README-4.5

	SQUASHFS-TOOLS 4.5 - A squashed read-only filesystem for Linux

	Copyright 2002-2021 Phillip Lougher <phillip@squashfs.org.uk>

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

Welcome to Squashfs-Tools 4.5.  The last release was just under 2 years
ago, and there are substantial improvements to the tools in this release.

Please see the INSTALL file for instructions on installing the tools,
and the USAGE file for additional documentation on how to use the tools.

Summary of changes (and sections below)
---------------------------------------

1. Major improvements in Mksquashfs
===================================

1.1 Mksquashfs now supports "Actions". These are modelled on "find"
    and allow compression, fragment packing, file exclusion and
    file attributes to be changed on a per-file basis (see section 4).

1.2 New sqfstar command which will create a Squashfs image from a tar
    archive (see section 5).

1.3 Tar style handling of source pathnames in Mksquashfs.  If option is
    specified, Mksquashfs will not strip leading directories (see
    section 6).

1.4 Cpio style handling of source pathnames in Mksquashfs.  If this option
    is specified, files will be read in from stdin.  No leading
    directories will be stripped (see section 7).

1.5 New option to throttle the amount of CPU and I/O Mksquashfs uses
    (see section 8).

1.6 New Pseudo file definitions which support timestamps (see section 9).

1.7 New Pseudo file definitions to create File references (Hard Links)
    (see section 10).

1.8 New Pseudo file definitions to create Sockets/Fifos (see section 11).

1.9 Mksquashfs now allows no source directory ( specified as - )
    to be specified, with all input from pseudo file definitions
    (see section 12).

1.10 New Pseudo file "R" definition which allows a Regular file to
    be created with data stored within the Pseudo file (see 2.5 below).


2. Major improvements in Unsquashfs
===================================

2.1 New Sqfscat command which outputs the contents of files to
    standard out (see section 13).

2.2 Symbolic links are now followed in extract files, if -follow-symlinks
    or -missing-symlinks option specified (see section 14).

2.3 Unsquashfs now supports "exclude" files in addition to "extract"
    files (see section 15).

2.4 Max depth traversal option added (see section 16).

2.5 Unsquashfs can now output a "Pseudo file" representing the
    input Squashfs filesystem.  This can be altered and given
    as input to Mksquashfs to re-generate the filesystem without
    unpacking it (see section 17).


3. Minor improvements and bug fixes
===================================

3.1 The progress bar is now displayed and updated whilst the input
    is being scanned (blocks to be compressed will increase).  This
    means on large (or slow to read) input sources, Mksquashfs will
    not appear to hang doing nothing.

3.2 New -one-file-system option in Mksquashfs, which tells
    Mksquashfs to not cross filesystem boundaries when
    scanning the sources.

3.3 New -no-hardlinks option in Mksquashfs, which makes
    Mksquashfs treat hardlinked files as duplicates instead.

3.4 New -help options in Mksquashfs and Unsquashfs which output
    to standard out, rather than standard error.

3.5 New -root-uid option in Mksquashfs, which allows the uid of the
    root directory to be set.

3.6 New -root-gid option in Mksquashfs, which allows the gid of the
    root directory to be set.

3.7 New -root-time option in Mksquashfs, which allows the time of the
    root directory to be set.

3.8 -no-exit-code option added to Unsquashfs which makes it
    not output an error exit code.

3.9 Exit code in Unsquashfs changed to distinguish between
    non-fatal errors (exit of 2), and fatal errors
    (exit code of 1).

3.10 Mksquashfs when appending, now writes the recovery file to
    the home directory ($HOME), rather than the current directory.
    A new option -recovery-path <name> has been added, which specifies
    that <name> is to be used as the recovery file directory.

3.11 Xattr id count added in Unsquashfs "-stat" output.

3.12 Unsquashfs "write outside directory" exploit fixed.

3.13 Error handling in Unsquashfs writer thread fixed.

3.14 Fix failure to truncate destination file when appending
    aborted in Mksquashfs.

3.15 Prevent Mksquashfs reading the destination file
     as input.


4. Mksquashfs Actions Introduction
==================================

The new Mksquashfs Actions code allows an "action" to be executed
on a file if one or more "tests" succeed.  If you're familiar
with the "find" command, then an action is similar to "-print",
and a test is similar to say "-name" or "-type".

To illustrate this it is useful to give two concrete examples.

example 1: the fragment action

% mksquashfs /home/phillip/github github.sqsh -action "fragment(cfiles) @ name(*.[ch])" -action "fragment(ofiles) @ name(*.o)"

This example defines two "fragment actions" which control the packing
of files within fragments.  Specifically, it creates a specialised fragment
called "cfiles" which packs files matching the wildcard name "*.[ch]".

It also creates another specialised fragment called "ofiles" which packs
files matching the wilcard name "*.o".

Producing specialised fragments which only pack files which match a range
of tests, can produce better compression and/or I/O performance as
it can optimise similarity or access patterns.  But it can also
produce worse compression, and so you should always test
the effect.

Additionally, you should be able to see that an action definition is split
into an action function before the "@", and one or more test functions
after the @.  Quoting is necessary here to protect it from interpretation
by the shell.  Also the spacing before and after the "@" isn't necessary
and is used here for added readability.

example 2: the uncompressed action

% mksquashfs /home/phillip backup.sqsh -action "uncompressed @ ( name(*.jpg) || name(*.mpg) ) || ( name(*.img) && filesize(+1G) )"

This is a more complex example.  It tells Mksquashfs to not try and
compress any file which matches the wildcard names "*.jpg" and "*.mpg".
But it also tells Mksquashfs not to try and compress files which match
the wildcard name "*.img" and are also 1 Gigabyte in size or larger.

This example introduces the fact that tests can be combined using
the logical operators && (and), || (or) and ! (not), and can be
bracketed.

Please see the ACTIONS-README file for syntax and extra information.


5. New Sqfstar command to create image from a tar archive
=========================================================

Mksquashfs when run as "sqfstar" (either named as such, or via
a hard or symbolic link) will read a tar archive from standard
in (stdin), and create a Squashfs image from it.

Syntax: sqfstar [options] image [list of exclude dirs/files]

Where <image> is the output Squashfs filesystem.

The tar archive should be uncompressed (pipe it though a decompressor
if compressed).

Sqfstar supports V7, ustar, bsdtar (libarchive), GNU tar and PAX extensions.
Sparse file extensions are supported, including the "old GNU format, type S",
and PAX formats, Versions 0.0, 0.1 and the current 1.0.

Sqfstar supports extended attributes, and recognises the SCHILY xattr
PAX extension (used by GNU tar), and the LIBARCHIVE xattr PAX extension
(used by bsdtar).

Files and directories may be excluded, and both anchored and non-anchored
exclude files are supported (see section 3.5.2 in USAGE file and examples
later).  Wildcards (globbing) is supported by default.

Sqfstar can also by invoked by running "mksquashfs source dest -tar", if
the Sqfstar link isn't available.

5.1 Usage examples
------------------

% sqfstar image.sqfs < archive.tar

    Create a Squashfs image from archive.tar, using defaults (gzip compression,
    128K blocks).

% sqfstar -comp xz -b 1M image.sqfs < archive.tar

    As previous, but use XZ compression and 1Mbyte block sizes.

% zcat archive.tgz | sqfstar image.sqfs

    Create a Squashfs image from a gzip compressed tar archive.

% sqfstar -root-uid 0 -root-gid 0 image.sqfs < archive.tar

    Tar files do not supply a definition for the root directory, and the
    default is to make the directory owned/group owned by the user running
    Sqfstar.  The above command sets the ownership/group ownership to root.

% sqfstar -root-mode 0755 image.sqfs < archive.tar

   The default permissions for the root directory is 0777 (rwxrwxrwx).  The
   above command sets the permissions to 0755 (rwxr-xr-x).

% sqfstar image.sqsh dir1/file1 dir2/file2 < archive.tar

   Create a Squashfs image but exclude the files "file1" and "file2".

% sqfstar image.sqsh "... *.[ch]" < archive.tar

   Create a Squashfs image but exclude any file matching "*.[ch]" anywhere
   in the archive.

 
5.2 Sqfstar options list
------------------------

The following is the full list of options supported by Sqfstar.

Filesystem build options:
-comp <comp>		select <comp> compression
			Compressors available:
				gzip (default)
-b <block_size>		set data block to <block_size>.  Default 128 Kbytes
			Optionally a suffix of K or M can be given to specify
			Kbytes or Mbytes respectively
-reproducible		build images that are reproducible (default)
-not-reproducible	build images that are not reproducible
-mkfs-time <time>	set mkfs time to <time> which is an unsigned int
-fstime <time>		synonym for mkfs-time
-all-time <time>	set all inode times to <time> which is an unsigned int
-exports		make the filesystem exportable via NFS
-no-sparse		don't detect sparse files
-no-xattrs		don't store extended attributes
-xattrs			store extended attributes (default)
-noI			do not compress inode table
-noId			do not compress the uid/gid table (implied by -noI)
-noD			do not compress data blocks
-noF			do not compress fragment blocks
-noX			do not compress extended attributes
-no-fragments		do not use fragments
-no-tailends		don't pack tail ends into fragments
-no-duplicates		do not perform duplicate checking
-no-hardlinks		do not hardlink files, instead store duplicates
-all-root		make all files owned by root
-root-time <time>	set root directory time to <time>
-root-mode <mode>	set root directory permissions to octal <mode>
-root-uid <uid>		set root directory owner to <uid>
-root-gid <gid>		set root directory group to <gid>
-force-uid <uid>	set all file uids to <uid>
-force-gid <gid>	set all file gids to <gid>
-nopad			do not pad filesystem to a multiple of 4K

Filesystem filter options:
-ef <exclude_file>	list of exclude dirs/files.  One per line
-regex			Allow POSIX regular expressions to be used in exclude
			dirs/files

Sqfstar runtime options:
-version		print version, licence and copyright message
-force			force Sqfstar to write to block device or file
-exit-on-error		treat normally ignored errors as fatal
-quiet			no verbose output
-info			print files written to filesystem
-no-progress		don't display the progress bar
-progress		display progress bar when using the -info option
-throttle <percentage>	throttle the I/O input rate by the given percentage.
			This can be used to reduce the I/O and CPU consumption
			of Mksquashfs
-limit <percentage>	limit the I/O input rate to the given percentage.
			This can be used to reduce the I/O and CPU consumption
			of Mksquashfs (alternative to -throttle)
-processors <number>	Use <number> processors.  By default will use number of
			processors available
-mem <size>		Use <size> physical memory.  Currently set to 4096M
			Optionally a suffix of K, M or G can be given to specify
			Kbytes, Mbytes or Gbytes respectively

Miscellaneous options:
-root-owned		alternative name for -all-root
-offset <offset>	Skip <offset> bytes at the beginning of <dest>.
			Optionally a suffix of K, M or G can be given to specify
			Kbytes, Mbytes or Gbytes respectively.
			Default 0 bytes.
-o <offset>		synonym for -offset
-noInodeCompression	alternative name for -noI
-noIdTableCompression	alternative name for -noId
-noDataCompression	alternative name for -noD
-noFragmentCompression	alternative name for -noF
-noXattrCompression	alternative name for -noX

-help			output this options text to stdout
-h			output this options text to stdout

-Xhelp			print compressor options for selected compressor


6. Tar style handling of source pathnames in Mksquashfs
=========================================================

Mksquashfs has always stripped the leading directories of any source
pathnames given on the command line.

For example given the command line

% mksquashfs dir-a/dir-b/dir-c/file1 dir-A/dir-B/file2 sqfs

Mksquashfs will strip the leading directories, and put file1 and file2
into the same root directory.  If file1 and file2 are directories it
will place the directories into the same root directory, but, if
only one directory is specified, it will put the contents of that
directory into the root directory.  Obviously, for a lot of the time
this is what you want.  But, if it isn't what you want it can be
quite difficult to get Mksquashfs to do something different.

A lot of people don't like this, and would prefer Mksquashfs acted
like "tar", which does not strip leading directories.  This allows
you to create a directory hierarchy from the pathnames of the supplied
files.  In the above example, the tar archive would contain the pathnames
"dir-a/dir-b/dir-c/file1" and "dir-A/dir-B/file2".

Mksquashfs will now act like tar if given the option -no-strip, or -tarstyle.


7. Cpio style handling of source pathnames in Mksquashfs
========================================================

Mksquashfs now allows you to pipe the set of files to be added to the
filesystem to standard in (stdin), if the -cpiostyle option is given.

As with cpio, directories are not recursively scanned, and their contents
added.  Evey file to be added to the filesystem must be explicitly
specified.

Typically the list of files to be added will be produced via find, or a
similar utility.

For example

% find /home/phillip/squashfs-tools | mksquashfs - img.sqfs -cpiostyle

Will create an image containing everything in squashfs-tools and its
sub-directories.  Note, "-" is given as the source pathname in Mksquashfs,
and indicates no commmand line sources.

and the following will add just the files ending in .c, .h and .o.

% find /home/phillip/squashfs-tools -name "*.[cho]" | mksquashfs - img.sqfs -cpiostyle


8. New option to throttle CPU and I/O Mksquashfs uses
=====================================================

People have sometimes complained that Mksquashfs uses too much CPU and too
much I/O, and can make a computer become very slow, or over-heat, and
setting the processors option to 1 often doesn't have much effect.

The reason for this is because Mksquashfs tries to run as fast as possible,
as this is usually what people want.  Due to this it is highly parallelised,
and spawns multiple compression threads (for both blocks and fragments),
and by default one each of those compression threads are spawned per core.
In addition it also has separate reader and writer threads and large caches
to ensure I/O through-put is maximised and performed in parallel with the
compression threads (called read-ahead and write-behind).

This means even when the "-processors 1" option is set, there is still a
lot of parallel processing being performed, because the minmimum set of
threads is a single block compression thread, a single fragment compression
thread, and both a reader and writer thread.  Unfortunately, these threads
can easily still max out a single core, so that it is never idle.  On a
multi-core machine the "-processors 1" option usually produces a CPU load
of around 130-180%.

The solution adopted is to add an I/O "throttle", which will regulate the
amount of I/O performed, and ultimately, how much CPU is consumed.  This
throttle can be set from 0 - 100, with the following two options added.

-throttle <percentage> has the semantics of throttle the reader by X%.  The
higher the percentage, the greater the throughput of the reader thread is
reduced.

-limit <percentage> has the opposite sense of limit the output to X%.  The
higher the percentage, the greater the throughput of the reader thread.


9. New Pseudo file definitions with timestamps
==============================================

The Pseudo file definitions previously supported did not allow the time
of the file to be set.  New extended pseudo file definitions have been added
which take a <time> timestamp.  These are distinquished from the previous
definitions by using an upper-case type character.  For example the "D"
definition is identical to the previous "d" definition (it creates
a directory), but it takes a <time> timestamp.

The list of extended definitions are:

	filename D time mode uid gid
	filename M time mode uid gid
	filename B time mode uid gid major minor
	filename C time mode uid gid major minor
	filename F time mode uid gid command
	filename S time mode uid gid symlink

<time> can be either an unsigned decimal integer (which represents the
seconds since the epoch of 1970-01-01 01:00 UTC), or a "date string"
which is parsed and converted into an integer since the epoch, by calling
"date".

Because most date strings have spaces, they will need to be quoted, and if
entered on the command line, these quotes will need to be protected from the
shell by backslashes, i.e.

% mksquashfs x x.sqsh -p "file D \"1 jan 1980\" 0777 phillip phillip"

Obviously anything "date" accepts as a valid string can be used, such as
"yesterday", "last week" etc.


10. New Pseudo file definition to create File references (Hard Links)
=====================================================================

The existing "f" Pseudo definition allows a regular file to be
created from the output of a command (or shell).

Often this is used to reference a file outside the source directories,
by executing "cat", e.g.

README f 0555 0 0 cat /home/phillip/latest-version/README

Because this is quite a frequent use of the definition, a new faster
"File reference" or Hard Link (in filesystem terms a File reference is a
Hard Link) Pseudo definition has been added named 'l'.

README l /home/phillip/latest-version/README

Will create a reference to "/home/phillip/latest-version/README",
and obviously the timestamp/mode and owership will be used.

The definition also be used to create additional references to
files within the source directories.  For instance if "phillip/latest/README"
was a file being added to the filesystem, then

README l phillip/latest/README

Will create a Hard Link (and increment the nlink count on the inode).

In both cases, the Path to the file being referenced is the system
filesystem path, and can be absolute (prefixed with /), or relative
to the current working directory.

There is an additional 'L' Pseudo definition, which closes a loophole in
the above 'l' definition.  The 'l' Pseudo definition cannot create references
or Hard Links to files created by Pseudo definitions, because by
definition they do not exist in the system filesystem.

with 'L' the referenced file is expected to be a Pseudo file, and in this case
the Path is taken to be from the root of the Squashfs filesystem being created,
e.g.

char-dev c 0555 0 0 1 2

link L char-dev

Will create a Hard Link named "link" to the character device called "char-dev"
created by the previous Pseudo definition.


11. New Pseudo file definitions to create Sockets/Fifos
=======================================================

Two new definitions have been added which allows Sockets and Fifos
to be created, which are:

filename i mode uid gid [s|f]
filename I time mode uid gid [s|f]

The 'i' definition doesn't take a time value, whereas the 'I'
definition does.

To create a Unix domain socket, 's' should be used, i.e.

filename i 0777 root root s

and to create a FIFO, 'f' should be used, i.e.

filename i 0777 root root f


12. Mksquashfs now allows no source directories ( specified as - )
==================================================================

Mksquashfs has removed the requirement to specify any physical source
directories, and a Squashfs filesystem can now be built entirely with
Pseudo definitions.

For example the following command line

% mksquashfs - img.sqfs -p "/ d 777 phillip users" -p "/hello l /tmp/typescript" -p "/link L hello"

Will create a root directory, a file created by a reference to a file
within the filesystem, and create a hard link to it.


13. New Sqfscat command which outputs the contents of files to standard out
===========================================================================

Unsquashfs when run as "sqfscat" (either named as such, or via a hard or
symbolic link) will output the contents of one or more files to standard
out.  The syntax and options supported are:

SYNTAX: sqfscat [options] filesystem [files to cat]
	-v[ersion]		print version, licence and copyright information
	-p[rocessors] <number>	use <number> processors.  By default will use
				number of processors available
	-o[ffset] <bytes>	skip <bytes> at start of <dest>.  Optionally a
				suffix of K, M or G can be given to specify
				Kbytes, Mbytes or Gbytes respectively (default
				0 bytes).
	-ig[nore-errors]	treat errors writing files to output as
				non-fatal
	-st[rict-errors]	treat all errors as fatal
	-no-exit[-code]		don't set exit code (to nonzero) on non-fatal
				errors
	-da[ta-queue] <size>	set data queue to <size> Mbytes.  Default 256
				Mbytes
	-fr[ag-queue] <size>	set fragment queue to <size> Mbytes.  Default
				256 Mbytes
	-no-wild[cards]		do not use wildcard matching in extract names
	-r[egex]		treat extract names as POSIX regular expressions
				rather than use the default shell wildcard
				expansion (globbing)
	-h[elp]			output this options text to stdout


The pathnames of the files to be output, like cat, can contain symbolic links,
and "." and ".." elements.

Sqfscat is a short-cut to using the equivalent Unsquashfs -cat option, i.e.
the following will behave the same:

% sqfscat img file

% unsquashfs -cat img file


14. Symbolic links are now "followed" in extract files
======================================================

Unsquashfs walks the extract file paths as it recursively extracts
the Squashfs filesystem  from top to bottom.  During that recursive
extraction symbolic links are obviously not followed (see below).

The consequence of this is if an extract file ends in a symbolic link
(leaf component) it is extracted and left as a dangling symlink,
unless the real file it links to has also specified as an extract file.

Additionally, if the extract file pathname traverses symbolic links
while walking the pathname (i.e. the extract file pathname has
embedded symbolic links), the extraction will stop at that point.

One way of solving this problem is by following (or dereferencing) the
symbolic link(s), and replacing them with what they actually link to,
in a similar way to "cp -L".  But this is dangerous, and can cause
Unsquashfs to produce output which does not match the input Squashfs
filesystem, which is something Unsquashfs should never do.

The reason for this is whereas "cp -L" is derefencing the *input*,
Unsquashfs is dereferencing the *output*.  If a filesystem has a real
filename, say

	a/b/c/hello_world

and two symlinks

	a/symlink1 ---> b/c/hello_world
	a/symlink2 ---> b/c

There are multiple different ways (or paths) to the *single* hello_world file.

	a/b/c/hello_world
	a/symlink1
	a/symlink2/hello_world

If Unsquashfs was given all three paths as extract files, and Unsquashfs
dereferenced them on output, you will get *three* copies of the hello_world
file, and two copies of directory "c".

Superficially the output may look the same, but, it may not work the same,
and obviously edits to the one hello_world file will not get reflected in
the other copies.  Any option that can be accidently or maliciously used
to produce such an output is too dangerous to be added.

Unsquashfs solves the problem in an equivalent way, but which does not alter
the output, and so it is a completely safe option.

If the -follow-symlinks option is specified, Unsquashfs will canonicalise
the extract files to produce the canonical pathname (that is the
"real" pathname without any symbolic links).  It will then add all
the symbolic links necessary to ensure that the extract file can be
resolved.

The -missing-symlinks option is similar to -follow-symlinks except it
will cause Unsquashfs to abort if any symbolic link cannot be resolved.

Note: as a side effect of the canonicalisation, with the above options
enabled extract filenames can also now have ".", ".." elements within
the pathnames.


15. Unsquashfs now supports "exclude" files in addition to "extract" files
==========================================================================

Unsquashfs now allows exclude files to be specified, either on their
own, or in addition to extract files.

An exclude file is, obviously, the opposite of an extract file.  Whereas
an extract file limits the output of Unsquashfs to the files/directories
matched by the extract file(s), exclude file(s) output the entire filesystem,
with the sub-set of files matched by the exclude file(s) excluded.

Often you want to output the filesystem where you're only interested in
some files, which is where extract files are useful.  But equally, you
often want to to output the entire filesystem, but, with some unwanted
files removed.  An example of this, perhaps, is where you've inadvertantly,
archived binaries (.o files etc.) and you're only interested in extracting
the source code.

In the above example, trying to remove the binaries (.o etc) when you've
only got "extract" files capability becomes extremely messy.  It is a
lot easier to do that with "exclude" files.

Unsquashfs now supports two ways of specifying exclude files.

The most straightforward way is to tell Unsquashfs to treat extract files as
exclude files.  That is extract files are specified on the command line as
a list of files after the options and filesystem image.  Giving the
-excludes option tells Unsquashfs to treat them as exclude files.

To make this clearer,

% unsquashfs img file1 file2 file3

Tells Unsquashfs to extract file1, file2 and file3.  But,

% unsquashfs -excludes img file1 file2 file3

Tells Unsquashfs to exclude file1, file2, and file3.

The perhaps obvious problem with this, is it allows you to
choose between "extract" files or "exclude" files.  But, it doesn't
allow you to have both "extract" files and "exclude" files on the
command line.

To get around this problem Unsquashfs supports another way of specifying
exclude files.  That is to use the option -exclude-list.  This
option allows a list of exclude files to specified, terminated by a
";".  The necessity of using ";" to terminate the list is because this
is a normal option, without it, any further entries on the command line
would be interpreted as being part of the list.

For example, the following are equivalent:

% unsquashfs -excludes img file1 file2 file3

% unsquashfs -exclude-list file1 file2 file3 \; img

Note the black-slashing of ";" to prevent it from being
interpreted by the shell as a special character.

Obviously, where the -exclude-list option comes into its own is
when it is mixed with extract files, for example:

% unsquashfs -exclude-list dir1/file1 dir2/file2 \; img dir1 dir2

This tells Unsquashfs to extract directories "dir1" and "dir2", and
then to exclude the files "dir1/file1" and "dir2/file2".

From this it should be clear that the precedence is extract files and
then exclude files, because it doesn't make any sense otherwise.
Extract files define the set of directories/files to be extracted,
and exclude files remove directories/files from that sub-set.


16. Max depth traversal option added
====================================

Unsquashfs has added a -max-depth <n> option, which will limit output
and/or file listing to <n> depth directories, where the top level
directory is depth 1, e.g.

% unsquashfs -max-depth 3 -ll img

Will limit the filesystem output listing to be 3 directories in depth.


17. Unsquashfs Pseudo file output
=================================

If the Pseudo file (-pf) option is given, Unsquashfs will output a
Pseudo file representation of the input filesystem.  This pseudo file
can be used as input to Mksquashfs to reproduce the Squashfs filesystem
without having to unpack the input filesystem image.

The Pseudo file output is designed to be editable, and the pseudo file
entries can be altered (name, date, ownership etc.) or added/deleted
before any new filesystem is rebuilt.

The format of the pseudo file, is obviously, the same as the pseudo file
definitions supported by Msquashfs.  In this release (see sections 8, 9
and 10) this has been extended to fully support any file type and
timestamp information (extended attributes are not yet supported).

Regular files (with data) are supported with the new "R" pseudo
definition, which is

filename R time mode uid gid length offset

<length> specifies the size of the file, and <offset> is a byte
offset into the pseudo file where the data is stored.  This offset
is taken from the start of the data section (see below), rather
than from the start of the file (this is allow the pseudo file
entries to be edited without altering the data offsets).

Data is deliberately stored out of line (i.e. unlike tar), to make
the file more easily editable manually.

An example pseudo file output for a filesystem consisting of a directory,
two regular files, and a symbolic link might be

	/ D 1625536033 1777 0 0
	test D 1625536969 755 1000 100
	test/hello_world R 1625535696 644 1000 100 12 0
	test/regfile R 1625536928 644 1000 100 37 12
	test/symlink S 1625535712 777 1000 100 hello_world
	#
	# START OF DATA - DO NOT MODIFY
	#
	Hello World
	This is the data contents of regfile

Here you should be able see the most important aspects of the layout,
the pseudo file entries appear at the start of the file, and the data
is stored at the end.  The two sections are separated by 3 special
marker lines, containing the words "START OF DATA - DO NOT MODIFY".

The following is a small example of how someone might edit and
rebuild a Squashfs image.

$ unsquashfs -pf pseudo test.sqsh
$ sed -i "0,/# START OF DATA/s/\([^ ]* . [0-9]* [0-7]* \)[0-9]*/\11234/g" pseudo
$ sed -i "0,/# START OF DATA/s/hello_world/hello/g" pseudo
$ mksquashfs - test2.sqsh -pf pseudo

This will change the ownership to uid 1234, and change the name of
"hello_word" to "hello".


18. Compatibility
=================

Mksquashfs 4.5 generates 4.0 filesystems.  These filesystems are fully
compatible/interchangable with filesystems generated by Mksquashfs 4.x and are
mountable on 2.6.29 and later kernels.