SQFSTAR 4.7.3 - A tool to create a Squashfs filesystem from a TAR archive

This file describes how to use Sqfstar, and it has the following sections:

1. Introduction and basic usage
2. Getting help and displaying Sqfstar options
3. Changing compression algorithm and compression specific options
4. Changing global compression defaults used in sqfstar
5. Building reproducible filesystem images
6. Specifying the UIDs/GIDs used in the filesystem
7. Specifying the file permissions used in the filesystem
8. Excluding files from the filesystem
9. Reducing CPU and I/O usage
10. Filtering and adding extended attributes (xattrs)
11. Pseudo file support
12. Extended pseudo file definitions with timestamps
13. Miscellaneous options

1. INTRODUCTION AND BASIC USAGE
-------------------------------

Sqfstar will read an uncompressed TAR archive from standard in, and it will
create a Squashfs filesystem from it.  If a TAR archive is compressed it
should be piped through a decompressor utility and then input into Sqfstar.

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).

Sqfstar uses the following arguments

sqfstar [OPTIONS] FILESYSTEM [list of exclude dirs/files]

Where FILESYSTEM is the name of the output filesystem.  This can be a file or a
block device.  If the file already exists or it is a block device Sqfstar will
refuse to write to it, unless the -force option is specified.

Most simple usage is to read an uncompressed tar file from stdin.

% sqfstar image.sqfs < archive.tar

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

example 2:

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

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

example 3:

Uncompress a tar archive to stdout, and pipe the result to Sqfstar.

% zcat archive.tgz | sqfstar image.sqfs

example 4:

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.  This
example explicitly sets the ownership/group ownership to root.

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

example 5:

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

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

example 6:

Files can be excluded from the Squashfs filesystem by putting them at the
end of the command line.

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

This creates a Squashfs image but excludes the files "dir1/file1" and
"dir2/file2".

example 7:

Files anywhere in the filesystem can be excluded by prefixing with ... (which
means not anchored to a particular directory)

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

This creates a Squashfs image with all files matching "*.[ch]" excluded.


2. GETTING HELP AND DISPLAYING SQFSTAR OPTIONS
----------------------------------------------

Sqfstar has fairly detailed built-in help information describing the available
options:  Running:

% sqfstar help

Will displaythe following summary of the help options and information
available:

*************************************************************************
SYNTAX: sqfstar [OPTIONS] FILESYSTEM [list of exclude dirs/files]

Run
  "sqfstar -help-option <regex>" to get help on all options matching <regex>

Or run
  "sqfstar -help-section <section-name>" to get help on these sections
        SECTION NAME            SECTION
        compression             Filesystem compression options:
        build                   Filesystem build options:
        time                    Filesystem time options:
        perms                   Filesystem permissions options:
        pseudo                  Filesystem pseudo options:
        xattrs                  Filesystem extended attribute (xattrs) options:
        runtime                 Sqfstar runtime options:
        expert                  Expert options (these may make the filesystem unmountable):
        help                    Help options:
        misc                    Miscellaneous options:
        pseudo-defs             Pseudo file definition format:
        symbolic                Symbolic mode specification:
        environment             Environment:
        exit                    Exit status:
        extra                   See also (extra information elsewhere):

Or run
  "sqfstar -help-all" to get help on all the sections
*************************************************************************

For example to get a list of all the options that operate on uids and gids, you
could do

% sqfstar -help-option "uid|gid"
-root-uid <user>        set root directory owner to specified <user>, <user> can
                        be either an integer uid or user name
-root-gid <group>       set root directory group to specified <group>, <group>
                        can be either an integer gid or group name
-force-uid <user>       set all file and directory uids to specified <user>,
                        <user> can be either an integer uid or user name
-force-gid <group>      set all file and directory gids to specified <group>,
                        <group> can be either an integer gid or group name
-uid-gid-offset <value> offset all uid and gids by specified <value>.  <value>
                        should be a positive integer
-default-uid <user>     tar files often do not store uids for intermediate
                        directories.  This option sets the default directory
                        owner to <user>, rather than the user running Sqfstar.
                        <user> can be either an integer uid or user name.  This
                        also sets the root directory uid
-default-gid <group>    tar files often do not store gids for intermediate
                        directories.  This option sets the default directory
                        group to <group>, rather than the group of the user
                        running Sqfstar.  <group> can be either an integer gid
                        or group name.  This also sets the root directory gid
-pd <d mode uid gid>    specify a default pseudo directory which will be used in
                        pseudo definitions if a directory in the pathname does
                        not exist.  This also allows pseudo definitions to be
                        specified without specifying all the directories in the
                        pathname.  The definition should be quoted
*************************************************************************

2.1 -help-section <section>
----------------------------

The -help-section option displays the section that matches the <section> name.
If <section> does not exactly match a section name, it is treated as a regular
expression, and all section names that match are displayed.  Finally, if
<section> is "list", a list of sections and their names is displayed.

For example:

*************************************************************************
% sqfstar -help-section compression
Filesystem compression options:
-b <block-size>         set data block to <block-size>.  Default 128 Kbytes.
                        Optionally a suffix of K, KB, Kbytes or M, MB, Mbytes
                        can be given to specify Kbytes or Mbytes respectively
-comp <comp>            select <comp> compression.  Run -help-comp <comp> to get
                        compressor options for <comp>, or <all> for all the
                        compressors.
                        Compressors available:
                                gzip (default)
                                lzo
                                lz4
                                xz
                                zstd
-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-compression         do not compress any of the data or metadata.  This is
                        equivalent to specifying -noI -noD -noF and -noX
*************************************************************************

Will display the compression options section.

Using regular expression matching section names can be abbreviated, for example
"comp" will also display the compression options section.  But, it also means
multiple sections can be displayed, for example:

*************************************************************************
% sqfstar -help-section "comp|build"
Filesystem compression options:
-b <block-size>         set data block to <block-size>.  Default 128 Kbytes.
                        Optionally a suffix of K, KB, Kbytes or M, MB, Mbytes
                        can be given to specify Kbytes or Mbytes respectively
-comp <comp>            select <comp> compression.  Run -help-comp <comp> to get
                        compressor options for <comp>, or <all> for all the
                        compressors.
                        Compressors available:
                                gzip (default)
                                lzo
                                lz4
                                xz
                                zstd
-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-compression         do not compress any of the data or metadata.  This is
                        equivalent to specifying -noI -noD -noF and -noX

Filesystem build options:
-reproducible           build filesystems that are reproducible (default)
-not-reproducible       build filesystems that are not reproducible
-exports                make the filesystem exportable via NFS
-no-sparse              do not detect sparse files
-no-fragments           do not use fragments
-no-tailends            do not pack tail ends into fragments
-no-duplicates          do not perform duplicate checking
-no-hardlinks           do not hardlink files, instead store duplicates
-regex                  allow POSIX regular expressions to be used in exclude
                        dirs/files
-ignore-zeros           allow tar files to be concatenated together and fed to
                        Sqfstar.  Normally a tarfile has two consecutive 512
                        byte blocks filled with zeros which means EOF and
                        Sqfstar will stop reading after the first tar file on
                        encountering them. This option makes Sqfstar ignore the
                        zero filled blocks
-ef <exclude-file>      list of exclude dirs/files.  One per line
*************************************************************************

Will display the compression options and build options sections.

2.2. PAGER environment variable
-------------------------------

By default the tools try pager, /usr/bin/pager, less, /usr/bin/less, more,
/usr/bin/more, cat and /usr/bin/cat in that order.

The pager used can be over-ridden using the PAGER environment variable.  If the
filename given by PAGER doesn't contain slashes, the PATH environment variable
will be used to locate it, otherwise it will be treated as a pathname.


3. CHANGING COMPRESSION ALGORITHM AND COMPRESSION SPECIFIC OPTIONS
------------------------------------------------------------------

By default Sqfstar will compress using the GZIP compression algorithm.  This
algorithm offers a good trade-off between compression ratio, and memory and time
taken to decompress.

Squashfs also supports LZ4, LZO, XZ and ZSTD compression.  LZO offers worse
compression ratio than GZIP, but is faster to decompress.  XZ offers better
compression ratio than GZIP, but at the expense of greater memory and time
to decompress (and significantly more time to compress).  LZ4 is similar
to LZO.  ZSTD has been developed by Facebook, and aims to compress well and
be fast to decompress.  You should experiment with the compressors to
see which is best for you.

If you're not building the squashfs-tools and kernel from source, then the tools
and kernel may or may not have been built with support for LZ4, LZO, XZ or ZSTD
compression.

The compression algorithms supported by the build of Sqfstar can be found by
typing:

% sqfstar -help-comp list
        gzip (default)
        lzo
        lz4
        xz
        zstd

The list of compressor specific options for a compressor can be found by typing
sqfstar -help-comp <compressor>, for example:

% sqfstar -help-comp xz
sqfstar: compressor "xz".  Options supported:
          -Xbcj filter1,filter2,...,filterN
                Compress using filter1,filter2,...,filterN in turn (in addition
                to no filter), and choose the best compression.  Available
                filters: x86, arm, armthumb, arm64, powerpc, sparc, ia64, riscv
          -Xdict-size <dict-size>
                Use <dict-size> as the XZ dictionary size.  The dictionary size
                can be specified as a percentage of the block size, or as an
                absolute value.  The dictionary size must be less than or equal
                to the block size and 8192 bytes or larger.  It must also be
                storable in the xz header as either 2^n or as 2^n+2^(n+1).
                Example dict-sizes are 75%, 50%, 37.5%, 25%, or 32K, 16K, 8K
                etc.

The compression specific options for all compressors can be found by typing
sqfstar -help-comp all

The compression specific options are, obviously, specific to the compressor in
question, and the compressor documentation and web sites should be consulted to
understand their behaviour.  In general the Sqfstar compression defaults for
each compressor are optimised to give the best performance for each compressor,
where what constitutes best depends on the compressor.  For GZIP/XZ best means
highest compression, for LZO/LZ4 best means a tradeoff between compression and
(de)-compression overhead (LZO/LZ4 by definition are intended for weaker
processors).


4. CHANGING GLOBAL COMPRESSION DEFAULTS USED IN SQFSTAR
-------------------------------------------------------

There are a large number of options that can be used to control the compression
in Sqfstar.  By and large the defaults are the most optimum settings and should
rarely need to be changed.

Note, this does not apply to the block size, increasing the block size from
the default of 128 Kbytes will increase compression (especially for the XZ and
ZSTD compressors) and should increase I/O performance too.  However, a block
size of greater than 128 Kbytes may increase latency in certain cases (where the
filesystem contains lots of fragments, and no locality of reference is
observed).  For this reason the block size default is configured to the less
optimal 128 Kbytes.  Users should experiment with 256 Kbyte sizes or above.

The -b option allows the block size to be selected, both "K" and "M" postfixes
are supported, this can be either 4K, 8K, 16K, 32K, 64K, 128K, 256K, 512K or
1M bytes.

The -noI, -noD, -noF and -noX options can be used to force Mksquashfs to not
compress inodes (metadata), data, fragments and extended attributes
respectively.  The -no-compression option generates an uncompressed filesystem,
and it is equivalent to specifying all of the -noI, -noD, -noF and -noX options.

The -no-fragments option tells Sqfstar to not generate fragment blocks.  A
fragment block consists of multiple small files (all smaller than the block
size) packed and compressed together.  This produces much better compression
than storing and compressing these files separately.  It also typically
improves I/O as multiple files in the same directory are read at the same time.
You don't want to enable this option unless you fully understand the effects.

The -no-tailends option tells Sqfstar to not pack file tailends into fragment
blocks.  Normally a file will not be a multiple of the block size, and so
there were always be a tail which doesn't fit fully into a data block.  This
tailend is by default packed into fragment blocks.  Enabling this option will
reduce compression.

The -no-duplicates option tells Sqfstar to not check the files being added to
the filesystem for duplicates.  This can result in quicker filesystem generation
although obviously compression will suffer badly if there is a lot of duplicate
files.


5. BUILDING REPRODUCIBLE FILESYSTEM IMAGES
------------------------------------------

If you want Sqfstar to generate an identical (byte for byte) filesystem on
every run, then the following conditions have to be true:

1. The filesystem source data has to be the same,
2. The timestamps (and other metadata such as permissions), must be the same,
3. The root directory timestamp (and other metadata), must be the same,
4. The filesystem make time (stored in the super-block) must be the same.

Due to point 4, every time you run Sqfstar, the filesystem will be different,
even if everything else is the same.  But less obviously, Sqfstar has to
fabricate a root directory because tar files do not contain one, and so the
timestamp of the root directory will also change on every run.

To avoid the above, previous versions introduced the -mkfs-time <time>, and
-root-time <time> options:

% sqfstar -mkfs-time 0 -root-time 0 image.sqfs < tarfile

Will generate a filesystem image where the timestamps (that will change) have
been set to 0 (the start of the epoch 1970-01-01).

But a problem with this (for many people) is that it ensures reproducibility by
losing information and functionality, akin to using a sledgehammer to crack a
nut.  With a filesystem make time of 0, it is no longer possible to discover the
difference between one filesystem and another without looking at the content, or
know how old the filesystem is without looking at the content either.

Due to this, this release introduces new variants of -mkfs-time, and -root-time.
It also introduces a new variant of -all-time, while also renaming it to
-inode-time.  Lastly, there are some new easy to remember shorthand options
added.

5.1 -mkfs-time, -root-time and -inode-time options with a timestamp
-------------------------------------------------------------------

5.1.1 -mkfs-time <time>

Set mkfs time to <time>.  Time can be an integer which is the seconds since
the epoch of 1970-01-01 00:00:00 UTC), or a date string as recognised by the
"date" command.

5.1.2 -root-time <time>

Set root directory timestamp to <time>.  Time can be an integer which is the
seconds since the epoch of 1970-01-01 00:00:00 UTC), or a date string as
recognised by the "date" command.

5.1.3 -inode-time <time>

This option has been renamed from -all-time [*] in previous versions because
all-time was a misnomer (it sets all the inode timestamps, but not also the
filesystem make time as the name suggests).

Set all file timestamps to <time>.  Time can be an integer which is the seconds
since the epoch of 1970-01-01 00:00:00 UTC), or a date string as recognised by
the "date" command.

[*] the name -all-time is still recognised for backwards compatibility.

5.2 New -mkfs-time, -root-time and -inode-time variants
-------------------------------------------------------

5.2.1 -mkfs-time inode

This sets the filesystem make time to the latest inode timestamp in the
tar file.  Because this is a relative value (rather than absolute), it ensures
the filesystem is identical on multiple runs of Sqfstar if the content
doesn't change, it also allows filesystems with newer content to be
distinquished using the filesystem make time, and if the timestamps are updated
(due to changed content) this will produce a newer filesystem make time.

In effect this is a more nuanced way of producing reproducibility than an
absolute value.  Also the latest inode timestamp is taken from the tar file,
ignoring any fabricated timestamps (e.g. root directory), and all fabricated
timestamps are set to the latest inode value too.  This means the -root-time
option is no longer necessary if the -mkfs-time inode option is used.

5.2.2 -root-time inode

This sets the root directory timestamp to the latest inode timestamp in the
tar file.  If -mkfs-time inode is specified this option is no longer
necessary.

5.2.3 -inode-time inode

This sets all the inode timestamps to the latest inode timestamp in the
tar file.  I doubt there are many use-cases for this, but it keeps the
functionality matching between options.


5.3 New easier to remember shorthand options
--------------------------------------------

5.3.1 -repro

This option makes Sqfstar build a reproducible filesystem image.  This is
equivalent to -mkfs-time inode, which achieves reproducibility by setting the
filesystem build time to the latest inode timestamp.  Obviously the image won't
be reproducible if the timestamps or content changes.

5.3.2 -repro-time <time>

This option makes Sqfstar build a reproducible filesystem image.  This is
equivalent to specifying -mkfs-time <time> and -inode-time <time>, which
achieves reproducibility by setting all timestamps to <time>.  This option can
be used in cases where timestamps may change, and where -repro cannot be used
for this reason.

5.4 the environment variable SOURCE_DATE_EPOCH
----------------------------------------------

As an alternative to the above command line options, you can set the environment
variable SOURCE_DATE_EPOCH to a time value.

This value will be used to set the mkfs time.  Also any file timestamps which
are after SOURCE_DATE_EPOCH will be clamped to SOURCE_DATE_EPOCH.

See https://reproducible-builds.org/docs/source-date-epoch/ for more
information.


6. SPECIFYING THE UIDs/GIDs USED IN THE FILESYSTEM
--------------------------------------------------

By default files in the generated filesystem use the ownership of the file
stored in the TAR archive.  TAR archives depending on the TAR format stores
ownership in two ways.  The early V7 format only stored a numeric UID and GID.
If Sqfstar is reading a V7 archive these are used.  Later ustar and PAX
archives can also store a string user name and group name.  If these are
present and are recognised by the system (i.e. can be mapped to a UID and GID),
their UID and GID is used.  If a user name or group name is not recognised
by the system, the numeric UID or GID is used.

The -all-root option forces all file UIDs/GIDs in the generated Squashfs
filesystem to be root.  This allows root owned filesystems to be built without
root access on the host machine.

The "-force-uid value" option forces all files in the generated Squashfs
filesystem to be owned by value, where value can either be a user name or a
numeric UID.

The "-force-gid value" option forces all files in the generated Squashfs
filesystem to be group owned by value, where value can be either a group name or
a numeric GID.


7. SPECIFYING THE FILE PERMISSIONS USED IN THE FILESYSTEM
---------------------------------------------------------

By default files and directories in the generated filesystem inherit the
permissions of the files and directories in the TAR archive.  However, Sqfstar
provides a number of options which can be used to override the permissions.

The "-force-file-mode <mode>" option sets all the file (non directories)
permissions to <mode>.  <Mode> can be symbolic or octal (see the next subsection
for the Symbolic mode specification).  The octal mode sets the permissions to
that value, and the symbolic mode specification can either set the permissions,
or add or subtract permissions from the existing file permissions.

The "-force-dir-mode <mode>" option sets all the directory permissions to
<mode>.  <Mode> can be symbolic or octal (see the next subsection for the
Symbolic mode specification).  The octal mode sets the permissions to that
value, and the symbolic mode specification can either set the permissions, or
add or subtract permissions from the existing directory permissions.

The reason why the options are split into two, one that acts on files and one
that acts on directories, is because the permission bits have different
semantics for files and directories.  For example the 'x' bit means execute for
files and search for directories.  Hence you might want to ensure the 'x'
bit is set for directories for user/group/other but not want to make all files
executable.

Some examples:

% sqfstar -force-dir-mode 0700 image.sqfs < archive.tar

Will set all directory permissions to drwx------, or read, write and search
for owner, and nothing for everyone else.

% sqfstar -force-file-mode 0666 image.sqfs < archive.tar

Will set all file permissions to -rw-rw-rw-, or read and write by everyone,
but executable by no-one.

% sqfstar -force-file-mode go-w,u+rw image.sqfs < archive.tar

Will modify each file's permissions, by removing write ('w') for group and
other, but also ensure the owner can read and write to the files by adding 'r'
and 'w'.

7.1 SYMBOLIC MODE SPECIFICATION
-------------------------------

The symbolic mode is of the format [ugoa]*[[+-=]PERMS]+.  PERMS = [rwxXst]+ or
[ugo], and the sequence can be repeated separated with commas.

A combination of the letters ugoa specify which permission bits will be
affected, u means user, g means group, o means other, and a means all or ugo.

The next letter is +, - or =.  The letter + means add to the existing permission
bits, - means remove the bits from the existing permission bits, and = means set
the permission bits.

The permission bits (PERMS) are a combination of [rwxXst] which
sets/adds/removes those bits for the specified ugoa combination, r means read, w
means write and x means execute for files or search for directories.  X has a
special meaning, if the file is a directory it is equivalent to x or search, but
if it is a non-directory, it only takes effect if execute is already set for
user, group or other.  The s flag sets user or group ID on execution, and the t
flag on a directory sets restricted deletion, or historically made the file
sticky if a non-directory.

The permission bits can also be u, g or o, which takes the permission bits from
the user, group or other of the file respectively.


8. EXCLUDING FILES FROM THE FILESYSTEM
--------------------------------------

Sqfstar can exclude files from the TAR archive, so that they don't appear in
the Squashfs filesystem.  Exclude files can be directly specified on
the command line (immediately after the FILESYSTEM argument), or the -ef
option can be used to specify an exclude file, with one exclude file per line.

Exclude files by default use wildcard matching (globbing) and can match on
more than one file (if wildcards are used).  Regular expression matching
can be used instead by specifying the -regex option.  In most cases wildcards
should be used rather than regular expressions because wildcard matching
behaviour is significantly easier to understand!

In addition to wildcards/regex expressions, exclude files can be "anchored" or
"non-anchored".  An anchored exclude is one which matches from the root of the
directory and nowhere else, a non-anchored exclude matches anywhere.  For
example given the directory hierarchy "a/b/c/a/b", the anchored exclude "a/b"
will match "a/b" at the root of the directory hierarchy, but it will not match
the "/a/b" sub-directory within directory "c", whereas a non-anchored exclude
would.

A couple of examples should make this clearer.

Anchored excludes

  1. sqfstar image.sqsh 'test/*.gz' < archive.tar

     Exclude all files matching "*.gz" in the top level directory "test".

  2. sqfstar image.sqsh '*/[Tt]est/example*' < archive.tar

     Exclude all files beginning with "example" inside directories called
     "Test" or "test", that occur inside any top level directory.

  Using extended wildcards, negative matching is also possible.

  3. sqfstar image.sqsh 'test/!(*data*).gz' < archive.tar

     Exclude all files matching "*.gz" in top level directory "test",
     except those with "data" in the name.

Non-anchored excludes

  By default excludes match from the top level directory, but it is
  often useful to exclude a file matching anywhere in the source directories.
  For this non-anchored excludes can be used, specified by pre-fixing the
  exclude with "...".

  Examples:

  1. sqfstar image.sqsh '... *.gz' < archive.tar

     Exclude files matching "*.gz" anywhere in the source directories.
     For example this will match "example.gz", "test/example.gz", and
     "test/test/example.gz".

  2. sqfstar image.sqsh '... [Tt]est/*.gz' < archive.tar

     Exclude files matching "*.gz" inside directories called "Test" or
     "test" that occur anywhere in the source directories.

  Again, using extended wildcards, negative matching is also possible.

  3. sqfstar image.sqsh '... !(*data*).gz' < archive.tar

     Exclude all files matching "*.gz" anywhere in the source directories,
     except those with "data" in the name.


9. REDUCING CPU AND I/O USAGE
-----------------------------

By default Sqfstar will use all the CPUs available to compress and create the
filesystem, and will read from the TAR archive and write to the output
filesystem as fast as possible.  This maximises both CPU usage and I/O.

Sometimes you don't want Sqfstar to use all CPU and I/O bandwidth.  For those
cases Sqfstar supports two complementary options, -processors and -throttle.

The -processors option can be used to reduce the number of parallel compression
threads used by Sqfstar.  Reducing this to 1 will create the minimum number of
threads, and this will reduce CPU usage, and that in turn will reduce I/O
(because CPUs are normally the bottleneck).

The -throttle option reduces the speed Sqfstar reads from the TAR archive.
The value is a percentage (obviously from 1 - 100), and 50 will reduce the
read rate by half (the read thread will spend half its time idling), and 75
by three quarters.  Reducing the speed of I/O will also reduce the CPU
usage as there is insufficient data rate to use all cores.

Which option should you use?  Both will effectively reduce CPU and I/O in
normal cases where intensive use is being made of both I/O and CPUs.  But
in edge cases there can be an imbalance where reducing one has no effect, or
it can't be reduced any further.  For instance when there is only 1 or 2 cores
available, setting -processors to the minimum of 1 may still use too much
CPU.  Additionally if your input source is slow Sqfstar may still max it out
with -processors set to the minimum of 1.  In this case you can use -throttle
in addition to -processors or on its own.


10. FILTERING AND ADDING EXTENDED ATTRIBUTES (XATTRs)
-----------------------------------------------------

Sqfstar has a number of options which allow extended attributes (xattrs) to be
filtered from the TAR archive or added to the created Squashfs filesystem.

The -no-xattrs option removes any extended attributes which may exist in the
TAR archive, and creates a filesystem without any extended attributes.

The -xattrs-exclude option specifies a regular expression (regex), which
removes any extended attribute that matches the regular expression from all
files.  For example the regex '^user.' will remove all User extended attributes.

The -xattrs-include option instead specifies a regular expression (regex)
which includes any extended attribute that matches, and removes anything
that does't match.  For example the regex '^user.' will only keep User
extended attributes and anything else will be removed.

Sqfstar also allows you to add extended attributes to files in the Squashfs
filesystem using the -xattrs-add option.  This option takes an xattr name and
value pair separated by the '=' character.

The extended attribute name can be any valid name and can be in the namespaces
security, system, trusted, or user.  User extended attributes are added to files
and directories (see man 7 xattr for explanation), and the others are added to
all files.

The extended attribute value by default will be treated as binary (i.e. an
uninterpreted byte sequence), but it can be prefixed with 0s, where it will be
treated as base64 encoded, or prefixed with 0x, where it will be treated as
hexidecimal.

Obviously using base64 or hexidecimal allows values to be used that cannot be
entered on the command line such as non-printable characters etc.  But it
renders the string non-human readable.  To keep readability and to allow
non-printable characters to be entered, the 0t prefix is supported.  This
encoding is similar to binary encoding, except backslashes are specially
treated, and a backslash followed by three octal digits can be used to encode
any ASCII character, which obviously can be used to encode non-printable values.

The following four command lines are equivalent

sqfstar -xattrs-add "user.comment=hello world" image.sqfs
sqfstar -xattrs-add "user.comment=0saGVsbG8gd29ybGQ=" image.sqfs
sqfstar -xattrs-add "user.comment=0x68656c6c6f20776f726c64"
sqfstar -xattrs-add "user.comment=0thello world" image.sqfs

Obviously in the above example there are no non-printable characters and so
the 0t prefixed string is identical to the first line.  The following three
command lines are identical, but where the space has been replaced by the
non-printable NUL '\0' (null character).

sqfstar -xattrs-add "user.comment=0thello\000world" image.sqfs
sqfstar -xattrs-add "user.comment=0saGVsbG8Ad29ybGQ=" image.sqfs
sqfstar -xattrs-add "user.comment=0x68656c6c6f00776f726c64" image.sqsh


11. PSEUDO FILE SUPPORT
-----------------------

Sqfstar supports pseudo files, these allow files, directories, character
devices, block devices, fifos, symbolic links, hard links and extended
attributes to be specified and added to the Squashfs filesystem being built,
rather than requiring them to be present in the TAR archive.  This, for example,
allows device nodes to be added to the filesystem without requiring root access.

Pseudo files also support "dynamic pseudo files" and a modify operation.
Dynamic pseudo files allow files to be dynamically created when Sqfstar
is run, their contents being the result of running a command or piece of
shell script.  The modifiy operation allows the mode/uid/gid of an existing
file in the source filesystem to be modified.

Two Sqfstar options are supported, -p allows one pseudo file to be specified
on the command line, and -pf allows a pseudo file to be specified containing a
list of pseduo definitions, one per line.

11.1 CREATING A DYNAMIC FILE
----------------------------

Pseudo definition

Filename f mode uid gid command

mode is the octal mode specifier, similar to that expected by chmod.

uid and gid can be either specified as a decimal number, or by name.

command can be an executable or a piece of shell script, and it is executed
by running "/bin/sh -c command".   The stdout becomes the contents of
"Filename".

Examples:

Running a basic command
-----------------------

/somedir/dmesg f 444 root root dmesg

creates a file "/somedir/dmesg" containing the output from dmesg.

Executing shell script
----------------------

RELEASE f 444 root root \
		if [ ! -e /tmp/ver ]; then \
			echo 0 > /tmp/ver; \
		fi; \
                ver=`cat /tmp/ver`; \
                ver=$((ver +1)); \
                echo $ver > /tmp/ver; \
                echo -n `cat /tmp/release`; \
                echo "-dev #"$ver `date` "Build host" `hostname`

Creates a file RELEASE containing the release name, date, build host, and
an incrementing version number.  The incrementing version is a side-effect
of executing the shell script, and ensures every time Sqfstar is run a
new version number is used without requiring any other shell scripting.

The above example also shows that commands can be split across multiple lines
using "\".  Obviously as the script will be presented to the shell as a single
line, a semicolon is need to separate individual shell commands within the
shell script.

Reading from a device (or fifo/named socket)
--------------------------------------------

input f 444 root root dd if=/dev/sda1 bs=1024 count=10

Copies 10K from the device /dev/sda1 into the file input.  Ordinarily Sqfstar
given a device, fifo, or named socket will place that special file within the
Squashfs filesystem, the above allows input from these special files to be
captured and placed in the Squashfs filesystem.

11.2 CREATING A BLOCK OR CHARACTER DEVICE
-----------------------------------------

Pseudo definition

Filename type mode uid gid major minor

Where type is either
	b - for block devices, and
	c - for character devices

mode is the octal mode specifier, similar to that expected by chmod.

uid and gid can be either specified as a decimal number, or by name.

For example:

/dev/chr_dev c 666 root root 100 1
/dev/blk_dev b 666 0 0 200 200

creates a character device "/dev/chr_dev" with major:minor 100:1 and
a block device "/dev/blk_dev" with major:minor 200:200, both with root
uid/gid and a mode of rw-rw-rw.

11.3 CREATING A DIRECTORY
-------------------------

Pseudo definition

Filename d mode uid gid

mode is the octal mode specifier, similar to that expected by chmod.

uid and gid can be either specified as a decimal number, or by name.

For example:

/pseudo_dir d 666 root root

creates a directory "/pseudo_dir" with root uid/gid and mode of rw-rw-rw.

11.4 CREATING A SYMBOLIC LINK
-----------------------------

Pseudo definition

Filename s mode uid gid symlink

uid and gid can be either specified as a decimal number, or by name.

Note mode is ignored, as symlinks always have "rwxrwxrwx" permissions.

For example:

symlink s 0 root root example

Creates a symlink "symlink" to file "example" with root uid/gid.

11.5 CREATING HARD LINKS (FILE REFERENCES)
------------------------------------------

The "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 a quite frequent use of the definition, alternative faster
"File reference" or Hard Link Pseudo definitions exist:

README l pathname

and

README h pathname

If pathname was "/home/phillip/latest-version/README", then both will create a
reference to "/home/phillip/latest-version/README", and obviously the
timestamp/mode and owership will be used.

The difference between the 'l' and 'h' definitions is in the handling of
symbolic links, if the pathname points to a symbolic link, the 'l' definition
won't follow it, and so you'll get a reference to the symbolic link, whereas
the 'h' definition will follow it, and you'll get a reference to whatever
the symbolic link points to.

The definition also can 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.

11.6 CREATING SOCKETS/FIFOS
---------------------------

Pseudo definition

filename i mode uid gid [s|f]

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

11.7 ADDING EXTENDED ATTRIBUTES TO FILES
----------------------------------------

Pseudo definition

filename x name=val

Will add the extended attribute <name> to <filename> with <val > contents.  See
Section 7 for a description of the <val> formats supported.

11.8 MODIFYING ATTRIBUTES OF AN EXISTING FILE
---------------------------------------------

Pseudo definition

Filename m mode uid gid

mode is the octal mode specifier, similar to that expected by chmod.

uid and gid can be either specified as a decimal number, or by name.

For example:

dmesg m 666 root root

Changes the attributes of the file "dmesg" in the filesystem to have
root uid/gid and a mode of rw-rw-rw, overriding the attributes obtained
from the TAR archive.


11.9 SPECIFYING A DEFAULT PSEUDO DIRECTORY DEFINITION
-----------------------------------------------------

The option

-pd <d mode uid gid>

Specifies a default pseudo directory which will be used in pseudo definitions if
a directory in the pathname does not exist.  This also allows pseudo definitions
to be specified without specifying all the directories in the pathname.  The
definition should be quoted.

For example this provides an alternative way of specifying any leading
directories in a pseudo file definition.  We could create a pseudo file
/dir1/dir2/file on the command line this:

% sqfstar -p "/dir1 d 0777 0 0" -p "/dir1/dir2 d 0777 0 0"
  -p "/dir1/dir2/file f 0777 0 0 echo hello_world" image.sqfs < archive.tar

Here we have to create each directory explicitly.

The new -pd option allows us to specify a default pseudo directory and then
just specify the pseudo definition for the file, leaving Sqfstar to
automatically create the leading directories:

% sqfstar -pd "d 0777 0 0" -p "/dir1/dir2/file f 0777 0 0 echo hello world" < archive.tar


12. EXTENDED PSEUDO FILE DEFINITIONS WITH TIMESTAMPS
----------------------------------------------------

The Pseudo file definitions described above do not allow the timestamp
of the created file to be specified, and the files will be timestamped
with the current time.

Extended versions of the Pseudo file definitions are supported 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 "d" definition, but it takes a <time>
timestamp.

The list of extended definitions are:

	filename F time mode uid gid command
	filename D time mode uid gid
	filename B time mode uid gid major minor
	filename C time mode uid gid major minor
	filename S time mode uid gid symlink
	filename I time mode uid gid [s|f]
	filename M time mode uid gid

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

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.

% sqfstar img.sqfs -p "file D \"1 jan 1980\" 0777 phillip phillip" < archive.tar

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

12.1 SPECIFYING A DEFAULT PSEUDO DIRECTORY DEFINITION WITH TIMESTAMP
--------------------------------------------------------------------

The option

-pd <D time mode uid gid>

Specifies a default pseudo directory which will be used in pseudo definitions if
a directory in the pathname does not exist.

The upper-case D indicates this is an extended pseudo definition which takes
a <time> timestamp.  <time> can be either an unsigned decimal integer or a
"date string" which is parsed and converted into an integer by calling the
"date" command.


13. MISCELLANEOUS OPTIONS
-------------------------

The -info option displays the files/directories as they are compressed and
added to the filesystem.  The original uncompressed size of each file is
printed, along with DUPLICATE if the file is a duplicate of a file in the
filesystem.

The -info-file option does the same except that the output is to a file rather
than to stdout.  This allows the -info-file option to be used together with the
progressbar.

The -nopad option informs Sqfstar to not pad the filesystem to a 4K multiple.
This is performed by default to enable the output filesystem file to be mounted
by loopback, which requires files to be a 4K multiple.  If the filesystem is
being written to a block device, or is to be stored in a bootimage, the extra
pad bytes are not needed.
