The p-Code Machine

I noticed this post on kerneltrap. It’s a new filesystemen designed by an ex SUSE employee if I’m not mistaken. I think Chris worked on ReiserFS a while back. Anyway btrfs has some really cool features, including checksums on everything just like ZFS. Another feature taken from ZFS are the subvolumes. A single volume can be divided into several subvolumes. Subvolumes are basically separately mountable volumes, except that are on the same volume. This is done by giving the filesystem a seperate B-trees for each subvolume. This is really neat, as this allows for more dynamics allocation of storage. However this comes at a price, theoretically this means that subvolume corruption could affect other subvolumes on the same master volume.

But why create a new filesystem when we can swipe the ZFS code if it might get relicensed soon. Well, even if we could legally swipe the ZFS code, filesystem ports from different operation systems generally involve nasty glue layers because of different VFS designs. In the past XFS has been known for this. By creating our own filesystems we can create clean code designed from scratch for the Linux VFS layer.