polder@ark.UUCP (04/10/87)
Below is the source code for `readfs', a utility to list
and/or extract the filesystem tree of a MINIX filesystem.
Effectively, it is the opposite of mkfs.
The listing it produces is in mkfs-format. Thus, with this listing
(and mkfs), you can reproduce the filesystem on another disk,
maybe after changing, adding or deleting some files/directories.
Also included is a "man-page" in the same style as in the book.
------------------ (start of readfs.man) ------------------
Command: readfs - read and extract a MINIX filesystem.
Syntax: readfs [-li] block-special [directory]
Flags: -l: extract all directories and files from the filesystem
and produce a mkfs-listing on standard output.
-i: give the listing, but do not extract files.
Examples: readfs /dev/fd1 # extract all files from /dev/fd1.
readfs -i /dev/hd2 # see what's on /dev/hd2.
readfs -l /dev/at0 rootfs # extract and list the filesystem
# of /dev/at0 and put the tree
# in the directory `rootfs'.
Readfs reads a MINIX filesystem and extracts recursively all directories
and files, and (optionally) produces a mkfs-listing of them on standard output.
The root directory contents are placed in the current directory, unless
a directory is given as argument, in which case the contents are put there.
Readfs tries to restore the attributes (mode/uid/gid/time) of the files
extracted to those of the original files.
Special files are created as ordinary files, but the mkfs-listing
enables mkfs to restore them to original.
----------------- (end of readfs.man, start of readfs.c) -------------------
#include <stdio.h>
#include "fs/const.h"
#include "h/type.h"
#include "h/const.h"
#include "fs/type.h"
#include "fs/buf.h"
#include "fs/super.h"
#undef printf /* Definition used only in the kernel */
extern file_pos lseek();
/*
* Compile with -I/user0/ast/minix
* (i.e. the directory containing the MINIX system sources)
*
* Author: Paul Polderman (polder@cs.vu.nl) April 1987
*/
char verbose = 0; /* give a mkfs-listing of the filesystem */
/* and extracts its contents. */
char noaction = 0; /* just give a mkfs-listing, do not extract files. */
struct super_block sb;
char pathname[1024];
main(argc, argv)
int argc;
char **argv;
{
switch (argc) {
case 2:
pathname[0] = '\0';
readfs(argv[1], pathname);
break;
case 3:
if (argv[1][0] == '-') {
get_flags(&argv[1][1]);
pathname[0] = '\0';
readfs(argv[2], pathname);
} else {
strcpy(pathname, argv[2]);
readfs(argv[1], pathname);
}
break;
case 4: if (argv[1][0] == '-') {
get_flags(&argv[1][1]);
strcpy(pathname, argv[3]);
readfs(argv[2], pathname);
break;
} /* else fall through .. */
default:
fprintf(stderr,"Usage: %s [-li] <special> [dirname]\n",argv[0]);
exit(1);
}
exit(0);
}
get_flags(flags)
register char *flags;
{
while (*flags) {
switch (*flags) {
case 'L':
case 'l':
verbose = 1; break;
case 'I':
case 'i':
noaction = 1; verbose = 1; break;
default:
fprintf(stderr, "Bad flag: %c\n", *flags);
break;
}
flags++;
}
}
readfs(special_file, directory)
char *special_file, *directory;
/*
* Readfs: opens the given special file (with MINIX filesystem),
* and extracts its contents into the given directory.
*/
{
d_inode root_inode;
int special, inum;
file_pos super_b;
umask(0);
/* Open the special file */
if ( (special = open(special_file, 0)) < 0) {
fprintf(stderr, "cannot open %s\n", special_file);
return;
}
/* Read the superblock */
super_b = (file_pos) SUPER_BLOCK * (file_pos) BLOCK_SIZE;
if (lseek(special, super_b, 0) != super_b) {
fprintf(stderr, "cannot seek to superblock\n");
return;
}
if (read(special, &sb, sizeof(struct super_block))
!= sizeof(struct super_block)) {
fprintf(stderr, "cannot read superblock\n");
return;
}
/* Is it really a MINIX filesystem ? */
if (sb.s_magic != SUPER_MAGIC) {
fprintf(stderr, "%s is not a valid MINIX filesystem\n",
special_file);
return;
}
/* Fetch the inode of the root directory */
if (get_inode(special, (inode_nr) ROOT_INODE, &root_inode) < 0) {
fprintf(stderr, "cannot get inode of root directory\n");
return;
}
/* Print number of zones and inodes */
if (verbose)
printf("boot\n%d %d\n", sb.s_nzones, sb.s_ninodes);
/* Extract (recursively) the root directory */
dump_dir(special, &root_inode, directory);
}
/*
* different type of blocks: (used in routine get_block for caching)
*/
#define B_INODE 0 /* Cache #0 is the inode cache */
#define B_INDIRECT 1 /* Cache #1 is the (dbl) indirect block cache */
#define B_DATA 2 /* No cache for data blocks (only read once) */
int get_inode(fd, inum, ip)
int fd;
inode_nr inum;
d_inode *ip;
/*
* Get inode `inum' from the MINIX filesystem. (Uses the inode-cache)
*/
{
struct buf bp;
block_nr block;
block_nr ino_block;
unshort ino_offset;
int r;
/* Calculate start of i-list */
block = SUPER_BLOCK + 1 + sb.s_imap_blocks + sb.s_zmap_blocks;
/* Calculate block with inode inum */
ino_block = ((inum - 1) / INODES_PER_BLOCK);
ino_offset = ((inum - 1) % INODES_PER_BLOCK);
block += ino_block;
/* Fetch the block */
if (get_block(fd, block, &bp, B_INODE) == 0) {
bcopy(&bp.b_inode[ino_offset], ip, sizeof(d_inode));
return(0);
}
/* Oeps, foutje .. */
fprintf(stderr, "cannot find inode %d\n", inum);
return(-1);
}
static int indent = 0; /* current indent (used for mkfs-listing) */
dump_dir(special, ip, directory)
int special;
d_inode *ip;
char *directory;
/*
* Make the given directory (if non-NULL),
* and recursively extract its contents.
*/
{
register dir_struct *dp;
register int n_entries;
register char *name;
block_nr b = 0;
d_inode dip;
struct buf bp;
if (verbose) {
show_info(directory, ip, "");
indent++;
}
if (!noaction && *directory) {
/* Try to make the directory if not already there */
if (mkdir(directory) != 0 || chdir(directory) < 0) {
fprintf(stderr,"mkdir %s failed\n", directory);
return;
}
}
for (name = directory; *name; name++) /* Find end of pathname */
;
*name++ = '/'; /* Add trailing slash */
n_entries = (int) (ip->i_size / (file_pos) sizeof(dir_struct));
while (n_entries > 0) {
/* Read next block of the directory */
if (get_fileblock(special, ip, b, &bp) < 0)
return(-1);
dp = &bp.b_dir[0];
if (b++ == (block_nr) 0) {
dp += 2; /* Skip "." and ".." */
n_entries -= 2;
}
/* Extract the files/directories listed in the block */
while (n_entries-- > 0 && dp < &bp.b_dir[NR_DIR_ENTRIES]) {
if (dp->d_inum != (inode_nr) 0) {
if (get_inode(special, dp->d_inum, &dip) < 0) {
/* Bad luck */
dp++;
continue;
}
/* Add new pathname-component to `pathname'. */
strncpy(name, dp->d_name, NAME_SIZE);
name[NAME_SIZE] = '\0';
/* Call the right routine */
if ((dip.i_mode & I_TYPE) == I_DIRECTORY)
dump_dir(special, &dip, name);
else
dump_file(special, &dip, name);
}
dp++; /* Next entry, please. */
}
}
*--name = '\0'; /* Restore `pathname' to what it was. */
if (!noaction && *directory) {
chdir(".."); /* Go back up. */
restore(directory, ip); /* Restore mode/owner/accesstime */
}
if (verbose) {
do_indent(--indent); /* Let mkfs know we are done */
printf("$\n"); /* with this directory. */
}
}
dump_file(special, ip, filename)
int special;
d_inode *ip;
char *filename;
/*
* Extract given filename from the MINIX-filesystem,
* and store it on the local filesystem.
*/
{
dir_struct *dp;
int file;
block_nr b = 0;
struct buf bp;
file_pos size;
if (verbose)
show_info(filename, ip, pathname);
if (noaction) return(0);
if (access(filename, 0) == 0) {
/* Should not happen, but just in case .. */
fprintf(stderr, "Will not create %s: file exists\n", filename);
return(-1);
}
if ( (file = creat(filename, (ip->i_mode & ALL_MODES) )) < 0) {
fprintf(stderr, "cannot create %s\n", filename);
return(-1);
}
/* Don't try to extract /dev/hd0 */
if ((ip->i_mode & I_TYPE) == I_REGULAR) {
size = ip->i_size;
while (size > (file_pos) 0) {
/* Get next block of file */
if (get_fileblock(special, ip, b++, &bp) < 0) {
close(file);
return(-1);
}
/* Write it to the file */
if (size > (file_pos) BLOCK_SIZE)
write(file, bp.b_data, BLOCK_SIZE);
else
write(file, bp.b_data, (int) size);
size -= (file_pos) BLOCK_SIZE;
}
}
close(file);
restore(filename, ip); /* Restore mode/owner/filetimes */
return(0);
}
#define zone_shift (sb.s_log_zone_size) /* zone to block ratio */
int get_fileblock(special, ip, b, bp)
int special;
d_inode *ip;
block_nr b;
struct buf *bp;
/*
* Read the `b'-th block from the file whose inode is `ip'.
*/
{
zone_nr zone, ind_zone;
block_nr z, zone_index;
int r;
/* Calculate zone in which the datablock number is contained */
zone = (zone_nr) (b >> zone_shift);
/* Calculate index of the block number in the zone */
zone_index = b - ((block_nr) zone << zone_shift);
/* Go get the zone */
if (zone < (zone_nr) NR_DZONE_NUM) { /* direct block */
zone = ip->i_zone[zone];
z = ((block_nr) zone << zone_shift) + zone_index;
r = get_block(special, z, bp, B_DATA);
return(r);
}
/* The zone is not a direct one */
zone -= (zone_nr) NR_DZONE_NUM;
/* Is it single indirect ? */
if (zone < (zone_nr) NR_INDIRECTS) { /* single indirect block */
ind_zone = ip->i_zone[NR_DZONE_NUM];
} else { /* double indirect block */
/* Fetch the double indirect block */
ind_zone = ip->i_zone[NR_DZONE_NUM + 1];
r = get_block(special, (block_nr) ind_zone, bp, B_INDIRECT);
if (r < 0) return(r);
/* Extract the indirect zone number from it */
zone -= (zone_nr) NR_INDIRECTS;
ind_zone = bp->b_ind[zone / (zone_nr) NR_INDIRECTS];
zone %= (zone_nr) NR_INDIRECTS;
}
/* Extract the datablock number from the indirect zone */
r = get_block(special, (block_nr) ind_zone, bp, B_INDIRECT);
if (r < 0) return(r);
zone = bp->b_ind[zone];
/* Calculate datablock number to be fetched */
z = ((block_nr) zone << zone_shift) + zone_index;
r = get_block(special, z, bp, B_DATA);
return(r);
}
/*
* The following routines simulate a LRU block cache.
*
* Definition of a cache block:
*/
struct cache_block {
block_nr b_block; /* block number of block */
long b_access; /* counter value of last access */
char b_buf[BLOCK_SIZE]; /* buffer for block */
};
#define NR_CACHES 2 /* total number of caches */
#define NR_CBLOCKS 5 /* number of blocks in a cache */
static struct cache_block cache[NR_CACHES][NR_CBLOCKS];
static long counter = 0L; /* Counter used as a sense of time. */
/* Incremented after each cache operation. */
get_block(fd, block, bp, type)
int fd;
block_nr block;
struct buf *bp;
int type;
/*
* Get the requested block from the device with filedescriptor fd.
* If it is in the cache, no (floppy-) disk access is needed,
* if not, allocate a cache block and read the block into it.
*/
{
register int i;
register struct cache_block *cache_p, *cp;
if (block == (block_nr) NO_ZONE) {
/* Should never happen in a good filesystem. */
fprintf(stderr, "get_block: NO_ZONE requested !\n");
return(-1);
}
if (type < 0 || type >= NR_CACHES) /* No cache for this type */
return( get_rawblock(fd, block, bp) );
cache_p = cache[type];
cp = (struct cache_block *) 0;
/* First find out if block requested is in the cache */
for (i = 0; i < NR_CBLOCKS; i++) {
if (cache_p[i].b_block == block) { /* found right block */
cp = &cache_p[i];
break;
}
}
if (cp == (struct cache_block *) 0) { /* block is not in cache */
cp = cache_p; /* go find oldest buffer */
for (i = 0; i < NR_CBLOCKS; i++) {
if (cache_p[i].b_access < cp->b_access)
cp = &cache_p[i];
}
/* Fill the buffer with the right block */
if (get_rawblock(fd, block, cp->b_buf) < 0)
return(-1);
}
/* Update/store last access counter */
cp->b_access = ++counter;
cp->b_block = block;
bcopy(cp->b_buf, bp, BLOCK_SIZE);
return(0);
}
int get_rawblock(special, blockno, bufp)
int special;
block_nr blockno;
char *bufp;
/*
* Read a block from the disk.
*/
{
int r;
file_pos pos;
/* Calculate the position of the block on the disk */
pos = (file_pos) blockno * (file_pos) BLOCK_SIZE;
/* Read the block from the disk */
if (lseek(special, pos, 0) == pos
&& read(special, bufp, BLOCK_SIZE) == BLOCK_SIZE)
return(0);
/* Should never get here .. */
fprintf(stderr, "read block %d failed\n", blockno);
return(-1);
}
restore(name, ip)
char *name;
d_inode *ip;
/*
* Restores given file's attributes.
* `ip' contains the attributes of the file on the MINIX filesystem,
* `name' is the filename of the extracted file on the local filesystem.
*/
{
long ttime[2];
chown(name, ip->i_uid, ip->i_gid); /* Fails if not superuser */
chmod(name, (ip->i_mode & ALL_MODES));
ttime[0] = ttime[1] = ip->i_modtime;
utime(name, ttime);
}
/*
* characters to use as prefix to `mkfs' mode field
*/
static char special_chars[] = {
'-', /* I_REGULAR */
'c', /* I_CHAR_SPECIAL */
'd', /* I_DIRECTORY */
'b' /* I_BLOCK_SPECIAL */
};
show_info(name, ip, path)
char *name;
d_inode *ip;
char *path;
/*
* Show information about the given file/dir in `mkfs'-format
*/
{
char c1, c2, c3;
c1 = special_chars[(ip->i_mode >> 13) & 03];
c2 = ((ip->i_mode & ALL_MODES & ~RWX_MODES)==I_SET_UID_BIT) ? 'u' : '-';
c3 = ((ip->i_mode & ALL_MODES & ~RWX_MODES)==I_SET_GID_BIT) ? 'g' : '-';
if (*name) {
do_indent(indent);
printf("%-14s ", name);
}
printf("%c%c%c%03o %d %d", c1, c2, c3,
(ip->i_mode & RWX_MODES), ip->i_uid, ip->i_gid);
switch (ip->i_mode & I_TYPE) {
case I_DIRECTORY:
break;
case I_CHAR_SPECIAL: /* Print major and minor dev numbers */
printf(" %d %d", (ip->i_zone[0] >> MAJOR) & 0377,
(ip->i_zone[0] >> MINOR) & 0377);
break;
case I_BLOCK_SPECIAL: /* Print major and minor dev numbers */
printf(" %d %d", (ip->i_zone[0] >> MAJOR) & 0377,
(ip->i_zone[0] >> MINOR) & 0377);
/* Also print the number of blocks on the device */
printf(" %D", (ip->i_size / (file_pos) BLOCK_SIZE));
break;
default: /* Just print the pathname */
printf(" %s", path);
break;
}
putchar('\n');
}
#define INDENT_SIZE 4
do_indent(i)
int i;
{
i *= INDENT_SIZE;
while (i-- > 0)
putchar(' ');
}
int mkdir(directory)
int directory;
/*
* Make a directory, return exit status.
* This routine is not necessary on systems that
* have a system call to make directories.
*/
{
int pid, status;
if ( (pid = fork()) == 0) {
execl("/bin/mkdir", "mkdir", directory, 0);
execl("/usr/bin/mkdir", "mkdir", directory, 0);
exit(1);
} else if (pid < 0)
return(-1);
while (wait(&status) != pid)
;
return(status);
}
bcopy(src, dest, bytes)
register char *src, *dest;
register int bytes;
/*
* Copy bytes from src to dest.
* May also be in the standard libc.a on some systems.
*/
{
while (bytes--)
*dest++ = *src++;
}
--------------------- (end of readfs.c) ----------------------
- Paul Polderman (polder@cs.vu.nl)