rawgrep 0.1.3

Grep at the speed of raw disk
Documentation

rawgrep

Grep at the speed of raw disk - search text by reading data directly from raw block devices.

How is rawgrep so fast?

  • rawgrep reads files DIRECTLY from your partition, completely bypassing the filesystem.
  • rawgrep is cache-friendly and INSANELY memory efficient, simply streaming through your device and outputting the matches.
  • rawgrep uses a sophisticated fragment-based caching system (inspired by nowgrep) that learns which files can be skipped for repeated searches.

Installation

Prerequisites

  • Linux (contribute to make rawgrep support Windows) system with ext4/ntfs filesystem
  • Rust toolchain (for building from source)
  • Root access or be able to set capabilities

Option 1: One-Time Setup with Capabilities (Recommended)

git clone https://github.com/rakivo/rawgrep
cd rawgrep

cargo build --profile=release-fast

# If you want maximum speed possible (requires nightly):
# cargo +nightly build --profile=release-fast --target=<your_target> --features=use_nightly

# Run the one-time setup command. Why? Read "Why Elevated Permissions?" section
sudo setcap cap_dac_read_search=eip ./target/release-fast/rawgrep

Now you can run it without sudo:

rawgrep "search pattern"

Option 2: Use sudo Every Time

If you prefer not to use capabilities, just build and run with sudo:

cargo build --profile=release-fast

# Again, if you want maximum speed possible (requires nightly):
# cargo +nightly build --profile=release-fast --target=<your_target> --features=use_nightly

# Run with sudo each time
sudo ./target/release-fast/rawgrep "search pattern"

Usage

Basic Search

# Search current directory
rawgrep "error"

# Search specific directory
rawgrep "TODO" /var/log

# Regex patterns
rawgrep "error|warning|critical" .

Advanced Options

# Specify device manually (auto-detected by default)
rawgrep "pattern" /home --device=/dev/sda1

# Print statistics at the end of the search
rawgrep "pattern" . --stats

# Disable filtering (search everything)
rawgrep "pattern" . -uuu
# or
rawgrep "pattern" . --all

# Disable specific filters
rawgrep "pattern" . --no-ignore # Don't use .gitignore
rawgrep "pattern" . --binary    # Search binary files

Filtering Levels

# Default: respects .gitignore, skips binaries and large files (> 5 MB)
rawgrep "pattern"

# -u: ignore .gitignore
rawgrep "pattern" -u

# -uu: also search binary files
rawgrep "pattern" -uu

# -uuu: search everything, including large files
rawgrep "pattern" -uuu

Why Elevated Permissions?

rawgrep reads raw block devices (e.g., /dev/sda1), which are protected by the OS. Instead of requiring full root access via sudo every time, we use Linux capabilities to grant only the specific permission needed.

What is CAP_DAC_READ_SEARCH?

This capability grants exactly one permission: bypass file read permission checks.

rawgrep only reads data, it never writes anything to disk.

Verifying Capabilities

You can verify what capabilities the binary has:

getcap ./target/release-fast/rawgrep
# Output: ./target/release-fast/rawgrep = cap_dac_read_search+eip

Removing Capabilities

If you want to revoke the capability and go back to using sudo:

sudo setcap -r ./target/release-fast/rawgrep

Limitations (IMPORTANT)

  • ext4/ntfs only: Currently only supports ext4/ntfs filesystems.

Development

Note: Capabilities are tied to the binary file itself, so you'll need to re-run setcap after each rebuild.

Why no automation script? I intentionally decide not to provide a script that runs sudo commands. If you want automation, write your own script, it's just a few lines of bash code and you'll understand exactly what it does.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Roadmap

  • Support for Windows. (Some physical partition stuff needs to get fixed on Windows, besides that everything should be already working)
  • Symlink support

Emacs Integration

To use rawgrep from Emacs with jumpable locations:

  1. Download emacs/rawgrep.el and place it in your load path
  2. Add to your .emacs or init.el:
(require 'rawgrep)
(global-set-key (kbd "M-e") 'rawgrep)

Or if you use use-package:

(use-package rawgrep
  :load-path "path/to/rawgrep.el"
  :bind ("M-e" . rawgrep))

Works exactly like 'grep-find but better.

FAQ

Q: Is this safe to use? A: Yes. The tool only reads data and never writes. The CAP_DAC_READ_SEARCH capability is narrowly scoped.

Q: Is rawgrep faster than ripgrep? A: Yeah.

Q: Why am I missing some matches? A: By default, rawgrep respects .gitignore and skips binary/large files. Use -u to ignore .gitignore, -uu to also search binaries, or -uuu to search everything. This matches ripgrep's behavior.

Q: Can I use this on other filesystems? A: Currently only ext4 is supported. Support for other filesystems may be added in the future. (Motivate me with stars)

Q: Will this damage my filesystem? A: No. The tool only performs read operations. It cannot modify your filesystem.

Q: What if partition auto-detection fails? A: Specify the device manually with --device=/dev/sdXY. Use df -Th to find your partition.

Acknowledgments

Inspired by ripgrep and nowgrep, and the need for high-quality software in the big 25.