What it is...
Another "Union File System". Basically a translucent file system! The idea is that you grab branches from other file systems, combine them (some parts read-only, some read-write), and having done that, present the result to the target Linux system (the usual way, using mount). The filesystem type is aufs which the kernel doesn't natively understand, so we need to compile support into the kernel itself, or make a kernel module.Why do we need it?
- So we can write to otherwise read-only filesystems (eg: CDRoms). While this isn't technically what happens, this is how it appears to the system user. Technically it's simulating a read-write filesystem (a RAMdisk) and writing to that instead.
- UnionFS was used in the past for this purpose, but it's been subjected to criticism for lack of stablilty and over-complexity. It was time to write Another Union FS!!
- To create custom filesystems from a variety of sources!!
How does it work?
AuFS consists of a kernel module (aufs.ko) and userspace tools (unionctl). AuFS can be enabled in the kernel by copying the compiled module to /lib/modules/`uname -r`/fs and it will be available after the next reboot, or immediately using:# depmod -a `uname -r`
With this new ability, the kernel can be told to mount an AuFS filesystem like this:
# mount -t aufs -o dirs=/path-to/ro-branch=ro:/path-to/rw-branch=rw none /mnt/point