Project Conncept is an experimental layer 4 app for Caddy. It facilitates composable handling of raw TCP/UDP connections based on properties of the connection or the beginning of the stream.
With it, you can listen on sockets/ports and express logic such as:
- "Echo all input back to the client."
- "Proxy all the raw bytes to 10.0.3.14:1592."
- "If connection is TLS, terminate TLS then proxy all bytes to :5000."
- "Terminate TLS; then if it is HTTP, proxy to localhost:80; otherwise echo."
- "If connection is TLS, proxy to :443 without terminating; if HTTP, proxy to :80; if SSH, proxy to :22."
- "If the HTTP Host is
example.com
or the TLS ServerName isexample.com
, then proxy to 192.168.0.4." - "Block connections from these IP ranges: ..."
- "Throttle data flow to simulate slow connections."
- And much more!
Because this is a caddy app, it can be used alongside other Caddy apps such as the HTTP server or TLS certificate manager.
Note that only JSON config is available at this time. More documentation will come soon. For now, please read the code, especially type definitions and their comments. It's actually a pretty simple code base, and the JSON config isn't that bad once you get used to it! See below for tips and examples writing config.
Note
This is not an official repository of the Caddy Web Server organization.
This app works similarly to the http
app. You define servers, and each server consists of routes. A route has a set of matchers and handlers; if a connection matches, the assoicated handlers are invoked.
Current matchers:
- layer4.matchers.http - matches connections that start with HTTP requests. In addition, any
http.matchers
modules can be used for matching on HTTP-specific properties of requests, such as header or path. Note that only the first request of each connection can be used for matching. - layer4.matchers.tls - matches connections that start with TLS handshakes. In addition, any
tls.handshake_match
modules can be used for matching on TLS-specific properties of the ClientHello, such as ServerName (SNI). - layer4.matchers.ssh - matches connections that look like SSH connections.
- layer4.matchers.postgres - matches connections that look like Postgres connections.
- layer4.matchers.ip - matches connections based on remote IP (or CIDR range).
- layer4.matchers.local_ip - matches connections based on local IP (or CIDR range).
- layer4.matchers.proxy_protocol - matches connections that start with HAPROXY proxy protocol.
- layer4.matchers.socks4 - matches connections that look like SOCKSv4.
- layer4.matchers.socks5 - matches connections that look like SOCKSv5.
Current handlers:
- layer4.handlers.echo - An echo server.
- layer4.handlers.proxy - Powerful layer 4 proxy, capable of multiple upstreams (with load balancing and health checks) and establishing new TLS connections to backends. Optionally supports sending the HAProxy proxy protocol.
- layer4.handlers.tee - Branches the handling of a connection into a concurrent handler chain.
- layer4.handlers.throttle - Throttle connections to simulate slowness and latency.
- layer4.handlers.tls - TLS termination.
- layer4.handlers.proxy_protocol - Accepts the HAPROXY proxy protocol on the receiving side.
- layer4.handlers.socks5 - Handles SOCKSv5 proxy protocol connections.
Like the http
app, some handlers are "terminal" meaning that they don't call the next handler in the chain. For example: echo
and proxy
are terminal handlers because they consume the client's input.
The recommended way is to use xcaddy:
$ xcaddy build --with github.com/mholt/caddy-l4
Alternatively, to hack on the plugin code, you can clone it down, then build and run like so:
- Download or clone this repo:
git clone https://github.com/mholt/caddy-l4.git
- In the project folder, run
xcaddy
just like you would runcaddy
. For example:xcaddy list-modules --versions
(you should see thelayer4
modules).
Since this app does not support Caddyfile (yet?), you will have to use Caddy's native JSON format to configure it. I highly recommend this caddy-json-schema plugin by @abiosoft which can give you auto-complete and documentation right in your editor as you write your config!
See below for some examples to help you get started.
A simple echo server:
{
"apps": {
"layer4": {
"servers": {
"example": {
"listen": ["127.0.0.1:5000"],
"routes": [
{
"handle": [
{"handler": "echo"}
]
}
]
}
}
}
}
}
A simple echo server with TLS termination that uses a self-signed cert for localhost
:
{
"apps": {
"layer4": {
"servers": {
"example": {
"listen": ["127.0.0.1:5000"],
"routes": [
{
"handle": [
{"handler": "tls"},
{"handler": "echo"}
]
}
]
}
}
},
"tls": {
"certificates": {
"automate": ["localhost"]
},
"automation": {
"policies": [
{
"issuers": [{"module": "internal"}]
}
]
}
}
}
}
A simple TCP reverse proxy that terminates TLS on 993, and sends the PROXY protocol header to 1143 through 143:
{
"apps": {
"layer4": {
"servers": {
"secure-imap": {
"listen": ["0.0.0.0:993"],
"routes": [
{
"handle": [
{
"handler": "tls"
},
{
"handler": "proxy",
"proxy_protocol": "v1",
"upstreams": [
{"dial": ["localhost:143"]}
]
}
]
}
]
},
"normal-imap": {
"listen": ["0.0.0.0:143"],
"routes": [
{
"handle": [
{
"handler": "proxy_protocol"
},
{
"handler": "proxy",
"proxy_protocol": "v2",
"upstreams": [
{"dial": ["localhost:1143"]}
]
}
]
}
]
}
}
}
}
}
A multiplexer that proxies HTTP to one backend, and TLS to another (without terminating TLS):
{
"apps": {
"layer4": {
"servers": {
"example": {
"listen": ["127.0.0.1:5000"],
"routes": [
{
"match": [
{
"http": []
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["localhost:80"]}
]
}
]
},
{
"match": [
{
"tls": {}
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["localhost:443"]}
]
}
]
}
]
}
}
}
}
}
Same as previous, but only applies to HTTP requests with specific hosts:
{
"apps": {
"layer4": {
"servers": {
"example": {
"listen": ["127.0.0.1:5000"],
"routes": [
{
"match": [
{
"http": [
{"host": ["example.com"]}
]
}
],
"handle": [
{
"handler": "subroute",
"routes": [
{
"match": [
{
"http": []
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["localhost:80"]}
]
}
]
},
{
"match": [
{
"tls": {}
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["localhost:443"]}
]
}
]
}
]
}
]
}
]
}
}
}
}
}
Same as previous, but filter by HTTP Host header and/or TLS ClientHello ServerName:
{
"apps": {
"layer4": {
"servers": {
"example": {
"listen": ["127.0.0.1:5000"],
"routes": [
{
"match": [
{
"http": [
{"host": ["example.com"]}
]
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["localhost:80"]}
]
}
]
},
{
"match": [
{
"tls": {
"sni": ["example.net"]
}
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["localhost:443"]}
]
}
]
}
]
}
}
}
}
}
Forwarding SOCKSv4 to a remote server and handling SOCKSv5 directly in caddy.
While only allowing connections from a specific network and requiring a username and password for SOCKSv5.
{
"apps": {
"layer4": {
"servers": {
"socks": {
"listen": ["0.0.0.0:1080"],
"routes": [
{
"match": [
{
"socks5": {},
"ip": {"ranges": ["10.0.0.0/24"]}
}
],
"handle": [
{
"handler": "socks5",
"credentials": {
"bob": "qHoEtVpGRM"
}
}
]
},
{
"match": [
{
"socks4": {}
}
],
"handle": [
{
"handler": "proxy",
"upstreams": [
{"dial": ["10.64.0.1:1080"]}
]
}
]
}
]
}
}
}
}
}