xref: /netbsd-src/sbin/fsck_lfs/lfs.c (revision bf1e9b32e27832f0c493206710fb8b58a980838a)
1 /* $NetBSD: lfs.c,v 1.16 2005/06/08 19:09:55 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@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[VNODE_HASH_MAX];
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 	metalbn = 0;	/* XXXGCC -Wuninitialized [dreamcast] */
244 
245 	if (nump)
246 		*nump = 0;
247 	numlevels = 0;
248 	realbn = bn;
249 	if (bn < 0)
250 		bn = -bn;
251 
252 	lognindir = -1;
253 	for (indir = fs->lfs_nindir; indir; indir >>= 1)
254 		++lognindir;
255 
256 	/* Determine the number of levels of indirection.  After this loop is
257 	 * done, blockcnt indicates the number of data blocks possible at the
258 	 * given level of indirection, and NIADDR - i is the number of levels
259 	 * of indirection needed to locate the requested block. */
260 
261 	bn -= NDADDR;
262 	for (lbc = 0, i = NIADDR;; i--, bn -= blockcnt) {
263 		if (i == 0)
264 			return (EFBIG);
265 
266 		lbc += lognindir;
267 		blockcnt = (int64_t) 1 << lbc;
268 
269 		if (bn < blockcnt)
270 			break;
271 	}
272 
273 	/* Calculate the address of the first meta-block. */
274 	if (realbn >= 0)
275 		metalbn = -(realbn - bn + NIADDR - i);
276 	else
277 		metalbn = -(-realbn - bn + NIADDR - i);
278 
279 	/* At each iteration, off is the offset into the bap array which is an
280 	 * array of disk addresses at the current level of indirection. The
281 	 * logical block number and the offset in that block are stored into
282 	 * the argument array. */
283 	ap->in_lbn = metalbn;
284 	ap->in_off = off = NIADDR - i;
285 	ap->in_exists = 0;
286 	ap++;
287 	for (++numlevels; i <= NIADDR; i++) {
288 		/* If searching for a meta-data block, quit when found. */
289 		if (metalbn == realbn)
290 			break;
291 
292 		lbc -= lognindir;
293 		blockcnt = (int64_t) 1 << lbc;
294 		off = (bn >> lbc) & (fs->lfs_nindir - 1);
295 
296 		++numlevels;
297 		ap->in_lbn = metalbn;
298 		ap->in_off = off;
299 		ap->in_exists = 0;
300 		++ap;
301 
302 		metalbn -= -1 + (off << lbc);
303 	}
304 	if (nump)
305 		*nump = numlevels;
306 	return (0);
307 }
308 
309 int
310 lfs_vop_bmap(struct uvnode * vp, daddr_t lbn, daddr_t * daddrp)
311 {
312 	return ufs_bmaparray(vp->v_fs, vp, lbn, daddrp, NULL, NULL);
313 }
314 
315 /* Search a block for a specific dinode. */
316 struct ufs1_dinode *
317 lfs_ifind(struct lfs * fs, ino_t ino, struct ubuf * bp)
318 {
319 	struct ufs1_dinode *dip = (struct ufs1_dinode *) bp->b_data;
320 	struct ufs1_dinode *ldip, *fin;
321 
322 	fin = dip + INOPB(fs);
323 
324 	/*
325 	 * Read the inode block backwards, since later versions of the
326 	 * inode will supercede earlier ones.  Though it is unlikely, it is
327 	 * possible that the same inode will appear in the same inode block.
328 	 */
329 	for (ldip = fin - 1; ldip >= dip; --ldip)
330 		if (ldip->di_inumber == ino)
331 			return (ldip);
332 	return NULL;
333 }
334 
335 /*
336  * lfs_raw_vget makes us a new vnode from the inode at the given disk address.
337  * XXX it currently loses atime information.
338  */
339 struct uvnode *
340 lfs_raw_vget(struct lfs * fs, ino_t ino, int fd, ufs_daddr_t daddr)
341 {
342 	struct uvnode *vp;
343 	struct inode *ip;
344 	struct ufs1_dinode *dip;
345 	struct ubuf *bp;
346 	int i, hash;
347 
348 	vp = (struct uvnode *) malloc(sizeof(*vp));
349 	memset(vp, 0, sizeof(*vp));
350 	vp->v_fd = fd;
351 	vp->v_fs = fs;
352 	vp->v_usecount = 0;
353 	vp->v_strategy_op = lfs_vop_strategy;
354 	vp->v_bwrite_op = lfs_vop_bwrite;
355 	vp->v_bmap_op = lfs_vop_bmap;
356 	LIST_INIT(&vp->v_cleanblkhd);
357 	LIST_INIT(&vp->v_dirtyblkhd);
358 
359 	ip = (struct inode *) malloc(sizeof(*ip));
360 	memset(ip, 0, sizeof(*ip));
361 
362 	ip->i_din.ffs1_din = (struct ufs1_dinode *)
363 	    malloc(sizeof(struct ufs1_dinode));
364 	memset(ip->i_din.ffs1_din, 0, sizeof (struct ufs1_dinode));
365 
366 	/* Initialize the inode -- from lfs_vcreate. */
367 	ip->inode_ext.lfs = malloc(sizeof(struct lfs_inode_ext));
368 	memset(ip->inode_ext.lfs, 0, sizeof(struct lfs_inode_ext));
369 	vp->v_data = ip;
370 	/* ip->i_vnode = vp; */
371 	ip->i_number = ino;
372 	ip->i_lockf = 0;
373 	ip->i_diroff = 0;
374 	ip->i_lfs_effnblks = 0;
375 	ip->i_flag = 0;
376 
377 	/* Load inode block and find inode */
378 	if (daddr > 0) {
379 		bread(fs->lfs_devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NULL, &bp);
380 		bp->b_flags |= B_AGE;
381 		dip = lfs_ifind(fs, ino, bp);
382 		if (dip == NULL) {
383 			brelse(bp);
384 			free(ip);
385 			free(vp);
386 			return NULL;
387 		}
388 		memcpy(ip->i_din.ffs1_din, dip, sizeof(*dip));
389 		brelse(bp);
390 	}
391 	ip->i_number = ino;
392 	/* ip->i_devvp = fs->lfs_devvp; */
393 	ip->i_lfs = fs;
394 
395 	ip->i_ffs_effnlink = ip->i_ffs1_nlink;
396 	ip->i_lfs_effnblks = ip->i_ffs1_blocks;
397 	ip->i_lfs_osize = ip->i_ffs1_size;
398 #if 0
399 	if (fs->lfs_version > 1) {
400 		ip->i_ffs1_atime = ts.tv_sec;
401 		ip->i_ffs1_atimensec = ts.tv_nsec;
402 	}
403 #endif
404 
405 	memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize));
406 	for (i = 0; i < NDADDR; i++)
407 		if (ip->i_ffs1_db[i] != 0)
408 			ip->i_lfs_fragsize[i] = blksize(fs, ip, i);
409 
410 	++nvnodes;
411 	hash = ((int)(intptr_t)fs + ino) & (VNODE_HASH_MAX - 1);
412 	LIST_INSERT_HEAD(&getvnodelist[hash], vp, v_getvnodes);
413 	LIST_INSERT_HEAD(&vnodelist, vp, v_mntvnodes);
414 
415 	return vp;
416 }
417 
418 static struct uvnode *
419 lfs_vget(void *vfs, ino_t ino)
420 {
421 	struct lfs *fs = (struct lfs *)vfs;
422 	ufs_daddr_t daddr;
423 	struct ubuf *bp;
424 	IFILE *ifp;
425 
426 	LFS_IENTRY(ifp, fs, ino, bp);
427 	daddr = ifp->if_daddr;
428 	brelse(bp);
429 	if (daddr <= 0 || dtosn(fs, daddr) >= fs->lfs_nseg)
430 		return NULL;
431 	return lfs_raw_vget(fs, ino, fs->lfs_ivnode->v_fd, daddr);
432 }
433 
434 /* Check superblock magic number and checksum */
435 static int
436 check_sb(struct lfs *fs)
437 {
438 	u_int32_t checksum;
439 
440 	if (fs->lfs_magic != LFS_MAGIC) {
441 		printf("Superblock magic number (0x%lx) does not match "
442 		       "expected 0x%lx\n", (unsigned long) fs->lfs_magic,
443 		       (unsigned long) LFS_MAGIC);
444 		return 1;
445 	}
446 	/* checksum */
447 	checksum = lfs_sb_cksum(&(fs->lfs_dlfs));
448 	if (fs->lfs_cksum != checksum) {
449 		printf("Superblock checksum (%lx) does not match computed checksum (%lx)\n",
450 		    (unsigned long) fs->lfs_cksum, (unsigned long) checksum);
451 		return 1;
452 	}
453 	return 0;
454 }
455 
456 /* Initialize LFS library; load superblocks and choose which to use. */
457 struct lfs *
458 lfs_init(int devfd, daddr_t sblkno, daddr_t idaddr, int dummy_read, int debug)
459 {
460 	struct uvnode *devvp;
461 	struct ubuf *bp;
462 	int tryalt;
463 	struct lfs *fs, *altfs;
464 	int error;
465 
466 	vfs_init();
467 
468 	devvp = (struct uvnode *) malloc(sizeof(*devvp));
469 	memset(devvp, 0, sizeof(*devvp));
470 	devvp->v_fs = NULL;
471 	devvp->v_fd = devfd;
472 	devvp->v_strategy_op = raw_vop_strategy;
473 	devvp->v_bwrite_op = raw_vop_bwrite;
474 	devvp->v_bmap_op = raw_vop_bmap;
475 	LIST_INIT(&devvp->v_cleanblkhd);
476 	LIST_INIT(&devvp->v_dirtyblkhd);
477 
478 	tryalt = 0;
479 	if (dummy_read) {
480 		if (sblkno == 0)
481 			sblkno = btodb(LFS_LABELPAD);
482 		fs = (struct lfs *) malloc(sizeof(*fs));
483 		memset(fs, 0, sizeof(*fs));
484 		fs->lfs_devvp = devvp;
485 	} else {
486 		if (sblkno == 0) {
487 			sblkno = btodb(LFS_LABELPAD);
488 			tryalt = 1;
489 		} else if (debug) {
490 			printf("No -b flag given, not attempting to verify checkpoint\n");
491 		}
492 		error = bread(devvp, sblkno, LFS_SBPAD, NOCRED, &bp);
493 		fs = (struct lfs *) malloc(sizeof(*fs));
494 		memset(fs, 0, sizeof(*fs));
495 		fs->lfs_dlfs = *((struct dlfs *) bp->b_data);
496 		fs->lfs_devvp = devvp;
497 		bp->b_flags |= B_INVAL;
498 		brelse(bp);
499 
500 		if (tryalt) {
501 			error = bread(devvp, fsbtodb(fs, fs->lfs_sboffs[1]),
502 		    	LFS_SBPAD, NOCRED, &bp);
503 			altfs = (struct lfs *) malloc(sizeof(*altfs));
504 			memset(altfs, 0, sizeof(*altfs));
505 			altfs->lfs_dlfs = *((struct dlfs *) bp->b_data);
506 			altfs->lfs_devvp = devvp;
507 			bp->b_flags |= B_INVAL;
508 			brelse(bp);
509 
510 			if (check_sb(fs) || fs->lfs_idaddr <= 0) {
511 				if (debug)
512 					printf("Primary superblock is no good, using first alternate\n");
513 				free(fs);
514 				fs = altfs;
515 			} else {
516 				/* If both superblocks check out, try verification */
517 				if (check_sb(altfs)) {
518 					if (debug)
519 						printf("First alternate superblock is no good, using primary\n");
520 					free(altfs);
521 				} else {
522 					if (lfs_verify(fs, altfs, devvp, debug) == fs) {
523 						free(altfs);
524 					} else {
525 						free(fs);
526 						fs = altfs;
527 					}
528 				}
529 			}
530 		}
531 		if (check_sb(fs)) {
532 			free(fs);
533 			return NULL;
534 		}
535 	}
536 
537 	/* Compatibility */
538 	if (fs->lfs_version < 2) {
539 		fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE;
540 		fs->lfs_ibsize = fs->lfs_bsize;
541 		fs->lfs_start = fs->lfs_sboffs[0];
542 		fs->lfs_tstamp = fs->lfs_otstamp;
543 		fs->lfs_fsbtodb = 0;
544 	}
545 
546 	if (!dummy_read) {
547 		fs->lfs_suflags = (u_int32_t **) malloc(2 * sizeof(u_int32_t *));
548 		fs->lfs_suflags[0] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
549 		fs->lfs_suflags[1] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t));
550 	}
551 
552 	if (idaddr == 0)
553 		idaddr = fs->lfs_idaddr;
554 	else
555 		fs->lfs_idaddr = idaddr;
556 	/* NB: If dummy_read!=0, idaddr==0 here so we get a fake inode. */
557 	fs->lfs_ivnode = lfs_raw_vget(fs,
558 		(dummy_read ? LFS_IFILE_INUM : fs->lfs_ifile), devvp->v_fd,
559 		idaddr);
560 
561 	register_vget((void *)fs, lfs_vget);
562 
563 	return fs;
564 }
565 
566 /*
567  * Check partial segment validity between fs->lfs_offset and the given goal.
568  *
569  * If goal == 0, just keep on going until the segments stop making sense,
570  * and return the address of the last valid partial segment.
571  *
572  * If goal != 0, return the address of the first partial segment that failed,
573  * or "goal" if we reached it without failure (the partial segment *at* goal
574  * need not be valid).
575  */
576 ufs_daddr_t
577 try_verify(struct lfs *osb, struct uvnode *devvp, ufs_daddr_t goal, int debug)
578 {
579 	ufs_daddr_t daddr, odaddr;
580 	SEGSUM *sp;
581 	int bc, flag;
582 	struct ubuf *bp;
583 	ufs_daddr_t nodirop_daddr;
584 	u_int64_t serial;
585 
586 	odaddr = -1;
587 	daddr = osb->lfs_offset;
588 	nodirop_daddr = daddr;
589 	serial = osb->lfs_serial;
590 	while (daddr != goal) {
591 		flag = 0;
592 oncemore:
593 		/* Read in summary block */
594 		bread(devvp, fsbtodb(osb, daddr), osb->lfs_sumsize, NULL, &bp);
595 		sp = (SEGSUM *)bp->b_data;
596 
597 		/*
598 		 * Could be a superblock instead of a segment summary.
599 		 * XXX should use gseguse, but right now we need to do more
600 		 * setup before we can...fix this
601 		 */
602 		if (sp->ss_magic != SS_MAGIC ||
603 		    sp->ss_ident != osb->lfs_ident ||
604 		    sp->ss_serial < serial ||
605 		    sp->ss_sumsum != cksum(&sp->ss_datasum, osb->lfs_sumsize -
606 			sizeof(sp->ss_sumsum))) {
607 			brelse(bp);
608 			if (flag == 0) {
609 				flag = 1;
610 				daddr += btofsb(osb, LFS_SBPAD);
611 				goto oncemore;
612 			}
613 			break;
614 		}
615 		++serial;
616 		bc = check_summary(osb, sp, daddr, debug, devvp, NULL);
617 		if (bc == 0) {
618 			brelse(bp);
619 			break;
620 		}
621 		assert (bc > 0);
622 		odaddr = daddr;
623 		daddr += btofsb(osb, osb->lfs_sumsize + bc);
624 		if (dtosn(osb, odaddr) != dtosn(osb, daddr) ||
625 		    dtosn(osb, daddr) != dtosn(osb, daddr +
626 			btofsb(osb, osb->lfs_sumsize + osb->lfs_bsize))) {
627 			daddr = sp->ss_next;
628 		}
629 		if (!(sp->ss_flags & SS_CONT))
630 			nodirop_daddr = daddr;
631 		brelse(bp);
632 	}
633 
634 	if (goal == 0)
635 		return nodirop_daddr;
636 	else
637 		return daddr;
638 }
639 
640 /* Use try_verify to check whether the newer superblock is valid. */
641 struct lfs *
642 lfs_verify(struct lfs *sb0, struct lfs *sb1, struct uvnode *devvp, int debug)
643 {
644 	ufs_daddr_t daddr;
645 	struct lfs *osb, *nsb;
646 
647 	/*
648 	 * Verify the checkpoint of the newer superblock,
649 	 * if the timestamp/serial number of the two superblocks is
650 	 * different.
651 	 */
652 
653 	osb = NULL;
654 	if (debug)
655 		printf("sb0 %lld, sb1 %lld\n", (long long) sb0->lfs_serial,
656 		    (long long) sb1->lfs_serial);
657 
658 	if ((sb0->lfs_version == 1 &&
659 		sb0->lfs_otstamp != sb1->lfs_otstamp) ||
660 	    (sb0->lfs_version > 1 &&
661 		sb0->lfs_serial != sb1->lfs_serial)) {
662 		if (sb0->lfs_version == 1) {
663 			if (sb0->lfs_otstamp > sb1->lfs_otstamp) {
664 				osb = sb1;
665 				nsb = sb0;
666 			} else {
667 				osb = sb0;
668 				nsb = sb1;
669 			}
670 		} else {
671 			if (sb0->lfs_serial > sb1->lfs_serial) {
672 				osb = sb1;
673 				nsb = sb0;
674 			} else {
675 				osb = sb0;
676 				nsb = sb1;
677 			}
678 		}
679 		if (debug) {
680 			printf("Attempting to verify newer checkpoint...");
681 			fflush(stdout);
682 		}
683 		daddr = try_verify(osb, devvp, nsb->lfs_offset, debug);
684 
685 		if (debug)
686 			printf("done.\n");
687 		if (daddr == nsb->lfs_offset) {
688 			pwarn("** Newer checkpoint verified, recovered %lld seconds of data\n",
689 			    (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
690 			sbdirty();
691 		} else {
692 			pwarn("** Newer checkpoint invalid, lost %lld seconds of data\n", (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp);
693 		}
694 		return (daddr == nsb->lfs_offset ? nsb : osb);
695 	}
696 	/* Nothing to check */
697 	return osb;
698 }
699 
700 /* Verify a partial-segment summary; return the number of bytes on disk. */
701 int
702 check_summary(struct lfs *fs, SEGSUM *sp, ufs_daddr_t pseg_addr, int debug,
703 	      struct uvnode *devvp, void (func(ufs_daddr_t, FINFO *)))
704 {
705 	FINFO *fp;
706 	int bc;			/* Bytes in partial segment */
707 	int nblocks;
708 	ufs_daddr_t seg_addr, daddr;
709 	ufs_daddr_t *dp, *idp;
710 	struct ubuf *bp;
711 	int i, j, k, datac, len;
712 	long sn;
713 	u_int32_t *datap;
714 	u_int32_t ccksum;
715 
716 	sn = dtosn(fs, pseg_addr);
717 	seg_addr = sntod(fs, sn);
718 
719 	/* We've already checked the sumsum, just do the data bounds and sum */
720 
721 	/* Count the blocks. */
722 	nblocks = howmany(sp->ss_ninos, INOPB(fs));
723 	bc = nblocks << (fs->lfs_version > 1 ? fs->lfs_ffshift : fs->lfs_bshift);
724 	assert(bc >= 0);
725 
726 	fp = (FINFO *) (sp + 1);
727 	for (i = 0; i < sp->ss_nfinfo; i++) {
728 		nblocks += fp->fi_nblocks;
729 		bc += fp->fi_lastlength + ((fp->fi_nblocks - 1)
730 					   << fs->lfs_bshift);
731 		assert(bc >= 0);
732 		fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
733 	}
734 	datap = (u_int32_t *) malloc(nblocks * sizeof(*datap));
735 	datac = 0;
736 
737 	dp = (ufs_daddr_t *) sp;
738 	dp += fs->lfs_sumsize / sizeof(ufs_daddr_t);
739 	dp--;
740 
741 	idp = dp;
742 	daddr = pseg_addr + btofsb(fs, fs->lfs_sumsize);
743 	fp = (FINFO *) (sp + 1);
744 	for (i = 0, j = 0;
745 	     i < sp->ss_nfinfo || j < howmany(sp->ss_ninos, INOPB(fs)); i++) {
746 		if (i >= sp->ss_nfinfo && *idp != daddr) {
747 			pwarn("Not enough inode blocks in pseg at 0x%" PRIx32
748 			      ": found %d, wanted %d\n",
749 			      pseg_addr, j, howmany(sp->ss_ninos, INOPB(fs)));
750 			if (debug)
751 				pwarn("*idp=%x, daddr=%" PRIx32 "\n", *idp,
752 				      daddr);
753 			break;
754 		}
755 		while (j < howmany(sp->ss_ninos, INOPB(fs)) && *idp == daddr) {
756 			bread(devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NOCRED, &bp);
757 			datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
758 			brelse(bp);
759 
760 			++j;
761 			daddr += btofsb(fs, fs->lfs_ibsize);
762 			--idp;
763 		}
764 		if (i < sp->ss_nfinfo) {
765 			if (func)
766 				func(daddr, fp);
767 			for (k = 0; k < fp->fi_nblocks; k++) {
768 				len = (k == fp->fi_nblocks - 1 ?
769 				       fp->fi_lastlength
770 				       : fs->lfs_bsize);
771 				bread(devvp, fsbtodb(fs, daddr), len, NOCRED, &bp);
772 				datap[datac++] = ((u_int32_t *) (bp->b_data))[0];
773 				brelse(bp);
774 				daddr += btofsb(fs, len);
775 			}
776 			fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks);
777 		}
778 	}
779 
780 	if (datac != nblocks) {
781 		pwarn("Partial segment at 0x%llx expected %d blocks counted %d\n",
782 		    (long long) pseg_addr, nblocks, datac);
783 	}
784 	ccksum = cksum(datap, nblocks * sizeof(u_int32_t));
785 	/* Check the data checksum */
786 	if (ccksum != sp->ss_datasum) {
787 		pwarn("Partial segment at 0x%" PRIx32 " data checksum"
788 		      " mismatch: given 0x%x, computed 0x%x\n",
789 		      pseg_addr, sp->ss_datasum, ccksum);
790 		free(datap);
791 		return 0;
792 	}
793 	free(datap);
794 	assert(bc >= 0);
795 	return bc;
796 }
797 
798 /* print message and exit */
799 void
800 my_vpanic(int fatal, const char *fmt, va_list ap)
801 {
802         (void) vprintf(fmt, ap);
803 	exit(8);
804 }
805 
806 void
807 call_panic(const char *fmt, ...)
808 {
809 	va_list ap;
810 
811 	va_start(ap, fmt);
812         panic_func(1, fmt, ap);
813 	va_end(ap);
814 }
815 
816 /* Allocate a new inode. */
817 struct uvnode *
818 lfs_valloc(struct lfs *fs, ino_t ino)
819 {
820 	struct ubuf *bp, *cbp;
821 	struct ifile *ifp;
822 	ino_t new_ino;
823 	int error;
824 	int new_gen;
825 	CLEANERINFO *cip;
826 
827 	/* Get the head of the freelist. */
828 	LFS_GET_HEADFREE(fs, cip, cbp, &new_ino);
829 
830 	/*
831 	 * Remove the inode from the free list and write the new start
832 	 * of the free list into the superblock.
833 	 */
834 	LFS_IENTRY(ifp, fs, new_ino, bp);
835 	if (ifp->if_daddr != LFS_UNUSED_DADDR)
836 		panic("lfs_valloc: inuse inode %d on the free list", new_ino);
837 	LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree);
838 
839 	new_gen = ifp->if_version; /* version was updated by vfree */
840 	brelse(bp);
841 
842 	/* Extend IFILE so that the next lfs_valloc will succeed. */
843 	if (fs->lfs_freehd == LFS_UNUSED_INUM) {
844 		if ((error = extend_ifile(fs)) != 0) {
845 			LFS_PUT_HEADFREE(fs, cip, cbp, new_ino);
846 			return NULL;
847 		}
848 	}
849 
850 	/* Set superblock modified bit and increment file count. */
851         sbdirty();
852 	++fs->lfs_nfiles;
853 
854         return lfs_raw_vget(fs, ino, fs->lfs_devvp->v_fd, 0x0);
855 }
856 
857 /*
858  * Add a new block to the Ifile, to accommodate future file creations.
859  */
860 int
861 extend_ifile(struct lfs *fs)
862 {
863 	struct uvnode *vp;
864 	struct inode *ip;
865 	IFILE *ifp;
866 	IFILE_V1 *ifp_v1;
867 	struct ubuf *bp, *cbp;
868 	daddr_t i, blkno, max;
869 	ino_t oldlast;
870 	CLEANERINFO *cip;
871 
872 	vp = fs->lfs_ivnode;
873 	ip = VTOI(vp);
874 	blkno = lblkno(fs, ip->i_ffs1_size);
875 
876 	bp = getblk(vp, blkno, fs->lfs_bsize);	/* XXX VOP_BALLOC() */
877 	ip->i_ffs1_size += fs->lfs_bsize;
878 
879 	i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) *
880 		fs->lfs_ifpb;
881 	LFS_GET_HEADFREE(fs, cip, cbp, &oldlast);
882 	LFS_PUT_HEADFREE(fs, cip, cbp, i);
883 	max = i + fs->lfs_ifpb;
884 	fs->lfs_bfree -= btofsb(fs, fs->lfs_bsize);
885 
886 	if (fs->lfs_version == 1) {
887 		for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) {
888 			ifp_v1->if_version = 1;
889 			ifp_v1->if_daddr = LFS_UNUSED_DADDR;
890 			ifp_v1->if_nextfree = ++i;
891 		}
892 		ifp_v1--;
893 		ifp_v1->if_nextfree = oldlast;
894 	} else {
895 		for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) {
896 			ifp->if_version = 1;
897 			ifp->if_daddr = LFS_UNUSED_DADDR;
898 			ifp->if_nextfree = ++i;
899 		}
900 		ifp--;
901 		ifp->if_nextfree = oldlast;
902 	}
903 	LFS_PUT_TAILFREE(fs, cip, cbp, max - 1);
904 
905 	LFS_BWRITE_LOG(bp);
906 
907 	return 0;
908 }
909 
910