xref: /netbsd-src/sbin/dump/traverse.c (revision ce0bb6e8d2e560ecacbe865a848624f94498063b)
1 /*	$NetBSD: traverse.c,v 1.11 1995/03/27 22:14:47 mycroft Exp $	*/
2 
3 /*-
4  * Copyright (c) 1980, 1988, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 #ifndef lint
37 #if 0
38 static char sccsid[] = "@(#)traverse.c	8.2 (Berkeley) 9/23/93";
39 #else
40 static char rcsid[] = "$NetBSD: traverse.c,v 1.11 1995/03/27 22:14:47 mycroft Exp $";
41 #endif
42 #endif /* not lint */
43 
44 #include <sys/param.h>
45 #include <sys/time.h>
46 #include <sys/stat.h>
47 #ifdef sunos
48 #include <sys/vnode.h>
49 
50 #include <ufs/fs.h>
51 #include <ufs/fsdir.h>
52 #include <ufs/inode.h>
53 #else
54 #include <ufs/ffs/fs.h>
55 #include <ufs/ufs/dir.h>
56 #include <ufs/ufs/dinode.h>
57 #endif
58 
59 #include <protocols/dumprestore.h>
60 
61 #include <ctype.h>
62 #include <stdio.h>
63 #ifdef __STDC__
64 #include <string.h>
65 #include <unistd.h>
66 #endif
67 
68 #include "dump.h"
69 
70 #define	HASDUMPEDFILE	0x1
71 #define	HASSUBDIRS	0x2
72 
73 #ifdef	FS_44INODEFMT
74 typedef	quad_t fsizeT;
75 #else
76 typedef	long fsizeT;
77 #endif
78 
79 static	int dirindir __P((ino_t ino, daddr_t blkno, int level, long *size));
80 static	void dmpindir __P((ino_t ino, daddr_t blk, int level, fsizeT *size));
81 static	int searchdir __P((ino_t ino, daddr_t blkno, long size, long filesize));
82 
83 /*
84  * This is an estimation of the number of TP_BSIZE blocks in the file.
85  * It estimates the number of blocks in files with holes by assuming
86  * that all of the blocks accounted for by di_blocks are data blocks
87  * (when some of the blocks are usually used for indirect pointers);
88  * hence the estimate may be high.
89  */
90 long
91 blockest(dp)
92 	register struct dinode *dp;
93 {
94 	long blkest, sizeest;
95 
96 	/*
97 	 * dp->di_size is the size of the file in bytes.
98 	 * dp->di_blocks stores the number of sectors actually in the file.
99 	 * If there are more sectors than the size would indicate, this just
100 	 *	means that there are indirect blocks in the file or unused
101 	 *	sectors in the last file block; we can safely ignore these
102 	 *	(blkest = sizeest below).
103 	 * If the file is bigger than the number of sectors would indicate,
104 	 *	then the file has holes in it.	In this case we must use the
105 	 *	block count to estimate the number of data blocks used, but
106 	 *	we use the actual size for estimating the number of indirect
107 	 *	dump blocks (sizeest vs. blkest in the indirect block
108 	 *	calculation).
109 	 */
110 	blkest = howmany(dbtob(dp->di_blocks), TP_BSIZE);
111 	sizeest = howmany(dp->di_size, TP_BSIZE);
112 	if (blkest > sizeest)
113 		blkest = sizeest;
114 	if (dp->di_size > sblock->fs_bsize * NDADDR) {
115 		/* calculate the number of indirect blocks on the dump tape */
116 		blkest +=
117 			howmany(sizeest - NDADDR * sblock->fs_bsize / TP_BSIZE,
118 			TP_NINDIR);
119 	}
120 	return (blkest + 1);
121 }
122 
123 /* Auxiliary macro to pick up files changed since previous dump. */
124 #ifdef FS_44INODEFMT
125 #define	CHANGEDSINCE(dp, t) \
126 	((dp)->di_mtime.ts_sec >= (t) || (dp)->di_ctime.ts_sec >= (t))
127 #else
128 #define	CHANGEDSINCE(dp, t) \
129 	((dp)->di_mtime >= (t) || (dp)->di_ctime >= (t))
130 #endif
131 
132 /* The WANTTODUMP macro decides whether a file should be dumped. */
133 #ifdef UF_NODUMP
134 #define	WANTTODUMP(dp) \
135 	(CHANGEDSINCE(dp, spcl.c_ddate) && \
136 	 (nonodump || ((dp)->di_flags & UF_NODUMP) != UF_NODUMP))
137 #else
138 #define	WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate)
139 #endif
140 
141 /*
142  * Dump pass 1.
143  *
144  * Walk the inode list for a filesystem to find all allocated inodes
145  * that have been modified since the previous dump time. Also, find all
146  * the directories in the filesystem.
147  */
148 int
149 mapfiles(maxino, tapesize)
150 	ino_t maxino;
151 	long *tapesize;
152 {
153 	register int mode;
154 	register ino_t ino;
155 	register struct dinode *dp;
156 	int anydirskipped = 0;
157 
158 	for (ino = ROOTINO; ino < maxino; ino++) {
159 		dp = getino(ino);
160 		if ((mode = (dp->di_mode & IFMT)) == 0)
161 			continue;
162 		SETINO(ino, usedinomap);
163 		if (mode == IFDIR)
164 			SETINO(ino, dumpdirmap);
165 		if (WANTTODUMP(dp)) {
166 			SETINO(ino, dumpinomap);
167 			if (mode != IFREG && mode != IFDIR && mode != IFLNK)
168 				*tapesize += 1;
169 			else
170 				*tapesize += blockest(dp);
171 			continue;
172 		}
173 		if (mode == IFDIR)
174 			anydirskipped = 1;
175 	}
176 	/*
177 	 * Restore gets very upset if the root is not dumped,
178 	 * so ensure that it always is dumped.
179 	 */
180 	SETINO(ROOTINO, dumpinomap);
181 	return (anydirskipped);
182 }
183 
184 /*
185  * Dump pass 2.
186  *
187  * Scan each directory on the filesystem to see if it has any modified
188  * files in it. If it does, and has not already been added to the dump
189  * list (because it was itself modified), then add it. If a directory
190  * has not been modified itself, contains no modified files and has no
191  * subdirectories, then it can be deleted from the dump list and from
192  * the list of directories. By deleting it from the list of directories,
193  * its parent may now qualify for the same treatment on this or a later
194  * pass using this algorithm.
195  */
196 int
197 mapdirs(maxino, tapesize)
198 	ino_t maxino;
199 	long *tapesize;
200 {
201 	register struct	dinode *dp;
202 	register int i, isdir;
203 	register char *map;
204 	register ino_t ino;
205 	long filesize;
206 	int ret, change = 0;
207 
208 	isdir = 0;		/* XXX just to get gcc to shut up */
209 	for (map = dumpdirmap, ino = 1; ino < maxino; ino++) {
210 		if (((ino - 1) % NBBY) == 0)	/* map is offset by 1 */
211 			isdir = *map++;
212 		else
213 			isdir >>= 1;
214 		if ((isdir & 1) == 0 || TSTINO(ino, dumpinomap))
215 			continue;
216 		dp = getino(ino);
217 		filesize = dp->di_size;
218 		for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) {
219 			if (dp->di_db[i] != 0)
220 				ret |= searchdir(ino, dp->di_db[i],
221 					(long)dblksize(sblock, dp, i),
222 					filesize);
223 			if (ret & HASDUMPEDFILE)
224 				filesize = 0;
225 			else
226 				filesize -= sblock->fs_bsize;
227 		}
228 		for (i = 0; filesize > 0 && i < NIADDR; i++) {
229 			if (dp->di_ib[i] == 0)
230 				continue;
231 			ret |= dirindir(ino, dp->di_ib[i], i, &filesize);
232 		}
233 		if (ret & HASDUMPEDFILE) {
234 			SETINO(ino, dumpinomap);
235 			*tapesize += blockest(dp);
236 			change = 1;
237 			continue;
238 		}
239 		if ((ret & HASSUBDIRS) == 0) {
240 			if (!TSTINO(ino, dumpinomap)) {
241 				CLRINO(ino, dumpdirmap);
242 				change = 1;
243 			}
244 		}
245 	}
246 	return (change);
247 }
248 
249 /*
250  * Read indirect blocks, and pass the data blocks to be searched
251  * as directories. Quit as soon as any entry is found that will
252  * require the directory to be dumped.
253  */
254 static int
255 dirindir(ino, blkno, ind_level, filesize)
256 	ino_t ino;
257 	daddr_t blkno;
258 	int ind_level;
259 	long *filesize;
260 {
261 	int ret = 0;
262 	register int i;
263 	daddr_t	idblk[MAXNINDIR];
264 
265 	bread(fsbtodb(sblock, blkno), (char *)idblk, (int)sblock->fs_bsize);
266 	if (ind_level <= 0) {
267 		for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
268 			blkno = idblk[i];
269 			if (blkno != 0)
270 				ret |= searchdir(ino, blkno, sblock->fs_bsize,
271 					*filesize);
272 			if (ret & HASDUMPEDFILE)
273 				*filesize = 0;
274 			else
275 				*filesize -= sblock->fs_bsize;
276 		}
277 		return (ret);
278 	}
279 	ind_level--;
280 	for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
281 		blkno = idblk[i];
282 		if (blkno != 0)
283 			ret |= dirindir(ino, blkno, ind_level, filesize);
284 	}
285 	return (ret);
286 }
287 
288 /*
289  * Scan a disk block containing directory information looking to see if
290  * any of the entries are on the dump list and to see if the directory
291  * contains any subdirectories.
292  */
293 static int
294 searchdir(ino, blkno, size, filesize)
295 	ino_t ino;
296 	daddr_t blkno;
297 	register long size;
298 	long filesize;
299 {
300 	register struct direct *dp;
301 	register long loc, ret = 0;
302 	char dblk[MAXBSIZE];
303 
304 	bread(fsbtodb(sblock, blkno), dblk, (int)size);
305 	if (filesize < size)
306 		size = filesize;
307 	for (loc = 0; loc < size; ) {
308 		dp = (struct direct *)(dblk + loc);
309 		if (dp->d_reclen == 0) {
310 			msg("corrupted directory, inumber %d\n", ino);
311 			break;
312 		}
313 		loc += dp->d_reclen;
314 		if (dp->d_ino == 0)
315 			continue;
316 		if (dp->d_name[0] == '.') {
317 			if (dp->d_name[1] == '\0')
318 				continue;
319 			if (dp->d_name[1] == '.' && dp->d_name[2] == '\0')
320 				continue;
321 		}
322 		if (TSTINO(dp->d_ino, dumpinomap)) {
323 			ret |= HASDUMPEDFILE;
324 			if (ret & HASSUBDIRS)
325 				break;
326 		}
327 		if (TSTINO(dp->d_ino, dumpdirmap)) {
328 			ret |= HASSUBDIRS;
329 			if (ret & HASDUMPEDFILE)
330 				break;
331 		}
332 	}
333 	return (ret);
334 }
335 
336 /*
337  * Dump passes 3 and 4.
338  *
339  * Dump the contents of an inode to tape.
340  */
341 void
342 dumpino(dp, ino)
343 	register struct dinode *dp;
344 	ino_t ino;
345 {
346 	int ind_level, cnt;
347 	fsizeT size;
348 	char buf[TP_BSIZE];
349 
350 	if (newtape) {
351 		newtape = 0;
352 		dumpmap(dumpinomap, TS_BITS, ino);
353 	}
354 	CLRINO(ino, dumpinomap);
355 	spcl.c_dinode = *dp;
356 	spcl.c_type = TS_INODE;
357 	spcl.c_count = 0;
358 	switch (dp->di_mode & IFMT) {
359 
360 	case 0:
361 		/*
362 		 * Freed inode.
363 		 */
364 		return;
365 
366 	case IFLNK:
367 		/*
368 		 * Check for short symbolic link.
369 		 */
370 		if (dp->di_size > 0 &&
371 #ifdef FS_44INODEFMT
372 		    (dp->di_size < sblock->fs_maxsymlinklen ||
373 		     (sblock->fs_maxsymlinklen == 0 && dp->di_blocks == 0))) {
374 #else
375 		    dp->di_blocks == 0) {
376 #endif
377 			spcl.c_addr[0] = 1;
378 			spcl.c_count = 1;
379 			writeheader(ino);
380 			memcpy(buf, dp->di_shortlink, (u_long)dp->di_size);
381 			buf[dp->di_size] = '\0';
382 			writerec(buf, 0);
383 			return;
384 		}
385 		/* fall through */
386 
387 	case IFDIR:
388 	case IFREG:
389 		if (dp->di_size > 0)
390 			break;
391 		/* fall through */
392 
393 	case IFIFO:
394 	case IFSOCK:
395 	case IFCHR:
396 	case IFBLK:
397 		writeheader(ino);
398 		return;
399 
400 	default:
401 		msg("Warning: undefined file type 0%o\n", dp->di_mode & IFMT);
402 		return;
403 	}
404 	if (dp->di_size > NDADDR * sblock->fs_bsize)
405 		cnt = NDADDR * sblock->fs_frag;
406 	else
407 		cnt = howmany(dp->di_size, sblock->fs_fsize);
408 	blksout(&dp->di_db[0], cnt, ino);
409 	if ((size = dp->di_size - NDADDR * sblock->fs_bsize) <= 0)
410 		return;
411 	for (ind_level = 0; ind_level < NIADDR; ind_level++) {
412 		dmpindir(ino, dp->di_ib[ind_level], ind_level, &size);
413 		if (size <= 0)
414 			return;
415 	}
416 }
417 
418 /*
419  * Read indirect blocks, and pass the data blocks to be dumped.
420  */
421 static void
422 dmpindir(ino, blk, ind_level, size)
423 	ino_t ino;
424 	daddr_t blk;
425 	int ind_level;
426 	fsizeT *size;
427 {
428 	int i, cnt;
429 	daddr_t idblk[MAXNINDIR];
430 
431 	if (blk != 0)
432 		bread(fsbtodb(sblock, blk), (char *)idblk, (int) sblock->fs_bsize);
433 	else
434 		memset(idblk, 0, (int)sblock->fs_bsize);
435 	if (ind_level <= 0) {
436 		if (*size < NINDIR(sblock) * sblock->fs_bsize)
437 			cnt = howmany(*size, sblock->fs_fsize);
438 		else
439 			cnt = NINDIR(sblock) * sblock->fs_frag;
440 		*size -= NINDIR(sblock) * sblock->fs_bsize;
441 		blksout(&idblk[0], cnt, ino);
442 		return;
443 	}
444 	ind_level--;
445 	for (i = 0; i < NINDIR(sblock); i++) {
446 		dmpindir(ino, idblk[i], ind_level, size);
447 		if (*size <= 0)
448 			return;
449 	}
450 }
451 
452 /*
453  * Collect up the data into tape record sized buffers and output them.
454  */
455 void
456 blksout(blkp, frags, ino)
457 	daddr_t *blkp;
458 	int frags;
459 	ino_t ino;
460 {
461 	register daddr_t *bp;
462 	int i, j, count, blks, tbperdb;
463 
464 	blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
465 	tbperdb = sblock->fs_bsize >> tp_bshift;
466 	for (i = 0; i < blks; i += TP_NINDIR) {
467 		if (i + TP_NINDIR > blks)
468 			count = blks;
469 		else
470 			count = i + TP_NINDIR;
471 		for (j = i; j < count; j++)
472 			if (blkp[j / tbperdb] != 0)
473 				spcl.c_addr[j - i] = 1;
474 			else
475 				spcl.c_addr[j - i] = 0;
476 		spcl.c_count = count - i;
477 		writeheader(ino);
478 		bp = &blkp[i / tbperdb];
479 		for (j = i; j < count; j += tbperdb, bp++)
480 			if (*bp != 0)
481 				if (j + tbperdb <= count)
482 					dumpblock(*bp, (int)sblock->fs_bsize);
483 				else
484 					dumpblock(*bp, (count - j) * TP_BSIZE);
485 		spcl.c_type = TS_ADDR;
486 	}
487 }
488 
489 /*
490  * Dump a map to the tape.
491  */
492 void
493 dumpmap(map, type, ino)
494 	char *map;
495 	int type;
496 	ino_t ino;
497 {
498 	register int i;
499 	char *cp;
500 
501 	spcl.c_type = type;
502 	spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE);
503 	writeheader(ino);
504 	for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE)
505 		writerec(cp, 0);
506 }
507 
508 /*
509  * Write a header record to the dump tape.
510  */
511 void
512 writeheader(ino)
513 	ino_t ino;
514 {
515 	register long sum, cnt, *lp;
516 
517 	spcl.c_inumber = ino;
518 	spcl.c_magic = NFS_MAGIC;
519 	spcl.c_checksum = 0;
520 	lp = (long *)&spcl;
521 	sum = 0;
522 	cnt = sizeof(union u_spcl) / (4 * sizeof(long));
523 	while (--cnt >= 0) {
524 		sum += *lp++;
525 		sum += *lp++;
526 		sum += *lp++;
527 		sum += *lp++;
528 	}
529 	spcl.c_checksum = CHECKSUM - sum;
530 	writerec((char *)&spcl, 1);
531 }
532 
533 struct dinode *
534 getino(inum)
535 	ino_t inum;
536 {
537 	static daddr_t minino, maxino;
538 	static struct dinode inoblock[MAXINOPB];
539 
540 	curino = inum;
541 	if (inum >= minino && inum < maxino)
542 		return (&inoblock[inum - minino]);
543 	bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), (char *)inoblock,
544 	    (int)sblock->fs_bsize);
545 	minino = inum - (inum % INOPB(sblock));
546 	maxino = minino + INOPB(sblock);
547 	return (&inoblock[inum - minino]);
548 }
549 
550 /*
551  * Read a chunk of data from the disk.
552  * Try to recover from hard errors by reading in sector sized pieces.
553  * Error recovery is attempted at most BREADEMAX times before seeking
554  * consent from the operator to continue.
555  */
556 int	breaderrors = 0;
557 #define	BREADEMAX 32
558 
559 void
560 bread(blkno, buf, size)
561 	daddr_t blkno;
562 	char *buf;
563 	int size;
564 {
565 	int cnt, i;
566 	extern int errno;
567 
568 loop:
569 	if (lseek(diskfd, ((off_t)blkno << dev_bshift), 0) < 0)
570 		msg("bread: lseek fails\n");
571 	if ((cnt = read(diskfd, buf, size)) == size)
572 		return;
573 	if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) {
574 		/*
575 		 * Trying to read the final fragment.
576 		 *
577 		 * NB - dump only works in TP_BSIZE blocks, hence
578 		 * rounds `dev_bsize' fragments up to TP_BSIZE pieces.
579 		 * It should be smarter about not actually trying to
580 		 * read more than it can get, but for the time being
581 		 * we punt and scale back the read only when it gets
582 		 * us into trouble. (mkm 9/25/83)
583 		 */
584 		size -= dev_bsize;
585 		goto loop;
586 	}
587 	if (cnt == -1)
588 		msg("read error from %s: %s: [block %d]: count=%d\n",
589 			disk, strerror(errno), blkno, size);
590 	else
591 		msg("short read error from %s: [block %d]: count=%d, got=%d\n",
592 			disk, blkno, size, cnt);
593 	if (++breaderrors > BREADEMAX) {
594 		msg("More than %d block read errors from %d\n",
595 			BREADEMAX, disk);
596 		broadcast("DUMP IS AILING!\n");
597 		msg("This is an unrecoverable error.\n");
598 		if (!query("Do you want to attempt to continue?")){
599 			dumpabort(0);
600 			/*NOTREACHED*/
601 		} else
602 			breaderrors = 0;
603 	}
604 	/*
605 	 * Zero buffer, then try to read each sector of buffer separately.
606 	 */
607 	memset(buf, 0, size);
608 	for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) {
609 		if (lseek(diskfd, ((off_t)blkno << dev_bshift), 0) < 0)
610 			msg("bread: lseek2 fails!\n");
611 		if ((cnt = read(diskfd, buf, (int)dev_bsize)) == dev_bsize)
612 			continue;
613 		if (cnt == -1) {
614 			msg("read error from %s: %s: [sector %d]: count=%d\n",
615 				disk, strerror(errno), blkno, dev_bsize);
616 			continue;
617 		}
618 		msg("short read error from %s: [sector %d]: count=%d, got=%d\n",
619 			disk, blkno, dev_bsize, cnt);
620 	}
621 }
622