Previously, the `libp2p-ping` module came with a policy to close a connection after X failed pings. This is only one of many possible policies on how users would want to do connection management.
We remove this policy without a replacement. If users wish to restore this functionality, they can easily implement such policy themselves: The default value of `max_failures` was 1. To restore the previous functionality users can simply close the connection upon the first received ping error.
In this same patch, we also simplify the API of `ping::Event` by removing the layer of `ping::Success` and instead reporting the RTT to the peer directly.
Related: #3591.
Pull-Request: #3947.
Previously, a `NetworkBehaviour` could report an `AddressScore` for an external address. This score was a `u32` and addresses would be ranked amongst those.
In reality, an address is either confirmed to be publicly reachable (via a protocol such as AutoNAT) or merely represents a candidate that might be an external address. In a way, addresses are guilty (private) until proven innocent (publicly reachable).
When a `NetworkBehaviour` reports an address candidate, we perform address translation on it to potentially correct for ephemeral ports of TCP. These candidates are then injected back into the `NetworkBehaviour`. Protocols such as AutoNAT can use these addresses as a source for probing their NAT status. Once confirmed, they can emit a `ToSwarm::ExternalAddrConfirmed` event which again will be passed to all `NetworkBehaviour`s.
This simplified approach will allow us implement Kademlia's client-mode (https://github.com/libp2p/rust-libp2p/issues/2032) without additional configuration options: As soon as an address is reported as publicly reachable, we can activate server-mode for that connection.
Related: https://github.com/libp2p/rust-libp2p/pull/3877.
Related: https://github.com/libp2p/rust-libp2p/issues/3953.
Related: https://github.com/libp2p/rust-libp2p/issues/2032.
Related: https://github.com/libp2p/go-libp2p/issues/2229.
Co-authored-by: Max Inden <mail@max-inden.de>
Pull-Request: #3954.
Previously, the associated types on `NetworkBehaviour` and `ConnectionHandler` carried generic names like `InEvent` and `OutEvent`. These names are _correct_ in that `OutEvent`s are passed out and `InEvent`s are passed in but they don't help users understand how these types are used.
In theory, a `ConnectionHandler` could be used separately from `NetworkBehaviour`s but that is highly unlikely. Thus, we rename these associated types to indicate, where the message is going to be sent to:
- `NetworkBehaviour::OutEvent` is renamed to `ToSwarm`: It describes the message(s) a `NetworkBehaviour` can emit to the `Swarm`. The user is going to receive those in `SwarmEvent::Behaviour`.
- `ConnectionHandler::InEvent` is renamed to `FromBehaviour`: It describes the message(s) a `ConnectionHandler` can receive from its behaviour via `ConnectionHandler::on_swarm_event`. The `NetworkBehaviour` can send it via the `ToSwarm::NotifyHandler` command.
- `ConnectionHandler::OutEvent` is renamed to `ToBehaviour`: It describes the message(s) a `ConnectionHandler` can send back to the behaviour via the now also renamed `ConnectionHandlerEvent::NotifyBehaviour` (previously `ConnectionHandlerEvent::Custom`)
Resolves: #2854.
Pull-Request: #3848.
`Transport::listen_on` is an asynchronous operation. It returns immediately but the actual process of establishing a listening socket happens as part of `Transport::poll` which will return one or more `TransportEvent`s related to a particular `listen_on` call.
Currently, `listen_on` returns a `ListenerId` which allows the user of the `Transport` interface to correlate the events with a particular `listen_on` call. This "user" is the `Swarm` runtime. Currently, a user of libp2p establishes a new listening socket by talking to the `Swarm::listen_on` interface and it is not possible to do the same thing via the `NetworkBehaviour` trait.
Within the `NetworkBehaviour` trait, we emit _commands_ to the `Swarm` like `ToSwarm::Dial`. These commands don't have a "return value" like a synchronous function does and thus, if we were to add a `ToSwarm::ListenOn` command, it could not receive the `ListenerId` from the `Transport`.
To fix this and to be consistent with our [coding guidelines](https://github.com/libp2p/rust-libp2p/blob/master/docs/coding-guidelines.md#allow-correlating-asynchronous-responses-to-their-requests) we change the interface of `Transport::listen_on` to require the user to pass in a `ListenerId`. This will allow us to construct a command in a `NetworkBehaviour` that remembers this ID which enables precise tracking of which events containing a `ListenerId` correlate which a particular `listen_on` command.
This is especially important in the context of listening on wildcard addresses like `0.0.0.0` because we end up binding to multiple network interfaces and thus emit multiple events for a single `listen_on` call.
Pull-Request: #3567.
This patch tackles two things at once that are fairly intertwined:
1. There is no such thing as a "substream" in libp2p, the spec and other implementations only talk about "streams". We fix this by deprecating `NegotiatedSubstream`.
2. Previously, `NegotiatedSubstream` was a type alias that pointed to a type from `multistream-select`, effectively leaking the version of `multistream-select` to all dependencies of `libp2p-swarm`. We fix this by introducing a `Stream` newtype.
Resolves: #3759.
Related: #3748.
Pull-Request: #3912.
With this patch, implementations of `ConnectionHandler` (which are typically composed in a tree) can exchange information about the supported protocols of a remote with each other via `ConnectionHandlerEvent::ReportRemoteProtocols`. The provided `ProtocolSupport` enum can describe either additions or removals of the remote peer's protocols.
This information is aggregated in the connection and passed down to the `ConnectionHandler` via `ConnectionEvent::RemoteProtocolsChange`.
Similarly, if the listen protocols of a connection change, all `ConnectionHandler`s on the connection will be notified via `ConnectionEvent::LocalProtocolsChange`. This will allow us to eventually remove `PollParameters` from `NetworkBehaviour`.
This pattern allows protocols on a connection to communicate with each other. For example, protocols like identify can share the list of (supposedly) supported protocols by the remote with all other handlers. A protocol like kademlia can accurately add and remove a remote from its routing table as a result.
Resolves: #2680.
Related: #3124.
Pull-Request: #3651.
The currently provided `ConnectionHandlerUpgrErr` is very hard to use. Not only does it have a long name, it also features 3 levels of nesting which results in a lot of boilerplate. Last but not least, it exposes `multistream-select` as a dependency to all protocols.
We fix all of the above by renaming the type to `StreamUpgradeError` and flattening out its interface. Unrecoverable errors during protocol selection are hidden within the `Io` variant.
Related: #3759.
Pull-Request: #3882.
Previously, a protocol could be any sequence of bytes as long as it started with `/`. Now, we directly parse a protocol as `String` which enforces it to be valid UTF8.
To notify users of this change, we delete the `ProtocolName` trait. The new requirement is that users need to provide a type that implements `AsRef<str>`.
We also add a `StreamProtocol` newtype in `libp2p-swarm` which provides an easy way for users to ensure their protocol strings are compliant. The newtype enforces that protocol strings start with `/`. `StreamProtocol` also implements `AsRef<str>`, meaning users can directly use it in their upgrades.
`multistream-select` by itself only changes marginally with this patch. The only thing we enforce in the type-system is that protocols must implement `AsRef<str>`.
Resolves: #2831.
Pull-Request: #3746.
The `unreachable_pub` lint makes us aware of uses of `pub` that are not actually reachable from the crate root. This is considered good because it means reading a `pub` somewhere means it is actually public API. Some of our crates are quite large and keeping their entire API surface in your head is difficult.
We should strive for most items being `pub(crate)`. This lint helps us enforce that.
Pull-Request: #3735.
Previously, a `ConnectionHandler` was immediately requested from the `NetworkBehaviour` as soon as a new dial was initiated or a new incoming connection accepted.
With this patch, we delay the creation of the handler until the connection is actually established and fully upgraded, i.e authenticated and multiplexed.
As a consequence, `NetworkBehaviour::new_handler` is now deprecated in favor of a new set of callbacks:
- `NetworkBehaviour::handle_pending_inbound_connection`
- `NetworkBehaviour::handle_pending_outbound_connection`
- `NetworkBehaviour::handle_established_inbound_connection`
- `NetworkBehaviour::handle_established_outbound_connection`
All callbacks are fallible, allowing the `NetworkBehaviour` to abort the connection either immediately or after it is fully established. All callbacks also receive a `ConnectionId` parameter which uniquely identifies the connection. For example, in case a `NetworkBehaviour` issues a dial via `NetworkBehaviourAction::Dial`, it can unambiguously detect this dial in these lifecycle callbacks via the `ConnectionId`.
Finally, `NetworkBehaviour::handle_pending_outbound_connection` also replaces `NetworkBehaviour::addresses_of_peer` by allowing the behaviour to return more addresses to be used for the dial.
Resolves#2824.
Pull-Request: #3254.
We create the `ConnectionId` for the new connection as part of `DialOpts`. This allows `NetworkBehaviour`s to accurately track state regarding their own dial attempts.
This patch is the main enabler of https://github.com/libp2p/rust-libp2p/pull/3254. Removing the `handler` field will allow us to deprecate the `NetworkBehaviour::new_handler` function in favor of four new ones that give more control over the connection lifecycle.
Previously, we used the full reference to the `OutEvent` of the `ConnectionHandler` in all implementations of `NetworkBehaviour`. Not only is this very verbose, it is also more brittle to changes. With the current implementation plan for #2824, we will be removing the `IntoConnectionHandler` abstraction. Using a type-alias to refer to the `OutEvent` makes the migration much easier.
As I do frequently, I corrected for the latest clippy warnings. This will make sure the CI won't complain in the future. We could automate this btw and maybe run the nightly version of clippy.
Previously, we had one callback for each kind of message that a `ConnectionHandler` would receive from either its `NetworkBehaviour` or the connection itself.
With this patch, we combine these functions, resulting in two callbacks:
- `on_behaviour_event`
- `on_connection_event`
Resolves#3080.
Previously, the `DummyConnectionHandler` offered a "keep alive" functionality,
i.e. it allowed users to set the value of what is returned from
`ConnectionHandler::keep_alive`. This handler is primarily used in tests or
`NetworkBehaviour`s that don't open any connections (like mDNS). In all of these
cases, it is statically known whether we want to keep connections alive. As
such, this functionality is better represented by a static
`KeepAliveConnectionHandler` that always returns `KeepAlive::Yes` and a
`DummyConnectionHandler` that always returns `KeepAlive::No`.
To follow the naming conventions described in
https://github.com/libp2p/rust-libp2p/issues/2217, we introduce a top-level
`keep_alive` and `dummy` behaviour in `libp2p-swarm` that contains both the
`NetworkBehaviour` and `ConnectionHandler` implementation for either case.
* misc/metrics: Explicitly delegate event recording to each recorder
This allows delegating a single event to multiple `Recorder`s. That enables e.g. the
`identify::Metrics` `Recorder` to act both on `IdentifyEvent` and `SwarmEvent`. The latter enables
it to garbage collect per peer data on disconnects.
* protocols/dcutr: Expose PROTOCOL_NAME
* protocols/identify: Expose PROTOCOL_NAME and PUSH_PROTOCOL_NAME
* protocols/ping: Expose PROTOCOL_NAME
* protocols/relay: Expose HOP_PROTOCOL_NAME and STOP_PROTOCOL_NAME
* misc/metrics: Track # connected nodes supporting specific protocol
An example metric exposed with this patch:
```
libp2p_identify_protocols{protocol="/ipfs/ping/1.0.0"} 10
```
This implies that 10 of the currently connected nodes support the ping protocol.
Remove the concept of individual `Transport::Listener` streams from `Transport`.
Instead the `Transport` is polled directly via `Transport::poll`. The
`Transport` is now responsible for driving its listeners.
A `ProtocolsHandler`, now `ConnectionHandler`, handels a connection, not
a protocol. Thus the name `CONNECTIONHandler` is more appropriate.
Next to the rename of `ProtocolsHandler` this commit renames the `mod
protocols_handler` to `mod handler`. Finally all combinators (e.g.
`ProtocolsHandlerSelect`) are renamed appropriately.
Changes needed to get libp2p to run via `wasm32-unknown-unknown` in the browser
(both main thread and inside web workers).
Replaces wasm-timer with futures-timer and instant.
Co-authored-by: Oliver Wangler <oliver@wngr.de>
Require `NetworkBehaviourAction::{DialPeer,DialAddress}` to contain a
`ProtocolsHandler`. This allows a behaviour to attach custom state to its
handler. The behaviour would no longer need to track this state separately
during connection establishment, thus reducing state required in a behaviour.
E.g. in the case of `libp2p-kad` the behaviour can include a `GetRecord` request
in its handler, or e.g. in the case of `libp2p-request-response` the behaviour
can include the first request in the handler.
Return `ProtocolsHandler` on connection error and close. This allows a behaviour
to extract its custom state previously included in the handler on connection
failure and connection closing. E.g. in the case of `libp2p-kad` the behaviour
could extract the attached `GetRecord` from the handler of the failed connection
and then start another connection attempt with a new handler with the same
`GetRecord` or bubble up an error to the user.
Co-authored-by: Thomas Eizinger <thomas@eizinger.io>
Don't close connection if ping protocol is unsupported by remote. Previously, a
failed protocol negotation for ping caused a force close of the connection. As a
result, all nodes in a network had to support ping. To allow networks where some
nodes don't support ping, we now emit `PingFailure::Unsupported` once for every
connection on which ping is not supported.
Co-authored-by: Max Inden <mail@max-inden.de>
Not all implementations of `NetworkBehaviour` need all callbacks.
We've have been adding new callbacks with default implementations
for a while now. There is no reason the initial ones cannot also
be defaulted, thus making it easier create new implementations.
Co-authored-by: Max Inden <mail@max-inden.de>
* Refactor the ping protocol.
Such that pings are sent over a single substream, as it is
done in other libp2p implementations. Note that, since each
peer sends its pings over a single, dedicated substream,
every peer that participates in the protocol has effectively
two open substreams.
* Cleanup
* Update ping changelog.
* Update protocols/ping/src/protocol.rs
Co-authored-by: Max Inden <mail@max-inden.de>
Co-authored-by: Max Inden <mail@max-inden.de>
although 32 is prefect fine in our case, it would be consistent to use the const value PING_SIZE.
Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com>
* [libp2p-swarm] Make the multiple connections per peer first-class.
This commit makes the notion of multiple connections per peer
first-class in the API of libp2p-swarm, introducing the new
callbacks `inject_connection_established` and
`inject_connection_closed`. The `endpoint` parameter from
`inject_connected` and `inject_disconnected` is removed,
since the first connection to open may not be the last
connection to close, i.e. it cannot be guaranteed,
as was previously the case, that the endpoints passed
to these callbacks match up.
* Have identify track all addresses.
So that identify requests can be answered with the correct
observed address of the connection on which the request
arrives.
* Cleanup
* Cleanup
* Improve the `Peer` state API.
* Remove connection ID from `SwarmEvent::Dialing`.
* Mark `DialPeerCondition` non-exhaustive.
* Re-encapsulate `NetworkConfig`.
To retain the possibility of not re-exposing all
network configuration choices, thereby providing
a more convenient API on the \`SwarmBuilder\`.
* Rework Swarm::dial API.
* Update CHANGELOG.
* Doc formatting tweaks.
* 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>
* Fix broken links in rustdoc
This fixes all of the rustdoc warnings on nightly.
* Check documentation intra-link
* Fix config
* Fix bad indent
* Make nightly explicit
* More links fixes
* Fix link broken after master merge
Co-authored-by: Demi Obenour <48690212+DemiMarie-parity@users.noreply.github.com>
* Simplify trait bounds requirements
* More work
* Moar
* Finish
* Fix final tests
* More simplification
* Use separate traits for Inbound/Outbound
* Update gossipsub and remove warnings
* Add documentation to swarm
* Remove BoxSubstream
* Fix tests not compiling
* Fix stack overflow
* Address concerns
* For some reason my IDE ignored libp2p-kad
- Pin `futures_codec` to version 0.3.3 as later versions require
at least bytes-0.5 which he have not upgraded to yet.
- Replace `futures::executor::block_on` with `async_std::task::block_on`
where `async-std` is already a dependency to work around an issue with
`park`/`unpark` behaviour.
- Use the published version of `quicksink`.
