1308 lines
53 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 behaviour;
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 protocols_handler;
pub mod toggle;
pub use behaviour::{
NetworkBehaviour,
NetworkBehaviourAction,
NetworkBehaviourEventProcess,
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
PollParameters,
NotifyHandler,
DialPeerCondition
};
pub use protocols_handler::{
IntoProtocolsHandler,
IntoProtocolsHandlerSelect,
KeepAlive,
ProtocolsHandler,
ProtocolsHandlerEvent,
ProtocolsHandlerSelect,
ProtocolsHandlerUpgrErr,
OneShotHandler,
OneShotHandlerConfig,
SubstreamProtocol
};
use protocols_handler::{
NodeHandlerWrapperBuilder,
NodeHandlerWrapperError,
};
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
use futures::{
prelude::*,
executor::ThreadPoolBuilder,
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::FusedStream,
};
use libp2p_core::{
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
Executor,
Transport,
Multiaddr,
Negotiated,
PeerId,
connection::{
ConnectionError,
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
ConnectionId,
ConnectionInfo,
ConnectionLimit,
ConnectedPoint,
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
EstablishedConnection,
IntoConnectionHandler,
ListenerId,
PendingConnectionError,
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
},
transport::{self, TransportError},
muxing::StreamMuxerBox,
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,
NetworkInfo,
NetworkEvent,
NetworkConfig,
peer::ConnectedPeer,
},
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
upgrade::ProtocolName,
};
use registry::{Addresses, AddressIntoIter};
use smallvec::SmallVec;
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
use std::{error, fmt, hash::Hash, io, ops::{Deref, DerefMut}, pin::Pin, task::{Context, Poll}};
use std::collections::HashSet;
use std::num::{NonZeroU32, NonZeroUsize};
use upgrade::UpgradeInfoSend as _;
/// Contains the state of the network, plus the way it should behave.
pub type Swarm<TBehaviour, TConnInfo = PeerId> = ExpandedSwarm<
2019-04-07 18:34:14 -03:00
TBehaviour,
<<<TBehaviour as NetworkBehaviour>::ProtocolsHandler as IntoProtocolsHandler>::Handler as ProtocolsHandler>::InEvent,
<<<TBehaviour as NetworkBehaviour>::ProtocolsHandler as IntoProtocolsHandler>::Handler as ProtocolsHandler>::OutEvent,
<TBehaviour as NetworkBehaviour>::ProtocolsHandler,
TConnInfo,
2019-04-07 18:34:14 -03:00
>;
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
/// 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 `Swarm`.
#[derive(Debug)]
pub enum SwarmEvent<TBvEv, THandleErr> {
/// Event generated by the `NetworkBehaviour`.
Behaviour(TBvEv),
/// 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,
},
[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<THandleErr>>>,
},
/// 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: PendingConnectionError<io::Error>,
},
/// 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,
},
/// Tried to dial an address but it ended up being unreachaable.
UnreachableAddr {
/// `PeerId` that we were trying to reach.
peer_id: PeerId,
/// Address that we failed to reach.
address: Multiaddr,
/// Error that has been encountered.
error: PendingConnectionError<io::Error>,
/// Number of remaining connection attempts that are being tried for this peer.
attempts_remaining: u32,
},
/// Tried to dial an address but it ended up being unreachaable.
/// Contrary to `UnreachableAddr`, we don't know the identity of the peer that we were trying
/// to reach.
UnknownPeerUnreachableAddr {
/// Address that we failed to reach.
address: Multiaddr,
/// Error that has been encountered.
error: PendingConnectionError<io::Error>,
},
/// One of our listeners has reported a new local listening address.
NewListenAddr(Multiaddr),
/// One of our listeners has reported the expiration of a listening address.
ExpiredListenAddr(Multiaddr),
/// One of the listeners gracefully closed.
ListenerClosed {
/// 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 error.
error: io::Error,
},
/// A new dialing attempt has been initiated.
///
/// A [`ConnectionEstablished`](SwarmEvent::ConnectionEstablished)
/// event is reported if the dialing attempt succeeds, otherwise a
/// [`UnreachableAddr`](SwarmEvent::UnreachableAddr) event is reported
/// with `attempts_remaining` equal to 0.
Dialing(PeerId),
}
2019-04-07 18:34:14 -03:00
/// Contains the state of the network, plus the way it should behave.
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
pub struct ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo = PeerId>
where
THandler: IntoProtocolsHandler,
TConnInfo: ConnectionInfo<PeerId = PeerId>,
{
network: Network<
transport::Boxed<(TConnInfo, StreamMuxerBox)>,
2019-04-07 18:34:14 -03:00
TInEvent,
TOutEvent,
NodeHandlerWrapperBuilder<THandler>,
TConnInfo,
PeerId,
>,
/// 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
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, TInEvent)>
}
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
impl<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo> Deref for
ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo>
where
THandler: IntoProtocolsHandler,
TConnInfo: ConnectionInfo<PeerId = PeerId>,
{
type Target = TBehaviour;
fn deref(&self) -> &Self::Target {
&self.behaviour
}
}
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
impl<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo> DerefMut for
ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo>
where
THandler: IntoProtocolsHandler,
TConnInfo: ConnectionInfo<PeerId = PeerId>,
{
fn deref_mut(&mut self) -> &mut Self::Target {
&mut self.behaviour
}
}
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
impl<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo> Unpin for
ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo>
where
THandler: IntoProtocolsHandler,
TConnInfo: ConnectionInfo<PeerId = PeerId>,
{
}
impl<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo, THandleErr>
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
ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo>
2019-04-07 18:34:14 -03:00
where TBehaviour: NetworkBehaviour<ProtocolsHandler = THandler>,
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
TInEvent: Clone + Send + 'static,
TOutEvent: Send + 'static,
TConnInfo: ConnectionInfo<PeerId = PeerId> + fmt::Debug + Clone + Send + 'static,
2019-04-07 18:34:14 -03:00
THandler: IntoProtocolsHandler + Send + 'static,
THandler::Handler: ProtocolsHandler<InEvent = TInEvent, OutEvent = TOutEvent, Error = THandleErr>,
THandleErr: error::Error + Send + 'static,
{
/// Builds a new `Swarm`.
pub fn new(
transport: transport::Boxed<(TConnInfo, 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(me: &Self) -> NetworkInfo {
me.network.info()
}
/// Starts listening on the given address.
///
/// Returns an error if the address is not supported.
pub fn listen_on(me: &mut Self, addr: Multiaddr) -> Result<ListenerId, TransportError<io::Error>> {
me.network.listen_on(addr)
}
/// Remove some listener.
///
/// Returns `Ok(())` if there was a listener with this ID.
pub fn remove_listener(me: &mut Self, id: ListenerId) -> Result<(), ()> {
me.network.remove_listener(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
/// Initiates a new dialing attempt to the given address.
pub fn dial_addr(me: &mut Self, addr: Multiaddr) -> Result<(), ConnectionLimit> {
let handler = me.behaviour.new_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
me.network.dial(&addr, handler.into_node_handler_builder()).map(|_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
/// Initiates a new dialing attempt to the given peer.
pub fn dial(me: &mut Self, peer_id: &PeerId) -> Result<(), 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
if me.banned_peers.contains(peer_id) {
me.behaviour.inject_dial_failure(peer_id);
return Err(DialError::Banned)
}
let self_listening = &me.listened_addrs;
let mut addrs = me.behaviour.addresses_of_peer(peer_id)
.into_iter()
.filter(|a| !self_listening.contains(a));
let result =
if let Some(first) = addrs.next() {
let handler = me.behaviour.new_handler().into_node_handler_builder();
me.network.peer(peer_id.clone())
.dial(first, addrs, handler)
.map(|_| ())
.map_err(DialError::ConnectionLimit)
} else {
Err(DialError::NoAddresses)
};
if let Err(error) = &result {
log::debug!(
"New dialing attempt to peer {:?} failed: {:?}.",
peer_id, error);
me.behaviour.inject_dial_failure(&peer_id);
}
result
}
/// Returns an iterator that produces the list of addresses we're listening on.
pub fn listeners(me: &Self) -> impl Iterator<Item = &Multiaddr> {
me.network.listen_addrs()
}
/// Returns an iterator that produces the list of addresses that other nodes can use to reach
/// us.
pub fn external_addresses(me: &Self) -> impl Iterator<Item = &Multiaddr> {
me.external_addrs.iter()
}
/// Returns the peer ID of the swarm passed as parameter.
pub fn local_peer_id(me: &Self) -> &PeerId {
&me.network.local_peer_id()
}
/// Adds an external address.
///
/// An external address is an address we are listening on but that accounts for things such as
/// NAT traversal.
pub fn add_external_address(me: &mut Self, addr: Multiaddr) {
me.external_addrs.add(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
/// Returns the connection info for an arbitrary connection with the peer, or `None`
/// if there is no connection to that peer.
// TODO: should take &self instead of &mut self, but the API in network requires &mut
pub fn connection_info(me: &mut Self, peer_id: &PeerId) -> Option<TConnInfo> {
if let Some(mut n) = me.network.peer(peer_id.clone()).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
Some(n.some_connection().info().clone())
} else {
None
}
}
/// 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.
2019-04-23 10:54:25 +02:00
pub fn ban_peer_id(me: &mut Self, peer_id: PeerId) {
[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 me.banned_peers.insert(peer_id.clone()) {
if let Some(peer) = me.network.peer(peer_id).into_connected() {
peer.disconnect();
}
2019-04-23 10:54:25 +02:00
}
}
/// Unbans a peer.
2019-04-23 10:54:25 +02:00
pub fn unban_peer_id(me: &mut Self, peer_id: PeerId) {
me.banned_peers.remove(&peer_id);
}
/// Returns the next event that happens in the `Swarm`.
///
/// Includes events from the `NetworkBehaviour` but also events about the connections status.
pub async fn next_event(&mut self) -> SwarmEvent<TBehaviour::OutEvent, THandleErr> {
future::poll_fn(move |cx| ExpandedSwarm::poll_next_event(Pin::new(self), cx)).await
}
/// Returns the next event produced by the [`NetworkBehaviour`].
pub async fn next(&mut self) -> TBehaviour::OutEvent {
future::poll_fn(move |cx| {
loop {
let event = futures::ready!(ExpandedSwarm::poll_next_event(Pin::new(self), cx));
if let SwarmEvent::Behaviour(event) = event {
return Poll::Ready(event);
}
}
}).await
}
/// Internal function used by everything event-related.
///
/// Polls the `Swarm` for the next event.
2020-07-27 20:27:33 +00:00
fn poll_next_event(mut self: Pin<&mut Self>, cx: &mut Context<'_>)
-> Poll<SwarmEvent<TBehaviour::OutEvent, THandleErr>>
{
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().clone();
let connection = connection.id();
this.behaviour.inject_event(peer, connection, event);
},
Poll::Ready(NetworkEvent::AddressChange { connection, new_endpoint, old_endpoint }) => {
let peer = connection.peer_id();
let connection = connection.id();
this.behaviour.inject_address_change(&peer, &connection, &old_endpoint, &new_endpoint);
},
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::ConnectionEstablished { connection, num_established }) => {
let peer_id = connection.peer_id().clone();
let endpoint = connection.endpoint().clone();
if this.banned_peers.contains(&peer_id) {
this.network.peer(peer_id.clone())
.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 {
log::debug!("Connection established: {:?}; Total (peer): {}.",
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
connection.connected(), num_established);
let endpoint = connection.endpoint().clone();
this.behaviour.inject_connection_established(&peer_id, &connection.id(), &endpoint);
if num_established.get() == 1 {
this.behaviour.inject_connected(&peer_id);
}
return Poll::Ready(SwarmEvent::ConnectionEstablished {
peer_id, num_established, endpoint
});
}
},
[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
Poll::Ready(NetworkEvent::ConnectionClosed { id, connected, error, num_established }) => {
if let Some(error) = error.as_ref() {
log::debug!("Connection {:?} closed: {:?}", connected, error);
} else {
log::debug!("Connection {:?} closed (active close).", connected);
}
let info = connected.info;
let endpoint = connected.endpoint;
this.behaviour.inject_connection_closed(info.peer_id(), &id, &endpoint);
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 num_established == 0 {
this.behaviour.inject_disconnected(info.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: info.peer_id().clone(),
endpoint,
cause: error,
num_established,
});
},
Poll::Ready(NetworkEvent::IncomingConnection { connection, .. }) => {
let handler = this.behaviour.new_handler();
let local_addr = connection.local_addr.clone();
let send_back_addr = connection.send_back_addr.clone();
if let Err(e) = this.network.accept(connection, handler.into_node_handler_builder()) {
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::warn!("Incoming connection rejected: {:?}", e);
}
return Poll::Ready(SwarmEvent::IncomingConnection {
local_addr,
send_back_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
Poll::Ready(NetworkEvent::NewListenerAddress { listener_id, listen_addr }) => {
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(&listen_addr);
return Poll::Ready(SwarmEvent::NewListenAddr(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
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(&listen_addr);
return Poll::Ready(SwarmEvent::ExpiredListenAddr(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(addr);
}
this.behaviour.inject_listener_closed(listener_id, match &reason {
Ok(()) => Ok(()),
Err(err) => Err(err),
});
return Poll::Ready(SwarmEvent::ListenerClosed {
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 {
error,
});
},
Poll::Ready(NetworkEvent::IncomingConnectionError { local_addr, send_back_addr, 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
log::debug!("Incoming connection failed: {:?}", error);
return Poll::Ready(SwarmEvent::IncomingConnectionError {
local_addr,
send_back_addr,
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
},
Poll::Ready(NetworkEvent::DialError { peer_id, multiaddr, error, attempts_remaining }) => {
log::debug!(
"Connection attempt to {:?} via {:?} failed with {:?}. Attempts remaining: {}.",
peer_id, multiaddr, error, attempts_remaining);
this.behaviour.inject_addr_reach_failure(Some(&peer_id), &multiaddr, &error);
if attempts_remaining == 0 {
this.behaviour.inject_dial_failure(&peer_id);
}
return Poll::Ready(SwarmEvent::UnreachableAddr {
peer_id,
address: multiaddr,
error,
attempts_remaining,
});
},
Poll::Ready(NetworkEvent::UnknownPeerDialError { multiaddr, 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
log::debug!("Connection attempt to address {:?} of unknown peer failed with {:?}",
multiaddr, error);
this.behaviour.inject_addr_reach_failure(None, &multiaddr, &error);
return Poll::Ready(SwarmEvent::UnknownPeerUnreachableAddr {
address: multiaddr,
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.clone()).into_connected() {
match handler {
PendingNotifyHandler::One(conn_id) =>
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));
return Poll::Pending
}
},
PendingNotifyHandler::Any(ids) => {
if let Some((event, ids)) = notify_any(ids, &mut peer, event, cx) {
let handler = PendingNotifyHandler::Any(ids);
this.pending_event = Some((peer_id, handler, event));
return Poll::Pending
}
}
PendingNotifyHandler::All(ids) => {
if let Some((event, ids)) = notify_all(ids, &mut peer, event, cx) {
let handler = PendingNotifyHandler::All(ids);
this.pending_event = Some((peer_id, handler, event));
return Poll::Pending
}
}
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::DialAddress { address }) => {
let _ = ExpandedSwarm::dial_addr(&mut *this, address);
},
Poll::Ready(NetworkBehaviourAction::DialPeer { peer_id, condition }) => {
if this.banned_peers.contains(&peer_id) {
this.behaviour.inject_dial_failure(&peer_id);
} else {
let condition_matched = match condition {
DialPeerCondition::Disconnected
if this.network.is_disconnected(&peer_id) => true,
DialPeerCondition::NotDialing
if !this.network.is_dialing(&peer_id) => true,
_ => false
};
if condition_matched {
if ExpandedSwarm::dial(this, &peer_id).is_ok() {
return Poll::Ready(SwarmEvent::Dialing(peer_id))
}
} else {
// Even if the condition for a _new_ dialing attempt is not met,
// we always add any potentially new addresses of the peer to an
// ongoing dialing attempt, if there is one.
log::trace!("Condition for new dialing attempt to {:?} not met: {:?}",
peer_id, condition);
let self_listening = &this.listened_addrs;
if let Some(mut peer) = this.network.peer(peer_id.clone()).into_dialing() {
let addrs = this.behaviour.addresses_of_peer(peer.id());
let mut attempt = peer.some_attempt();
for a in addrs {
if !self_listening.contains(&a) {
attempt.add_address(a);
}
}
}
}
}
},
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(NetworkBehaviourAction::NotifyHandler { peer_id, handler, event }) => {
if let Some(mut peer) = this.network.peer(peer_id.clone()).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));
return Poll::Pending
}
}
}
NotifyHandler::Any => {
let ids = peer.connections().into_ids().collect();
if let Some((event, ids)) = notify_any(ids, &mut peer, event, cx) {
let handler = PendingNotifyHandler::Any(ids);
this.pending_event = Some((peer_id, handler, event));
return Poll::Pending
}
}
NotifyHandler::All => {
let ids = peer.connections().into_ids().collect();
if let Some((event, ids)) = notify_all(ids, &mut peer, event, cx) {
let handler = PendingNotifyHandler::All(ids);
this.pending_event = Some((peer_id, handler, event));
return Poll::Pending
}
}
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 }) => {
for addr in this.network.address_translation(&address) {
if this.external_addrs.iter().all(|a| *a != addr) {
this.behaviour.inject_new_external_addr(&addr);
}
this.external_addrs.add(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
/// Connections to notify of a pending event.
///
/// The connection IDs to notify of an event are captured at the time
/// the behaviour emits the event, in order not to forward the event
/// to new connections which the behaviour may not have been aware of
/// at the time it issued the request for sending it.
enum PendingNotifyHandler {
One(ConnectionId),
Any(SmallVec<[ConnectionId; 10]>),
All(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.
[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 notify_one<'a, TInEvent, TConnInfo>(
conn: &mut EstablishedConnection<'a, TInEvent, TConnInfo>,
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: TInEvent,
2020-07-27 20:27:33 +00:00
cx: &mut Context<'_>,
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
) -> Option<TInEvent>
where
[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
TConnInfo: ConnectionInfo
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, TInEvent, TOutEvent, THandler, TConnInfo, TPeerId>(
ids: SmallVec<[ConnectionId; 10]>,
peer: &mut ConnectedPeer<'a, TTrans, TInEvent, TOutEvent, THandler, TConnInfo, TPeerId>,
event: TInEvent,
2020-07-27 20:27:33 +00:00
cx: &mut Context<'_>,
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
) -> Option<(TInEvent, SmallVec<[ConnectionId; 10]>)>
where
TTrans: Transport,
THandler: IntoConnectionHandler<TConnInfo>,
TPeerId: Eq + Hash + Clone,
TConnInfo: ConnectionInfo<PeerId = TPeerId>
{
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
}
}
}
}
}
event.and_then(|e|
if !pending.is_empty() {
Some((e, pending))
} else {
None
})
}
/// Notify all of the given connections of a peer of an event.
///
/// Returns `Some` with the given event and a new list of connections if
/// at least one of the given connections is currently not able to receive
/// the event, in which case the current task is scheduled to be woken up and
/// the returned connections are those which still need to receive the 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
///
/// Returns `None` if all connections are either closing or the event
/// was successfully sent to all handlers whose connections are not closing,
/// in either case the event is consumed.
fn notify_all<'a, TTrans, TInEvent, TOutEvent, THandler, TConnInfo, TPeerId>(
ids: SmallVec<[ConnectionId; 10]>,
peer: &mut ConnectedPeer<'a, TTrans, TInEvent, TOutEvent, THandler, TConnInfo, TPeerId>,
event: TInEvent,
2020-07-27 20:27:33 +00:00
cx: &mut Context<'_>,
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
) -> Option<(TInEvent, SmallVec<[ConnectionId; 10]>)>
where
TTrans: Transport,
TInEvent: Clone,
THandler: IntoConnectionHandler<TConnInfo>,
TPeerId: Eq + Hash + Clone,
TConnInfo: ConnectionInfo<PeerId = TPeerId>
{
if ids.len() == 1 {
if let Some(mut conn) = peer.connection(ids[0]) {
return notify_one(&mut conn, event, cx).map(|e| (e, ids))
}
}
let mut pending = SmallVec::new();
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(Ok(())) => {
// Can now only fail due to the connection suddenly closing,
// which we ignore.
let _ = conn.notify_handler(event.clone());
},
Poll::Ready(Err(())) => {} // connection is closing
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() {
return Some((event, pending))
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
}
None
}
impl<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo> Stream for
ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo>
where TBehaviour: NetworkBehaviour<ProtocolsHandler = THandler>,
THandler: IntoProtocolsHandler + Send + 'static,
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
TInEvent: Clone + Send + 'static,
TOutEvent: Send + 'static,
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
THandler::Handler: ProtocolsHandler<InEvent = TInEvent, OutEvent = TOutEvent>,
TConnInfo: ConnectionInfo<PeerId = PeerId> + fmt::Debug + Clone + Send + 'static,
{
type Item = TBehaviour::OutEvent;
2020-07-27 20:27:33 +00:00
fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
loop {
let event = futures::ready!(ExpandedSwarm::poll_next_event(self.as_mut(), cx));
if let SwarmEvent::Behaviour(event) = event {
return Poll::Ready(Some(event));
}
}
}
}
/// the stream of behaviour events never terminates, so we can implement fused for 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
impl<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo> FusedStream for
ExpandedSwarm<TBehaviour, TInEvent, TOutEvent, THandler, TConnInfo>
where TBehaviour: NetworkBehaviour<ProtocolsHandler = THandler>,
THandler: IntoProtocolsHandler + Send + 'static,
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
TInEvent: Clone + Send + 'static,
TOutEvent: Send + 'static,
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
THandler::Handler: ProtocolsHandler<InEvent = TInEvent, OutEvent = TOutEvent>,
TConnInfo: ConnectionInfo<PeerId = PeerId> + fmt::Debug + Clone + Send + 'static,
{
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 {
self.local_peer_id
}
}
/// A `SwarmBuilder` provides an API for configuring and constructing a `Swarm`,
/// including the underlying [`Network`].
pub struct SwarmBuilder<TBehaviour, TConnInfo> {
local_peer_id: PeerId,
transport: transport::Boxed<(TConnInfo, StreamMuxerBox)>,
behaviour: TBehaviour,
network_config: NetworkConfig,
}
impl<TBehaviour, TConnInfo> SwarmBuilder<TBehaviour, TConnInfo>
where TBehaviour: NetworkBehaviour,
TConnInfo: ConnectionInfo<PeerId = PeerId> + fmt::Debug + Clone + Send + 'static,
{
/// 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<(TConnInfo, StreamMuxerBox)>,
behaviour: TBehaviour,
local_peer_id: PeerId
) -> Self {
SwarmBuilder {
local_peer_id,
transport: transport,
behaviour,
network_config: Default::default(),
}
}
/// 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.set_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.set_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.set_connection_event_buffer_size(n);
self
}
/// Configures a limit for the number of simultaneous incoming
/// connection attempts.
pub fn incoming_connection_limit(mut self, n: usize) -> Self {
self.network_config.set_incoming_limit(n);
self
}
/// Configures a limit for the number of simultaneous outgoing
/// connection attempts.
pub fn outgoing_connection_limit(mut self, n: usize) -> Self {
self.network_config.set_outgoing_limit(n);
self
}
/// Configures a limit for the number of simultaneous
/// established connections per peer.
pub fn peer_connection_limit(mut self, n: usize) -> Self {
self.network_config.set_established_per_peer_limit(n);
self
}
/// Builds a `Swarm` with the current configuration.
pub fn build(mut self) -> Swarm<TBehaviour, TConnInfo> {
let supported_protocols = self.behaviour
.new_handler()
.inbound_protocol()
.protocol_info()
.into_iter()
.map(|info| info.protocol_name().to_vec())
.collect();
let mut network_cfg = self.network_config;
// If no executor has been explicitly configured, try to set up a thread pool.
if network_cfg.executor().is_none() {
match ThreadPoolBuilder::new()
.name_prefix("libp2p-swarm-task-")
.create()
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
{
Ok(tp) => {
network_cfg.set_executor(Box::new(move |f| tp.spawn_ok(f)));
},
Err(err) => log::warn!("Failed to create executor thread pool: {:?}", err)
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 network = Network::new(self.transport, self.local_peer_id, network_cfg);
2019-04-07 18:34:14 -03:00
ExpandedSwarm {
network,
behaviour: self.behaviour,
supported_protocols,
listened_addrs: SmallVec::new(),
external_addrs: Addresses::default(),
banned_peers: HashSet::new(),
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: None
}
}
}
/// The possible failures of [`ExpandedSwarm::dial`].
#[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),
/// [`NetworkBehaviour::addresses_of_peer`] returned no addresses
/// for the peer to dial.
NoAddresses
}
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::Banned => write!(f, "Dial error: peer is banned.")
}
}
}
impl error::Error for DialError {
fn source(&self) -> Option<&(dyn error::Error + 'static)> {
match self {
DialError::ConnectionLimit(err) => Some(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 => None,
DialError::Banned => None
}
}
}
/// Dummy implementation of [`NetworkBehaviour`] that doesn't do anything.
#[derive(Clone, Default)]
pub struct DummyBehaviour {
}
impl NetworkBehaviour for DummyBehaviour {
type ProtocolsHandler = protocols_handler::DummyProtocolsHandler;
type OutEvent = void::Void;
fn new_handler(&mut self) -> Self::ProtocolsHandler {
protocols_handler::DummyProtocolsHandler::default()
}
fn addresses_of_peer(&mut self, _: &PeerId) -> Vec<Multiaddr> {
Vec::new()
}
fn inject_connected(&mut self, _: &PeerId) {}
fn inject_connection_established(&mut self, _: &PeerId, _: &ConnectionId, _: &ConnectedPoint) {}
fn inject_disconnected(&mut self, _: &PeerId) {}
fn inject_connection_closed(&mut self, _: &PeerId, _: &ConnectionId, _: &ConnectedPoint) {}
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
fn inject_event(&mut self, _: PeerId, _: ConnectionId,
_: <Self::ProtocolsHandler as ProtocolsHandler>::OutEvent) {}
2020-07-27 20:27:33 +00:00
fn poll(&mut self, _: &mut Context<'_>, _: &mut impl PollParameters) ->
Poll<NetworkBehaviourAction<<Self::ProtocolsHandler as
ProtocolsHandler>::InEvent, Self::OutEvent>>
{
Poll::Pending
}
}
#[cfg(test)]
mod tests {
[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::{MockBehaviour, CallTraceBehaviour};
use futures::{future, executor};
use libp2p_core::{
identity,
[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,
multiaddr,
transport
};
use libp2p_noise as noise;
[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 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
fn new_test_swarm<T, O>(handler_proto: T) -> Swarm<CallTraceBehaviour<MockBehaviour<T, O>>>
where
T: ProtocolsHandler + Clone,
T::OutEvent: Clone,
O: Send + 'static
{
let id_keys = identity::Keypair::generate_ed25519();
let pubkey = id_keys.public();
let noise_keys = noise::Keypair::<noise::X25519Spec>::new().into_authentic(&id_keys).unwrap();
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(noise::NoiseConfig::xx(noise_keys).into_authenticated())
.multiplex(libp2p_mplex::MplexConfig::new())
.boxed();
let behaviour = CallTraceBehaviour::new(MockBehaviour::new(handler_proto));
SwarmBuilder::new(transport, behaviour, pubkey.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
}
/// Establishes a number of connections between two peers,
/// 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.
#[test]
fn test_connect_disconnect_ban() {
// Since the test does not try to open any substreams, we can
// use the dummy protocols handler.
let mut handler_proto = DummyProtocolsHandler::default();
handler_proto.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();
Swarm::listen_on(&mut swarm1, addr1.clone().into()).unwrap();
Swarm::listen_on(&mut swarm2, addr2.clone().into()).unwrap();
// Test execution state. Connection => Disconnecting => Connecting.
enum State {
Connecting,
Disconnecting,
}
let swarm1_id = Swarm::local_peer_id(&swarm1).clone();
let mut banned = false;
let mut unbanned = false;
let num_connections = 10;
for _ in 0 .. num_connections {
Swarm::dial_addr(&mut swarm1, 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 => {
for s in &[&swarm1, &swarm2] {
if s.behaviour.inject_connection_established.len() > 0 {
assert_eq!(s.behaviour.inject_connected.len(), 1);
} else {
assert_eq!(s.behaviour.inject_connected.len(), 0);
}
assert!(s.behaviour.inject_connection_closed.len() == 0);
assert!(s.behaviour.inject_disconnected.len() == 0);
}
if [&swarm1, &swarm2].iter().all(|s| {
s.behaviour.inject_connection_established.len() == num_connections
}) {
if banned {
return Poll::Ready(())
}
Swarm::ban_peer_id(&mut swarm2, swarm1_id.clone());
swarm1.behaviour.reset();
swarm2.behaviour.reset();
banned = true;
state = State::Disconnecting;
}
}
State::Disconnecting => {
for s in &[&swarm1, &swarm2] {
if s.behaviour.inject_connection_closed.len() < num_connections {
assert_eq!(s.behaviour.inject_disconnected.len(), 0);
} else {
assert_eq!(s.behaviour.inject_disconnected.len(), 1);
}
assert_eq!(s.behaviour.inject_connection_established.len(), 0);
assert_eq!(s.behaviour.inject_connected.len(), 0);
}
if [&swarm1, &swarm2].iter().all(|s| {
s.behaviour.inject_connection_closed.len() == num_connections
}) {
if unbanned {
return Poll::Ready(())
}
// Unban the first peer and reconnect.
Swarm::unban_peer_id(&mut swarm2, swarm1_id.clone());
swarm1.behaviour.reset();
swarm2.behaviour.reset();
unbanned = true;
for _ in 0 .. num_connections {
Swarm::dial_addr(&mut swarm2, addr1.clone()).unwrap();
}
state = State::Connecting;
}
}
}
if poll1.is_pending() && poll2.is_pending() {
return Poll::Pending
}
}
}))
}
}