mkfs_ext2: new applet by Vladimir
authorVladimir Dronnikov <dronnikov@gmail.com>
Sat, 17 Oct 2009 19:38:19 +0000 (21:38 +0200)
committerDenys Vlasenko <vda.linux@googlemail.com>
Sat, 17 Oct 2009 19:38:19 +0000 (21:38 +0200)
function                                             old     new   delta
mkfs_ext2_main                                         -    2011   +2011
static.supers                                          -     184    +184
allocate                                               -     118    +118
packed_usage                                       26692   26791     +99
PUT                                                    -      52     +52
has_super                                              -      28     +28
applet_names                                        2143    2160     +17
applet_main                                         1264    1272      +8
applet_nameofs                                       632     636      +4
applet_install_loc                                   158     159      +1
------------------------------------------------------------------------------
(add/remove: 5/0 grow/shrink: 5/0 up/down: 2522/0)           Total: 2522 bytes

Signed-off-by: Vladimir Dronnikov <dronnikov@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
include/applets.h
include/usage.h
util-linux/Config.in
util-linux/Kbuild
util-linux/mkfs_ext2.c [new file with mode: 0644]

index 9b6b54e63a45feaa1920fe782d9bde061d3d6d2d..9ca84623b5169904f103bb5a54d7eb65576b32ec 100644 (file)
@@ -258,9 +258,9 @@ IF_MESG(APPLET(mesg, _BB_DIR_USR_BIN, _BB_SUID_DROP))
 IF_MICROCOM(APPLET(microcom, _BB_DIR_USR_BIN, _BB_SUID_DROP))
 IF_MKDIR(APPLET_NOFORK(mkdir, mkdir, _BB_DIR_BIN, _BB_SUID_DROP, mkdir))
 IF_MKFS_VFAT(APPLET_ODDNAME(mkdosfs, mkfs_vfat, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_vfat))
-//IF_MKE2FS(APPLET(mke2fs, _BB_DIR_SBIN, _BB_SUID_DROP))
+IF_MKFS_EXT2(APPLET_ODDNAME(mke2fs, mkfs_ext2, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_ext2))
 IF_MKFIFO(APPLET(mkfifo, _BB_DIR_USR_BIN, _BB_SUID_DROP))
-//IF_MKE2FS(APPLET_ODDNAME(mkfs.ext2, mke2fs, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_ext2))
+IF_MKFS_EXT2(APPLET_ODDNAME(mkfs.ext2, mkfs_ext2, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_ext2))
 //IF_MKE2FS(APPLET_ODDNAME(mkfs.ext3, mke2fs, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_ext3))
 IF_MKFS_MINIX(APPLET_ODDNAME(mkfs.minix, mkfs_minix, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_minix))
 IF_MKFS_VFAT(APPLET_ODDNAME(mkfs.vfat, mkfs_vfat, _BB_DIR_SBIN, _BB_SUID_DROP, mkfs_vfat))
index 3cb05e0904c3d81ed6a511160a8e4f06617b04f2..0215fa5bfc9cb850b657bc9162ddf2ad92fa39cb 100644 (file)
        "/tmp/foo/bar/baz: No such file or directory\n" \
        "$ mkdir -p /tmp/foo/bar/baz\n"
 
-#define mke2fs_trivial_usage \
-       "[-c|-l filename] [-b block-size] [-f fragment-size] [-g blocks-per-group] " \
-       "[-i bytes-per-inode] [-j] [-J journal-options] [-N number-of-inodes] [-n] " \
-       "[-m reserved-blocks-percentage] [-o creator-os] [-O feature[,...]] [-q] " \
-       "[r fs-revision-level] [-E extended-options] [-v] [-F] [-L volume-label] " \
-       "[-M last-mounted-directory] [-S] [-T filesystem-type] " \
-       "device [blocks-count]"
-#define mke2fs_full_usage "\n\n" \
-       "       -b size         Block size in bytes" \
-     "\n       -c              Check for bad blocks before creating" \
-     "\n       -E opts         Set extended options" \
-     "\n       -f size         Fragment size in bytes" \
-     "\n       -F              Force (ignore sanity checks)" \
-     "\n       -g num          Number of blocks in a block group" \
-     "\n       -i ratio        The bytes/inode ratio" \
-     "\n       -j              Create a journal (ext3)" \
-     "\n       -J opts         Set journal options (size/device)" \
-     "\n       -l file         Read bad blocks list from file" \
-     "\n       -L lbl          Set the volume label" \
-     "\n       -m percent      Percent of fs blocks to reserve for admin" \
-     "\n       -M dir          Set last mounted directory" \
-     "\n       -n              Don't actually create anything" \
-     "\n       -N num          Number of inodes to create" \
-     "\n       -o os           Set the 'creator os' field" \
-     "\n       -O features     Dir_index/filetype/has_journal/journal_dev/sparse_super" \
-     "\n       -q              Quiet" \
-     "\n       -r rev          Set filesystem revision" \
-     "\n       -S              Write superblock and group descriptors only" \
-     "\n       -T fs-type      Set usage type (news/largefile/largefile4)" \
-     "\n       -v              Verbose" \
-
 #define mkfifo_trivial_usage \
        "[OPTIONS] name"
 #define mkfifo_full_usage "\n\n" \
      "\n       -Z      Set security context" \
        )
 
+#define mkfs_ext2_trivial_usage \
+       /* "[-c|-l filename] " */ \
+       "[-b BLK_SIZE] " \
+       /* "[-f fragment-size] [-g blocks-per-group] " */ \
+       "[-i BYTES_PER_INODE] " \
+       /* "[-j] [-J journal-options] [-N number-of-inodes] [-n] " */ \
+       "[-m RESERVED_PERCENT] " \
+       /* "[-o creator-os] [-O feature[,...]] [-q] " */ \
+       /* "[r fs-revision-level] [-E extended-options] [-v] [-F] " */ \
+       "[-L LABEL] " \
+       /* "[-M last-mounted-directory] [-S] [-T filesystem-type] " */ \
+       "DEVICE [BLK_COUNT]"
+#define mkfs_ext2_full_usage "\n" \
+     "\n       -b BLK_SIZE     Block size in bytes" \
+/*   "\n       -c              Check for bad blocks before creating" */ \
+/*   "\n       -E opts         Set extended options" */ \
+/*   "\n       -f size         Fragment size in bytes" */ \
+/*   "\n       -F              Force (ignore sanity checks)" */ \
+/*   "\n       -g num          Number of blocks in a block group" */ \
+     "\n       -i BYTES        The bytes/inode ratio" \
+/*   "\n       -j              Create a journal (ext3)" */ \
+/*   "\n       -J opts         Set journal options (size/device)" */ \
+/*   "\n       -l file         Read bad blocks list from file" */ \
+     "\n       -L LABEL        Set the volume label" \
+     "\n       -m PERCENT      Percent of blocks to reserve for admin" \
+/*   "\n       -M dir          Set last mounted directory" */ \
+/*   "\n       -n              Don't actually create anything" */ \
+/*   "\n       -N num          Number of inodes to create" */ \
+/*   "\n       -o os           Set the 'creator os' field" */ \
+/*   "\n       -O features     Dir_index/filetype/has_journal/journal_dev/sparse_super" */ \
+/*   "\n       -q              Quiet" */ \
+/*   "\n       -r rev          Set filesystem revision" */ \
+/*   "\n       -S              Write superblock and group descriptors only" */ \
+/*   "\n       -T fs-type      Set usage type (news/largefile/largefile4)" */ \
+/*   "\n       -v              Verbose" */ \
+
 #define mkfs_minix_trivial_usage \
        "[-c | -l filename] [-nXX] [-iXX] /dev/name [blocks]"
 #define mkfs_minix_full_usage "\n\n" \
index 7cf17575c21e46665a6f11355983bf3f556b5494..578d059c12ce0c6a7cf62a90f499cc5879751ba5 100644 (file)
@@ -239,6 +239,12 @@ config MKFS_VFAT
        help
          Utility to create FAT32 filesystems.
 
+config MKFS_EXT2
+       bool "mkfs_ext2"
+       default n
+       help
+         Utility to create EXT2 filesystems.
+
 config GETOPT
        bool "getopt"
        default n
index eaad3319d13a437b76aeefb8fa4ff41e0191dd48..7befe06787ce1066c569e529439794cab5856244 100644 (file)
@@ -22,6 +22,7 @@ lib-$(CONFIG_IPCRM)             += ipcrm.o
 lib-$(CONFIG_IPCS)              += ipcs.o
 lib-$(CONFIG_LOSETUP)           += losetup.o
 lib-$(CONFIG_MDEV)              += mdev.o
+lib-$(CONFIG_MKFS_EXT2)         += mkfs_ext2.o
 lib-$(CONFIG_MKFS_MINIX)        += mkfs_minix.o
 lib-$(CONFIG_MKFS_VFAT)         += mkfs_vfat.o
 lib-$(CONFIG_MKSWAP)            += mkswap.o
diff --git a/util-linux/mkfs_ext2.c b/util-linux/mkfs_ext2.c
new file mode 100644 (file)
index 0000000..0f5e014
--- /dev/null
@@ -0,0 +1,479 @@
+/* vi: set sw=4 ts=4: */
+/*
+ * mkfs_ext2: utility to create EXT2 filesystem
+ * inspired by genext2fs
+ *
+ * Busybox'ed (2009) by Vladimir Dronnikov <dronnikov@gmail.com>
+ *
+ * Licensed under GPLv2, see file LICENSE in this tarball for details.
+ */
+#include "libbb.h"
+#include <linux/fs.h>
+#include <linux/ext2_fs.h>
+#include "volume_id/volume_id_internal.h"
+
+#define        ENABLE_FEATURE_MKFS_EXT2_RESERVED_GDT 0
+#define        ENABLE_FEATURE_MKFS_EXT2_DIR_INDEX 1
+
+// from e2fsprogs
+#define s_reserved_gdt_blocks s_padding1
+#define s_mkfs_time           s_reserved[0]
+#define s_flags               s_reserved[22]
+#define EXT2_HASH_HALF_MD4     1
+#define EXT2_FLAGS_SIGNED_HASH 0x0001
+
+// whiteout: for writable overlays
+//#define LINUX_S_IFWHT                  0160000
+//#define EXT2_FEATURE_INCOMPAT_WHITEOUT 0x0020
+
+// storage helper
+void BUG_unsupported_field_size(void);
+#define STORE_LE(field, value) \
+do { \
+       if (sizeof(field) == 4) \
+               field = cpu_to_le32(value); \
+       else if (sizeof(field) == 2) \
+               field = cpu_to_le16(value); \
+       else if (sizeof(field) == 1) \
+               field = (value); \
+       else \
+               BUG_unsupported_field_size(); \
+} while (0)
+
+// All fields are little-endian
+struct ext2_dir {
+       uint32_t inode1;
+       uint16_t rec_len1;
+       uint8_t  name_len1;
+       uint8_t  file_type1;
+       char     name1[4];
+       uint32_t inode2;
+       uint16_t rec_len2;
+       uint8_t  name_len2;
+       uint8_t  file_type2;
+       char     name2[4];
+       uint32_t inode3;
+       uint16_t rec_len3;
+       uint8_t  name_len3;
+       uint8_t  file_type3;
+       char     name3[12];
+};
+
+static inline int int_log2(int arg)
+{
+       int r = 0;
+       while ((arg >>= 1) != 0)
+               r++;
+       return r;
+}
+
+// taken from mkfs_minix.c. libbb candidate?
+static ALWAYS_INLINE unsigned div_roundup(uint32_t size, uint32_t n)
+{
+       return (size + n-1) / n;
+}
+
+static void allocate(uint8_t *bitmap, uint32_t blocksize, uint32_t start, uint32_t end)
+{
+       uint32_t i;
+       memset(bitmap, 0, blocksize);
+       i = start/8;
+       memset(bitmap, 0xFF, i);
+       bitmap[i] = 0xFF >> (8-(start&7));
+//bb_info_msg("ALLOC: [%u][%u][%u]: [%u]:=[%x]", blocksize, start, end, blocksize - end/8 - 1, (uint8_t)(0xFF << (8-(end&7))));
+       i = end/8;
+       bitmap[blocksize - i - 1] = 0xFF << (8-(end&7));
+       memset(bitmap + blocksize - i, 0xFF, i); // N.B. no overflow here!
+}
+
+#if 0
+// TODO: get rid of FPU
+static bool is_power_of(uint32_t x, uint16_t n)
+{
+//     return (!(x % n) && is_power_of(x / n, n));
+       double z = logf(x)/logf(n);
+       return (z == (int)z);
+}
+
+static bool has_super(uint32_t x)
+{
+       return (0 == x || 1 == x || is_power_of(x, 3) || is_power_of(x, 5) || is_power_of(x, 7));
+}
+
+#else
+
+static bool has_super(uint32_t x)
+{
+       static const uint32_t supers[] = {
+               0, 1, 3, 5, 7, 9, 25, 27, 49, 81, 125, 243, 343, 625, 729,
+               2187, 2401, 3125, 6561, 15625, 16807, 19683, 59049, 78125,
+               117649, 177147, 390625, 531441, 823543, 1594323, 1953125,
+               4782969, 5764801, 9765625, 14348907, 40353607, 43046721,
+               48828125, 129140163, 244140625, 282475249, 387420489,
+               1162261467, 1220703125, 1977326743, 3486784401/* >2^31 */,
+       };
+       for (int i = sizeof(supers)/sizeof(supers[0]); --i >= 0; )
+               if (x == supers[i])
+                       return 1;
+       return 0;
+}
+#endif
+
+/* Standard mke2fs 1.41.9:
+ * Usage: mke2fs [-c|-l filename] [-b block-size] [-f fragment-size]
+ *     [-i bytes-per-inode] [-I inode-size] [-J journal-options]
+ *     [-G meta group size] [-N number-of-inodes]
+ *     [-m reserved-blocks-percentage] [-o creator-os]
+ *     [-g blocks-per-group] [-L volume-label] [-M last-mounted-directory]
+ *     [-O feature[,...]] [-r fs-revision] [-E extended-option[,...]]
+ *     [-T fs-type] [-U UUID] [-jnqvFSV] device [blocks-count]
+*/
+// N.B. not commented below options are taken and silently ignored
+enum {
+       OPT_c = 1 << 0,
+       OPT_l = 1 << 1,
+       OPT_b = 1 << 2,         // block size, in bytes
+       OPT_f = 1 << 3,
+       OPT_i = 1 << 4,         // bytes per inode
+       OPT_I = 1 << 5,
+       OPT_J = 1 << 6,
+       OPT_G = 1 << 7,
+       OPT_N = 1 << 8,
+       OPT_m = 1 << 9,         // percentage of blocks reserved for superuser
+       OPT_o = 1 << 10,
+       OPT_g = 1 << 11,
+       OPT_L = 1 << 12,        // label
+       OPT_M = 1 << 13,
+       OPT_O = 1 << 14,
+       OPT_r = 1 << 15,
+       OPT_E = 1 << 16,
+       OPT_T = 1 << 17,
+       OPT_U = 1 << 18,
+       OPT_j = 1 << 19,
+       OPT_n = 1 << 20,
+       OPT_q = 1 << 21,
+       OPT_v = 1 << 22,
+       OPT_F = 1 << 23,
+       OPT_S = 1 << 24,
+       //OPT_V = 1 << 25,      // -V version. bbox applets don't support that
+};
+
+#define fd 3   /* predefined output descriptor */
+
+static void PUT(uint64_t off, void *buf, uint32_t size)
+{
+       if (!(option_mask32 & OPT_n)) {
+               xlseek(fd, off, SEEK_SET);
+               xwrite(fd, buf, size);
+       }
+}
+
+int mkfs_ext2_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
+int mkfs_ext2_main(int argc UNUSED_PARAM, char **argv)
+{
+       unsigned i, pos, n;
+       unsigned bs, blocksize;
+       unsigned nreserved = 5;
+       uint32_t nblocks;
+       uint32_t ngroups;
+       unsigned bytes_per_inode;
+       uint32_t nblocks_per_group;
+       uint32_t first_data_block;
+       uint32_t ninodes;
+       uint32_t ninodes_per_group;
+       uint32_t gdtsz, rgdtsz, itsz;
+       time_t timestamp;
+       unsigned opts;
+       const char *label;
+       struct stat st;
+       struct ext2_super_block *sb; // superblock
+       struct ext2_group_desc *gd; // group descriptors
+       struct ext2_inode *inode;
+       struct ext2_dir *dir;
+       uint8_t *buf;
+
+       bs = EXT2_MIN_BLOCK_SIZE;
+       opt_complementary = "-1:b+:m+:i+";
+       opts = getopt32(argv, "cl:b:f:i:I:J:G:N:m:o:g:L:M:O:r:E:T:U:jnqvFS",
+               NULL, &bs, NULL, &bytes_per_inode, NULL, NULL, NULL, NULL,
+               &nreserved, NULL, NULL, &label, NULL, NULL, NULL, NULL, NULL, NULL);
+       argv += optind; // argv[0] -- device
+
+       // block size minimax, block size is a multiple of minimum
+       blocksize = bs;
+       if (blocksize < EXT2_MIN_BLOCK_SIZE
+        || blocksize > EXT2_MAX_BLOCK_SIZE
+        || (blocksize & (blocksize - 1)) // not power of 2
+       ) {
+               bb_error_msg_and_die("-%c is bad", 'b');
+       }
+
+       // reserved blocks count
+       if (nreserved > 50)
+               bb_error_msg_and_die("-%c is bad", 'm');
+
+       // check the device is a block device
+       xstat(argv[0], &st);
+       if (!S_ISBLK(st.st_mode) && !(opts & OPT_F))
+               bb_error_msg_and_die("not a block device");
+
+       // check if it is mounted
+       // N.B. what if we format a file? find_mount_point will return false negative since
+       // it is loop block device which mounted!
+       if (find_mount_point(argv[0], 0))
+               bb_error_msg_and_die("can't format mounted filesystem");
+
+       // TODO: 5?/5 WE MUST NOT DEPEND ON WHETHER DEVICE IS /dev/zero 'ed OR NOT
+       // TODO: 3/5 refuse if mounted
+       // TODO: 4/5 compat options
+       // TODO: 1/5 sanity checks
+       // TODO: 0/5 more verbose error messages
+       // TODO: 0/5 info printing
+       // TODO: 2/5 bigendianness! Spot where it comes to play! sb->, gd->
+       // TODO: 2/5 reserved GDT: how to mark but not allocate?
+       // TODO: 0/5 dir_index?
+
+       // fill the superblock
+       sb = xzalloc(blocksize);
+       sb->s_rev_level = 1; // revision 1 filesystem
+       sb->s_magic = EXT2_SUPER_MAGIC;
+       sb->s_inode_size = sizeof(*inode);
+       sb->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
+       sb->s_log_block_size = sb->s_log_frag_size = int_log2(blocksize >> EXT2_MIN_BLOCK_LOG_SIZE);
+       // first 1024 bytes of the device are for boot record. If block size is 1024 bytes, then
+       // the first block available for data is 1, otherwise 0
+       first_data_block = sb->s_first_data_block = (EXT2_MIN_BLOCK_SIZE == blocksize);
+       // block and inode bitmaps occupy no more than one block, so maximum number of blocks is
+       // number of bits in one block, i.e. 8*blocksize
+       nblocks_per_group = sb->s_blocks_per_group = sb->s_frags_per_group = sb->s_inodes_per_group = 8*blocksize;
+       timestamp = time(NULL);
+       sb->s_mkfs_time = sb->s_wtime = sb->s_lastcheck = timestamp;
+       sb->s_state = 1;
+       sb->s_creator_os = EXT2_OS_LINUX;
+       sb->s_max_mnt_count = EXT2_DFL_MAX_MNT_COUNT;
+       sb->s_checkinterval = 24*60*60 * 180; // 180 days
+       sb->s_errors = EXT2_ERRORS_DEFAULT;
+       sb->s_feature_compat = EXT2_FEATURE_COMPAT_SUPP
+               | (EXT2_FEATURE_COMPAT_RESIZE_INO * ENABLE_FEATURE_MKFS_EXT2_RESERVED_GDT)
+               | (EXT2_FEATURE_COMPAT_DIR_INDEX * ENABLE_FEATURE_MKFS_EXT2_DIR_INDEX)
+               ;
+       // e2fsprogs-1.41.9 doesn't like EXT2_FEATURE_INCOMPAT_WHITEOUT
+       sb->s_feature_incompat = EXT2_FEATURE_INCOMPAT_FILETYPE;// | EXT2_FEATURE_INCOMPAT_WHITEOUT;
+       sb->s_feature_ro_compat = EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER;
+       sb->s_flags = EXT2_FLAGS_SIGNED_HASH * ENABLE_FEATURE_MKFS_EXT2_DIR_INDEX;
+       generate_uuid(sb->s_uuid);
+#if ENABLE_FEATURE_MKFS_EXT2_DIR_INDEX
+       sb->s_def_hash_version = EXT2_HASH_HALF_MD4;
+       generate_uuid((uint8_t *)sb->s_hash_seed);
+#endif
+       /*
+        * From e2fsprogs: add "jitter" to the superblock's check interval so that we
+        * don't check all the filesystems at the same time.  We use a
+        * kludgy hack of using the UUID to derive a random jitter value.
+        */
+       for (i = 0, n = 0; i < sizeof(sb->s_uuid); i++)
+               n += sb->s_uuid[i];
+       sb->s_max_mnt_count += n % EXT2_DFL_MAX_MNT_COUNT;
+
+       // open the device, get number of blocks
+       xmove_fd(xopen3(argv[0], O_WRONLY | O_CREAT, 0666), fd);
+       if (argv[1]) {
+               nblocks = xatou(argv[1]);
+       } else {
+               nblocks = ((uoff_t)xlseek(fd, 0, SEEK_END)) / blocksize;
+               xlseek(fd, 0, SEEK_SET);
+       }
+       sb->s_blocks_count = nblocks;
+
+       // nblocks is the total number of blocks in the filesystem
+       if (nblocks < 8)
+               bb_error_msg_and_die("nblocks");
+       // reserve blocks for superuser
+       sb->s_r_blocks_count = ((uint64_t) nblocks * nreserved) / 100;
+
+       // N.B. a block group can have no more than nblocks_per_group blocks
+       ngroups = div_roundup(nblocks - first_data_block, nblocks_per_group);
+       if (0 == ngroups)
+               bb_error_msg_and_die("ngroups");
+       gdtsz = div_roundup(ngroups, EXT2_DESC_PER_BLOCK(sb));
+       /*
+        * From e2fsprogs: Calculate the number of GDT blocks to reserve for online
+        * filesystem growth.
+        * The absolute maximum number of GDT blocks we can reserve is determined by
+        * the number of block pointers that can fit into a single block.
+        */
+       /* We set it at 1024x the current filesystem size, or
+        * the upper block count limit (2^32), whichever is lower.
+        */
+#if ENABLE_FEATURE_MKFS_EXT2_RESERVED_GDT
+       rgdtsz = 0xFFFFFFFF; // maximum block number
+       if (nblocks < rgdtsz / 1024)
+               rgdtsz = nblocks * 1024;
+       rgdtsz = div_roundup(rgdtsz - first_data_block, nblocks_per_group);
+       rgdtsz = div_roundup(rgdtsz, EXT2_DESC_PER_BLOCK(sb)) - gdtsz;
+       if (rgdtsz > EXT2_ADDR_PER_BLOCK(sb))
+               rgdtsz = EXT2_ADDR_PER_BLOCK(sb);
+       sb->s_reserved_gdt_blocks = rgdtsz;
+       //bb_info_msg("RSRVD[%u]", n);
+#else
+       rgdtsz = 0;
+#endif
+
+       // ninodes is the total number of inodes (files) in the file system
+       if (!(opts & OPT_i)) {
+               bytes_per_inode = 16384;
+               if (nblocks < 512*1024)
+                       bytes_per_inode = 4096;
+               if (nblocks < 3*1024)
+                       bytes_per_inode = 8192;
+       }
+       ninodes = nblocks / (bytes_per_inode / blocksize);
+       if (ninodes < EXT2_GOOD_OLD_FIRST_INO+1)
+               ninodes = EXT2_GOOD_OLD_FIRST_INO+1;
+       ninodes_per_group = div_roundup(ninodes, ngroups);
+       if (ninodes_per_group < 16)
+               ninodes_per_group = 16; // minimum number because the first 10 are reserved
+       // N.B. a block group can have no more than 8*blocksize = sb->s_inodes_per_group inodes
+       if (ninodes_per_group > sb->s_inodes_per_group)
+               ninodes_per_group = sb->s_inodes_per_group;
+       // adjust inodes per group so they completely fill the inode table blocks in the descriptor
+       ninodes_per_group = ((div_roundup(ninodes_per_group * EXT2_INODE_SIZE(sb), blocksize) * blocksize) / EXT2_INODE_SIZE(sb));
+       // make sure the number of inodes per group is a multiple of 8
+       ninodes_per_group &= ~7;
+       sb->s_inodes_per_group = ninodes_per_group;// = div_roundup(ninodes_per_group * sb->s_inode_size, blocksize);
+       // total ninodes
+       ninodes = sb->s_inodes_count = ninodes_per_group * ngroups;
+
+       itsz = ninodes_per_group * sb->s_inode_size / blocksize;
+       sb->s_free_inodes_count = sb->s_inodes_count - EXT2_GOOD_OLD_FIRST_INO;
+
+       // write the label, if any
+       if (opts & OPT_L)
+               safe_strncpy((char *)sb->s_volume_name, label, sizeof(sb->s_volume_name));
+
+       // fill group descriptors
+       gd = xzalloc((gdtsz + rgdtsz) * blocksize);
+       sb->s_free_blocks_count = 0;
+       for (i = 0, pos = first_data_block, n = nblocks;
+               i < ngroups;
+               i++, pos += nblocks_per_group, n -= nblocks_per_group
+       ) {
+               uint32_t overhead = pos + has_super(i) * (1/*sb*/ + gdtsz + rgdtsz);
+               gd[i].bg_block_bitmap = overhead + 0;
+               gd[i].bg_inode_bitmap = overhead + 1;
+               gd[i].bg_inode_table  = overhead + 2;
+               overhead = overhead - pos + 1/*bbmp*/ + 1/*ibmp*/ + itsz;
+               gd[i].bg_free_inodes_count = ninodes_per_group;
+               // N.B. both root and lost+found dirs are within the first block group, thus +2
+               //gd[i].bg_used_dirs_count = 0;
+               if (0 == i) {
+                       overhead += 2;
+                       gd[i].bg_used_dirs_count = 2;
+                       gd[i].bg_free_inodes_count -= EXT2_GOOD_OLD_FIRST_INO;
+               }
+               // N.B. the following is pure heuristics!
+               // Likely to cope with 1024-byte blocks, when first block is for boot sectors
+               if (ngroups-1 == i) {
+                       overhead += first_data_block;
+               }
+               gd[i].bg_free_blocks_count = (n < nblocks_per_group ? n : nblocks_per_group) - overhead;
+               sb->s_free_blocks_count += gd[i].bg_free_blocks_count;
+       }
+       STORE_LE(sb->s_free_blocks_count, sb->s_free_blocks_count);
+
+       // dump filesystem skeleton structures
+       buf = xmalloc(blocksize);
+       for (i = 0, pos = first_data_block; i < ngroups; i++, pos += nblocks_per_group) {
+               uint32_t overhead = has_super(i) * (1/*sb*/ + gdtsz + rgdtsz);
+               uint32_t start;// = has_super(i) * (1/*sb*/ + gdtsz + rgdtsz);
+               uint32_t end;
+
+               // dump superblock and group descriptors and their backups
+               if (overhead) { // N.B. in fact, we want (has_super(i)) condition, but it is equal to (overhead != 0) and is cheaper
+//bb_info_msg("SUPER@[%d]", pos);
+                       // N.B. 1024 byte blocks are special
+                       PUT(blocksize * pos + 1024 * (0 == i && 0 == first_data_block), sb, blocksize);
+                       PUT(blocksize * pos + blocksize, gd, (gdtsz + rgdtsz) * blocksize);
+               }
+
+               start = overhead + 1/*bbmp*/ + 1/*ibmp*/ + itsz + (0 == i) * 2; // +2: /, /lost+found
+               end = nblocks_per_group - (start + gd[i].bg_free_blocks_count);
+               // mark preallocated blocks as allocated
+               allocate(buf, blocksize, start, end);
+               // dump block bitmap
+               PUT((pos + overhead) * blocksize, buf, blocksize);
+
+               // mark preallocated inodes as allocated
+               allocate(buf, blocksize,
+                       ninodes_per_group - gd[i].bg_free_inodes_count,
+                       8*blocksize - ninodes_per_group
+               );
+               // dump inode bitmap
+               PUT((pos + overhead + 1) * blocksize, buf, blocksize);
+
+       }
+
+       // zero boot sectors
+       memset(buf, 0, blocksize);
+       PUT(0, buf, 1024); // N.B. 1024 <= blocksize
+       // zero inode tables
+       for (i = 0; i < ngroups; ++i)
+               for (n = 0; n < itsz; ++n)
+                       PUT((gd[i].bg_inode_table + n) * blocksize, buf, blocksize);
+
+       // prepare directory inode
+       inode = (struct ext2_inode *)buf;
+       STORE_LE(inode->i_mode, S_IFDIR | S_IRWXU | S_IRGRP | S_IROTH | S_IXGRP | S_IXOTH);
+       inode->i_mtime = inode->i_atime = timestamp;
+       STORE_LE(inode->i_ctime, timestamp);
+       STORE_LE(inode->i_size, blocksize);
+       // N.B. inode->i_blocks stores the number of 512 byte data blocks. Why on Earth?!
+       STORE_LE(inode->i_blocks, blocksize / 512);
+
+       // dump root dir inode
+       STORE_LE(inode->i_links_count, 3); // "/.", "/..", "/lost+found/.." point to this inode
+       STORE_LE(inode->i_block[0], gd[0].bg_inode_table + itsz);
+       PUT(gd[0].bg_inode_table * blocksize + (EXT2_ROOT_INO-1) * sizeof(*inode), buf, sizeof(*inode));
+
+       // dump lost+found dir inode
+       STORE_LE(inode->i_links_count, 2); // both "/lost+found" and "/lost+found/." point to this inode
+       STORE_LE(inode->i_block[0], inode->i_block[0]+1); // use next block //= gd[0].bg_inode_table + itsz + 1;
+       PUT(gd[0].bg_inode_table * blocksize + (EXT2_GOOD_OLD_FIRST_INO-1) * sizeof(*inode), buf, sizeof(*inode));
+
+       // dump directories
+       memset(buf, 0, blocksize);
+       dir = (struct ext2_dir *)buf;
+
+       // dump lost+found dir block
+       STORE_LE(dir->inode1, EXT2_GOOD_OLD_FIRST_INO);
+       STORE_LE(dir->rec_len1, 12);
+       STORE_LE(dir->name_len1, 1);
+       STORE_LE(dir->file_type1, EXT2_FT_DIR);
+       dir->name1[0] = '.';
+       STORE_LE(dir->inode2, EXT2_ROOT_INO);
+       STORE_LE(dir->rec_len2, blocksize - 12);
+       STORE_LE(dir->name_len2, 2);
+       STORE_LE(dir->file_type2, EXT2_FT_DIR);
+       dir->name2[0] = '.'; dir->name2[1] = '.';
+       PUT((gd[0].bg_inode_table + itsz + 1) * blocksize, buf, blocksize);
+
+       // dump root dir block
+       STORE_LE(dir->inode1, EXT2_ROOT_INO);
+       STORE_LE(dir->rec_len2, 12);
+       STORE_LE(dir->inode3, EXT2_GOOD_OLD_FIRST_INO);
+       STORE_LE(dir->rec_len3, blocksize - 12 - 12);
+       STORE_LE(dir->name_len3, 10);
+       STORE_LE(dir->file_type3, EXT2_FT_DIR);
+       strcpy(dir->name3, "lost+found");
+       PUT((gd[0].bg_inode_table + itsz + 0) * blocksize, buf, blocksize);
+
+       // cleanup
+       if (ENABLE_FEATURE_CLEAN_UP) {
+               free(buf);
+               free(gd);
+               free(sb);
+               close(fd);
+       }
+
+       return EXIT_SUCCESS;
+}