Fuzzing GitPython

This directory contains files related to GitPython's suite of fuzz tests that are executed daily on automated infrastructure provided by OSS-Fuzz. This document aims to provide necessary information for working with fuzzing in GitPython.

The latest details regarding OSS-Fuzz test status, including build logs and coverage reports, is available on the Open Source Fuzzing Introspection website.

How to Contribute

There are many ways to contribute to GitPython's fuzzing efforts! Contributions are welcomed through issues, discussions, or pull requests on this repository.

Areas that are particularly appreciated include:

• Tackling the existing backlog of open issues. While fuzzing is an effective way to identify bugs, that information isn't useful unless they are fixed. If you are not sure where to start, the issues tab is a great place to get ideas!
• Improvements to this (or other) documentation make it easier for new contributors to get involved, so even small improvements can have a large impact over time. If you see something that could be made easier by a documentation update of any size, please consider suggesting it!

For everything else, such as expanding test coverage, optimizing test performance, or enhancing error detection capabilities, jump into the "Getting Started" section below.

Getting Started with Fuzzing GitPython

Tip

New to fuzzing or unfamiliar with OSS-Fuzz?

These resources are an excellent place to start:

• OSS-Fuzz documentation - Continuous fuzzing service for open source software.
• Google/fuzzing - Tutorials, examples, discussions, research proposals, and other resources related to fuzzing.
• CNCF Fuzzing Handbook - A comprehensive guide for fuzzing open source software.
• Efficient Fuzzing Guide by The Chromium Project - Explores strategies to enhance the effectiveness of your fuzz tests, recommended for those looking to optimize their testing efforts.

Setting Up Your Local Environment

Before contributing to fuzzing efforts, ensure Python and Docker are installed on your machine. Docker is required for running fuzzers in containers provided by OSS-Fuzz and for safely executing test files directly. Install Docker following the official guide if you do not already have it.

Understanding Existing Fuzz Targets

Review the fuzz-targets/ directory to familiarize yourself with how existing tests are implemented. See the Files & Directories Overview for more details on the directory structure.

Contributing to Fuzz Tests

Start by reviewing the Atheris documentation and the section on Running Fuzzers Locally to begin writing or improving fuzz tests.

Files & Directories Overview

The fuzzing/ directory is organized into three key areas:

Fuzz Targets (fuzz-targets/)

Contains Python files for each fuzz test.

Things to Know:

• Each fuzz test targets a specific part of GitPython's functionality.
• Test files adhere to the naming convention: fuzz_<API Under Test>.py, where <API Under Test> indicates the functionality targeted by the test.
• Any functionality that involves performing operations on input data is a possible candidate for fuzz testing, but features that involve processing untrusted user input or parsing operations are typically going to be the most interesting.
• The goal of these tests is to identify previously unknown or unexpected error cases caused by a given input. For that reason, fuzz tests should gracefully handle anticipated exception cases with a try/except block to avoid false positives that halt the fuzzing engine.

OSS-Fuzz Scripts (oss-fuzz-scripts/)

Includes scripts for building and integrating fuzz targets with OSS-Fuzz:

• container-environment-bootstrap.sh - Sets up the execution environment. It is responsible for fetching default dictionary entries and ensuring all required build dependencies are installed and up-to-date.
• build.sh - Executed within the Docker container, this script builds fuzz targets with necessary instrumentation and prepares seed corpora and dictionaries for use.

Where to learn more:

• OSS-Fuzz documentation on the build.sh
• See GitPython's build.sh and Dockerfile in the OSS-Fuzz repository

Local Development Helpers (local-dev-helpers/)

Contains tools to make local development tasks easier. See the "Running Fuzzers Locally" section below for further documentation and use cases related to files found here.

Running Fuzzers Locally

Warning

Some fuzz targets in this repository write to the filesystem during execution. For that reason, it is strongly recommended to always use Docker when executing fuzz targets, even when it may be possible to do so without it.

Although I/O operations such as writing to disk are not considered best practice, the current implementation of at least one test requires it. See the "Setting Up Your Local Environment" section above if you do not already have Docker installed on your machine.

PRs that replace disk I/O with in-memory alternatives are very much welcomed!

Direct Execution of Fuzz Targets

Directly executing fuzz targets allows for quick iteration and testing of changes which can be helpful during early development of new fuzz targets or for validating changes made to an existing test. The Dockerfile located in the local-dev-helpers/ subdirectory provides a lightweight container environment preconfigured with Atheris that makes it easy to execute a fuzz target directly.

From the root directory of your GitPython repository clone:

1. Build the local development helper image:

docker build -f fuzzing/local-dev-helpers/Dockerfile -t gitpython-fuzzdev .

2. Then execute a fuzz target inside the image, for example:

 docker run -it -v "$PWD":/src gitpython-fuzzdev python fuzzing/fuzz-targets/fuzz_config.py -atheris_runs=10000

The above command executes fuzz_config.py and exits after 10000 runs, or earlier if the fuzzer finds an error.

Docker CLI's -v flag specifies a volume mount in Docker that maps the directory in which the command is run (which should be the root directory of your local GitPython clone) to a directory inside the container, so any modifications made between invocations will be reflected immediately without the need to rebuild the image each time.

Running OSS-Fuzz Locally

This approach uses Docker images provided by OSS-Fuzz for building and running fuzz tests locally. It offers comprehensive features but requires a local clone of the OSS-Fuzz repository and sufficient disk space for Docker containers.

Build the Execution Environment

Clone the OSS-Fuzz repository and prepare the Docker environment:

git clone --depth 1 https://github.com/google/oss-fuzz.git oss-fuzz
cd oss-fuzz
python infra/helper.py build_image gitpython
python infra/helper.py build_fuzzers --sanitizer address gitpython

Tip

The build_fuzzers command above accepts a local file path pointing to your GitPython repository clone as the last argument. This makes it easy to build fuzz targets you are developing locally in this repository without changing anything in the OSS-Fuzz repo! For example, if you have cloned this repository (or a fork of it) into: ~/code/GitPython Then running this command would build new or modified fuzz targets using the ~/code/GitPython/fuzzing/fuzz-targets directory:

python infra/helper.py build_fuzzers --sanitizer address gitpython ~/code/GitPython

Verify the build of your fuzzers with the optional check_build command:

python infra/helper.py check_build gitpython

Run a Fuzz Target

Setting an environment variable for the fuzz target argument of the execution command makes it easier to quickly select a different target between runs:

# specify the fuzz target without the .py extension:
export FUZZ_TARGET=fuzz_config

Execute the desired fuzz target:

python infra/helper.py run_fuzzer gitpython $FUZZ_TARGET -- -max_total_time=60 -print_final_stats=1

Tip

In the example above, the "-- -max_total_time=60 -print_final_stats=1" portion of the command is optional but quite useful.

Every argument provided after "--" in the above command is passed to the fuzzing engine directly. In this case:

• -max_total_time=60 tells the LibFuzzer to stop execution after 60 seconds have elapsed.
• -print_final_stats=1 tells the LibFuzzer to print a summary of useful metrics about the target run upon completion.

But almost any LibFuzzer option listed in the documentation should work as well.

Next Steps

For detailed instructions on advanced features like reproducing OSS-Fuzz issues or using the Fuzz Introspector, refer to the official OSS-Fuzz documentation.

LICENSE

All files located within the fuzzing/ directory are subject to the same license as the other files in this repository with one exception:

fuzz_config.py was migrated to this repository from the OSS-Fuzz project's repository where it was originally created. As such, fuzz_config.py retains its original license and copyright notice (Apache License, Version 2.0 and Copyright 2023 Google LLC respectively) as in a header comment, followed by a notice stating that it has have been modified contributors to GitPython. LICENSE-APACHE contains the original license used by the OSS-Fuzz project repository at the time the file was migrated.