* Add listener ID and error event.
Report listener errors to client code so they are aware that an error
occurred within a listener. By default we continue to poll listeners
which produced an error, but clients can remove listeners by ID.
* tcp: Report errors.
Instead of silently waiting after errors we return all errors, but pause
after each error, before continuing.
* Add a test.
To ease testing, `Listener` is made generic and we test that no values
and errors are lost. Elapsed time between item generation is not
measured.
* Support the new methods in core-derive.
* Address review concerns.
* Remove `Display` impl of `ListenerId`.
* Add 'static bound to `on_listener_error` error.
* Remove tokio-codec dependency from multistream-select.
In preparation for the eventual switch from tokio to std futures.
Includes some initial refactoring in preparation for further work
in the context of https://github.com/libp2p/rust-libp2p/issues/659.
* Reduce default buffer sizes.
* Allow more than one frame to be buffered for sending.
* Doc tweaks.
* Remove superfluous (duplicated) Message types.
* Reduce roundtrips in multistream-select negotiation.
1. Enable 0-RTT: If the dialer only supports a single protocol, it can send
protocol data (e.g. the actual application request) together with
the multistream-select header and protocol proposal. Similarly,
if the listener supports a proposed protocol, it can send protocol
data (e.g. the actual application response) together with the
multistream-select header and protocol confirmation.
2. In general, the dialer "settles on" an expected protocol as soon
as it runs out of alternatives. Furthermore, both dialer and listener
do not immediately flush the final protocol confirmation, allowing it
to be sent together with application protocol data. Attempts to read
from the negotiated I/O stream implicitly flushes any pending data.
3. A clean / graceful shutdown of an I/O stream always completes protocol
negotiation.
The publich API of multistream-select changed slightly, requiring both
AsyncRead and AsyncWrite bounds for async reading and writing due to
the implicit buffering and "lazy" negotiation. The error types have
also been changed, but they were not previously fully exported.
Includes some general refactoring with simplifications and some more tests,
e.g. there was an edge case relating to a possible ambiguity when parsing
multistream-select protocol messages.
* Further missing commentary.
* Remove unused test dependency.
* Adjust commentary.
* Cleanup NegotiatedComplete::poll()
* Fix deflate protocol tests.
* Stabilise network_simult test.
The test implicitly relied on "slow" connection establishment
in order to have a sufficient probability of passing.
With the removal of roundtrips in multistream-select, it is now
more likely that within the up to 50ms duration between swarm1
and swarm2 dialing, the connection is already established, causing
the expectation of step == 1 to fail when receiving a Connected event,
since the step may then still be 0.
This commit aims to avoid these spurious errors by detecting runs
during which a connection is established "too quickly", repeating
the test run.
It still seems theoretically possible that, if connections are always
established "too quickly", the test runs forever. However, given that
the delta between swarm1 and swarm2 dialing is 0-50ms and that the
TCP transport is used, that seems probabilistically unlikely.
Nevertheless, the purpose of the artificial dialing delay between
swarm1 and swarm2 should be re-evaluated and possibly at least
the maximum delay further reduced.
* Complete negotiation between upgrades in libp2p-core.
While multistream-select, as a standalone library and providing
an API at the granularity of a single negotiation, supports
lazy negotiation (and in particular 0-RTT negotiation), in the
context of libp2p-core where any number of negotiations are
composed generically within the concept of composable "upgrades",
it is necessary to wait for protocol negotiation between upgrades
to complete.
* Clarify docs. Simplify listener upgrades.
Since reading from a Negotiated I/O stream implicitly flushes any pending
negotiation data, there is no pitfall involved in not waiting for completion.
* Remove pending RPCs on query completion.
Ensure that any still pending RPCs related to a query are removed
once the query terminates (successfully or through timeout) by
scoping pending RPCs to the lifetime of a query.
* Cleanup.
* Remove tokio-codec dependency from multistream-select.
In preparation for the eventual switch from tokio to std futures.
Includes some initial refactoring in preparation for further work
in the context of https://github.com/libp2p/rust-libp2p/issues/659.
* Reduce default buffer sizes.
* Allow more than one frame to be buffered for sending.
* Doc tweaks.
* Remove superfluous (duplicated) Message types.
Addresses https://github.com/libp2p/rust-libp2p/issues/1206 by always
registering the current task before calling poll_*_notify functions.
This is in the same spirit as the corresponding fix for yamux
in https://github.com/paritytech/yamux/pull/54.
Also adds missing registration of the current task in close()
and flush_all(), which have been observed to cause stalls
when trying to do a graceful connection shutdown / close.
* core/src/translation: Add unit tests
* core/src/translation: Support dns4 and dns6
Add dns4 and dns6 as valid protocol replacements for the origin address
to construct external addresses of a given node.
* core/nodes/network: %s/nat_translation/address_translation/
When receiving an observed address on a tcp connection that we initiated, the
observed address contains our tcp dial port, not our tcp listen port. We know
which port we are listening on, thereby we can replace the port within the
observed address.
When receiving an observed address on a tcp connection that we did **not**
initiated, the observed address should contain our listening port. In case it
differs from our listening port there might be a NAT along the path.
With the above in mind, the function name `nat_translation` is misleading.
* Somewhat complete the implementation of Kademlia records.
This commit relates to [libp2p-146] and [libp2p-1089].
* All records expire (by default, configurable).
* Provider records are also stored in the RecordStore, and the RecordStore
API extended.
* Background jobs for periodic (re-)replication and (re-)publication
of records. Regular (value-)records are subject to re-replication and
re-publication as per standard Kademlia. Provider records are only
subject to re-publication.
* For standard Kademlia value lookups (quorum = 1), the record is cached
at the closest peer to the key that did not return the value, as per
standard Kademlia.
* Expiration times of regular (value-)records is computed exponentially
inversely proportional to the number of nodes between the local node
and the closest node known to the key (beyond the k closest), as per
standard Kademlia.
The protobuf messages are extended with two fields: `ttl` and `publisher`
in order to implement the different semantics of re-replication (by any
of the k closest peers to the key, not affecting expiry) and re-publication
(by the original publisher, resetting the expiry). This is not done yet in
other libp2p Kademlia implementations, see e.g. [libp2p-go-323]. The new protobuf fields
have been given somewhat unique identifiers to prevent future collision.
Similarly, periodic re-publication of provider records does not seem to
be done yet in other implementations, see e.g. [libp2p-js-98].
[libp2p-146]: https://github.com/libp2p/rust-libp2p/issues/146
[libp2p-1089]: https://github.com/libp2p/rust-libp2p/issues/1089
[libp2p-go-323]: https://github.com/libp2p/go-libp2p-kad-dht/issues/323
[libp2p-js-98]: https://github.com/libp2p/js-libp2p-kad-dht/issues/98
* Tweak kad-ipfs example.
* Add missing files.
* Ensure new delays are polled immediately.
To ensure task notification, since `NotReady` is returned right after.
* Fix ipfs-kad example and use wasm_timer.
* Small cleanup.
* Incorporate some feedback.
* Adjustments after rebase.
* Distinguish events further.
In order for a user to easily distinguish the result of e.g.
a `put_record` operation from the result of a later republication,
different event constructors are used. Furthermore, for now,
re-replication and "caching" of records (at the closest peer to
the key that did not return a value during a successful lookup)
do not yield events for now as they are less interesting.
* Speed up tests for CI.
* Small refinements and more documentation.
* Guard a node against overriding records for which it considers
itself to be the publisher.
* Document the jobs module more extensively.
* More inline docs around removal of "unreachable" addresses.
* Remove wildcard re-exports.
* Use NonZeroUsize for the constants.
* Re-add method lost on merge.
* Add missing 'pub'.
* Further increase the timeout in the ipfs-kad example.
* Readd log dependency to libp2p-kad.
* Simplify RecordStore API slightly.
* Some more commentary.
* Change Addresses::remove to return Result<(),()>.
Change the semantics of `Addresses::remove` so that the error case
is unambiguous, instead of the success case. Use the `Result` for
clearer semantics to that effect.
* Add some documentation to .
* Rename RawSwarm* to Network*.
To complete the cut performed in [1].
The only remaining mention of a "swarm" in libp2p-core is in some tests
which actually depend on libp2p-swarm.
[1]: https://github.com/libp2p/rust-libp2p/pull/1188
* Post-merge corrections.
The example typically runs into a lot of connection timeouts,
which may cause the query to time out. A query timeout currently
results in the example to be considered failed, but the example
should only be considered failed if no closest peers are found,
whether the query timed out or not.
* 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.
* Adds a retain method to kademlia.
* Add a store_mut getter to Kademlia
* Removes a blank line
* Changes store_mut comment appropriately
* Fixes build
* Return a type, not a trait
* Address some TODOs, refactor queries and public API.
The following left-over issues are addressed:
* The key for FIND_NODE requests is generalised to any Multihash,
instead of just peer IDs.
* All queries get a (configurable) timeout.
* Finishing queries as soon as enough results have been received is simplified
to avoid code duplication.
* No more panics in provider-API-related code paths. The provider API is
however still untested and (I think) still incomplete (e.g. expiration
of provider records).
* Numerous smaller TODOs encountered in the code.
The following public API changes / additions are made:
* Introduce a `KademliaConfig` with new configuration options for
the replication factor and query timeouts.
* Rename `find_node` to `get_closest_peers`.
* Rename `get_value` to `get_record` and `put_value` to `put_record`,
introducing a `Quorum` parameter for both functions, replacing the
existing `num_results` parameter with clearer semantics.
* Rename `add_providing` to `start_providing` and `remove_providing`
to `stop_providing`.
* Add a `bootstrap` function that implements a (almost) standard
Kademlia bootstrapping procedure.
* Rename `KademliaOut` to `KademliaEvent` with an updated list of
constructors (some renaming). All events that report query results
now report a `Result` to uniformly permit reporting of errors.
The following refactorings are made:
* Introduce some constants.
* Consolidate `query.rs` and `write.rs` behind a common query interface
to reduce duplication and facilitate better code reuse, introducing
the notion of a query peer iterator. `query/peers/closest.rs`
contains the code that was formerly in `query.rs`. `query/peers/fixed.rs` contains
a modified variant of `write.rs` (which is removed). The new `query.rs`
provides an interface for working with a collection of queries, taking
over some code from `behaviour.rs`.
* Reduce code duplication in tests and use the current_thread runtime for
polling swarms to avoid spurious errors in the test output due to aborted
connections when a test finishes prematurely (e.g. because a quorum of
results has been collected).
* Some additions / improvements to the existing tests.
* Fix test.
* Fix rebase.
* Tweak kad-ipfs example.
* Incorporate some feedback.
* Provide easy access and conversion to keys in error results.
multihash: Use `Bytes` instead of `Vec<u8>` internally.
To improve the efficiency of cloning multi-hashes (e.g. as the
representation of `PeerId`s), this PR replaces the `Vec<u8>`
representation with `Bytes`. The API is kept backwards
compatible and does not leak the representation type.
The transport should be able to continue processing other connections.
An error to determine a socket's address is not a fatal condition but
may happen when a connection is immediately reset after being
established. By the time the programme asks for the remote address, the
socket may already be gone.
Fixes#1182.
* Make `<SubstreamRef as AsyncWrite>::shutdown` imply flush
* Use try_ready
* Apply suggestions from code review
Co-Authored-By: Toralf Wittner <tw@dtex.org>
Refactoring of iterative queries (`query.rs`) to improve both
correctness and performance (for larger DHTs):
Correctness:
1. Queries no longer terminate prematurely due to counting results
from peers farther from the target while results from closer
peers are still pending. (#1105).
2. Queries no longer ignore reported closer peers that are not duplicates
just because they are currently not among the `num_results` closest.
The currently `max_results` closest may contain peers marked as failed
or pending / waiting. Hence all reported closer peers that are not
duplicates must be considered candidates that may still end up
among the `num_results` closest that successfully responded.
3. Bounded parallelism based on the `active_counter` was not working
correctly, as new (not yet contacted) peers closer to the target
may be discovered at any time and thus appear in `closer_peers`
before the already active / pending peers.
4. The `Frozen` query mechanism allowed all remaining not-yet contacted
peers to be contacted, but their results were discarded, because
`inject_rpc_result` would only incorporate results while the
query is `Iterating`. The `Frozen` state has been reworked into
a `Stalled` state that implements a slightly more permissive
variant of the following from the paper / specs: "If a round of
FIND_NODEs fails to return a node any closer than the closest
already seen, the initiator resends the FIND_NODE to all of the
k closest nodes it has not already queried.". Importantly, though
not explicitly mentioned, the query can move back to `Iterating`
if it makes further progress again as a result of these requests.
The `Stalled` state thus allows (temporarily) higher parallelism
in an effort to make progress and bring the query to an end.
Performance:
1. Repeated distance calculations between the same peers and the
target is avoided.
2. Enabled by #1108, use of a more appropriate data structure (`BTreeMap`) for
the incrementally updated list of closer peers. The data structure needs
efficient lookups (to avoid duplicates) and insertions at any position,
both of which large(r) vectors are not that good at. Unscientific benchmarks
showed a ~40-60% improvement in somewhat pathological scenarios with at least
20 healthy nodes, each possibly returning a distinct list of closer 20 peers
to the requestor. A previous assumption may have been that the vector always
stays very small, but that is not the case in larger clusters: Even if the
lists of closer peers reported by the 20 contacted peers are heavily overlapping,
typically a lot more than 20 peers have to be (at least temporarily) considered
as closest peers until the query completes. See also issue (2) above.
New tests are added for:
* Query termination conditions.
* Bounded parallelism.
* Absence of duplicates.
* Update soketto and enable deflate extension.
* libp2p-deflate and libp2p-websocket share flate2.
Due to the way feature resolution works in cargo today, the `deflate`
feature of `soketto` will include `flate2` with feature `zlib` which is
then also active for the `flate2` that `libp2p-deflate` depends on. This
leads to compilation failures for WASM targets. This PR therefore moves
libp2p-deflate to the crates which are not available on WASM.
* initial implementation of the records
* move to multihash keys
* correctly process query results
* comments and formatting
* correctly return closer_peers in query
* checking wrong peer id in test
* Apply suggestions from code review
Co-Authored-By: Roman Borschel <romanb@users.noreply.github.com>
* Fix changes from suggestions
* Send responses to PUT_VALUE requests
* Shortcut in get_value
* Update protocols/kad/src/behaviour.rs
Co-Authored-By: Roman Borschel <romanb@users.noreply.github.com>
* Revert "Update protocols/kad/src/behaviour.rs"
This reverts commit 579ce742a7f4c94587f1e1f0866d2a3a37418efb.
* Remove duplicate insertion
* Adds a record to a PUT_VALUE response
* Fix a racy put_value test
* Store value ourselves only if we are in K closest
* Abstract over storage
* Revert "Abstract over storage": bad take
This reverts commit eaebf5b6d915712eaf3b05929577fdf697f204d8.
* Abstract over records storage using hashmap as default
* Constructor for custom records
* New Record type and its traits
* Fix outdated storage name
* Fixes returning an event
* Change FindNodeReq key type to Multihash
* WriteState for a second stage of a PUT_VALUE request
* GET_VALUE should not have a record
* Refactor a match arm
* Add successes and failures counters to PutValueRes
* If value is found no need to return closer peers
* Remove a custo storage from tests
* Rename a test to get_value_not_found
* Adds a TODO to change FindNode request key to Multihash
Co-Authored-By: Roman Borschel <romanb@users.noreply.github.com>
* Move MemoryRecordStorage to record.rs
* Return a Cow-ed Record from get
* Fix incorrect GET_VALUE parsing
* Various fixes with review
* Fixes get_value_not_found
* Fix peerids names in test
* another fix
* PutValue correctly distributes values
* Simplify the test
* Check that results are actually the closest
* Reverts changes to tests
* Fix the test topology and checking the results
* Run put_value test ten times
* Adds a get_value test
* Apply suggestions from code review
Co-Authored-By: Roman Borschel <romanb@users.noreply.github.com>
* Make Record fields public
* Moves WriteState to write.rs
* A couple of minor fixes
* Another few fixes of review
* Simplify the put_value test
* Dont synchronously return an error from put_value
* Formatting fixes and comments
* Collect a bunch of results
* Take exactly as much elements as neede
* Check if the peer is still connected
* Adds a multiple GetValueResults results number test
* Unnecessary mut iterators in put_value
* Ask for num_results in get_value
* Dont allocate twice in get_value
* Dont count same errored peer multiple times
* Apply suggestions from code review
Co-Authored-By: Roman Borschel <romanb@users.noreply.github.com>
* Fix another review
* Apply suggestions from code review
Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com>
* Bring back FromIterator and improve a panic message
* Update protocols/kad/src/behaviour.rs
Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com>
* Begin reimplementing the websocket transport.
* Add TLS support.
* Add support for redirects during handshake.
* Cosmetics.
* Remove unused error cases in tls module.
Left-overs from a previous implementation.
* No libp2p-websocket for wasm targets.
* Change tls::Config to make the server optional.
* Update transports/websocket/src/lib.rs
Co-Authored-By: Pierre Krieger <pierre.krieger1708@gmail.com>
* Duplicate config methods.
As per PR review feedback.
* Some improvements to the docs of NetworkBehaviour
* Apply suggestions from code review
Co-Authored-By: Roman Borschel <romanb@users.noreply.github.com>
* Map all 127.0.0.0/8 addresses to 127.0.0.1.
Since every local socket address in the 127.0.0.0/8 space is looped back
to 127.0.0.1/32 we should only have to report the later as the listen
address. For other addresses we still attempt to discover host addresses
when we encounter an unknown local address. We now also check that after
the host addresses have been reset that the address is now found,
otherwise we produce an error.
* Change listen address lookup.
Perform multiple steps:
1. Check for exact address match.
2. Else consider netmask and check for containment.
3. Else re-check host addresses and try 1 & 2 again.
4. Else report an error.
* Small fixes.
* Test and improve prefix_len.
* Simplify and inline the prefix_len logic.
