c4af7464e5
The priority of && and || was borked.
1271 lines
41 KiB
Rust
1271 lines
41 KiB
Rust
// Copyright 2015-2018 Parity Technologies (UK) Ltd.
|
|
// This file is part of Parity.
|
|
|
|
// Parity is free software: you can redistribute it and/or modify
|
|
// it under the terms of the GNU General Public License as published by
|
|
// the Free Software Foundation, either version 3 of the License, or
|
|
// (at your option) any later version.
|
|
|
|
// Parity is distributed in the hope that it will be useful,
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
// GNU General Public License for more details.
|
|
|
|
// You should have received a copy of the GNU General Public License
|
|
// along with Parity. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
use std::net::{SocketAddr, SocketAddrV4, Ipv4Addr};
|
|
use std::collections::{HashMap, HashSet};
|
|
use std::str::FromStr;
|
|
use std::sync::Arc;
|
|
use std::sync::atomic::{AtomicBool, Ordering as AtomicOrdering};
|
|
use std::ops::*;
|
|
use std::cmp::{min, max};
|
|
use std::path::{Path, PathBuf};
|
|
use std::io::{Read, Write, self};
|
|
use std::fs;
|
|
use std::time::Duration;
|
|
use ethkey::{KeyPair, Secret, Random, Generator};
|
|
use hash::keccak;
|
|
use mio::*;
|
|
use mio::deprecated::{EventLoop};
|
|
use mio::tcp::*;
|
|
use mio::udp::*;
|
|
use ethereum_types::H256;
|
|
use rlp::{RlpStream, Encodable};
|
|
|
|
use session::{Session, SessionData};
|
|
use io::*;
|
|
use PROTOCOL_VERSION;
|
|
use node_table::*;
|
|
use network::{NetworkConfiguration, NetworkIoMessage, ProtocolId, PeerId, PacketId};
|
|
use network::{NonReservedPeerMode, NetworkContext as NetworkContextTrait};
|
|
use network::{SessionInfo, Error, ErrorKind, DisconnectReason, NetworkProtocolHandler};
|
|
use discovery::{Discovery, TableUpdates, NodeEntry, MAX_DATAGRAM_SIZE};
|
|
use ip_utils::{map_external_address, select_public_address};
|
|
use parity_path::restrict_permissions_owner;
|
|
use parking_lot::{Mutex, RwLock};
|
|
use network::{ConnectionFilter, ConnectionDirection};
|
|
|
|
type Slab<T> = ::slab::Slab<T, usize>;
|
|
|
|
const MAX_SESSIONS: usize = 2048 + MAX_HANDSHAKES;
|
|
const MAX_HANDSHAKES: usize = 1024;
|
|
|
|
const DEFAULT_PORT: u16 = 30303;
|
|
|
|
// StreamToken/TimerToken
|
|
const TCP_ACCEPT: StreamToken = SYS_TIMER + 1;
|
|
const IDLE: TimerToken = SYS_TIMER + 2;
|
|
const DISCOVERY: StreamToken = SYS_TIMER + 3;
|
|
const DISCOVERY_REFRESH: TimerToken = SYS_TIMER + 4;
|
|
const FAST_DISCOVERY_REFRESH: TimerToken = SYS_TIMER + 5;
|
|
const DISCOVERY_ROUND: TimerToken = SYS_TIMER + 6;
|
|
const NODE_TABLE: TimerToken = SYS_TIMER + 7;
|
|
const FIRST_SESSION: StreamToken = 0;
|
|
const LAST_SESSION: StreamToken = FIRST_SESSION + MAX_SESSIONS - 1;
|
|
const USER_TIMER: TimerToken = LAST_SESSION + 256;
|
|
const SYS_TIMER: TimerToken = LAST_SESSION + 1;
|
|
|
|
// Timeouts
|
|
// for IDLE TimerToken
|
|
const MAINTENANCE_TIMEOUT: Duration = Duration::from_secs(1);
|
|
// for DISCOVERY_REFRESH TimerToken
|
|
const DISCOVERY_REFRESH_TIMEOUT: Duration = Duration::from_secs(60);
|
|
// for FAST_DISCOVERY_REFRESH TimerToken
|
|
const FAST_DISCOVERY_REFRESH_TIMEOUT: Duration = Duration::from_secs(10);
|
|
// for DISCOVERY_ROUND TimerToken
|
|
const DISCOVERY_ROUND_TIMEOUT: Duration = Duration::from_millis(300);
|
|
// for NODE_TABLE TimerToken
|
|
const NODE_TABLE_TIMEOUT: Duration = Duration::from_secs(300);
|
|
|
|
#[derive(Debug, PartialEq, Eq)]
|
|
/// Protocol info
|
|
pub struct CapabilityInfo {
|
|
/// Protocol ID
|
|
pub protocol: ProtocolId,
|
|
/// Protocol version
|
|
pub version: u8,
|
|
/// Total number of packet IDs this protocol support.
|
|
pub packet_count: u8,
|
|
}
|
|
|
|
impl Encodable for CapabilityInfo {
|
|
fn rlp_append(&self, s: &mut RlpStream) {
|
|
s.begin_list(2);
|
|
s.append(&&self.protocol[..]);
|
|
s.append(&self.version);
|
|
}
|
|
}
|
|
|
|
/// IO access point. This is passed to all IO handlers and provides an interface to the IO subsystem.
|
|
pub struct NetworkContext<'s> {
|
|
io: &'s IoContext<NetworkIoMessage>,
|
|
protocol: ProtocolId,
|
|
sessions: Arc<RwLock<Slab<SharedSession>>>,
|
|
session: Option<SharedSession>,
|
|
session_id: Option<StreamToken>,
|
|
reserved_peers: &'s HashSet<NodeId>,
|
|
}
|
|
|
|
impl<'s> NetworkContext<'s> {
|
|
/// Create a new network IO access point. Takes references to all the data that can be updated within the IO handler.
|
|
fn new(
|
|
io: &'s IoContext<NetworkIoMessage>,
|
|
protocol: ProtocolId,
|
|
session: Option<SharedSession>,
|
|
sessions: Arc<RwLock<Slab<SharedSession>>>,
|
|
reserved_peers: &'s HashSet<NodeId>,
|
|
) -> NetworkContext<'s> {
|
|
let id = session.as_ref().map(|s| s.lock().token());
|
|
NetworkContext {
|
|
io,
|
|
protocol,
|
|
session_id: id,
|
|
session,
|
|
sessions,
|
|
reserved_peers: reserved_peers,
|
|
}
|
|
}
|
|
|
|
fn resolve_session(&self, peer: PeerId) -> Option<SharedSession> {
|
|
match self.session_id {
|
|
Some(id) if id == peer => self.session.clone(),
|
|
_ => self.sessions.read().get(peer).cloned(),
|
|
}
|
|
}
|
|
}
|
|
|
|
impl<'s> NetworkContextTrait for NetworkContext<'s> {
|
|
fn send(&self, peer: PeerId, packet_id: PacketId, data: Vec<u8>) -> Result<(), Error> {
|
|
self.send_protocol(self.protocol, peer, packet_id, data)
|
|
}
|
|
|
|
fn send_protocol(&self, protocol: ProtocolId, peer: PeerId, packet_id: PacketId, data: Vec<u8>) -> Result<(), Error> {
|
|
let session = self.resolve_session(peer);
|
|
if let Some(session) = session {
|
|
session.lock().send_packet(self.io, Some(protocol), packet_id as u8, &data)?;
|
|
} else {
|
|
trace!(target: "network", "Send: Peer no longer exist")
|
|
}
|
|
Ok(())
|
|
}
|
|
|
|
fn respond(&self, packet_id: PacketId, data: Vec<u8>) -> Result<(), Error> {
|
|
assert!(self.session.is_some(), "Respond called without network context");
|
|
self.session_id.map_or_else(|| Err(ErrorKind::Expired.into()), |id| self.send(id, packet_id, data))
|
|
}
|
|
|
|
fn disable_peer(&self, peer: PeerId) {
|
|
self.io.message(NetworkIoMessage::DisablePeer(peer))
|
|
.unwrap_or_else(|e| warn!("Error sending network IO message: {:?}", e));
|
|
}
|
|
|
|
fn disconnect_peer(&self, peer: PeerId) {
|
|
self.io.message(NetworkIoMessage::Disconnect(peer))
|
|
.unwrap_or_else(|e| warn!("Error sending network IO message: {:?}", e));
|
|
}
|
|
|
|
fn is_expired(&self) -> bool {
|
|
self.session.as_ref().map_or(false, |s| s.lock().expired())
|
|
}
|
|
|
|
fn register_timer(&self, token: TimerToken, delay: Duration) -> Result<(), Error> {
|
|
self.io.message(NetworkIoMessage::AddTimer {
|
|
token,
|
|
delay,
|
|
protocol: self.protocol,
|
|
}).unwrap_or_else(|e| warn!("Error sending network IO message: {:?}", e));
|
|
Ok(())
|
|
}
|
|
|
|
fn peer_client_version(&self, peer: PeerId) -> String {
|
|
self.resolve_session(peer).map_or("unknown".to_owned(), |s| s.lock().info.client_version.clone())
|
|
}
|
|
|
|
fn session_info(&self, peer: PeerId) -> Option<SessionInfo> {
|
|
self.resolve_session(peer).map(|s| s.lock().info.clone())
|
|
}
|
|
|
|
fn protocol_version(&self, protocol: ProtocolId, peer: PeerId) -> Option<u8> {
|
|
let session = self.resolve_session(peer);
|
|
session.and_then(|s| s.lock().capability_version(protocol))
|
|
}
|
|
|
|
fn subprotocol_name(&self) -> ProtocolId { self.protocol }
|
|
|
|
fn is_reserved_peer(&self, peer: PeerId) -> bool {
|
|
self.session_info(peer)
|
|
.and_then(|info| info.id)
|
|
.map(|node| self.reserved_peers.contains(&node))
|
|
.unwrap_or(false)
|
|
}
|
|
}
|
|
|
|
/// Shared host information
|
|
pub struct HostInfo {
|
|
/// Our private and public keys.
|
|
keys: KeyPair,
|
|
/// Current network configuration
|
|
config: NetworkConfiguration,
|
|
/// Connection nonce.
|
|
nonce: H256,
|
|
/// RLPx protocol version
|
|
pub protocol_version: u32,
|
|
/// Registered capabilities (handlers)
|
|
pub capabilities: Vec<CapabilityInfo>,
|
|
/// Local address + discovery port
|
|
pub local_endpoint: NodeEndpoint,
|
|
/// Public address + discovery port
|
|
pub public_endpoint: Option<NodeEndpoint>,
|
|
}
|
|
|
|
impl HostInfo {
|
|
fn next_nonce(&mut self) -> H256 {
|
|
self.nonce = keccak(&self.nonce);
|
|
self.nonce
|
|
}
|
|
|
|
pub(crate) fn client_version(&self) -> &str {
|
|
&self.config.client_version
|
|
}
|
|
|
|
pub(crate) fn secret(&self) -> &Secret {
|
|
self.keys.secret()
|
|
}
|
|
|
|
pub(crate) fn id(&self) -> &NodeId {
|
|
self.keys.public()
|
|
}
|
|
}
|
|
|
|
type SharedSession = Arc<Mutex<Session>>;
|
|
|
|
#[derive(Copy, Clone)]
|
|
struct ProtocolTimer {
|
|
pub protocol: ProtocolId,
|
|
pub token: TimerToken, // Handler level token
|
|
}
|
|
|
|
/// Root IO handler. Manages protocol handlers, IO timers and network connections.
|
|
pub struct Host {
|
|
pub info: RwLock<HostInfo>,
|
|
udp_socket: Mutex<Option<UdpSocket>>,
|
|
tcp_listener: Mutex<TcpListener>,
|
|
sessions: Arc<RwLock<Slab<SharedSession>>>,
|
|
discovery: Mutex<Option<Discovery<'static>>>,
|
|
nodes: RwLock<NodeTable>,
|
|
handlers: RwLock<HashMap<ProtocolId, Arc<NetworkProtocolHandler + Sync>>>,
|
|
timers: RwLock<HashMap<TimerToken, ProtocolTimer>>,
|
|
timer_counter: RwLock<usize>,
|
|
reserved_nodes: RwLock<HashSet<NodeId>>,
|
|
stopping: AtomicBool,
|
|
filter: Option<Arc<ConnectionFilter>>,
|
|
}
|
|
|
|
impl Host {
|
|
/// Create a new instance
|
|
pub fn new(mut config: NetworkConfiguration, filter: Option<Arc<ConnectionFilter>>) -> Result<Host, Error> {
|
|
let mut listen_address = match config.listen_address {
|
|
None => SocketAddr::V4(SocketAddrV4::new(Ipv4Addr::new(0, 0, 0, 0), DEFAULT_PORT)),
|
|
Some(addr) => addr,
|
|
};
|
|
|
|
let keys = if let Some(ref secret) = config.use_secret {
|
|
KeyPair::from_secret(secret.clone())?
|
|
} else {
|
|
config.config_path.clone().and_then(|ref p| load_key(Path::new(&p)))
|
|
.map_or_else(|| {
|
|
let key = Random.generate().expect("Error generating random key pair");
|
|
if let Some(path) = config.config_path.clone() {
|
|
save_key(Path::new(&path), key.secret());
|
|
}
|
|
key
|
|
},
|
|
|s| KeyPair::from_secret(s).expect("Error creating node secret key"))
|
|
};
|
|
let path = config.net_config_path.clone();
|
|
// Setup the server socket
|
|
let tcp_listener = TcpListener::bind(&listen_address)?;
|
|
listen_address = SocketAddr::new(listen_address.ip(), tcp_listener.local_addr()?.port());
|
|
debug!(target: "network", "Listening at {:?}", listen_address);
|
|
let udp_port = config.udp_port.unwrap_or_else(|| listen_address.port());
|
|
let local_endpoint = NodeEndpoint { address: listen_address, udp_port };
|
|
|
|
let boot_nodes = config.boot_nodes.clone();
|
|
let reserved_nodes = config.reserved_nodes.clone();
|
|
config.max_handshakes = min(config.max_handshakes, MAX_HANDSHAKES as u32);
|
|
|
|
let mut host = Host {
|
|
info: RwLock::new(HostInfo {
|
|
keys,
|
|
config,
|
|
nonce: H256::random(),
|
|
protocol_version: PROTOCOL_VERSION,
|
|
capabilities: Vec::new(),
|
|
public_endpoint: None,
|
|
local_endpoint,
|
|
}),
|
|
discovery: Mutex::new(None),
|
|
udp_socket: Mutex::new(None),
|
|
tcp_listener: Mutex::new(tcp_listener),
|
|
sessions: Arc::new(RwLock::new(Slab::new_starting_at(FIRST_SESSION, MAX_SESSIONS))),
|
|
nodes: RwLock::new(NodeTable::new(path)),
|
|
handlers: RwLock::new(HashMap::new()),
|
|
timers: RwLock::new(HashMap::new()),
|
|
timer_counter: RwLock::new(USER_TIMER),
|
|
reserved_nodes: RwLock::new(HashSet::new()),
|
|
stopping: AtomicBool::new(false),
|
|
filter,
|
|
};
|
|
|
|
for n in boot_nodes {
|
|
host.add_node(&n);
|
|
}
|
|
|
|
for n in reserved_nodes {
|
|
if let Err(e) = host.add_reserved_node(&n) {
|
|
debug!(target: "network", "Error parsing node id: {}: {:?}", n, e);
|
|
}
|
|
}
|
|
Ok(host)
|
|
}
|
|
|
|
pub fn add_node(&mut self, id: &str) {
|
|
match Node::from_str(id) {
|
|
Err(e) => { debug!(target: "network", "Could not add node {}: {:?}", id, e); },
|
|
Ok(n) => {
|
|
let entry = NodeEntry { endpoint: n.endpoint.clone(), id: n.id };
|
|
|
|
self.nodes.write().add_node(n);
|
|
if let Some(ref mut discovery) = *self.discovery.lock() {
|
|
discovery.add_node(entry);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
pub fn add_reserved_node(&self, id: &str) -> Result<(), Error> {
|
|
let n = Node::from_str(id)?;
|
|
|
|
let entry = NodeEntry { endpoint: n.endpoint.clone(), id: n.id };
|
|
self.reserved_nodes.write().insert(n.id);
|
|
self.nodes.write().add_node(Node::new(entry.id, entry.endpoint.clone()));
|
|
|
|
if let Some(ref mut discovery) = *self.discovery.lock() {
|
|
discovery.add_node(entry);
|
|
}
|
|
|
|
Ok(())
|
|
}
|
|
|
|
pub fn set_non_reserved_mode(&self, mode: NonReservedPeerMode, io: &IoContext<NetworkIoMessage>) {
|
|
let mut info = self.info.write();
|
|
|
|
if info.config.non_reserved_mode != mode {
|
|
info.config.non_reserved_mode = mode;
|
|
drop(info);
|
|
if let NonReservedPeerMode::Deny = mode {
|
|
// disconnect all non-reserved peers here.
|
|
let reserved: HashSet<NodeId> = self.reserved_nodes.read().clone();
|
|
let mut to_kill = Vec::new();
|
|
for e in self.sessions.read().iter() {
|
|
let mut s = e.lock();
|
|
{
|
|
let id = s.id();
|
|
if id.map_or(false, |id| reserved.contains(id)) {
|
|
continue;
|
|
}
|
|
}
|
|
|
|
s.disconnect(io, DisconnectReason::ClientQuit);
|
|
to_kill.push(s.token());
|
|
}
|
|
for p in to_kill {
|
|
trace!(target: "network", "Disconnecting on reserved-only mode: {}", p);
|
|
self.kill_connection(p, io, false);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
pub fn remove_reserved_node(&self, id: &str) -> Result<(), Error> {
|
|
let n = Node::from_str(id)?;
|
|
self.reserved_nodes.write().remove(&n.id);
|
|
|
|
Ok(())
|
|
}
|
|
|
|
pub fn external_url(&self) -> Option<String> {
|
|
let info = self.info.read();
|
|
info.public_endpoint.as_ref().map(|e| format!("{}", Node::new(*info.id(), e.clone())))
|
|
}
|
|
|
|
pub fn local_url(&self) -> String {
|
|
let info = self.info.read();
|
|
format!("{}", Node::new(*info.id(), info.local_endpoint.clone()))
|
|
}
|
|
|
|
pub fn stop(&self, io: &IoContext<NetworkIoMessage>) {
|
|
self.stopping.store(true, AtomicOrdering::Release);
|
|
let mut to_kill = Vec::new();
|
|
for e in self.sessions.read().iter() {
|
|
let mut s = e.lock();
|
|
s.disconnect(io, DisconnectReason::ClientQuit);
|
|
to_kill.push(s.token());
|
|
}
|
|
for p in to_kill {
|
|
trace!(target: "network", "Disconnecting on shutdown: {}", p);
|
|
self.kill_connection(p, io, true);
|
|
}
|
|
io.unregister_handler();
|
|
}
|
|
|
|
/// Get all connected peers.
|
|
pub fn connected_peers(&self) -> Vec<PeerId> {
|
|
let sessions = self.sessions.read();
|
|
let sessions = &*sessions;
|
|
|
|
let mut peers = Vec::with_capacity(sessions.count());
|
|
for i in (0..MAX_SESSIONS).map(|x| x + FIRST_SESSION) {
|
|
if sessions.get(i).is_some() {
|
|
peers.push(i);
|
|
}
|
|
}
|
|
peers
|
|
}
|
|
|
|
fn init_public_interface(&self, io: &IoContext<NetworkIoMessage>) -> Result<(), Error> {
|
|
if self.info.read().public_endpoint.is_some() {
|
|
return Ok(());
|
|
}
|
|
let local_endpoint = self.info.read().local_endpoint.clone();
|
|
let public_address = self.info.read().config.public_address;
|
|
let allow_ips = self.info.read().config.ip_filter.clone();
|
|
let public_endpoint = match public_address {
|
|
None => {
|
|
let public_address = select_public_address(local_endpoint.address.port());
|
|
let public_endpoint = NodeEndpoint { address: public_address, udp_port: local_endpoint.udp_port };
|
|
if self.info.read().config.nat_enabled {
|
|
match map_external_address(&local_endpoint) {
|
|
Some(endpoint) => {
|
|
info!("NAT mapped to external address {}", endpoint.address);
|
|
endpoint
|
|
},
|
|
None => public_endpoint
|
|
}
|
|
} else {
|
|
public_endpoint
|
|
}
|
|
}
|
|
Some(addr) => NodeEndpoint { address: addr, udp_port: local_endpoint.udp_port }
|
|
};
|
|
|
|
self.info.write().public_endpoint = Some(public_endpoint.clone());
|
|
|
|
if let Some(url) = self.external_url() {
|
|
io.message(NetworkIoMessage::NetworkStarted(url)).unwrap_or_else(|e| warn!("Error sending IO notification: {:?}", e));
|
|
}
|
|
|
|
// Initialize discovery.
|
|
let discovery = {
|
|
let info = self.info.read();
|
|
if info.config.discovery_enabled && info.config.non_reserved_mode == NonReservedPeerMode::Accept {
|
|
Some(Discovery::new(&info.keys, public_endpoint, allow_ips))
|
|
} else { None }
|
|
};
|
|
|
|
if let Some(mut discovery) = discovery {
|
|
let mut udp_addr = local_endpoint.address;
|
|
udp_addr.set_port(local_endpoint.udp_port);
|
|
let socket = UdpSocket::bind(&udp_addr).expect("Error binding UDP socket");
|
|
*self.udp_socket.lock() = Some(socket);
|
|
|
|
discovery.add_node_list(self.nodes.read().entries());
|
|
*self.discovery.lock() = Some(discovery);
|
|
io.register_stream(DISCOVERY)?;
|
|
io.register_timer(FAST_DISCOVERY_REFRESH, FAST_DISCOVERY_REFRESH_TIMEOUT)?;
|
|
io.register_timer(DISCOVERY_REFRESH, DISCOVERY_REFRESH_TIMEOUT)?;
|
|
io.register_timer(DISCOVERY_ROUND, DISCOVERY_ROUND_TIMEOUT)?;
|
|
}
|
|
io.register_timer(NODE_TABLE, NODE_TABLE_TIMEOUT)?;
|
|
io.register_stream(TCP_ACCEPT)?;
|
|
Ok(())
|
|
}
|
|
|
|
fn maintain_network(&self, io: &IoContext<NetworkIoMessage>) {
|
|
self.keep_alive(io);
|
|
self.connect_peers(io);
|
|
}
|
|
|
|
fn have_session(&self, id: &NodeId) -> bool {
|
|
self.sessions.read().iter().any(|e| e.lock().info.id == Some(*id))
|
|
}
|
|
|
|
// returns (handshakes, egress, ingress)
|
|
fn session_count(&self) -> (usize, usize, usize) {
|
|
let mut handshakes = 0;
|
|
let mut egress = 0;
|
|
let mut ingress = 0;
|
|
for s in self.sessions.read().iter() {
|
|
match s.try_lock() {
|
|
Some(ref s) if s.is_ready() && s.info.originated => egress += 1,
|
|
Some(ref s) if s.is_ready() && !s.info.originated => ingress += 1,
|
|
_ => handshakes +=1,
|
|
}
|
|
}
|
|
(handshakes, egress, ingress)
|
|
}
|
|
|
|
fn connecting_to(&self, id: &NodeId) -> bool {
|
|
self.sessions.read().iter().any(|e| e.lock().id() == Some(id))
|
|
}
|
|
|
|
fn keep_alive(&self, io: &IoContext<NetworkIoMessage>) {
|
|
let mut to_kill = Vec::new();
|
|
for e in self.sessions.read().iter() {
|
|
let mut s = e.lock();
|
|
if !s.keep_alive(io) {
|
|
s.disconnect(io, DisconnectReason::PingTimeout);
|
|
to_kill.push(s.token());
|
|
}
|
|
}
|
|
for p in to_kill {
|
|
trace!(target: "network", "Ping timeout: {}", p);
|
|
self.kill_connection(p, io, true);
|
|
}
|
|
}
|
|
|
|
fn has_enough_peers(&self) -> bool {
|
|
let min_peers = {
|
|
let info = self.info.read();
|
|
let config = &info.config;
|
|
|
|
config.min_peers
|
|
};
|
|
let (_, egress_count, ingress_count) = self.session_count();
|
|
|
|
return egress_count + ingress_count >= min_peers as usize;
|
|
}
|
|
|
|
fn connect_peers(&self, io: &IoContext<NetworkIoMessage>) {
|
|
let (min_peers, mut pin, max_handshakes, allow_ips, self_id) = {
|
|
let info = self.info.read();
|
|
if info.capabilities.is_empty() {
|
|
return;
|
|
}
|
|
let config = &info.config;
|
|
|
|
(config.min_peers, config.non_reserved_mode == NonReservedPeerMode::Deny, config.max_handshakes as usize, config.ip_filter.clone(), *info.id())
|
|
};
|
|
|
|
let (handshake_count, egress_count, ingress_count) = self.session_count();
|
|
let reserved_nodes = self.reserved_nodes.read();
|
|
if egress_count + ingress_count >= min_peers as usize + reserved_nodes.len() {
|
|
// check if all pinned nodes are connected.
|
|
if reserved_nodes.iter().all(|n| self.have_session(n) && self.connecting_to(n)) {
|
|
return;
|
|
}
|
|
|
|
// if not, only attempt connect to reserved peers
|
|
pin = true;
|
|
}
|
|
|
|
// allow 16 slots for incoming connections
|
|
if handshake_count >= max_handshakes {
|
|
return;
|
|
}
|
|
|
|
// iterate over all nodes, reserved ones coming first.
|
|
// if we are pinned to only reserved nodes, ignore all others.
|
|
let nodes = reserved_nodes.iter().cloned().chain(if !pin {
|
|
self.nodes.read().nodes(&allow_ips)
|
|
} else {
|
|
Vec::new()
|
|
});
|
|
|
|
let max_handshakes_per_round = max_handshakes / 2;
|
|
let mut started: usize = 0;
|
|
for id in nodes.filter(|id|
|
|
!self.have_session(id) &&
|
|
!self.connecting_to(id) &&
|
|
*id != self_id &&
|
|
self.filter.as_ref().map_or(true, |f| f.connection_allowed(&self_id, &id, ConnectionDirection::Outbound))
|
|
).take(min(max_handshakes_per_round, max_handshakes - handshake_count)) {
|
|
self.connect_peer(&id, io);
|
|
started += 1;
|
|
}
|
|
debug!(target: "network", "Connecting peers: {} sessions, {} pending + {} started", egress_count + ingress_count, handshake_count, started);
|
|
}
|
|
|
|
fn connect_peer(&self, id: &NodeId, io: &IoContext<NetworkIoMessage>) {
|
|
if self.have_session(id) {
|
|
trace!(target: "network", "Aborted connect. Node already connected.");
|
|
return;
|
|
}
|
|
if self.connecting_to(id) {
|
|
trace!(target: "network", "Aborted connect. Node already connecting.");
|
|
return;
|
|
}
|
|
|
|
let socket = {
|
|
let address = {
|
|
let mut nodes = self.nodes.write();
|
|
if let Some(node) = nodes.get_mut(id) {
|
|
node.endpoint.address
|
|
} else {
|
|
debug!(target: "network", "Connection to expired node aborted");
|
|
return;
|
|
}
|
|
};
|
|
match TcpStream::connect(&address) {
|
|
Ok(socket) => {
|
|
trace!(target: "network", "{}: Connecting to {:?}", id, address);
|
|
socket
|
|
},
|
|
Err(e) => {
|
|
debug!(target: "network", "{}: Can't connect to address {:?}: {:?}", id, address, e);
|
|
self.nodes.write().note_failure(&id);
|
|
return;
|
|
}
|
|
}
|
|
};
|
|
|
|
if let Err(e) = self.create_connection(socket, Some(id), io) {
|
|
debug!(target: "network", "Can't create connection: {:?}", e);
|
|
}
|
|
}
|
|
|
|
fn create_connection(&self, socket: TcpStream, id: Option<&NodeId>, io: &IoContext<NetworkIoMessage>) -> Result<(), Error> {
|
|
let nonce = self.info.write().next_nonce();
|
|
let mut sessions = self.sessions.write();
|
|
|
|
let token = sessions.insert_with_opt(|token| {
|
|
trace!(target: "network", "{}: Initiating session {:?}", token, id);
|
|
match Session::new(io, socket, token, id, &nonce, &self.info.read()) {
|
|
Ok(s) => Some(Arc::new(Mutex::new(s))),
|
|
Err(e) => {
|
|
debug!(target: "network", "Session create error: {:?}", e);
|
|
None
|
|
}
|
|
}
|
|
});
|
|
|
|
match token {
|
|
Some(t) => io.register_stream(t).map(|_| ()).map_err(Into::into),
|
|
None => {
|
|
debug!(target: "network", "Max sessions reached");
|
|
Ok(())
|
|
}
|
|
}
|
|
}
|
|
|
|
fn accept(&self, io: &IoContext<NetworkIoMessage>) {
|
|
trace!(target: "network", "Accepting incoming connection");
|
|
loop {
|
|
let socket = match self.tcp_listener.lock().accept() {
|
|
Ok((sock, _addr)) => sock,
|
|
Err(e) => {
|
|
if e.kind() != io::ErrorKind::WouldBlock {
|
|
debug!(target: "network", "Error accepting connection: {:?}", e);
|
|
}
|
|
break
|
|
},
|
|
};
|
|
if let Err(e) = self.create_connection(socket, None, io) {
|
|
debug!(target: "network", "Can't accept connection: {:?}", e);
|
|
}
|
|
}
|
|
}
|
|
|
|
fn session_writable(&self, token: StreamToken, io: &IoContext<NetworkIoMessage>) {
|
|
let session = { self.sessions.read().get(token).cloned() };
|
|
|
|
if let Some(session) = session {
|
|
let mut s = session.lock();
|
|
if let Err(e) = s.writable(io, &self.info.read()) {
|
|
trace!(target: "network", "Session write error: {}: {:?}", token, e);
|
|
}
|
|
if s.done() {
|
|
io.deregister_stream(token).unwrap_or_else(|e| debug!("Error deregistering stream: {:?}", e));
|
|
}
|
|
}
|
|
}
|
|
|
|
fn connection_closed(&self, token: StreamToken, io: &IoContext<NetworkIoMessage>) {
|
|
trace!(target: "network", "Connection closed: {}", token);
|
|
self.kill_connection(token, io, true);
|
|
}
|
|
|
|
fn session_readable(&self, token: StreamToken, io: &IoContext<NetworkIoMessage>) {
|
|
let mut ready_data: Vec<ProtocolId> = Vec::new();
|
|
let mut packet_data: Vec<(ProtocolId, PacketId, Vec<u8>)> = Vec::new();
|
|
let mut kill = false;
|
|
let session = { self.sessions.read().get(token).cloned() };
|
|
let mut ready_id = None;
|
|
if let Some(session) = session.clone() {
|
|
{
|
|
loop {
|
|
let session_result = session.lock().readable(io, &self.info.read());
|
|
match session_result {
|
|
Err(e) => {
|
|
let s = session.lock();
|
|
trace!(target: "network", "Session read error: {}:{:?} ({:?}) {:?}", token, s.id(), s.remote_addr(), e);
|
|
match *e.kind() {
|
|
ErrorKind::Disconnect(DisconnectReason::IncompatibleProtocol) | ErrorKind::Disconnect(DisconnectReason::UselessPeer) => {
|
|
if let Some(id) = s.id() {
|
|
if !self.reserved_nodes.read().contains(id) {
|
|
let mut nodes = self.nodes.write();
|
|
nodes.note_failure(&id);
|
|
nodes.mark_as_useless(id);
|
|
}
|
|
}
|
|
},
|
|
_ => {},
|
|
}
|
|
kill = true;
|
|
break;
|
|
},
|
|
Ok(SessionData::Ready) => {
|
|
let (_, egress_count, ingress_count) = self.session_count();
|
|
let mut s = session.lock();
|
|
let (min_peers, mut max_peers, reserved_only, self_id) = {
|
|
let info = self.info.read();
|
|
let mut max_peers = info.config.max_peers;
|
|
for cap in &s.info.capabilities {
|
|
if let Some(num) = info.config.reserved_protocols.get(&cap.protocol) {
|
|
max_peers += *num;
|
|
break;
|
|
}
|
|
}
|
|
(info.config.min_peers as usize, max_peers as usize, info.config.non_reserved_mode == NonReservedPeerMode::Deny, *info.id())
|
|
};
|
|
|
|
max_peers = max(max_peers, min_peers);
|
|
|
|
let id = *s.id().expect("Ready session always has id");
|
|
|
|
// Check for the session limit.
|
|
// Outgoing connections are allowed as long as their count is <= min_peers
|
|
// Incoming connections are allowed to take all of the max_peers reserve, or at most half of the slots.
|
|
let max_ingress = max(max_peers - min_peers, min_peers / 2);
|
|
if reserved_only ||
|
|
(s.info.originated && egress_count > min_peers) ||
|
|
(!s.info.originated && ingress_count > max_ingress) {
|
|
if !self.reserved_nodes.read().contains(&id) {
|
|
// only proceed if the connecting peer is reserved.
|
|
trace!(target: "network", "Disconnecting non-reserved peer {:?}", id);
|
|
s.disconnect(io, DisconnectReason::TooManyPeers);
|
|
kill = true;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if !self.filter.as_ref().map_or(true, |f| f.connection_allowed(&self_id, &id, ConnectionDirection::Inbound)) {
|
|
trace!(target: "network", "Inbound connection not allowed for {:?}", id);
|
|
s.disconnect(io, DisconnectReason::UnexpectedIdentity);
|
|
kill = true;
|
|
break;
|
|
}
|
|
|
|
ready_id = Some(id);
|
|
|
|
// Add it to the node table
|
|
if !s.info.originated {
|
|
if let Ok(address) = s.remote_addr() {
|
|
// We can't know remote listening ports, so just assume defaults and hope for the best.
|
|
let endpoint = NodeEndpoint { address: SocketAddr::new(address.ip(), DEFAULT_PORT), udp_port: DEFAULT_PORT };
|
|
let entry = NodeEntry { id, endpoint };
|
|
let mut nodes = self.nodes.write();
|
|
if !nodes.contains(&entry.id) {
|
|
nodes.add_node(Node::new(entry.id, entry.endpoint.clone()));
|
|
let mut discovery = self.discovery.lock();
|
|
if let Some(ref mut discovery) = *discovery {
|
|
discovery.add_node(entry);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Note connection success
|
|
self.nodes.write().note_success(&id);
|
|
|
|
for (p, _) in self.handlers.read().iter() {
|
|
if s.have_capability(*p) {
|
|
ready_data.push(*p);
|
|
}
|
|
}
|
|
},
|
|
Ok(SessionData::Packet {
|
|
data,
|
|
protocol,
|
|
packet_id,
|
|
}) => {
|
|
match self.handlers.read().get(&protocol) {
|
|
None => { warn!(target: "network", "No handler found for protocol: {:?}", protocol) },
|
|
Some(_) => packet_data.push((protocol, packet_id, data)),
|
|
}
|
|
},
|
|
Ok(SessionData::Continue) => (),
|
|
Ok(SessionData::None) => break,
|
|
}
|
|
}
|
|
}
|
|
|
|
if kill {
|
|
self.kill_connection(token, io, true);
|
|
}
|
|
|
|
let handlers = self.handlers.read();
|
|
if !ready_data.is_empty() {
|
|
let duplicate = self.sessions.read().iter().any(|e| {
|
|
let session = e.lock();
|
|
session.token() != token && session.info.id == ready_id
|
|
});
|
|
if duplicate {
|
|
trace!(target: "network", "Rejected duplicate connection: {}", token);
|
|
session.lock().disconnect(io, DisconnectReason::DuplicatePeer);
|
|
self.kill_connection(token, io, false);
|
|
return;
|
|
}
|
|
for p in ready_data {
|
|
let reserved = self.reserved_nodes.read();
|
|
if let Some(h) = handlers.get(&p) {
|
|
h.connected(&NetworkContext::new(io, p, Some(session.clone()), self.sessions.clone(), &reserved), &token);
|
|
// accumulate pending packets.
|
|
let mut session = session.lock();
|
|
packet_data.extend(session.mark_connected(p));
|
|
}
|
|
}
|
|
}
|
|
|
|
for (p, packet_id, data) in packet_data {
|
|
let reserved = self.reserved_nodes.read();
|
|
if let Some(h) = handlers.get(&p) {
|
|
h.read(&NetworkContext::new(io, p, Some(session.clone()), self.sessions.clone(), &reserved), &token, packet_id, &data);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
fn discovery_readable(&self, io: &IoContext<NetworkIoMessage>) {
|
|
let node_changes = match (self.udp_socket.lock().as_ref(), self.discovery.lock().as_mut()) {
|
|
(Some(udp_socket), Some(discovery)) => {
|
|
let mut buf = [0u8; MAX_DATAGRAM_SIZE];
|
|
let writable = discovery.any_sends_queued();
|
|
let res = match udp_socket.recv_from(&mut buf) {
|
|
Ok(Some((len, address))) => discovery.on_packet(&buf[0..len], address).unwrap_or_else(|e| {
|
|
debug!(target: "network", "Error processing UDP packet: {:?}", e);
|
|
None
|
|
}),
|
|
Ok(_) => None,
|
|
Err(e) => {
|
|
debug!(target: "network", "Error reading UPD socket: {:?}", e);
|
|
None
|
|
}
|
|
};
|
|
let new_writable = discovery.any_sends_queued();
|
|
if writable != new_writable {
|
|
io.update_registration(DISCOVERY)
|
|
.unwrap_or_else(|e| {
|
|
debug!(target: "network" ,"Error updating discovery registration: {:?}", e)
|
|
});
|
|
}
|
|
res
|
|
},
|
|
_ => None,
|
|
};
|
|
if let Some(node_changes) = node_changes {
|
|
self.update_nodes(io, node_changes);
|
|
}
|
|
}
|
|
|
|
fn discovery_writable(&self, io: &IoContext<NetworkIoMessage>) {
|
|
if let (Some(udp_socket), Some(discovery)) = (self.udp_socket.lock().as_ref(), self.discovery.lock().as_mut()) {
|
|
while let Some(data) = discovery.dequeue_send() {
|
|
match udp_socket.send_to(&data.payload, &data.address) {
|
|
Ok(Some(size)) if size == data.payload.len() => {
|
|
},
|
|
Ok(Some(_)) => {
|
|
warn!(target: "network", "UDP sent incomplete datagram");
|
|
},
|
|
Ok(None) => {
|
|
discovery.requeue_send(data);
|
|
return;
|
|
}
|
|
Err(e) => {
|
|
debug!(target: "network", "UDP send error: {:?}, address: {:?}", e, &data.address);
|
|
return;
|
|
}
|
|
}
|
|
}
|
|
io.update_registration(DISCOVERY)
|
|
.unwrap_or_else(|e| {
|
|
debug!(target: "network", "Error updating discovery registration: {:?}", e)
|
|
});
|
|
}
|
|
}
|
|
|
|
fn connection_timeout(&self, token: StreamToken, io: &IoContext<NetworkIoMessage>) {
|
|
trace!(target: "network", "Connection timeout: {}", token);
|
|
self.kill_connection(token, io, true)
|
|
}
|
|
|
|
fn kill_connection(&self, token: StreamToken, io: &IoContext<NetworkIoMessage>, remote: bool) {
|
|
let mut to_disconnect: Vec<ProtocolId> = Vec::new();
|
|
let mut failure_id = None;
|
|
let mut deregister = false;
|
|
let mut expired_session = None;
|
|
if let FIRST_SESSION ... LAST_SESSION = token {
|
|
let sessions = self.sessions.read();
|
|
if let Some(session) = sessions.get(token).cloned() {
|
|
expired_session = Some(session.clone());
|
|
let mut s = session.lock();
|
|
if !s.expired() {
|
|
if s.is_ready() {
|
|
for (p, _) in self.handlers.read().iter() {
|
|
if s.have_capability(*p) {
|
|
to_disconnect.push(*p);
|
|
}
|
|
}
|
|
}
|
|
s.set_expired();
|
|
failure_id = s.id().cloned();
|
|
}
|
|
deregister = remote || s.done();
|
|
}
|
|
}
|
|
if let Some(id) = failure_id {
|
|
if remote {
|
|
self.nodes.write().note_failure(&id);
|
|
}
|
|
}
|
|
for p in to_disconnect {
|
|
let reserved = self.reserved_nodes.read();
|
|
if let Some(h) = self.handlers.read().get(&p) {
|
|
h.disconnected(&NetworkContext::new(io, p, expired_session.clone(), self.sessions.clone(), &reserved), &token);
|
|
}
|
|
}
|
|
if deregister {
|
|
io.deregister_stream(token).unwrap_or_else(|e| debug!("Error deregistering stream: {:?}", e));
|
|
}
|
|
}
|
|
|
|
fn update_nodes(&self, _io: &IoContext<NetworkIoMessage>, node_changes: TableUpdates) {
|
|
let mut to_remove: Vec<PeerId> = Vec::new();
|
|
{
|
|
let sessions = self.sessions.read();
|
|
for c in sessions.iter() {
|
|
let s = c.lock();
|
|
if let Some(id) = s.id() {
|
|
if node_changes.removed.contains(id) {
|
|
to_remove.push(s.token());
|
|
}
|
|
}
|
|
}
|
|
}
|
|
for i in to_remove {
|
|
trace!(target: "network", "Removed from node table: {}", i);
|
|
}
|
|
self.nodes.write().update(node_changes, &*self.reserved_nodes.read());
|
|
}
|
|
|
|
pub fn with_context<F>(&self, protocol: ProtocolId, io: &IoContext<NetworkIoMessage>, action: F) where F: FnOnce(&NetworkContextTrait) {
|
|
let reserved = { self.reserved_nodes.read() };
|
|
|
|
let context = NetworkContext::new(io, protocol, None, self.sessions.clone(), &reserved);
|
|
action(&context);
|
|
}
|
|
|
|
pub fn with_context_eval<F, T>(&self, protocol: ProtocolId, io: &IoContext<NetworkIoMessage>, action: F) -> T where F: FnOnce(&NetworkContextTrait) -> T {
|
|
let reserved = { self.reserved_nodes.read() };
|
|
|
|
let context = NetworkContext::new(io, protocol, None, self.sessions.clone(), &reserved);
|
|
action(&context)
|
|
}
|
|
}
|
|
|
|
impl IoHandler<NetworkIoMessage> for Host {
|
|
/// Initialize networking
|
|
fn initialize(&self, io: &IoContext<NetworkIoMessage>) {
|
|
io.register_timer(IDLE, MAINTENANCE_TIMEOUT).expect("Error registering Network idle timer");
|
|
io.message(NetworkIoMessage::InitPublicInterface).unwrap_or_else(|e| warn!("Error sending IO notification: {:?}", e));
|
|
self.maintain_network(io)
|
|
}
|
|
|
|
fn stream_hup(&self, io: &IoContext<NetworkIoMessage>, stream: StreamToken) {
|
|
trace!(target: "network", "Hup: {}", stream);
|
|
match stream {
|
|
FIRST_SESSION ... LAST_SESSION => self.connection_closed(stream, io),
|
|
_ => warn!(target: "network", "Unexpected hup"),
|
|
};
|
|
}
|
|
|
|
fn stream_readable(&self, io: &IoContext<NetworkIoMessage>, stream: StreamToken) {
|
|
if self.stopping.load(AtomicOrdering::Acquire) {
|
|
return;
|
|
}
|
|
match stream {
|
|
FIRST_SESSION ... LAST_SESSION => self.session_readable(stream, io),
|
|
DISCOVERY => self.discovery_readable(io),
|
|
TCP_ACCEPT => self.accept(io),
|
|
_ => panic!("Received unknown readable token"),
|
|
}
|
|
}
|
|
|
|
fn stream_writable(&self, io: &IoContext<NetworkIoMessage>, stream: StreamToken) {
|
|
if self.stopping.load(AtomicOrdering::Acquire) {
|
|
return;
|
|
}
|
|
match stream {
|
|
FIRST_SESSION ... LAST_SESSION => self.session_writable(stream, io),
|
|
DISCOVERY => self.discovery_writable(io),
|
|
_ => panic!("Received unknown writable token"),
|
|
}
|
|
}
|
|
|
|
fn timeout(&self, io: &IoContext<NetworkIoMessage>, token: TimerToken) {
|
|
if self.stopping.load(AtomicOrdering::Acquire) {
|
|
return;
|
|
}
|
|
match token {
|
|
IDLE => self.maintain_network(io),
|
|
FIRST_SESSION ... LAST_SESSION => self.connection_timeout(token, io),
|
|
DISCOVERY_REFRESH => {
|
|
// Run the _slow_ discovery if enough peers are connected
|
|
if !self.has_enough_peers() {
|
|
return;
|
|
}
|
|
self.discovery.lock().as_mut().map(|d| d.refresh());
|
|
io.update_registration(DISCOVERY).unwrap_or_else(|e| debug!("Error updating discovery registration: {:?}", e));
|
|
},
|
|
FAST_DISCOVERY_REFRESH => {
|
|
// Run the fast discovery if not enough peers are connected
|
|
if self.has_enough_peers() {
|
|
return;
|
|
}
|
|
self.discovery.lock().as_mut().map(|d| d.refresh());
|
|
io.update_registration(DISCOVERY).unwrap_or_else(|e| debug!("Error updating discovery registration: {:?}", e));
|
|
},
|
|
DISCOVERY_ROUND => {
|
|
self.discovery.lock().as_mut().map(|d| d.round());
|
|
io.update_registration(DISCOVERY).unwrap_or_else(|e| debug!("Error updating discovery registration: {:?}", e));
|
|
},
|
|
NODE_TABLE => {
|
|
trace!(target: "network", "Refreshing node table");
|
|
self.nodes.write().clear_useless();
|
|
self.nodes.write().save();
|
|
},
|
|
_ => match self.timers.read().get(&token).cloned() {
|
|
Some(timer) => match self.handlers.read().get(&timer.protocol).cloned() {
|
|
None => { warn!(target: "network", "No handler found for protocol: {:?}", timer.protocol) },
|
|
Some(h) => {
|
|
let reserved = self.reserved_nodes.read();
|
|
h.timeout(&NetworkContext::new(io, timer.protocol, None, self.sessions.clone(), &reserved), timer.token);
|
|
}
|
|
},
|
|
None => { warn!("Unknown timer token: {}", token); } // timer is not registerd through us
|
|
}
|
|
}
|
|
}
|
|
|
|
fn message(&self, io: &IoContext<NetworkIoMessage>, message: &NetworkIoMessage) {
|
|
if self.stopping.load(AtomicOrdering::Acquire) {
|
|
return;
|
|
}
|
|
match *message {
|
|
NetworkIoMessage::AddHandler {
|
|
ref handler,
|
|
ref protocol,
|
|
ref versions,
|
|
} => {
|
|
let h = handler.clone();
|
|
let reserved = self.reserved_nodes.read();
|
|
h.initialize(
|
|
&NetworkContext::new(io, *protocol, None, self.sessions.clone(), &reserved),
|
|
);
|
|
self.handlers.write().insert(*protocol, h);
|
|
let mut info = self.info.write();
|
|
for &(version, packet_count) in versions {
|
|
info.capabilities.push(CapabilityInfo {
|
|
protocol: *protocol,
|
|
version,
|
|
packet_count,
|
|
});
|
|
}
|
|
},
|
|
NetworkIoMessage::AddTimer {
|
|
ref protocol,
|
|
ref delay,
|
|
ref token,
|
|
} => {
|
|
let handler_token = {
|
|
let mut timer_counter = self.timer_counter.write();
|
|
let counter = &mut *timer_counter;
|
|
let handler_token = *counter;
|
|
*counter += 1;
|
|
handler_token
|
|
};
|
|
self.timers.write().insert(handler_token, ProtocolTimer { protocol: *protocol, token: *token });
|
|
io.register_timer(handler_token, *delay).unwrap_or_else(|e| debug!("Error registering timer {}: {:?}", token, e));
|
|
},
|
|
NetworkIoMessage::Disconnect(ref peer) => {
|
|
let session = { self.sessions.read().get(*peer).cloned() };
|
|
if let Some(session) = session {
|
|
session.lock().disconnect(io, DisconnectReason::DisconnectRequested);
|
|
}
|
|
trace!(target: "network", "Disconnect requested {}", peer);
|
|
self.kill_connection(*peer, io, false);
|
|
},
|
|
NetworkIoMessage::DisablePeer(ref peer) => {
|
|
let session = { self.sessions.read().get(*peer).cloned() };
|
|
if let Some(session) = session {
|
|
session.lock().disconnect(io, DisconnectReason::DisconnectRequested);
|
|
if let Some(id) = session.lock().id() {
|
|
let mut nodes = self.nodes.write();
|
|
nodes.note_failure(&id);
|
|
nodes.mark_as_useless(id);
|
|
}
|
|
}
|
|
trace!(target: "network", "Disabling peer {}", peer);
|
|
self.kill_connection(*peer, io, false);
|
|
},
|
|
NetworkIoMessage::InitPublicInterface =>
|
|
self.init_public_interface(io).unwrap_or_else(|e| warn!("Error initializing public interface: {:?}", e)),
|
|
_ => {} // ignore others.
|
|
}
|
|
}
|
|
|
|
fn register_stream(&self, stream: StreamToken, reg: Token, event_loop: &mut EventLoop<IoManager<NetworkIoMessage>>) {
|
|
match stream {
|
|
FIRST_SESSION ... LAST_SESSION => {
|
|
let session = { self.sessions.read().get(stream).cloned() };
|
|
if let Some(session) = session {
|
|
session.lock().register_socket(reg, event_loop).expect("Error registering socket");
|
|
}
|
|
}
|
|
DISCOVERY => match self.udp_socket.lock().as_ref() {
|
|
Some(udp_socket) => {
|
|
event_loop.register(udp_socket, reg, Ready::all(), PollOpt::edge())
|
|
.expect("Error registering UDP socket");
|
|
},
|
|
_ => panic!("Error registering discovery socket"),
|
|
}
|
|
TCP_ACCEPT => event_loop.register(&*self.tcp_listener.lock(), Token(TCP_ACCEPT), Ready::all(), PollOpt::edge()).expect("Error registering stream"),
|
|
_ => warn!("Unexpected stream registration")
|
|
}
|
|
}
|
|
|
|
fn deregister_stream(&self, stream: StreamToken, event_loop: &mut EventLoop<IoManager<NetworkIoMessage>>) {
|
|
match stream {
|
|
FIRST_SESSION ... LAST_SESSION => {
|
|
let mut connections = self.sessions.write();
|
|
if let Some(connection) = connections.get(stream).cloned() {
|
|
let c = connection.lock();
|
|
if c.expired() { // make sure it is the same connection that the event was generated for
|
|
c.deregister_socket(event_loop).expect("Error deregistering socket");
|
|
connections.remove(stream);
|
|
}
|
|
}
|
|
}
|
|
DISCOVERY => (),
|
|
_ => warn!("Unexpected stream deregistration")
|
|
}
|
|
}
|
|
|
|
fn update_stream(&self, stream: StreamToken, reg: Token, event_loop: &mut EventLoop<IoManager<NetworkIoMessage>>) {
|
|
match stream {
|
|
FIRST_SESSION ... LAST_SESSION => {
|
|
let connection = { self.sessions.read().get(stream).cloned() };
|
|
if let Some(connection) = connection {
|
|
connection.lock().update_socket(reg, event_loop).expect("Error updating socket");
|
|
}
|
|
}
|
|
DISCOVERY => match (self.udp_socket.lock().as_ref(), self.discovery.lock().as_ref()) {
|
|
(Some(udp_socket), Some(discovery)) => {
|
|
let registration = if discovery.any_sends_queued() {
|
|
Ready::readable() | Ready::writable()
|
|
} else {
|
|
Ready::readable()
|
|
};
|
|
event_loop.reregister(udp_socket, reg, registration, PollOpt::edge())
|
|
.expect("Error reregistering UDP socket");
|
|
},
|
|
_ => panic!("Error reregistering discovery socket"),
|
|
}
|
|
TCP_ACCEPT => event_loop.reregister(&*self.tcp_listener.lock(), Token(TCP_ACCEPT), Ready::all(), PollOpt::edge()).expect("Error reregistering stream"),
|
|
_ => warn!("Unexpected stream update")
|
|
}
|
|
}
|
|
}
|
|
|
|
fn save_key(path: &Path, key: &Secret) {
|
|
let mut path_buf = PathBuf::from(path);
|
|
if let Err(e) = fs::create_dir_all(path_buf.as_path()) {
|
|
warn!("Error creating key directory: {:?}", e);
|
|
return;
|
|
};
|
|
path_buf.push("key");
|
|
let path = path_buf.as_path();
|
|
let mut file = match fs::File::create(&path) {
|
|
Ok(file) => file,
|
|
Err(e) => {
|
|
warn!("Error creating key file: {:?}", e);
|
|
return;
|
|
}
|
|
};
|
|
if let Err(e) = restrict_permissions_owner(path, true, false) {
|
|
warn!(target: "network", "Failed to modify permissions of the file ({})", e);
|
|
}
|
|
if let Err(e) = file.write(&key.hex().into_bytes()[2..]) {
|
|
warn!("Error writing key file: {:?}", e);
|
|
}
|
|
}
|
|
|
|
fn load_key(path: &Path) -> Option<Secret> {
|
|
let mut path_buf = PathBuf::from(path);
|
|
path_buf.push("key");
|
|
let mut file = match fs::File::open(path_buf.as_path()) {
|
|
Ok(file) => file,
|
|
Err(e) => {
|
|
debug!("Error opening key file: {:?}", e);
|
|
return None;
|
|
}
|
|
};
|
|
let mut buf = String::new();
|
|
match file.read_to_string(&mut buf) {
|
|
Ok(_) => {},
|
|
Err(e) => {
|
|
warn!("Error reading key file: {:?}", e);
|
|
return None;
|
|
}
|
|
}
|
|
match Secret::from_str(&buf) {
|
|
Ok(key) => Some(key),
|
|
Err(e) => {
|
|
warn!("Error parsing key file: {:?}", e);
|
|
None
|
|
}
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn key_save_load() {
|
|
use tempdir::TempDir;
|
|
|
|
let tempdir = TempDir::new("").unwrap();
|
|
let key = H256::random().into();
|
|
save_key(tempdir.path(), &key);
|
|
let r = load_key(tempdir.path());
|
|
assert_eq!(key, r.unwrap());
|
|
}
|
|
|
|
#[test]
|
|
fn host_client_url() {
|
|
let mut config = NetworkConfiguration::new_local();
|
|
let key = "6f7b0d801bc7b5ce7bbd930b84fd0369b3eb25d09be58d64ba811091046f3aa2".parse().unwrap();
|
|
config.use_secret = Some(key);
|
|
let host: Host = Host::new(config, None).unwrap();
|
|
assert!(host.local_url().starts_with("enode://101b3ef5a4ea7a1c7928e24c4c75fd053c235d7b80c22ae5c03d145d0ac7396e2a4ffff9adee3133a7b05044a5cee08115fd65145e5165d646bde371010d803c@"));
|
|
}
|