lfs.c revision 1.14
1/* $NetBSD: lfs.c,v 1.14 2005/06/02 01:02:21 lukem 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
102extern u_int32_t cksum(void *, size_t);
103extern u_int32_t lfs_sb_cksum(struct dlfs *);
104extern void pwarn(const char *, ...);
105
106extern struct uvnodelst vnodelist;
107extern struct uvnodelst getvnodelist[VNODE_HASH_MAX];
108extern int nvnodes;
109
110int fsdirty = 0;
111void (*panic_func)(int, const char *, va_list) = my_vpanic;
112
113/*
114 * LFS buffer and uvnode operations
115 */
116
117int
118lfs_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
140int
141lfs_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 */
161int
162ufs_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 */
234int
235ufs_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
307int
308lfs_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. */
314struct ufs1_dinode *
315lfs_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 */
337struct uvnode *
338lfs_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, hash;
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	hash = ((int)(intptr_t)fs + ino) & (VNODE_HASH_MAX - 1);
410	LIST_INSERT_HEAD(&getvnodelist[hash], vp, v_getvnodes);
411	LIST_INSERT_HEAD(&vnodelist, vp, v_mntvnodes);
412
413	return vp;
414}
415
416static struct uvnode *
417lfs_vget(void *vfs, ino_t ino)
418{
419	struct lfs *fs = (struct lfs *)vfs;
420	ufs_daddr_t daddr;
421	struct ubuf *bp;
422	IFILE *ifp;
423
424	LFS_IENTRY(ifp, fs, ino, bp);
425	daddr = ifp->if_daddr;
426	brelse(bp);
427	if (daddr <= 0 || dtosn(fs, daddr) >= fs->lfs_nseg)
428		return NULL;
429	return lfs_raw_vget(fs, ino, fs->lfs_ivnode->v_fd, daddr);
430}
431
432/* Check superblock magic number and checksum */
433static int
434check_sb(struct lfs *fs)
435{
436	u_int32_t checksum;
437
438	if (fs->lfs_magic != LFS_MAGIC) {
439		printf("Superblock magic number (0x%lx) does not match "
440		       "expected 0x%lx\n", (unsigned long) fs->lfs_magic,
441		       (unsigned long) LFS_MAGIC);
442		return 1;
443	}
444	/* checksum */
445	checksum = lfs_sb_cksum(&(fs->lfs_dlfs));
446	if (fs->lfs_cksum != checksum) {
447		printf("Superblock checksum (%lx) does not match computed checksum (%lx)\n",
448		    (unsigned long) fs->lfs_cksum, (unsigned long) checksum);
449		return 1;
450	}
451	return 0;
452}
453
454/* Initialize LFS library; load superblocks and choose which to use. */
455struct lfs *
456lfs_init(int devfd, daddr_t sblkno, daddr_t idaddr, int dummy_read, int debug)
457{
458	struct uvnode *devvp;
459	struct ubuf *bp;
460	int tryalt;
461	struct lfs *fs, *altfs;
462	int error;
463
464	vfs_init();
465
466	devvp = (struct uvnode *) malloc(sizeof(*devvp));
467	memset(devvp, 0, sizeof(*devvp));
468	devvp->v_fs = NULL;
469	devvp->v_fd = devfd;
470	devvp->v_strategy_op = raw_vop_strategy;
471	devvp->v_bwrite_op = raw_vop_bwrite;
472	devvp->v_bmap_op = raw_vop_bmap;
473	LIST_INIT(&devvp->v_cleanblkhd);
474	LIST_INIT(&devvp->v_dirtyblkhd);
475
476	tryalt = 0;
477	if (dummy_read) {
478		if (sblkno == 0)
479			sblkno = btodb(LFS_LABELPAD);
480		fs = (struct lfs *) malloc(sizeof(*fs));
481		memset(fs, 0, sizeof(*fs));
482		fs->lfs_devvp = devvp;
483	} else {
484		if (sblkno == 0) {
485			sblkno = btodb(LFS_LABELPAD);
486			tryalt = 1;
487		} else if (debug) {
488			printf("No -b flag given, not attempting to verify checkpoint\n");
489		}
490		error = bread(devvp, sblkno, LFS_SBPAD, NOCRED, &bp);
491		fs = (struct lfs *) malloc(sizeof(*fs));
492		memset(fs, 0, sizeof(*fs));
493		fs->lfs_dlfs = *((struct dlfs *) bp->b_data);
494		fs->lfs_devvp = devvp;
495		bp->b_flags |= B_INVAL;
496		brelse(bp);
497
498		if (tryalt) {
499			error = bread(devvp, fsbtodb(fs, fs->lfs_sboffs[1]),
500		    	LFS_SBPAD, NOCRED, &bp);
501			altfs = (struct lfs *) malloc(sizeof(*altfs));
502			memset(altfs, 0, sizeof(*altfs));
503			altfs->lfs_dlfs = *((struct dlfs *) bp->b_data);
504			altfs->lfs_devvp = devvp;
505			bp->b_flags |= B_INVAL;
506			brelse(bp);
507
508			if (check_sb(fs) || fs->lfs_idaddr <= 0) {
509				if (debug)
510					printf("Primary superblock is no good, using first alternate\n");
511				free(fs);
512				fs = altfs;
513			} else {
514				/* If both superblocks check out, try verification */
515				if (check_sb(altfs)) {
516					if (debug)
517						printf("First alternate superblock is no good, using primary\n");
518					free(altfs);
519				} else {
520					if (lfs_verify(fs, altfs, devvp, debug) == fs) {
521						free(altfs);
522					} else {
523						free(fs);
524						fs = altfs;
525					}
526				}
527			}
528		}
529		if (check_sb(fs)) {
530			free(fs);
531			return NULL;
532		}
533	}
534
535	/* Compatibility */
536	if (fs->lfs_version < 2) {
537		fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE;
538		fs->lfs_ibsize = fs->lfs_bsize;
539		fs->lfs_start = fs->lfs_sboffs[0];
540		fs->lfs_tstamp = fs->lfs_otstamp;
541		fs->lfs_fsbtodb = 0;
542	}
543
544	if (!dummy_read) {
545		fs->lfs_suflags = (u_int32_t **) malloc(2 * sizeof(u_int32_t *));
546		fs->lfs_suflags[0] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
547		fs->lfs_suflags[1] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
548	}
549
550	if (idaddr == 0)
551		idaddr = fs->lfs_idaddr;
552	else
553		fs->lfs_idaddr = idaddr;
554	/* NB: If dummy_read!=0, idaddr==0 here so we get a fake inode. */
555	fs->lfs_ivnode = lfs_raw_vget(fs,
556		(dummy_read ? LFS_IFILE_INUM : fs->lfs_ifile), devvp->v_fd,
557		idaddr);
558
559	register_vget((void *)fs, lfs_vget);
560
561	return fs;
562}
563
564/*
565 * Check partial segment validity between fs->lfs_offset and the given goal.
566 *
567 * If goal == 0, just keep on going until the segments stop making sense,
568 * and return the address of the last valid partial segment.
569 *
570 * If goal != 0, return the address of the first partial segment that failed,
571 * or "goal" if we reached it without failure (the partial segment *at* goal
572 * need not be valid).
573 */
574ufs_daddr_t
575try_verify(struct lfs *osb, struct uvnode *devvp, ufs_daddr_t goal, int debug)
576{
577	ufs_daddr_t daddr, odaddr;
578	SEGSUM *sp;
579	int bc, flag;
580	struct ubuf *bp;
581	ufs_daddr_t nodirop_daddr;
582	u_int64_t serial;
583
584	odaddr = -1;
585	daddr = osb->lfs_offset;
586	nodirop_daddr = daddr;
587	serial = osb->lfs_serial;
588	while (daddr != goal) {
589		flag = 0;
590oncemore:
591		/* Read in summary block */
592		bread(devvp, fsbtodb(osb, daddr), osb->lfs_sumsize, NULL, &bp);
593		sp = (SEGSUM *)bp->b_data;
594
595		/*
596		 * Could be a superblock instead of a segment summary.
597		 * XXX should use gseguse, but right now we need to do more
598		 * setup before we can...fix this
599		 */
600		if (sp->ss_magic != SS_MAGIC ||
601		    sp->ss_ident != osb->lfs_ident ||
602		    sp->ss_serial < serial ||
603		    sp->ss_sumsum != cksum(&sp->ss_datasum, osb->lfs_sumsize -
604			sizeof(sp->ss_sumsum))) {
605			brelse(bp);
606			if (flag == 0) {
607				flag = 1;
608				daddr += btofsb(osb, LFS_SBPAD);
609				goto oncemore;
610			}
611			break;
612		}
613		++serial;
614		bc = check_summary(osb, sp, daddr, debug, devvp, NULL);
615		if (bc == 0) {
616			brelse(bp);
617			break;
618		}
619		assert (bc > 0);
620		odaddr = daddr;
621		daddr += btofsb(osb, osb->lfs_sumsize + bc);
622		if (dtosn(osb, odaddr) != dtosn(osb, daddr) ||
623		    dtosn(osb, daddr) != dtosn(osb, daddr +
624			btofsb(osb, osb->lfs_sumsize + osb->lfs_bsize))) {
625			daddr = sp->ss_next;
626		}
627		if (!(sp->ss_flags & SS_CONT))
628			nodirop_daddr = daddr;
629		brelse(bp);
630	}
631
632	if (goal == 0)
633		return nodirop_daddr;
634	else
635		return daddr;
636}
637
638/* Use try_verify to check whether the newer superblock is valid. */
639struct lfs *
640lfs_verify(struct lfs *sb0, struct lfs *sb1, struct uvnode *devvp, int debug)
641{
642	ufs_daddr_t daddr;
643	struct lfs *osb, *nsb;
644
645	/*
646	 * Verify the checkpoint of the newer superblock,
647	 * if the timestamp/serial number of the two superblocks is
648	 * different.
649	 */
650
651	osb = NULL;
652	if (debug)
653		printf("sb0 %lld, sb1 %lld\n", (long long) sb0->lfs_serial,
654		    (long long) sb1->lfs_serial);
655
656	if ((sb0->lfs_version == 1 &&
657		sb0->lfs_otstamp != sb1->lfs_otstamp) ||
658	    (sb0->lfs_version > 1 &&
659		sb0->lfs_serial != sb1->lfs_serial)) {
660		if (sb0->lfs_version == 1) {
661			if (sb0->lfs_otstamp > sb1->lfs_otstamp) {
662				osb = sb1;
663				nsb = sb0;
664			} else {
665				osb = sb0;
666				nsb = sb1;
667			}
668		} else {
669			if (sb0->lfs_serial > sb1->lfs_serial) {
670				osb = sb1;
671				nsb = sb0;
672			} else {
673				osb = sb0;
674				nsb = sb1;
675			}
676		}
677		if (debug) {
678			printf("Attempting to verify newer checkpoint...");
679			fflush(stdout);
680		}
681		daddr = try_verify(osb, devvp, nsb->lfs_offset, debug);
682
683		if (debug)
684			printf("done.\n");
685		if (daddr == nsb->lfs_offset) {
686			pwarn("** Newer checkpoint verified, recovered %lld seconds of data\n",
687			    (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
688			sbdirty();
689		} else {
690			pwarn("** Newer checkpoint invalid, lost %lld seconds of data\n", (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
691		}
692		return (daddr == nsb->lfs_offset ? nsb : osb);
693	}
694	/* Nothing to check */
695	return osb;
696}
697
698/* Verify a partial-segment summary; return the number of bytes on disk. */
699int
700check_summary(struct lfs *fs, SEGSUM *sp, ufs_daddr_t pseg_addr, int debug,
701	      struct uvnode *devvp, void (func(ufs_daddr_t, FINFO *)))
702{
703	FINFO *fp;
704	int bc;			/* Bytes in partial segment */
705	int nblocks;
706	ufs_daddr_t seg_addr, daddr;
707	ufs_daddr_t *dp, *idp;
708	struct ubuf *bp;
709	int i, j, k, datac, len;
710	long sn;
711	u_int32_t *datap;
712	u_int32_t ccksum;
713
714	sn = dtosn(fs, pseg_addr);
715	seg_addr = sntod(fs, sn);
716
717	/* We've already checked the sumsum, just do the data bounds and sum */
718
719	/* Count the blocks. */
720	nblocks = howmany(sp->ss_ninos, INOPB(fs));
721	bc = nblocks << (fs->lfs_version > 1 ? fs->lfs_ffshift : fs->lfs_bshift);
722	assert(bc >= 0);
723
724	fp = (FINFO *) (sp + 1);
725	for (i = 0; i < sp->ss_nfinfo; i++) {
726		nblocks += fp->fi_nblocks;
727		bc += fp->fi_lastlength + ((fp->fi_nblocks - 1)
728					   << fs->lfs_bshift);
729		assert(bc >= 0);
730		fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
731	}
732	datap = (u_int32_t *) malloc(nblocks * sizeof(*datap));
733	datac = 0;
734
735	dp = (ufs_daddr_t *) sp;
736	dp += fs->lfs_sumsize / sizeof(ufs_daddr_t);
737	dp--;
738
739	idp = dp;
740	daddr = pseg_addr + btofsb(fs, fs->lfs_sumsize);
741	fp = (FINFO *) (sp + 1);
742	for (i = 0, j = 0;
743	     i < sp->ss_nfinfo || j < howmany(sp->ss_ninos, INOPB(fs)); i++) {
744		if (i >= sp->ss_nfinfo && *idp != daddr) {
745			pwarn("Not enough inode blocks in pseg at 0x%" PRIx32
746			      ": found %d, wanted %d\n",
747			      pseg_addr, j, howmany(sp->ss_ninos, INOPB(fs)));
748			if (debug)
749				pwarn("*idp=%x, daddr=%" PRIx32 "\n", *idp,
750				      daddr);
751			break;
752		}
753		while (j < howmany(sp->ss_ninos, INOPB(fs)) && *idp == daddr) {
754			bread(devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NOCRED, &bp);
755			datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
756			brelse(bp);
757
758			++j;
759			daddr += btofsb(fs, fs->lfs_ibsize);
760			--idp;
761		}
762		if (i < sp->ss_nfinfo) {
763			if (func)
764				func(daddr, fp);
765			for (k = 0; k < fp->fi_nblocks; k++) {
766				len = (k == fp->fi_nblocks - 1 ?
767				       fp->fi_lastlength
768				       : fs->lfs_bsize);
769				bread(devvp, fsbtodb(fs, daddr), len, NOCRED, &bp);
770				datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
771				brelse(bp);
772				daddr += btofsb(fs, len);
773			}
774			fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
775		}
776	}
777
778	if (datac != nblocks) {
779		pwarn("Partial segment at 0x%llx expected %d blocks counted %d\n",
780		    (long long) pseg_addr, nblocks, datac);
781	}
782	ccksum = cksum(datap, nblocks * sizeof(u_int32_t));
783	/* Check the data checksum */
784	if (ccksum != sp->ss_datasum) {
785		pwarn("Partial segment at 0x%" PRIx32 " data checksum"
786		      " mismatch: given 0x%x, computed 0x%x\n",
787		      pseg_addr, sp->ss_datasum, ccksum);
788		free(datap);
789		return 0;
790	}
791	free(datap);
792	assert(bc >= 0);
793	return bc;
794}
795
796/* print message and exit */
797void
798my_vpanic(int fatal, const char *fmt, va_list ap)
799{
800        (void) vprintf(fmt, ap);
801	exit(8);
802}
803
804void
805call_panic(const char *fmt, ...)
806{
807	va_list ap;
808
809	va_start(ap, fmt);
810        panic_func(1, fmt, ap);
811	va_end(ap);
812}
813