* Set peer's `last_active` even if they are being throttled
* Make sure not to break out of loop in wire module
* Add some more logging
* Use different variable for session filtering to make it clearer
Ref announcements contain the namespace (NID) and the SHA of the new
`rad/sigrefs`. This information can be used by the fetch protocol to
fetch only those `Oid`s for those namespaces.
Add a new stage `SigrefsAnnouncement` that captures this behaviour by
only asking for the advertised `Oid`s and updating the corresponding
`rad/sigrefs` for each namespace. Note that this is a special case of
the `SpecialRefs` stage and so the protocol chooses between these two
by checking if there was announcement passed through. The following
stage, `DataRefs`, stays the same.
Note that this stage can only be activated via a `pull`, and `clone`
stays the same.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
The RefsAnnouncement previous announced its complete set of
`rad/sigrefs`. This was useful in that verification can occur on the
receiveing side immediately by verifying the refs and proceeding with
the fetch -- note that it did nothing with the ref information it
received.
This can be minimised to only advertising the namespace, i.e. the
NID, and the new SHA of the `rad/sigrefs`. This is represented by a
new message type, `RefsAt`, which contains the NID and the Oid.
For checking if an announcement is stale or fresh, the receiving
side can check the existence of the SHA in its Git repository.
The tests are appropriately modified to use NIDs and Oids.
To prepare for a follow-up change, the announcement data is also
threaded through for the fetch protocol to focus fetching only the
advertised set of Oids.
N.B.: because the refs aren't communicated, the verification happens
when fetching since the data needs to be fetched before
verifying. This is a trade-off with minimising the message payload for
verifying earlier.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
Wire up the `radicle-fetch` crate to the fetch logic in the
`radicle-node` worker.
Any required traits for storage, identity, tracking, and transport
used by `radicle-fetch`.
The worker adds a new module to run the `git upload-pack` process and
pipes the `stdout`/`stdin` to the respective channels.
The `Worker` type required adding a `FetchConfig` to easily allow the
configuration of fetches and passing the signer and tracking store
location through to the fetch `Handle`.
The channels code is adapted to have a new writer that always flushes
when a call to `write_all` is made. This is necessary since the
Gitoxide code never calls `flush`, and so it will hang since no data
is sent until `flush` is called.
Since `UserInfo` is used to set up a newly cloned repository, this
changed some of the SHA1 hashes output in the CLI test examples. It
was also necessary to update the CLI's clone code to be deterministic
in its output for tests.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
When a fetch is user-requested, a timeout can be supplied that is passed
down to the worker.
When a fetch is service-requested, a default timeout is used.
This is a breaking protocol change that prefixes all gossip messages
with their length.
This is to allow message extensions, as we'll be able to read the whole
message inclulding the extensions, without having to decode the
extensions.
It also makes it easier to check the message length before we start
decoding, so that we can reject messages that are too big.
It turns out it's possible to create a deadlock when using
`child.wait()` with `Stdio::piped()`, if we don't read the output while
the child process is running. This is because the pipe has an internal
kernel buffer that will eventually fill up, causing the child process to
block on writing to stdout and never exit.
To remedy this situation, we read the stdout in a thread while the child
process is running.
When attempting to connect to a peer, and when succeeding, write the timestamp
to the database.
This will allow us to better select addresses and not retry peers to
soon.
It's currently possible for a channel to be blocked forever while
waiting for data, if the remote peer keeps the connection alive.
This can occupy a worker indefinitely.
With these timeouts, we ensure that workers exit if there is no
activity after a certain amount of time.
Instead of returning the namespaces that we fetched with,
we return the remotes that were actually fetched from.
This can differ if some trusted peers were not available
on the remote node.
Previously, if that was the case, it would cause errors
since the remotes were looked up and that lookup failed.
Instead of parsing the worker requests and responses, which was causing issues
around packfile transmission, we use threads and pass the bytes
transparently.
To improve the reliability and flexibility of the protocol, we introduce
multiplexing over peer connections. This involves a new `Frame` type
that carries a stream-id and payload.
Three stream types are made available:
1. Control
2. Gossip
3. Git
This change brings the following improvements:
* Removed need to queue fetch requests
* Removed need to queue gossip messages
* Removed need for `Fetch` and `FetchOk` messages
* Service doesn't need to know about inbound fetches
* Removed one round-trip for fetch negotiation
* Removed special `done` git packet
* Removed session logic and state around Git/Fetch protocols
* Removed code around upgrading/downgrading transport
* Worker in responder mode is able to process any number of fetches
* Connections support any number of concurrent fetches
We had to introduce a few extra things however to make it all work:
* A `VarInt` type for variable-length integers, since we want the frames
to be lightweight.
* A custom "tunnel" implementation, since we couldn't use the existing
one anymore.
Overall the change removes more complexity than it adds, while improving
the protocol along the way.
For whatever reason, we weren't using the deserializer for message
decoding. This meant that if a message came in two pieces, it would
cause decoding problems, since the first piece would be discarded.
Due to never removing peers from the peer map in the wire
protocol module, it was possible for there to be multiple
connected peers with the same Node ID. Some might be in a
"connected" state, while others might be "disconnected".
This caused a bug where a peer, although connected would
be considered disconnected by the wire module, because
when searching for a peer with that Node ID, the disconnected
one would return first.
To fix this, we make sure to remove peers from the map once
the connection is dropped.
A fetch is a client-server interaction and so there are two sides to
this interaction. These sides are generally referred to as receive --
the client side -- and upload -- the server side.
The `Fetch` type indicated whether it was doing a receive or upload by
using a field `initiated: bool`. However, this can be confusing when
you pair it with the `namespaces: Namespaces` field, since only the
receive side should be sending what refspecs it wants and the upload
side responds with those matching refspecs.
A better way to represent this is using an enum for which direction
the fetch is being considered, where the receive side contains the
Namespaces. To borrow from Noise, the variant names are 'Initiator'
for the client and 'Responder' for the server.
The resulting code then makes decisions based on what variant of the
enum was passed.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
* Read git request line before starting daemon, so that we don't leave
git data in the socket read buffer in case of an error connecting to
the daemon.
* Send a response on the fetch channel even if we disconnect the remote.
* Don't send `done` packet if there's an error; it won't be expected.
Rust 1.67 was announced[0]. Update the necessary files for running
1.67 and fix the clippy suggestions for string interpolation.
[0]: https://blog.rust-lang.org/2023/01/26/Rust-1.67.0.html
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
This change makes sure that nodes don't gossip announcements back to
the announcers.
Since the tests were relying on this mechanism to populate the routing
table, we had to change a few more things, such as adding the inventory
to the node's routing table.
Signed-off-by: Alexis Sellier <self@cloudhead.io>
There are few unrelated changes in this commit that I wasn't able to
split out, mostly due to laziness, so I'll list them here:
* The node startup logic is moved out of the main function, and into a
dedicated functionin `client.rs`.
* There is a tentative worker implementation that is untested, and
handles git replication both ways. It's likely lacking in various ways
but I wanted to get it in there anyway.
* `Address` was moved from `radicle-node` to `radicle`, because it's
needed to send `Connect` commands from the node handle.
* The `netservices` dependency was updated.
* Finally, there is a first end-to-end node test, which currently
doesn't use NoiseXK, but uses the new reactor in `netservices`.
Signed-off-by: Alexis Sellier <self@cloudhead.io>
This is a large change that introduces transport encryption via
a Noise_XK handshake. The inner `Service` now works directly
with `NodeId` and doesn't need to interact with socket addresses.
The transport logic takes place in `wire::transport`, and the
established sessions are passed down to `Service`.
Note that the i/o implementation in the `poll-reactor` crate still
needs testing, and there is still work to be done to integrate the
Git transport as part of this architecture.
Signed-off-by: Alexis Sellier <alexis@radicle.xyz>
Use the BoundedVec to limit the supported number of refs in
RefAnnouncement messages. The size (235) is the maximum supported by
message wire size.
To reduce the complexity of the changes, BoundedVec is used again
instead of using a new bounded Refs or BTreeMap. This requires
defining new Decode/Encode logic for Ref's element types, and
workarounds for arbitrary instances.
Signed-off-by: Slack Coder <slackcoder@server.ky>
The latter is a fork of the former, which adds const generics
and a bunch of other quality of life improvements that aren't
in the original crate due to it being unmaintained.
Signed-off-by: Alexis Sellier <alexis@radicle.xyz>
Restrict message size decoding by limiting Message's vector size.
Introduce the BoundedVec type to do this on the type level for
convenience and clarity.
Signed-off-by: Slack Coder <slackcoder@server.ky>
The changes in this commit are all intertwined and it was hard to
split them into smaller commits:
* Use custom transports everywhere
Instead of `file://` and instead of using a child process for fetching
and pushing; we go through our custom transports, which use the `rad://`
and `heartwood://` schemes.
* Introduce a 'mock' remote transport
To test the remote transport without needing to spawn TCP streams,
we add a mock transport that works between storages on the file
system.
* Get rid of node's `git-url`
Instead of git urls, nodes simply use their node-ids to replicate.
This type that was wrapping `Message` isn't very useful here.
If we decide that we do need something like that, it's best handled one
layer below.
Signed-off-by: Alexis Sellier <alexis@radicle.xyz>
To build an address book, peers communicate via gossip, sending
`NodeAnnouncement` messages.
These need to be handled by storing the addresses in an address book
and the addresses need to be persisted to disk.
Node announcement messages should also be relayed to peers like other
gossip messages.
cloudhead
Fintan Halpenny
Alexis Sellier
xphoniex
Alexis Sellier
Dr Maxim Orlovsky
Dr Maxim Orlovsky
Slack Coder