xref: /netbsd-src/sbin/dump/traverse.c (revision 76dfffe33547c37f8bdd446e3e4ab0f3c16cea4b)
1 /*	$NetBSD: traverse.c,v 1.14 1995/06/18 21:35:33 cgd 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 1.14 1995/06/18 21:35:33 cgd 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 #define	CHANGEDSINCE(dp, t) \
125 	((dp)->di_mtime >= (t) || (dp)->di_ctime >= (t))
126 
127 /* The WANTTODUMP macro decides whether a file should be dumped. */
128 #ifdef UF_NODUMP
129 #define	WANTTODUMP(dp) \
130 	(CHANGEDSINCE(dp, spcl.c_ddate) && \
131 	 (nonodump || ((dp)->di_flags & UF_NODUMP) != UF_NODUMP))
132 #else
133 #define	WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate)
134 #endif
135 
136 /*
137  * Dump pass 1.
138  *
139  * Walk the inode list for a filesystem to find all allocated inodes
140  * that have been modified since the previous dump time. Also, find all
141  * the directories in the filesystem.
142  */
143 int
144 mapfiles(maxino, tapesize)
145 	ino_t maxino;
146 	long *tapesize;
147 {
148 	register int mode;
149 	register ino_t ino;
150 	register struct dinode *dp;
151 	int anydirskipped = 0;
152 
153 	for (ino = ROOTINO; ino < maxino; ino++) {
154 		dp = getino(ino);
155 		if ((mode = (dp->di_mode & IFMT)) == 0)
156 			continue;
157 		SETINO(ino, usedinomap);
158 		if (mode == IFDIR)
159 			SETINO(ino, dumpdirmap);
160 		if (WANTTODUMP(dp)) {
161 			SETINO(ino, dumpinomap);
162 			if (mode != IFREG && mode != IFDIR && mode != IFLNK)
163 				*tapesize += 1;
164 			else
165 				*tapesize += blockest(dp);
166 			continue;
167 		}
168 		if (mode == IFDIR)
169 			anydirskipped = 1;
170 	}
171 	/*
172 	 * Restore gets very upset if the root is not dumped,
173 	 * so ensure that it always is dumped.
174 	 */
175 	SETINO(ROOTINO, dumpinomap);
176 	return (anydirskipped);
177 }
178 
179 /*
180  * Dump pass 2.
181  *
182  * Scan each directory on the filesystem to see if it has any modified
183  * files in it. If it does, and has not already been added to the dump
184  * list (because it was itself modified), then add it. If a directory
185  * has not been modified itself, contains no modified files and has no
186  * subdirectories, then it can be deleted from the dump list and from
187  * the list of directories. By deleting it from the list of directories,
188  * its parent may now qualify for the same treatment on this or a later
189  * pass using this algorithm.
190  */
191 int
192 mapdirs(maxino, tapesize)
193 	ino_t maxino;
194 	long *tapesize;
195 {
196 	register struct	dinode *dp;
197 	register int i, isdir;
198 	register char *map;
199 	register ino_t ino;
200 	long filesize;
201 	int ret, change = 0;
202 
203 	isdir = 0;		/* XXX just to get gcc to shut up */
204 	for (map = dumpdirmap, ino = 1; ino < maxino; ino++) {
205 		if (((ino - 1) % NBBY) == 0)	/* map is offset by 1 */
206 			isdir = *map++;
207 		else
208 			isdir >>= 1;
209 		if ((isdir & 1) == 0 || TSTINO(ino, dumpinomap))
210 			continue;
211 		dp = getino(ino);
212 		filesize = dp->di_size;
213 		for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) {
214 			if (dp->di_db[i] != 0)
215 				ret |= searchdir(ino, dp->di_db[i],
216 					(long)dblksize(sblock, dp, i),
217 					filesize);
218 			if (ret & HASDUMPEDFILE)
219 				filesize = 0;
220 			else
221 				filesize -= sblock->fs_bsize;
222 		}
223 		for (i = 0; filesize > 0 && i < NIADDR; i++) {
224 			if (dp->di_ib[i] == 0)
225 				continue;
226 			ret |= dirindir(ino, dp->di_ib[i], i, &filesize);
227 		}
228 		if (ret & HASDUMPEDFILE) {
229 			SETINO(ino, dumpinomap);
230 			*tapesize += blockest(dp);
231 			change = 1;
232 			continue;
233 		}
234 		if ((ret & HASSUBDIRS) == 0) {
235 			if (!TSTINO(ino, dumpinomap)) {
236 				CLRINO(ino, dumpdirmap);
237 				change = 1;
238 			}
239 		}
240 	}
241 	return (change);
242 }
243 
244 /*
245  * Read indirect blocks, and pass the data blocks to be searched
246  * as directories. Quit as soon as any entry is found that will
247  * require the directory to be dumped.
248  */
249 static int
250 dirindir(ino, blkno, ind_level, filesize)
251 	ino_t ino;
252 	daddr_t blkno;
253 	int ind_level;
254 	long *filesize;
255 {
256 	int ret = 0;
257 	register int i;
258 	daddr_t	idblk[MAXNINDIR];
259 
260 	bread(fsbtodb(sblock, blkno), (char *)idblk, (int)sblock->fs_bsize);
261 	if (ind_level <= 0) {
262 		for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
263 			blkno = idblk[i];
264 			if (blkno != 0)
265 				ret |= searchdir(ino, blkno, sblock->fs_bsize,
266 					*filesize);
267 			if (ret & HASDUMPEDFILE)
268 				*filesize = 0;
269 			else
270 				*filesize -= sblock->fs_bsize;
271 		}
272 		return (ret);
273 	}
274 	ind_level--;
275 	for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
276 		blkno = idblk[i];
277 		if (blkno != 0)
278 			ret |= dirindir(ino, blkno, ind_level, filesize);
279 	}
280 	return (ret);
281 }
282 
283 /*
284  * Scan a disk block containing directory information looking to see if
285  * any of the entries are on the dump list and to see if the directory
286  * contains any subdirectories.
287  */
288 static int
289 searchdir(ino, blkno, size, filesize)
290 	ino_t ino;
291 	daddr_t blkno;
292 	register long size;
293 	long filesize;
294 {
295 	register struct direct *dp;
296 	register long loc, ret = 0;
297 	char dblk[MAXBSIZE];
298 
299 	bread(fsbtodb(sblock, blkno), dblk, (int)size);
300 	if (filesize < size)
301 		size = filesize;
302 	for (loc = 0; loc < size; ) {
303 		dp = (struct direct *)(dblk + loc);
304 		if (dp->d_reclen == 0) {
305 			msg("corrupted directory, inumber %d\n", ino);
306 			break;
307 		}
308 		loc += dp->d_reclen;
309 		if (dp->d_ino == 0)
310 			continue;
311 		if (dp->d_name[0] == '.') {
312 			if (dp->d_name[1] == '\0')
313 				continue;
314 			if (dp->d_name[1] == '.' && dp->d_name[2] == '\0')
315 				continue;
316 		}
317 		if (TSTINO(dp->d_ino, dumpinomap)) {
318 			ret |= HASDUMPEDFILE;
319 			if (ret & HASSUBDIRS)
320 				break;
321 		}
322 		if (TSTINO(dp->d_ino, dumpdirmap)) {
323 			ret |= HASSUBDIRS;
324 			if (ret & HASDUMPEDFILE)
325 				break;
326 		}
327 	}
328 	return (ret);
329 }
330 
331 /*
332  * Dump passes 3 and 4.
333  *
334  * Dump the contents of an inode to tape.
335  */
336 void
337 dumpino(dp, ino)
338 	register struct dinode *dp;
339 	ino_t ino;
340 {
341 	int ind_level, cnt;
342 	fsizeT size;
343 	char buf[TP_BSIZE];
344 
345 	if (newtape) {
346 		newtape = 0;
347 		dumpmap(dumpinomap, TS_BITS, ino);
348 	}
349 	CLRINO(ino, dumpinomap);
350 	spcl.c_dinode = *dp;
351 	spcl.c_type = TS_INODE;
352 	spcl.c_count = 0;
353 	switch (dp->di_mode & IFMT) {
354 
355 	case 0:
356 		/*
357 		 * Freed inode.
358 		 */
359 		return;
360 
361 	case IFLNK:
362 		/*
363 		 * Check for short symbolic link.
364 		 */
365 		if (dp->di_size > 0 &&
366 #ifdef FS_44INODEFMT
367 		    (dp->di_size < sblock->fs_maxsymlinklen ||
368 		     (sblock->fs_maxsymlinklen == 0 && dp->di_blocks == 0))) {
369 #else
370 		    dp->di_blocks == 0) {
371 #endif
372 			spcl.c_addr[0] = 1;
373 			spcl.c_count = 1;
374 			writeheader(ino);
375 			memcpy(buf, dp->di_shortlink, (u_long)dp->di_size);
376 			buf[dp->di_size] = '\0';
377 			writerec(buf, 0);
378 			return;
379 		}
380 		/* fall through */
381 
382 	case IFDIR:
383 	case IFREG:
384 		if (dp->di_size > 0)
385 			break;
386 		/* fall through */
387 
388 	case IFIFO:
389 	case IFSOCK:
390 	case IFCHR:
391 	case IFBLK:
392 		writeheader(ino);
393 		return;
394 
395 	default:
396 		msg("Warning: undefined file type 0%o\n", dp->di_mode & IFMT);
397 		return;
398 	}
399 	if (dp->di_size > NDADDR * sblock->fs_bsize)
400 		cnt = NDADDR * sblock->fs_frag;
401 	else
402 		cnt = howmany(dp->di_size, sblock->fs_fsize);
403 	blksout(&dp->di_db[0], cnt, ino);
404 	if ((size = dp->di_size - NDADDR * sblock->fs_bsize) <= 0)
405 		return;
406 	for (ind_level = 0; ind_level < NIADDR; ind_level++) {
407 		dmpindir(ino, dp->di_ib[ind_level], ind_level, &size);
408 		if (size <= 0)
409 			return;
410 	}
411 }
412 
413 /*
414  * Read indirect blocks, and pass the data blocks to be dumped.
415  */
416 static void
417 dmpindir(ino, blk, ind_level, size)
418 	ino_t ino;
419 	daddr_t blk;
420 	int ind_level;
421 	fsizeT *size;
422 {
423 	int i, cnt;
424 	daddr_t idblk[MAXNINDIR];
425 
426 	if (blk != 0)
427 		bread(fsbtodb(sblock, blk), (char *)idblk, (int) sblock->fs_bsize);
428 	else
429 		memset(idblk, 0, (int)sblock->fs_bsize);
430 	if (ind_level <= 0) {
431 		if (*size < NINDIR(sblock) * sblock->fs_bsize)
432 			cnt = howmany(*size, sblock->fs_fsize);
433 		else
434 			cnt = NINDIR(sblock) * sblock->fs_frag;
435 		*size -= NINDIR(sblock) * sblock->fs_bsize;
436 		blksout(&idblk[0], cnt, ino);
437 		return;
438 	}
439 	ind_level--;
440 	for (i = 0; i < NINDIR(sblock); i++) {
441 		dmpindir(ino, idblk[i], ind_level, size);
442 		if (*size <= 0)
443 			return;
444 	}
445 }
446 
447 /*
448  * Collect up the data into tape record sized buffers and output them.
449  */
450 void
451 blksout(blkp, frags, ino)
452 	daddr_t *blkp;
453 	int frags;
454 	ino_t ino;
455 {
456 	register daddr_t *bp;
457 	int i, j, count, blks, tbperdb;
458 
459 	blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
460 	tbperdb = sblock->fs_bsize >> tp_bshift;
461 	for (i = 0; i < blks; i += TP_NINDIR) {
462 		if (i + TP_NINDIR > blks)
463 			count = blks;
464 		else
465 			count = i + TP_NINDIR;
466 		for (j = i; j < count; j++)
467 			if (blkp[j / tbperdb] != 0)
468 				spcl.c_addr[j - i] = 1;
469 			else
470 				spcl.c_addr[j - i] = 0;
471 		spcl.c_count = count - i;
472 		writeheader(ino);
473 		bp = &blkp[i / tbperdb];
474 		for (j = i; j < count; j += tbperdb, bp++)
475 			if (*bp != 0)
476 				if (j + tbperdb <= count)
477 					dumpblock(*bp, (int)sblock->fs_bsize);
478 				else
479 					dumpblock(*bp, (count - j) * TP_BSIZE);
480 		spcl.c_type = TS_ADDR;
481 	}
482 }
483 
484 /*
485  * Dump a map to the tape.
486  */
487 void
488 dumpmap(map, type, ino)
489 	char *map;
490 	int type;
491 	ino_t ino;
492 {
493 	register int i;
494 	char *cp;
495 
496 	spcl.c_type = type;
497 	spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE);
498 	writeheader(ino);
499 	for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE)
500 		writerec(cp, 0);
501 }
502 
503 /*
504  * Write a header record to the dump tape.
505  */
506 void
507 writeheader(ino)
508 	ino_t ino;
509 {
510 	register long sum, cnt, *lp;
511 
512 	spcl.c_inumber = ino;
513 	spcl.c_magic = NFS_MAGIC;
514 	spcl.c_checksum = 0;
515 	lp = (long *)&spcl;
516 	sum = 0;
517 	cnt = sizeof(union u_spcl) / (4 * sizeof(long));
518 	while (--cnt >= 0) {
519 		sum += *lp++;
520 		sum += *lp++;
521 		sum += *lp++;
522 		sum += *lp++;
523 	}
524 	spcl.c_checksum = CHECKSUM - sum;
525 	writerec((char *)&spcl, 1);
526 }
527 
528 struct dinode *
529 getino(inum)
530 	ino_t inum;
531 {
532 	static daddr_t minino, maxino;
533 	static struct dinode inoblock[MAXINOPB];
534 
535 	curino = inum;
536 	if (inum >= minino && inum < maxino)
537 		return (&inoblock[inum - minino]);
538 	bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), (char *)inoblock,
539 	    (int)sblock->fs_bsize);
540 	minino = inum - (inum % INOPB(sblock));
541 	maxino = minino + INOPB(sblock);
542 	return (&inoblock[inum - minino]);
543 }
544 
545 /*
546  * Read a chunk of data from the disk.
547  * Try to recover from hard errors by reading in sector sized pieces.
548  * Error recovery is attempted at most BREADEMAX times before seeking
549  * consent from the operator to continue.
550  */
551 int	breaderrors = 0;
552 #define	BREADEMAX 32
553 
554 void
555 bread(blkno, buf, size)
556 	daddr_t blkno;
557 	char *buf;
558 	int size;
559 {
560 	int cnt, i;
561 	extern int errno;
562 
563 loop:
564 	if (lseek(diskfd, ((off_t)blkno << dev_bshift), 0) < 0)
565 		msg("bread: lseek fails\n");
566 	if ((cnt = read(diskfd, buf, size)) == size)
567 		return;
568 	if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) {
569 		/*
570 		 * Trying to read the final fragment.
571 		 *
572 		 * NB - dump only works in TP_BSIZE blocks, hence
573 		 * rounds `dev_bsize' fragments up to TP_BSIZE pieces.
574 		 * It should be smarter about not actually trying to
575 		 * read more than it can get, but for the time being
576 		 * we punt and scale back the read only when it gets
577 		 * us into trouble. (mkm 9/25/83)
578 		 */
579 		size -= dev_bsize;
580 		goto loop;
581 	}
582 	if (cnt == -1)
583 		msg("read error from %s: %s: [block %d]: count=%d\n",
584 			disk, strerror(errno), blkno, size);
585 	else
586 		msg("short read error from %s: [block %d]: count=%d, got=%d\n",
587 			disk, blkno, size, cnt);
588 	if (++breaderrors > BREADEMAX) {
589 		msg("More than %d block read errors from %d\n",
590 			BREADEMAX, disk);
591 		broadcast("DUMP IS AILING!\n");
592 		msg("This is an unrecoverable error.\n");
593 		if (!query("Do you want to attempt to continue?")){
594 			dumpabort(0);
595 			/*NOTREACHED*/
596 		} else
597 			breaderrors = 0;
598 	}
599 	/*
600 	 * Zero buffer, then try to read each sector of buffer separately.
601 	 */
602 	memset(buf, 0, size);
603 	for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) {
604 		if (lseek(diskfd, ((off_t)blkno << dev_bshift), 0) < 0)
605 			msg("bread: lseek2 fails!\n");
606 		if ((cnt = read(diskfd, buf, (int)dev_bsize)) == dev_bsize)
607 			continue;
608 		if (cnt == -1) {
609 			msg("read error from %s: %s: [sector %d]: count=%d\n",
610 				disk, strerror(errno), blkno, dev_bsize);
611 			continue;
612 		}
613 		msg("short read error from %s: [sector %d]: count=%d, got=%d\n",
614 			disk, blkno, dev_bsize, cnt);
615 	}
616 }
617