Here's a little introduction on how to use FuseCompress.

Basic use

There are two basic ways to use FuseCompress. The first is

fusecompress /foo

This starts FuseCompress and mounts /foo as a compressed filesystem, meaning that (almost) all files you write to /foo will be compressed and transparently decompressed when you read them again. Once you unmount the compressed filesystem using

fusermount -u /foo

you will see the compressed files in directory /foo as they are actually stored on disk.

The other way to use FuseCompress is

fusecompress /foo /bar

This mounts a compressed filesystem at /bar where you can store and retrieve transparently compressed data, but stores the actual files on disk in directory /foo and thus allows you to see the compressed data while the filesystem is mounted. There are not many cases where this is useful, and you will usually want to use the first method.

WARNING: Mounting FuseCompress filesystems over their backing directory (first method) can lead to lockups when using libfuse 2.7.x. (See issue #14.) This problem does not seem to occur when using separate mount points and backing directories (second method) or with libfuse 2.8.0pre1.

Choosing a compression method

By default, FuseCompress runs using LZO compression. It is also possible to use different methods:

fusecompress -c lzma /foo

This mounts a compressed filesystem at /foo using LZMA compression. This means that all new files created (and all files that had to be uncompressed for some reason) will be compressed using LZMA. Existing files will retain whatever compression method they were created with. The other methods available are Deflate (-c gz), bzip2 (-c bz2), and null (-c null, only useful for debugging).

It is also possible to give the compression level as a filesystem option (see below), which is helpful when mounting FuseCompress filesystems from /etc/fstab.

WARNING: The LZMA Utils library used in FuseCompress is not stable yet, and different revisions may create mutually incompatible data streams, potentially leaving you unable to read your previously written data in case of an update.

Compression level

Some compression methods (LZMA, Deflate, and bzip2) support different compression levels (1 to 9), with the lower levels being faster with lower compression ratio, and the higher levels slower (sometimes dramatically so) with higher compression ratio. The defaults are 4 for LZMA, and 6 for Deflate and bzip2. If you want to use a different level, you can use the -l option:

fusecompress -c lzma -l 7 /foo

There is a wiki page that tries to help you in choosing your compression method and level.

FUSE options

FuseCompress allows you to pass options to FUSE using the -o option. Commonly used options are:

• allow_other - allows other users to access the files inside; by default, only the user mounting the filesystem can see the files inside. For the root user, allow_other is on by default.
• suid and dev - allows SUID binaries and device files to be used inside your compressed filesystem (default when mounting as root)
• lzo, gz, lzma, and bzip2 select the corresponding compression method. This way of specifying it works better in /etc/fstab.
• cache or cache_skipped causes FuseCompress to cache the decompressed data it comes along while seeking for another bit of data. This avoids repeated decompression (and possibly recompression) of files on random read accesses. This is a new feature and has not yet been extensively tested, but its use is highly recommended anyway, especially in combination with advanced algorithms such as LZMA, where decompression and especially recompression take an awful long time. The maximum cache size defaults to 100 MB and can be changed using the cache_size option. For instance, cache_size=32 reduces the maximum size to 32 MB.
• uncompressed_binaries tells FuseCompress not to compress files in the standard binary directories (such as /bin, /usr/bin, /sbin, etc.) or shared objects (files with .so in their names). This option is probably only useful if you do not want to use cache_skipped. The idea is to prevent the permanent decompression and recompression of frequently used binaries, which will not usually occur when using cache_skipped.
• detach and nodetach force FuseCompress to run in the background or in the foreground, respectively. If these options are not given, release builds will detach, while debug builds will not.
• noterm keeps FuseCompress from exiting when it receives a SIGTERM signal. This option is vital if you are using FuseCompress as your root filesystem. Most (if not all) distributions send a SIGTERM to all processes when shutting down. If FuseCompress heeded that signal you would no longer have access to your root filesystem and your system would hang. noterm is the default when mounting a filesystem as root. If you want the terminating behavior instead, use the option term.

(There is a related issue with the SIGKILL signal that is usually sent to processes that won't die voluntarily when sent a SIGTERM. SIGKILL cannot be handled and will in all likelihood result in data corruption if it catches FuseCompress - or indeed any other FUSE filesystem - with its pants down. This issue has to be fixed in your distributions init system before you can use FUSE filesystems for / safely.)

• maxcompress specifies the file size above which FuseCompress will no longer try to compress a file. There is no such limit by default. Note that FuseCompress cannot know the final size of a file that is created from scratch, so it is still possible that newly created files above that size will be written with compression enabled.
• nocompext allows the user to extend the list of uncompressible file types built into FuseCompress. For instance, nocompext=foo,nocompext=bla will keep files like file.foo and image.bla from being compressed.

Example:

fusecompress -o gz,cache_skipped /

This mounts a compressed root filesystem using Deflate compression and with caching of uncompressed data. (Caveat: Do not do this unless you can make sure that your system automatically executes that command on the next reboot, before using any of the files in the root filesystem.)

Access Control

FuseCompress filesystems mounted by the root user exhibit the same default behavior as regular kernel filesystems. For instance, they allow SUID binaries and devices. When mounting a filesystem as a regular user, however, none of these are enabled by default, and no other user (including root!) can access the filesystem. If you want to enable other users to see your mounted filesystem, you need to give the option allow_other. This is only possible if it has been explicitly enabled in /etc/fuse.conf. SUID binaries and devices cannot be enabled by regular users.