RELEASE-README

SQUASHFS - A squashed read-only filesystem for Linux

Copyright 2002 Phillip Lougher (phillip@lougher.demon.co.uk)

Released under the GPL licence (version 2 or later).

Squashfs is a highly compressed read-only filesystem for Linux (kernel 2.4.x).
It uses zlib compression to compress both files, inodes and directories.
Inodes in the system are very small and all blocks are packed to minimise
data overhead. Block sizes greater than 4K are supported up to a maximum
of 32K.

Squashfs is intended for general read-only filesystem use, for archival
use (i.e. in cases where a .tar.gz file may be used), and in constrained
block device/memory systems (e.g. embedded systems) where low overhead is
needed.

The filesystem is currently stable, and has been tested on PowerPC, i586
and Sparc architectures.

Squashfs overview
-----------------

1. Data, inodes and directories are compressed.

2. Squashfs stores full uid/gids (32 bits), and file creation time.

3. Files up to 2^32 bytes are supported. Filesystems can be up to
 2^32 bytes.

4. Inode and directory data are highly compacted, and packed on byte
 boundaries. Each compressed inode is on average 8 bytes in length
 (the exact length varies on file type, i.e. regular file, directory,
 symbolic link, and block/char device inodes have different sizes).

5. Squashfs can use block sizes up to 32K (the default size is 32K).
 Using 32K blocks achieves greater compression ratios than the normal
 4K block size.

6. File duplicates are detected and removed.

7. Both big and little endian architectures are supported. The mksquashfs
 program can generate filesystems for different endian architectures for
 cases where the host byte ordering is different to the target. This is
 useful for embedded systems.


Installing squashfs
-------------------

The squashfs1.0.tar.gz file contains this README, the squashfs patch file
(squashfs-patch), and the squashfs-tools directory (mksquashfs).

The squashfs-patch file has been generated against linux-2.4.19, but may work
with older 2.4.x kernels. It assumes the kernel has inflatefs support, and
therefore definately will not work with anything much older than 2.4.17.

The squashfs patch patches the relevant kernel files to add configure support,
initrd support, include files, and the squashfs directory under linux/fs/.

Once patched, the kernel must be reconfigured, with squashfs support turned on
(either Y/M) to ensure that inflatefs is built into the kernel.


mksquashfs
----------

As squashfs is a read-only filesystem, the mksquashfs program must be used to
create populated squashfs filesystems.

mksquashfs usage: mksquashfs source dest [options]

Options are
-info	print files written to filesystem
-b block size	size of blocks in filesystem, default 32768
-noI -noInodeCompression	do not compress inode table
-noD -noDataCompression	do not compress data blocks
-check_data	add checkdata for greater filesystem integrity checks
-le	create a little endian filesystem
-be	create a big endian filesystem

Source is the source directory containing the files/directories that will
form the squashfs filesystem.

Dest is the destination where the squashfs filesystem will be written. This
can either be a conventional file or a block device. If the file doesn't exist
it will be created, if it does exist it will be truncated.

The -info option displays the files/directories as they are compressed and
added to the filesystem. The compression percentage achieved is printed, with
the original uncompressed size. If the compression percentage is listed as
0% it means the file is a duplicate.

The -b option allows the block size to be selected, this can be either
512, 1024, 2048, 4096, 8192, 16384, or 32768 bytes.

The -noI and -noD options (also -noInodeCompression and -noDataCompression)
can be used to force mksquashfs to not compress inodes/directories and data
respectively. Giving both options generates an uncompressed filesystem.

The -le and -be options can be used to force mksquashfs to generate a little
endian or big endian filesystem. Normally mksquashfs will generate a
filesystem in the host byte order.


Filesystem layout
-----------------

Brief filesystem design notes follow.

A squashfs filesystem consists of five parts, packed together on a byte alignment:

 ---------------
| superblock	|
|---------------|
| data	|
| blocks	|
|---------------|
| inodes	|
|---------------|
| directories	|
|---------------|
| uid/gid	|
| lookup table	|
 ---------------

Compressed data blocks are written to the filesystem as files are read from
the source directory, and checked for duplicates. Once all file data has been
written the completed inode, directory and uid/gid lookup tables are written.

Metadata
--------

Metadata (inodes and directories) are compressed in 8Kbyte blocks. Each
compressed block is prefixed by a two byte length, the top bit is set if the
block is uncompressed. A block will be uncompressed if the -noI option is set,
or if the compressed block was larger than the uncompressed block.

Inodes are packed into the metadata blocks, and are not aligned to block
boundaries, therefore inodes overlap compressed blocks. An inode is
identified by a two field tuple <start address of compressed block : offset
into de-compressed block>.

Inode contents vary depending on the file type. The base inode consists of:

base inode:
Inode type
Mode
uid index
gid index

The inode type is 4 bits in size, and the mode is 12 bits.

The uid and gid indexes are 4 bits in length. Ordinarily, this will allow 16
unique indexes into the uid table. To minimise overhead, the uid index is
used in conjunction with the spare bit in the file type to form a 48 entry
index as follows:

inode type 1 - 5: uid index = uid
inode type 5 -10: uid index = 16 + uid
inode type 11 - 15: uid index = 32 + uid

In this way 48 unique uids are supported using 4 bits, minimising data inode
overhead. The 4 bit gid index is used to index into a 15 entry gid table.
Gid index 15 is used to indicate that the gid is the same as the uid.
This prevents the 15 entry gid table filling up with the common case where
the uid/gid is the same.

The data contents of symbolic links are stored immediately after the symbolic
link inode, inside the inode table. This allows the normally small symbolic
link to be compressed as part of the inode table, achieving much greater
compression than if the symbolic link was compressed individually.

Similarly, the block index for regular files is stored immediately after the
regular file inode. The block index is a list of block lengths (two bytes
each), rather than block addresses, saving two bytes per block. The block
address for a given block is computed by the summation of the previous
block lengths. This takes advantage of the fact that the blocks making up a
file are stored contiguously in the filesystem. The top bit of each block
length is set if the block is uncompressed, either because the -noD option is
set, or if the compressed block was larger than the uncompressed block.

Directories
-----------

Like inodes, directories are packed into the metadata blocks, and are not
aligned on block boundaries, therefore directories can overlap compressed
blocks. A directory is, again, identified by a two field tuple
<start address of compressed block containing directory start : offset
into de-compressed block>.

Directories are organised in a slightly complex way, and are not simply
a list of file names and inode tuples. The organisation takes advantage of the
observation that in most cases, the inodes of the files in the directory
will be in the same compressed metadata block, and therefore, the
inode tuples will have the same start block.

Directories are therefore organised in a two level list, a directory
header containing the shared start block value, and a sequence of
directory entries, each of which share the shared start block. A
new directory header is written once/if the inode start block
changes. The directory header/directory entry list is repeated as many times
as necessary. The organisation is as follows:

directory_header:
count (8 bits)
inode start block (24 bits)

directory entry: * count
inode offset (13 bits)
inode type (3 bits)
filename size (8 bits)
filename

This organisation saves on average 3 bytes per filename.

File data
---------

File data is compressed on a block by block basis and written to the
filesystem. The filesystem supports up to 32K blocks, which achieves
greater compression ratios than the Linux 4K page size.

The disadvantage with using greater than 4K blocks (and the reason why
most filesystems do not), is that the VFS reads data in 4K pages.
The filesystem reads and decompresses a larger block containing that page
(e.g. 32K). However, only 4K can be returned to the VFS, resulting in a
very inefficient filesystem, as 28K must be thrown away. Squashfs,
solves this problem by explicitly pushing the extra pages into the page
cache.