quic: ability to listen on ipv4 and v6

[mxinden]

Summary

With libp2p-quic v0.8.0-alpha one can not listen on both a v4 and v6 address.

Unconfirmed whether #3454 has the same issue.

Expected behaviour

Calling listen_on with a v4 and v6 address to work.

Actual behaviour

Apply the following diff:

modified   examples/dcutr/src/main.rs
@@ -172,7 +172,10 @@ fn main() -> Result<(), Box<dyn Error>> {
     .build();
 
     swarm
-        .listen_on("/ip4/0.0.0.0/udp/0/quic-v1".parse().unwrap())
+        .listen_on("/ip4/0.0.0.0/udp/4001/quic-v1".parse().unwrap())
+        .unwrap();
+    swarm
+        .listen_on("/ip6/::/udp/4001/quic-v1".parse().unwrap())
         .unwrap();
     swarm
         .listen_on("/ip4/0.0.0.0/tcp/0".parse().unwrap())

and then run with:

cargo run -- --mode listen --secret-key-seed 0 --relay-address /ip4/127.0.0.1

resulting in:

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Other(Custom { kind: Other, error: Transport(Right(Io(Os { code: 98, kind: AddrInUse, message: "Address already in use" }))) })', examples/dcutr/src/main.rs:179:10

Debug Output

<output>

Possible Solution

My suspicion is that the ipv6 UdpSocket binds to both v4 and v6 in order to accept IPv4-mapped IPv6 traffic e.g. ::ffff:192.0.2.128.

In such case we would need to call set_only_v6 on the socket.

See same code on the tcp side:

rust-libp2p/transports/tcp/src/lib.rs

Lines 356 to 358 in 2671375

	if socket_addr.is_ipv6() {
	socket.set_only_v6(true)?;
	}

Version

libp2p-quic v0.8.0-alpha

Would you like to work on fixing this bug?

No

[eserilev]

hi, i'd like to work on this if its still relevant

[eserilev]

so far im having a bit of trouble recreating this bug

applying the diff you mentioned i'm getting a:
thread 'main' panicked at 'OutgoingConnectionError { connection_id: ConnectionId(1), peer_id: None, error: Transport([("/ip4/127.0.0.1", Other(Custom { kind: Other, error: MultiaddrNotSupported("/ip4/127.0.0.1") }))]) }'

[mxinden]

Help is much appreciated. Also since #3454 is merged now, I don't suspect any larger merge conflicts to come up.

so far im having a bit of trouble recreating this bug

applying the diff you mentioned i'm getting a: thread 'main' panicked at 'OutgoingConnectionError { connection_id: ConnectionId(1), peer_id: None, error: Transport([("/ip4/127.0.0.1", Other(Custom { kind: Other, error: MultiaddrNotSupported("/ip4/127.0.0.1") }))]) }'

I suspect this to be due to an error I made above. --relay-address /ip4/127.0.0.1 should instead be a full QUIC address, e.g. --relay-address /ip4/127.0.0.1/udp/<listen-port/quic-v1. Does that fix the issue you are seeing?

[eserilev]

Thanks for the additional info

I took a step back and just tried to run this example as is, with the example command found in the readme, but I'm getting a Connection Refused error message

applying the diff you mentioned and running with the following command:

cargo run -- --mode listen --secret-key-seed 0 --relay-address /ip4/127.0.0.1/udp/4005/quic-v1

I'm getting a HandshakeTimedOut error. I'm probably missing something here, so if you have any additional tips/hints please let me know. In the meantime I'll keep playing around with this and see how far I can get

[jxs]

Have you tried this after #3454 @mxinden?

UPD:

was able to reproduce your error with the ping example.