A good overview of the changes to Btrfs is available in Chris Mason's Git-Pull request. For instance, Red Hat developer Josef Bacik introduced some changes that make the "Next Generation Linux File System" cache a list of free blocks in a special inode if the file system was integrated with "-o space_cache". This makes Btrfs "dramatically faster" after a fresh mount, because its structures no longer need to be scanned for free blocks. Just before the release of RC4 of 2.6.37, the developers also added various corrections to the Btrfs code to fix, for instance, problems in connection with the use of Direct I / O and the interaction with NFS.
As briefly explained in the pertaining Git-Pull request, the developers have extensively restructured the code for the Ceph cluster file system, which is partially based on Btrfs. They moved parts of the file system code to the libceph kernel library, which is also the basis for the new and experimental Rados Block Device (RDB). Similar to nbd or iSCSI, RDB allows the creation of block devices that store data on the network – in this case in a Ceph cluster. Further background details and a short description of the function can be found in a blog posting written by the Ceph developers and in the Ceph wiki.
NFS, XFS etc.
The CIFS code for accessing Samba or Windows shares now supports the multi-user and mfsymlinks mount options. A new idmapper allows the NFS client to be quicker and more flexible when mapping user and group IDs to names – details about this function can be found in the pertaining documentation. Various optimisations to the Readdir code of NFSv4 can considerably speed up the listing of directories that contain a large number of files, as demonstrated by test results the patch developer included in his commit comment. In his Git-Pull request, J. Bruce Fields describes one of the most important changes to the NFSD code. A good overview of the changes to XFS is available in the "XFS Status Update For October 2010" – for instance, various metadata optimisations (such as 1) are designed to further improve scalability, reportedly speeding up by over 100% the removal of 50 million files in a test.
The GFS2 cluster file system is no longer classified as experimental and now supports Fallocate. Various other changes to GFS can be found in Steven Whitehouse's Git-Pull request; the request by Joel Becker lists the most important changes to OCFS2, while the request by Ryusuke Konishi describes the changes to NILFS2.