From f0bbad2db95b00ec429f498fe15f1007bd87da5e Mon Sep 17 00:00:00 2001 From: Alex Auvolat Date: Mon, 19 Feb 2024 10:58:54 +0100 Subject: [PATCH 1/3] [networking-fixes] use rpc_public_addr in netapp's HelloMessage --- src/net/netapp.rs | 19 ++++++++++++------- src/rpc/system.rs | 10 +++++----- 2 files changed, 17 insertions(+), 12 deletions(-) diff --git a/src/net/netapp.rs b/src/net/netapp.rs index b1ad9db8..e572dc11 100644 --- a/src/net/netapp.rs +++ b/src/net/netapp.rs @@ -38,6 +38,11 @@ pub(crate) type VersionTag = [u8; 16]; /// Value of the Netapp version used in the version tag pub(crate) const NETAPP_VERSION_TAG: u64 = 0x6e65746170700005; // netapp 0x0005 +/// HelloMessage is sent by the client on a Netapp connection to indicate +/// that they are also a server and ready to recieve incoming connections +/// at the specified address and port. If the client doesn't know their +/// public address, they don't need to specify it and we look at the +/// remote address of the socket is used instead. #[derive(Serialize, Deserialize, Debug)] pub(crate) struct HelloMessage { pub server_addr: Option, @@ -56,9 +61,6 @@ type OnDisconnectHandler = Box; /// NetApp can be used in a stand-alone fashion or together with a peering strategy. /// If using it alone, you will want to set `on_connect` and `on_disconnect` events /// in order to manage information about the current peer list. -/// -/// It is generally not necessary to use NetApp stand-alone, as the provided full mesh -/// and RPS peering strategies take care of the most common use cases. pub struct NetApp { listen_params: ArcSwapOption, @@ -83,7 +85,7 @@ pub struct NetApp { struct ListenParams { listen_addr: SocketAddr, - public_addr: Option, + public_addr: Option, } impl NetApp { @@ -180,7 +182,7 @@ impl NetApp { pub async fn listen( self: Arc, listen_addr: SocketAddr, - public_addr: Option, + public_addr: Option, mut must_exit: watch::Receiver, ) { let listen_params = ListenParams { @@ -396,8 +398,11 @@ impl NetApp { } if let Some(lp) = self.listen_params.load_full() { - let server_addr = lp.public_addr; - let server_port = lp.listen_addr.port(); + let server_addr = lp.public_addr.map(|x| x.ip()); + let server_port = lp + .public_addr + .map(|x| x.port()) + .unwrap_or(lp.listen_addr.port()); let hello_endpoint = self.hello_endpoint.load_full().unwrap(); tokio::spawn(async move { hello_endpoint diff --git a/src/rpc/system.rs b/src/rpc/system.rs index de44e656..5c6e07b1 100644 --- a/src/rpc/system.rs +++ b/src/rpc/system.rs @@ -98,7 +98,6 @@ pub struct System { system_endpoint: Arc>, rpc_listen_addr: SocketAddr, - #[cfg(any(feature = "consul-discovery", feature = "kubernetes-discovery"))] rpc_public_addr: Option, bootstrap_peers: Vec, @@ -369,7 +368,6 @@ impl System { replication_mode, replication_factor, rpc_listen_addr: config.rpc_bind_addr, - #[cfg(any(feature = "consul-discovery", feature = "kubernetes-discovery"))] rpc_public_addr, bootstrap_peers: config.bootstrap_peers.clone(), #[cfg(feature = "consul-discovery")] @@ -390,9 +388,11 @@ impl System { /// Perform bootstraping, starting the ping loop pub async fn run(self: Arc, must_exit: watch::Receiver) { join!( - self.netapp - .clone() - .listen(self.rpc_listen_addr, None, must_exit.clone()), + self.netapp.clone().listen( + self.rpc_listen_addr, + self.rpc_public_addr, + must_exit.clone() + ), self.peering.clone().run(must_exit.clone()), self.discovery_loop(must_exit.clone()), self.status_exchange_loop(must_exit.clone()), From b96f84b894684ed43e281a3aa2f391b424414a84 Mon Sep 17 00:00:00 2001 From: Alex Auvolat Date: Mon, 19 Feb 2024 11:24:33 +0100 Subject: [PATCH 2/3] [networking-fixes] add option to bind outgoing RPC sockets (fix #638) Thanks to yuka for the original patch. --- src/garage/main.rs | 2 +- src/net/netapp.rs | 26 ++++++++++++++++++++++---- src/net/test.rs | 2 +- src/rpc/system.rs | 5 ++++- src/util/config.rs | 3 +++ 5 files changed, 31 insertions(+), 7 deletions(-) diff --git a/src/garage/main.rs b/src/garage/main.rs index 4d8dcc67..e489fff0 100644 --- a/src/garage/main.rs +++ b/src/garage/main.rs @@ -203,7 +203,7 @@ async fn cli_command(opt: Opt) -> Result<(), Error> { // Generate a temporary keypair for our RPC client let (_pk, sk) = sodiumoxide::crypto::sign::ed25519::gen_keypair(); - let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, sk); + let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, sk, None); // Find and parse the address of the target host let (id, addr, is_default_addr) = if let Some(h) = opt.rpc_host { diff --git a/src/net/netapp.rs b/src/net/netapp.rs index e572dc11..faa51a99 100644 --- a/src/net/netapp.rs +++ b/src/net/netapp.rs @@ -13,7 +13,7 @@ use sodiumoxide::crypto::sign::ed25519; use futures::stream::futures_unordered::FuturesUnordered; use futures::stream::StreamExt; -use tokio::net::{TcpListener, TcpStream}; +use tokio::net::{TcpListener, TcpSocket, TcpStream}; use tokio::select; use tokio::sync::{mpsc, watch}; @@ -62,6 +62,7 @@ type OnDisconnectHandler = Box; /// If using it alone, you will want to set `on_connect` and `on_disconnect` events /// in order to manage information about the current peer list. pub struct NetApp { + bind_outgoing_to: Option, listen_params: ArcSwapOption, /// Version tag, 8 bytes for netapp version, 8 bytes for app version @@ -94,13 +95,19 @@ impl NetApp { /// using `.listen()` /// /// Our Peer ID is the public key associated to the secret key given here. - pub fn new(app_version_tag: u64, netid: auth::Key, privkey: ed25519::SecretKey) -> Arc { + pub fn new( + app_version_tag: u64, + netid: auth::Key, + privkey: ed25519::SecretKey, + bind_outgoing_to: Option, + ) -> Arc { let mut version_tag = [0u8; 16]; version_tag[0..8].copy_from_slice(&u64::to_be_bytes(NETAPP_VERSION_TAG)[..]); version_tag[8..16].copy_from_slice(&u64::to_be_bytes(app_version_tag)[..]); let id = privkey.public_key(); let netapp = Arc::new(Self { + bind_outgoing_to, listen_params: ArcSwapOption::new(None), version_tag, netid, @@ -300,9 +307,20 @@ impl NetApp { return Ok(()); } - let socket = TcpStream::connect(ip).await?; + let stream = match self.bind_outgoing_to { + Some(addr) => { + let socket = if addr.is_ipv4() { + TcpSocket::new_v4()? + } else { + TcpSocket::new_v6()? + }; + socket.bind(SocketAddr::new(addr, 0))?; + socket.connect(ip).await? + } + None => TcpStream::connect(ip).await?, + }; info!("Connected to {}, negotiating handshake...", ip); - ClientConn::init(self, socket, id).await?; + ClientConn::init(self, stream, id).await?; Ok(()) } diff --git a/src/net/test.rs b/src/net/test.rs index c6259752..5a3f236d 100644 --- a/src/net/test.rs +++ b/src/net/test.rs @@ -102,7 +102,7 @@ fn run_netapp( Arc, Arc, ) { - let netapp = NetApp::new(0u64, netid, sk); + let netapp = NetApp::new(0u64, netid, sk, None); let peering = PeeringManager::new(netapp.clone(), bootstrap_peers, None); let peering2 = peering.clone(); diff --git a/src/rpc/system.rs b/src/rpc/system.rs index 5c6e07b1..147ec4d6 100644 --- a/src/rpc/system.rs +++ b/src/rpc/system.rs @@ -324,7 +324,10 @@ impl System { warn!("This Garage node does not know its publicly reachable RPC address, this might hamper intra-cluster communication."); } - let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, node_key); + let bind_outgoing_to = Some(config) + .filter(|x| x.rpc_bind_outgoing) + .map(|x| x.rpc_bind_addr.ip()); + let netapp = NetApp::new(GARAGE_VERSION_TAG, network_key, node_key, bind_outgoing_to); let peering = PeeringManager::new(netapp.clone(), vec![], rpc_public_addr); if let Some(ping_timeout) = config.rpc_ping_timeout_msec { peering.set_ping_timeout_millis(ping_timeout); diff --git a/src/util/config.rs b/src/util/config.rs index a9a72110..056c625d 100644 --- a/src/util/config.rs +++ b/src/util/config.rs @@ -55,6 +55,9 @@ pub struct Config { pub rpc_secret_file: Option, /// Address to bind for RPC pub rpc_bind_addr: SocketAddr, + /// Bind outgoing sockets to rpc_bind_addr's IP address as well + #[serde(default)] + pub rpc_bind_outgoing: bool, /// Public IP address of this node pub rpc_public_addr: Option, From eab54b37981073651712a15334b60b2c409f2576 Mon Sep 17 00:00:00 2001 From: Alex Auvolat Date: Mon, 19 Feb 2024 11:31:09 +0100 Subject: [PATCH 3/3] [networking-fixes] add doc for rpc_bind_outgoing --- doc/book/reference-manual/configuration.md | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/doc/book/reference-manual/configuration.md b/doc/book/reference-manual/configuration.md index 27d9fd54..f1474613 100644 --- a/doc/book/reference-manual/configuration.md +++ b/doc/book/reference-manual/configuration.md @@ -27,6 +27,7 @@ compression_level = 1 rpc_secret = "4425f5c26c5e11581d3223904324dcb5b5d5dfb14e5e7f35e38c595424f5f1e6" rpc_bind_addr = "[::]:3901" +rpc_bind_outgoing = false rpc_public_addr = "[fc00:1::1]:3901" bootstrap_peers = [ @@ -91,6 +92,7 @@ Top-level configuration options: [`metadata_fsync`](#metadata_fsync), [`replication_mode`](#replication_mode), [`rpc_bind_addr`](#rpc_bind_addr), +[`rpc_bind_outgoing`](#rpc_bind_outgoing), [`rpc_public_addr`](#rpc_public_addr), [`rpc_secret`/`rpc_secret_file`](#rpc_secret), [`sled_cache_capacity`](#sled_cache_capacity), @@ -415,6 +417,17 @@ the node, even in the case of a NAT: the NAT should be configured to forward the port number to the same internal port nubmer. This means that if you have several nodes running behind a NAT, they should each use a different RPC port number. +#### `rpc_bind_outgoing` {#rpc_bind_outgoing} (since v0.9.2) + +If enabled, pre-bind all sockets for outgoing connections to the same IP address +used for listening (the IP address specified in `rpc_bind_addr`) before +trying to connect to remote nodes. +This can be necessary if a node has multiple IP addresses, +but only one is allowed or able to reach the other nodes, +for instance due to firewall rules or specific routing configuration. + +Disabled by default. + #### `rpc_public_addr` {#rpc_public_addr} The address and port that other nodes need to use to contact this node for