Scalability of Ext4

One of the features of the file system Ext4 i开发者_如何学编程s its new scalability features documented here. I was wondering what changes at the source code level would enable Ext4 to support such features, like the virtually unlimited number of sub directories which Ext3 didn't. Pointers to the relevant areas of the Kernel source tree would be greatly appreciated.

I have tried to find a clue in fs/ext4/ sources but haven't been able to yet.

The paper "The new ext4 filesystem: current status and future plans" by Mathur et al. describes the changes from ext3 to ext4 from a file system developer perspective.

Prove me wrong, but I think that in ext4 as in ext2/ext a directory has blocks using direct, indirect and multiple-leve indirect pointers. dentry entries are stored in such block. So as a file as a maximal file size in ext4 a directory as a maximal number of dentries. However, as the paper describes, using directory indexing using a special B-Tree variant is active by default.