Index: HelenOS.config
===================================================================
--- HelenOS.config	(revision a2271a385d521c53eb7137d1142971451edc6aaa)
+++ HelenOS.config	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -88,4 +88,5 @@
 @ "tmpfs" TMPFS image
 @ "fat" FAT16 image
+@ "ext2fs" EXT2 image
 ! RDFMT (choice)
 
Index: boot/Makefile
===================================================================
--- boot/Makefile	(revision a2271a385d521c53eb7137d1142971451edc6aaa)
+++ boot/Makefile	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -50,4 +50,7 @@
 	$(MKFAT) 1048576 $(DIST_PATH) $@
 endif
+ifeq ($(RDFMT),ext2fs)
+	$(MKEXT2) 1048576 $(DIST_PATH) $@
+endif
 
 build_dist: clean_dist
Index: boot/Makefile.common
===================================================================
--- boot/Makefile.common	(revision a2271a385d521c53eb7137d1142971451edc6aaa)
+++ boot/Makefile.common	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -56,4 +56,5 @@
 MKTMPFS = $(TOOLS_PATH)/mktmpfs.py
 MKFAT = $(TOOLS_PATH)/mkfat.py
+MKEXT2 = $(TOOLS_PATH)/mkext2.py
 MKUIMAGE = $(TOOLS_PATH)/mkuimage.py
 
@@ -82,4 +83,8 @@
 ifeq ($(RDFMT),fat)
 	INIT_TASKS += $(USPACE_PATH)/srv/fs/fat/fat
+endif
+
+ifeq ($(RDFMT),ext2fs)
+	INIT_TASKS += $(USPACE_PATH)/srv/fs/ext2fs/ext2fs
 endif
 
Index: tools/imgutil.py
===================================================================
--- tools/imgutil.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
+++ tools/imgutil.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -0,0 +1,108 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2008 Martin Decky
+# Copyright (c) 2011 Martin Sucha
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#
+# - Redistributions of source code must retain the above copyright
+#   notice, this list of conditions and the following disclaimer.
+# - 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.
+# - The name of the author may not be used to endorse or promote products
+#   derived from this software without specific prior written permission.
+#
+# 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.
+#
+
+"""
+Utilities for filesystem image creators
+"""
+
+import os
+import stat
+
+exclude_names = set(['.svn', '.bzr'])
+
+def align_up(size, alignment):
+	"Return size aligned up to alignment"
+	
+	if (size % alignment == 0):
+		return size
+	
+	return ((size // alignment) + 1) * alignment
+
+def count_up(size, alignment):
+	"Return units necessary to fit the size"
+	
+	if (size % alignment == 0):
+		return (size // alignment)
+	
+	return ((size // alignment) + 1)
+
+def num_of_trailing_bin_zeros(num):
+	"Return number of trailing zeros in binary representation"
+	
+	i = 0
+	if (num == 0): raise ValueError()
+	while num & 1 == 0:
+		i += 1
+		num = num >> 1
+	return i
+
+def get_bit(number, n):
+	"Return True if n-th least-significant bit is set in the given number"
+	
+	return bool((number >> n) & 1)
+
+def set_bit(number, n):
+	"Return the number with n-th least-significant bit set"
+	
+	return number | (1 << n)
+
+class ItemToPack:
+	"Stores information about one directory item to be added to the image"
+	
+	def __init__(self, parent, name):
+		self.parent = parent
+		self.name = name
+		self.path = os.path.join(parent, name)
+		self.stat = os.stat(self.path)
+		self.is_dir = stat.S_ISDIR(self.stat.st_mode)
+		self.is_file = stat.S_ISREG(self.stat.st_mode)
+		self.size = self.stat.st_size
+
+def listdir_items(path):
+	"Return a list of items to be packed inside a fs image"
+	
+	for name in os.listdir(path):
+		if name in exclude_names:
+			continue
+		item = ItemToPack(path, name)
+		if not (item.is_dir or item.is_file):
+			continue
+		yield item
+
+def chunks(item, chunk_size):
+	"Iterate contents of a file in chunks of a given size"
+	
+	inf = open(item.path, 'rb')
+	rd = 0
+	while (rd < item.size):
+		data = bytes(inf.read(chunk_size))
+		yield data
+		rd += len(data)
+	inf.close()
Index: tools/mkext2.py
===================================================================
--- tools/mkext2.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
+++ tools/mkext2.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -0,0 +1,676 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2011 Martin Sucha
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#
+# - Redistributions of source code must retain the above copyright
+#   notice, this list of conditions and the following disclaimer.
+# - 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.
+# - The name of the author may not be used to endorse or promote products
+#   derived from this software without specific prior written permission.
+#
+# 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.
+#
+
+"""
+EXT2 creator
+"""
+
+import sys
+import os
+import xstruct
+import array
+import time
+import uuid
+from imgutil import *
+
+GDE_SIZE = 32
+
+STRUCT_DIR_ENTRY_HEAD = """little:
+	uint32_t inode       /* inode number */
+	uint16_t skip        /* byte offset to the next record */
+	uint8_t  name_length /* file attributes */
+	uint8_t  inode_type  /* type of the referenced inode */
+"""
+
+STRUCT_SUPERBLOCK = """little:
+	uint32_t total_inode_count    /* Total number of inodes */
+	uint32_t total_block_count    /* Total number of blocks */
+	uint32_t reserved_block_count /* Total number of reserved blocks */
+	uint32_t free_block_count     /* Total number of free blocks */
+	uint32_t free_inode_count     /* Total number of free inodes */
+	uint32_t first_block          /* Block containing the superblock */
+	uint32_t block_size_log2      /* log_2(block_size) */
+	int32_t  fragment_size_log2   /* log_2(fragment size) */
+	uint32_t blocks_per_group     /* Number of blocks in one block group */
+	uint32_t fragments_per_group  /* Number of fragments per block group */
+	uint32_t inodes_per_group     /* Number of inodes per block group */
+	uint32_t mount_time           /* Time when the filesystem was last mounted */
+	uint32_t write_time           /* Time when the filesystem was last written */
+	uint16_t mount_count          /* Mount count since last full filesystem check */
+	uint16_t max_mount_count      /* Number of mounts after which the fs must be checked */
+	uint16_t magic                /* Magic value */
+	uint16_t state                /* State (mounted/unmounted) */
+	uint16_t error_behavior       /* What to do when errors are encountered */
+	uint16_t rev_minor            /* Minor revision level */
+	uint32_t last_check_time      /* Unix time of last fs check */
+	uint32_t max_check_interval   /* Max unix time interval between checks */
+	uint32_t os                   /* OS that created the filesystem */
+	uint32_t rev_major            /* Major revision level */
+	padding[4] /* default reserved uid and gid */
+	
+	/* Following is for ext2 revision 1 only */
+	uint32_t first_inode
+	uint16_t inode_size
+	uint16_t block_group_number /* Block group where this SB is stored */
+	uint32_t features_compatible
+	uint32_t features_incompatible
+	uint32_t features_read_only
+	char     uuid[16]
+	char     volume_name[16]
+"""
+
+STRUCT_BLOCK_GROUP_DESCRIPTOR = """little:
+	uint32_t block_bitmap_block      /* Block ID for block bitmap */
+	uint32_t inode_bitmap_block      /* Block ID for inode bitmap */
+	uint32_t inode_table_first_block /* Block ID of first block of inode table */
+	uint16_t free_block_count        /* Count of free blocks */
+	uint16_t free_inode_count        /* Count of free inodes */
+	uint16_t directory_inode_count   /* Number of inodes allocated to directories */
+"""
+
+STRUCT_INODE = """little:
+	uint16_t mode
+	uint16_t user_id
+	uint32_t size
+	uint32_t access_time
+	uint32_t creation_time
+	uint32_t modification_time
+	uint32_t deletion_time
+	uint16_t group_id
+	uint16_t usage_count /* Hard link count, when 0 the inode is to be freed */
+	uint32_t reserved_512_blocks /* Size of this inode in 512-byte blocks */
+	uint32_t flags
+	padding[4]
+	uint32_t direct_blocks[12] /* Direct block ids stored in this inode */
+	uint32_t indirect_blocks[3]
+	uint32_t version
+	uint32_t file_acl
+	uint32_t size_high /* For regular files in version >= 1, dir_acl if dir */
+	padding[6]
+	uint16_t mode_high /* Hurd only */
+	uint16_t user_id_high /* Linux/Hurd only */
+	uint16_t group_id_high /* Linux/Hurd only */
+"""
+
+# The following is here to handle byte-order conversion in indirect block lookups
+STRUCT_BLOCK_REFERENCE = """little:
+	uint32_t block_id /* Referenced block ID */
+"""
+
+class Filesystem:	
+	def __init__(self, filename, block_groups, blocks_per_group, inodes_per_group, block_size, inode_size, reserved_inode_count):
+		"Initialize the filesystem writer"
+		
+		outf = open(filename, "w+b")
+		# Set the correct size of the image, so that we can read arbitrary position
+		outf.truncate(block_size * blocks_per_group * block_groups)
+		self.outf = outf
+		self.uuid = uuid.uuid4()
+		self.block_size = block_size
+		self.inode_size = inode_size
+		self.block_groups = block_groups
+		self.blocks_per_group = blocks_per_group
+		self.inodes_per_group = inodes_per_group
+		self.reserved_inode_count = reserved_inode_count
+		self.gdt_blocks = count_up(block_groups * GDE_SIZE, block_size)
+		self.inode_table_blocks_per_group = (inodes_per_group * inode_size) // block_size
+		self.inode_bitmap_blocks_per_group = count_up(inodes_per_group // 8, block_size)
+		self.block_bitmap_blocks_per_group = count_up(blocks_per_group // 8, block_size)
+		self.total_block_count = self.blocks_per_group * self.block_groups
+		self.total_inode_count = self.inodes_per_group * self.block_groups
+		self.inode_table_blocks = count_up(self.total_inode_count * inode_size, block_size)
+		self.inodes = {}
+		self.superblock_at_block = 2047 // block_size
+		self.block_ids_per_block = self.block_size // 4
+		self.indirect_limits = [12, None, None, None]
+		self.indirect_blocks_per_level = [1, None, None, None]
+		for i in range(1,4):
+			self.indirect_blocks_per_level[i] = self.indirect_blocks_per_level[i-1] * self.block_ids_per_block
+			self.indirect_limits[i] = self.indirect_limits[i-1] + self.indirect_blocks_per_level[i]
+		self.block_allocator = Allocator(0, self.total_block_count)
+		self.inode_allocator = Allocator(1, self.total_inode_count)
+		self.init_gdt()
+		# Set the callbacks after GDT has been initialized
+		self.block_allocator.mark_cb = self.mark_block_cb
+		self.inode_allocator.mark_cb = self.mark_inode_cb
+		self.block_allocator.mark_used_all(range(self.superblock_at_block))
+		self.inode_allocator.mark_used_all(range(1, self.reserved_inode_count + 1))
+		self.root_inode = Inode(self, 2, Inode.TYPE_DIR)
+		self.gdt[0].directory_inode_count += 1
+		self.lost_plus_found = Inode(self, self.inode_allocator.alloc(directory=True), Inode.TYPE_DIR)
+		lpf_dir = DirWriter(self.lost_plus_found)
+		lpf_dir.add(self.lost_plus_found.as_dirent('.'))
+		lpf_dir.add(self.root_inode.as_dirent('..'))
+		lpf_dir.finish()
+	
+	def init_gdt(self):
+		"Initialize block group descriptor table"
+		
+		self.superblock_positions = []
+		self.gdt = []
+		consumed_blocks_per_group = (1 + self.gdt_blocks +
+			    self.inode_bitmap_blocks_per_group +
+			    self.block_bitmap_blocks_per_group +
+			    self.inode_table_blocks_per_group)
+		initial_free_blocks = self.blocks_per_group - consumed_blocks_per_group
+		for bg in range(self.block_groups):
+			base = bg * self.blocks_per_group
+			if (bg == 0):
+				base = self.superblock_at_block
+				self.superblock_positions.append(1024)
+			else:
+				self.superblock_positions.append(base * self.block_size)
+			self.block_allocator.mark_used_all(range(base, base+consumed_blocks_per_group))
+			pos = base + 1 + self.gdt_blocks
+			gde = xstruct.create(STRUCT_BLOCK_GROUP_DESCRIPTOR)
+			gde.block_bitmap_block = pos
+			pos += self.block_bitmap_blocks_per_group
+			gde.inode_bitmap_block = pos
+			pos += self.inode_bitmap_blocks_per_group
+			gde.inode_table_first_block = pos
+			gde.free_block_count = initial_free_blocks
+			gde.free_inode_count = self.inodes_per_group
+			gde.directory_inode_count = 0
+			self.gdt.append(gde)
+	
+	def mark_block_cb(self, block):
+		"Called after a block has been allocated"
+		
+		self.gdt[block // self.blocks_per_group].free_block_count -= 1
+	
+	def mark_inode_cb(self, index, directory=False):
+		"Called after an inode has been allocated"
+		
+		index -= 1
+		gde = self.gdt[index // self.inodes_per_group]
+		gde.free_inode_count -= 1
+		if directory:
+			gde.directory_inode_count += 1		
+	
+	def seek_to_block(self, block, offset=0):
+		"Seek to offset bytes after the start of the given block"
+		
+		if offset < 0 or offset > self.block_size:
+			raise Exception("Invalid in-block offset")
+		self.outf.seek(block * self.block_size + offset)
+	
+	def seek_to_inode(self, index):
+		"Seek to the start of the inode structure for the inode number index"
+		
+		index -= 1
+		if index < 0:
+			raise Exception("Invalid inode number")
+		gde = self.gdt[index // self.inodes_per_group]
+		base_block = gde.inode_table_first_block
+		offset = (index % self.inodes_per_group) * self.inode_size
+		block = base_block + (offset // self.block_size)
+		self.seek_to_block(block, offset % self.block_size)	
+	
+	def subtree_add(self, inode, parent_inode, dirpath, is_root=False):
+		"Recursively add files to the filesystem"
+		
+		dir_writer = DirWriter(inode)	
+		dir_writer.add(inode.as_dirent('.'))
+		dir_writer.add(parent_inode.as_dirent('..'))
+		
+		if is_root:
+			dir_writer.add(self.lost_plus_found.as_dirent('lost+found'))
+
+		for item in listdir_items(dirpath):
+			newidx = self.inode_allocator.alloc(directory = item.is_dir)
+			if item.is_file:
+				child_inode = Inode(self, newidx, Inode.TYPE_FILE)
+				for data in chunks(item, self.block_size):
+					child_inode.write(data)
+			elif item.is_dir:
+				child_inode = Inode(self, newidx, Inode.TYPE_DIR)
+				self.subtree_add(child_inode, inode, item.path)
+		
+			dir_writer.add(child_inode.as_dirent(item.name))
+			self.write_inode(child_inode)
+
+		dir_writer.finish()
+	
+	def write_inode(self, inode):
+		"Write inode information into the inode table"
+		
+		self.seek_to_inode(inode.index)
+		self.outf.write(inode.pack())
+
+	def write_gdt(self):
+		"Write group descriptor table at the current file position"
+		
+		for gde in self.gdt:
+			data = bytes(gde.pack())
+			self.outf.write(data)
+			self.outf.seek(GDE_SIZE-len(data), os.SEEK_CUR)
+	
+	def write_superblock(self, block_group):
+		"Write superblock at the current position"
+		
+		sb = xstruct.create(STRUCT_SUPERBLOCK)
+		sb.total_inode_count = self.total_inode_count
+		sb.total_block_count = self.total_block_count
+		sb.reserved_block_count = 0
+		sb.free_block_count = self.block_allocator.free
+		sb.free_inode_count = self.inode_allocator.free
+		sb.first_block = self.superblock_at_block
+		sb.block_size_log2 = num_of_trailing_bin_zeros(self.block_size) - 10
+		sb.fragment_size_log2 = sb.block_size_log2
+		sb.blocks_per_group = self.blocks_per_group
+		sb.fragments_per_group = self.blocks_per_group
+		sb.inodes_per_group = self.inodes_per_group
+		curtime = int(time.time())
+		sb.mount_time = curtime
+		sb.write_time = curtime
+		sb.mount_count = 0
+		sb.max_mount_count = 21
+		sb.magic = 0xEF53
+		sb.state = 5 # clean
+		sb.error_behavior = 1 # continue on errors
+		sb.rev_minor = 0
+		sb.last_check_time = curtime
+		sb.max_check_interval = 15552000 # 6 months
+		sb.os = 0 # Linux
+		sb.rev_major = 1
+		sb.first_inode = self.reserved_inode_count + 1
+		sb.inode_size = self.inode_size
+		sb.block_group_number = block_group
+		sb.features_compatible = 0
+		sb.features_incompatible = 2 # filetype
+		sb.features_read_only = 0
+		sb.uuid = self.uuid.bytes_le
+		sb.volume_name = 'HelenOS rdimage\0'
+		self.outf.write(bytes(sb.pack()))
+	
+	def write_all_metadata(self):
+		"Write superblocks, block group tables, block and inode bitmaps"
+		
+		bbpg = self.blocks_per_group // 8
+		bipg = self.inodes_per_group // 8
+		def window(arr, index, size):
+			return arr[index * size:(index + 1) * size]
+		
+		for bg_index in xrange(len(self.gdt)):
+			sbpos = self.superblock_positions[bg_index]
+			sbblock = (sbpos + 1023) // self.block_size
+			gde = self.gdt[bg_index]
+			
+			self.outf.seek(sbpos)
+			self.write_superblock(bg_index)
+			
+			self.seek_to_block(sbblock+1)
+			self.write_gdt()
+			
+			self.seek_to_block(gde.block_bitmap_block)
+			self.outf.write(window(self.block_allocator.bitmap, bg_index, bbpg))
+			
+			self.seek_to_block(gde.inode_bitmap_block)
+			self.outf.write(window(self.inode_allocator.bitmap, bg_index, bipg))
+	
+	def close(self):
+		"Write all remaining data to the filesystem and close the file"
+		
+		self.write_inode(self.root_inode)
+		self.write_inode(self.lost_plus_found)
+		self.write_all_metadata()
+		self.outf.close()
+
+class Allocator:
+	def __init__(self, base, count):
+		self.count = count
+		self.base = base
+		self.free = count
+		self.nextidx = 0
+		self.bitmap = array.array('B', [0] * (count // 8))
+		self.mark_cb = None
+	
+	def __contains__(self, item):
+		"Check if the item is already used"
+		
+		bitidx = item - self.base
+		return get_bit(self.bitmap[bitidx // 8], bitidx % 8)
+	
+	def alloc(self, **options):
+		"Allocate a new item"
+		
+		while self.nextidx < self.count and (self.base + self.nextidx) in self:
+			self.nextidx += 1
+		if self.nextidx == self.count:
+			raise Exception("No free item available")
+		item = self.base + self.nextidx
+		self.nextidx += 1
+		self.mark_used(item, **options)
+		return item
+	
+	def mark_used(self, item, **options):
+		"Mark the specified item as used"
+		
+		bitidx = item - self.base
+		if item in self:
+			raise Exception("Item already used: " + str(item))
+		self.free -= 1
+		index = bitidx // 8
+		self.bitmap[index] = set_bit(self.bitmap[index], bitidx % 8)
+		if self.mark_cb:
+			self.mark_cb(item, **options)
+	
+	def mark_used_all(self, items, **options):
+		"Mark all specified items as used"
+		
+		for item in items:
+			self.mark_used(item, **options)
+
+class Inode:
+	TYPE_FILE = 1
+	TYPE_DIR = 2
+	TYPE2MODE = {TYPE_FILE: 8, TYPE_DIR: 4}
+	
+	def __init__(self, fs, index, typ):
+		self.fs = fs
+		self.direct = [None] * 12
+		self.indirect = [None] * 3
+		self.pos = 0
+		self.size = 0
+		self.blocks = 0
+		self.index = index
+		self.type = typ
+		self.refcount = 0
+	
+	def as_dirent(self, name):
+		"Return a DirEntry corresponding to this inode"
+		self.refcount += 1
+		return DirEntry(name, self.index, self.type)
+	
+	def new_block(self, data=True):
+		"Get a new block index from allocator and count it here as belonging to the file"
+		
+		block = self.fs.block_allocator.alloc()
+		self.blocks += 1
+		return block
+	
+	def get_or_add_block(self, block):
+		"Get or add a real block to the file"
+		
+		if block < 12:
+			return self.get_or_add_block_direct(block)
+		return self.get_or_add_block_indirect(block)
+	
+	def get_or_add_block_direct(self, block):
+		"Get or add a real block to the file (direct blocks)"
+		
+		if self.direct[block] == None:
+			self.direct[block] = self.new_block()
+		return self.direct[block]
+	
+	def get_or_add_block_indirect(self, block):
+		"Get or add a real block to the file (indirect blocks)"
+		
+		# Determine the indirection level for the desired block
+		level = None
+		for i in range(4):
+			if block < self.fs.indirect_limits[i]:
+				level = i
+				break
+
+		assert level != None
+	
+		# Compute offsets for the topmost level
+		block_offset_in_level = block - self.fs.indirect_limits[level-1];
+		if self.indirect[level-1] == None:
+			self.indirect[level-1] = self.new_block(data = False)
+		current_block = xstruct.create(STRUCT_BLOCK_REFERENCE)
+		current_block.block_id = self.indirect[level-1]
+		offset_in_block = block_offset_in_level // self.fs.indirect_blocks_per_level[level-1]
+	
+		# Navigate through other levels
+		while level > 0:		
+			assert offset_in_block < self.fs.block_ids_per_block
+			
+			level -= 1
+			
+			self.fs.seek_to_block(current_block.block_id, offset_in_block*4)
+			current_block.unpack(self.fs.outf.read(4))
+			
+			if current_block.block_id == 0:
+				# The block does not exist, so alloc one and write it there
+				self.fs.outf.seek(-4, os.SEEK_CUR)
+				current_block.block_id = self.new_block(data=(level==0))
+				self.fs.outf.write(current_block.pack())
+		
+			# If we are on the last level, break here as
+			# there is no next level to visit
+			if level == 0:
+				break
+		
+			# Visit the next level
+			block_offset_in_level %= self.fs.indirect_blocks_per_level[level];
+			offset_in_block = block_offset_in_level // self.fs.indirect_blocks_per_level[level-1]
+
+		return current_block.block_id
+	
+	def do_seek(self):
+		"Perform a seek to the position indicated by self.pos"
+		
+		block = self.pos // self.fs.block_size
+		real_block = self.get_or_add_block(block)
+		offset = self.pos % self.fs.block_size
+		self.fs.seek_to_block(real_block, offset)
+		
+	def write(self, data):
+		"Write a piece of data (arbitrarily long) as the contents of the inode"
+		
+		data_pos = 0		
+		while data_pos < len(data):
+			bytes_remaining_in_block = self.fs.block_size - (self.pos % self.fs.block_size)
+			bytes_to_write = min(bytes_remaining_in_block, len(data)-data_pos)
+			self.do_seek()
+			self.fs.outf.write(data[data_pos:data_pos + bytes_to_write])
+			self.pos += bytes_to_write
+			data_pos += bytes_to_write
+			self.size = max(self.pos, self.size)
+	
+	def align_size_to_block(self):
+		"Align the size of the inode up to block size"
+		
+		self.size = align_up(self.size, self.fs.block_size)
+	
+	def align_pos(self, bytes):
+		"Align the current position up to bytes boundary"
+		
+		self.pos = align_up(self.pos, bytes)
+	
+	def pack(self):
+		"Pack the inode structure and return the result"
+		
+		data = xstruct.create(STRUCT_INODE)
+		data.mode = (Inode.TYPE2MODE[self.type] << 12)
+		data.mode |= 0x1ff # ugo+rwx
+		data.user_id = 0
+		data.size = self.size & 0xFFFFFFFF
+		data.group_id = 0
+		curtime = int(time.time())
+		data.access_time = curtime
+		data.modification_time = curtime
+		data.creation_time = curtime
+		data.deletion_time = 0
+		data.usage_count = self.refcount
+		data.reserved_512_blocks = self.blocks * (self.fs.block_size // 512)
+		data.flags = 0
+		blockconv = lambda x: 0 if x == None else x
+		data.direct_blocks = map(blockconv, self.direct)
+		data.indirect_blocks = map(blockconv, self.indirect)
+		data.version = 0
+		data.file_acl = 0
+		if self.type == Inode.TYPE_FILE:
+			data.size_high = (self.size >> 32)
+		else:
+			# size_high represents dir_acl in this case
+			data.size_high = 0
+		data.mode_high = 0
+		data.user_id_high = 0
+		data.group_id_high = 0
+		return data.pack()
+		
+class DirEntry:
+	"Represents a linked list directory entry"
+	
+	def __init__(self, name, inode, typ):
+		self.name = name.encode('UTF-8')
+		self.inode = inode
+		self.skip = None
+		self.type = typ
+	
+	def size(self):
+		"Return size of the entry in bytes"
+		
+		return align_up(8 + len(self.name)+1, 4)
+	
+	def write(self, inode):
+		"Write the directory entry into the inode"
+		
+		head = xstruct.create(STRUCT_DIR_ENTRY_HEAD)
+		head.inode = self.inode
+		head.skip = self.skip
+		head.name_length = len(self.name)
+		head.inode_type = self.type
+		inode.write(head.pack())
+		inode.write(self.name+'\0')
+		inode.align_pos(4)
+
+class DirWriter:
+	"Manages writing directory entries into an inode (alignment, etc.)"
+	
+	def __init__(self, inode):
+		self.pos = 0
+		self.inode = inode
+		self.prev_entry = None
+		self.prev_pos = None
+	
+	def prev_write(self):
+		"Write a previously remembered entry"
+		
+		if self.prev_entry:
+			self.prev_entry.skip = self.pos - self.prev_pos
+			if self.inode:
+				self.prev_entry.write(self.inode)
+	
+	def add(self, entry):
+		"Add a directory entry to the directory"
+		
+		size = entry.size()
+		block_size = self.inode.fs.block_size
+		if align_up(self.pos, block_size) < align_up(self.pos + size, block_size):
+			self.pos = align_up(self.pos, block_size)
+		self.prev_write()
+		self.prev_entry = entry
+		self.prev_pos = self.pos
+		self.pos += size
+	
+	def finish(self):
+		"Write the last entry and finish writing the directory contents"
+		
+		if not self.inode:
+			return
+		self.pos = align_up(self.pos, self.inode.fs.block_size)
+		self.prev_write()
+		self.inode.align_size_to_block()
+
+def subtree_stats(root, block_size):
+	"Recursively calculate statistics"
+	
+	blocks = 0
+	inodes = 1
+	dir_writer = DirWriter(None)
+	
+	for item in listdir_items(root):
+		inodes += 1
+		if item.is_file:
+			blocks += count_up(item.size, block_size)
+		elif item.is_dir:
+			subtree_blocks, subtree_inodes = subtree_stats(item.path, block_size)
+			blocks += subtree_blocks
+			inodes += subtree_inodes
+	
+	dir_writer.finish()
+	blocks += count_up(dir_writer.pos, block_size)
+	return (blocks, inodes)
+
+def usage(prname):
+	"Print usage syntax"
+	print(prname + " <EXTRA_BYTES> <PATH> <IMAGE>")
+
+def main():
+	if (len(sys.argv) < 4):
+		usage(sys.argv[0])
+		return
+	
+	if (not sys.argv[1].isdigit()):
+		print("<EXTRA_BYTES> must be a number")
+		return
+	
+	extra_bytes = int(sys.argv[1])
+	
+	path = os.path.abspath(sys.argv[2])
+	if (not os.path.isdir(path)):
+		print("<PATH> must be a directory")
+		return
+	
+	block_size = 4096
+	inode_size = 128
+	reserved_inode_count = 10
+	blocks_per_group = 1024
+	inodes_per_group = 512
+	
+	blocks, inodes = subtree_stats(path, block_size)
+	blocks += count_up(extra_bytes, block_size)
+	inodes += reserved_inode_count
+	
+	inodes_per_group = align_up(inodes_per_group, 8)
+	blocks_per_group = align_up(blocks_per_group, 8)
+	
+	inode_table_blocks_per_group = (inodes_per_group * inode_size) // block_size
+	inode_bitmap_blocks_per_group = count_up(inodes_per_group // 8, block_size)
+	block_bitmap_blocks_per_group = count_up(blocks_per_group // 8, block_size)
+	free_blocks_per_group = blocks_per_group
+	free_blocks_per_group -= inode_table_blocks_per_group
+	free_blocks_per_group -= inode_bitmap_blocks_per_group
+	free_blocks_per_group -= block_bitmap_blocks_per_group
+	free_blocks_per_group -= 10 # one for SB and some reserve for GDT
+	
+	block_groups = max(count_up(inodes, inodes_per_group), count_up(blocks, free_blocks_per_group))
+	
+	fs = Filesystem(sys.argv[3], block_groups, blocks_per_group, inodes_per_group,
+	                        block_size, inode_size, reserved_inode_count)
+	
+	fs.subtree_add(fs.root_inode, fs.root_inode, path, is_root=True)
+	fs.close()
+	
+if __name__ == '__main__':
+	main()
Index: tools/mkfat.py
===================================================================
--- tools/mkfat.py	(revision a2271a385d521c53eb7137d1142971451edc6aaa)
+++ tools/mkfat.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -37,14 +37,5 @@
 import xstruct
 import array
-
-exclude_names = set(['.svn', '.bzr'])
-
-def align_up(size, alignment):
-	"Return size aligned up to alignment"
-	
-	if (size % alignment == 0):
-		return size
-	
-	return ((size // alignment) + 1) * alignment
+from imgutil import *
 
 def subtree_size(root, cluster_size, dirent_size):
@@ -54,13 +45,10 @@
 	files = 2
 	
-	for name in os.listdir(root):
-		canon = os.path.join(root, name)
-		
-		if (os.path.isfile(canon) and (not name in exclude_names)):
-			size += align_up(os.path.getsize(canon), cluster_size)
+	for item in listdir_items(root):
+		if item.is_file:
+			size += align_up(item.size, cluster_size)
 			files += 1
-		
-		if (os.path.isdir(canon) and (not name in exclude_names)):
-			size += subtree_size(canon, cluster_size, dirent_size)
+		elif item.is_dir:
+			size += subtree_size(item.path, cluster_size, dirent_size)
 			files += 1
 	
@@ -72,14 +60,11 @@
 	return len(os.listdir(root))
 
-def write_file(path, outf, cluster_size, data_start, fat, reserved_clusters):
+def write_file(item, outf, cluster_size, data_start, fat, reserved_clusters):
 	"Store the contents of a file"
 	
-	size = os.path.getsize(path)
 	prev = -1
 	first = 0
 	
-	inf = open(path, "rb")
-	rd = 0;
-	while (rd < size):
+	for data in chunks(item, cluster_size):
 		empty_cluster = fat.index(0)
 		fat[empty_cluster] = 0xffff
@@ -92,11 +77,8 @@
 		prev = empty_cluster
 		
-		data = bytes(inf.read(cluster_size));
 		outf.seek(data_start + (empty_cluster - reserved_clusters) * cluster_size)
 		outf.write(data)
-		rd += len(data)
-	inf.close()
-	
-	return first, size
+	
+	return first, item.size
 
 def write_directory(directory, outf, cluster_size, data_start, fat, reserved_clusters, dirent_size, empty_cluster):
@@ -303,14 +285,11 @@
 		empty_cluster = 0
 	
-	for name in os.listdir(root):
-		canon = os.path.join(root, name)
-		
-		if (os.path.isfile(canon) and (not name in exclude_names)):
-			rv = write_file(canon, outf, cluster_size, data_start, fat, reserved_clusters)
-			directory.append(create_dirent(name, False, rv[0], rv[1]))
-		
-		if (os.path.isdir(canon) and (not name in exclude_names)):
-			rv = recursion(False, canon, outf, cluster_size, root_start, data_start, fat, reserved_clusters, dirent_size, empty_cluster)
-			directory.append(create_dirent(name, True, rv[0], rv[1]))
+	for item in listdir_items(root):		
+		if item.is_file:
+			rv = write_file(item, outf, cluster_size, data_start, fat, reserved_clusters)
+			directory.append(create_dirent(item.name, False, rv[0], rv[1]))
+		elif item.is_dir:
+			rv = recursion(False, item.path, outf, cluster_size, root_start, data_start, fat, reserved_clusters, dirent_size, empty_cluster)
+			directory.append(create_dirent(item.name, True, rv[0], rv[1]))
 	
 	if (head):
Index: tools/mktmpfs.py
===================================================================
--- tools/mktmpfs.py	(revision a2271a385d521c53eb7137d1142971451edc6aaa)
+++ tools/mktmpfs.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -35,6 +35,5 @@
 import os
 import xstruct
-
-exclude_names = set(['.svn', '.bzr'])
+from imgutil import listdir_items, chunks
 
 HEADER = """little:
@@ -71,35 +70,26 @@
 	"Recursive directory walk"
 	
-	for name in os.listdir(root):
-		canon = os.path.join(root, name)
-		
-		if (os.path.isfile(canon) and (not name in exclude_names)):
-			size = os.path.getsize(canon)
-			
-			dentry = xstruct.create(DENTRY_FILE % len(name))
+	for item in listdir_items(root):		
+		if item.is_file:			
+			dentry = xstruct.create(DENTRY_FILE % len(item.name))
 			dentry.kind = TMPFS_FILE
-			dentry.fname_len = len(name)
-			dentry.fname = name.encode('ascii')
-			dentry.flen = size
+			dentry.fname_len = len(item.name)
+			dentry.fname = item.name.encode('ascii')
+			dentry.flen = item.size
 			
 			outf.write(dentry.pack())
 			
-			inf = open(canon, "rb")
-			rd = 0;
-			while (rd < size):
-				data = inf.read(4096);
+			for data in chunks(item, 4096):
 				outf.write(data)
-				rd += len(data)
-			inf.close()
 		
-		if (os.path.isdir(canon) and (not name in exclude_names)):
-			dentry = xstruct.create(DENTRY_DIRECTORY % len(name))
+		elif item.is_dir:
+			dentry = xstruct.create(DENTRY_DIRECTORY % len(item.name))
 			dentry.kind = TMPFS_DIRECTORY
-			dentry.fname_len = len(name)
-			dentry.fname = name.encode('ascii')
+			dentry.fname_len = len(item.name)
+			dentry.fname = item.name.encode('ascii')
 			
 			outf.write(dentry.pack())
 			
-			recursion(canon, outf)
+			recursion(item.path, outf)
 			
 			dentry = xstruct.create(DENTRY_NONE)
Index: tools/xstruct.py
===================================================================
--- tools/xstruct.py	(revision a2271a385d521c53eb7137d1142971451edc6aaa)
+++ tools/xstruct.py	(revision b0d309f4768e78374754c1e98aa8d1a4a1eadd6c)
@@ -1,4 +1,5 @@
 #
 # Copyright (c) 2008 Martin Decky
+# Copyright (c) 2011 Martin Sucha
 # All rights reserved.
 #
@@ -31,4 +32,29 @@
 
 import struct
+import types
+
+ranges = {
+	'B': ((int, long), 0x00, 0xff),
+	'H': ((int, long), 0x0000, 0xffff),
+	'L': ((int, long), 0x00000000, 0xffffffff),
+	'Q': ((int, long), 0x0000000000000000, 0xffffffffffffffff),
+	'b': ((int, long), -0x80, 0x7f),
+	'h': ((int, long), -0x8000, 0x7fff),
+	'l': ((int, long), -0x80000000, 0x7fffffff) ,
+	'q': ((int, long), -0x8000000000000000, 0x7fffffffffffffff),
+}
+
+def check_range(varname, fmt, value):
+	if value == None:
+		raise ValueError('Variable "%s" not set' % varname)
+	if not fmt in ranges:
+		return
+	vartype, varmin, varmax = ranges[fmt]
+	if not isinstance(value, vartype):
+		raise ValueError('Variable "%s" is %s but should be %s' %
+		                 (varname, str(type(value)), str(vartype)))
+	if value < varmin or value > varmax:
+		raise ValueError('Variable "%s" value %s out of range %s..%s' % 
+		                 (varname, repr(value), repr(varmin), repr(varmax)))
 
 class Struct:
@@ -38,12 +64,24 @@
 	def pack(self):
 		args = []
-		for variable in self._args_:
-			if (isinstance(self.__dict__[variable], list)):
-				for item in self.__dict__[variable]:
+		for variable, fmt, length in self._args_:
+			value = self.__dict__[variable]
+			if isinstance(value, list):
+				if length != None and length != len(value):
+					raise ValueError('Variable "%s" length %u does not match %u' %
+				                      (variable, len(value), length))
+				for index, item in enumerate(value):
+					check_range(variable + '[' + repr(index) + ']', fmt, item)
 					args.append(item)
 			else:
-				args.append(self.__dict__[variable])
-		
+				check_range(variable, fmt, value)
+				args.append(value)		
 		return struct.pack(self._format_, *args)
+	
+	def unpack(self, data):
+		values = struct.unpack(self._format_, data)
+		i = 0
+		for variable, fmt, length in self._args_:
+			self.__dict__[variable] = values[i]
+			i += 1
 
 def create(definition):
@@ -77,11 +115,13 @@
 			subtokens = token.split("[")
 			
+			length = None
 			if (len(subtokens) > 1):
-				format += "%d" % int(subtokens[1].split("]")[0])
+				length = int(subtokens[1].split("]")[0])
+				format += "%d" % length
 			
 			format += variable
 			
 			inst.__dict__[subtokens[0]] = None
-			args.append(subtokens[0])
+			args.append((subtokens[0], variable, length))
 			
 			variable = None
