diff --git a/libsparse/output_file.c b/libsparse/output_file.c
new file mode 100644
index 0000000..2c4b557
--- /dev/null
+++ b/libsparse/output_file.c
@@ -0,0 +1,613 @@
+/*
+ * Copyright (C) 2010 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#define _FILE_OFFSET_BITS 64
+#define _LARGEFILE64_SOURCE 1
+
+#include <fcntl.h>
+#include <stdbool.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <zlib.h>
+
+#include "output_file.h"
+#include "sparse_format.h"
+#include "sparse_crc32.h"
+
+#ifndef USE_MINGW
+#include <sys/mman.h>
+#define O_BINARY 0
+#endif
+
+#if defined(__APPLE__) && defined(__MACH__)
+#define lseek64 lseek
+#define ftruncate64 ftruncate
+#define mmap64 mmap
+#define off64_t off_t
+#endif
+
+#ifdef __BIONIC__
+extern void*  __mmap2(void *, size_t, int, int, int, off_t);
+static inline void *mmap64(void *addr, size_t length, int prot, int flags,
+        int fd, off64_t offset)
+{
+    return __mmap2(addr, length, prot, flags, fd, offset >> 12);
+}
+#endif
+
+#define SPARSE_HEADER_MAJOR_VER 1
+#define SPARSE_HEADER_MINOR_VER 0
+#define SPARSE_HEADER_LEN       (sizeof(sparse_header_t))
+#define CHUNK_HEADER_LEN (sizeof(chunk_header_t))
+
+struct output_file_ops {
+	int (*seek)(struct output_file *, int64_t);
+	int (*write)(struct output_file *, u8 *, int);
+	void (*close)(struct output_file *);
+};
+
+struct output_file {
+	int fd;
+	gzFile gz_fd;
+	bool close_fd;
+	int sparse;
+	int64_t cur_out_ptr;
+	u32 chunk_cnt;
+	u32 crc32;
+	struct output_file_ops *ops;
+	int use_crc;
+	unsigned int block_size;
+	int64_t len;
+};
+
+static int file_seek(struct output_file *out, int64_t off)
+{
+	off64_t ret;
+
+	ret = lseek64(out->fd, off, SEEK_SET);
+	if (ret < 0) {
+		error_errno("lseek64");
+		return -1;
+	}
+	return 0;
+}
+
+static int file_write(struct output_file *out, u8 *data, int len)
+{
+	int ret;
+	ret = write(out->fd, data, len);
+	if (ret < 0) {
+		error_errno("write");
+		return -1;
+	} else if (ret < len) {
+		error("incomplete write");
+		return -1;
+	}
+
+	return 0;
+}
+
+static void file_close(struct output_file *out)
+{
+	if (out->close_fd) {
+		close(out->fd);
+	}
+}
+
+
+static struct output_file_ops file_ops = {
+	.seek = file_seek,
+	.write = file_write,
+	.close = file_close,
+};
+
+static int gz_file_seek(struct output_file *out, int64_t off)
+{
+	off64_t ret;
+
+	ret = gzseek(out->gz_fd, off, SEEK_SET);
+	if (ret < 0) {
+		error_errno("gzseek");
+		return -1;
+	}
+	return 0;
+}
+
+static int gz_file_write(struct output_file *out, u8 *data, int len)
+{
+	int ret;
+	ret = gzwrite(out->gz_fd, data, len);
+	if (ret < 0) {
+		error_errno("gzwrite");
+		return -1;
+	} else if (ret < len) {
+		error("incomplete gzwrite");
+		return -1;
+	}
+
+	return 0;
+}
+
+static void gz_file_close(struct output_file *out)
+{
+	gzclose(out->gz_fd);
+}
+
+static struct output_file_ops gz_file_ops = {
+	.seek = gz_file_seek,
+	.write = gz_file_write,
+	.close = gz_file_close,
+};
+
+static sparse_header_t sparse_header = {
+	.magic = SPARSE_HEADER_MAGIC,
+	.major_version = SPARSE_HEADER_MAJOR_VER,
+	.minor_version = SPARSE_HEADER_MINOR_VER,
+	.file_hdr_sz = SPARSE_HEADER_LEN,
+	.chunk_hdr_sz = CHUNK_HEADER_LEN,
+	.blk_sz = 0,
+	.total_blks = 0,
+	.total_chunks = 0,
+	.image_checksum = 0
+};
+
+static u8 *zero_buf;
+
+static int emit_skip_chunk(struct output_file *out, u64 skip_len)
+{
+	chunk_header_t chunk_header;
+	int ret, chunk;
+
+	//DBG printf("skip chunk: 0x%llx bytes\n", skip_len);
+
+	if (skip_len % out->block_size) {
+		error("don't care size %llu is not a multiple of the block size %u",
+				skip_len, out->block_size);
+		return -1;
+	}
+
+	/* We are skipping data, so emit a don't care chunk. */
+	chunk_header.chunk_type = CHUNK_TYPE_DONT_CARE;
+	chunk_header.reserved1 = 0;
+	chunk_header.chunk_sz = skip_len / out->block_size;
+	chunk_header.total_sz = CHUNK_HEADER_LEN;
+	ret = out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header));
+	if (ret < 0)
+		return -1;
+
+	out->cur_out_ptr += skip_len;
+	out->chunk_cnt++;
+
+	return 0;
+}
+
+static int write_chunk_fill(struct output_file *out, int64_t off, u32 fill_val, int len)
+{
+	chunk_header_t chunk_header;
+	int rnd_up_len, zero_len, count;
+	int ret;
+	unsigned int i;
+	u32 fill_buf[4096/sizeof(u32)]; /* Maximum size of a block */
+
+	/* We can assume that all the chunks to be written are in
+	 * ascending order, block-size aligned, and non-overlapping.
+	 * So, if the offset is less than the current output pointer,
+	 * throw an error, and if there is a gap, emit a "don't care"
+	 * chunk.  The first write (of the super block) may not be
+	 * blocksize aligned, so we need to deal with that too.
+	 */
+	//DBG printf("write chunk: offset 0x%llx, length 0x%x bytes\n", off, len);
+
+	if (off < out->cur_out_ptr) {
+		error("offset %llu is less than the current output offset %llu",
+				off, out->cur_out_ptr);
+		return -1;
+	}
+
+	if (off > out->cur_out_ptr) {
+		emit_skip_chunk(out, off - out->cur_out_ptr);
+	}
+
+	if (off % out->block_size) {
+		error("write chunk offset %llu is not a multiple of the block size %u",
+				off, out->block_size);
+		return -1;
+	}
+
+	if (off != out->cur_out_ptr) {
+		error("internal error, offset accounting screwy in write_chunk_raw()");
+		return -1;
+	}
+
+	/* Round up the file length to a multiple of the block size */
+	rnd_up_len = (len + (out->block_size - 1)) & (~(out->block_size -1));
+
+	/* Finally we can safely emit a chunk of data */
+	chunk_header.chunk_type = CHUNK_TYPE_FILL;
+	chunk_header.reserved1 = 0;
+	chunk_header.chunk_sz = rnd_up_len / out->block_size;
+	chunk_header.total_sz = CHUNK_HEADER_LEN + sizeof(fill_val);
+	ret = out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header));
+
+	if (ret < 0)
+		return -1;
+	ret = out->ops->write(out, (u8 *)&fill_val, sizeof(fill_val));
+	if (ret < 0)
+		return -1;
+
+	if (out->use_crc) {
+                /* Initialize fill_buf with the fill_val */
+		for (i = 0; i < (out->block_size / sizeof(u32)); i++) {
+			fill_buf[i] = fill_val;
+		}
+
+		count = chunk_header.chunk_sz;
+		while (count) {
+			out->crc32 = sparse_crc32(out->crc32, fill_buf, out->block_size);
+			count--;
+		}
+	}
+
+	out->cur_out_ptr += rnd_up_len;
+	out->chunk_cnt++;
+
+	return 0;
+}
+
+static int write_chunk_raw(struct output_file *out, int64_t off, u8 *data, int len)
+{
+	chunk_header_t chunk_header;
+	int rnd_up_len, zero_len;
+	int ret;
+
+	/* We can assume that all the chunks to be written are in
+	 * ascending order, block-size aligned, and non-overlapping.
+	 * So, if the offset is less than the current output pointer,
+	 * throw an error, and if there is a gap, emit a "don't care"
+	 * chunk.  The first write (of the super block) may not be
+	 * blocksize aligned, so we need to deal with that too.
+	 */
+	//DBG printf("write chunk: offset 0x%llx, length 0x%x bytes\n", off, len);
+
+	if (off < out->cur_out_ptr) {
+		error("offset %llu is less than the current output offset %llu",
+				off, out->cur_out_ptr);
+		return -1;
+	}
+
+	if (off > out->cur_out_ptr) {
+		emit_skip_chunk(out, off - out->cur_out_ptr);
+	}
+
+	if (off % out->block_size) {
+		error("write chunk offset %llu is not a multiple of the block size %u",
+				off, out->block_size);
+		return -1;
+	}
+
+	if (off != out->cur_out_ptr) {
+		error("internal error, offset accounting screwy in write_chunk_raw()");
+		return -1;
+	}
+
+	/* Round up the file length to a multiple of the block size */
+	rnd_up_len = (len + (out->block_size - 1)) & (~(out->block_size -1));
+	zero_len = rnd_up_len - len;
+
+	/* Finally we can safely emit a chunk of data */
+	chunk_header.chunk_type = CHUNK_TYPE_RAW;
+	chunk_header.reserved1 = 0;
+	chunk_header.chunk_sz = rnd_up_len / out->block_size;
+	chunk_header.total_sz = CHUNK_HEADER_LEN + rnd_up_len;
+	ret = out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header));
+
+	if (ret < 0)
+		return -1;
+	ret = out->ops->write(out, data, len);
+	if (ret < 0)
+		return -1;
+	if (zero_len) {
+		ret = out->ops->write(out, zero_buf, zero_len);
+		if (ret < 0)
+			return -1;
+	}
+
+	if (out->use_crc) {
+		out->crc32 = sparse_crc32(out->crc32, data, len);
+		if (zero_len)
+			out->crc32 = sparse_crc32(out->crc32, zero_buf, zero_len);
+	}
+
+	out->cur_out_ptr += rnd_up_len;
+	out->chunk_cnt++;
+
+	return 0;
+}
+
+void close_output_file(struct output_file *out)
+{
+	int ret;
+	chunk_header_t chunk_header;
+
+	if (out->sparse) {
+		if (out->use_crc) {
+			chunk_header.chunk_type = CHUNK_TYPE_CRC32;
+			chunk_header.reserved1 = 0;
+			chunk_header.chunk_sz = 0;
+			chunk_header.total_sz = CHUNK_HEADER_LEN + 4;
+
+			out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header));
+			out->ops->write(out, (u8 *)&out->crc32, 4);
+
+			out->chunk_cnt++;
+		}
+
+		if (out->chunk_cnt != sparse_header.total_chunks)
+			error("sparse chunk count did not match: %d %d", out->chunk_cnt,
+					sparse_header.total_chunks);
+	}
+	out->ops->close(out);
+}
+
+struct output_file *open_output_fd(int fd, unsigned int block_size, int64_t len,
+		int gz, int sparse, int chunks, int crc)
+{
+	int ret;
+	struct output_file *out = malloc(sizeof(struct output_file));
+	if (!out) {
+		error_errno("malloc struct out");
+		return NULL;
+	}
+	zero_buf = malloc(out->block_size);
+	if (!zero_buf) {
+		error_errno("malloc zero_buf");
+		free(out);
+		return NULL;
+	}
+	memset(zero_buf, '\0', out->block_size);
+
+	if (gz) {
+		out->ops = &gz_file_ops;
+		out->gz_fd = gzdopen(fd, "wb9");
+		if (!out->gz_fd) {
+			error_errno("gzopen");
+			free(out);
+			return NULL;
+		}
+	} else {
+		out->fd = fd;
+		out->ops = &file_ops;
+	}
+	out->close_fd = false;
+	out->sparse = sparse;
+	out->cur_out_ptr = 0ll;
+	out->chunk_cnt = 0;
+
+	/* Initialize the crc32 value */
+	out->crc32 = 0;
+	out->use_crc = crc;
+
+	out->len = len;
+	out->block_size = block_size;
+
+	if (out->sparse) {
+		sparse_header.blk_sz = out->block_size,
+		sparse_header.total_blks = out->len / out->block_size,
+		sparse_header.total_chunks = chunks;
+		if (out->use_crc)
+			sparse_header.total_chunks++;
+
+		ret = out->ops->write(out, (u8 *)&sparse_header, sizeof(sparse_header));
+		if (ret < 0)
+			return NULL;
+	}
+
+	return out;
+}
+
+struct output_file *open_output_file(const char *filename,
+		unsigned int block_size, int64_t len,
+		int gz, int sparse, int chunks, int crc)
+{
+	int fd;
+	struct output_file *file;
+
+	if (strcmp(filename, "-")) {
+		fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
+		if (fd < 0) {
+			error_errno("open");
+			return NULL;
+		}
+	} else {
+		fd = STDOUT_FILENO;
+	}
+
+	file = open_output_fd(fd, block_size, len, gz, sparse, chunks, crc);
+	if (!file) {
+		close(fd);
+		return NULL;
+	}
+
+	file->close_fd = true; // we opened descriptor thus we responsible for closing it
+
+	return file;
+}
+
+void pad_output_file(struct output_file *out, int64_t len)
+{
+	int ret;
+
+	if (len > out->len) {
+		error("attempted to pad file %llu bytes past end of filesystem",
+				len - out->len);
+		return;
+	}
+	if (out->sparse) {
+		/* We need to emit a DONT_CARE chunk to pad out the file if the
+		 * cur_out_ptr is not already at the end of the filesystem.
+		 */
+		if (len < out->cur_out_ptr) {
+			error("attempted to pad file %llu bytes less than the current output pointer",
+					out->cur_out_ptr - len);
+			return;
+		}
+		if (len > out->cur_out_ptr) {
+			emit_skip_chunk(out, len - out->cur_out_ptr);
+		}
+	} else {
+		//KEN TODO: Fixme.  If the filesystem image needs no padding,
+		//          this will overwrite the last byte in the file with 0
+		//          The answer is to do accounting like the sparse image
+		//          code does and know if there is already data there.
+		ret = out->ops->seek(out, len - 1);
+		if (ret < 0)
+			return;
+
+		ret = out->ops->write(out, (u8*)"", 1);
+		if (ret < 0)
+			return;
+	}
+}
+
+/* Write a contiguous region of data blocks from a memory buffer */
+void write_data_block(struct output_file *out, int64_t off, void *data, int len)
+{
+	int ret;
+
+	if (off + len > out->len) {
+		error("attempted to write block %llu past end of filesystem",
+				off + len - out->len);
+		return;
+	}
+
+	if (out->sparse) {
+		write_chunk_raw(out, off, data, len);
+	} else {
+		ret = out->ops->seek(out, off);
+		if (ret < 0)
+			return;
+
+		ret = out->ops->write(out, data, len);
+		if (ret < 0)
+			return;
+	}
+}
+
+/* Write a contiguous region of data blocks with a fill value */
+void write_fill_block(struct output_file *out, int64_t off, unsigned int fill_val, int len)
+{
+	int ret;
+	unsigned int i;
+	int write_len;
+	u32 fill_buf[4096/sizeof(u32)]; /* Maximum size of a block */
+
+	if (off + len > out->len) {
+		error("attempted to write block %llu past end of filesystem",
+				off + len - out->len);
+		return;
+	}
+
+	if (out->sparse) {
+		write_chunk_fill(out, off, fill_val, len);
+	} else {
+		/* Initialize fill_buf with the fill_val */
+		for (i = 0; i < sizeof(fill_buf)/sizeof(u32); i++) {
+			fill_buf[i] = fill_val;
+		}
+
+		ret = out->ops->seek(out, off);
+		if (ret < 0)
+			return;
+
+		while (len) {
+			write_len = (len > (int)sizeof(fill_buf) ? (int)sizeof(fill_buf) : len);
+			ret = out->ops->write(out, (u8 *)fill_buf, write_len);
+			if (ret < 0) {
+				return;
+			} else {
+				len -= write_len;
+			}
+		}
+	}
+}
+
+/* Write a contiguous region of data blocks from a file */
+void write_data_file(struct output_file *out, int64_t off, const char *file,
+		int64_t offset, int len)
+{
+	int ret;
+	int64_t aligned_offset;
+	int aligned_diff;
+	int buffer_size;
+
+	if (off + len >= out->len) {
+		error("attempted to write block %llu past end of filesystem",
+				off + len - out->len);
+		return;
+	}
+
+	int file_fd = open(file, O_RDONLY | O_BINARY);
+	if (file_fd < 0) {
+		error_errno("open");
+		return;
+	}
+
+	aligned_offset = offset & ~(4096 - 1);
+	aligned_diff = offset - aligned_offset;
+	buffer_size = len + aligned_diff;
+
+#ifndef USE_MINGW
+	u8 *data = mmap64(NULL, buffer_size, PROT_READ, MAP_SHARED, file_fd,
+			aligned_offset);
+	if (data == MAP_FAILED) {
+		error_errno("mmap64");
+		close(file_fd);
+		return;
+	}
+#else
+	u8 *data = malloc(buffer_size);
+	if (!data) {
+		error_errno("malloc");
+		close(file_fd);
+		return;
+	}
+	memset(data, 0, buffer_size);
+#endif
+
+	if (out->sparse) {
+		write_chunk_raw(out, off, data + aligned_diff, len);
+	} else {
+		ret = out->ops->seek(out, off);
+		if (ret < 0)
+			goto err;
+
+		ret = out->ops->write(out, data + aligned_diff, len);
+		if (ret < 0)
+			goto err;
+	}
+
+err:
+#ifndef USE_MINGW
+	munmap(data, buffer_size);
+#else
+	write(file_fd, data, buffer_size);
+	free(data);
+#endif
+	close(file_fd);
+}