squash_dir is an init-script for OpenRC (used e.g. by Gentoo) which allows to keep a directory compressed by squashfs but simultaneously allows to write on it
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



(C) Martin Väth (martin at mvath.de). The license of this package is the GNU Public License GPL-2.


This project is no longer maintained. Use instead the successor project(s)

squashmount is more generic, support also overlay(fs) from linux-3.18 and newer and squashfuse and works without problems with any init system. Moreover, squashmount has a highly improved control interface.

The above openrc-wrapper script might receive further development, while the corresponding script from squash_dir is frozen and no longer maintained.

What is this project?

This is essentially an initscript for openrc and/or systemd which allows to keep a directory compressed by squashfs but simultaneously allows to write on it using some of (depending on the configuration and what is available):

The idea is that on shutdown the data is recompressed (and the temporary modified data removed). This approach is originally due to synss' script from http://forums.gentoo.org/viewtopic-t-465367-highlight-.html

In that forum thread you can also ask for help about this project.

If the Gentoo portage tree /usr/portage is compressed this way (without $DISTDIR which you should store somewhere else when using this script), the required disk space is only about 50 MB (instead of about 180 MB, the actual space requirement depending essentially on the filesystem), and usually the access is even faster.

In addition, this project contains an openrc-wrapper script which can be used to use simple initscripts of openrc even if openrc is not running: In particular, these scripts can then be used with systemd (and also this script is sometimes handy from the command line for administrative purposes). If the initscripts are simple enough, openrc need not even be installed for the `openrc-wrapper``to work: For this reason, this project will also work with only systemd, although natively it is meant for openrc.


The script requires of course that squashfs support is activated in the kernel (and supports the COMPRESSION method), that the mksquashfs tool is available, and also that some of the above mentioned unionfs-type tools is available and supported by the kernel. Moreover, a POSIX shell and find are needed with the following extensions (which are not yet POSIX):

  1. The find command must know the -path option
  2. A mktemp program must be available in $PATH if you make use of the temporary file name feature (described later)
  3. flock is needed unless you disable file locking (see LOCKFILE below).

Some standard tools like /bin/false are also assumed (if /bin/false is missing, it is also ok, but it must not do something else than return nonzero status). Of course, for the case that you use the FILE_TBZ feature, also the corresponding TARCMD and its requirement (like tar and bzip2) must exist.

If you want that the hard status line is set, also the title script from https://github.com/vaeth/runtitle (version 2.3 or newer) is required in your $PATH.

It is strongly recommended to put

alias squash_dir='noglob squash_dir'

into your ~/.zshrc, /etc/zsh/zshrc, or /etc/zshrc, so that things like

squash_dir start *

will work in your zsh as intended without the need to quote *. (I assume that you do not use a poor shell instead of zsh.)

Main Example

In this example, it is assumed that you have already installed/copied the /etc/init.d/squash_dir script and that you want to keep the three directories /usr/portage, /var/db, and /usr/share/texmf-dist compressed (as a remark: other good candidates for compression are the kernel sources and on some systems also /usr/share/games). Create three symbolic links:

  • ln -s squash_dir /etc/init.d/squash_portage
  • ln -s squash_dir /etc/init.d/squash_db
  • ln -s squash_dir /etc/init.d/squash_tex

You have to create the corresponding three config-files of the same name in /etc/conf.d/. Here are typical examples of the content of these files:







Optionally, you can also modify global defaults in /etc/conf.d/squash_dir; see the comments in the same conf.d/squash_dir file for details

The meaning of this configuration data is the following: The first line determines the directory you want to keep compressed and be writable; the corresponding compressed data will be expected/created in the file $DIRECTORY.sqfs (e.g. in /usr/share/portage.sqfs). Moreover, later, there will be auxiliary directories $DIRECTORY.changes and $DIRECTORIES.readonly available. The latter is the content of the compressed data without any modifications (this directory is readonly), and the former contains the modified files (and some metadata, depending on which of overlayfs | aufs | unionfs-fuse | unionfs | funionfs you are using). The THRESHOLD value means that your changes will not be compressed on every shutdown but only if $DIRECTORY.changes gets large enough. Moreover, the IGNORETOUCH-value in squash_tex means that changes only in the date of the file /usr/share/texmf-dist/ls-R or /usr/share/texmf-dist/generic/config/language.{dat{,.lua},def} should not cause a recompression on shutdown (these files are usually often recreated with the same content, and so if nothing else happens, it is reasonable to ignore this change). The FILE_SQFS_OLD value in /etc/conf.d/squash_db means that a backup should always be kept of the previous compressed data for /var/db; this is for security reasons, since that directory is rather vital to your system (the other two directories could be re-synced resp. re-emerged in case something bad happens, but if /var/db is lost, you have practically completely messed up your system). The settings of COMPRESSION mean the following:

  • squash_tex: Use the usually best compressing algorithm (currently: xz)
  • squash_portage: Use the default mksquash algorithm (currently: gzip)
  • squash_db: Use the lzo algorithm (usually the fastest)

Of course, you should make sure that the squashfs in your kernel supports the corresponding algorithms.

Now to actually make these settings active, start the initscripts. You can do this manually in the classical way using (in case of openrc)

  • /etc/init.d/squash_portage start
  • /etc/init.d/squash_tex start
  • /etc/init.d/squash_db start

or using (in case of systemd)

  • systemctl start squash_dir@portage.service
  • systemctl start squash_dir@tex.service
  • systemctl start squash_dir@db.service

or in a more convenient way using the /usr/sbin/squash_dir command line interface to start them all with only one command:

  • squash_dir start or
  • squash_dir start portage tex db (this assumes of course that /usr/sbin/squash_dir will be found in your $PATH).

Be very careful that you made no mistake before, when you make this step! The starting of the initscripts will create the corresponding squashed files and clean the directories (i.e. if you want to undo this change for some reason using unsquashfs on the created files, you might have lost some information permanently like e.g. hard-links within the cleaned directories).

If the above step appears too dangerous to you, you can of course also create the file manually

  • mksquashfs /usr/portage /usr/portage.sqfs
  • mksquashfs /usr/share/texmf-dist /usr/share/texmf-dist.sqfs
  • mksquashfs /var/db /var/db.sqfs

and if this was successful, for testing only move the corresponding directories /usr/portage, /usr/share/texmf-dist, /var/db somewhere else; actually, the latter is not mandatory: The directories will just get "over-mounted", i.e. you will not see the previous content anymore.

If the squashed file exists (i.e. if you created them manually using 'mksquashfs), you can start the initscript safely, i.e. in this case$DIRECTORY` will not get cleaned automatically.

In order to start the initscripts also on the next boot, add them to your default runlevel (if you use openrc):

  • rc-config add squash_portage default
  • rc-config add squash_tex default
  • rc-config add squash_db default

If you use systemd, to start all configured scripts at once on system start:

  • systemctl enable squash_dir

When you shutdown the computer (more precisely: When the above initscripts are stopped) any modifications to the directories are recompressed (unless the THRESHOLD is not reached or the touching of /usr/share/texmf-dist/ls-R was the only modification).

In some situations you might want to recompress the directories even before you shutdown the system (or perhaps even if the THRESHOLD is not yet reached or no serious modifications were done). In this case you can call e.g. (in openrc)

  • /etc/init.d/squash_portage restart
  • /etc/init.d/squash_tex restart
  • /etc/init.d/squash_db restart

or (in systemd)

  • systemctl restart squash_dir@portage.service
  • systemctl restart squash_dir@tex.service
  • systemctl restart squash_dir@db.service

or you can use the /usr/sbin/squash_dir command line interface with corresponding parameters to execute (some or all) of these commands in a row (calling the interface without parameters shows the options; the interface offers with -s resp. -sf also the possibility to e.g. create magic files to override the THRESHOLD option).

If you plan to squash the Gentoo portage tree /usr/portage, you should keep $DISTDIR in a different directory in advance: It makes no sense (i.e. it costs enormous time and needs lot of temporary diskspace without much gain) to compress compressed tarballs. Hence: Move /usr/portage/distdir to some other place (outside the /usr/portage hierarchy) and modify $DISTDIR in /etc/portage/make.conf correspondingly, before you start the initscripts for the first time.

A Word of None-Warning

It is in general rather safe to squash a directory, even a rather vital one: Even if e.g. you boot from a kernel which has no support for some of aufs | overlayfs | unionfs-fuse | unionfs | funionfs to make the directory writable, the script will mount it at least as read-only (using mount --bind if necessary). Moreover, if everything goes wrong you can still use unsquashfs to unpack the directory manually. Probably the only danger in packing "strange" directories are special files like hard links (this information will usually get lost) or special devices which are perhaps not supported by the used tools.

Modules and Mounting

If you compiled squashfs, aufs, overlayfs, or fuse as modules, you should modprobe these modules first (or better put them into /etc/conf.d/modules). squash_dir will make no attempt to load the modules (unless this happens automatically by the corresponding mount program): The script will just attempt to mount the directories using the corresponding tool and will check its return status for success. If the mount fails, the next tool is attempted for mounting, until one succeeds (the order and which tools are attempted can be influenced using the ORDER variable described below). If no tool succeeds, it is attempted to use mount --bind to get the directory at least readonly on the expected place, so even in this bad situation (which probably only happens if you boot from an experimental kernel or a brand new kernel without corresponding support) you can still access the directory read-only. Hence, also rather vital directories can be compressed as long as it is not vital to write to them (and as long as the relevant programs for mounting etc. are not contained withing these directories, of course).

Patching squashfs-tools

It is recommended to use a patched version of squashfs-tools which redirects its progress bar to stderr (instead of stdout). Cf. the description of the VERBOSE_MODE variable for details.

Call of the Initscripts

Besides the usual options like start/stop/restart, there are also helper options for the /etc/init.d/squash_* scripts which are meant to be used from other scripts (for example, they are used by the /usr/sbin/squash_dir command line interface).

  • STOP

    Executes the actions of stop without actually stopping. Moreover, this can be called even if the script was not started. Use this only in case of fatal problems and with extreme care!


    Execute the actions of start without actually starting. For instance, it is not required that dependencies are started. Moreover, this can be called even if the script was started. Use this only in case of fatal problems and with extreme care!


    Execute the actions STOP && START without actually restarting. For instance, dependencies are not restarted. Moreover, this can be called even if the script was stopped. Use this only in case of fatal problems and with extreme care!

  • will_squash

    returns true/false (and a corresponding message unless -q is used) depending on whether the stop operation would re-squash or kill $DIR_CHANGE (the latter means that $KILL_FILE exists).

  • need_squash

    returns true/false depending on whether new data was written. (If no $MAGIC_FILE exists - see below - this is the same as will_squash).

  • have_magic

    returns true/false depending on whether $MAGIC_FILE exists.

  • have_kill

    returns true/false depending on whether $KILL_FILE exists.

  • print_dir_change

    outputs the effective path to DIR_CHANGE to stderr, taking the temporary name feature into account (see below).

  • print_dir_squash

    outputs the effective path to DIR_SQUASH to stderr, taking the temporary name feature into account (see below).

  • print_ignore_threshold

    outputs the effective value of IGNORE_THRESHOLD, taking the temporary name feature into account (see below). If THRESHOLD is not used, the empty string is output.

  • print_magic_file

    outputs the effective value of MAGIC_FILE, taking the temporary name feature into account (see below).

  • print_kill_file

    outputs the effective value of KILL_FILE, taking the temporary name feature into account (see below).

(It may be faulty to use the print_* commands if the temporary name feature is used in the corresponding paths and the script was not started).

Unfortunately, the return value of initscripts is ignored with openrc. Therefore, the answer (1 for true, 0 for false) for the first three functions is written to stderr instead of being passed as an exit status. For the directory names of the print_dir_* options a / is appended to make it easier for calling scripts to deal with trailing spaces.

Variables (Configuration in /etc/conf.d/squash_*)

The subsequent variables can be defined in /etc/conf.d/squash_* Three values are mandatory:

DIRECTORY must be defined, and usually you will want to specify at least DIR_CHANGE and probably also DIR_SQUASH; for the others, default values are assumed as described below.

The corresponding directories will be created if they do not exist. Some variables support the "temporary name feature". This means that a path ending with XXXXXX (like e.g. /tmp/squash_foo.XXXXXXXX) is transformed into a corresponding file/directoryname using mktemp (or mktemp -d, respectively) when it is used. Using paths in world-writable directories (e.g. /tmp or /dev/shm) for some data is a security risk and therefore strongly discouraged. An exception of this rule is of course for those paths for which you use the temporary name feature which is handled safely - that's why this feature was introduced.


    The directory where the squashed filesystem should finally be mounted, i.e. the path of the "original" directory. If you want to use this script to compress your portage tree, it might be a good idea to set this variables with

    DIRECTORY=`. /etc/portage/make.conf 2>/dev/null; printf '%s' "$PORTDIR"`

    or the slower but more portable

    DIRECTORY="`portageq portdir`"

    The only restriction concerning DIRECTORY is that none of the following files or directories should reside within this directory. Note that you will get problems if you squash a directory containing data needed to run this script (e.g. the kernel module for squashfs or - if you ever plan to upgrade your kernel - tools to build the kernel modules) because you can of course not access the squash'ed directory until this script was successfully started...


    This variable is only used in the defaults of the following variables. If you leave it empty, it defaults to /tmp.


    The directory where the modifications of DIRECTORY will go. This should be on a partition where you have sufficient space. This variable supports the temporary name feature described above. However, you should use a temporary directory or a ramdisk only if your changes to DIRECTORY will always only be temporary, i.e. if you are really prepared to loose your changes after a cleanup of the temporary directory or after a reboot. Normally (i.e. if you do not use the temporary name feature and DIR_CHANGE is not cleaned) all your changes will survive a reboot even if you make use of the MAGIC_FILE feature below.

    Note that some earlier versions of squash_dir failed to empty DIR_CHANGE if it used the temporary file name feature, but now it should be save to use it.

    You can leave this variable empty: If you do this, squash_dir will assume that you do not want to be able to write to DIRECTORY, i.e. DIRECTORY will be mounted readonly. In this case you will probably also want to leave the following variable empty.


    If this variable is true (not empty, not 0 or - and does not start with f, F, n, or N), the script will remove DIR_CHANGE at stopping. If it contains p or P also empty parents are removed. RM_DIR_CHANGE defaults to parents if the temporary name feature is used for DIR_CHANGE.


    This is only used in the defaults of DIR_SQUASH. It defaults to /run, /var/run, or /, depending on what exists.


    This is a directory which is needed for technical reasons when DIR_CHANGE is non-empty (i.e. if you really want to mount DIRECTORY writable). This directory will contain a read-only version of the contents of FILE_SQFS. In case things go wrong when making DIRECTORY writable (e.g. if aufs/overlayfs/unionfs-fuse/unionfs/funionfs all fail due to missing support by the kernel) you can still access DIR_SQUASH; you can also use this directory to compare the content of currently changed files with their "original" stored in the most current version of FILE_SQFS.

    With some earlier versions of squash_dir, the temporary name feature was not or not properly supported for this variable, but now it is safe to use it.

    If you leave this variable empty and DIR_CHANGE is nonempty, DIR_SQUASH defaults to $RUNPATH/$SVCNAME.readonly.

    If DIR_CHANGE is empty, DIR_SQUASH is not really needed and should usually be left empty. Nevertheless, if you set DIR_SQUASH anyway, the DIR_SQUASH directory will contain an identical copy of DIRECTORY (this is useful if you use some scripts which rely on DIR_SQUASH, and you change DIR_CHANGE temporarily from nonempty to empty).


    If this variable is true, the script will remove DIR_SQUASH at stopping (if DIR_SQUASH is set and the directory does not contain files). If it contains the letter p, also empty parents are removed. RM_DIR_SQUASH defaults to true if the default (and nonempty) value is used for DIR_SQUASH. It defaults to parents if the temporary name feature is used for DIR_SQUASH.


    If you use the temporary name feature for DIR_CHANGE or DIR_SQUASH, this variable must contain the name of a file which is used to store these temporary names. If no errors occur, this file is created/removed on start/stop. If NAME_FILE is empty, it defaults to $RUNPATH/$SVCNAME.


    This is a filename (which supports the temporary name feature) which is used during shutdown for the squashfs-file which is created from DIRECTORY. After successfull creation that file is moved to FILE_SQFS. On the partition of this file there must be enough space to store the new FILE_SQFS. If possible, you should use the same partition on which also FILE_SQFS is stored (then the move command does not have to copy the data once more). Alternatively, if you have sufficient ram, you can also use a ramdisk like /dev/shm/dir_squash_tmp_sqfs.XXXXXXXX (but you will not get much speed increase from doing so). If TMP_SQFS is empty, it defaults to $TMPDIR/$SVCNAME.sqfs.XXXXXXXX.


    This is the file which contains the actual squashfs data of DIRECTORY (this file must exist when you start this script. See the output message of the script on how to create it.) If you leave this empty, it defaults to $DIRECTORY.sqfs.


    If defined, this file will be an (automatically updated) copy of your previous FILE_SQFS file.


    Here you can define the mv command which is used to create FILE_SQFS_OLD. You might want to set this e.g. to mv --backup --, mv --version-control --, or similar things. The default is mv --.


    If defined, this file will contain a copy of FILE_SQFS, but in .tar.bz2 format (updated during stopping). This may be handy in case of problems when you do not have access to a kernel with an sqfs-module or to unsquashfs.


    This is the command (including options) used to create the FILE_TBZ archive. It defaults to tar -cjf, but you might want set it to /home/bin/tbzd -R if you installed my compression scripts there (this will save slightly more space by stripping the parent directory names and using bzip -9). Of course, you can also use your own script analogously or use another compression program.


    If defined, this file will be an (automatically updated) copy of your previous FILE_SQFS_TBZ file.


    Here you can define the mv command which is used to create FILE_TBZ_OLD. See MV_FILE_SQFS.


    This is a list of paths relative to the top-level $DIRECTORY whose changes (which actually occur in $DIR_CHANGE) are ignored when squash_dir decides whether the directory needs to be re-squashed. So, for example if you define

    IGNORETOP='\"A \\\"magic\\\" link\" \"subdir/device-*\"'

    then changes in the files/links/devices/whatever called

    • $DIR_CHANGE/A "magic" file
    • $DIR_CHANGE/subdir/device-something may follow here

    will get ignored. Do not use this for directories if you also want to ignore the content of the directories - use IGNOREDIR for the latter. The format of the list is space-separated but quoting is allowed. More precisely, the list is first eval'ed, and then each item is eval'ed within "..." context (and interpreted as a filenane pattern), therefore the included " needs to be triple-quoted and the space and * needs not be quoted if used within " ... ".


    This is like IGNORETOP, but it is explicitly checked that the match is a file (a link to a file does not count either); those of other type are ignored.


    This is like IGNORETOPFILE, but it is explicitly checked that the match is a directory. In this case, also the content of the directory is ignored.


    This is like IGNORETOP with the difference that the content is not a list of names but an expression which is passed to find in an eval "..." context. Here you can check also for other files which are not in the top-level directory. If you use -path, note that the pathname starts with $DIR_CHANGE, so e.g. to ignore the link .bad_link in the top-level of $DIRECTORY, but only if this is really a link) you can define

    IGNORE='-path "\$DIR_CHANGE"/.bad_link -type l'

    If your match is a directory and you call -prune, then also the content of the directory is ignored.


    This is similar to IGNORE, but it is additionally (automatically) added code for find to check that the match is a file.


    This is similar to IGNORE, but it is additionally (automatically) added code for find to check that the match is a directory. Moreover, in this case also -prune is called to ignore the content of this directory.


    This is a list of paths relative to the top-level $DIRECTORY whose changes (which actually occur in $DIR_CHANGE) are ignored, provided that the diff utility shows that the file was not modified compared to the corresponding original file in $DIR_SQUASH. The format of the list is that of a case statement, i.e. you can use file globbing and | to separate alternatives and have to quote | or ) if you use it as symbols. Note that no pruning of directories is done, i.e. with each directory also its whole content is checked against IGNORETOUCH; directories are by definition always equal if they existed before. Typical example:


    Even if the file $DIRECTORY/ls-R was recreated by one of your system tools, no re-squashing will be done if the new file is identical to the old one and if there were no other changes. Note that also parent directories must be explicitly listed to ignore the possible changes in them.

  • KEEP

    If $KILL_FILE does not exist but $DIR_CHANGE is cleaned without re-squashing (this can only happen if no other files than those in IGNORE or IGNORETOUCH were changed) then the files in this variable are kept anyway. Currently, only files/subdirs of $DIRECTORY are supported, no deeper nesting is allowed (i.e. you must not use / here). The format is that of case, i.e. you can use wildcards and separate several names with |. Quoting is of course supported. You must not use ) in the name unless you quote it.

    Example: KEEP=".*|'*)'" will keep all files/subdirs of the form $DIRECTORY/.* or $DIRECTORY/*)"


    If $KILL_FILE exists (so that $DIR_CHANGE is cleaned without re-squashing) then the files in this variable are kept anyway. The syntax is the same as that of KEEP.


    If the variable DIR_CHANGE is empty, this variable is ignored. Otherwise, its value should be a string consisting of all (or some) of the words

    overlayfs aufs unionfs-fuse funionfs unionfs (*)

    Then the corresponding tool is used to make DIRECTORY writable. The tools are attempted in the given order: If the first tool fails, a warning message is printed, and the next tool is attempted and so on. If ORDER is empty, it defaults to (*).


    Normally, if the directory $DIR_CHANGE is nonempty on shutdown, the squash file and the tbz file are created (and the backups are made), and afterwards directory $DIR_CHANGE is deleted. However, if the file $MAGIC_FILE exists during shutdown, nothing of these things happens, i.e. the information in directory $DIR_CHANGE is simply kept for the next start of this script (or for your manual interaction). In particular, creating the file $MAGIC_FILE temporarily, you can "delay" the update during shutdown until you delete $MAGIC_FILE again. This is useful e.g. for /usr/src/linux which you usually will not want to recompress after recompilation of the kernel. If you leave the variable MAGIC_FILE empty, it defaults to $DIR_CHANGE/.no-save.


    This is similar to MAGIC_FILE, but with the difference that if this file exists during shutdown, directory $DIR_CHANGE is deleted nevertheless; only the files in $KILL_NOT are kept. So be very careful when creating this file: It means that your changes will be lost completely on shutdown or restart! If you leave the variable KILL_FILE empty, it defaults to $DIR_CHANGE/.No-save.


    If this variable is a positive number, and directory $DIR_CHANGE is not larger than THRESHOLD kilobytes, then the effect is as if MAGIC_FILE exists, i.e. the update during shutdown will not happen. You can override this with:


    If this file exists, then THRESHOLD is ignored during shutdown. If this file resides in directory $DIR_CHANGE, then it is removed before the compression. Of you leave IGNORE_THRESHOLD empty, it defaults to $DIR_CHANGE/.do-save.


    Specify the compression method used by mksquashfs. If empty, the default mksquashfs algorithm (currently: gzip) is used. Other possible values are gzip, xz, lzo. If this variable is not specified, it defaults to that algorithm which is presumably best compressing (which is currently xz).


    These options are used for mksquashfs (e.g. -check_data). The option -noappend is used automatically. Depending on VERBOSE_MODE and COMPRESSION also the options -no-progress or -comp ... might be used automatically.


    If this variable is

    • 0

      then -no-progress is appended to mksquashfs options and the output of mksquashfs is redirected to /dev/null.

    • 1 or `` (empty)

      then stdout of mksquashfs is redirected to /dev/null Usually, this is the same as 0, only marginally slower. However, you might apply a patch for mksquashfs to redirect the progress bar to stderr. This produces the nicest output with no redundant information.

    • 2 then mksquashfs is called without additional options or redirections.


    These are additional options used for the mount-command for aufs. The default is -o noatime. You might want to add options like -o rdblk=0 -o rdhash=0 which help you if you have setup aufs for userspace (RDU), see man aufs (e.g. from sys-fs/aufs*[-util]) for details.


    These are additional options used for the mount-command for overlayfs. The default is -o noatime.


    These are additional mount options used for unionfs-fuse. The default is:

    -o cow -o allow_other -o use_ino -o nonempty -o noatime -o hide_meta_files

    So if you modify this variable, you should probably include these options. Omit them only if you know what you are doing.


    This variable is only used in the default value of MOUNT_UNIONFS_FUSE. If you do not set it, the default is either -o hide_meta_files or empty, depending on whether unionfs-fuse is new enough. If you use unionfs-fuse, you might want to set this variable explicitly in order to save time for determining whether unionfs-fuse is new enough (or if you do not agree with the default).


    These are additional mount options used for unionfs. The default is empty.


    These are additional mount options used for funionfs. The default is -o allow_other -o nonempty. So if you modify this variable, you should probably include these options. Omit them only if you know what you are doing.


    These are options which are used for umounting. This defaults to -i because /sbin/umount.aufs causes problems in some cases.


    If this variable is unset or true, a lazy umount is attempted after a failed umount. This means that the chances are good that this script will successfully stop even if the directory is still in use. However, it might happen that e.g. loop devices are not freed correctly.


    Since mount/umount calls cannot properly be done in parallel (since apparently they do not properly lock /etc/mtab), we use flock for these calls to create a lockfile (which is never erased) to allow rc_parallel=YES. You can set this variable to the following values:

    1. to a path of the lockfile. The path must be absolute, i.e. starting with /. It is dangerous to use a user-writable directory here (like e.g. /tmp).
    2. to the magic value auto (or if you do not set this variable): In this case the lockfile /etc/mtab.lock is used unless /etc/mtab is a symbolic link. In the latter case the lockfile feature is switched off.
    3. In all other cases (e.g. if you set LOCKFILE= or LOCKFILE=no) the lockfile feature is switched off.