size_t newalloced, i;
    736 		newalloced = MIN(sblock->fs_ipg,
    737 			MAX(2 * inostathead[cg].il_numalloced, 10));
    738 		info = calloc(newalloced, sizeof(*info));
    739 		if (info == NULL) {
    740 			pwarn("cannot alloc %zu bytes to extend inoinfo\n",
    741 				sizeof(struct inostat) * newalloced);
    742 			return 0;
    743 		}
    744 		memmove(info, inostathead[cg].il_stat,
    745 			inostathead[cg].il_numalloced * sizeof(*info));
    746 		for (i = inostathead[cg].il_numalloced; i < newalloced; i++) {
    747 			info[i].ino_state = USTATE;
    748 		}
    749 		if (inostathead[cg].il_numalloced)
    750 			free(inostathead[cg].il_stat);
    751 		inostathead[cg].il_stat = info;
    752 		inostathead[cg].il_numalloced = newalloced;
    753 		info = inoinfo(ino);
    754 	}
    755 	getblk(&cgblk, cgtod(sblock, cg), sblock->fs_cgsize);
    756 	memcpy(cgp, cgblk.b_un.b_cg, sblock->fs_cgsize);
    757 	if ((doswap && !needswap) || (!doswap && needswap))
    758 		ffs_cg_swap(cgblk.b_un.b_cg, cgp, sblock);
    759 	if (!cg_chkmagic(cgp, 0))
    760 		pfatal("CG %d: ALLOCINO: BAD MAGIC NUMBER\n", cg);
    761 	if (doswap)
    762 		cgdirty();
    763 	setbit(cg_inosused(cgp, 0), ino % sblock->fs_ipg);
    764 	cgp->cg_cs.cs_nifree--;
    765 	sblock->fs_cstotal.cs_nifree--;
    766 	sblock->fs_cs(fs, cg).cs_nifree--;
    767 	sbdirty();
    768 	switch (type & IFMT) {
    769 	case IFDIR:
    770 		info->ino_state = DSTATE;
    771 		cgp->cg_cs.cs_ndir++;
    772 		nfrags = 1;
    773 		break;
    774 	case IFREG:
    775 		info->ino_state = FSTATE;
    776 		nfrags = sblock->fs_frag;
    777 		break;
    778 	case IFLNK:
    779 		info->ino_state = FSTATE;
    780 		nfrags = 1;
    781 		break;
    782 	default:
    783 		return (0);
    784 	}
    785 	cgdirty();
    786 	dp = ginode(ino);
    787 	if (is_ufs2) {
    788 		dp2 = &dp->dp2;
    789 		dp2->di_db[0] = iswap64(allocblk(nfrags));
    790 		if (dp2->di_db[0] == 0) {
    791 			info->ino_state = USTATE;
    792 			return (0);
    793 		}
    794 		dp2->di_mode = iswap16(type);
    795 		dp2->di_flags = 0;
    796 		(void)time(&t);
    797 		dp2->di_atime = iswap64(t);
    798 		dp2->di_mtime = dp2->di_ctime = dp2->di_atime;
    799 		dp2->di_size = iswap64(ffs_lfragtosize(sblock, nfrags));
    800 		dp2->di_blocks = iswap64(btodb(ffs_lfragtosize(sblock, nfrags)));
    801 	} else {
    802 		dp1 = &dp->dp1;
    803 		dp1->di_db[0] = iswap32(allocblk(nfrags));
    804 		if (dp1->di_db[0] == 0) {
    805 			info->ino_state = USTATE;
    806 			return (0);
    807 		}
    808 		dp1->di_mode = iswap16(type);
    809 		dp1->di_flags = 0;
    810 		(void)time(&t);
    811 		dp1->di_atime = iswap32(t);
    812 		dp1->di_mtime = dp1->di_ctime = dp1->di_atime;
    813 		dp1->di_size = iswap64(ffs_lfragtosize(sblock, nfrags));
    814 		dp1->di_blocks = iswap32(btodb(ffs_lfragtosize(sblock, nfrags)));
    815 	}
    816 	n_files++;
    817 	inodirty();
    818 	if (newinofmt)
    819 		info->ino_type = IFTODT(type);
    820 	return (ino);
    821 }
    822 
    823 /*
    824  * deallocate an inode
    825  */
    826 void
    827 freeino(ino_t ino)
    828 {
    829 	struct inodesc idesc;
    830 	union dinode *dp;
    831 	struct cg *cgp = cgrp;
    832 	int cg;
    833 
    834 	cg = ino_to_cg(sblock, ino);
    835 	getblk(&cgblk, cgtod(sblock, cg), sblock->fs_cgsize);
    836 	memcpy(cgp, cgblk.b_un.b_cg, sblock->fs_cgsize);
    837 	if ((doswap && !needswap) || (!doswap && needswap))
    838 		ffs_cg_swap(cgblk.b_un.b_cg, cgp, sblock);
    839 	if (!cg_chkmagic(cgp, 0)) {
    840 		pwarn("CG %d: FREEINO: BAD MAGIC NUMBER\n", cg);
    841 		cgp = NULL;
    842 	}
    843 
    844 	memset(&idesc, 0, sizeof(struct inodesc));
    845 	idesc.id_func = pass4check;
    846 	idesc.id_number = ino;
    847 	dp = ginode(ino);
    848 	idesc.id_uid = iswap32(DIP(dp, uid));
    849 	idesc.id_gid = iswap32(DIP(dp, gid));
    850 	if (iswap32(DIP(dp, flags)) & SF_SNAPSHOT)
    851 		idesc.id_type = SNAP;
    852 	else
    853 		idesc.id_type = ADDR;
    854 	(void)ckinode(dp, &idesc);
    855 	clearinode(dp);
    856 	inodirty();
    857 	inoinfo(ino)->ino_state = USTATE;
    858 	if (idesc.id_type != SNAP)
    859 		update_uquot(idesc.id_number,
    860 		    idesc.id_uid, idesc.id_gid, 0, -1);
    861 	n_files--;
    862 	if (cgp) {
    863 		clrbit(cg_inosused(cgp, 0), ino % sblock->fs_ipg);
    864 		cgp->cg_cs.cs_nifree++;
    865 		sblock->fs_cstotal.cs_nifree++;
    866 		sblock->fs_cs(fs, cg).cs_nifree++;
    867 		sbdirty();
    868 		cgdirty();
    869 	}
    870 }
    871 
    872 /* read a data block from inode */
    873 ssize_t
    874 readblk(union dinode *dp, off_t offset, struct bufarea **bp)
    875 {
    876 	daddr_t blkno = ffs_lblkno(sblock, offset);
    877 	daddr_t iblkno;
    878 	int type = IFMT & iswap16(DIP(dp, mode));
    879 	ssize_t filesize = iswap64(DIP(dp, size));
    880 	int ilevel;
    881 	daddr_t nblks;
    882 	const daddr_t naddrperblk = sblock->fs_bsize /
    883 	    (is_ufs2 ? sizeof(uint64_t) : sizeof(uint32_t));
    884 	struct bufarea *ibp;
    885 
    886 	*bp = NULL;
    887 	offset &= ~(sblock->fs_bsize - 1);
    888 
    889 	if (type != IFREG)
    890 		return 0;
    891 	if (offset >= filesize)
    892 		return 0; /* short read */
    893 	if (blkno < UFS_NDADDR) {
    894 		blkno = is_ufs2 ? iswap64(dp->dp2.di_db[blkno]) :
    895 		    iswap32(dp->dp1.di_db[blkno]);
    896 		if (blkno == 0)
    897 			return 0;
    898 		*bp = getdatablk(blkno, sblock->fs_bsize);
    899 		return (bp != NULL) ? sblock->fs_bsize : 0;
    900 	}
    901 	blkno -= UFS_NDADDR;
    902 	/* find indir level */
    903 	for (ilevel = 1, nblks = naddrperblk;
    904 	     ilevel <= UFS_NIADDR;
    905 	     ilevel++, nblks *= naddrperblk) {
    906 		if (blkno < nblks)
    907 			break;
    908 		else
    909 			blkno -= nblks;
    910 	}
    911 	if (ilevel > UFS_NIADDR)
    912 		errexit("bad offset %" PRIu64 " to readblk", offset);
    913 
    914 	/* get the first indirect block */
    915 	iblkno = is_ufs2 ? iswap64(dp->dp2.di_ib[ilevel - 1]) :
    916 		    iswap32(dp->dp1.di_ib[ilevel - 1]);
    917 	if (iblkno == 0)
    918 		return 0;
    919 	ibp = getdatablk(iblkno, sblock->fs_bsize);
    920 	/* walk indirect blocks up to the data block */
    921 	for (; ilevel >0 ; ilevel--) {
    922 		nblks = nblks / naddrperblk;
    923 		if (is_ufs2)
    924 			iblkno = iswap64(ibp->b_un.b_indir2[blkno / nblks]);
    925 		else
    926 			iblkno = iswap32(ibp->b_un.b_indir1[blkno / nblks]);
    927 		if (iblkno == 0)
    928 			return 0;
    929 		blkno = blkno % nblks;
    930 		ibp->b_flags &= ~B_INUSE;
    931 		ibp = getdatablk(iblkno, sblock->fs_bsize);
    932 	}
    933 	*bp = ibp;
    934 	return sblock->fs_bsize;
    935 }
    936 
    937 static struct bufarea * getnewblk(daddr_t *);
    938 static struct bufarea *
    939 getnewblk(daddr_t *blkno)
    940 {
    941 	struct bufarea *bp;
    942 	*blkno = allocblk(sblock->fs_frag);
    943 	if (*blkno == 0)
    944 		return NULL;
    945 	bp = getdatablk(*blkno, sblock->fs_bsize);
    946 	memset(bp->b_un.b_buf, 0, sblock->fs_bsize);
    947 	return bp;
    948 }
    949 
    950 /* expand given inode by one full fs block */
    951 struct bufarea *
    952 expandfile(union dinode *dp)
    953 {
    954 	uint64_t filesize = iswap64(DIP(dp, size));
    955 	daddr_t newblk, blkno, iblkno, nblks;
    956 	daddr_t di_blocks;
    957 	int ilevel;
    958 	const daddr_t naddrperblk = sblock->fs_bsize /
    959 	    (is_ufs2 ? sizeof(uint64_t) : sizeof(uint32_t));
    960 	struct bufarea *ibp, *bp = NULL;
    961 
    962 	di_blocks = is_ufs2 ? iswap64(dp->dp2.di_blocks) :
    963 	    iswap32(dp->dp1.di_blocks);
    964 	/* compute location of new block */
    965 	blkno = ffs_lblkno(sblock, filesize);
    966 
    967 	if (blkno < UFS_NDADDR) {
    968 		/* easy way: allocate a direct block */
    969 		if ((bp = getnewblk(&newblk)) == NULL) {
    970 			return NULL;
    971 		}
    972 		di_blocks += btodb(sblock->fs_bsize);
    973 
    974 		if (is_ufs2) {
    975 			dp->dp2.di_db[blkno] = iswap64(newblk);
    976 		} else {
    977 			dp->dp1.di_db[blkno] = iswap32(newblk);
    978 		}
    979 		goto out;
    980 	}
    981 	blkno -= UFS_NDADDR;
    982 	/* find indir level */
    983 	for (ilevel = 1, nblks = naddrperblk;
    984 	     ilevel <= UFS_NIADDR;
    985 	     ilevel++, nblks *= naddrperblk) {
    986 		if (blkno < nblks)
    987 			break;
    988 		else
    989 			blkno -= nblks;
    990 	}
    991 	if (ilevel > UFS_NIADDR)
    992 		errexit("bad filesize %" PRIu64 " to expandfile", filesize);
    993 
    994 	/* get the first indirect block, allocating if needed */
    995 	if ((is_ufs2 ? iswap64(dp->dp2.di_ib[ilevel - 1]) :
    996 		iswap32(dp->dp1.di_ib[ilevel - 1])) == 0) {
    997 		if ((ibp = getnewblk(&newblk)) == NULL)
    998 			return 0;
    999 		di_blocks += btodb(sblock->fs_bsize);
   1000 		if (is_ufs2)
   1001 			dp->dp2.di_ib[ilevel - 1] = iswap64(newblk);
   1002 		else
   1003 			dp->dp1.di_ib[ilevel - 1] = iswap32(newblk);
   1004 	} else {
   1005 		ibp = getdatablk(is_ufs2 ? iswap64(dp->dp2.di_ib[ilevel - 1]) :
   1006 		    iswap32(dp->dp1.di_ib[ilevel - 1]), sblock->fs_bsize);
   1007 	}
   1008 	/* walk indirect blocks up to the data block */
   1009 	for (; ilevel >0 ; ilevel--) {
   1010 		nblks = nblks / naddrperblk;
   1011 		if (is_ufs2)
   1012 			iblkno = iswap64(ibp->b_un.b_indir2[blkno / nblks]);
   1013 		else
   1014 			iblkno = iswap32(ibp->b_un.b_indir1[blkno / nblks]);
   1015 		if (iblkno == 0) {
   1016 			if ((bp = getnewblk(&newblk)) == NULL)
   1017 				return NULL;
   1018 			di_blocks += btodb(sblock->fs_bsize);
   1019 			if (is_ufs2)
   1020 				ibp->b_un.b_indir2[blkno / nblks] =
   1021 				    iswap64(newblk);
   1022 			else
   1023 				ibp->b_un.b_indir1[blkno / nblks] =
   1024 				    iswap32(newblk);
   1025 			dirty(ibp);
   1026 			ibp->b_flags &= ~B_INUSE;
   1027 			ibp = bp;
   1028 		} else {
   1029 			ibp->b_flags &= ~B_INUSE;
   1030 			ibp = getdatablk(iblkno, sblock->fs_bsize);
   1031 			bp = NULL;
   1032 		}
   1033 		blkno = blkno % nblks;
   1034 	}
   1035 	if (bp == NULL) {
   1036 		errexit("INTERNAL ERROR: "
   1037 		    "expandfile() failed to allocate a new block\n");
   1038 	}
   1039 
   1040 out:
   1041 	filesize += sblock->fs_bsize;
   1042 	if (is_ufs2) {
   1043 		dp->dp2.di_size = iswap64(filesize);
   1044 		dp->dp2.di_blocks = iswap64(di_blocks);
   1045 	} else {
   1046 		dp->dp1.di_size = iswap64(filesize);
   1047 		dp->dp1.di_blocks = iswap32(di_blocks);
   1048 	}
   1049 	inodirty();
   1050 	return bp;
   1051 }
   1052