blob: a2b6d8a6e5547dc196627bac05aee12a21b88a79 [file] [log] [blame]
/* vi: set sw=4 ts=4: */
/*
* mkfs.c - make a linux (minix) file-system.
*
* (C) 1991 Linus Torvalds. This file may be redistributed as per
* the Linux copyright.
*/
/*
* DD.MM.YY
*
* 24.11.91 - Time began. Used the fsck sources to get started.
*
* 25.11.91 - Corrected some bugs. Added support for ".badblocks"
* The algorithm for ".badblocks" is a bit weird, but
* it should work. Oh, well.
*
* 25.01.92 - Added the -l option for getting the list of bad blocks
* out of a named file. (Dave Rivers, rivers@ponds.uucp)
*
* 28.02.92 - Added %-information when using -c.
*
* 28.02.93 - Added support for other namelengths than the original
* 14 characters so that I can test the new kernel routines..
*
* 09.10.93 - Make exit status conform to that required by fsutil
* (Rik Faith, faith@cs.unc.edu)
*
* 31.10.93 - Added inode request feature, for backup floppies: use
* 32 inodes, for a news partition use more.
* (Scott Heavner, sdh@po.cwru.edu)
*
* 03.01.94 - Added support for file system valid flag.
* (Dr. Wettstein, greg%wind.uucp@plains.nodak.edu)
*
* 30.10.94 - added support for v2 filesystem
* (Andreas Schwab, schwab@issan.informatik.uni-dortmund.de)
*
* 09.11.94 - Added test to prevent overwrite of mounted fs adapted
* from Theodore Ts'o's (tytso@athena.mit.edu) mke2fs
* program. (Daniel Quinlan, quinlan@yggdrasil.com)
*
* 03.20.95 - Clear first 512 bytes of filesystem to make certain that
* the filesystem is not misidentified as a MS-DOS FAT filesystem.
* (Daniel Quinlan, quinlan@yggdrasil.com)
*
* 02.07.96 - Added small patch from Russell King to make the program a
* good deal more portable (janl@math.uio.no)
*
* Usage: mkfs [-c | -l filename ] [-v] [-nXX] [-iXX] device [size-in-blocks]
*
* -c for readablility checking (SLOW!)
* -l for getting a list of bad blocks from a file.
* -n for namelength (currently the kernel only uses 14 or 30)
* -i for number of inodes
* -v for v2 filesystem
*
* The device may be a block device or a image of one, but this isn't
* enforced (but it's not much fun on a character device :-).
*
* Modified for BusyBox by Erik Andersen <andersen@debian.org> --
* removed getopt based parser and added a hand rolled one.
*/
#include "busybox.h"
#include <stdio.h>
#include <time.h>
#include <unistd.h>
#include <string.h>
#include <signal.h>
#include <fcntl.h>
#include <ctype.h>
#include <stdlib.h>
#include <termios.h>
#include <sys/ioctl.h>
#include <sys/param.h>
#include <mntent.h>
typedef unsigned char u8;
typedef unsigned short u16;
typedef unsigned int u32;
#define MINIX_ROOT_INO 1
#define MINIX_LINK_MAX 250
#define MINIX2_LINK_MAX 65530
#define MINIX_I_MAP_SLOTS 8
#define MINIX_Z_MAP_SLOTS 64
#define MINIX_SUPER_MAGIC 0x137F /* original minix fs */
#define MINIX_SUPER_MAGIC2 0x138F /* minix fs, 30 char names */
#define MINIX2_SUPER_MAGIC 0x2468 /* minix V2 fs */
#define MINIX2_SUPER_MAGIC2 0x2478 /* minix V2 fs, 30 char names */
#define MINIX_VALID_FS 0x0001 /* Clean fs. */
#define MINIX_ERROR_FS 0x0002 /* fs has errors. */
#define MINIX_INODES_PER_BLOCK ((BLOCK_SIZE)/(sizeof (struct minix_inode)))
#define MINIX2_INODES_PER_BLOCK ((BLOCK_SIZE)/(sizeof (struct minix2_inode)))
#define MINIX_V1 0x0001 /* original minix fs */
#define MINIX_V2 0x0002 /* minix V2 fs */
#define INODE_VERSION(inode) inode->i_sb->u.minix_sb.s_version
/*
* This is the original minix inode layout on disk.
* Note the 8-bit gid and atime and ctime.
*/
struct minix_inode {
u16 i_mode;
u16 i_uid;
u32 i_size;
u32 i_time;
u8 i_gid;
u8 i_nlinks;
u16 i_zone[9];
};
/*
* The new minix inode has all the time entries, as well as
* long block numbers and a third indirect block (7+1+1+1
* instead of 7+1+1). Also, some previously 8-bit values are
* now 16-bit. The inode is now 64 bytes instead of 32.
*/
struct minix2_inode {
u16 i_mode;
u16 i_nlinks;
u16 i_uid;
u16 i_gid;
u32 i_size;
u32 i_atime;
u32 i_mtime;
u32 i_ctime;
u32 i_zone[10];
};
/*
* minix super-block data on disk
*/
struct minix_super_block {
u16 s_ninodes;
u16 s_nzones;
u16 s_imap_blocks;
u16 s_zmap_blocks;
u16 s_firstdatazone;
u16 s_log_zone_size;
u32 s_max_size;
u16 s_magic;
u16 s_state;
u32 s_zones;
};
struct minix_dir_entry {
u16 inode;
char name[0];
};
#define BLOCK_SIZE_BITS 10
#define BLOCK_SIZE (1<<BLOCK_SIZE_BITS)
#define NAME_MAX 255 /* # chars in a file name */
#define MINIX_INODES_PER_BLOCK ((BLOCK_SIZE)/(sizeof (struct minix_inode)))
#define MINIX_VALID_FS 0x0001 /* Clean fs. */
#define MINIX_ERROR_FS 0x0002 /* fs has errors. */
#define MINIX_SUPER_MAGIC 0x137F /* original minix fs */
#define MINIX_SUPER_MAGIC2 0x138F /* minix fs, 30 char names */
#ifndef BLKGETSIZE
#define BLKGETSIZE _IO(0x12,96) /* return device size */
#endif
#ifndef __linux__
#define volatile
#endif
#define MINIX_ROOT_INO 1
#define MINIX_BAD_INO 2
#define TEST_BUFFER_BLOCKS 16
#define MAX_GOOD_BLOCKS 512
#define UPPER(size,n) (((size)+((n)-1))/(n))
#define INODE_SIZE (sizeof(struct minix_inode))
#ifdef BB_FEATURE_MINIX2
#define INODE_SIZE2 (sizeof(struct minix2_inode))
#define INODE_BLOCKS UPPER(INODES, (version2 ? MINIX2_INODES_PER_BLOCK \
: MINIX_INODES_PER_BLOCK))
#else
#define INODE_BLOCKS UPPER(INODES, (MINIX_INODES_PER_BLOCK))
#endif
#define INODE_BUFFER_SIZE (INODE_BLOCKS * BLOCK_SIZE)
#define BITS_PER_BLOCK (BLOCK_SIZE<<3)
static char *device_name = NULL;
static int DEV = -1;
static long BLOCKS = 0;
static int check = 0;
static int badblocks = 0;
static int namelen = 30; /* default (changed to 30, per Linus's
suggestion, Sun Nov 21 08:05:07 1993) */
static int dirsize = 32;
static int magic = MINIX_SUPER_MAGIC2;
static int version2 = 0;
static char root_block[BLOCK_SIZE] = "\0";
static char *inode_buffer = NULL;
#define Inode (((struct minix_inode *) inode_buffer)-1)
#ifdef BB_FEATURE_MINIX2
#define Inode2 (((struct minix2_inode *) inode_buffer)-1)
#endif
static char super_block_buffer[BLOCK_SIZE];
static char boot_block_buffer[512];
#define Super (*(struct minix_super_block *)super_block_buffer)
#define INODES ((unsigned long)Super.s_ninodes)
#ifdef BB_FEATURE_MINIX2
#define ZONES ((unsigned long)(version2 ? Super.s_zones : Super.s_nzones))
#else
#define ZONES ((unsigned long)(Super.s_nzones))
#endif
#define IMAPS ((unsigned long)Super.s_imap_blocks)
#define ZMAPS ((unsigned long)Super.s_zmap_blocks)
#define FIRSTZONE ((unsigned long)Super.s_firstdatazone)
#define ZONESIZE ((unsigned long)Super.s_log_zone_size)
#define MAXSIZE ((unsigned long)Super.s_max_size)
#define MAGIC (Super.s_magic)
#define NORM_FIRSTZONE (2+IMAPS+ZMAPS+INODE_BLOCKS)
static char *inode_map;
static char *zone_map;
static unsigned short good_blocks_table[MAX_GOOD_BLOCKS];
static int used_good_blocks = 0;
static unsigned long req_nr_inodes = 0;
static inline int bit(char * a,unsigned int i)
{
return (a[i >> 3] & (1<<(i & 7))) != 0;
}
#define inode_in_use(x) (bit(inode_map,(x)))
#define zone_in_use(x) (bit(zone_map,(x)-FIRSTZONE+1))
#define mark_inode(x) (setbit(inode_map,(x)))
#define unmark_inode(x) (clrbit(inode_map,(x)))
#define mark_zone(x) (setbit(zone_map,(x)-FIRSTZONE+1))
#define unmark_zone(x) (clrbit(zone_map,(x)-FIRSTZONE+1))
static __attribute__ ((noreturn)) void show_usage()
{
usage(mkfs_minix_usage);
}
/*
* Check to make certain that our new filesystem won't be created on
* an already mounted partition. Code adapted from mke2fs, Copyright
* (C) 1994 Theodore Ts'o. Also licensed under GPL.
*/
static void check_mount(void)
{
FILE *f;
struct mntent *mnt;
if ((f = setmntent(MOUNTED, "r")) == NULL)
return;
while ((mnt = getmntent(f)) != NULL)
if (strcmp(device_name, mnt->mnt_fsname) == 0)
break;
endmntent(f);
if (!mnt)
return;
error_msg_and_die("%s is mounted; will not make a filesystem here!", device_name);
}
static long valid_offset(int fd, int offset)
{
char ch;
if (lseek(fd, offset, 0) < 0)
return 0;
if (read(fd, &ch, 1) < 1)
return 0;
return 1;
}
static int count_blocks(int fd)
{
int high, low;
low = 0;
for (high = 1; valid_offset(fd, high); high *= 2)
low = high;
while (low < high - 1) {
const int mid = (low + high) / 2;
if (valid_offset(fd, mid))
low = mid;
else
high = mid;
}
valid_offset(fd, 0);
return (low + 1);
}
static int get_size(const char *file)
{
int fd;
long size;
if ((fd = open(file, O_RDWR)) < 0)
perror_msg_and_die("%s", file);
if (ioctl(fd, BLKGETSIZE, &size) >= 0) {
close(fd);
return (size * 512);
}
size = count_blocks(fd);
close(fd);
return size;
}
static void write_tables(void)
{
/* Mark the super block valid. */
Super.s_state |= MINIX_VALID_FS;
Super.s_state &= ~MINIX_ERROR_FS;
if (lseek(DEV, 0, SEEK_SET))
error_msg_and_die("seek to boot block failed in write_tables");
if (512 != write(DEV, boot_block_buffer, 512))
error_msg_and_die("unable to clear boot sector");
if (BLOCK_SIZE != lseek(DEV, BLOCK_SIZE, SEEK_SET))
error_msg_and_die("seek failed in write_tables");
if (BLOCK_SIZE != write(DEV, super_block_buffer, BLOCK_SIZE))
error_msg_and_die("unable to write super-block");
if (IMAPS * BLOCK_SIZE != write(DEV, inode_map, IMAPS * BLOCK_SIZE))
error_msg_and_die("unable to write inode map");
if (ZMAPS * BLOCK_SIZE != write(DEV, zone_map, ZMAPS * BLOCK_SIZE))
error_msg_and_die("unable to write zone map");
if (INODE_BUFFER_SIZE != write(DEV, inode_buffer, INODE_BUFFER_SIZE))
error_msg_and_die("unable to write inodes");
}
static void write_block(int blk, char *buffer)
{
if (blk * BLOCK_SIZE != lseek(DEV, blk * BLOCK_SIZE, SEEK_SET))
error_msg_and_die("seek failed in write_block");
if (BLOCK_SIZE != write(DEV, buffer, BLOCK_SIZE))
error_msg_and_die("write failed in write_block");
}
static int get_free_block(void)
{
int blk;
if (used_good_blocks + 1 >= MAX_GOOD_BLOCKS)
error_msg_and_die("too many bad blocks");
if (used_good_blocks)
blk = good_blocks_table[used_good_blocks - 1] + 1;
else
blk = FIRSTZONE;
while (blk < ZONES && zone_in_use(blk))
blk++;
if (blk >= ZONES)
error_msg_and_die("not enough good blocks");
good_blocks_table[used_good_blocks] = blk;
used_good_blocks++;
return blk;
}
static void mark_good_blocks(void)
{
int blk;
for (blk = 0; blk < used_good_blocks; blk++)
mark_zone(good_blocks_table[blk]);
}
inline int next(int zone)
{
if (!zone)
zone = FIRSTZONE - 1;
while (++zone < ZONES)
if (zone_in_use(zone))
return zone;
return 0;
}
static void make_bad_inode(void)
{
struct minix_inode *inode = &Inode[MINIX_BAD_INO];
int i, j, zone;
int ind = 0, dind = 0;
unsigned short ind_block[BLOCK_SIZE >> 1];
unsigned short dind_block[BLOCK_SIZE >> 1];
#define NEXT_BAD (zone = next(zone))
if (!badblocks)
return;
mark_inode(MINIX_BAD_INO);
inode->i_nlinks = 1;
inode->i_time = time(NULL);
inode->i_mode = S_IFREG + 0000;
inode->i_size = badblocks * BLOCK_SIZE;
zone = next(0);
for (i = 0; i < 7; i++) {
inode->i_zone[i] = zone;
if (!NEXT_BAD)
goto end_bad;
}
inode->i_zone[7] = ind = get_free_block();
memset(ind_block, 0, BLOCK_SIZE);
for (i = 0; i < 512; i++) {
ind_block[i] = zone;
if (!NEXT_BAD)
goto end_bad;
}
inode->i_zone[8] = dind = get_free_block();
memset(dind_block, 0, BLOCK_SIZE);
for (i = 0; i < 512; i++) {
write_block(ind, (char *) ind_block);
dind_block[i] = ind = get_free_block();
memset(ind_block, 0, BLOCK_SIZE);
for (j = 0; j < 512; j++) {
ind_block[j] = zone;
if (!NEXT_BAD)
goto end_bad;
}
}
error_msg_and_die("too many bad blocks");
end_bad:
if (ind)
write_block(ind, (char *) ind_block);
if (dind)
write_block(dind, (char *) dind_block);
}
#ifdef BB_FEATURE_MINIX2
static void make_bad_inode2(void)
{
struct minix2_inode *inode = &Inode2[MINIX_BAD_INO];
int i, j, zone;
int ind = 0, dind = 0;
unsigned long ind_block[BLOCK_SIZE >> 2];
unsigned long dind_block[BLOCK_SIZE >> 2];
if (!badblocks)
return;
mark_inode(MINIX_BAD_INO);
inode->i_nlinks = 1;
inode->i_atime = inode->i_mtime = inode->i_ctime = time(NULL);
inode->i_mode = S_IFREG + 0000;
inode->i_size = badblocks * BLOCK_SIZE;
zone = next(0);
for (i = 0; i < 7; i++) {
inode->i_zone[i] = zone;
if (!NEXT_BAD)
goto end_bad;
}
inode->i_zone[7] = ind = get_free_block();
memset(ind_block, 0, BLOCK_SIZE);
for (i = 0; i < 256; i++) {
ind_block[i] = zone;
if (!NEXT_BAD)
goto end_bad;
}
inode->i_zone[8] = dind = get_free_block();
memset(dind_block, 0, BLOCK_SIZE);
for (i = 0; i < 256; i++) {
write_block(ind, (char *) ind_block);
dind_block[i] = ind = get_free_block();
memset(ind_block, 0, BLOCK_SIZE);
for (j = 0; j < 256; j++) {
ind_block[j] = zone;
if (!NEXT_BAD)
goto end_bad;
}
}
/* Could make triple indirect block here */
error_msg_and_die("too many bad blocks");
end_bad:
if (ind)
write_block(ind, (char *) ind_block);
if (dind)
write_block(dind, (char *) dind_block);
}
#endif
static void make_root_inode(void)
{
struct minix_inode *inode = &Inode[MINIX_ROOT_INO];
mark_inode(MINIX_ROOT_INO);
inode->i_zone[0] = get_free_block();
inode->i_nlinks = 2;
inode->i_time = time(NULL);
if (badblocks)
inode->i_size = 3 * dirsize;
else {
root_block[2 * dirsize] = '\0';
root_block[2 * dirsize + 1] = '\0';
inode->i_size = 2 * dirsize;
}
inode->i_mode = S_IFDIR + 0755;
inode->i_uid = getuid();
if (inode->i_uid)
inode->i_gid = getgid();
write_block(inode->i_zone[0], root_block);
}
#ifdef BB_FEATURE_MINIX2
static void make_root_inode2(void)
{
struct minix2_inode *inode = &Inode2[MINIX_ROOT_INO];
mark_inode(MINIX_ROOT_INO);
inode->i_zone[0] = get_free_block();
inode->i_nlinks = 2;
inode->i_atime = inode->i_mtime = inode->i_ctime = time(NULL);
if (badblocks)
inode->i_size = 3 * dirsize;
else {
root_block[2 * dirsize] = '\0';
root_block[2 * dirsize + 1] = '\0';
inode->i_size = 2 * dirsize;
}
inode->i_mode = S_IFDIR + 0755;
inode->i_uid = getuid();
if (inode->i_uid)
inode->i_gid = getgid();
write_block(inode->i_zone[0], root_block);
}
#endif
static void setup_tables(void)
{
int i;
unsigned long inodes;
memset(super_block_buffer, 0, BLOCK_SIZE);
memset(boot_block_buffer, 0, 512);
MAGIC = magic;
ZONESIZE = 0;
MAXSIZE = version2 ? 0x7fffffff : (7 + 512 + 512 * 512) * 1024;
ZONES = BLOCKS;
/* some magic nrs: 1 inode / 3 blocks */
if (req_nr_inodes == 0)
inodes = BLOCKS / 3;
else
inodes = req_nr_inodes;
/* Round up inode count to fill block size */
#ifdef BB_FEATURE_MINIX2
if (version2)
inodes = ((inodes + MINIX2_INODES_PER_BLOCK - 1) &
~(MINIX2_INODES_PER_BLOCK - 1));
else
#endif
inodes = ((inodes + MINIX_INODES_PER_BLOCK - 1) &
~(MINIX_INODES_PER_BLOCK - 1));
if (inodes > 65535)
inodes = 65535;
INODES = inodes;
IMAPS = UPPER(INODES + 1, BITS_PER_BLOCK);
ZMAPS = 0;
i = 0;
while (ZMAPS !=
UPPER(BLOCKS - (2 + IMAPS + ZMAPS + INODE_BLOCKS) + 1,
BITS_PER_BLOCK) && i < 1000) {
ZMAPS =
UPPER(BLOCKS - (2 + IMAPS + ZMAPS + INODE_BLOCKS) + 1,
BITS_PER_BLOCK);
i++;
}
/* Real bad hack but overwise mkfs.minix can be thrown
* in infinite loop...
* try:
* dd if=/dev/zero of=test.fs count=10 bs=1024
* /sbin/mkfs.minix -i 200 test.fs
* */
if (i >= 999) {
error_msg_and_die("unable to allocate buffers for maps");
}
FIRSTZONE = NORM_FIRSTZONE;
inode_map = xmalloc(IMAPS * BLOCK_SIZE);
zone_map = xmalloc(ZMAPS * BLOCK_SIZE);
memset(inode_map, 0xff, IMAPS * BLOCK_SIZE);
memset(zone_map, 0xff, ZMAPS * BLOCK_SIZE);
for (i = FIRSTZONE; i < ZONES; i++)
unmark_zone(i);
for (i = MINIX_ROOT_INO; i <= INODES; i++)
unmark_inode(i);
inode_buffer = xmalloc(INODE_BUFFER_SIZE);
memset(inode_buffer, 0, INODE_BUFFER_SIZE);
printf("%ld inodes\n", INODES);
printf("%ld blocks\n", ZONES);
printf("Firstdatazone=%ld (%ld)\n", FIRSTZONE, NORM_FIRSTZONE);
printf("Zonesize=%d\n", BLOCK_SIZE << ZONESIZE);
printf("Maxsize=%ld\n\n", MAXSIZE);
}
/*
* Perform a test of a block; return the number of
* blocks readable/writeable.
*/
long do_check(char *buffer, int try, unsigned int current_block)
{
long got;
/* Seek to the correct loc. */
if (lseek(DEV, current_block * BLOCK_SIZE, SEEK_SET) !=
current_block * BLOCK_SIZE) {
error_msg_and_die("seek failed during testing of blocks");
}
/* Try the read */
got = read(DEV, buffer, try * BLOCK_SIZE);
if (got < 0)
got = 0;
if (got & (BLOCK_SIZE - 1)) {
printf("Weird values in do_check: probably bugs\n");
}
got /= BLOCK_SIZE;
return got;
}
static unsigned int currently_testing = 0;
static void alarm_intr(int alnum)
{
if (currently_testing >= ZONES)
return;
signal(SIGALRM, alarm_intr);
alarm(5);
if (!currently_testing)
return;
printf("%d ...", currently_testing);
fflush(stdout);
}
static void check_blocks(void)
{
int try, got;
static char buffer[BLOCK_SIZE * TEST_BUFFER_BLOCKS];
currently_testing = 0;
signal(SIGALRM, alarm_intr);
alarm(5);
while (currently_testing < ZONES) {
if (lseek(DEV, currently_testing * BLOCK_SIZE, SEEK_SET) !=
currently_testing * BLOCK_SIZE)
error_msg_and_die("seek failed in check_blocks");
try = TEST_BUFFER_BLOCKS;
if (currently_testing + try > ZONES)
try = ZONES - currently_testing;
got = do_check(buffer, try, currently_testing);
currently_testing += got;
if (got == try)
continue;
if (currently_testing < FIRSTZONE)
error_msg_and_die("bad blocks before data-area: cannot make fs");
mark_zone(currently_testing);
badblocks++;
currently_testing++;
}
if (badblocks > 1)
printf("%d bad blocks\n", badblocks);
else if (badblocks == 1)
printf("one bad block\n");
}
static void get_list_blocks(filename)
char *filename;
{
FILE *listfile;
unsigned long blockno;
listfile = fopen(filename, "r");
if (listfile == (FILE *) NULL) {
error_msg_and_die("can't open file of bad blocks");
}
while (!feof(listfile)) {
fscanf(listfile, "%ld\n", &blockno);
mark_zone(blockno);
badblocks++;
}
if (badblocks > 1)
printf("%d bad blocks\n", badblocks);
else if (badblocks == 1)
printf("one bad block\n");
}
extern int mkfs_minix_main(int argc, char **argv)
{
int i=1;
char *tmp;
struct stat statbuf;
char *listfile = NULL;
int stopIt=FALSE;
if (INODE_SIZE * MINIX_INODES_PER_BLOCK != BLOCK_SIZE)
error_msg_and_die("bad inode size");
#ifdef BB_FEATURE_MINIX2
if (INODE_SIZE2 * MINIX2_INODES_PER_BLOCK != BLOCK_SIZE)
error_msg_and_die("bad inode size");
#endif
/* Parse options */
argv++;
while (--argc >= 0 && *argv && **argv) {
if (**argv == '-') {
stopIt=FALSE;
while (i > 0 && *++(*argv) && stopIt==FALSE) {
switch (**argv) {
case 'c':
check = 1;
break;
case 'i':
{
char *cp=NULL;
if (*(*argv+1) != 0) {
cp = ++(*argv);
} else {
if (--argc == 0) {
goto goodbye;
}
cp = *(++argv);
}
req_nr_inodes = strtoul(cp, &tmp, 0);
if (*tmp)
show_usage();
stopIt=TRUE;
break;
}
case 'l':
if (--argc == 0) {
goto goodbye;
}
listfile = *(++argv);
break;
case 'n':
{
char *cp=NULL;
if (*(*argv+1) != 0) {
cp = ++(*argv);
} else {
if (--argc == 0) {
goto goodbye;
}
cp = *(++argv);
}
i = strtoul(cp, &tmp, 0);
if (*tmp)
show_usage();
if (i == 14)
magic = MINIX_SUPER_MAGIC;
else if (i == 30)
magic = MINIX_SUPER_MAGIC2;
else
show_usage();
namelen = i;
dirsize = i + 2;
stopIt=TRUE;
break;
}
case 'v':
#ifdef BB_FEATURE_MINIX2
version2 = 1;
#else
error_msg("%s: not compiled with minix v2 support",
device_name);
exit(-1);
#endif
break;
case '-':
case 'h':
default:
goodbye:
show_usage();
}
}
} else {
if (device_name == NULL)
device_name = *argv;
else if (BLOCKS == 0)
BLOCKS = strtol(*argv, &tmp, 0);
else {
goto goodbye;
}
}
argv++;
}
if (device_name && !BLOCKS)
BLOCKS = get_size(device_name) / 1024;
if (!device_name || BLOCKS < 10) {
show_usage();
}
#ifdef BB_FEATURE_MINIX2
if (version2) {
if (namelen == 14)
magic = MINIX2_SUPER_MAGIC;
else
magic = MINIX2_SUPER_MAGIC2;
} else
#endif
if (BLOCKS > 65535)
BLOCKS = 65535;
check_mount(); /* is it already mounted? */
tmp = root_block;
*(short *) tmp = 1;
strcpy(tmp + 2, ".");
tmp += dirsize;
*(short *) tmp = 1;
strcpy(tmp + 2, "..");
tmp += dirsize;
*(short *) tmp = 2;
strcpy(tmp + 2, ".badblocks");
DEV = open(device_name, O_RDWR);
if (DEV < 0)
error_msg_and_die("unable to open %s", device_name);
if (fstat(DEV, &statbuf) < 0)
error_msg_and_die("unable to stat %s", device_name);
if (!S_ISBLK(statbuf.st_mode))
check = 0;
else if (statbuf.st_rdev == 0x0300 || statbuf.st_rdev == 0x0340)
error_msg_and_die("will not try to make filesystem on '%s'", device_name);
setup_tables();
if (check)
check_blocks();
else if (listfile)
get_list_blocks(listfile);
#ifdef BB_FEATURE_MINIX2
if (version2) {
make_root_inode2();
make_bad_inode2();
} else
#endif
{
make_root_inode();
make_bad_inode();
}
mark_good_blocks();
write_tables();
return( 0);
}