Writing a FUSE Filesystem: a Tutorial

Joseph J. Pfeiffer, Jr., Ph.D.
Emeritus Professor
Department of Computer Science
New Mexico State University

Version of 2012-11-08

One of the real contributions of Unix has been the view that "everything is a file". A tremendous number of radically different sorts of objects, from data storage to file format conversions to internal operating system data structures, have been mapped to the file abstraction.

One of the more recent directions this view has taken has been Filesystems in User Space, or FUSE (no, the acronym really doesn't work. Oh well). The idea here is that if you can envision your interaction with an object in terms of a directory structure and filesystem operations, you can write a FUSE file system to provide that interaction. You just write code that implements file operations like open(), read(), and write(); when your filesystem is mounted, programs are able to access the data using the standard file operation system calls, which call your code.

There are many documents on the web describing how FUSE works and how to install and use a FUSE filesystem, but I haven't come across any that try to describe how to go about actually writing one. The goal of this tutorial is to meet what I see as a need for such a document.

This tutorial introduces FUSE using a filesystem I call the "Big Brother File System" (the reason for the name is that "Big Brother is watching." The filesystem simply passes every operation down to an underlying directory, but logs the operation.

This tutorial, together with its associated example filesystem, is available as a tarball at http://www.cs.nmsu.edu/~pfeiffer/fuse-tutorial.tgz.

Audience: This tutorial is aimed at developers who have some familiarity with general programming in Linux (and Unix-like operating systems in general), so you know how to untar a tarball, how Makefiles work, and so forth. I won't be going through the details of how to perform those tasks; I'll be focussing on what you need to know that's specific to using FUSE filesystems.

I am not affiliated with the FUSE project in any way, except as a user. My descriptions of the interface to fuse, and of techniques to work with it, are a distillation of my reading of the existing documentation, and my experience working with it. Consequently, any errors are mine (and corrections are welcome!).


The tutorial is divided into the following sections:

Files and Naming Conventions in This Tutorial
This section describes the files distributed as a part of this tutorial, and the naming conventions for the functions in the bbfs filesystem. It's really there to provide an overview of the whole tutorial and filesystem, not to directly provide information on FUSE.

Running BBFS
A little bit of information on mounting a filesystem with bbfs, watching the log, and unmounting it.

Callbacks and struct fuse_operations
This is the heart of a FUSE filesystem, and of this tutorial. The central concepts are discussed here.

Parsing the Command Line and Initializing FUSE
Getting your program started.

Private Data and Related Topics
Maintaining filesystem state.

Extra Information on Unclear FUSE Functions
The intent of this section is to give some extra information on FUSE functions that seem a little unclear. Write now, this section covers readdir() and FUSE's handling of the file creation flags to open().

Security Considerations and Race Conditions
Running a FUSE filesystem may raise some security issues that you should be aware of. Also, since the FUSE server is multithreaded, there may be some race conditions you need to think about.

Thanks and Other Resources
I'd like to acknowledge people who've helped by noticing bugs (or in any other way). Also, this is not the only resource on FUSE on the web.

Next: Files and Naming Conventions in This Tutorial

Last modified: Thu Nov 8 16:44:32 MST 2012