Lib.rs

Onyums is a simple axum wrapper for serving tor onion services. It provides the ConnectionInfo extractor to differentiate connections made over tor since the concept of SocketAddrs does not exist on private connections.

NEW! - Onyums now supports websockets over Tor!

Hello world example

use axum::{routing::get, Router};
use native_tls::Identity;
use tokio_native_tls::TlsAcceptor;
use onyums::serve;

#[tokio::main]
async fn main() {
        // standard axum router
        let app = Router::new().route("/", get(|| async { "Hello, World!" }));

        // create tls identity & acceptor
        let c = include_bytes!("../self_signed_certs/cert.pem");
        let k = include_bytes!("../self_signed_certs/key.pem");
        let identity = Identity::from_pem(c, k).unwrap();
        let tls_acceptor = TlsAcceptor::from(native_tls::TlsAcceptor::builder(identity).build().unwrap());

        // start the serve
        // pass in the router, tls acceptor, and the server nickname (used to generate an onion address).
        serve(app, tls_acceptor, "my_onion").await.unwrap();
}

Websocket example

use axum::{
    extract::ws::{Message, WebSocket, WebSocketUpgrade},
    response::IntoResponse,
    routing::get,
    Router,
};
use axum_extra::TypedHeader;

use std::borrow::Cow;
use std::ops::ControlFlow;
use native_tls::Identity;
use tokio_native_tls::TlsAcceptor;
use onyums::{serve, ConnectionInfo};

#[tokio::main]
async fn main() {
        // build our application with some routes
        let app = Router::new().route("/ws", get(ws_handler));

        // create tls identity & acceptor
        let c = include_bytes!("../self_signed_certs/cert.pem");
        let k = include_bytes!("../self_signed_certs/key.pem");
        let identity = Identity::from_pem(c, k).unwrap();
        let tls_acceptor = TlsAcceptor::from(native_tls::TlsAcceptor::builder(identity).build().unwrap());

        // start the serve
        // pass in the router, tls acceptor, and the server nickname (used to generate an onion address).
        serve(app, tls_acceptor, "my_onion").await.unwrap();
}

/// The handler for the HTTP request (this gets called when the HTTP GET lands at the start
/// of websocket negotiation). After this completes, the actual switching from HTTP to
/// websocket protocol will occur.
/// This is the last point where we can extract TCP/IP metadata such as IP address of the client
/// as well as things from HTTP headers such as user-agent of the browser etc.
async fn ws_handler(ws: WebSocketUpgrade, user_agent: Option<TypedHeader<headers::UserAgent>>, ConnectInfo(connection_info): ConnectInfo<ConnectionInfo>) -> impl IntoResponse {
        let default_addr = SocketAddr::new(IpAddr::V4(Ipv4Addr::new(0, 0, 0, 0)), 0);
	let parsed_addr = connection_info.socket_addr.unwrap_or(default_addr).to_string();
	let circuit_id = connection_info.circuit_id.unwrap_or_default();

	let addr = if parsed_addr == default_addr.to_string() { circuit_id } else { parsed_addr };

        let user_agent = if let Some(TypedHeader(user_agent)) = user_agent { user_agent.to_string() } else { String::from("Unknown browser") };
        println!("`{user_agent}` at {addr} connected.");
        // finalize the upgrade process by returning upgrade callback.
        // we can customize the callback by sending additional info such as address.
        ws.on_upgrade(move |socket| handle_socket(socket, addr))
}

/// Actual websocket statemachine (one will be spawned per connection)
async fn handle_socket(mut socket: WebSocket, who: String) {
        // send a ping (unsupported by some browsers) just to kick things off and get a response
        if socket.send(Message::Ping(vec![1, 2, 3])).await.is_ok() {
                println!("Pinged {who}...");
        } else {
                println!("Could not send ping {who}!");
                // no Error here since the only thing we can do is to close the connection.
                // If we can not send messages, there is no way to salvage the statemachine anyway.
                return;
        }

        // receive single message from a client (we can either receive or send with socket).
        // this will likely be the Pong for our Ping or a hello message from client.
        // waiting for message from a client will block this task, but will not block other client's
        // connections.
        if let Some(msg) = socket.recv().await {
                if let Ok(msg) = msg {
                        if process_message(msg, who).is_break() {
                                return;
                        }
                } else {
                        println!("client {who} abruptly disconnected");
                        return;
                }
        }

        // Since each client gets individual statemachine, we can pause handling
        // when necessary to wait for some external event (in this case illustrated by sleeping).
        // Waiting for this client to finish getting its greetings does not prevent other clients from
        // connecting to server and receiving their greetings.
        for i in 1..5 {
                if socket.send(Message::Text(format!("Hi {i} times!"))).await.is_err() {
                        println!("client {who} abruptly disconnected");
                        return;
                }
                tokio::time::sleep(std::time::Duration::from_millis(100)).await;
        }

        // By splitting socket we can send and receive at the same time. In this example we will send
        // unsolicited messages to client based on some sort of server's internal event (i.e .timer).
        let (mut sender, mut receiver) = socket.split();

        // Spawn a task that will push several messages to the client (does not matter what client does)
        let mut send_task = tokio::spawn(async move {
                let n_msg = 20;
                for i in 0..n_msg {
                        // In case of any websocket error, we exit.
                        if sender.send(Message::Text(format!("Server message {i} ..."))).await.is_err() {
                                return i;
                        }

                        tokio::time::sleep(std::time::Duration::from_millis(300)).await;
                }

                println!("Sending close to {who}...");
                if let Err(e) = sender.send(Message::Close(Some(CloseFrame {code: axum::extract::ws::close_code::NORMAL, reason: Cow::from("Goodbye"),}))).await {
                        println!("Could not send Close due to {e}, probably it is ok?");
                }
                n_msg
        });

        // This second task will receive messages from client and print them on server console
        let mut recv_task = tokio::spawn(async move {
                let mut cnt = 0;
                while let Some(Ok(msg)) = receiver.next().await {
                        cnt += 1;
                        // print message and break if instructed to do so
                        if process_message(msg, who).is_break() {
                                break;
                        }
                }
                cnt
        });

        // If any one of the tasks exit, abort the other.
        tokio::select! {
                rv_a = (&mut send_task) => {
                        match rv_a {
                                Ok(a) => println!("{a} messages sent to {who}"),
                                Err(a) => println!("Error sending messages {a:?}")
                        }
                        recv_task.abort();
                },
                rv_b = (&mut recv_task) => {
                        match rv_b {
                                Ok(b) => println!("Received {b} messages"),
                                Err(b) => println!("Error receiving messages {b:?}")
                        }
                        send_task.abort();
                }
        }

        // returning from the handler closes the websocket connection
        println!("Websocket context {who} destroyed");
}

/// helper to print contents of messages to stdout. Has special treatment for Close.
fn process_message(msg: Message, who: String) -> ControlFlow<(), ()> {
        match msg {
                Message::Text(t) => {
                        println!(">>> {who} sent str: {t:?}");
                }
                Message::Binary(d) => {
                        println!(">>> {} sent {} bytes: {:?}", who, d.len(), d);
                }
                Message::Close(c) => {
                        if let Some(cf) = c {
                                println!(">>> {} sent close with code {} and reason `{}`", who, cf.code, cf.reason);
                        } else {
                                println!(">>> {who} somehow sent close message without CloseFrame");
                        }
                        return ControlFlow::Break(());
                }
                Message::Pong(v) => {
                        println!(">>> {who} sent pong with {v:?}");
                }

                // You should never need to manually handle Message::Ping, as axum's websocket library
                // will do so for you automagically by replying with Pong and copying the v according to
                // spec. But if you need the contents of the pings you can see them here.
                Message::Ping(v) => {
                        println!(">>> {who} sent ping with {v:?}");
                }
        }
        ControlFlow::Continue(())
}