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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 static char sccsid[] = "@(#)traverse.c 8.7 (Berkeley) 6/15/95";
34 static const char rcsid[] =
38 #include <sys/param.h>
41 #include <ufs/ufs/dir.h>
42 #include <ufs/ufs/dinode.h>
43 #include <ufs/ffs/fs.h>
45 #include <protocols/dumprestore.h>
60 struct ufs1_dinode dp1;
61 struct ufs2_dinode dp2;
63 #define DIP(dp, field) \
64 ((sblock->fs_magic == FS_UFS1_MAGIC) ? \
65 (dp)->dp1.field : (dp)->dp2.field)
66 #define DIP_SET(dp, field, val) do {\
67 if (sblock->fs_magic == FS_UFS1_MAGIC) \
68 (dp)->dp1.field = (val); \
70 (dp)->dp2.field = (val); \
73 #define HASDUMPEDFILE 0x1
74 #define HASSUBDIRS 0x2
76 static int dirindir(ino_t ino, ufs2_daddr_t blkno, int level, long *size,
77 long *tapesize, int nodump);
78 static void dmpindir(ino_t ino, ufs2_daddr_t blk, int level, off_t *size);
79 static int searchdir(ino_t ino, ufs2_daddr_t blkno, long size, long filesize,
80 long *tapesize, int nodump);
81 static long blockest(union dinode *dp);
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.
91 blockest(union dinode *dp)
96 * dp->di_size is the size of the file in bytes.
97 * dp->di_blocks stores the number of sectors actually in the file.
98 * If there are more sectors than the size would indicate, this just
99 * means that there are indirect blocks in the file or unused
100 * sectors in the last file block; we can safely ignore these
101 * (blkest = sizeest below).
102 * If the file is bigger than the number of sectors would indicate,
103 * then the file has holes in it. In this case we must use the
104 * block count to estimate the number of data blocks used, but
105 * we use the actual size for estimating the number of indirect
106 * dump blocks (sizeest vs. blkest in the indirect block
109 if ((DIP(dp, di_flags) & SF_SNAPSHOT) != 0)
111 blkest = howmany(dbtob(DIP(dp, di_blocks)), TP_BSIZE);
112 sizeest = howmany(DIP(dp, di_size), TP_BSIZE);
113 if (blkest > sizeest)
115 if (DIP(dp, di_size) > sblock->fs_bsize * NDADDR) {
116 /* calculate the number of indirect blocks on the dump tape */
118 howmany(sizeest - NDADDR * sblock->fs_bsize / TP_BSIZE,
124 /* Auxiliary macro to pick up files changed since previous dump. */
125 #define CHANGEDSINCE(dp, t) \
126 (DIP(dp, di_mtime) >= (t) || DIP(dp, di_ctime) >= (t))
128 /* The WANTTODUMP macro decides whether a file should be dumped. */
130 #define WANTTODUMP(dp) \
131 (CHANGEDSINCE(dp, spcl.c_ddate) && \
132 (nonodump || (DIP(dp, di_flags) & UF_NODUMP) != UF_NODUMP))
134 #define WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate)
140 * Walk the inode list for a file system to find all allocated inodes
141 * that have been modified since the previous dump time. Also, find all
142 * the directories in the file system.
145 mapfiles(ino_t maxino, long *tapesize)
147 int i, cg, mode, inosused;
148 int anydirskipped = 0;
154 if ((cgp = malloc(sblock->fs_cgsize)) == NULL)
155 quit("mapfiles: cannot allocate memory.\n");
156 for (cg = 0; cg < sblock->fs_ncg; cg++) {
157 ino = cg * sblock->fs_ipg;
158 bread(fsbtodb(sblock, cgtod(sblock, cg)), (char *)cgp,
160 if (sblock->fs_magic == FS_UFS2_MAGIC)
161 inosused = cgp->cg_initediblk;
163 inosused = sblock->fs_ipg;
165 * If we are using soft updates, then we can trust the
166 * cylinder group inode allocation maps to tell us which
167 * inodes are allocated. We will scan the used inode map
168 * to find the inodes that are really in use, and then
169 * read only those inodes in from disk.
171 if (sblock->fs_flags & FS_DOSOFTDEP) {
172 if (!cg_chkmagic(cgp))
173 quit("mapfiles: cg %d: bad magic number\n", cg);
174 cp = &cg_inosused(cgp)[(inosused - 1) / CHAR_BIT];
175 for ( ; inosused > 0; inosused -= CHAR_BIT, cp--) {
178 for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
188 for (i = 0; i < inosused; i++, ino++) {
190 (dp = getino(ino, &mode)) == NULL ||
194 * Everything must go in usedinomap so that a check
195 * for "in dumpdirmap but not in usedinomap" to detect
196 * dirs with nodump set has a chance of succeeding
197 * (this is used in mapdirs()).
199 SETINO(ino, usedinomap);
201 SETINO(ino, dumpdirmap);
202 if (WANTTODUMP(dp)) {
203 SETINO(ino, dumpinomap);
209 *tapesize += blockest(dp);
214 (DIP(dp, di_flags) & UF_NODUMP))
215 CLRINO(ino, usedinomap);
221 * Restore gets very upset if the root is not dumped,
222 * so ensure that it always is dumped.
224 SETINO(ROOTINO, dumpinomap);
225 return (anydirskipped);
231 * Scan each directory on the file system to see if it has any modified
232 * files in it. If it does, and has not already been added to the dump
233 * list (because it was itself modified), then add it. If a directory
234 * has not been modified itself, contains no modified files and has no
235 * subdirectories, then it can be deleted from the dump list and from
236 * the list of directories. By deleting it from the list of directories,
237 * its parent may now qualify for the same treatment on this or a later
238 * pass using this algorithm.
241 mapdirs(ino_t maxino, long *tapesize)
244 int i, isdir, nodump;
251 isdir = 0; /* XXX just to get gcc to shut up */
252 for (map = dumpdirmap, ino = 1; ino < maxino; ino++) {
253 if (((ino - 1) % CHAR_BIT) == 0) /* map is offset by 1 */
258 * If a directory has been removed from usedinomap, it
259 * either has the nodump flag set, or has inherited
260 * it. Although a directory can't be in dumpinomap if
261 * it isn't in usedinomap, we have to go through it to
262 * propagate the nodump flag.
264 nodump = !nonodump && (TSTINO(ino, usedinomap) == 0);
265 if ((isdir & 1) == 0 || (TSTINO(ino, dumpinomap) && !nodump))
267 dp = getino(ino, &i);
269 * inode buf may change in searchdir().
271 if (sblock->fs_magic == FS_UFS1_MAGIC)
275 filesize = DIP(&di, di_size);
276 for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) {
277 if (DIP(&di, di_db[i]) != 0)
278 ret |= searchdir(ino, DIP(&di, di_db[i]),
279 (long)sblksize(sblock, DIP(&di, di_size),
280 i), filesize, tapesize, nodump);
281 if (ret & HASDUMPEDFILE)
284 filesize -= sblock->fs_bsize;
286 for (i = 0; filesize > 0 && i < NIADDR; i++) {
287 if (DIP(&di, di_ib[i]) == 0)
289 ret |= dirindir(ino, DIP(&di, di_ib[i]), i, &filesize,
292 if (ret & HASDUMPEDFILE) {
293 SETINO(ino, dumpinomap);
294 *tapesize += blockest(&di);
299 if (ret & HASSUBDIRS)
300 change = 1; /* subdirs inherit nodump */
301 CLRINO(ino, dumpdirmap);
302 } else if ((ret & HASSUBDIRS) == 0)
303 if (!TSTINO(ino, dumpinomap)) {
304 CLRINO(ino, dumpdirmap);
312 * Read indirect blocks, and pass the data blocks to be searched
313 * as directories. Quit as soon as any entry is found that will
314 * require the directory to be dumped.
326 ufs1_daddr_t ufs1[MAXBSIZE / sizeof(ufs1_daddr_t)];
327 ufs2_daddr_t ufs2[MAXBSIZE / sizeof(ufs2_daddr_t)];
332 bread(fsbtodb(sblock, blkno), (char *)&idblk, (int)sblock->fs_bsize);
333 if (ind_level <= 0) {
334 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
335 if (sblock->fs_magic == FS_UFS1_MAGIC)
336 blkno = idblk.ufs1[i];
338 blkno = idblk.ufs2[i];
340 ret |= searchdir(ino, blkno, sblock->fs_bsize,
341 *filesize, tapesize, nodump);
342 if (ret & HASDUMPEDFILE)
345 *filesize -= sblock->fs_bsize;
350 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
351 if (sblock->fs_magic == FS_UFS1_MAGIC)
352 blkno = idblk.ufs1[i];
354 blkno = idblk.ufs2[i];
356 ret |= dirindir(ino, blkno, ind_level, filesize,
363 * Scan a disk block containing directory information looking to see if
364 * any of the entries are on the dump list and to see if the directory
365 * contains any subdirectories.
382 if (dblk == NULL && (dblk = malloc(sblock->fs_bsize)) == NULL)
383 quit("searchdir: cannot allocate indirect memory.\n");
384 bread(fsbtodb(sblock, blkno), dblk, (int)size);
387 for (loc = 0; loc < size; ) {
388 dp = (struct direct *)(dblk + loc);
389 if (dp->d_reclen == 0) {
390 msg("corrupted directory, inumber %d\n", ino);
396 if (dp->d_name[0] == '.') {
397 if (dp->d_name[1] == '\0')
399 if (dp->d_name[1] == '.' && dp->d_name[2] == '\0')
403 ip = getino(dp->d_ino, &mode);
404 if (TSTINO(dp->d_ino, dumpinomap)) {
405 CLRINO(dp->d_ino, dumpinomap);
406 *tapesize -= blockest(ip);
409 * Add back to dumpdirmap and remove from usedinomap
410 * to propagate nodump.
413 SETINO(dp->d_ino, dumpdirmap);
414 CLRINO(dp->d_ino, usedinomap);
418 if (TSTINO(dp->d_ino, dumpinomap)) {
419 ret |= HASDUMPEDFILE;
420 if (ret & HASSUBDIRS)
423 if (TSTINO(dp->d_ino, dumpdirmap)) {
425 if (ret & HASDUMPEDFILE)
434 * Dump passes 3 and 4.
436 * Dump the contents of an inode to tape.
439 dumpino(union dinode *dp, ino_t ino)
447 dumpmap(dumpinomap, TS_BITS, ino);
449 CLRINO(ino, dumpinomap);
451 * Zero out the size of a snapshot so that it will be dumped
452 * as a zero length file.
454 if ((DIP(dp, di_flags) & SF_SNAPSHOT) != 0) {
455 DIP_SET(dp, di_size, 0);
456 DIP_SET(dp, di_flags, DIP(dp, di_flags) & ~SF_SNAPSHOT);
458 if (sblock->fs_magic == FS_UFS1_MAGIC) {
459 spcl.c_mode = dp->dp1.di_mode;
460 spcl.c_size = dp->dp1.di_size;
461 spcl.c_atime = _time32_to_time(dp->dp1.di_atime);
462 spcl.c_atimensec = dp->dp1.di_atimensec;
463 spcl.c_mtime = _time32_to_time(dp->dp1.di_mtime);
464 spcl.c_mtimensec = dp->dp1.di_mtimensec;
465 spcl.c_birthtime = 0;
466 spcl.c_birthtimensec = 0;
467 spcl.c_rdev = dp->dp1.di_rdev;
468 spcl.c_file_flags = dp->dp1.di_flags;
469 spcl.c_uid = dp->dp1.di_uid;
470 spcl.c_gid = dp->dp1.di_gid;
472 spcl.c_mode = dp->dp2.di_mode;
473 spcl.c_size = dp->dp2.di_size;
474 spcl.c_atime = _time64_to_time(dp->dp2.di_atime);
475 spcl.c_atimensec = dp->dp2.di_atimensec;
476 spcl.c_mtime = _time64_to_time(dp->dp2.di_mtime);
477 spcl.c_mtimensec = dp->dp2.di_mtimensec;
478 spcl.c_birthtime = _time64_to_time(dp->dp2.di_birthtime);
479 spcl.c_birthtimensec = dp->dp2.di_birthnsec;
480 spcl.c_rdev = dp->dp2.di_rdev;
481 spcl.c_file_flags = dp->dp2.di_flags;
482 spcl.c_uid = dp->dp2.di_uid;
483 spcl.c_gid = dp->dp2.di_gid;
485 spcl.c_type = TS_INODE;
487 switch (DIP(dp, di_mode) & S_IFMT) {
497 * Check for short symbolic link.
499 if (DIP(dp, di_size) > 0 &&
500 DIP(dp, di_size) < sblock->fs_maxsymlinklen) {
504 if (sblock->fs_magic == FS_UFS1_MAGIC)
505 memmove(buf, (caddr_t)dp->dp1.di_db,
506 (u_long)DIP(dp, di_size));
508 memmove(buf, (caddr_t)dp->dp2.di_db,
509 (u_long)DIP(dp, di_size));
510 buf[DIP(dp, di_size)] = '\0';
518 if (DIP(dp, di_size) > 0)
530 msg("Warning: undefined file type 0%o\n",
531 DIP(dp, di_mode) & IFMT);
534 if (DIP(dp, di_size) > NDADDR * sblock->fs_bsize)
535 cnt = NDADDR * sblock->fs_frag;
537 cnt = howmany(DIP(dp, di_size), sblock->fs_fsize);
538 if (sblock->fs_magic == FS_UFS1_MAGIC)
539 ufs1_blksout(&dp->dp1.di_db[0], cnt, ino);
541 ufs2_blksout(&dp->dp2.di_db[0], cnt, ino);
542 if ((size = DIP(dp, di_size) - NDADDR * sblock->fs_bsize) <= 0)
544 for (ind_level = 0; ind_level < NIADDR; ind_level++) {
545 dmpindir(ino, DIP(dp, di_ib[ind_level]), ind_level, &size);
552 * Read indirect blocks, and pass the data blocks to be dumped.
555 dmpindir(ino_t ino, ufs2_daddr_t blk, int ind_level, off_t *size)
558 ufs1_daddr_t ufs1[MAXBSIZE / sizeof(ufs1_daddr_t)];
559 ufs2_daddr_t ufs2[MAXBSIZE / sizeof(ufs2_daddr_t)];
564 bread(fsbtodb(sblock, blk), (char *)&idblk,
565 (int)sblock->fs_bsize);
567 memset(&idblk, 0, sblock->fs_bsize);
568 if (ind_level <= 0) {
569 if (*size < NINDIR(sblock) * sblock->fs_bsize)
570 cnt = howmany(*size, sblock->fs_fsize);
572 cnt = NINDIR(sblock) * sblock->fs_frag;
573 *size -= NINDIR(sblock) * sblock->fs_bsize;
574 if (sblock->fs_magic == FS_UFS1_MAGIC)
575 ufs1_blksout(idblk.ufs1, cnt, ino);
577 ufs2_blksout(idblk.ufs2, cnt, ino);
581 for (i = 0; i < NINDIR(sblock); i++) {
582 if (sblock->fs_magic == FS_UFS1_MAGIC)
583 dmpindir(ino, idblk.ufs1[i], ind_level, size);
585 dmpindir(ino, idblk.ufs2[i], ind_level, size);
592 * Collect up the data into tape record sized buffers and output them.
595 ufs1_blksout(ufs1_daddr_t *blkp, int frags, ino_t ino)
598 int i, j, count, blks, tbperdb;
600 blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
601 tbperdb = sblock->fs_bsize >> tp_bshift;
602 for (i = 0; i < blks; i += TP_NINDIR) {
603 if (i + TP_NINDIR > blks)
606 count = i + TP_NINDIR;
607 for (j = i; j < count; j++)
608 if (blkp[j / tbperdb] != 0)
609 spcl.c_addr[j - i] = 1;
611 spcl.c_addr[j - i] = 0;
612 spcl.c_count = count - i;
614 bp = &blkp[i / tbperdb];
615 for (j = i; j < count; j += tbperdb, bp++)
617 if (j + tbperdb <= count)
618 dumpblock(*bp, (int)sblock->fs_bsize);
620 dumpblock(*bp, (count - j) * TP_BSIZE);
622 spcl.c_type = TS_ADDR;
627 * Collect up the data into tape record sized buffers and output them.
630 ufs2_blksout(ufs2_daddr_t *blkp, int frags, ino_t ino)
633 int i, j, count, blks, tbperdb;
635 blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
636 tbperdb = sblock->fs_bsize >> tp_bshift;
637 for (i = 0; i < blks; i += TP_NINDIR) {
638 if (i + TP_NINDIR > blks)
641 count = i + TP_NINDIR;
642 for (j = i; j < count; j++)
643 if (blkp[j / tbperdb] != 0)
644 spcl.c_addr[j - i] = 1;
646 spcl.c_addr[j - i] = 0;
647 spcl.c_count = count - i;
649 bp = &blkp[i / tbperdb];
650 for (j = i; j < count; j += tbperdb, bp++)
652 if (j + tbperdb <= count)
653 dumpblock(*bp, (int)sblock->fs_bsize);
655 dumpblock(*bp, (count - j) * TP_BSIZE);
657 spcl.c_type = TS_ADDR;
662 * Dump a map to the tape.
665 dumpmap(char *map, int type, ino_t ino)
671 spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE);
673 for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE)
678 * Write a header record to the dump tape.
681 writeheader(ino_t ino)
683 int32_t sum, cnt, *lp;
685 spcl.c_inumber = ino;
686 spcl.c_magic = FS_UFS2_MAGIC;
688 lp = (int32_t *)&spcl;
690 cnt = sizeof(union u_spcl) / (4 * sizeof(int32_t));
697 spcl.c_checksum = CHECKSUM - sum;
698 writerec((char *)&spcl, 1);
702 getino(ino_t inum, int *modep)
704 static ino_t minino, maxino;
705 static caddr_t inoblock;
706 struct ufs1_dinode *dp1;
707 struct ufs2_dinode *dp2;
709 if (inoblock == NULL && (inoblock = malloc(sblock->fs_bsize)) == NULL)
710 quit("cannot allocate inode memory.\n");
712 if (inum >= minino && inum < maxino)
714 bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), inoblock,
715 (int)sblock->fs_bsize);
716 minino = inum - (inum % INOPB(sblock));
717 maxino = minino + INOPB(sblock);
719 if (sblock->fs_magic == FS_UFS1_MAGIC) {
720 dp1 = &((struct ufs1_dinode *)inoblock)[inum - minino];
721 *modep = (dp1->di_mode & IFMT);
722 return ((union dinode *)dp1);
724 dp2 = &((struct ufs2_dinode *)inoblock)[inum - minino];
725 *modep = (dp2->di_mode & IFMT);
726 return ((union dinode *)dp2);
730 * Read a chunk of data from the disk.
731 * Try to recover from hard errors by reading in sector sized pieces.
732 * Error recovery is attempted at most BREADEMAX times before seeking
733 * consent from the operator to continue.
739 bread(ufs2_daddr_t blkno, char *buf, int size)
741 int secsize, bytes, resid, xfer, base, cnt, i;
746 offset = blkno << dev_bshift;
747 secsize = sblock->fs_fsize;
748 base = offset % secsize;
749 resid = size % secsize;
751 * If the transfer request starts or ends on a non-sector
752 * boundary, we must read the entire sector and copy out
753 * just the part that we need.
755 if (base == 0 && resid == 0) {
756 cnt = cread(diskfd, buf, size, offset);
760 if (tmpbuf == NULL && (tmpbuf = malloc(secsize)) == 0)
761 quit("buffer malloc failed\n");
765 cnt = cread(diskfd, tmpbuf, secsize, offset - base);
768 xfer = secsize - base;
771 resid = bytes % secsize;
772 memcpy(buf, &tmpbuf[base], xfer);
774 if (bytes >= secsize) {
775 cnt = cread(diskfd, &buf[xfer], bytes - resid, offset);
776 if (cnt != bytes - resid)
783 cnt = cread(diskfd, tmpbuf, secsize, offset);
784 if (cnt == secsize) {
785 memcpy(&buf[xfer], tmpbuf, resid);
790 if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) {
792 * Trying to read the final fragment.
794 * NB - dump only works in TP_BSIZE blocks, hence
795 * rounds `dev_bsize' fragments up to TP_BSIZE pieces.
796 * It should be smarter about not actually trying to
797 * read more than it can get, but for the time being
798 * we punt and scale back the read only when it gets
799 * us into trouble. (mkm 9/25/83)
805 msg("read error from %s: %s: [block %jd]: count=%d\n",
806 disk, strerror(errno), (intmax_t)blkno, size);
808 msg("short read error from %s: [block %jd]: count=%d, got=%d\n",
809 disk, (intmax_t)blkno, size, cnt);
810 if (++breaderrors > BREADEMAX) {
811 msg("More than %d block read errors from %s\n",
813 broadcast("DUMP IS AILING!\n");
814 msg("This is an unrecoverable error.\n");
815 if (!query("Do you want to attempt to continue?")){
822 * Zero buffer, then try to read each sector of buffer separately,
823 * and bypass the cache.
825 memset(buf, 0, size);
826 for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) {
827 if ((cnt = pread(diskfd, buf, (int)dev_bsize,
828 ((off_t)blkno << dev_bshift))) == dev_bsize)
831 msg("read error from %s: %s: [sector %jd]: count=%ld\n",
832 disk, strerror(errno), (intmax_t)blkno, dev_bsize);
835 msg("short read from %s: [sector %jd]: count=%ld, got=%d\n",
836 disk, (intmax_t)blkno, dev_bsize, cnt);