2 * Copyright (c) 1980, 1988, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)traverse.c 8.7 (Berkeley) 6/15/95
34 * $FreeBSD: src/sbin/dump/traverse.c,v 1.10.2.6 2003/04/14 20:10:35 johan Exp $
35 * $DragonFly: src/sbin/dump/traverse.c,v 1.11 2005/04/13 15:21:36 joerg Exp $
38 #include <sys/param.h>
41 #include <sys/vnode.h>
44 #include <ufs/fsdir.h>
45 #include <ufs/inode.h>
47 #include <vfs/ufs/dir.h>
48 #include <vfs/ufs/dinode.h>
49 #include <vfs/ufs/fs.h>
52 #include <protocols/dumprestore.h>
64 #define HASDUMPEDFILE 0x1
65 #define HASSUBDIRS 0x2
68 typedef quad_t fsizeT;
73 static int dirindir(ino_t, daddr_t, int, long *, long *, int);
74 static void dmpindir(ino_t, daddr_t, int level, fsizeT *);
75 static int searchdir(ino_t, daddr_t, long, long, long *, int);
78 * This is an estimation of the number of TP_BSIZE blocks in the file.
79 * It estimates the number of blocks in files with holes by assuming
80 * that all of the blocks accounted for by di_blocks are data blocks
81 * (when some of the blocks are usually used for indirect pointers);
82 * hence the estimate may be high.
85 blockest(struct dinode *dp)
90 * dp->di_size is the size of the file in bytes.
91 * dp->di_blocks stores the number of sectors actually in the file.
92 * If there are more sectors than the size would indicate, this just
93 * means that there are indirect blocks in the file or unused
94 * sectors in the last file block; we can safely ignore these
95 * (blkest = sizeest below).
96 * If the file is bigger than the number of sectors would indicate,
97 * then the file has holes in it. In this case we must use the
98 * block count to estimate the number of data blocks used, but
99 * we use the actual size for estimating the number of indirect
100 * dump blocks (sizeest vs. blkest in the indirect block
103 blkest = howmany(dbtob(dp->di_blocks), TP_BSIZE);
104 sizeest = howmany(dp->di_size, TP_BSIZE);
105 if (blkest > sizeest)
107 if (dp->di_size > sblock->fs_bsize * NDADDR) {
108 /* calculate the number of indirect blocks on the dump tape */
110 howmany(sizeest - NDADDR * sblock->fs_bsize / TP_BSIZE,
116 /* Auxiliary macro to pick up files changed since previous dump. */
117 #define CHANGEDSINCE(dp, t) \
118 ((dp)->di_mtime >= (t) || (dp)->di_ctime >= (t))
120 /* The WANTTODUMP macro decides whether a file should be dumped. */
122 #define WANTTODUMP(dp) \
123 (CHANGEDSINCE(dp, spcl.c_ddate) && \
124 (nonodump || ((dp)->di_flags & UF_NODUMP) != UF_NODUMP))
126 #define WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate)
132 * Walk the inode list for a filesystem to find all allocated inodes
133 * that have been modified since the previous dump time. Also, find all
134 * the directories in the filesystem.
137 mapfiles(ino_t maxino, long *tape_size)
142 int anydirskipped = 0;
144 for (ino = ROOTINO; ino < maxino; ino++) {
146 if ((mode = (dp->di_mode & IFMT)) == 0)
149 * Everything must go in usedinomap so that a check
150 * for "in dumpdirmap but not in usedinomap" to detect
151 * dirs with nodump set has a chance of succeeding
152 * (this is used in mapdirs()).
154 SETINO(ino, usedinomap);
156 SETINO(ino, dumpdirmap);
157 if (WANTTODUMP(dp)) {
158 SETINO(ino, dumpinomap);
159 if (mode != IFREG && mode != IFDIR && mode != IFLNK)
162 *tape_size += blockest(dp);
166 if (!nonodump && (dp->di_flags & UF_NODUMP))
167 CLRINO(ino, usedinomap);
172 * Restore gets very upset if the root is not dumped,
173 * so ensure that it always is dumped.
175 SETINO(ROOTINO, dumpinomap);
176 return (anydirskipped);
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.
192 mapdirs(ino_t maxino, long *tape_size)
195 int i, isdir, nodump;
202 isdir = 0; /* XXX just to get gcc to shut up */
203 for (map = dumpdirmap, ino = 1; ino < maxino; ino++) {
204 if (((ino - 1) % NBBY) == 0) /* map is offset by 1 */
209 * If a directory has been removed from usedinomap, it
210 * either has the nodump flag set, or has inherited
211 * it. Although a directory can't be in dumpinomap if
212 * it isn't in usedinomap, we have to go through it to
213 * propagate the nodump flag.
215 nodump = !nonodump && (TSTINO(ino, usedinomap) == 0);
216 if ((isdir & 1) == 0 || (TSTINO(ino, dumpinomap) && !nodump))
219 di = *dp; /* inode buf may change in searchdir(). */
220 filesize = di.di_size;
221 for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) {
222 if (di.di_db[i] != 0) {
223 ret |= searchdir(ino, di.di_db[i],
224 (long)dblksize(sblock, &di, i),
225 filesize, tape_size, nodump);
227 if (ret & HASDUMPEDFILE)
230 filesize -= sblock->fs_bsize;
232 for (i = 0; filesize > 0 && i < NIADDR; i++) {
233 if (di.di_ib[i] == 0)
235 ret |= dirindir(ino, di.di_ib[i], i, &filesize,
238 if (ret & HASDUMPEDFILE) {
239 SETINO(ino, dumpinomap);
240 *tape_size += blockest(&di);
245 if (ret & HASSUBDIRS)
246 change = 1; /* subdirs inherit nodump */
247 CLRINO(ino, dumpdirmap);
248 } else if ((ret & HASSUBDIRS) == 0)
249 if (!TSTINO(ino, dumpinomap)) {
250 CLRINO(ino, dumpdirmap);
258 * Read indirect blocks, and pass the data blocks to be searched
259 * as directories. Quit as soon as any entry is found that will
260 * require the directory to be dumped.
263 dirindir(ino_t ino, daddr_t blkno, int ind_level, long *filesize,
264 long *tape_size, int nodump)
268 daddr_t idblk[MAXNINDIR];
270 bread(fsbtodb(sblock, blkno), (char *)idblk, (int)sblock->fs_bsize);
271 if (ind_level <= 0) {
272 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
275 ret |= searchdir(ino, blkno, sblock->fs_bsize,
276 *filesize, tape_size, nodump);
278 if (ret & HASDUMPEDFILE)
281 *filesize -= sblock->fs_bsize;
286 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
289 ret |= dirindir(ino, blkno, ind_level, filesize,
297 * Scan a disk block containing directory information looking to see if
298 * any of the entries are on the dump list and to see if the directory
299 * contains any subdirectories.
302 searchdir(ino_t ino, daddr_t blkno, long size, long filesize,
303 long *tape_size, int nodump)
310 bread(fsbtodb(sblock, blkno), dblk, (int)size);
313 for (loc = 0; loc < size; ) {
314 dp = (struct direct *)(dblk + loc);
315 if (dp->d_reclen == 0) {
316 msg("corrupted directory, inumber %d\n", ino);
322 if (dp->d_name[0] == '.') {
323 if (dp->d_name[1] == '\0')
325 if (dp->d_name[1] == '.' && dp->d_name[2] == '\0')
329 ip = getino(dp->d_ino);
330 if (TSTINO(dp->d_ino, dumpinomap)) {
331 CLRINO(dp->d_ino, dumpinomap);
332 *tape_size -= blockest(ip);
335 * Add back to dumpdirmap and remove from usedinomap
336 * to propagate nodump.
338 if ((ip->di_mode & IFMT) == IFDIR) {
339 SETINO(dp->d_ino, dumpdirmap);
340 CLRINO(dp->d_ino, usedinomap);
344 if (TSTINO(dp->d_ino, dumpinomap)) {
345 ret |= HASDUMPEDFILE;
346 if (ret & HASSUBDIRS)
349 if (TSTINO(dp->d_ino, dumpdirmap)) {
351 if (ret & HASDUMPEDFILE)
360 * Dump passes 3 and 4.
362 * Dump the contents of an inode to tape.
365 dumpino(struct dinode *dp, ino_t ino)
373 dumpmap(dumpinomap, TS_BITS, ino);
375 CLRINO(ino, dumpinomap);
377 spcl.c_type = TS_INODE;
379 switch (dp->di_mode & S_IFMT) {
389 * Check for short symbolic link.
392 if (dp->di_size > 0 &&
393 dp->di_size < sblock->fs_maxsymlinklen) {
397 memmove(buf, dp->di_shortlink, (u_long)dp->di_size);
398 buf[dp->di_size] = '\0';
419 msg("Warning: undefined file type 0%o\n", dp->di_mode & IFMT);
422 if (dp->di_size > NDADDR * sblock->fs_bsize)
423 cnt = NDADDR * sblock->fs_frag;
425 cnt = howmany(dp->di_size, sblock->fs_fsize);
426 blksout(&dp->di_db[0], cnt, ino);
427 if ((size = dp->di_size - NDADDR * sblock->fs_bsize) <= 0)
429 for (ind_level = 0; ind_level < NIADDR; ind_level++) {
430 dmpindir(ino, dp->di_ib[ind_level], ind_level, &size);
437 * Read indirect blocks, and pass the data blocks to be dumped.
440 dmpindir(ino_t ino, daddr_t blk, int ind_level, fsizeT *size)
443 daddr_t idblk[MAXNINDIR];
446 bread(fsbtodb(sblock, blk), (char *)idblk, (int) sblock->fs_bsize);
448 memset(idblk, 0, (int)sblock->fs_bsize);
449 if (ind_level <= 0) {
450 if (*size < NINDIR(sblock) * sblock->fs_bsize)
451 cnt = howmany(*size, sblock->fs_fsize);
453 cnt = NINDIR(sblock) * sblock->fs_frag;
454 *size -= NINDIR(sblock) * sblock->fs_bsize;
455 blksout(&idblk[0], cnt, ino);
459 for (i = 0; i < NINDIR(sblock); i++) {
460 dmpindir(ino, idblk[i], ind_level, size);
467 * Collect up the data into tape record sized buffers and output them.
470 blksout(daddr_t *blkp, int frags, ino_t ino)
473 int i, j, count, blks, tbperdb;
475 blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
476 tbperdb = sblock->fs_bsize >> tp_bshift;
477 for (i = 0; i < blks; i += TP_NINDIR) {
478 if (i + TP_NINDIR > blks)
481 count = i + TP_NINDIR;
482 for (j = i; j < count; j++)
483 if (blkp[j / tbperdb] != 0)
484 spcl.c_addr[j - i] = 1;
486 spcl.c_addr[j - i] = 0;
487 spcl.c_count = count - i;
489 bp = &blkp[i / tbperdb];
490 for (j = i; j < count; j += tbperdb, bp++)
492 if (j + tbperdb <= count)
493 dumpblock(*bp, (int)sblock->fs_bsize);
495 dumpblock(*bp, (count - j) * TP_BSIZE);
497 spcl.c_type = TS_ADDR;
502 * Dump a map to the tape.
505 dumpmap(const char *map, int type, ino_t ino)
511 spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE);
513 for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE)
518 * Write a header record to the dump tape.
521 writeheader(ino_t ino)
523 int32_t sum, cnt, *lp;
525 spcl.c_inumber = ino;
526 spcl.c_magic = NFS_MAGIC;
528 lp = (int32_t *)&spcl;
530 cnt = sizeof(union u_spcl) / (4 * sizeof(int32_t));
537 spcl.c_checksum = CHECKSUM - sum;
544 static daddr_t minino, maxino;
545 static struct dinode inoblock[MAXINOPB];
548 if (inum >= minino && inum < maxino)
549 return (&inoblock[inum - minino]);
550 bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), (char *)inoblock,
551 (int)sblock->fs_bsize);
552 minino = inum - (inum % INOPB(sblock));
553 maxino = minino + INOPB(sblock);
554 return (&inoblock[inum - minino]);
558 * Read a chunk of data from the disk.
559 * Try to recover from hard errors by reading in sector sized pieces.
560 * Error recovery is attempted at most BREADEMAX times before seeking
561 * consent from the operator to continue.
567 bread(daddr_t blkno, char *buf, int size)
572 cnt = cread(diskfd, buf, size, ((off_t)blkno << dev_bshift));
575 if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) {
577 * Trying to read the final fragment.
579 * NB - dump only works in TP_BSIZE blocks, hence
580 * rounds `dev_bsize' fragments up to TP_BSIZE pieces.
581 * It should be smarter about not actually trying to
582 * read more than it can get, but for the time being
583 * we punt and scale back the read only when it gets
584 * us into trouble. (mkm 9/25/83)
590 msg("read error from %s: %s: [block %d]: count=%d\n",
591 disk, strerror(errno), blkno, size);
593 msg("short read error from %s: [block %d]: count=%d, got=%d\n",
594 disk, blkno, size, cnt);
595 if (++breaderrors > BREADEMAX) {
596 msg("More than %d block read errors from %s\n",
598 broadcast("DUMP IS AILING!\n");
599 msg("This is an unrecoverable error.\n");
600 if (!query("Do you want to attempt to continue?")){
607 * Zero buffer, then try to read each sector of buffer separately,
608 * and bypass the cache.
610 memset(buf, 0, size);
611 for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) {
612 if ((cnt = pread(diskfd, buf, (int)dev_bsize,
613 ((off_t)blkno << dev_bshift))) == dev_bsize)
616 msg("read error from %s: %s: [sector %d]: count=%ld\n",
617 disk, strerror(errno), blkno, dev_bsize);
620 msg("short read error from %s: [sector %d]: count=%ld, got=%d\n",
621 disk, blkno, dev_bsize, cnt);