Open up terminal/shell on remote server via tcp request - node.js

Question

I have this:
const http = require('http');
const cp = require('child_process');
const server = http.createServer((req,res) => {
const bash = cp.spawn('bash');
req.pipe(bash.stdin, {end:false);
bash.stdout.pipe(res);
bash.stderr.pipe(res);
});
server.listen('4004');
when I hit the server with:
curl localhost:4004
and I type bash commands, nothing gets outputed to my console, anybody know why?
Note: To address security I plan to run this in a docker container, use https/ssl, and implement authentication (any recommendations on auth schemes lmk).
More importantly, I am looking for shell prompts to appear ... apparently bash by itself doesn't open up a shell/prompt?

Answer 1

It is possible to do this "over the web" so to speak. However, your approach will not work, because you are mixing paradigms (batch vs. interactive), and you are missing large chunks of setup that's needed to run terminal applications.
Normally I would show you how to program this, however, that's really involved. Have a look at:
https://github.com/chjj/tty.js
and,
https://github.com/xtermjs/xterm.js
as starting points to create your solution.
Both are usable directly from node.js to serve up terminal applications over HTTP.

Answer 2

This is a partial answer, but I started a bounty because I am looking for something better. I was able to create something rudimentary with TCP like so:
const net = require('net'); // !use net package not http
const cp = require('child_process');
const server = net.createServer(s => {
const bash = cp.spawn('bash');
s.pipe(bash.stdin, {end:false});
bash.stdout.pipe(s);
bash.stderr.pipe(s);
});
server.listen('4004');
not sure why it won't work with HTTP though. I connect to it using netcat:
nc localhost 4004
but this isn't opening a terminal, just a bash process. the experience is not ideal, as described here:
https://unix.stackexchange.com/questions/519364/bash-shell-modes-how-to-pipe-request-to-shell-on-remote-server
however I am looking to replicate the shell experience you have when you do something like:
docker exec -ti <container> /bin/bash
when I run my script it "works", but I don't get any shell prompts or anything like that. (One way to solve this might be with ssh, but I am trying to figure out a different way).

Answer 3

You can connect to an http server with telnet. It depends on how you're starting the http server. Here's an example
Start an http server with the npm package http-server
npm install -g http-server
cd ~/ <Any directory>
http-server
Now seperately start a telnet session
telnet localhost 8080
OR
nc localhost 8080
And then type something like GET /
Use the telnet client instead of nc
Check this: https://www.the-art-of-web.com/system/telnet-http11/
Update: Running an ssh server over nodejs. It allows you to run an ssh server
I found this at https://github.com/mscdex/ssh2
var fs = require('fs');
var crypto = require('crypto');
var inspect = require('util').inspect;
var ssh2 = require('ssh2');
var utils = ssh2.utils;
var allowedUser = Buffer.from('foo');
var allowedPassword = Buffer.from('bar');
var allowedPubKey = utils.parseKey(fs.readFileSync('foo.pub'));
new ssh2.Server({
hostKeys: [fs.readFileSync('host.key')]
}, function(client) {
console.log('Client connected!');
client.on('authentication', function(ctx) {
var user = Buffer.from(ctx.username);
if (user.length !== allowedUser.length
|| !crypto.timingSafeEqual(user, allowedUser)) {
return ctx.reject();
}
switch (ctx.method) {
case 'password':
var password = Buffer.from(ctx.password);
if (password.length !== allowedPassword.length
|| !crypto.timingSafeEqual(password, allowedPassword)) {
return ctx.reject();
}
break;
case 'publickey':
var allowedPubSSHKey = allowedPubKey.getPublicSSH();
if (ctx.key.algo !== allowedPubKey.type
|| ctx.key.data.length !== allowedPubSSHKey.length
|| !crypto.timingSafeEqual(ctx.key.data, allowedPubSSHKey)
|| (ctx.signature && !allowedPubKey.verify(ctx.blob, ctx.signature))) {
return ctx.reject();
}
break;
default:
return ctx.reject();
}
ctx.accept();
}).on('ready', function() {
console.log('Client authenticated!');
client.on('session', function(accept, reject) {
var session = accept();
session.once('exec', function(accept, reject, info) {
console.log('Client wants to execute: ' + inspect(info.command));
var stream = accept();
stream.stderr.write('Oh no, the dreaded errors!\n');
stream.write('Just kidding about the errors!\n');
stream.exit(0);
stream.end();
});
});
}).on('end', function() {
console.log('Client disconnected');
});
}).listen(0, '127.0.0.1', function() {
console.log('Listening on port ' + this.address().port);
});

Answer 4

Your approaches are quite mixed, nonetheless, when ever you finally connect to the remote server do not use 'bash' as a method to start the connection, BASH is just born again shell with other commands & stuff in it,
Rather use some of the following program, command-line names: i.e :
~ $ 'gnome-terminal'
~ $ 'xterm'
there you will now be referencing a true program in the system, even kernel level C code has its own recognition of these, if not changed.