ext4-rs
can extract the basic stat
information, directory listings, and file content
from real images generated by other tools, and by the Linux kernel.
This operates directly on partitions. To read actual disc images, probably need to handle
partition tables. This can be accomplished with the bootsector
crate.
All basic file types are represented: files, directories, symlinks, char and block devices, fifos and sockets. Hard links are not a type of thing that makes sense: the item is just in multiple directories.
- No support for extended flags (e.g.
immutable
,append-only
). - Probably contains overflows and fencepost errors, many of which Rust will translate to panics for you. At least, in debug mode.
This is not a filesystem driver. It does not support efficiently modifying real filesystems. Currently, it doesn't support modifying anything at all, but that may change.
I'm not especially interested in resource-constrained platforms: memory and IO are not used efficiently.
0.7.0
: userust 1.34
features to removecast
dependency0.6.0
: usefailure
for error handling0.5.0
: updatebitflags
for associated constants, and rename some public constants.0.4.1
: fix for an infinite loop parsing directory entries
Rust 1.34 (TryFrom
) is supported, and checked by Travis.
Updating this is a semver bump.
Some of the tests read generated image files. These images are not directly checked into git,
and will be unpacked by build.rs
. These files are apparently very large, but should take
very little space. This requires a decent tar
to be on the path.
These test assets can be rebuilt (on Linux, with root) by running
./extract-test-data --refresh
.
The code is licensed under the super-permissive MIT license.
However, a number of struct and bitfield definitions, and some maths expressions, are lifted directly from the Linux, or e2fsprogs, source code. These code-bases are under the GPLv2. I believe this to be fair use: these places are the documentation, and only interface definitions have been extracted, no code. I leave the final decision to you.
This would have been practically impossible without the work a Djwong
has done on
the ext4 disk layout page on
the kernel wiki. I believe this person is Darrick Wong. Thanks, man.