NetBSD/sbin/scan_ffs/scan_ffs.c

486 lines
13 KiB
C

/* $NetBSD: scan_ffs.c,v 1.9 2005/09/02 17:48:40 xtraeme Exp $ */
/*
* Copyright (c) 2005 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Juan Romero Pardines.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Juan Romero Pardines
* for the NetBSD Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Copyright (c) 1998 Niklas Hallqvist, Tobias Weingartner
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Currently it can detect:
* o FFSv1 fsize/bsize: 512/4096, 1024/8192, 2048/16384.
* o FFSv2 fsize/bsize: 512/4096, 1024/8192, 2048/16384,
* 4096/32768, 8192/65536.
* o LFSv[12] fsize/bsize: 512/4096, 1024/8192, 2048/16384,
* 4096/32768, 8192/65536.
*
* TODO:
* o Detect FFSv1 partitions with fsize/bsize > 2048/16384.
*
* -- xtraeme --
*/
#include <sys/cdefs.h>
#ifndef lint
__RCSID("$NetBSD: scan_ffs.c,v 1.9 2005/09/02 17:48:40 xtraeme Exp $");
#endif /* not lint */
#include <sys/types.h>
#include <sys/param.h>
#include <sys/disklabel.h>
#include <sys/dkio.h>
#include <sys/ioctl.h>
#include <sys/fcntl.h>
#include <sys/queue.h>
#include <sys/mount.h>
#include <ufs/ufs/dinode.h>
#include <ufs/lfs/lfs.h>
/* Undefine macros defined by both lfs/lfs.h and ffs/fs.h */
#undef fsbtodb
#undef dbtofsb
#undef blkoff
#undef fragoff
#undef lblktosize
#undef lblkno
#undef numfrags
#undef blkroundup
#undef fragroundup
#undef fragstoblks
#undef blkstofrags
#undef fragnum
#undef blknum
#undef blksize
#undef INOPB
#undef INOPF
#undef NINDIR
#include <ufs/ffs/fs.h>
/* Undefine macros defined by both lfs/lfs.h and ffs/fs.h */
/* ...to make sure we don't later depend on their (ambigious) definition */
#undef fsbtodb
#undef dbtofsb
#undef blkoff
#undef fragoff
#undef lblktosize
#undef lblkno
#undef numfrags
#undef blkroundup
#undef fragroundup
#undef fragstoblks
#undef blkstofrags
#undef fragnum
#undef blknum
#undef blksize
#undef INOPB
#undef INOPF
#undef NINDIR
#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <err.h>
#include <util.h>
/* common struct for FFS/LFS */
struct sblockinfo {
struct lfs *lfs;
struct fs *ffs;
u_int64_t lfs_off;
u_int64_t ffs_off;
char lfs_path[MAXMNTLEN];
char ffs_path[MAXMNTLEN];
} sbinfo;
static daddr_t blk, lastblk;
static int eflag = 0;
static int flags = 0;
static int sbaddr = 0; /* counter for the LFS superblocks */
static char device[MAXPATHLEN];
static const char *fstypes[] = { "NONE", "FFSv1", "FFSv2", "LFS" };
#define FSTYPE_NONE 0
#define FSTYPE_FFSV1 1
#define FSTYPE_FFSV2 2
#define SBCOUNT 64 /* may be changed */
#define SBPASS (SBCOUNT * SBLOCKSIZE / 512)
/* This is only useful for LFS */
/* first sblock address contains the correct offset */
#define FIRST_SBLOCK_ADDRESS 1
/* second and third sblock address contain lfs_fsmnt[MAXMNTLEN] */
#define SECOND_SBLOCK_ADDRESS 2
/* last sblock address in a LFS partition */
#define MAX_SBLOCK_ADDRESS 10
enum { NADA, VERBOSE, LABELS };
/* FFS functions */
static void ffs_printpart(int, size_t, int);
static void ffs_scan(int);
static int ffs_checkver(void);
/* LFS functions */
static void lfs_printpart(int, int, struct sblockinfo *);
static void lfs_scan(int);
/* common functions */
static void usage(void) __attribute__((__noreturn__));
static int scan_disk(int, daddr_t, daddr_t, int);
static int
ffs_checkver(void)
{
switch (sbinfo.ffs->fs_magic) {
case FS_UFS1_MAGIC:
case FS_UFS1_MAGIC_SWAPPED:
sbinfo.ffs->fs_size = sbinfo.ffs->fs_old_size;
return FSTYPE_FFSV1;
case FS_UFS2_MAGIC:
case FS_UFS2_MAGIC_SWAPPED:
return FSTYPE_FFSV2;
default:
return FSTYPE_NONE;
}
}
static void
ffs_printpart(int flag, size_t ffsize, int n)
{
int fstype = ffs_checkver();
switch (flag) {
case VERBOSE:
(void)printf("block: %" PRIu64 " id %x,%x size %" PRIu64 "\n",
blk + (n / 512),
sbinfo.ffs->fs_id[0],
sbinfo.ffs->fs_id[1], sbinfo.ffs->fs_size);
break;
case LABELS:
(void)printf("X: %9" PRIu64,
(uint64_t)((off_t)sbinfo.ffs->fs_size *
sbinfo.ffs->fs_fsize / 512));
switch (fstype) {
case FSTYPE_FFSV1:
(void)printf(" %9" PRIu64,
blk + (n / 512) - (2 * SBLOCKSIZE / 512));
break;
case FSTYPE_FFSV2:
(void)printf(" %9" PRIu64,
blk + (n / 512) -
(ffsize * SBLOCKSIZE / 512 + 128));
break;
default:
break;
}
(void)printf(" 4.2BSD %6d %5d %7d # %s [%s]\n",
sbinfo.ffs->fs_fsize, sbinfo.ffs->fs_bsize,
sbinfo.ffs->fs_old_cpg,
sbinfo.ffs_path, fstypes[fstype]);
break;
default:
(void)printf("%s ", fstypes[fstype]);
switch (fstype) {
case FSTYPE_FFSV1:
(void)printf("at %" PRIu64,
blk + (n / 512) - (2 * SBLOCKSIZE / 512));
break;
case FSTYPE_FFSV2:
(void)printf("at %" PRIu64,
blk + (n / 512) -
(ffsize * SBLOCKSIZE / 512 + 128));
break;
default:
break;
}
(void)printf(" size %" PRIu64 ", last mounted on %s\n",
(uint64_t)((off_t)sbinfo.ffs->fs_size *
sbinfo.ffs->fs_fsize / 512), sbinfo.ffs_path);
break;
}
}
static void
ffs_scan(int n)
{
int fstype = ffs_checkver();
size_t i;
/*
* XXX:
* It cannot find FFSv1 partitions with fsize/bsize > 2048/16384,
* same problem found in the original program that comes from
* OpenBSD (scan_ffs(8)).
*/
if (flags & VERBOSE)
ffs_printpart(VERBOSE, NADA, n);
switch (fstype) {
case FSTYPE_FFSV1:
if (((blk + (n / 512)) - lastblk) == (SBLOCKSIZE / 512)) {
if (flags & LABELS)
ffs_printpart(LABELS, NADA, n);
else
ffs_printpart(NADA, NADA, n);
}
break;
case FSTYPE_FFSV2:
/*
* That checks for FFSv2 partitions with fragsize/blocksize:
* 512/4096, 1024/8192, 2048/16384, 4096/32768 and 8192/65536.
* Really enough for now.
*/
for (i = 1; i < 16; i <<= 1)
if (((blk + (n / 512)) - lastblk) ==
(i * SBLOCKSIZE / 512)) {
if (flags & LABELS)
ffs_printpart(LABELS, i, n);
else
ffs_printpart(NADA, i, n);
}
break;
}
}
static void
lfs_printpart(int flag, int n, struct sblockinfo *sbi)
{
if (flags & VERBOSE)
(void)printf("block: %" PRIu64 " size %" PRIu32 "\n",
blk + (n / 512), sbi->lfs->lfs_size);
switch (flag) {
case LABELS:
(void)printf("X: %9" PRIu64,
(uint64_t)((off_t)sbi->lfs->lfs_size *
sbi->lfs->lfs_fsize / 512));
(void)printf(" %9" PRIu64, sbi->lfs_off);
(void)printf(" 4.4LFS %6d %5d %7d # %s [LFSv%d]\n",
sbi->lfs->lfs_fsize, sbi->lfs->lfs_bsize,
sbi->lfs->lfs_nseg, sbi->lfs_path,
sbi->lfs->lfs_version);
break;
default:
(void)printf("LFSv%d ", sbinfo.lfs->lfs_version);
(void)printf("at %" PRIu64, sbinfo.lfs_off);
(void)printf(" size %" PRIu64 ", last mounted on %s\n",
(uint64_t)((off_t)sbinfo.lfs->lfs_size *
sbinfo.lfs->lfs_fsize / 512), sbinfo.lfs_path);
break;
}
}
static void
lfs_scan(int n)
{
/* backup offset */
lastblk = blk + (n / 512) - (LFS_SBPAD / 512);
/* increment counter */
++sbaddr;
switch (sbaddr) {
/*
* first superblock contains the right offset, but lfs_fsmnt is
* empty... afortunately the next superblock address has it.
*/
case FIRST_SBLOCK_ADDRESS:
/* copy partition offset */
if (sbinfo.lfs_off != lastblk)
sbinfo.lfs_off = blk + (n / 512) - (LFS_SBPAD / 512);
break;
case SECOND_SBLOCK_ADDRESS:
/* copy the path of last mount */
(void)memcpy(sbinfo.lfs_path,
sbinfo.lfs->lfs_fsmnt, MAXMNTLEN);
/* print now that we have the info */
if (flags & LABELS)
lfs_printpart(LABELS, n, &sbinfo);
else
lfs_printpart(NADA, n, &sbinfo);
/* clear our struct */
(void)memset(&sbinfo, 0, sizeof(sbinfo));
break;
case MAX_SBLOCK_ADDRESS:
/*
* reset the counter, this is the last superblock address,
* the next one will be another partition maybe.
*/
sbaddr = 0;
break;
default:
break;
}
}
static int
scan_disk(int fd, daddr_t beg, daddr_t end, int fflags)
{
u_int8_t buf[SBLOCKSIZE * SBCOUNT];
int n, fstype;
n = fstype = 0;
lastblk = -1;
/* clear our struct before using it */
(void)memset(&sbinfo, 0, sizeof(sbinfo));
if (fflags & LABELS)
(void)printf(
"# size offset fstype [fsize bsize cpg/sgs]\n");
for (blk = beg; blk <= ((end < 0) ? blk: end); blk += SBPASS) {
(void)memset(buf, 0, sizeof(buf));
if (pread(fd, buf, sizeof(buf), (off_t)blk * 512) == (off_t)-1)
err(1, "pread");
for (n = 0; n < (SBLOCKSIZE * SBCOUNT); n += 512) {
sbinfo.ffs = (struct fs *)(void *)&buf[n];
sbinfo.lfs = (struct lfs *)(void *)&buf[n];
fstype = ffs_checkver();
switch (fstype) {
case FSTYPE_FFSV1:
case FSTYPE_FFSV2:
ffs_scan(n);
lastblk = blk + (n / 512);
(void)memcpy(sbinfo.ffs_path,
sbinfo.ffs->fs_fsmnt, MAXMNTLEN);
break;
case FSTYPE_NONE:
/* maybe LFS? */
if (sbinfo.lfs->lfs_magic == LFS_MAGIC)
lfs_scan(n);
break;
default:
break;
}
}
}
return EXIT_SUCCESS;
}
static void
usage(void)
{
(void)fprintf(stderr,
"Usage: %s [-lv] [-s start] [-e end] device\n", getprogname());
exit(EXIT_FAILURE);
}
int
main(int argc, char **argv)
{
int ch, fd;
daddr_t end = -1, beg = 0;
struct disklabel dl;
setprogname(*argv);
while ((ch = getopt(argc, argv, "e:ls:v")) != -1)
switch(ch) {
case 'e':
eflag = 1;
end = atoi(optarg);
break;
case 'l':
flags |= LABELS;
break;
case 's':
beg = atoi(optarg);
break;
case 'v':
flags |= VERBOSE;
break;
default:
usage();
/* NOTREACHED */
}
argc -= optind;
argv += optind;
if (argc != 1)
usage();
fd = opendisk(argv[0], O_RDONLY, device, sizeof(device), 0);
if (fd == -1)
err(1, "Cannot open `%s'", device);
/* NOTREACHED */
if (ioctl(fd, DIOCGDINFO, &dl) == -1) {
warn("Couldn't retrieve disklabel");
(void)memset(&dl, 0, sizeof(dl));
dl.d_secperunit = 0x7fffffff;
} else {
(void)printf("Disk: %s\n", dl.d_typename);
(void)printf("Total sectors on disk: %" PRIu32 "\n\n",
dl.d_secperunit);
}
if (!eflag)
end = dl.d_secperunit; /* default to max sectors */
return scan_disk(fd, beg, end, flags);
}