Home | History | Annotate | Line # | Download | only in fsck_lfs
lfs.c revision 1.9
      1 /* $NetBSD: lfs.c,v 1.9 2005/03/25 20:16:37 perseant Exp $ */
      2 /*-
      3  * Copyright (c) 2003 The NetBSD Foundation, Inc.
      4  * All rights reserved.
      5  *
      6  * This code is derived from software contributed to The NetBSD Foundation
      7  * by Konrad E. Schroder <perseant (at) hhhh.org>.
      8  *
      9  * Redistribution and use in source and binary forms, with or without
     10  * modification, are permitted provided that the following conditions
     11  * are met:
     12  * 1. Redistributions of source code must retain the above copyright
     13  *    notice, this list of conditions and the following disclaimer.
     14  * 2. Redistributions in binary form must reproduce the above copyright
     15  *    notice, this list of conditions and the following disclaimer in the
     16  *    documentation and/or other materials provided with the distribution.
     17  * 3. All advertising materials mentioning features or use of this software
     18  *    must display the following acknowledgement:
     19  *	This product includes software developed by the NetBSD
     20  *	Foundation, Inc. and its contributors.
     21  * 4. Neither the name of The NetBSD Foundation nor the names of its
     22  *    contributors may be used to endorse or promote products derived
     23  *    from this software without specific prior written permission.
     24  *
     25  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     26  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     35  * POSSIBILITY OF SUCH DAMAGE.
     36  */
     37 /*
     38  * Copyright (c) 1989, 1991, 1993
     39  *	The Regents of the University of California.  All rights reserved.
     40  * (c) UNIX System Laboratories, Inc.
     41  * All or some portions of this file are derived from material licensed
     42  * to the University of California by American Telephone and Telegraph
     43  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     44  * the permission of UNIX System Laboratories, Inc.
     45  *
     46  * Redistribution and use in source and binary forms, with or without
     47  * modification, are permitted provided that the following conditions
     48  * are met:
     49  * 1. Redistributions of source code must retain the above copyright
     50  *    notice, this list of conditions and the following disclaimer.
     51  * 2. Redistributions in binary form must reproduce the above copyright
     52  *    notice, this list of conditions and the following disclaimer in the
     53  *    documentation and/or other materials provided with the distribution.
     54  * 3. Neither the name of the University nor the names of its contributors
     55  *    may be used to endorse or promote products derived from this software
     56  *    without specific prior written permission.
     57  *
     58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     68  * SUCH DAMAGE.
     69  *
     70  *	@(#)ufs_bmap.c	8.8 (Berkeley) 8/11/95
     71  */
     72 
     73 
     74 #include <sys/types.h>
     75 #include <sys/param.h>
     76 #include <sys/time.h>
     77 #include <sys/buf.h>
     78 #include <sys/mount.h>
     79 
     80 #include <ufs/ufs/inode.h>
     81 #include <ufs/ufs/ufsmount.h>
     82 #define vnode uvnode
     83 #include <ufs/lfs/lfs.h>
     84 #undef vnode
     85 
     86 #include <assert.h>
     87 #include <err.h>
     88 #include <errno.h>
     89 #include <stdarg.h>
     90 #include <stdio.h>
     91 #include <stdlib.h>
     92 #include <string.h>
     93 #include <unistd.h>
     94 
     95 #include "bufcache.h"
     96 #include "vnode.h"
     97 #include "lfs.h"
     98 #include "segwrite.h"
     99 
    100 #define panic call_panic
    101 
    102 extern u_int32_t cksum(void *, size_t);
    103 extern u_int32_t lfs_sb_cksum(struct dlfs *);
    104 extern void pwarn(const char *, ...);
    105 
    106 extern struct uvnodelst vnodelist;
    107 extern struct uvnodelst getvnodelist;
    108 extern int nvnodes;
    109 
    110 int fsdirty = 0;
    111 void (*panic_func)(int, const char *, va_list) = my_vpanic;
    112 
    113 /*
    114  * LFS buffer and uvnode operations
    115  */
    116 
    117 int
    118 lfs_vop_strategy(struct ubuf * bp)
    119 {
    120 	int count;
    121 
    122 	if (bp->b_flags & B_READ) {
    123 		count = pread(bp->b_vp->v_fd, bp->b_data, bp->b_bcount,
    124 		    dbtob(bp->b_blkno));
    125 		if (count == bp->b_bcount)
    126 			bp->b_flags |= B_DONE;
    127 	} else {
    128 		count = pwrite(bp->b_vp->v_fd, bp->b_data, bp->b_bcount,
    129 		    dbtob(bp->b_blkno));
    130 		if (count == 0) {
    131 			perror("pwrite");
    132 			return -1;
    133 		}
    134 		bp->b_flags &= ~B_DELWRI;
    135 		reassignbuf(bp, bp->b_vp);
    136 	}
    137 	return 0;
    138 }
    139 
    140 int
    141 lfs_vop_bwrite(struct ubuf * bp)
    142 {
    143 	struct lfs *fs;
    144 
    145 	fs = bp->b_vp->v_fs;
    146 	if (!(bp->b_flags & B_DELWRI)) {
    147 		fs->lfs_avail -= btofsb(fs, bp->b_bcount);
    148 	}
    149 	bp->b_flags |= B_DELWRI | B_LOCKED;
    150 	reassignbuf(bp, bp->b_vp);
    151 	brelse(bp);
    152 	return 0;
    153 }
    154 
    155 /*
    156  * ufs_bmaparray does the bmap conversion, and if requested returns the
    157  * array of logical blocks which must be traversed to get to a block.
    158  * Each entry contains the offset into that block that gets you to the
    159  * next block and the disk address of the block (if it is assigned).
    160  */
    161 int
    162 ufs_bmaparray(struct lfs * fs, struct uvnode * vp, daddr_t bn, daddr_t * bnp, struct indir * ap, int *nump)
    163 {
    164 	struct inode *ip;
    165 	struct ubuf *bp;
    166 	struct indir a[NIADDR + 1], *xap;
    167 	daddr_t daddr;
    168 	daddr_t metalbn;
    169 	int error, num;
    170 
    171 	ip = VTOI(vp);
    172 
    173 	if (bn >= 0 && bn < NDADDR) {
    174 		if (nump != NULL)
    175 			*nump = 0;
    176 		*bnp = fsbtodb(fs, ip->i_ffs1_db[bn]);
    177 		if (*bnp == 0)
    178 			*bnp = -1;
    179 		return (0);
    180 	}
    181 	xap = ap == NULL ? a : ap;
    182 	if (!nump)
    183 		nump = &num;
    184 	if ((error = ufs_getlbns(fs, vp, bn, xap, nump)) != 0)
    185 		return (error);
    186 
    187 	num = *nump;
    188 
    189 	/* Get disk address out of indirect block array */
    190 	daddr = ip->i_ffs1_ib[xap->in_off];
    191 
    192 	for (bp = NULL, ++xap; --num; ++xap) {
    193 		/* Exit the loop if there is no disk address assigned yet and
    194 		 * the indirect block isn't in the cache, or if we were
    195 		 * looking for an indirect block and we've found it. */
    196 
    197 		metalbn = xap->in_lbn;
    198 		if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn)
    199 			break;
    200 		/*
    201 		 * If we get here, we've either got the block in the cache
    202 		 * or we have a disk address for it, go fetch it.
    203 		 */
    204 		if (bp)
    205 			brelse(bp);
    206 
    207 		xap->in_exists = 1;
    208 		bp = getblk(vp, metalbn, fs->lfs_bsize);
    209 
    210 		if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
    211 			bp->b_blkno = fsbtodb(fs, daddr);
    212 			bp->b_flags |= B_READ;
    213 			VOP_STRATEGY(bp);
    214 		}
    215 		daddr = ((ufs_daddr_t *) bp->b_data)[xap->in_off];
    216 	}
    217 	if (bp)
    218 		brelse(bp);
    219 
    220 	daddr = fsbtodb(fs, (ufs_daddr_t) daddr);
    221 	*bnp = daddr == 0 ? -1 : daddr;
    222 	return (0);
    223 }
    224 
    225 /*
    226  * Create an array of logical block number/offset pairs which represent the
    227  * path of indirect blocks required to access a data block.  The first "pair"
    228  * contains the logical block number of the appropriate single, double or
    229  * triple indirect block and the offset into the inode indirect block array.
    230  * Note, the logical block number of the inode single/double/triple indirect
    231  * block appears twice in the array, once with the offset into the i_ffs1_ib and
    232  * once with the offset into the page itself.
    233  */
    234 int
    235 ufs_getlbns(struct lfs * fs, struct uvnode * vp, daddr_t bn, struct indir * ap, int *nump)
    236 {
    237 	daddr_t metalbn, realbn;
    238 	int64_t blockcnt;
    239 	int lbc;
    240 	int i, numlevels, off;
    241 	int lognindir, indir;
    242 
    243 	if (nump)
    244 		*nump = 0;
    245 	numlevels = 0;
    246 	realbn = bn;
    247 	if (bn < 0)
    248 		bn = -bn;
    249 
    250 	lognindir = -1;
    251 	for (indir = fs->lfs_nindir; indir; indir >>= 1)
    252 		++lognindir;
    253 
    254 	/* Determine the number of levels of indirection.  After this loop is
    255 	 * done, blockcnt indicates the number of data blocks possible at the
    256 	 * given level of indirection, and NIADDR - i is the number of levels
    257 	 * of indirection needed to locate the requested block. */
    258 
    259 	bn -= NDADDR;
    260 	for (lbc = 0, i = NIADDR;; i--, bn -= blockcnt) {
    261 		if (i == 0)
    262 			return (EFBIG);
    263 
    264 		lbc += lognindir;
    265 		blockcnt = (int64_t) 1 << lbc;
    266 
    267 		if (bn < blockcnt)
    268 			break;
    269 	}
    270 
    271 	/* Calculate the address of the first meta-block. */
    272 	if (realbn >= 0)
    273 		metalbn = -(realbn - bn + NIADDR - i);
    274 	else
    275 		metalbn = -(-realbn - bn + NIADDR - i);
    276 
    277 	/* At each iteration, off is the offset into the bap array which is an
    278 	 * array of disk addresses at the current level of indirection. The
    279 	 * logical block number and the offset in that block are stored into
    280 	 * the argument array. */
    281 	ap->in_lbn = metalbn;
    282 	ap->in_off = off = NIADDR - i;
    283 	ap->in_exists = 0;
    284 	ap++;
    285 	for (++numlevels; i <= NIADDR; i++) {
    286 		/* If searching for a meta-data block, quit when found. */
    287 		if (metalbn == realbn)
    288 			break;
    289 
    290 		lbc -= lognindir;
    291 		blockcnt = (int64_t) 1 << lbc;
    292 		off = (bn >> lbc) & (fs->lfs_nindir - 1);
    293 
    294 		++numlevels;
    295 		ap->in_lbn = metalbn;
    296 		ap->in_off = off;
    297 		ap->in_exists = 0;
    298 		++ap;
    299 
    300 		metalbn -= -1 + (off << lbc);
    301 	}
    302 	if (nump)
    303 		*nump = numlevels;
    304 	return (0);
    305 }
    306 
    307 int
    308 lfs_vop_bmap(struct uvnode * vp, daddr_t lbn, daddr_t * daddrp)
    309 {
    310 	return ufs_bmaparray(vp->v_fs, vp, lbn, daddrp, NULL, NULL);
    311 }
    312 
    313 /* Search a block for a specific dinode. */
    314 struct ufs1_dinode *
    315 lfs_ifind(struct lfs * fs, ino_t ino, struct ubuf * bp)
    316 {
    317 	struct ufs1_dinode *dip = (struct ufs1_dinode *) bp->b_data;
    318 	struct ufs1_dinode *ldip, *fin;
    319 
    320 	fin = dip + INOPB(fs);
    321 
    322 	/*
    323 	 * Read the inode block backwards, since later versions of the
    324 	 * inode will supercede earlier ones.  Though it is unlikely, it is
    325 	 * possible that the same inode will appear in the same inode block.
    326 	 */
    327 	for (ldip = fin - 1; ldip >= dip; --ldip)
    328 		if (ldip->di_inumber == ino)
    329 			return (ldip);
    330 	return NULL;
    331 }
    332 
    333 /*
    334  * lfs_raw_vget makes us a new vnode from the inode at the given disk address.
    335  * XXX it currently loses atime information.
    336  */
    337 struct uvnode *
    338 lfs_raw_vget(struct lfs * fs, ino_t ino, int fd, ufs_daddr_t daddr)
    339 {
    340 	struct uvnode *vp;
    341 	struct inode *ip;
    342 	struct ufs1_dinode *dip;
    343 	struct ubuf *bp;
    344 	int i;
    345 
    346 	vp = (struct uvnode *) malloc(sizeof(*vp));
    347 	memset(vp, 0, sizeof(*vp));
    348 	vp->v_fd = fd;
    349 	vp->v_fs = fs;
    350 	vp->v_usecount = 0;
    351 	vp->v_strategy_op = lfs_vop_strategy;
    352 	vp->v_bwrite_op = lfs_vop_bwrite;
    353 	vp->v_bmap_op = lfs_vop_bmap;
    354 	LIST_INIT(&vp->v_cleanblkhd);
    355 	LIST_INIT(&vp->v_dirtyblkhd);
    356 
    357 	ip = (struct inode *) malloc(sizeof(*ip));
    358 	memset(ip, 0, sizeof(*ip));
    359 
    360 	ip->i_din.ffs1_din = (struct ufs1_dinode *)
    361 	    malloc(sizeof(struct ufs1_dinode));
    362 	memset(ip->i_din.ffs1_din, 0, sizeof (struct ufs1_dinode));
    363 
    364 	/* Initialize the inode -- from lfs_vcreate. */
    365 	ip->inode_ext.lfs = malloc(sizeof(struct lfs_inode_ext));
    366 	memset(ip->inode_ext.lfs, 0, sizeof(struct lfs_inode_ext));
    367 	vp->v_data = ip;
    368 	/* ip->i_vnode = vp; */
    369 	ip->i_number = ino;
    370 	ip->i_lockf = 0;
    371 	ip->i_diroff = 0;
    372 	ip->i_lfs_effnblks = 0;
    373 	ip->i_flag = 0;
    374 
    375 	/* Load inode block and find inode */
    376 	if (daddr > 0) {
    377 		bread(fs->lfs_devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NULL, &bp);
    378 		bp->b_flags |= B_AGE;
    379 		dip = lfs_ifind(fs, ino, bp);
    380 		if (dip == NULL) {
    381 			brelse(bp);
    382 			free(ip);
    383 			free(vp);
    384 			return NULL;
    385 		}
    386 		memcpy(ip->i_din.ffs1_din, dip, sizeof(*dip));
    387 		brelse(bp);
    388 	}
    389 	ip->i_number = ino;
    390 	/* ip->i_devvp = fs->lfs_devvp; */
    391 	ip->i_lfs = fs;
    392 
    393 	ip->i_ffs_effnlink = ip->i_ffs1_nlink;
    394 	ip->i_lfs_effnblks = ip->i_ffs1_blocks;
    395 	ip->i_lfs_osize = ip->i_ffs1_size;
    396 #if 0
    397 	if (fs->lfs_version > 1) {
    398 		ip->i_ffs1_atime = ts.tv_sec;
    399 		ip->i_ffs1_atimensec = ts.tv_nsec;
    400 	}
    401 #endif
    402 
    403 	memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize));
    404 	for (i = 0; i < NDADDR; i++)
    405 		if (ip->i_ffs1_db[i] != 0)
    406 			ip->i_lfs_fragsize[i] = blksize(fs, ip, i);
    407 
    408 	++nvnodes;
    409 	LIST_INSERT_HEAD(&getvnodelist, vp, v_getvnodes);
    410 	LIST_INSERT_HEAD(&vnodelist, vp, v_mntvnodes);
    411 
    412 	return vp;
    413 }
    414 
    415 static struct uvnode *
    416 lfs_vget(void *vfs, ino_t ino)
    417 {
    418 	struct lfs *fs = (struct lfs *)vfs;
    419 	ufs_daddr_t daddr;
    420 	struct ubuf *bp;
    421 	IFILE *ifp;
    422 
    423 	LFS_IENTRY(ifp, fs, ino, bp);
    424 	daddr = ifp->if_daddr;
    425 	brelse(bp);
    426 	if (daddr == 0)
    427 		return NULL;
    428 	return lfs_raw_vget(fs, ino, fs->lfs_ivnode->v_fd, daddr);
    429 }
    430 
    431 /* Check superblock magic number and checksum */
    432 static int
    433 check_sb(struct lfs *fs)
    434 {
    435 	u_int32_t checksum;
    436 
    437 	if (fs->lfs_magic != LFS_MAGIC) {
    438 		printf("Superblock magic number (0x%lx) does not match "
    439 		       "expected 0x%lx\n", (unsigned long) fs->lfs_magic,
    440 		       (unsigned long) LFS_MAGIC);
    441 		return 1;
    442 	}
    443 	/* checksum */
    444 	checksum = lfs_sb_cksum(&(fs->lfs_dlfs));
    445 	if (fs->lfs_cksum != checksum) {
    446 		printf("Superblock checksum (%lx) does not match computed checksum (%lx)\n",
    447 		    (unsigned long) fs->lfs_cksum, (unsigned long) checksum);
    448 		return 1;
    449 	}
    450 	return 0;
    451 }
    452 
    453 /* Initialize LFS library; load superblocks and choose which to use. */
    454 struct lfs *
    455 lfs_init(int devfd, daddr_t sblkno, daddr_t idaddr, int dummy_read, int debug)
    456 {
    457 	struct uvnode *devvp;
    458 	struct ubuf *bp;
    459 	int tryalt;
    460 	struct lfs *fs, *altfs;
    461 	int error;
    462 
    463 	vfs_init();
    464 
    465 	devvp = (struct uvnode *) malloc(sizeof(*devvp));
    466 	memset(devvp, 0, sizeof(*devvp));
    467 	devvp->v_fs = NULL;
    468 	devvp->v_fd = devfd;
    469 	devvp->v_strategy_op = raw_vop_strategy;
    470 	devvp->v_bwrite_op = raw_vop_bwrite;
    471 	devvp->v_bmap_op = raw_vop_bmap;
    472 	LIST_INIT(&devvp->v_cleanblkhd);
    473 	LIST_INIT(&devvp->v_dirtyblkhd);
    474 
    475 	tryalt = 0;
    476 	if (dummy_read) {
    477 		if (sblkno == 0)
    478 			sblkno = btodb(LFS_LABELPAD);
    479 		fs = (struct lfs *) malloc(sizeof(*fs));
    480 		memset(fs, 0, sizeof(*fs));
    481 		fs->lfs_devvp = devvp;
    482 	} else {
    483 		if (sblkno == 0) {
    484 			sblkno = btodb(LFS_LABELPAD);
    485 			tryalt = 1;
    486 		} else if (debug) {
    487 			printf("No -b flag given, not attempting to verify checkpoint\n");
    488 		}
    489 		error = bread(devvp, sblkno, LFS_SBPAD, NOCRED, &bp);
    490 		fs = (struct lfs *) malloc(sizeof(*fs));
    491 		memset(fs, 0, sizeof(*fs));
    492 		fs->lfs_dlfs = *((struct dlfs *) bp->b_data);
    493 		fs->lfs_devvp = devvp;
    494 		bp->b_flags |= B_INVAL;
    495 		brelse(bp);
    496 
    497 		if (tryalt) {
    498 			error = bread(devvp, fsbtodb(fs, fs->lfs_sboffs[1]),
    499 		    	LFS_SBPAD, NOCRED, &bp);
    500 			altfs = (struct lfs *) malloc(sizeof(*altfs));
    501 			memset(altfs, 0, sizeof(*altfs));
    502 			altfs->lfs_dlfs = *((struct dlfs *) bp->b_data);
    503 			altfs->lfs_devvp = devvp;
    504 			bp->b_flags |= B_INVAL;
    505 			brelse(bp);
    506 
    507 			if (check_sb(fs) || fs->lfs_idaddr <= 0) {
    508 				if (debug)
    509 					printf("Primary superblock is no good, using first alternate\n");
    510 				free(fs);
    511 				fs = altfs;
    512 			} else {
    513 				/* If both superblocks check out, try verification */
    514 				if (check_sb(altfs)) {
    515 					if (debug)
    516 						printf("First alternate superblock is no good, using primary\n");
    517 					free(altfs);
    518 				} else {
    519 					if (lfs_verify(fs, altfs, devvp, debug) == fs) {
    520 						free(altfs);
    521 					} else {
    522 						free(fs);
    523 						fs = altfs;
    524 					}
    525 				}
    526 			}
    527 		}
    528 		if (check_sb(fs)) {
    529 			free(fs);
    530 			return NULL;
    531 		}
    532 	}
    533 
    534 	/* Compatibility */
    535 	if (fs->lfs_version < 2) {
    536 		fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE;
    537 		fs->lfs_ibsize = fs->lfs_bsize;
    538 		fs->lfs_start = fs->lfs_sboffs[0];
    539 		fs->lfs_tstamp = fs->lfs_otstamp;
    540 		fs->lfs_fsbtodb = 0;
    541 	}
    542 
    543 	if (!dummy_read) {
    544 		fs->lfs_suflags = (u_int32_t **) malloc(2 * sizeof(u_int32_t *));
    545 		fs->lfs_suflags[0] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
    546 		fs->lfs_suflags[1] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
    547 	}
    548 
    549 	if (idaddr == 0)
    550 		idaddr = fs->lfs_idaddr;
    551 	/* NB: If dummy_read!=0, idaddr==0 here so we get a fake inode. */
    552 	fs->lfs_ivnode = lfs_raw_vget(fs,
    553 		(dummy_read ? LFS_IFILE_INUM : fs->lfs_ifile), devvp->v_fd,
    554 		idaddr);
    555 
    556 	register_vget((void *)fs, lfs_vget);
    557 
    558 	return fs;
    559 }
    560 
    561 /*
    562  * Check partial segment validity between fs->lfs_offset and the given goal.
    563  * If goal == 0, just keep on going until the segments stop making sense.
    564  * Return the address of the first partial segment that failed.
    565  */
    566 ufs_daddr_t
    567 try_verify(struct lfs *osb, struct uvnode *devvp, ufs_daddr_t goal, int debug)
    568 {
    569 	ufs_daddr_t daddr, odaddr;
    570 	SEGSUM *sp;
    571 	int bc, flag;
    572 	struct ubuf *bp;
    573 	ufs_daddr_t nodirop_daddr;
    574 	u_int64_t serial;
    575 
    576 	daddr = osb->lfs_offset;
    577 	nodirop_daddr = daddr;
    578 	serial = osb->lfs_serial;
    579 	while (daddr != goal) {
    580 		flag = 0;
    581 oncemore:
    582 		/* Read in summary block */
    583 		bread(devvp, fsbtodb(osb, daddr), osb->lfs_sumsize, NULL, &bp);
    584 		sp = (SEGSUM *)bp->b_data;
    585 
    586 		/*
    587 		 * Could be a superblock instead of a segment summary.
    588 		 * XXX should use gseguse, but right now we need to do more
    589 		 * setup before we can...fix this
    590 		 */
    591 		if (sp->ss_magic != SS_MAGIC ||
    592 		    sp->ss_ident != osb->lfs_ident ||
    593 		    sp->ss_serial < serial ||
    594 		    sp->ss_sumsum != cksum(&sp->ss_datasum, osb->lfs_sumsize -
    595 			sizeof(sp->ss_sumsum))) {
    596 			brelse(bp);
    597 			if (flag == 0) {
    598 				flag = 1;
    599 				daddr += btofsb(osb, LFS_SBPAD);
    600 				goto oncemore;
    601 			}
    602 			break;
    603 		}
    604 		++serial;
    605 		bc = check_summary(osb, sp, daddr, debug, devvp, NULL);
    606 		if (bc == 0) {
    607 			brelse(bp);
    608 			break;
    609 		}
    610 		assert (bc > 0);
    611 		odaddr = daddr;
    612 		daddr += btofsb(osb, osb->lfs_sumsize + bc);
    613 		if (dtosn(osb, odaddr) != dtosn(osb, daddr) ||
    614 		    dtosn(osb, daddr) != dtosn(osb, daddr +
    615 			btofsb(osb, osb->lfs_sumsize + osb->lfs_bsize))) {
    616 			daddr = sp->ss_next;
    617 		}
    618 		if (!(sp->ss_flags & SS_CONT))
    619 			nodirop_daddr = daddr;
    620 		brelse(bp);
    621 	}
    622 
    623 	if (goal == 0)
    624 		return nodirop_daddr;
    625 	else
    626 		return daddr;
    627 }
    628 
    629 /* Use try_verify to check whether the newer superblock is valid. */
    630 struct lfs *
    631 lfs_verify(struct lfs *sb0, struct lfs *sb1, struct uvnode *devvp, int debug)
    632 {
    633 	ufs_daddr_t daddr;
    634 	struct lfs *osb, *nsb;
    635 
    636 	/*
    637 	 * Verify the checkpoint of the newer superblock,
    638 	 * if the timestamp/serial number of the two superblocks is
    639 	 * different.
    640 	 */
    641 
    642 	if (debug)
    643 		printf("sb0 %lld, sb1 %lld\n", (long long) sb0->lfs_serial,
    644 		    (long long) sb1->lfs_serial);
    645 
    646 	if ((sb0->lfs_version == 1 &&
    647 		sb0->lfs_otstamp != sb1->lfs_otstamp) ||
    648 	    (sb0->lfs_version > 1 &&
    649 		sb0->lfs_serial != sb1->lfs_serial)) {
    650 		if (sb0->lfs_version == 1) {
    651 			if (sb0->lfs_otstamp > sb1->lfs_otstamp) {
    652 				osb = sb1;
    653 				nsb = sb0;
    654 			} else {
    655 				osb = sb0;
    656 				nsb = sb1;
    657 			}
    658 		} else {
    659 			if (sb0->lfs_serial > sb1->lfs_serial) {
    660 				osb = sb1;
    661 				nsb = sb0;
    662 			} else {
    663 				osb = sb0;
    664 				nsb = sb1;
    665 			}
    666 		}
    667 		if (debug) {
    668 			printf("Attempting to verify newer checkpoint...");
    669 			fflush(stdout);
    670 		}
    671 		daddr = try_verify(osb, devvp, nsb->lfs_offset, debug);
    672 
    673 		if (debug)
    674 			printf("done.\n");
    675 		if (daddr == nsb->lfs_offset) {
    676 			pwarn("** Newer checkpoint verified, recovered %lld seconds of data\n",
    677 			    (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
    678 			sbdirty();
    679 		} else {
    680 			pwarn("** Newer checkpoint invalid, lost %lld seconds of data\n", (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
    681 		}
    682 		return (daddr == nsb->lfs_offset ? nsb : osb);
    683 	}
    684 	/* Nothing to check */
    685 	return osb;
    686 }
    687 
    688 /* Verify a partial-segment summary; return the number of bytes on disk. */
    689 int
    690 check_summary(struct lfs *fs, SEGSUM *sp, ufs_daddr_t pseg_addr, int debug,
    691 	      struct uvnode *devvp, void (func(ufs_daddr_t, FINFO *)))
    692 {
    693 	FINFO *fp;
    694 	int bc;			/* Bytes in partial segment */
    695 	int nblocks;
    696 	ufs_daddr_t seg_addr, daddr;
    697 	ufs_daddr_t *dp, *idp;
    698 	struct ubuf *bp;
    699 	int i, j, k, datac, len;
    700 	long sn;
    701 	u_int32_t *datap;
    702 	u_int32_t ccksum;
    703 
    704 	sn = dtosn(fs, pseg_addr);
    705 	seg_addr = sntod(fs, sn);
    706 
    707 	/* We've already checked the sumsum, just do the data bounds and sum */
    708 
    709 	/* Count the blocks. */
    710 	nblocks = howmany(sp->ss_ninos, INOPB(fs));
    711 	bc = nblocks << (fs->lfs_version > 1 ? fs->lfs_ffshift : fs->lfs_bshift);
    712 	assert(bc >= 0);
    713 
    714 	fp = (FINFO *) (sp + 1);
    715 	for (i = 0; i < sp->ss_nfinfo; i++) {
    716 		nblocks += fp->fi_nblocks;
    717 		bc += fp->fi_lastlength + ((fp->fi_nblocks - 1)
    718 					   << fs->lfs_bshift);
    719 		assert(bc >= 0);
    720 		fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
    721 	}
    722 	datap = (u_int32_t *) malloc(nblocks * sizeof(*datap));
    723 	datac = 0;
    724 
    725 	dp = (ufs_daddr_t *) sp;
    726 	dp += fs->lfs_sumsize / sizeof(ufs_daddr_t);
    727 	dp--;
    728 
    729 	idp = dp;
    730 	daddr = pseg_addr + btofsb(fs, fs->lfs_sumsize);
    731 	fp = (FINFO *) (sp + 1);
    732 	for (i = 0, j = 0;
    733 	     i < sp->ss_nfinfo || j < howmany(sp->ss_ninos, INOPB(fs)); i++) {
    734 		if (i >= sp->ss_nfinfo && *idp != daddr) {
    735 			pwarn("Not enough inode blocks in pseg at 0x%" PRIx32
    736 			      ": found %d, wanted %d\n",
    737 			      pseg_addr, j, howmany(sp->ss_ninos, INOPB(fs)));
    738 			if (debug)
    739 				pwarn("*idp=%x, daddr=%" PRIx32 "\n", *idp,
    740 				      daddr);
    741 			break;
    742 		}
    743 		while (j < howmany(sp->ss_ninos, INOPB(fs)) && *idp == daddr) {
    744 			bread(devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NOCRED, &bp);
    745 			datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
    746 			brelse(bp);
    747 
    748 			++j;
    749 			daddr += btofsb(fs, fs->lfs_ibsize);
    750 			--idp;
    751 		}
    752 		if (i < sp->ss_nfinfo) {
    753 			if (func)
    754 				func(daddr, fp);
    755 			for (k = 0; k < fp->fi_nblocks; k++) {
    756 				len = (k == fp->fi_nblocks - 1 ?
    757 				       fp->fi_lastlength
    758 				       : fs->lfs_bsize);
    759 				bread(devvp, fsbtodb(fs, daddr), len, NOCRED, &bp);
    760 				datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
    761 				brelse(bp);
    762 				daddr += btofsb(fs, len);
    763 			}
    764 			fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
    765 		}
    766 	}
    767 
    768 	if (datac != nblocks) {
    769 		pwarn("Partial segment at 0x%llx expected %d blocks counted %d\n",
    770 		    (long long) pseg_addr, nblocks, datac);
    771 	}
    772 	ccksum = cksum(datap, nblocks * sizeof(u_int32_t));
    773 	/* Check the data checksum */
    774 	if (ccksum != sp->ss_datasum) {
    775 		pwarn("Partial segment at 0x%" PRIx32 " data checksum"
    776 		      " mismatch: given 0x%x, computed 0x%x\n",
    777 		      pseg_addr, sp->ss_datasum, ccksum);
    778 		free(datap);
    779 		return 0;
    780 	}
    781 	free(datap);
    782 	assert(bc >= 0);
    783 	return bc;
    784 }
    785 
    786 /* print message and exit */
    787 void
    788 my_vpanic(int fatal, const char *fmt, va_list ap)
    789 {
    790         (void) vprintf(fmt, ap);
    791 	exit(8);
    792 }
    793 
    794 void
    795 call_panic(const char *fmt, ...)
    796 {
    797 	va_list ap;
    798 
    799 	va_start(ap, fmt);
    800         panic_func(1, fmt, ap);
    801 	va_end(ap);
    802 }
    803