Skip to content

engineer-man/piston

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

engineer-man piston Piston

A high performance general purpose code execution engine.


GitHub last commit GitHub issues GitHub pull requests


About • Public API • Getting Started • Usage • Supported Languages • Principles • Security • License • Documentation



About

Piston is a high performance general purpose code execution engine. It excels at running untrusted and possibly malicious code without fear from any harmful effects.


It's used in numerous places including:


Official Extensions

The following are approved and endorsed extensions/utilities to the core Piston offering.

  • I Run Code, a Discord bot used in 4100+ servers to handle arbitrary code evaluation in Discord. To get this bot in your own server, go here: https://emkc.org/run.
  • Piston CLI, a universal shell supporting code highlighting, files, and interpretation without the need to download a language.
  • Node Piston Client, a Node.js wrapper for accessing the Piston API.
  • Piston4J, a Java wrapper for accessing the Piston API.
  • Pyston, a Python wrapper for accessing the Piston API.
  • Go-Piston, a Golang wrapper for accessing the Piston API.
  • piston_rs, a Rust wrapper for accessing the Piston API.
  • piston_rspy, Python bindings for accessing the Piston API via piston_rs.

Public API

  • Requires no installation and you can use it immediately.
  • Reference the Runtimes/Execute sections below to learn about the request and response formats.

When using the public Piston API, use the following two URLs:

GET  https://emkc.org/api/v2/piston/runtimes
POST https://emkc.org/api/v2/piston/execute

Important Note: The Piston API is rate limited to 5 requests per second. Effective May 7, 2024, no additional unlimited keys will be granted and existing keys will be revoked on Jan 1, 2025. The public instance is at capacity and the public limit is already very generous. For usage beyond 5 requests/second, you should consider self hosting.


Getting Started

All In One

Host System Package Dependencies

  • Docker
  • Docker Compose
  • Node JS (>= 15)
  • cgroup v2 enabled, and cgroup v1 disabled

After system dependencies are installed, clone this repository:

# clone and enter repo
git clone https://github.com/engineer-man/piston

Note

Ensure the repository is cloned with LF line endings

Installation

# Start the API container
docker-compose up -d api

# Install all the dependencies for the cli
cd cli && npm i && cd -

The API will now be online with no language runtimes installed. To install runtimes, use the CLI.

Just Piston (no CLI)

Host System Package Dependencies

  • Docker

Installation

docker run \
    --privileged \
    -v $PWD:'/piston' \
    -dit \
    -p 2000:2000 \
    --name piston_api \
    ghcr.io/engineer-man/piston

Piston for testing packages locally

Host System Package Dependencies

Installation

# Build the Docker containers
./piston start

# For more help
./piston help

Usage

CLI

The CLI is the main tool used for installing packages within piston, but also supports running code.

You can execute the cli with cli/index.js.

# List all available packages
cli/index.js ppman list

# Install latest python
cli/index.js ppman install python

# Install specific version of python
cli/index.js ppman install python=3.9.4

# Run a python script using the latest version
echo 'print("Hello world!")' > test.py
cli/index.js run python test.py

# Run a python script using a specific version
echo 'print("Hello world!")' > test.py
cli/index.js run python test.py -l 3.9.4
cli/index.js run python test.py -l 3.x
cli/index.js run python test.py -l 3

If you are operating on a remote machine, add the -u flag like so:

cli/index.js -u http://piston.server:2000 ppman list

API

The container exposes an API on port 2000 by default. This is used by the CLI to carry out running jobs and package management.

Runtimes Endpoint

GET /api/v2/runtimes This endpoint will return the supported languages along with the current version and aliases. To execute code for a particular language using the /api/v2/execute endpoint, either the name or one of the aliases must be provided, along with the version. Multiple versions of the same language may be present at the same time, and may be selected when running a job.

HTTP/1.1 200 OK
Content-Type: application/json

[
    {
        "language": "bash",
        "version": "5.1.0",
        "aliases": [
            "sh"
        ]
    },
    {
        "language": "brainfuck",
        "version": "2.7.3",
        "aliases": [
            "bf"
        ]
    },
    ...
]

Execute Endpoint

POST /api/v2/execute This endpoint requests execution of some arbitrary code.

  • language (required) The language to use for execution, must be a string and must be installed.
  • version (required) The version of the language to use for execution, must be a string containing a SemVer selector for the version or the specific version number to use.
  • files (required) An array of files containing code or other data that should be used for execution. The first file in this array is considered the main file.
  • files[].name (optional) The name of the file to upload, must be a string containing no path or left out.
  • files[].content (required) The content of the files to upload, must be a string containing text to write.
  • files[].encoding (optional) The encoding scheme used for the file content. One of base64, hex or utf8. Defaults to utf8.
  • stdin (optional) The text to pass as stdin to the program. Must be a string or left out. Defaults to blank string.
  • args (optional) The arguments to pass to the program. Must be an array or left out. Defaults to [].
  • compile_timeout (optional) The maximum wall-time allowed for the compile stage to finish before bailing out in milliseconds. Must be a number or left out. Defaults to 10000 (10 seconds).
  • run_timeout (optional) The maximum wall-time allowed for the run stage to finish before bailing out in milliseconds. Must be a number or left out. Defaults to 3000 (3 seconds).
  • compile_cpu_time (optional) The maximum CPU-time allowed for the compile stage to finish before bailing out in milliseconds. Must be a number or left out. Defaults to 10000 (10 seconds).
  • run_cpu_time (optional) The maximum CPU-time allowed for the run stage to finish before bailing out in milliseconds. Must be a number or left out. Defaults to 3000 (3 seconds).
  • compile_memory_limit (optional) The maximum amount of memory the compile stage is allowed to use in bytes. Must be a number or left out. Defaults to -1 (no limit)
  • run_memory_limit (optional) The maximum amount of memory the run stage is allowed to use in bytes. Must be a number or left out. Defaults to -1 (no limit)
{
    "language": "js",
    "version": "15.10.0",
    "files": [
        {
            "name": "my_cool_code.js",
            "content": "console.log(process.argv)"
        }
    ],
    "stdin": "",
    "args": ["1", "2", "3"],
    "compile_timeout": 10000,
    "run_timeout": 3000,
    "compile_cpu_time": 10000,
    "run_cpu_time": 3000,
    "compile_memory_limit": -1,
    "run_memory_limit": -1
}

A typical response upon successful execution will contain 1 or 2 keys run and compile. compile will only be present if the language requested requires a compile stage.

Each of these keys has an identical structure, containing both a stdout and stderr key, which is a string containing the text outputted during the stage into each buffer. It also contains the code and signal which was returned from each process. It also includes a nullable human-readable message which is a description of why a stage has failed and a two-letter status that is either:

  • RE for runtime error
  • SG for dying on a signal
  • TO for timeout (either via timeout or cpu_time)
  • OL for stdout length exceeded
  • EL for stderr length exceeded
  • XX for internal error
HTTP/1.1 200 OK
Content-Type: application/json

{
    "language": "js",
    "version": "15.10.0",
    "run": {
        "stdout": "[\n  '/piston/packages/node/15.10.0/bin/node',\n  '/piston/jobs/9501b09d-0105-496b-b61a-e5148cf66384/my_cool_code.js',\n  '1',\n  '2',\n  '3'\n]\n",
        "stderr": "",
        "output": "[\n  '/piston/packages/node/15.10.0/bin/node',\n  '/piston/jobs/9501b09d-0105-496b-b61a-e5148cf66384/my_cool_code.js',\n  '1',\n  '2',\n  '3'\n]\n",
        "code": 0,
        "signal": null,
        "message": null,
        "status": null,
        "cpu_time": 8,
        "wall_time": 154,
        "memory": 1160000
    }
}

If a problem exists with the request, a 400 status code is returned and the reason in the message key.

HTTP/1.1 400 Bad Request
Content-Type: application/json

{
    "message": "html-5.0.0 runtime is unknown"
}

Interactive execution endpoint (not available through the public API)

To interact with running processes in real time, you can establish a WebSocket connection at /api/v2/connect. This allows you to both receive output and send input to active processes.

Each message is structured as a JSON object with a type key, which indicates the action to perform. Below is a list of message types, their directions, and descriptions:

  • init (client -> server): Initializes a job with the same parameters as the /execute endpoint, except that stdin is discarded.
  • runtime (server -> client): Provides details on the runtime environment, including the version and language.
  • stage (server -> client): Indicates the current execution stage, either "compile" or "run."
  • data (server <-> client): Exchanges data between the client and server, such as stdin, stdout, or stderr streams.
  • signal (client -> server): Sends a signal (e.g., for termination) to the running process, whether it's in the "compile" or "run" stage.
  • exit (server -> client): Signals the end of a stage, along with the exit code or signal.
  • error (server -> client): Reports an error, typically right before the WebSocket is closed.

An example of this endpoint in use is depicted below (< = client to server, > = server to client)

  1. Client establishes WebSocket connection to /api/v2/connect
  2. < {"type":"init", "language":"bash", "version":"*", "files":[{"content": "cat"}]}
  3. > {"type":"runtime","language": "bash", "version": "5.1.0"}
  4. > {"type":"stage", "stage":"run"}
  5. < {"type":"data", "stream":"stdin", "data":"Hello World!"}
  6. > {"type":"data", "stream":"stdout", "data":"Hello World!"}
  7. time passes
  8. > {"type":"exit", "stage":"run", "code":null, "signal": "SIGKILL"}

Errors may return status codes as follows:

  • 4000: Already Initialized: Sent when a second init command is issued.
  • 4001: Initialization Timeout: No init command was sent within 1 second of connection.
  • 4002: Notified Error: A fatal error occurred, and an error packet was transmitted.
  • 4003: Not yet Initialized: A non-init command was sent without a job context.
  • 4004: Can only write to stdin: The client attempted to write to a stream other than stdin.
  • 4005: Invalid Signal: An invalid signal was sent in a signal packet.

Supported Languages

awk, bash, befunge93, brachylog, brainfuck, bqn, c, c++, cjam, clojure, cobol, coffeescript, cow, crystal, csharp, csharp.net, d, dart, dash, dragon, elixir, emacs, emojicode, erlang, file, forte, forth, fortran, freebasic, fsharp.net, fsi, go, golfscript, groovy, haskell, husk, iverilog, japt, java, javascript, jelly, julia, kotlin, lisp, llvm_ir, lolcode, lua, matl, nasm, nasm64, nim, ocaml, octave, osabie, paradoc, pascal, perl, php, ponylang, powershell, prolog, pure, pyth, python, python2, racket, raku, retina, rockstar, rscript, ruby, rust, samarium, scala, smalltalk, sqlite3, swift, typescript, basic, basic.net, vlang, vyxal, yeethon, zig,


Principle of Operation

Piston uses Isolate inside Docker as the primary mechanism for sandboxing. There is an API within the container written in Node which takes in execution requests and executes them within the container safely. High level, the API writes any source code and executes it inside an Isolate sandbox. The source file is either ran or compiled and ran (in the case of languages like c, c++, c#, go, etc.).


Security

Piston uses Isolate which makes use of Linux namespaces, chroot, multiple unprivileged users, and cgroup for sandboxing and resource limiting. Code execution submissions on Piston shall not be aware of each other, shall not affect each other and shall not affect the underlying host system. This is ensured through multiple steps including:

  • Disabling outgoing network interaction by default
  • Capping max processes at 256 by default (resists :(){ :|: &}:;, while True: os.fork(), etc.)
  • Capping max files at 2048 (resists various file based attacks)
  • Cleaning up all temp space after each execution (resists out of drive space attacks)
  • Running each submission as a different unprivileged user
  • Running each submission with its own isolated Linux namespaces
  • Capping runtime execution at 3 seconds by default (CPU-time and wall-time)
  • Capping the peak memory that all the submission's processes can use
  • Capping stdout to 1024 characters by default (resists yes/no bombs and runaway output)
  • SIGKILLing misbehaving code

License

Piston is licensed under the MIT license.