s2e-env

A command-line tool for creating and administering isolated development environments for S2E. Each environment contains all the tools required to run S2E plus one or more "projects". A project is essentially an analysis target. For example, one project might be the analysis of a CGC binary, while another project might be the analysis of the file program from Coreutils.

Prerequisites

We assume that you are working on an Ubuntu 14.04 or 16.04 64-bit OS. Repo only works with Python 2.7, so you should use Python 2.7 too. You will also need gcc and python-dev installed.

Some commands (such as basic block coverage) require a disassembler. Supported disassemblers include:

• IDA Pro
• Radare. Requires r2pipe
• Binary Ninja. Requires GUI-less processing

Install

git clone https://github.com/S2E/s2e-env.git
cd s2e-env
pip install .

# Note: if you use pip earlier than version 19, you must add --process-dependency-links:
pip install . --process-dependency-links

If you wish to install s2e-env to a Python virtual environment, please create and activate this virtualenv before installing s2e-env with pip.

Tests

Tests can be run via

python setup.py test

Configuring

s2e-env is configurable in two ways. Firstly, there is a global YAML configuration file located in s2e_env/dat/config.yaml. This configuration file controls how all environments are created. You are not normally required to modify the settings in this file. If you wish to customize how environments are created, you should edit this file before running pip install to install s2e-env.

For example, you may want to clone the S2E source repos via SSH rather than HTTPS, in which case you would set the repos, url option to [email protected]:S2E.

A second YAML configuration file, s2e.yaml, is created in each S2E environment. This contains settings that are local to each S2E environment. For example, if you want to generate basic block coverage, you will also have to set the ida, path option.

Usage

The package can be installed via pip, thus making the s2e command available.

To list the available commands:

s2e help --commands

To get help on a particular command:

s2e <subcommand> --help

Most commands use the S2EDIR environment variable so that commands can be run from any directory. S2EDIR can be set by sourcing s2e_activate in your environment directory. Sourcing this file also makes s2e_deactivate available, which unsets the S2E environment variables.

Alternatively, most commands take an optional --env /path/to/env argument. This argument can be used to specify the path to the S2E environment you want to execute the command in.

Note that one of the S2EDIR environment variable or --env option must be used.

Workflow

Each command follows the Unix philosophy that each command ("tool") consists of a small program designed to accomplish a single, particular task, rather than trying to develop monolithic commands to do a number of tasks.

A typical workflow is therefore:

1. Run s2e init $DIR to create a new S2E environment in $DIR. This will create the environment, install dependencies (unless --skip-dependencies is used) and fetch all of the S2E engine code.
2. Activate the environment via . $DIR/install/bin/s2e_activate.
3. Look around the source code, make some modifications, etc. Then when you are ready to build run s2e build.
4. You'll need some images to analyze your software in! See what images are available with s2e image_build.
5. Run s2e image_build $TEMPLATE to build one of the images listed in the previous step. This will create the image in the images directory.
6. Use s2e new_project to create a new analysis project. This will create all the launch scripts, configuration files and bootstrap scripts necessary to perform the analysis on a given target. Currently Linux ELF executables, Decree CGC binaries, Windows PE executables and Windows DLLs can be targeted with the new_project command.
7. Change into the project directory and run the S2E analysis with the launch-s2e.sh script.
8. After your analysis has finished, a number of subcommands exist to analyze and summarize your results, e.g. the coverage and execution_trace subcommands.

Other useful commands:

• s2e info can be used to display a summary of the S2E environment.
• To download the latest changes from the git repositories, run s2e update.
• Projects can be shared using s2e export_project and s2e import_project.

Environment structure

s2e init generates the following directory structure in your S2E environment.

.
├── build/
├── images/
├── install/
├── projects/
├── s2e.yaml
├── source/

• build: Staging directory for builds
• images: Images created with s2e image_build go here
• install: Installed executables, libraries, header files, etc.
• projects: Analysis projects created with s2e new_project go here
• s2e.yaml: A per-environment configuration file. This file is also used to "mark" the directory as an S2E environment, so please do not delete it!
• source: Source code repositories

Extending

Extending with new commands is relatively simple. s2e-env is heavily influenced by Django's command subsystem, so there is a wealth of documentation already available (for example, here).

For example, to create a command foo:

1. Create a new Python module s2e_env/commands/foo.py

2. In foo.py define a Command class that extends

   • s2e_env.command.BaseCommand - The base class. Probably not that useful to inherit directly from this class
   • s2e_env.command.EnvCommand - For commands that operate on an existing S2E environment
   • s2e_env.command.ProjectCommand - For commands that operate on an existing analysis project

3. The only method required in your Command class is handle(self, *args, **options). This method contains your command logic

4. You may optionally define an add_arguments(self, parser) method for parsing command-line arguments specific to the foo command. The parser argument is essentially an ArgumentParser from the argparse library.

   If you extend from EnvCommand you must call the super add_arguments, i.e.:

   def add_arguments(self, parser):
       super(Command, self).add_arguments(parser)
       # Add your arguments/options here

5. On error, an s2e_env.command.CommandError should be raised

6. Use the logging module for printing messages. When calling logging.getLogger the command name should be provided as the logger name.

Running commands from your code

Like Django's command subsystem (see here), s2e-env also allows you to call commands programatically via the call_command function.

Example:

from s2e_env.commands.new_project import Command as NewProjectCommand
from s2e_env.manage import call_command


def create_s2e_project(target_path, s2e_env_path):
    call_command(NewProjectCommand(), target_path, env=s2e_env_path, force=True)

Custom projects

Occasionally the default analysis projects (e.g., Windows driver, Linux application, etc.) may not meet your requirements. In these cases, a custom project may be created by extending the s2e_env.commands.project_creation.abstract_project.AbstractProject class. This child class must implement the following methods:

• _configure: Generates a configuration dictionary that describes the project. The contents of this dictionary are up to the user; and
• _create: Creates the actual project on disk. This should including, making the project directory, and creating the files necessary to run the project in this project directory. The project creation is guided by the configuration dictionary generated in _make_config. The path to the project should be returned from this method.

Optionally, the child class may also implement:

• _get_instructions: Return a string that is displayed to the user upon successful creation of a project; and
• _is_valid_image: If an image is not specified, this method is used as a predicate when automatically selecting an image.

Currently, custom projects can only be used programmatically as follows:

import os

from s2e_env.commands.new_project import Command as NewProjectCommand
from s2e_env.commands.project_creation import AbstractProject
from s2e_env.manage import call_command


class MyProject(AbstractProject):
    def _configure(self, target, *args, **kwargs):
        return dict(project_dir='/path/to/my/project')

    def _create(self, config, force=False):
        os.mkdir(config['project_dir'])

        return config['project_dir']

    def _get_instructions(self, config):
        return 'Your project has been successfully created in %s' % config['project_dir']


call_command(NewProjectCommand(), env='/path/to/s2e', project_class=MyProject)