findnode(self) should return multiple peers

[guillaumemichel]

When receiving a kademlia FIND_NODE request for its own peer id, a go-libp2p-kad-dht node will respond with its own peer record only. According to libp2p kademlia spec, it should reply with the k closest nodes.

The libp2p Kademlia DHT offers the following types of operations:

• Peer routing
  • Finding the closest nodes to a given key via FIND_NODE.

3. Upon a response:
   1. If successful the response will contain the k closest nodes the peer knows to the key Key. Add them to the candidate list Pn, except for those that have already been queried.

Depending on libp2p/specs#609, nodes shouldn't even include their own peer records in the response (already known to the requester). This will be implemented in a future PR.

[sukunrt]

Should we add a test for this?

[sukunrt]

dht.betterPeersToQuery errors if the list contains self.
https://github.com/libp2p/go-libp2p-kad-dht/blob/master/dht.go#L765-L769

		if clp == dht.self {
			logger.Error("BUG betterPeersToQuery: attempted to return self! this shouldn't happen...")
			return nil
		}
		// Dont send a peer back themselves
		if clp == from {
			continue
		}

It also handles the case that we don't inform the peer about itself.

[guillaumemichel]

This snippet handles the output of routingTable.NearestPeers, which never contains itself.

The peer adds it own peer id, only if its own peer id is the target of the FIND_NODE request at

go-libp2p-kad-dht/handlers.go

Lines 263 to 285 in 7d6120f

	// if looking for self... special case where we send it on CloserPeers.
	targetPid := peer.ID(pmes.GetKey())
	closest = dht.betterPeersToQuery(pmes, from, dht.bucketSize)
	// Never tell a peer about itself.
	if targetPid != from {
	// Add the target peer to the set of closest peers if
	// not already present in our routing table.
	//
	// Later, when we lookup known addresses for all peers
	// in this set, we'll prune this peer if we don't
	// _actually_ know where it is.
	found := false
	for _, p := range closest {
	if targetPid == p {
	found = true
	break
	}
	}
	if !found {
	closest = append(closest, targetPid)
	}
	}

[sukunrt]

Why do we do this? If the peerID isn't found, should we return it in findPeer?

[guillaumemichel]

dht.betterPeersToQuery never includes itself (nor from) in the results.

Currently if A sends FIND_NODE(B) to B, B will reply [B], and we now want it to return [B, C, D, ...]. In this case targetPid is our own peer id, it wasn't returned by dht.betterPeersToQuery, so we need to add it at this step.

We will be able to get rid of this when the response to FIND_NODE(self) will be not include self.

[sukunrt]

But if the target node doesnt exist we will include it in the output, right?

[guillaumemichel]

Peers without addresses are excluded from the response

go-libp2p-kad-dht/handlers.go

Lines 292 to 301 in 7d6120f

	closestinfos := pstore.PeerInfos(dht.peerstore, closest)
	// possibly an over-allocation but this array is temporary anyways.
	withAddresses := make([]peer.AddrInfo, 0, len(closestinfos))
	for _, pi := range closestinfos {
	if len(pi.Addrs) > 0 {
	withAddresses = append(withAddresses, pi)
	}
	}
	resp.CloserPeers = pb.PeerInfosToPBPeers(dht.host.Network(), withAddresses)