[GSoC 2019] Improving the btrfs filesystem

Blog post by brj on Tue, 2019-05-07 23:57

Introduction

Hello, world!

As some of you might be aware, I'm one of the students selected for GSoC 2019. My name is Bharathi Ramana Joshi. You may know me as brj, my initials, from the mailing lists and IRC channel. I'm pursuing an underguate degree in Computer Science and Engineering from Keshav Memorial Institute of Technology, India.

Project: Improving the btrfs filesystem

As the title suggests, I shall be working on Haiku's btrfs implementation for most of the summer. Currently, it is possible to read and write directories but only read from files. By the end of GSoC 2019, I aim to implement file writing.

Overview

btrfs uses a modified b-tree (without linkage) suitable for high-performance object store that can perform copy-on-write snapshots, while maintaining good concurrency. These modified trees store generic items sorted on a 136-bit key. The first 64 bits of the key are a unique object id, the middle 8 bits are an item type field and the last 64 bits are used in type specific ways.

User-visible files and directories are contained in a file system tree. Within each filesystem tree, each file and directory object has an inode item. Within each directory, directory entires appear as directory items, whose right-hand key values are a CRC32C hash of their filename and their data is a location key, or the key of the inode item it points to.

File data is kept outside the trees in extents, which are contiguous runs of disk blocks. Extent blocks default to 4KiB in size, do not have headers and contain only file data. Files have extent data items to track the extents which hold their contents. The item's right-hand key value is the starting byte offset of the extent.

The extent allocation tree acts as an allocation map for the file system. Their left-hand and right-hand key values are the starting offsets and lengths of the regions they represent. The filesystem zones its allocated space into block groups, which are variable-sized allocation regions that altrenate successively between preferring metadata extents (tree nodes) and data extents (file contents).

The project can be conveniently divided into three parts for the three phases.

Phase 1

Small files that occupy less than one block will be packed into the btree inside the extent item.

During phase 1, writing will be implemented for such files, along with Copy-on-Write (CoW) logging.

Testing will be done using btrfs_shell. This overcomes the tedious task of rebuilding entire source, generating an iso and booting it each time changes are made.

Another task during phase 1 would be to properly document the current implementation. Some documentation has already been written during initial contributions.

Phase 2

The main task during phase 2 is implementing write supports for large files, albiet without Copy-on-Write de-duplication.

Extensive testing using btrfs_shell must be performed with extreme operating parameters. btrfs_shell will have to be modified to perform such testing.

Documentation must also be written for all work done upto this point

Phase 3

Phase 3 aims at implementing what phase 2 doesn't - Copy-on-Write. Transaction handling must also be implemented.

Testing should be done to make sure transactions are correctly handled and CoW works without any problems. btrfs_shell must be modified appropriately.

Finally, documentation should be added, the project should be made more presentable for final evaluation and any left over work must be wrapped up.

Miscellaneous

  • I have set up a blog on blogger where I shall be updated my progress on a daily basis, like a diary. My weekly posts on this blog are the crux of all the work done in that week.
  • For more details about my project, see my proposal