1873 lines
76 KiB
Rust
Raw Normal View History

// Copyright 2019 Parity Technologies (UK) Ltd.
//
// Permission is hereby granted, free of charge, to any person obtaining a
// copy of this software and associated documentation files (the "Software"),
// to deal in the Software without restriction, including without limitation
// the rights to use, copy, modify, merge, publish, distribute, sublicense,
// and/or sell copies of the Software, and to permit persons to whom the
// Software is furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
// DEALINGS IN THE SOFTWARE.
//! High level manager of the network.
//!
//! A [`Swarm`] contains the state of the network as a whole. The entire
//! behaviour of a libp2p network can be controlled through the `Swarm`.
//! The `Swarm` struct contains all active and pending connections to
//! remotes and manages the state of all the substreams that have been
//! opened, and all the upgrades that were built upon these substreams.
//!
//! # Initializing a Swarm
//!
//! Creating a `Swarm` requires three things:
//!
//! 1. A network identity of the local node in form of a [`PeerId`].
//! 2. An implementation of the [`Transport`] trait. This is the type that
//! will be used in order to reach nodes on the network based on their
//! address. See the `transport` module for more information.
//! 3. An implementation of the [`NetworkBehaviour`] trait. This is a state
//! machine that defines how the swarm should behave once it is connected
//! to a node.
//!
//! # Network Behaviour
//!
//! The [`NetworkBehaviour`] trait is implemented on types that indicate to
//! the swarm how it should behave. This includes which protocols are supported
//! and which nodes to try to connect to. It is the `NetworkBehaviour` that
//! controls what happens on the network. Multiple types that implement
//! `NetworkBehaviour` can be composed into a single behaviour.
//!
//! # Protocols Handler
//!
//! The [`ProtocolsHandler`] trait defines how each active connection to a
//! remote should behave: how to handle incoming substreams, which protocols
//! are supported, when to open a new outbound substream, etc.
//!
mod registry;
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
#[cfg(test)]
mod test;
mod upgrade;
pub mod behaviour;
pub mod dial_opts;
pub mod protocols_handler;
pub use behaviour::{
CloseConnection, NetworkBehaviour, NetworkBehaviourAction, NetworkBehaviourEventProcess,
NotifyHandler, PollParameters,
};
pub use protocols_handler::{
IntoProtocolsHandler, IntoProtocolsHandlerSelect, KeepAlive, OneShotHandler,
OneShotHandlerConfig, ProtocolsHandler, ProtocolsHandlerEvent, ProtocolsHandlerSelect,
ProtocolsHandlerUpgrErr, SubstreamProtocol,
};
pub use registry::{AddAddressResult, AddressRecord, AddressScore};
use dial_opts::{DialOpts, PeerCondition};
use futures::{executor::ThreadPoolBuilder, prelude::*, stream::FusedStream};
use libp2p_core::{
connection::{
ConnectedPoint, ConnectionError, ConnectionHandler, ConnectionId, ConnectionLimit,
EstablishedConnection, IntoConnectionHandler, ListenerId, PendingConnectionError,
PendingInboundConnectionError, PendingOutboundConnectionError, Substream,
},
muxing::StreamMuxerBox,
network::{
self, peer::ConnectedPeer, ConnectionLimits, Network, NetworkConfig, NetworkEvent,
NetworkInfo,
},
transport::{self, TransportError},
upgrade::ProtocolName,
Executor, Multiaddr, Negotiated, PeerId, Transport,
};
use protocols_handler::{NodeHandlerWrapperBuilder, NodeHandlerWrapperError};
use registry::{AddressIntoIter, Addresses};
use smallvec::SmallVec;
use std::collections::HashSet;
use std::num::{NonZeroU32, NonZeroU8, NonZeroUsize};
use std::{
convert::TryFrom,
error, fmt, io,
pin::Pin,
task::{Context, Poll},
};
use upgrade::UpgradeInfoSend as _;
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
/// Substream for which a protocol has been chosen.
///
/// Implements the [`AsyncRead`](futures::io::AsyncRead) and
/// [`AsyncWrite`](futures::io::AsyncWrite) traits.
pub type NegotiatedSubstream = Negotiated<Substream<StreamMuxerBox>>;
/// Event generated by the [`NetworkBehaviour`] that the swarm will report back.
type TBehaviourOutEvent<TBehaviour> = <TBehaviour as NetworkBehaviour>::OutEvent;
/// [`ProtocolsHandler`] of the [`NetworkBehaviour`] for all the protocols the [`NetworkBehaviour`]
/// supports.
type THandler<TBehaviour> = <TBehaviour as NetworkBehaviour>::ProtocolsHandler;
/// Custom event that can be received by the [`ProtocolsHandler`] of the
/// [`NetworkBehaviour`].
type THandlerInEvent<TBehaviour> =
<<THandler<TBehaviour> as IntoProtocolsHandler>::Handler as ProtocolsHandler>::InEvent;
/// Custom event that can be produced by the [`ProtocolsHandler`] of the [`NetworkBehaviour`].
type THandlerOutEvent<TBehaviour> =
<<THandler<TBehaviour> as IntoProtocolsHandler>::Handler as ProtocolsHandler>::OutEvent;
/// Custom error that can be produced by the [`ProtocolsHandler`] of the [`NetworkBehaviour`].
type THandlerErr<TBehaviour> =
<<THandler<TBehaviour> as IntoProtocolsHandler>::Handler as ProtocolsHandler>::Error;
/// Event generated by the `Swarm`.
#[derive(Debug)]
pub enum SwarmEvent<TBehaviourOutEvent, THandlerErr> {
/// Event generated by the `NetworkBehaviour`.
Behaviour(TBehaviourOutEvent),
/// A connection to the given peer has been opened.
ConnectionEstablished {
/// Identity of the peer that we have connected to.
peer_id: PeerId,
/// Endpoint of the connection that has been opened.
endpoint: ConnectedPoint,
/// Number of established connections to this peer, including the one that has just been
/// opened.
num_established: NonZeroU32,
/// [`Some`] when the new connection is an outgoing connection.
/// Addresses are dialed concurrently. Contains the addresses and errors
/// of dial attempts that failed before the one successful dial.
concurrent_dial_errors: Option<Vec<(Multiaddr, TransportError<io::Error>)>>,
},
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
/// A connection with the given peer has been closed,
/// possibly as a result of an error.
ConnectionClosed {
/// Identity of the peer that we have connected to.
peer_id: PeerId,
/// Endpoint of the connection that has been closed.
endpoint: ConnectedPoint,
/// Number of other remaining connections to this same peer.
num_established: u32,
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
/// Reason for the disconnection, if it was not a successful
/// active close.
cause: Option<ConnectionError<NodeHandlerWrapperError<THandlerErr>>>,
},
/// A new connection arrived on a listener and is in the process of protocol negotiation.
///
/// A corresponding [`ConnectionEstablished`](SwarmEvent::ConnectionEstablished),
/// [`BannedPeer`](SwarmEvent::BannedPeer), or
/// [`IncomingConnectionError`](SwarmEvent::IncomingConnectionError) event will later be
/// generated for this connection.
IncomingConnection {
/// Local connection address.
/// This address has been earlier reported with a [`NewListenAddr`](SwarmEvent::NewListenAddr)
/// event.
local_addr: Multiaddr,
/// Address used to send back data to the remote.
send_back_addr: Multiaddr,
},
/// An error happened on a connection during its initial handshake.
///
/// This can include, for example, an error during the handshake of the encryption layer, or
/// the connection unexpectedly closed.
IncomingConnectionError {
/// Local connection address.
/// This address has been earlier reported with a [`NewListenAddr`](SwarmEvent::NewListenAddr)
/// event.
local_addr: Multiaddr,
/// Address used to send back data to the remote.
send_back_addr: Multiaddr,
/// The error that happened.
error: PendingInboundConnectionError<io::Error>,
},
/// Outgoing connection attempt failed.
OutgoingConnectionError {
/// If known, [`PeerId`] of the peer we tried to reach.
peer_id: Option<PeerId>,
/// Error that has been encountered.
error: DialError,
},
/// We connected to a peer, but we immediately closed the connection because that peer is banned.
BannedPeer {
/// Identity of the banned peer.
peer_id: PeerId,
/// Endpoint of the connection that has been closed.
endpoint: ConnectedPoint,
},
/// One of our listeners has reported a new local listening address.
NewListenAddr {
/// The listener that is listening on the new address.
listener_id: ListenerId,
/// The new address that is being listened on.
address: Multiaddr,
},
/// One of our listeners has reported the expiration of a listening address.
ExpiredListenAddr {
/// The listener that is no longer listening on the address.
listener_id: ListenerId,
/// The expired address.
address: Multiaddr,
},
/// One of the listeners gracefully closed.
ListenerClosed {
/// The listener that closed.
listener_id: ListenerId,
/// The addresses that the listener was listening on. These addresses are now considered
/// expired, similar to if a [`ExpiredListenAddr`](SwarmEvent::ExpiredListenAddr) event
/// has been generated for each of them.
addresses: Vec<Multiaddr>,
/// Reason for the closure. Contains `Ok(())` if the stream produced `None`, or `Err`
/// if the stream produced an error.
reason: Result<(), io::Error>,
},
/// One of the listeners reported a non-fatal error.
ListenerError {
/// The listener that errored.
listener_id: ListenerId,
/// The listener error.
error: io::Error,
},
/// A new dialing attempt has been initiated by the [`NetworkBehaviour`]
/// implementation.
///
/// A [`ConnectionEstablished`](SwarmEvent::ConnectionEstablished) event is
/// reported if the dialing attempt succeeds, otherwise a
/// [`OutgoingConnectionError`](SwarmEvent::OutgoingConnectionError) event
/// is reported.
Dialing(PeerId),
}
2019-04-07 18:34:14 -03:00
/// Contains the state of the network, plus the way it should behave.
///
/// Note: Needs to be polled via `<Swarm as Stream>` in order to make
/// progress.
pub struct Swarm<TBehaviour>
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
where
TBehaviour: NetworkBehaviour,
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
{
network: Network<
transport::Boxed<(PeerId, StreamMuxerBox)>,
NodeHandlerWrapperBuilder<THandler<TBehaviour>>,
>,
/// Handles which nodes to connect to and how to handle the events sent back by the protocol
/// handlers.
behaviour: TBehaviour,
/// List of protocols that the behaviour says it supports.
supported_protocols: SmallVec<[Vec<u8>; 16]>,
/// List of multiaddresses we're listening on.
listened_addrs: SmallVec<[Multiaddr; 8]>,
/// List of multiaddresses we're listening on, after account for external IP addresses and
/// similar mechanisms.
external_addrs: Addresses,
/// List of nodes for which we deny any incoming connection.
banned_peers: HashSet<PeerId>,
Replace unbounded channels with bounded ones. (#1191) * Replace unbounded channels with bounded ones. To remove the unbounded channels used for communicating with node tasks an API similar to `futures::Sink` is used, i.e. sending is split into a start and complete phase. The start phase returns `StartSend` and first attempts to complete any pending send operations. Completing the send means polling until `Poll::Ready(())` is returned. In addition this PR has split the `handled_node_tasks` module into several smaller ones (cf. `nodes::tasks`) and renamed some types: - `nodes::handled_node_tasks::NodeTask` -> `nodes::tasks::task::Task` - `nodes::handled_node_tasks::NodeTaskInner` -> `nodes::tasks::task::State` - `nodes::handled_node_tasks::NodeTasks` -> `nodes::tasks::Manager` - `nodes::handled_node_tasks::TaskClosedEvent` -> `nodes::tasks::Error` - `nodes::handled_node_tasks::HandledNodesEvent` -> `nodes::tasks::Event` - `nodes::handled_node_tasks::Task` -> `nodes::tasks::TaskEntry` - `nodes::handled_node_tasks::ExtToInMessage` -> `nodes::tasks::task::ToTaskMessage` - `nodes::handled_node_tasks::InToExtMessage` -> `nodes::tasks::task::FromTaskMessage` * `take_over_to_complete` can be an `Option`. Since it is always holding just a single pending message. * `send_event_to_complete` can be an `Option`. * Update core/src/nodes/tasks/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/nodes/tasks/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Add comments to explain the need to flush sends ... of take-over and event messages delivered over Sinks.
2019-07-09 16:47:24 +02:00
/// Connections for which we withhold any reporting. These belong to banned peers.
///
/// Note: Connections to a peer that are established at the time of banning that peer
/// are not added here. Instead they are simply closed.
banned_peer_connections: HashSet<ConnectionId>,
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
/// Pending event to be delivered to connection handlers
/// (or dropped if the peer disconnected) before the `behaviour`
/// can be polled again.
pending_event: Option<(PeerId, PendingNotifyHandler, THandlerInEvent<TBehaviour>)>,
/// The configured override for substream protocol upgrades, if any.
substream_upgrade_protocol_override: Option<libp2p_core::upgrade::Version>,
}
impl<TBehaviour> Unpin for Swarm<TBehaviour> where TBehaviour: NetworkBehaviour {}
impl<TBehaviour> Swarm<TBehaviour>
where
TBehaviour: NetworkBehaviour,
{
/// Builds a new `Swarm`.
pub fn new(
transport: transport::Boxed<(PeerId, StreamMuxerBox)>,
behaviour: TBehaviour,
local_peer_id: PeerId,
) -> Self {
SwarmBuilder::new(transport, behaviour, local_peer_id).build()
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
/// Returns information about the [`Network`] underlying the `Swarm`.
pub fn network_info(&self) -> NetworkInfo {
self.network.info()
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
/// Starts listening on the given address.
/// Returns an error if the address is not supported.
///
/// Listeners report their new listening addresses as [`SwarmEvent::NewListenAddr`].
/// Depending on the underlying transport, one listener may have multiple listening addresses.
pub fn listen_on(&mut self, addr: Multiaddr) -> Result<ListenerId, TransportError<io::Error>> {
let id = self.network.listen_on(addr)?;
self.behaviour.inject_new_listener(id);
Ok(id)
}
/// Remove some listener.
///
/// Returns `true` if there was a listener with this ID, `false`
/// otherwise.
pub fn remove_listener(&mut self, id: ListenerId) -> bool {
self.network.remove_listener(id)
}
/// Dial a known or unknown peer.
///
/// See also [`DialOpts`].
///
/// ```
/// # use libp2p_swarm::Swarm;
/// # use libp2p_swarm::dial_opts::{DialOpts, PeerCondition};
/// # use libp2p_core::{Multiaddr, PeerId, Transport};
/// # use libp2p_core::transport::dummy::DummyTransport;
/// # use libp2p_swarm::DummyBehaviour;
/// #
/// let mut swarm = Swarm::new(
/// DummyTransport::new().boxed(),
/// DummyBehaviour::default(),
/// PeerId::random(),
/// );
///
/// // Dial a known peer.
/// swarm.dial(PeerId::random());
///
/// // Dial an unknown peer.
/// swarm.dial("/ip6/::1/tcp/12345".parse::<Multiaddr>().unwrap());
/// ```
pub fn dial(&mut self, opts: impl Into<DialOpts>) -> Result<(), DialError> {
let handler = self.behaviour.new_handler();
self.dial_with_handler(opts.into(), handler)
}
fn dial_with_handler(
&mut self,
swarm_dial_opts: DialOpts,
handler: <TBehaviour as NetworkBehaviour>::ProtocolsHandler,
) -> Result<(), DialError> {
let core_dial_opts = match swarm_dial_opts.0 {
// Dial a known peer.
dial_opts::Opts::WithPeerId(dial_opts::WithPeerId {
peer_id,
condition,
dial_concurrency_factor_override,
})
| dial_opts::Opts::WithPeerIdWithAddresses(dial_opts::WithPeerIdWithAddresses {
peer_id,
condition,
dial_concurrency_factor_override,
..
}) => {
// Check [`PeerCondition`] if provided.
let condition_matched = match condition {
PeerCondition::Disconnected => self.network.is_disconnected(&peer_id),
PeerCondition::NotDialing => !self.network.is_dialing(&peer_id),
PeerCondition::Always => true,
};
if !condition_matched {
self.behaviour.inject_dial_failure(
Some(peer_id),
handler,
&DialError::DialPeerConditionFalse(condition),
);
return Err(DialError::DialPeerConditionFalse(condition));
}
// Check if peer is banned.
if self.banned_peers.contains(&peer_id) {
let error = DialError::Banned;
self.behaviour
.inject_dial_failure(Some(peer_id), handler, &error);
return Err(error);
}
// Retrieve the addresses to dial.
let addresses = {
let mut addresses = match swarm_dial_opts.0 {
dial_opts::Opts::WithPeerId(dial_opts::WithPeerId { .. }) => {
self.behaviour.addresses_of_peer(&peer_id)
}
dial_opts::Opts::WithPeerIdWithAddresses(
dial_opts::WithPeerIdWithAddresses {
peer_id,
mut addresses,
extend_addresses_through_behaviour,
..
},
) => {
if extend_addresses_through_behaviour {
addresses.extend(self.behaviour.addresses_of_peer(&peer_id))
}
addresses
}
dial_opts::Opts::WithoutPeerIdWithAddress { .. } => {
unreachable!("Due to outer match.")
}
};
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
let mut unique_addresses = HashSet::new();
addresses.retain(|a| {
!self.listened_addrs.contains(a) && unique_addresses.insert(a.clone())
});
if addresses.is_empty() {
let error = DialError::NoAddresses;
self.behaviour
.inject_dial_failure(Some(peer_id), handler, &error);
return Err(error);
};
addresses
};
let mut opts = libp2p_core::DialOpts::peer_id(peer_id).addresses(addresses);
if let Some(f) = dial_concurrency_factor_override {
opts = opts.override_dial_concurrency_factor(f);
}
opts.build()
}
// Dial an unknown peer.
dial_opts::Opts::WithoutPeerIdWithAddress(dial_opts::WithoutPeerIdWithAddress {
address,
}) => libp2p_core::DialOpts::unknown_peer_id()
.address(address)
.build(),
};
let handler = handler
.into_node_handler_builder()
.with_substream_upgrade_protocol_override(self.substream_upgrade_protocol_override);
match self.network.dial(handler, core_dial_opts).map(|_id| ()) {
Ok(_connection_id) => Ok(()),
Err(error) => {
let (error, handler) = DialError::from_network_dial_error(error);
self.behaviour
.inject_dial_failure(None, handler.into_protocols_handler(), &error);
return Err(error);
}
}
}
/// Returns an iterator that produces the list of addresses we're listening on.
pub fn listeners(&self) -> impl Iterator<Item = &Multiaddr> {
self.network.listen_addrs()
}
/// Returns the peer ID of the swarm passed as parameter.
pub fn local_peer_id(&self) -> &PeerId {
self.network.local_peer_id()
}
/// Returns an iterator for [`AddressRecord`]s of external addresses
/// of the local node, in decreasing order of their current
/// [score](AddressScore).
pub fn external_addresses(&self) -> impl Iterator<Item = &AddressRecord> {
self.external_addrs.iter()
}
/// Adds an external address record for the local node.
///
/// An external address is an address of the local node known to
/// be (likely) reachable for other nodes, possibly taking into
/// account NAT. The external addresses of the local node may be
/// shared with other nodes by the `NetworkBehaviour`.
///
/// The associated score determines both the position of the address
/// in the list of external addresses (which can determine the
/// order in which addresses are used to connect to) as well as
/// how long the address is retained in the list, depending on
/// how frequently it is reported by the `NetworkBehaviour` via
/// [`NetworkBehaviourAction::ReportObservedAddr`] or explicitly
/// through this method.
pub fn add_external_address(&mut self, a: Multiaddr, s: AddressScore) -> AddAddressResult {
let result = self.external_addrs.add(a.clone(), s);
let expired = match &result {
AddAddressResult::Inserted { expired } => {
self.behaviour.inject_new_external_addr(&a);
expired
}
AddAddressResult::Updated { expired } => expired,
};
for a in expired {
self.behaviour.inject_expired_external_addr(&a.addr);
}
result
}
/// Removes an external address of the local node, regardless of
/// its current score. See [`Swarm::add_external_address`]
/// for details.
///
/// Returns `true` if the address existed and was removed, `false`
/// otherwise.
pub fn remove_external_address(&mut self, addr: &Multiaddr) -> bool {
if self.external_addrs.remove(addr) {
self.behaviour.inject_expired_external_addr(addr);
true
} else {
false
}
}
/// Bans a peer by its peer ID.
///
/// Any incoming connection and any dialing attempt will immediately be rejected.
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
/// This function has no effect if the peer is already banned.
pub fn ban_peer_id(&mut self, peer_id: PeerId) {
if self.banned_peers.insert(peer_id) {
if let Some(peer) = self.network.peer(peer_id).into_connected() {
// Note that established connections to the now banned peer are closed but not
// added to [`Swarm::banned_peer_connections`]. They have been previously reported
// as open to the behaviour and need be reported as closed once closing the
// connection finishes.
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
peer.disconnect();
}
2019-04-23 10:54:25 +02:00
}
}
/// Unbans a peer.
pub fn unban_peer_id(&mut self, peer_id: PeerId) {
self.banned_peers.remove(&peer_id);
}
/// Disconnects a peer by its peer ID, closing all connections to said peer.
///
/// Returns `Ok(())` if there was one or more established connections to the peer.
///
/// Note: Closing a connection via [`Swarm::disconnect_peer_id`] does
/// not inform the corresponding [`ProtocolsHandler`].
/// Closing a connection via a [`ProtocolsHandler`] can be done either in a
/// collaborative manner across [`ProtocolsHandler`]s
/// with [`ProtocolsHandler::connection_keep_alive`] or directly with
/// [`ProtocolsHandlerEvent::Close`].
pub fn disconnect_peer_id(&mut self, peer_id: PeerId) -> Result<(), ()> {
if let Some(peer) = self.network.peer(peer_id).into_connected() {
peer.disconnect();
return Ok(());
}
Err(())
}
/// Checks whether the [`Network`] has an established connection to a peer.
pub fn is_connected(&self, peer_id: &PeerId) -> bool {
self.network.is_connected(peer_id)
}
/// Returns the currently connected peers.
pub fn connected_peers(&self) -> impl Iterator<Item = &PeerId> {
self.network.connected_peers()
}
/// Returns a reference to the provided [`NetworkBehaviour`].
pub fn behaviour(&self) -> &TBehaviour {
&self.behaviour
}
/// Returns a mutable reference to the provided [`NetworkBehaviour`].
pub fn behaviour_mut(&mut self) -> &mut TBehaviour {
&mut self.behaviour
}
/// Internal function used by everything event-related.
///
/// Polls the `Swarm` for the next event.
fn poll_next_event(
mut self: Pin<&mut Self>,
cx: &mut Context<'_>,
) -> Poll<SwarmEvent<TBehaviour::OutEvent, THandlerErr<TBehaviour>>> {
2019-09-20 10:46:13 +02:00
// We use a `this` variable because the compiler can't mutably borrow multiple times
// across a `Deref`.
let this = &mut *self;
loop {
let mut network_not_ready = false;
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
// First let the network make progress.
match this.network.poll(cx) {
Poll::Pending => network_not_ready = true,
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
Poll::Ready(NetworkEvent::ConnectionEvent { connection, event }) => {
let peer = connection.peer_id();
let conn_id = connection.id();
if this.banned_peer_connections.contains(&conn_id) {
log::debug!("Ignoring event from banned peer: {} {:?}.", peer, conn_id);
} else {
this.behaviour.inject_event(peer, conn_id, event);
}
}
Poll::Ready(NetworkEvent::AddressChange {
connection,
new_endpoint,
old_endpoint,
}) => {
let peer = connection.peer_id();
let conn_id = connection.id();
if !this.banned_peer_connections.contains(&conn_id) {
this.behaviour.inject_address_change(
&peer,
&conn_id,
&old_endpoint,
&new_endpoint,
);
}
}
Poll::Ready(NetworkEvent::ConnectionEstablished {
connection,
other_established_connection_ids,
concurrent_dial_errors,
}) => {
let peer_id = connection.peer_id();
let endpoint = connection.endpoint().clone();
if this.banned_peers.contains(&peer_id) {
// Mark the connection for the banned peer as banned, thus withholding any
// future events from the connection to the behaviour.
this.banned_peer_connections.insert(connection.id());
this.network
.peer(peer_id)
.into_connected()
.expect("the Network just notified us that we were connected; QED")
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
.disconnect();
return Poll::Ready(SwarmEvent::BannedPeer { peer_id, endpoint });
} else {
let num_established = NonZeroU32::new(
u32::try_from(other_established_connection_ids.len() + 1).unwrap(),
)
.expect("n + 1 is always non-zero; qed");
log::debug!(
"Connection established: {:?} {:?}; Total (peer): {}.",
connection.peer_id(),
connection.endpoint(),
num_established
);
let endpoint = connection.endpoint().clone();
let failed_addresses = concurrent_dial_errors
.as_ref()
.map(|es| es.iter().map(|(a, _)| a).cloned().collect());
this.behaviour.inject_connection_established(
&peer_id,
&connection.id(),
&endpoint,
failed_addresses.as_ref(),
);
// The peer is not banned, but there could be previous banned connections
// if the peer was just unbanned. Check if this is the first non-banned
// connection.
let first_non_banned = other_established_connection_ids
.into_iter()
.all(|conn_id| this.banned_peer_connections.contains(&conn_id));
if first_non_banned {
this.behaviour.inject_connected(&peer_id);
}
return Poll::Ready(SwarmEvent::ConnectionEstablished {
peer_id,
num_established,
endpoint,
concurrent_dial_errors,
});
}
}
Poll::Ready(NetworkEvent::ConnectionClosed {
id,
connected,
error,
remaining_established_connection_ids,
handler,
}) => {
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
if let Some(error) = error.as_ref() {
log::debug!(
"Connection closed with error {:?}: {:?}; Total (peer): {}.",
error,
connected,
remaining_established_connection_ids.len()
);
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
} else {
log::debug!(
"Connection closed: {:?}; Total (peer): {}.",
connected,
remaining_established_connection_ids.len()
);
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
let peer_id = connected.peer_id;
let endpoint = connected.endpoint;
let num_established =
u32::try_from(remaining_established_connection_ids.len()).unwrap();
let conn_was_reported = !this.banned_peer_connections.remove(&id);
if conn_was_reported {
this.behaviour.inject_connection_closed(
&peer_id,
&id,
&endpoint,
handler.into_protocols_handler(),
);
// This connection was reported as open to the behaviour. Check if this is
// the last non-banned connection for the peer.
let last_non_banned = remaining_established_connection_ids
.into_iter()
.all(|conn_id| this.banned_peer_connections.contains(&conn_id));
if last_non_banned {
this.behaviour.inject_disconnected(&peer_id)
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
return Poll::Ready(SwarmEvent::ConnectionClosed {
peer_id,
endpoint,
cause: error,
num_established,
});
}
Poll::Ready(NetworkEvent::IncomingConnection { connection, .. }) => {
let handler = this
.behaviour
.new_handler()
.into_node_handler_builder()
.with_substream_upgrade_protocol_override(
this.substream_upgrade_protocol_override,
);
let local_addr = connection.local_addr.clone();
let send_back_addr = connection.send_back_addr.clone();
match this.network.accept(connection, handler) {
Ok(_connection_id) => {
return Poll::Ready(SwarmEvent::IncomingConnection {
local_addr,
send_back_addr,
});
}
Err((connection_limit, handler)) => {
this.behaviour.inject_listen_failure(
&local_addr,
&send_back_addr,
handler.into_protocols_handler(),
);
log::warn!("Incoming connection rejected: {:?}", connection_limit);
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
}
Poll::Ready(NetworkEvent::NewListenerAddress {
listener_id,
listen_addr,
}) => {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
log::debug!("Listener {:?}; New address: {:?}", listener_id, listen_addr);
if !this.listened_addrs.contains(&listen_addr) {
this.listened_addrs.push(listen_addr.clone())
}
this.behaviour
.inject_new_listen_addr(listener_id, &listen_addr);
return Poll::Ready(SwarmEvent::NewListenAddr {
listener_id,
address: listen_addr,
});
}
Poll::Ready(NetworkEvent::ExpiredListenerAddress {
listener_id,
listen_addr,
}) => {
log::debug!(
"Listener {:?}; Expired address {:?}.",
listener_id,
listen_addr
);
this.listened_addrs.retain(|a| a != &listen_addr);
this.behaviour
.inject_expired_listen_addr(listener_id, &listen_addr);
return Poll::Ready(SwarmEvent::ExpiredListenAddr {
listener_id,
address: listen_addr,
});
}
Poll::Ready(NetworkEvent::ListenerClosed {
listener_id,
addresses,
reason,
}) => {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
log::debug!("Listener {:?}; Closed by {:?}.", listener_id, reason);
for addr in addresses.iter() {
this.behaviour.inject_expired_listen_addr(listener_id, addr);
}
this.behaviour.inject_listener_closed(
listener_id,
match &reason {
Ok(()) => Ok(()),
Err(err) => Err(err),
},
);
return Poll::Ready(SwarmEvent::ListenerClosed {
listener_id,
addresses,
reason,
});
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
Poll::Ready(NetworkEvent::ListenerError { listener_id, error }) => {
this.behaviour.inject_listener_error(listener_id, &error);
return Poll::Ready(SwarmEvent::ListenerError { listener_id, error });
}
Poll::Ready(NetworkEvent::IncomingConnectionError {
local_addr,
send_back_addr,
error,
handler,
}) => {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
log::debug!("Incoming connection failed: {:?}", error);
this.behaviour.inject_listen_failure(
&local_addr,
&send_back_addr,
handler.into_protocols_handler(),
);
return Poll::Ready(SwarmEvent::IncomingConnectionError {
local_addr,
send_back_addr,
error,
});
}
Poll::Ready(NetworkEvent::DialError {
peer_id,
error,
handler,
}) => {
let error = error.into();
this.behaviour.inject_dial_failure(
Some(peer_id),
handler.into_protocols_handler(),
&error,
);
log::debug!(
"Connection attempt to {:?} failed with {:?}.",
peer_id,
error,
);
return Poll::Ready(SwarmEvent::OutgoingConnectionError {
peer_id: Some(peer_id),
error,
});
}
Poll::Ready(NetworkEvent::UnknownPeerDialError { error, handler }) => {
log::debug!("Connection attempt to unknown peer failed with {:?}", error);
let error = error.into();
this.behaviour.inject_dial_failure(
None,
handler.into_protocols_handler(),
&error,
);
return Poll::Ready(SwarmEvent::OutgoingConnectionError {
peer_id: None,
error: error,
});
}
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
// After the network had a chance to make progress, try to deliver
// the pending event emitted by the behaviour in the previous iteration
// to the connection handler(s). The pending event must be delivered
// before polling the behaviour again. If the targeted peer
// meanwhie disconnected, the event is discarded.
if let Some((peer_id, handler, event)) = this.pending_event.take() {
if let Some(mut peer) = this.network.peer(peer_id).into_connected() {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
match handler {
PendingNotifyHandler::One(conn_id) => {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
if let Some(mut conn) = peer.connection(conn_id) {
if let Some(event) = notify_one(&mut conn, event, cx) {
this.pending_event = Some((peer_id, handler, event));
if network_not_ready {
return Poll::Pending;
} else {
continue;
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
}
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
PendingNotifyHandler::Any(ids) => {
if let Some((event, ids)) =
notify_any::<_, _, TBehaviour>(ids, &mut peer, event, cx)
{
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
let handler = PendingNotifyHandler::Any(ids);
this.pending_event = Some((peer_id, handler, event));
if network_not_ready {
return Poll::Pending;
} else {
continue;
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
}
Replace unbounded channels with bounded ones. (#1191) * Replace unbounded channels with bounded ones. To remove the unbounded channels used for communicating with node tasks an API similar to `futures::Sink` is used, i.e. sending is split into a start and complete phase. The start phase returns `StartSend` and first attempts to complete any pending send operations. Completing the send means polling until `Poll::Ready(())` is returned. In addition this PR has split the `handled_node_tasks` module into several smaller ones (cf. `nodes::tasks`) and renamed some types: - `nodes::handled_node_tasks::NodeTask` -> `nodes::tasks::task::Task` - `nodes::handled_node_tasks::NodeTaskInner` -> `nodes::tasks::task::State` - `nodes::handled_node_tasks::NodeTasks` -> `nodes::tasks::Manager` - `nodes::handled_node_tasks::TaskClosedEvent` -> `nodes::tasks::Error` - `nodes::handled_node_tasks::HandledNodesEvent` -> `nodes::tasks::Event` - `nodes::handled_node_tasks::Task` -> `nodes::tasks::TaskEntry` - `nodes::handled_node_tasks::ExtToInMessage` -> `nodes::tasks::task::ToTaskMessage` - `nodes::handled_node_tasks::InToExtMessage` -> `nodes::tasks::task::FromTaskMessage` * `take_over_to_complete` can be an `Option`. Since it is always holding just a single pending message. * `send_event_to_complete` can be an `Option`. * Update core/src/nodes/tasks/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/nodes/tasks/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Add comments to explain the need to flush sends ... of take-over and event messages delivered over Sinks.
2019-07-09 16:47:24 +02:00
}
}
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
debug_assert!(this.pending_event.is_none());
let behaviour_poll = {
let mut parameters = SwarmPollParameters {
local_peer_id: &mut this.network.local_peer_id(),
supported_protocols: &this.supported_protocols,
listened_addrs: &this.listened_addrs,
external_addrs: &this.external_addrs,
};
this.behaviour.poll(cx, &mut parameters)
};
match behaviour_poll {
Poll::Pending if network_not_ready => return Poll::Pending,
Poll::Pending => (),
Poll::Ready(NetworkBehaviourAction::GenerateEvent(event)) => {
return Poll::Ready(SwarmEvent::Behaviour(event))
}
Poll::Ready(NetworkBehaviourAction::Dial { opts, handler }) => {
let peer_id = opts.get_peer_id();
if let Ok(()) = this.dial_with_handler(opts, handler) {
if let Some(peer_id) = peer_id {
return Poll::Ready(SwarmEvent::Dialing(peer_id));
}
}
}
Poll::Ready(NetworkBehaviourAction::NotifyHandler {
peer_id,
handler,
event,
}) => {
if let Some(mut peer) = this.network.peer(peer_id).into_connected() {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
match handler {
NotifyHandler::One(connection) => {
if let Some(mut conn) = peer.connection(connection) {
if let Some(event) = notify_one(&mut conn, event, cx) {
let handler = PendingNotifyHandler::One(connection);
this.pending_event = Some((peer_id, handler, event));
if network_not_ready {
return Poll::Pending;
} else {
continue;
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
}
}
NotifyHandler::Any => {
let ids = peer.connections().into_ids().collect();
if let Some((event, ids)) =
notify_any::<_, _, TBehaviour>(ids, &mut peer, event, cx)
{
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
let handler = PendingNotifyHandler::Any(ids);
this.pending_event = Some((peer_id, handler, event));
if network_not_ready {
return Poll::Pending;
} else {
continue;
}
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
}
Replace unbounded channels with bounded ones. (#1191) * Replace unbounded channels with bounded ones. To remove the unbounded channels used for communicating with node tasks an API similar to `futures::Sink` is used, i.e. sending is split into a start and complete phase. The start phase returns `StartSend` and first attempts to complete any pending send operations. Completing the send means polling until `Poll::Ready(())` is returned. In addition this PR has split the `handled_node_tasks` module into several smaller ones (cf. `nodes::tasks`) and renamed some types: - `nodes::handled_node_tasks::NodeTask` -> `nodes::tasks::task::Task` - `nodes::handled_node_tasks::NodeTaskInner` -> `nodes::tasks::task::State` - `nodes::handled_node_tasks::NodeTasks` -> `nodes::tasks::Manager` - `nodes::handled_node_tasks::TaskClosedEvent` -> `nodes::tasks::Error` - `nodes::handled_node_tasks::HandledNodesEvent` -> `nodes::tasks::Event` - `nodes::handled_node_tasks::Task` -> `nodes::tasks::TaskEntry` - `nodes::handled_node_tasks::ExtToInMessage` -> `nodes::tasks::task::ToTaskMessage` - `nodes::handled_node_tasks::InToExtMessage` -> `nodes::tasks::task::FromTaskMessage` * `take_over_to_complete` can be an `Option`. Since it is always holding just a single pending message. * `send_event_to_complete` can be an `Option`. * Update core/src/nodes/tasks/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/nodes/tasks/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Add comments to explain the need to flush sends ... of take-over and event messages delivered over Sinks.
2019-07-09 16:47:24 +02:00
}
}
}
Poll::Ready(NetworkBehaviourAction::ReportObservedAddr { address, score }) => {
for addr in this.network.address_translation(&address) {
this.add_external_address(addr, score);
}
}
Poll::Ready(NetworkBehaviourAction::CloseConnection {
peer_id,
connection,
}) => {
if let Some(mut peer) = this.network.peer(peer_id).into_connected() {
match connection {
CloseConnection::One(connection_id) => {
if let Some(conn) = peer.connection(connection_id) {
conn.start_close();
}
}
CloseConnection::All => {
peer.disconnect();
}
}
}
}
}
}
}
}
/// Connection to notify of a pending event.
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
///
/// The connection IDs out of which to notify one of an event are captured at
/// the time the behaviour emits the event, in order not to forward the event to
/// a new connection which the behaviour may not have been aware of at the time
/// it issued the request for sending it.
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
enum PendingNotifyHandler {
One(ConnectionId),
Any(SmallVec<[ConnectionId; 10]>),
}
/// Notify a single connection of an event.
///
/// Returns `Some` with the given event if the connection is not currently
/// ready to receive another event, in which case the current task is
/// scheduled to be woken up.
///
/// Returns `None` if the connection is closing or the event has been
/// successfully sent, in either case the event is consumed.
fn notify_one<'a, THandlerInEvent>(
conn: &mut EstablishedConnection<'a, THandlerInEvent>,
event: THandlerInEvent,
2020-07-27 20:27:33 +00:00
cx: &mut Context<'_>,
) -> Option<THandlerInEvent> {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
match conn.poll_ready_notify_handler(cx) {
Poll::Pending => Some(event),
Poll::Ready(Err(())) => None, // connection is closing
Poll::Ready(Ok(())) => {
// Can now only fail if connection is closing.
let _ = conn.notify_handler(event);
None
}
}
}
/// Notify any one of a given list of connections of a peer of an event.
///
/// Returns `Some` with the given event and a new list of connections if
/// none of the given connections was able to receive the event but at
/// least one of them is not closing, in which case the current task
/// is scheduled to be woken up. The returned connections are those which
/// may still become ready to receive another event.
///
/// Returns `None` if either all connections are closing or the event
/// was successfully sent to a handler, in either case the event is consumed.
fn notify_any<'a, TTrans, THandler, TBehaviour>(
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
ids: SmallVec<[ConnectionId; 10]>,
peer: &mut ConnectedPeer<'a, TTrans, THandler>,
event: THandlerInEvent<TBehaviour>,
2020-07-27 20:27:33 +00:00
cx: &mut Context<'_>,
) -> Option<(THandlerInEvent<TBehaviour>, SmallVec<[ConnectionId; 10]>)>
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
where
TTrans: Transport,
TTrans::Error: Send + 'static,
TBehaviour: NetworkBehaviour,
THandler: IntoConnectionHandler,
THandler::Handler: ConnectionHandler<
InEvent = THandlerInEvent<TBehaviour>,
OutEvent = THandlerOutEvent<TBehaviour>,
>,
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
{
let mut pending = SmallVec::new();
let mut event = Some(event); // (1)
for id in ids.into_iter() {
if let Some(mut conn) = peer.connection(id) {
match conn.poll_ready_notify_handler(cx) {
Poll::Pending => pending.push(id),
Poll::Ready(Err(())) => {} // connection is closing
Poll::Ready(Ok(())) => {
let e = event.take().expect("by (1),(2)");
if let Err(e) = conn.notify_handler(e) {
event = Some(e) // (2)
} else {
break;
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
}
}
}
}
event.and_then(|e| {
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
if !pending.is_empty() {
Some((e, pending))
} else {
None
}
})
Multiple connections per peer (#1440) * Allow multiple connections per peer in libp2p-core. Instead of trying to enforce a single connection per peer, which involves quite a bit of additional complexity e.g. to prioritise simultaneously opened connections and can have other undesirable consequences [1], we now make multiple connections per peer a feature. The gist of these changes is as follows: The concept of a "node" with an implicit 1-1 correspondence to a connection has been replaced with the "first-class" concept of a "connection". The code from `src/nodes` has moved (with varying degrees of modification) to `src/connection`. A `HandledNode` has become a `Connection`, a `NodeHandler` a `ConnectionHandler`, the `CollectionStream` was the basis for the new `connection::Pool`, and so forth. Conceptually, a `Network` contains a `connection::Pool` which in turn internally employs the `connection::Manager` for handling the background `connection::manager::Task`s, one per connection, as before. These are all considered implementation details. On the public API, `Peer`s are managed as before through the `Network`, except now the API has changed with the shift of focus to (potentially multiple) connections per peer. The `NetworkEvent`s have accordingly also undergone changes. The Swarm APIs remain largely unchanged, except for the fact that `inject_replaced` is no longer called. It may now practically happen that multiple `ProtocolsHandler`s are associated with a single `NetworkBehaviour`, one per connection. If implementations of `NetworkBehaviour` rely somehow on communicating with exactly one `ProtocolsHandler`, this may cause issues, but it is unlikely. [1]: https://github.com/paritytech/substrate/issues/4272 * Fix intra-rustdoc links. * Update core/src/connection/pool.rs Co-Authored-By: Max Inden <mail@max-inden.de> * Address some review feedback and fix doc links. * Allow responses to be sent on the same connection. * Remove unnecessary remainders of inject_replaced. * Update swarm/src/behaviour.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update swarm/src/lib.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/manager.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Update core/src/connection/pool.rs Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com> * Incorporate more review feedback. * Move module declaration below imports. * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Update core/src/connection/manager.rs Co-Authored-By: Toralf Wittner <tw@dtex.org> * Simplify as per review. * Fix rustoc link. * Add try_notify_handler and simplify. * Relocate DialingConnection and DialingAttempt. For better visibility constraints. * Small cleanup. * Small cleanup. More robust EstablishedConnectionIter. * Clarify semantics of `DialingPeer::connect`. * Don't call inject_disconnected on InvalidPeerId. To preserve the previous behavior and ensure calls to `inject_disconnected` are always paired with calls to `inject_connected`. * Provide public ConnectionId constructor. Mainly needed for testing purposes, e.g. in substrate. * Move the established connection limit check to the right place. * Clean up connection error handling. Separate connection errors into those occuring during connection setup or upon rejecting a newly established connection (the `PendingConnectionError`) and those errors occurring on previously established connections, i.e. for which a `ConnectionEstablished` event has been emitted by the connection pool earlier. * Revert change in log level and clarify an invariant. * Remove inject_replaced entirely. * Allow notifying all connection handlers. Thereby simplify by introducing a new enum `NotifyHandler`, used with a single constructor `NetworkBehaviourAction::NotifyHandler`. * Finishing touches. Small API simplifications and code deduplication. Some more useful debug logging. Co-authored-by: Max Inden <mail@max-inden.de> Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com> Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-03-04 13:49:25 +01:00
}
/// Stream of events returned by [`Swarm`].
///
/// Includes events from the [`NetworkBehaviour`] as well as events about
/// connection and listener status. See [`SwarmEvent`] for details.
///
/// Note: This stream is infinite and it is guaranteed that
/// [`Stream::poll_next`] will never return `Poll::Ready(None)`.
impl<TBehaviour> Stream for Swarm<TBehaviour>
where
TBehaviour: NetworkBehaviour,
{
type Item = SwarmEvent<TBehaviourOutEvent<TBehaviour>, THandlerErr<TBehaviour>>;
2020-07-27 20:27:33 +00:00
fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
self.as_mut().poll_next_event(cx).map(Some)
}
}
/// The stream of swarm events never terminates, so we can implement fused for it.
impl<TBehaviour> FusedStream for Swarm<TBehaviour>
where
TBehaviour: NetworkBehaviour,
{
fn is_terminated(&self) -> bool {
false
}
}
/// Parameters passed to `poll()`, that the `NetworkBehaviour` has access to.
// TODO: #[derive(Debug)]
pub struct SwarmPollParameters<'a> {
local_peer_id: &'a PeerId,
supported_protocols: &'a [Vec<u8>],
listened_addrs: &'a [Multiaddr],
external_addrs: &'a Addresses,
}
impl<'a> PollParameters for SwarmPollParameters<'a> {
type SupportedProtocolsIter = std::vec::IntoIter<Vec<u8>>;
type ListenedAddressesIter = std::vec::IntoIter<Multiaddr>;
type ExternalAddressesIter = AddressIntoIter;
fn supported_protocols(&self) -> Self::SupportedProtocolsIter {
self.supported_protocols.to_vec().into_iter()
}
fn listened_addresses(&self) -> Self::ListenedAddressesIter {
self.listened_addrs.to_vec().into_iter()
}
fn external_addresses(&self) -> Self::ExternalAddressesIter {
self.external_addrs.clone().into_iter()
}
fn local_peer_id(&self) -> &PeerId {
2021-09-14 16:00:05 +03:00
self.local_peer_id
}
}
/// A `SwarmBuilder` provides an API for configuring and constructing a `Swarm`,
/// including the underlying [`Network`].
pub struct SwarmBuilder<TBehaviour> {
local_peer_id: PeerId,
transport: transport::Boxed<(PeerId, StreamMuxerBox)>,
behaviour: TBehaviour,
network_config: NetworkConfig,
substream_upgrade_protocol_override: Option<libp2p_core::upgrade::Version>,
}
impl<TBehaviour> SwarmBuilder<TBehaviour>
where
TBehaviour: NetworkBehaviour,
{
/// Creates a new `SwarmBuilder` from the given transport, behaviour and
/// local peer ID. The `Swarm` with its underlying `Network` is obtained
/// via [`SwarmBuilder::build`].
pub fn new(
transport: transport::Boxed<(PeerId, StreamMuxerBox)>,
behaviour: TBehaviour,
local_peer_id: PeerId,
) -> Self {
SwarmBuilder {
local_peer_id,
transport,
behaviour,
network_config: Default::default(),
substream_upgrade_protocol_override: None,
}
}
/// Configures the `Executor` to use for spawning background tasks.
///
/// By default, unless another executor has been configured,
/// [`SwarmBuilder::build`] will try to set up a `ThreadPool`.
pub fn executor(mut self, e: Box<dyn Executor + Send>) -> Self {
self.network_config = self.network_config.with_executor(e);
self
}
/// Configures the number of events from the [`NetworkBehaviour`] in
/// destination to the [`ProtocolsHandler`] that can be buffered before
/// the [`Swarm`] has to wait. An individual buffer with this number of
/// events exists for each individual connection.
///
/// The ideal value depends on the executor used, the CPU speed, and the
/// volume of events. If this value is too low, then the [`Swarm`] will
/// be sleeping more often than necessary. Increasing this value increases
/// the overall memory usage.
pub fn notify_handler_buffer_size(mut self, n: NonZeroUsize) -> Self {
self.network_config = self.network_config.with_notify_handler_buffer_size(n);
self
}
/// Configures the number of extra events from the [`ProtocolsHandler`] in
/// destination to the [`NetworkBehaviour`] that can be buffered before
/// the [`ProtocolsHandler`] has to go to sleep.
///
/// There exists a buffer of events received from [`ProtocolsHandler`]s
/// that the [`NetworkBehaviour`] has yet to process. This buffer is
/// shared between all instances of [`ProtocolsHandler`]. Each instance of
/// [`ProtocolsHandler`] is guaranteed one slot in this buffer, meaning
/// that delivering an event for the first time is guaranteed to be
/// instantaneous. Any extra event delivery, however, must wait for that
/// first event to be delivered or for an "extra slot" to be available.
///
/// This option configures the number of such "extra slots" in this
/// shared buffer. These extra slots are assigned in a first-come,
/// first-served basis.
///
/// The ideal value depends on the executor used, the CPU speed, the
/// average number of connections, and the volume of events. If this value
/// is too low, then the [`ProtocolsHandler`]s will be sleeping more often
/// than necessary. Increasing this value increases the overall memory
/// usage, and more importantly the latency between the moment when an
/// event is emitted and the moment when it is received by the
/// [`NetworkBehaviour`].
pub fn connection_event_buffer_size(mut self, n: usize) -> Self {
self.network_config = self.network_config.with_connection_event_buffer_size(n);
self
}
/// Number of addresses concurrently dialed for a single outbound connection attempt.
pub fn dial_concurrency_factor(mut self, factor: NonZeroU8) -> Self {
self.network_config = self.network_config.with_dial_concurrency_factor(factor);
self
}
/// Configures the connection limits.
pub fn connection_limits(mut self, limits: ConnectionLimits) -> Self {
self.network_config = self.network_config.with_connection_limits(limits);
self
}
/// Configures an override for the substream upgrade protocol to use.
///
/// The subtream upgrade protocol is the multistream-select protocol
/// used for protocol negotiation on substreams. Since a listener
/// supports all existing versions, the choice of upgrade protocol
/// only effects the "dialer", i.e. the peer opening a substream.
///
/// > **Note**: If configured, specific upgrade protocols for
/// > individual [`SubstreamProtocol`]s emitted by the `NetworkBehaviour`
/// > are ignored.
pub fn substream_upgrade_protocol_override(mut self, v: libp2p_core::upgrade::Version) -> Self {
self.substream_upgrade_protocol_override = Some(v);
self
}
/// Builds a `Swarm` with the current configuration.
pub fn build(mut self) -> Swarm<TBehaviour> {
let supported_protocols = self
.behaviour
.new_handler()
.inbound_protocol()
.protocol_info()
.into_iter()
.map(|info| info.protocol_name().to_vec())
.collect();
// If no executor has been explicitly configured, try to set up a thread pool.
let network_cfg =
self.network_config.or_else_with_executor(|| {
match ThreadPoolBuilder::new()
.name_prefix("libp2p-swarm-task-")
.create()
{
Ok(tp) => Some(Box::new(move |f| tp.spawn_ok(f))),
Err(err) => {
log::warn!("Failed to create executor thread pool: {:?}", err);
None
}
}
});
let network = Network::new(self.transport, self.local_peer_id, network_cfg);
Swarm {
network,
behaviour: self.behaviour,
supported_protocols,
listened_addrs: SmallVec::new(),
external_addrs: Addresses::default(),
banned_peers: HashSet::new(),
banned_peer_connections: HashSet::new(),
pending_event: None,
substream_upgrade_protocol_override: self.substream_upgrade_protocol_override,
}
}
}
/// The possible failures of dialing.
#[derive(Debug)]
pub enum DialError {
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
/// The peer is currently banned.
Banned,
/// The configured limit for simultaneous outgoing connections
/// has been reached.
ConnectionLimit(ConnectionLimit),
/// The peer being dialed is the local peer and thus the dial was aborted.
LocalPeerId,
/// [`NetworkBehaviour::addresses_of_peer`] returned no addresses
/// for the peer to dial.
NoAddresses,
/// The provided [`dial_opts::PeerCondition`] evaluated to false and thus
/// the dial was aborted.
DialPeerConditionFalse(dial_opts::PeerCondition),
/// Pending connection attempt has been aborted.
Aborted,
/// The provided peer identity is invalid or the peer identity obtained on
/// the connection did not match the one that was expected or is otherwise
/// invalid.
InvalidPeerId,
/// An I/O error occurred on the connection.
ConnectionIo(io::Error),
/// An error occurred while negotiating the transport protocol(s) on a connection.
Transport(Vec<(Multiaddr, TransportError<io::Error>)>),
}
impl DialError {
fn from_network_dial_error<THandler>(error: network::DialError<THandler>) -> (Self, THandler) {
match error {
network::DialError::ConnectionLimit { limit, handler } => {
(DialError::ConnectionLimit(limit), handler)
}
network::DialError::LocalPeerId { handler } => (DialError::LocalPeerId, handler),
network::DialError::InvalidPeerId { handler } => (DialError::InvalidPeerId, handler),
}
}
}
impl From<PendingOutboundConnectionError<io::Error>> for DialError {
fn from(error: PendingOutboundConnectionError<io::Error>) -> Self {
match error {
PendingConnectionError::ConnectionLimit(limit) => DialError::ConnectionLimit(limit),
PendingConnectionError::Aborted => DialError::Aborted,
PendingConnectionError::InvalidPeerId => DialError::InvalidPeerId,
PendingConnectionError::IO(e) => DialError::ConnectionIo(e),
PendingConnectionError::Transport(e) => DialError::Transport(e),
}
}
}
impl fmt::Display for DialError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
DialError::ConnectionLimit(err) => write!(f, "Dial error: {}", err),
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
DialError::NoAddresses => write!(f, "Dial error: no addresses for peer."),
DialError::LocalPeerId => write!(f, "Dial error: tried to dial local peer id."),
DialError::Banned => write!(f, "Dial error: peer is banned."),
DialError::DialPeerConditionFalse(c) => {
write!(
f,
"Dial error: condition {:?} for dialing peer was false.",
c
)
}
DialError::Aborted => write!(
f,
"Dial error: Pending connection attempt has been aborted."
),
DialError::InvalidPeerId => write!(f, "Dial error: Invalid peer ID."),
DialError::ConnectionIo(e) => write!(
f,
"Dial error: An I/O error occurred on the connection: {:?}.", e
),
DialError::Transport(e) => write!(f, "An error occurred while negotiating the transport protocol(s) on a connection: {:?}.", e),
}
}
}
impl error::Error for DialError {
fn source(&self) -> Option<&(dyn error::Error + 'static)> {
match self {
DialError::ConnectionLimit(err) => Some(err),
DialError::LocalPeerId => None,
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
DialError::NoAddresses => None,
DialError::Banned => None,
DialError::DialPeerConditionFalse(_) => None,
DialError::Aborted => None,
DialError::InvalidPeerId => None,
DialError::ConnectionIo(_) => None,
DialError::Transport(_) => None,
}
}
}
/// Dummy implementation of [`NetworkBehaviour`] that doesn't do anything.
#[derive(Clone)]
pub struct DummyBehaviour {
keep_alive: KeepAlive,
}
impl DummyBehaviour {
pub fn with_keep_alive(keep_alive: KeepAlive) -> Self {
Self { keep_alive }
}
pub fn keep_alive_mut(&mut self) -> &mut KeepAlive {
&mut self.keep_alive
}
}
impl Default for DummyBehaviour {
fn default() -> Self {
Self {
keep_alive: KeepAlive::No,
}
}
}
impl NetworkBehaviour for DummyBehaviour {
type ProtocolsHandler = protocols_handler::DummyProtocolsHandler;
type OutEvent = void::Void;
fn new_handler(&mut self) -> Self::ProtocolsHandler {
protocols_handler::DummyProtocolsHandler {
keep_alive: self.keep_alive,
}
}
fn inject_event(
&mut self,
_: PeerId,
_: ConnectionId,
event: <Self::ProtocolsHandler as ProtocolsHandler>::OutEvent,
) {
void::unreachable(event)
}
fn poll(
&mut self,
_: &mut Context<'_>,
_: &mut impl PollParameters,
) -> Poll<NetworkBehaviourAction<Self::OutEvent, Self::ProtocolsHandler>> {
Poll::Pending
}
}
#[cfg(test)]
mod tests {
use super::*;
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
use crate::protocols_handler::DummyProtocolsHandler;
use crate::test::{CallTraceBehaviour, MockBehaviour};
use futures::{executor, future};
use libp2p::core::{identity, multiaddr, transport, upgrade};
use libp2p::plaintext;
use libp2p::yamux;
// Test execution state.
// Connection => Disconnecting => Connecting.
enum State {
Connecting,
Disconnecting,
}
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
fn new_test_swarm<T, O>(handler_proto: T) -> Swarm<CallTraceBehaviour<MockBehaviour<T, O>>>
where
T: ProtocolsHandler + Clone,
T::OutEvent: Clone,
O: Send + 'static,
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
{
let id_keys = identity::Keypair::generate_ed25519();
let local_public_key = id_keys.public();
let transport = transport::MemoryTransport::default()
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
.upgrade(upgrade::Version::V1)
.authenticate(plaintext::PlainText2Config {
local_public_key: local_public_key.clone(),
})
.multiplex(yamux::YamuxConfig::default())
.boxed();
let behaviour = CallTraceBehaviour::new(MockBehaviour::new(handler_proto));
SwarmBuilder::new(transport, behaviour, local_public_key.into()).build()
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
fn swarms_connected<TBehaviour>(
swarm1: &Swarm<CallTraceBehaviour<TBehaviour>>,
swarm2: &Swarm<CallTraceBehaviour<TBehaviour>>,
num_connections: usize,
) -> bool
where
TBehaviour: NetworkBehaviour,
<<TBehaviour::ProtocolsHandler as IntoProtocolsHandler>::Handler as ProtocolsHandler>::OutEvent: Clone,
{
[swarm1, swarm2]
.iter()
.all(|s| s.behaviour.inject_connection_established.len() == num_connections)
}
fn swarms_disconnected<TBehaviour: NetworkBehaviour>(
swarm1: &Swarm<CallTraceBehaviour<TBehaviour>>,
swarm2: &Swarm<CallTraceBehaviour<TBehaviour>>,
num_connections: usize,
) -> bool
where
TBehaviour: NetworkBehaviour,
<<TBehaviour::ProtocolsHandler as IntoProtocolsHandler>::Handler as ProtocolsHandler>::OutEvent: Clone
{
[swarm1, swarm2]
.iter()
.all(|s| s.behaviour.inject_connection_closed.len() == num_connections)
&& [swarm1, swarm2]
.iter()
.all(|s| s.behaviour.inject_disconnected.len() == 1)
}
/// Establishes multiple connections between two peers,
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
/// after which one peer bans the other.
///
/// The test expects both behaviours to be notified via pairs of
/// inject_connected / inject_disconnected as well as
/// inject_connection_established / inject_connection_closed calls
/// while unbanned.
///
/// While the ban is in effect, further dials occur. For these connections no
/// `inject_connected`, `inject_connection_established`, `inject_disconnected`,
/// `inject_connection_closed` calls should be registered.
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
#[test]
fn test_connect_disconnect_ban() {
// Since the test does not try to open any substreams, we can
// use the dummy protocols handler.
let handler_proto = DummyProtocolsHandler {
keep_alive: KeepAlive::Yes,
};
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
let mut swarm1 = new_test_swarm::<_, ()>(handler_proto.clone());
let mut swarm2 = new_test_swarm::<_, ()>(handler_proto);
let addr1: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
let addr2: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
swarm1.listen_on(addr1.clone()).unwrap();
swarm2.listen_on(addr2.clone()).unwrap();
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
let swarm1_id = *swarm1.local_peer_id();
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
enum Stage {
/// Waiting for the peers to connect. Banning has not occurred.
Connecting,
/// Ban occurred.
Banned,
// Ban is in place and a dial is ongoing.
BannedDial,
// Mid-ban dial was registered and the peer was unbanned.
Unbanned,
// There are dial attempts ongoing for the no longer banned peers.
Reconnecting,
}
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
let num_connections = 10;
for _ in 0..num_connections {
swarm1.dial(addr2.clone()).unwrap();
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
let mut s1_expected_conns = num_connections;
let mut s2_expected_conns = num_connections;
let mut stage = Stage::Connecting;
executor::block_on(future::poll_fn(move |cx| loop {
let poll1 = Swarm::poll_next_event(Pin::new(&mut swarm1), cx);
let poll2 = Swarm::poll_next_event(Pin::new(&mut swarm2), cx);
match stage {
Stage::Connecting => {
if swarm1.behaviour.assert_connected(s1_expected_conns, 1)
&& swarm2.behaviour.assert_connected(s2_expected_conns, 1)
{
// Setup to test that already established connections are correctly closed
// and reported as such after the peer is banned.
swarm2.ban_peer_id(swarm1_id);
stage = Stage::Banned;
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
}
Stage::Banned => {
if swarm1.behaviour.assert_disconnected(s1_expected_conns, 1)
&& swarm2.behaviour.assert_disconnected(s2_expected_conns, 1)
{
// Setup to test that new connections of banned peers are not reported.
swarm1.dial(addr2.clone()).unwrap();
s1_expected_conns += 1;
stage = Stage::BannedDial;
}
}
Stage::BannedDial => {
if swarm2.network_info().num_peers() == 1 {
// The banned connection was established. Check that it was not reported to
// the behaviour of the banning swarm.
assert_eq!(
swarm2.behaviour.inject_connection_established.len(), s2_expected_conns,
"No additional closed connections should be reported for the banned peer"
);
// Setup to test that the banned connection is not reported upon closing
// even if the peer is unbanned.
swarm2.unban_peer_id(swarm1_id);
stage = Stage::Unbanned;
}
}
Stage::Unbanned => {
if swarm2.network_info().num_peers() == 0 {
// The banned connection has closed. Check that it was not reported.
assert_eq!(
swarm2.behaviour.inject_connection_closed.len(), s2_expected_conns,
"No additional closed connections should be reported for the banned peer"
);
assert!(swarm2.banned_peer_connections.is_empty());
// Setup to test that a ban lifted does not affect future connections.
for _ in 0..num_connections {
swarm1.dial(addr2.clone()).unwrap();
}
s1_expected_conns += num_connections;
s2_expected_conns += num_connections;
stage = Stage::Reconnecting;
}
}
Stage::Reconnecting => {
if swarm1.behaviour.inject_connection_established.len() == s1_expected_conns
&& swarm2.behaviour.assert_connected(s2_expected_conns, 2)
{
return Poll::Ready(());
}
}
}
if poll1.is_pending() && poll2.is_pending() {
return Poll::Pending;
}
}))
}
/// Establishes multiple connections between two peers,
/// after which one peer disconnects the other using [`Swarm::disconnect_peer_id`].
///
/// The test expects both behaviours to be notified via pairs of
/// inject_connected / inject_disconnected as well as
/// inject_connection_established / inject_connection_closed calls.
#[test]
fn test_swarm_disconnect() {
// Since the test does not try to open any substreams, we can
// use the dummy protocols handler.
let handler_proto = DummyProtocolsHandler {
keep_alive: KeepAlive::Yes,
};
let mut swarm1 = new_test_swarm::<_, ()>(handler_proto.clone());
let mut swarm2 = new_test_swarm::<_, ()>(handler_proto);
let addr1: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
let addr2: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
swarm1.listen_on(addr1.clone()).unwrap();
swarm2.listen_on(addr2.clone()).unwrap();
let swarm1_id = *swarm1.local_peer_id();
let mut reconnected = false;
let num_connections = 10;
for _ in 0..num_connections {
swarm1.dial(addr2.clone()).unwrap();
}
let mut state = State::Connecting;
executor::block_on(future::poll_fn(move |cx| loop {
let poll1 = Swarm::poll_next_event(Pin::new(&mut swarm1), cx);
let poll2 = Swarm::poll_next_event(Pin::new(&mut swarm2), cx);
match state {
State::Connecting => {
if swarms_connected(&swarm1, &swarm2, num_connections) {
if reconnected {
return Poll::Ready(());
}
swarm2
.disconnect_peer_id(swarm1_id)
.expect("Error disconnecting");
state = State::Disconnecting;
}
}
State::Disconnecting => {
if swarms_disconnected(&swarm1, &swarm2, num_connections) {
if reconnected {
return Poll::Ready(());
}
reconnected = true;
for _ in 0..num_connections {
swarm2.dial(addr1.clone()).unwrap();
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
state = State::Connecting;
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
}
}
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
if poll1.is_pending() && poll2.is_pending() {
return Poll::Pending;
[core/swarm] Emit events for active connection close and fix `disconnect()`. (#1619) * Emit events for active connection close and fix `disconnect()`. The `Network` does currently not emit events for actively closed connections, e.g. via `EstablishedConnection::close` or `ConnectedPeer::disconnect()`. As a result, when actively closing connections, there will be `ConnectionEstablished` events emitted without eventually a matching `ConnectionClosed` event. This seems undesirable and has the consequence that the `Swarm::ban_peer_id` feature in `libp2p-swarm` does not result in appropriate calls to `NetworkBehaviour::inject_connection_closed` and `NetworkBehaviour::inject_disconnected`. Furthermore, the `disconnect()` functionality in `libp2p-core` is currently broken as it leaves the `Pool` in an inconsistent state. This commit does the following: 1. When connection background tasks are dropped (i.e. removed from the `Manager`), they always terminate immediately, without attempting an orderly close of the connection. 2. An orderly close is sent to the background task of a connection as a regular command. The background task emits a `Closed` event before terminating. 3. `Pool::disconnect()` removes all connection tasks for the affected peer from the `Manager`, i.e. without an orderly close, thereby also fixing the discovered state inconsistency due to not removing the corresponding entries in the `Pool` itself after removing them from the `Manager`. 4. A new test is added to `libp2p-swarm` that exercises the ban/unban functionality and places assertions on the number and order of calls to the `NetworkBehaviour`. In that context some new testing utilities have been added to `libp2p-swarm`. This addresses https://github.com/libp2p/rust-libp2p/issues/1584. * Update swarm/src/lib.rs Co-authored-by: Toralf Wittner <tw@dtex.org> * Incorporate some review feedback. * Adapt to changes in master. * More verbose panic messages. * Simplify There is no need for a `StartClose` future. * Fix doc links. * Further small cleanup. * Update CHANGELOGs and versions. Co-authored-by: Toralf Wittner <tw@dtex.org>
2020-08-04 11:30:09 +02:00
}
}))
}
/// Establishes multiple connections between two peers,
/// after which one peer disconnects the other
/// using [`NetworkBehaviourAction::CloseConnection`] returned by a [`NetworkBehaviour`].
///
/// The test expects both behaviours to be notified via pairs of
/// inject_connected / inject_disconnected as well as
/// inject_connection_established / inject_connection_closed calls.
#[test]
fn test_behaviour_disconnect_all() {
// Since the test does not try to open any substreams, we can
// use the dummy protocols handler.
let handler_proto = DummyProtocolsHandler {
keep_alive: KeepAlive::Yes,
};
let mut swarm1 = new_test_swarm::<_, ()>(handler_proto.clone());
let mut swarm2 = new_test_swarm::<_, ()>(handler_proto);
let addr1: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
let addr2: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
swarm1.listen_on(addr1.clone()).unwrap();
swarm2.listen_on(addr2.clone()).unwrap();
let swarm1_id = *swarm1.local_peer_id();
let mut reconnected = false;
let num_connections = 10;
for _ in 0..num_connections {
swarm1.dial(addr2.clone()).unwrap();
}
let mut state = State::Connecting;
executor::block_on(future::poll_fn(move |cx| loop {
let poll1 = Swarm::poll_next_event(Pin::new(&mut swarm1), cx);
let poll2 = Swarm::poll_next_event(Pin::new(&mut swarm2), cx);
match state {
State::Connecting => {
if swarms_connected(&swarm1, &swarm2, num_connections) {
if reconnected {
return Poll::Ready(());
}
swarm2.behaviour.inner().next_action.replace(
NetworkBehaviourAction::CloseConnection {
peer_id: swarm1_id,
connection: CloseConnection::All,
},
);
state = State::Disconnecting;
}
}
State::Disconnecting => {
if swarms_disconnected(&swarm1, &swarm2, num_connections) {
if reconnected {
return Poll::Ready(());
}
reconnected = true;
for _ in 0..num_connections {
swarm2.dial(addr1.clone()).unwrap();
}
state = State::Connecting;
}
}
}
if poll1.is_pending() && poll2.is_pending() {
return Poll::Pending;
}
}))
}
/// Establishes multiple connections between two peers,
/// after which one peer closes a single connection
/// using [`NetworkBehaviourAction::CloseConnection`] returned by a [`NetworkBehaviour`].
///
/// The test expects both behaviours to be notified via pairs of
/// inject_connected / inject_disconnected as well as
/// inject_connection_established / inject_connection_closed calls.
#[test]
fn test_behaviour_disconnect_one() {
// Since the test does not try to open any substreams, we can
// use the dummy protocols handler.
let handler_proto = DummyProtocolsHandler {
keep_alive: KeepAlive::Yes,
};
let mut swarm1 = new_test_swarm::<_, ()>(handler_proto.clone());
let mut swarm2 = new_test_swarm::<_, ()>(handler_proto);
let addr1: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
let addr2: Multiaddr = multiaddr::Protocol::Memory(rand::random::<u64>()).into();
swarm1.listen_on(addr1.clone()).unwrap();
swarm2.listen_on(addr2.clone()).unwrap();
let swarm1_id = *swarm1.local_peer_id();
let num_connections = 10;
for _ in 0..num_connections {
swarm1.dial(addr2.clone()).unwrap();
}
let mut state = State::Connecting;
let mut disconnected_conn_id = None;
executor::block_on(future::poll_fn(move |cx| loop {
let poll1 = Swarm::poll_next_event(Pin::new(&mut swarm1), cx);
let poll2 = Swarm::poll_next_event(Pin::new(&mut swarm2), cx);
match state {
State::Connecting => {
if swarms_connected(&swarm1, &swarm2, num_connections) {
disconnected_conn_id = {
let conn_id = swarm2.behaviour.inject_connection_established
[num_connections / 2]
.1;
swarm2.behaviour.inner().next_action.replace(
NetworkBehaviourAction::CloseConnection {
peer_id: swarm1_id,
connection: CloseConnection::One(conn_id),
},
);
Some(conn_id)
};
state = State::Disconnecting;
}
}
State::Disconnecting => {
for s in &[&swarm1, &swarm2] {
assert_eq!(s.behaviour.inject_disconnected.len(), 0);
assert_eq!(
s.behaviour.inject_connection_established.len(),
num_connections
);
assert_eq!(s.behaviour.inject_connected.len(), 1);
}
if [&swarm1, &swarm2]
.iter()
.all(|s| s.behaviour.inject_connection_closed.len() == 1)
{
let conn_id = swarm2.behaviour.inject_connection_closed[0].1;
assert_eq!(Some(conn_id), disconnected_conn_id);
return Poll::Ready(());
}
}
}
if poll1.is_pending() && poll2.is_pending() {
return Poll::Pending;
}
}))
}
}