Block-Based Filesystem with RAID using FUSE

This project is a user-level block-based filesystem developed using FUSE in C, implementing core features such as file creation, reading, writing, directory management, and RAID-level storage (RAID 0, RAID 1, RAID 1v). Designed as part of the Operating Systems curriculum (CS537), it demonstrates low-level filesystem logic including inode and block bitmap management, memory-mapped file I/O, and page-aligned metadata structures.

Contributors

• Hrishikesh Shinde
• Ritik Singh

Table of Contents

• Overview
• Key Features
• Usage
• Mounting Behavior
• Error Handling
• Testing
• Structure
• Debugging Tips
• Further References
• Contributors

Overview

The filesystem supports:

• Traditional block-based layout with inodes, bitmaps, and a superblock.
• RAID 0 (striping), RAID 1 (mirroring), and RAID 1v (verified mirroring).
• Verified mirroring performs majority-read validation across disks.
• Full integration with FUSE to support mkdir, rmdir, read, write, unlink, and more.

This project is composed of:

• mkfs.c – Initializes a new filesystem on given disk images.
• wfs.c – Entry point for the FUSE-based filesystem.
• fuse_operations.c – Contains the FUSE operation implementations.
• wfs.h – Contains all the filesystem structure definitions.
• Utility scripts: create_disk.sh, umount.sh, Makefile

Key Features

• Block-aligned superblock, inode structures, and data blocks (512 bytes each)
• RAID 0 and RAID 1 support with metadata mirroring
• RAID 1v: Majority-based data verification during reads
• Lazy directory parsing and inode-based file structure
• Supports the following FUSE callbacks:
  • getattr, mknod, mkdir, unlink, rmdir, read, write, readdir

Usage

Build

make

Create Disks

./create_disk.sh     # Modify this script to create multiple disks

Format Filesystem

./mkfs -r 1 -d disk1 -d disk2 -i 32 -b 200

Mount Filesystem

mkdir mnt
./wfs disk1 disk2 -f -s mnt

Interact

mkdir mnt/dir
echo "data" > mnt/file
cat mnt/file
ls mnt

Mounting Behavior

• Filesystem must be mounted with the same number of disks used during formatting.
• Disk order during mount does not matter; disk image filenames can be changed.
• Filesystem mode (RAID 0, 1, 1v) is stored in the superblock.
• Valid modes: -r 0, -r 1, -r 1v

Error Handling

The filesystem returns standard Linux error codes when appropriate:

• -ENOENT: File or directory does not exist
• -EEXIST: File or directory already exists
• -ENOSPC: No space left on device

Testing

This repository includes a tests/ directory with functional tests to validate:

• File and directory creation
• RAID behavior (RAID 0, 1, and 1v)
• Reading and writing across block boundaries
• Mount and unmount correctness
• Edge cases like full disk, invalid flags, and max file size

To run tests:

make test

Structure

• mkfs.c – Formats disks with a fresh filesystem and metadata layout
• wfs.c – Main function for FUSE mounting
• fuse_operations.c – Core filesystem logic and FUSE callbacks
• wfs.h – Structs for superblock, inodes, dirents, and constants
• create_disk.sh – Script to create zeroed disk images
• umount.sh – Script to unmount the filesystem
• Makefile – Compiles all parts of the project

Debugging Tips

• Add printf() statements inside each FUSE callback to trace execution.
• Inspect raw disk contents with:

xxd -e -g 4 disk1 | less

• Run the filesystem in gdb with:

gdb --args ./wfs disk1 disk2 -f -s mnt

• Always make clean before testing or submitting.