Scroll to navigation

fusefile(8) System Manager's Manual fusefile(8)

NAME

fusefile, fusedisk - FUSE file mount for combining file fragments

SYNOPSIS

fusefile [fuse-opts] mountpoint [overlay] filename/from-to ...
fusefile -dump [fuse-opts] mountpoint [overlay] filename/from-to ...
fusefile -push [fuse-opts] mountpoint [overlay] filename/from-to ...
fusedisk [fuse-opts] mountpoint [overlay] filename/from-to ...

DESCRIPTION

fusefile is a FUSE file mount that presents a series of fragments of other files as a contiguous concatenation. Technically it bind mounts a driver on top of the filename mountpoint to provide access to the given file fragments as if in a single, contiguous file.

fusefile accepts over-writing on the fused file (i.e. the mountpoint) which gets distributed accordingly to the fragments. But neither the fused file nor the fragments can change size; any writing thus merely over-writes content without truncating fragments. All fragment files are held open while fusefile is active.

By using the optional -overlay:filename argument between the mount point and the fragments, an overlay file may be set up. The overlay file will then be used by fusefile for capturing writes to the fused file (i.e. the mountpoint). The overlay file will contain any new written fused file regions followed by meta data to distinguish between new, written content and old content that comes from the fragments.

By instead using the -overlay:list argument where list is a colon-separated list of filenames, fusefile will use those as an ordered stack of overlays and "inject" them as fragments on top of each other.

The option -dump as first argument together with a fusefile setup will print the setup to standard output rather than establishing a fusefile mount. This is of most use with a prior overlay setup where then the printout includes the portions of updates that have been captured in the overlay. The printout is the series of fusefile fragment argments to give in order to intersperse the captured overlay portions according to the overlay table.

The option -push as first argument together with a fusefile setup will push the overlay into the sources (except for write-protected fragments). This is only of use with a prior overlay setup where then the updates that have been captured in the overlay get pushed into the fragments.

fusedisk is a helper script to set up a fusefile as a block device (via fuseblk) by using the device mapper (dmsetup) to manage an empty block device mapping where content is handled at the mountpoint via fusefile. (Note that the same thing may be done with the device manager directly, but then all fragments need to be in sectors of N*512 bytes whereas with fusedisk, only the fused file as a whole is "clipped" at nearest N*512 bytes below actual size)

FRAGMENT ARGUMENTS

The fragment arguments include the filename of a source file, and optionally start and end byte positions. All in all there five variations:

include all of the file. A pathname that includes "/" must be ended with an extra "/" since that last "/" separates the filename from the range detail.

include the range from the given start to end. Either "start" or "end" or both may be omitted, to mean the beginning and end of the file respectively. If "start" or "end" are less than 0 then it means relative to the end of the file.

include "length" bytes from the given start. A negative "start" means relative to the end of the file. If "length" is negative or omitted it means that position relative to the end.

same as "/start+"

Note that a negative start position is clipped to 0 and a too large end position is clipped to the end of the file.

Character devices are treated as being of any given finite size, but have size 0 by default. For example, "/dev/zero/:100" means a fragment of 100 NUL bytes.

OPTIONS

This section enumerates the most interesting options to use with fuesfile. See "man fuse" and "man mount" for more options.

The -dump "option" tells fusefile to print out the applicable fragment sequence for the current setup, including the overlay table, if any. The printout is done instead of setting up a mount point.

The fuse option -oallow_other is needed for sharing the fused file with other users who otherwise will not have access to it (including "root"). Note however that this must first be enabled in /etc/fuse.conf.

The fuse option -ononempty may need to be used when reusing an existing file as mountpoint.

These mount options, where ... is a user or group id respectively, are useful for root when using fusedisk and thereby give user or group ownership for the mount to the nominated user or group.

EXAMPLES

This section illustrates uses of fusefile.

Exanple 1.

Insert a file "y" into a file "x" at position 1200.

$ fusefile -ononempty x x/:1200 y x/1200:
This also shadows the original file "x" and presents the fused file instead.

Example 2.

Make fused file y be a swap of the beginning and end of file "x", at position 2442.

$ fusefile y x/2442: x/:2442

Example 3.

Replace partition 2 of an image file, A, with a different file, X. For this example the partition table looks as follows.

$ partx -oNR,START,SECTORS A
NR START SECTORS
1 2048 2097152
2 2099200 409600
3 2508800 14268383

As each sector is 512 bytes the clipping points around partition 2 are 1074790400 and 1284505600 and the insertion size is 209715200 bytes. The fusefile comman will therefore be as follows.
$ fusefile -ononempty A A/:1074790400 X/:209715200 A/1284505600
Note that the fused file shadows the original file A.

Example 4.

Protect raw disk image file with an overlay:

$ fusefile -ononempty disk.raw -overlay:today disk.raw
By that set up, the overlay file, "today", will protect the disk image file, "disk.raw" from changes, and also override the pathname "disk.raw" to be the fused file.

Example 5.

A fusefile mount with an overlay file is writable regardless of the fused fragments, but all updates are written to the overlay file instead of to the fragments.

$ fusefile -ononempty A -overlay:B A

The overlay file, B in the example, contains all changes to the shadowed original file, A. The overlay file contains only the newly written regions and content is otherwise obtained from the original file.

To that end, the overlay file also contains a "marker table" at the end as if appended to the fused file. That part of the file is outside of the fused file; and it's simply an element count followed by pairs of byte addresses that tell which regions of the fused file have been captured into the overlay file. (The marker table is of course maintained so that adjoining regions are collapsed)

Thus, an overlay file may be reused to later re-establish the same fused file with overlay as previously, to continue capturing more changes.

Example 6.

As final example, we set up a fused block device y as a swap of the beginning and end of file "x", at position 2442:

$ sudo fusedisk -ouid=1000 y x/2442: x/:2442
Note the use of sudo for becoming root, which is required for block device handling, and also the -ouid=1000 option so as to make the block device y be owned by the user with id 1000.

NOTES

fusefile opens the nominated source files before any bind mounting. With the fuse option -ononempty it will bind over an non-empty file, which may be useful. The source files remain open, but the source fragments are not recomputed. If a source file changes the fused file will present the new content. If a source is reduced in size, access will be inconsistent.

If the mountpoint file doesn't exist, then fusefile creates it.

Unmounting is done with "fusermount -u mountpoint" as usual. A fusedisk mount is unmounted by root using umount.

SEE ALSO

fuse, fusermount, mount, dmsetup

AUTHOR

Ralph Rönnquist <ralph.ronnquist@gmail.com>.