xref: /netbsd-src/sbin/fsck_lfs/lfs.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /* $NetBSD: lfs.c,v 1.7 2003/08/07 10:04:23 agc 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@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 
105 extern struct uvnodelst vnodelist;
106 extern struct uvnodelst getvnodelist;
107 extern int nvnodes;
108 
109 int fsdirty = 0;
110 void (*panic_func)(int, const char *, va_list) = my_vpanic;
111 
112 /*
113  * LFS buffer and uvnode operations
114  */
115 
116 int
117 lfs_vop_strategy(struct ubuf * bp)
118 {
119 	int count;
120 
121 	if (bp->b_flags & B_READ) {
122 		count = pread(bp->b_vp->v_fd, bp->b_data, bp->b_bcount,
123 		    dbtob(bp->b_blkno));
124 		if (count == bp->b_bcount)
125 			bp->b_flags |= B_DONE;
126 	} else {
127 		count = pwrite(bp->b_vp->v_fd, bp->b_data, bp->b_bcount,
128 		    dbtob(bp->b_blkno));
129 		if (count == 0) {
130 			perror("pwrite");
131 			return -1;
132 		}
133 		bp->b_flags &= ~B_DELWRI;
134 		reassignbuf(bp, bp->b_vp);
135 	}
136 	return 0;
137 }
138 
139 int
140 lfs_vop_bwrite(struct ubuf * bp)
141 {
142 	struct lfs *fs;
143 
144 	fs = bp->b_vp->v_fs;
145 	if (!(bp->b_flags & B_DELWRI)) {
146 		fs->lfs_avail -= btofsb(fs, bp->b_bcount);
147 	}
148 	bp->b_flags |= B_DELWRI | B_LOCKED;
149 	reassignbuf(bp, bp->b_vp);
150 	brelse(bp);
151 	return 0;
152 }
153 
154 /*
155  * ufs_bmaparray does the bmap conversion, and if requested returns the
156  * array of logical blocks which must be traversed to get to a block.
157  * Each entry contains the offset into that block that gets you to the
158  * next block and the disk address of the block (if it is assigned).
159  */
160 int
161 ufs_bmaparray(struct lfs * fs, struct uvnode * vp, daddr_t bn, daddr_t * bnp, struct indir * ap, int *nump)
162 {
163 	struct inode *ip;
164 	struct ubuf *bp;
165 	struct indir a[NIADDR + 1], *xap;
166 	daddr_t daddr;
167 	daddr_t metalbn;
168 	int error, num;
169 
170 	ip = VTOI(vp);
171 
172 	if (bn >= 0 && bn < NDADDR) {
173 		if (nump != NULL)
174 			*nump = 0;
175 		*bnp = fsbtodb(fs, ip->i_ffs1_db[bn]);
176 		if (*bnp == 0)
177 			*bnp = -1;
178 		return (0);
179 	}
180 	xap = ap == NULL ? a : ap;
181 	if (!nump)
182 		nump = &num;
183 	if ((error = ufs_getlbns(fs, vp, bn, xap, nump)) != 0)
184 		return (error);
185 
186 	num = *nump;
187 
188 	/* Get disk address out of indirect block array */
189 	daddr = ip->i_ffs1_ib[xap->in_off];
190 
191 	for (bp = NULL, ++xap; --num; ++xap) {
192 		/* Exit the loop if there is no disk address assigned yet and
193 		 * the indirect block isn't in the cache, or if we were
194 		 * looking for an indirect block and we've found it. */
195 
196 		metalbn = xap->in_lbn;
197 		if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn)
198 			break;
199 		/*
200 		 * If we get here, we've either got the block in the cache
201 		 * or we have a disk address for it, go fetch it.
202 		 */
203 		if (bp)
204 			brelse(bp);
205 
206 		xap->in_exists = 1;
207 		bp = getblk(vp, metalbn, fs->lfs_bsize);
208 
209 		if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
210 			bp->b_blkno = fsbtodb(fs, daddr);
211 			bp->b_flags |= B_READ;
212 			VOP_STRATEGY(bp);
213 		}
214 		daddr = ((ufs_daddr_t *) bp->b_data)[xap->in_off];
215 	}
216 	if (bp)
217 		brelse(bp);
218 
219 	daddr = fsbtodb(fs, (ufs_daddr_t) daddr);
220 	*bnp = daddr == 0 ? -1 : daddr;
221 	return (0);
222 }
223 
224 /*
225  * Create an array of logical block number/offset pairs which represent the
226  * path of indirect blocks required to access a data block.  The first "pair"
227  * contains the logical block number of the appropriate single, double or
228  * triple indirect block and the offset into the inode indirect block array.
229  * Note, the logical block number of the inode single/double/triple indirect
230  * block appears twice in the array, once with the offset into the i_ffs1_ib and
231  * once with the offset into the page itself.
232  */
233 int
234 ufs_getlbns(struct lfs * fs, struct uvnode * vp, daddr_t bn, struct indir * ap, int *nump)
235 {
236 	daddr_t metalbn, realbn;
237 	int64_t blockcnt;
238 	int lbc;
239 	int i, numlevels, off;
240 	int lognindir, indir;
241 
242 	if (nump)
243 		*nump = 0;
244 	numlevels = 0;
245 	realbn = bn;
246 	if (bn < 0)
247 		bn = -bn;
248 
249 	lognindir = -1;
250 	for (indir = fs->lfs_nindir; indir; indir >>= 1)
251 		++lognindir;
252 
253 	/* Determine the number of levels of indirection.  After this loop is
254 	 * done, blockcnt indicates the number of data blocks possible at the
255 	 * given level of indirection, and NIADDR - i is the number of levels
256 	 * of indirection needed to locate the requested block. */
257 
258 	bn -= NDADDR;
259 	for (lbc = 0, i = NIADDR;; i--, bn -= blockcnt) {
260 		if (i == 0)
261 			return (EFBIG);
262 
263 		lbc += lognindir;
264 		blockcnt = (int64_t) 1 << lbc;
265 
266 		if (bn < blockcnt)
267 			break;
268 	}
269 
270 	/* Calculate the address of the first meta-block. */
271 	if (realbn >= 0)
272 		metalbn = -(realbn - bn + NIADDR - i);
273 	else
274 		metalbn = -(-realbn - bn + NIADDR - i);
275 
276 	/* At each iteration, off is the offset into the bap array which is an
277 	 * array of disk addresses at the current level of indirection. The
278 	 * logical block number and the offset in that block are stored into
279 	 * the argument array. */
280 	ap->in_lbn = metalbn;
281 	ap->in_off = off = NIADDR - i;
282 	ap->in_exists = 0;
283 	ap++;
284 	for (++numlevels; i <= NIADDR; i++) {
285 		/* If searching for a meta-data block, quit when found. */
286 		if (metalbn == realbn)
287 			break;
288 
289 		lbc -= lognindir;
290 		blockcnt = (int64_t) 1 << lbc;
291 		off = (bn >> lbc) & (fs->lfs_nindir - 1);
292 
293 		++numlevels;
294 		ap->in_lbn = metalbn;
295 		ap->in_off = off;
296 		ap->in_exists = 0;
297 		++ap;
298 
299 		metalbn -= -1 + (off << lbc);
300 	}
301 	if (nump)
302 		*nump = numlevels;
303 	return (0);
304 }
305 
306 int
307 lfs_vop_bmap(struct uvnode * vp, daddr_t lbn, daddr_t * daddrp)
308 {
309 	return ufs_bmaparray(vp->v_fs, vp, lbn, daddrp, NULL, NULL);
310 }
311 
312 /* Search a block for a specific dinode. */
313 struct ufs1_dinode *
314 lfs_ifind(struct lfs * fs, ino_t ino, struct ubuf * bp)
315 {
316 	struct ufs1_dinode *dip = (struct ufs1_dinode *) bp->b_data;
317 	struct ufs1_dinode *ldip, *fin;
318 
319 	fin = dip + INOPB(fs);
320 
321 	/*
322 	 * Read the inode block backwards, since later versions of the
323 	 * inode will supercede earlier ones.  Though it is unlikely, it is
324 	 * possible that the same inode will appear in the same inode block.
325 	 */
326 	for (ldip = fin - 1; ldip >= dip; --ldip)
327 		if (ldip->di_inumber == ino)
328 			return (ldip);
329 	return NULL;
330 }
331 
332 /*
333  * lfs_raw_vget makes us a new vnode from the inode at the given disk address.
334  * XXX it currently loses atime information.
335  */
336 struct uvnode *
337 lfs_raw_vget(struct lfs * fs, ino_t ino, int fd, ufs_daddr_t daddr)
338 {
339 	struct uvnode *vp;
340 	struct inode *ip;
341 	struct ufs1_dinode *dip;
342 	struct ubuf *bp;
343 	int i;
344 
345 	vp = (struct uvnode *) malloc(sizeof(*vp));
346 	memset(vp, 0, sizeof(*vp));
347 	vp->v_fd = fd;
348 	vp->v_fs = fs;
349 	vp->v_usecount = 0;
350 	vp->v_strategy_op = lfs_vop_strategy;
351 	vp->v_bwrite_op = lfs_vop_bwrite;
352 	vp->v_bmap_op = lfs_vop_bmap;
353 	LIST_INIT(&vp->v_cleanblkhd);
354 	LIST_INIT(&vp->v_dirtyblkhd);
355 
356 	ip = (struct inode *) malloc(sizeof(*ip));
357 	memset(ip, 0, sizeof(*ip));
358 
359 	ip->i_din.ffs1_din = (struct ufs1_dinode *)
360 	    malloc(sizeof(struct ufs1_dinode));
361 	memset(ip->i_din.ffs1_din, 0, sizeof (struct ufs1_dinode));
362 
363 	/* Initialize the inode -- from lfs_vcreate. */
364 	ip->inode_ext.lfs = malloc(sizeof(struct lfs_inode_ext));
365 	memset(ip->inode_ext.lfs, 0, sizeof(struct lfs_inode_ext));
366 	vp->v_data = ip;
367 	/* ip->i_vnode = vp; */
368 	ip->i_number = ino;
369 	ip->i_lockf = 0;
370 	ip->i_diroff = 0;
371 	ip->i_lfs_effnblks = 0;
372 	ip->i_flag = 0;
373 
374 	/* Load inode block and find inode */
375 	bread(fs->lfs_unlockvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NULL, &bp);
376 	bp->b_flags |= B_AGE;
377 	dip = lfs_ifind(fs, ino, bp);
378 	if (dip == NULL) {
379 		brelse(bp);
380 		free(ip);
381 		free(vp);
382 		return NULL;
383 	}
384 	memcpy(ip->i_din.ffs1_din, dip, sizeof(*dip));
385 	brelse(bp);
386 	ip->i_number = ino;
387 	/* ip->i_devvp = fs->lfs_unlockvp; */
388 	ip->i_lfs = fs;
389 
390 	ip->i_ffs_effnlink = ip->i_ffs1_nlink;
391 	ip->i_lfs_effnblks = ip->i_ffs1_blocks;
392 	ip->i_lfs_osize = ip->i_ffs1_size;
393 #if 0
394 	if (fs->lfs_version > 1) {
395 		ip->i_ffs1_atime = ts.tv_sec;
396 		ip->i_ffs1_atimensec = ts.tv_nsec;
397 	}
398 #endif
399 
400 	memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize));
401 	for (i = 0; i < NDADDR; i++)
402 		if (ip->i_ffs1_db[i] != 0)
403 			ip->i_lfs_fragsize[i] = blksize(fs, ip, i);
404 
405 	++nvnodes;
406 	LIST_INSERT_HEAD(&getvnodelist, vp, v_getvnodes);
407 	LIST_INSERT_HEAD(&vnodelist, vp, v_mntvnodes);
408 
409 	return vp;
410 }
411 
412 static struct uvnode *
413 lfs_vget(void *vfs, ino_t ino)
414 {
415 	struct lfs *fs = (struct lfs *)vfs;
416 	ufs_daddr_t daddr;
417 	struct ubuf *bp;
418 	IFILE *ifp;
419 
420 	LFS_IENTRY(ifp, fs, ino, bp);
421 	daddr = ifp->if_daddr;
422 	brelse(bp);
423 	if (daddr == 0)
424 		return NULL;
425 	return lfs_raw_vget(fs, ino, fs->lfs_ivnode->v_fd, daddr);
426 }
427 
428 /* Check superblock magic number and checksum */
429 static int
430 check_sb(struct lfs *fs)
431 {
432 	u_int32_t checksum;
433 
434 	if (fs->lfs_magic != LFS_MAGIC) {
435 		printf("Superblock magic number (0x%lx) does not match "
436 		       "expected 0x%lx\n", (unsigned long) fs->lfs_magic,
437 		       (unsigned long) LFS_MAGIC);
438 		return 1;
439 	}
440 	/* checksum */
441 	checksum = lfs_sb_cksum(&(fs->lfs_dlfs));
442 	if (fs->lfs_cksum != checksum) {
443 		printf("Superblock checksum (%lx) does not match computed checksum (%lx)\n",
444 		    (unsigned long) fs->lfs_cksum, (unsigned long) checksum);
445 		return 1;
446 	}
447 	return 0;
448 }
449 
450 /* Initialize LFS library; load superblocks and choose which to use. */
451 struct lfs *
452 lfs_init(int devfd, daddr_t sblkno, daddr_t idaddr, int debug)
453 {
454 	struct uvnode *devvp;
455 	struct ubuf *bp;
456 	int tryalt;
457 	struct lfs *fs, *altfs;
458 	int error;
459 
460 	vfs_init();
461 
462 	devvp = (struct uvnode *) malloc(sizeof(*devvp));
463 	memset(devvp, 0, sizeof(*devvp));
464 	devvp->v_fs = NULL;
465 	devvp->v_fd = devfd;
466 	devvp->v_strategy_op = raw_vop_strategy;
467 	devvp->v_bwrite_op = raw_vop_bwrite;
468 	devvp->v_bmap_op = raw_vop_bmap;
469 	LIST_INIT(&devvp->v_cleanblkhd);
470 	LIST_INIT(&devvp->v_dirtyblkhd);
471 
472 	tryalt = 0;
473 	if (sblkno == 0) {
474 		sblkno = btodb(LFS_LABELPAD);
475 		tryalt = 1;
476 	} else if (debug) {
477 		printf("No -b flag given, not attempting to verify checkpoint\n");
478 	}
479 	error = bread(devvp, sblkno, LFS_SBPAD, NOCRED, &bp);
480 	fs = (struct lfs *) malloc(sizeof(*fs));
481 	memset(fs, 0, sizeof(*fs));
482 	fs->lfs_dlfs = *((struct dlfs *) bp->b_data);
483 	fs->lfs_unlockvp = devvp;
484 	bp->b_flags |= B_INVAL;
485 	brelse(bp);
486 
487 	if (tryalt) {
488 		error = bread(devvp, fsbtodb(fs, fs->lfs_sboffs[1]),
489 		    LFS_SBPAD, NOCRED, &bp);
490 		altfs = (struct lfs *) malloc(sizeof(*altfs));
491 		memset(altfs, 0, sizeof(*altfs));
492 		altfs->lfs_dlfs = *((struct dlfs *) bp->b_data);
493 		altfs->lfs_unlockvp = devvp;
494 		bp->b_flags |= B_INVAL;
495 		brelse(bp);
496 
497 		if (check_sb(fs)) {
498 			if (debug)
499 				printf("Primary superblock is no good, using first alternate\n");
500 			free(fs);
501 			fs = altfs;
502 		} else {
503 			/* If both superblocks check out, try verification */
504 			if (check_sb(altfs)) {
505 				if (debug)
506 					printf("First alternate superblock is no good, using primary\n");
507 				free(altfs);
508 			} else {
509 				if (lfs_verify(fs, altfs, devvp, debug) == fs) {
510 					free(altfs);
511 				} else {
512 					free(fs);
513 					fs = altfs;
514 				}
515 			}
516 		}
517 	}
518 	if (check_sb(fs)) {
519 		free(fs);
520 		return NULL;
521 	}
522 	/* Compatibility */
523 	if (fs->lfs_version < 2) {
524 		fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE;
525 		fs->lfs_ibsize = fs->lfs_bsize;
526 		fs->lfs_start = fs->lfs_sboffs[0];
527 		fs->lfs_tstamp = fs->lfs_otstamp;
528 		fs->lfs_fsbtodb = 0;
529 	}
530 	fs->lfs_suflags = (u_int32_t **) malloc(2 * sizeof(u_int32_t *));
531 	fs->lfs_suflags[0] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
532 	fs->lfs_suflags[1] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
533 
534 	if (idaddr == 0)
535 		idaddr = fs->lfs_idaddr;
536 	fs->lfs_ivnode = lfs_raw_vget(fs, fs->lfs_ifile, devvp->v_fd, idaddr);
537 
538 	register_vget((void *)fs, lfs_vget);
539 
540 	return fs;
541 }
542 
543 /*
544  * Check partial segment validity between fs->lfs_offset and the given goal.
545  * If goal == 0, just keep on going until the segments stop making sense.
546  * Return the address of the first partial segment that failed.
547  */
548 ufs_daddr_t
549 try_verify(struct lfs *osb, struct uvnode *devvp, ufs_daddr_t goal, int debug)
550 {
551 	ufs_daddr_t daddr, odaddr;
552 	SEGSUM *sp;
553 	int bc, flag;
554 	struct ubuf *bp;
555 	ufs_daddr_t nodirop_daddr;
556 	u_int64_t serial;
557 
558 	daddr = osb->lfs_offset;
559 	nodirop_daddr = daddr;
560 	serial = osb->lfs_serial;
561 	while (daddr != goal) {
562 		flag = 0;
563 oncemore:
564 		/* Read in summary block */
565 		bread(devvp, fsbtodb(osb, daddr), osb->lfs_sumsize, NULL, &bp);
566 		sp = (SEGSUM *)bp->b_data;
567 
568 		/*
569 		 * Could be a superblock instead of a segment summary.
570 		 * XXX should use gseguse, but right now we need to do more
571 		 * setup before we can...fix this
572 		 */
573 		if (sp->ss_magic != SS_MAGIC ||
574 		    sp->ss_ident != osb->lfs_ident ||
575 		    sp->ss_serial < serial ||
576 		    sp->ss_sumsum != cksum(&sp->ss_datasum, osb->lfs_sumsize -
577 			sizeof(sp->ss_sumsum))) {
578 			brelse(bp);
579 			if (flag == 0) {
580 				flag = 1;
581 				daddr += btofsb(osb, LFS_SBPAD);
582 				goto oncemore;
583 			}
584 			break;
585 		}
586 		++serial;
587 		bc = check_summary(osb, sp, daddr, debug, devvp, NULL);
588 		if (bc == 0) {
589 			brelse(bp);
590 			break;
591 		}
592 		assert (bc > 0);
593 		odaddr = daddr;
594 		daddr += btofsb(osb, osb->lfs_sumsize + bc);
595 		if (dtosn(osb, odaddr) != dtosn(osb, daddr) ||
596 		    dtosn(osb, daddr) != dtosn(osb, daddr +
597 			btofsb(osb, osb->lfs_sumsize + osb->lfs_bsize))) {
598 			daddr = sp->ss_next;
599 		}
600 		if (!(sp->ss_flags & SS_CONT))
601 			nodirop_daddr = daddr;
602 		brelse(bp);
603 	}
604 
605 	if (goal == 0)
606 		return nodirop_daddr;
607 	else
608 		return daddr;
609 }
610 
611 /* Use try_verify to check whether the newer superblock is valid. */
612 struct lfs *
613 lfs_verify(struct lfs *sb0, struct lfs *sb1, struct uvnode *devvp, int debug)
614 {
615 	ufs_daddr_t daddr;
616 	struct lfs *osb, *nsb;
617 
618 	/*
619 	 * Verify the checkpoint of the newer superblock,
620 	 * if the timestamp/serial number of the two superblocks is
621 	 * different.
622 	 */
623 
624 	if (debug)
625 		printf("sb0 %lld, sb1 %lld\n", (long long) sb0->lfs_serial,
626 		    (long long) sb1->lfs_serial);
627 
628 	if ((sb0->lfs_version == 1 &&
629 		sb0->lfs_otstamp != sb1->lfs_otstamp) ||
630 	    (sb0->lfs_version > 1 &&
631 		sb0->lfs_serial != sb1->lfs_serial)) {
632 		if (sb0->lfs_version == 1) {
633 			if (sb0->lfs_otstamp > sb1->lfs_otstamp) {
634 				osb = sb1;
635 				nsb = sb0;
636 			} else {
637 				osb = sb0;
638 				nsb = sb1;
639 			}
640 		} else {
641 			if (sb0->lfs_serial > sb1->lfs_serial) {
642 				osb = sb1;
643 				nsb = sb0;
644 			} else {
645 				osb = sb0;
646 				nsb = sb1;
647 			}
648 		}
649 		if (debug) {
650 			printf("Attempting to verify newer checkpoint...");
651 			fflush(stdout);
652 		}
653 		daddr = try_verify(osb, devvp, nsb->lfs_offset, debug);
654 
655 		if (debug)
656 			printf("done.\n");
657 		if (daddr == nsb->lfs_offset) {
658 			warnx("** Newer checkpoint verified, recovered %lld seconds of data\n",
659 			    (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
660 			sbdirty();
661 		} else {
662 			warnx("** Newer checkpoint invalid, lost %lld seconds of data\n", (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
663 		}
664 		return (daddr == nsb->lfs_offset ? nsb : osb);
665 	}
666 	/* Nothing to check */
667 	return osb;
668 }
669 
670 /* Verify a partial-segment summary; return the number of bytes on disk. */
671 int
672 check_summary(struct lfs *fs, SEGSUM *sp, ufs_daddr_t pseg_addr, int debug,
673 	      struct uvnode *devvp, void (func(ufs_daddr_t, FINFO *)))
674 {
675 	FINFO *fp;
676 	int bc;			/* Bytes in partial segment */
677 	int nblocks;
678 	ufs_daddr_t seg_addr, daddr;
679 	ufs_daddr_t *dp, *idp;
680 	struct ubuf *bp;
681 	int i, j, k, datac, len;
682 	long sn;
683 	u_int32_t *datap;
684 	u_int32_t ccksum;
685 
686 	sn = dtosn(fs, pseg_addr);
687 	seg_addr = sntod(fs, sn);
688 
689 	/* We've already checked the sumsum, just do the data bounds and sum */
690 
691 	/* Count the blocks. */
692 	nblocks = howmany(sp->ss_ninos, INOPB(fs));
693 	bc = nblocks << (fs->lfs_version > 1 ? fs->lfs_ffshift : fs->lfs_bshift);
694 	assert(bc >= 0);
695 
696 	fp = (FINFO *) (sp + 1);
697 	for (i = 0; i < sp->ss_nfinfo; i++) {
698 		nblocks += fp->fi_nblocks;
699 		bc += fp->fi_lastlength + ((fp->fi_nblocks - 1)
700 					   << fs->lfs_bshift);
701 		assert(bc >= 0);
702 		fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
703 	}
704 	datap = (u_int32_t *) malloc(nblocks * sizeof(*datap));
705 	datac = 0;
706 
707 	dp = (ufs_daddr_t *) sp;
708 	dp += fs->lfs_sumsize / sizeof(ufs_daddr_t);
709 	dp--;
710 
711 	idp = dp;
712 	daddr = pseg_addr + btofsb(fs, fs->lfs_sumsize);
713 	fp = (FINFO *) (sp + 1);
714 	for (i = 0, j = 0;
715 	     i < sp->ss_nfinfo || j < howmany(sp->ss_ninos, INOPB(fs)); i++) {
716 		if (i >= sp->ss_nfinfo && *idp != daddr) {
717 			warnx("Not enough inode blocks in pseg at 0x%" PRIx32
718 			      ": found %d, wanted %d\n",
719 			      pseg_addr, j, howmany(sp->ss_ninos, INOPB(fs)));
720 			if (debug)
721 				warnx("*idp=%x, daddr=%" PRIx32 "\n", *idp,
722 				      daddr);
723 			break;
724 		}
725 		while (j < howmany(sp->ss_ninos, INOPB(fs)) && *idp == daddr) {
726 			bread(devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NOCRED, &bp);
727 			datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
728 			brelse(bp);
729 
730 			++j;
731 			daddr += btofsb(fs, fs->lfs_ibsize);
732 			--idp;
733 		}
734 		if (i < sp->ss_nfinfo) {
735 			if (func)
736 				func(daddr, fp);
737 			for (k = 0; k < fp->fi_nblocks; k++) {
738 				len = (k == fp->fi_nblocks - 1 ?
739 				       fp->fi_lastlength
740 				       : fs->lfs_bsize);
741 				bread(devvp, fsbtodb(fs, daddr), len, NOCRED, &bp);
742 				datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
743 				brelse(bp);
744 				daddr += btofsb(fs, len);
745 			}
746 			fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
747 		}
748 	}
749 
750 	if (datac != nblocks) {
751 		warnx("Partial segment at 0x%llx expected %d blocks counted %d\n",
752 		    (long long) pseg_addr, nblocks, datac);
753 	}
754 	ccksum = cksum(datap, nblocks * sizeof(u_int32_t));
755 	/* Check the data checksum */
756 	if (ccksum != sp->ss_datasum) {
757 		warnx("Partial segment at 0x%" PRIx32 " data checksum"
758 		      " mismatch: given 0x%x, computed 0x%x\n",
759 		      pseg_addr, sp->ss_datasum, ccksum);
760 		free(datap);
761 		return 0;
762 	}
763 	free(datap);
764 	assert(bc >= 0);
765 	return bc;
766 }
767 
768 /* print message and exit */
769 void
770 my_vpanic(int fatal, const char *fmt, va_list ap)
771 {
772         (void) vprintf(fmt, ap);
773 	exit(8);
774 }
775 
776 void
777 call_panic(const char *fmt, ...)
778 {
779 	va_list ap;
780 
781 	va_start(ap, fmt);
782         panic_func(1, fmt, ap);
783 	va_end(ap);
784 }
785