No statistics and timeout when using dual-stack wireguard tunnel

[dupondje]

Issue

No network statistics are shown in the UI. And the tunnel times out because there is no activity.
This while the logs give the following:

{ message: "Network activity statistics for interface wg0 sent to the client" }
{ message: "Found 1 peers configured on WireGuard interface" }
{ message: "Peer xxxxxxxx= statistics didn't change; ignoring it." }

The cause

To get the data, the client calls read_interface_data.
Now on dual-stack tunnels, netlink seems to return 2 peers. One for IPv4, one for IPv6.
But only the IPv4 one returns the TxBytes etc:

                                        LastHandshake(
                                            SystemTime {
                                                tv_sec: 1760018513,
                                                tv_nsec: 47243476,
                                            },
                                        ),
                                        PersistentKeepalive(
                                            25,
                                        ),
                                        TxBytes(
                                            245028,
                                        ),
                                        RxBytes(
                                            905264,
                                        ),

The IPv6 one does NOT.
But only the IPv6 one is returned by read_interface_data, causing incomplete data, causing this to fail :)

[dupondje]

The code in https://github.com/DefGuard/wireguard-rs/blob/42d478b980cb3f7549f20f2f492c896a7389a79d/src/host.rs#L349 will need to be extended.
This as both IPv4 and IPv6 public_key's are the same, the IPv4 peer values are now overwritten by the IPv6 values.

Some (incomplete) code:

diff --git a/src/host.rs b/src/host.rs
index 6dcdcd1..fb4080f 100644
--- a/src/host.rs
+++ b/src/host.rs
@@ -355,7 +355,35 @@ impl Host {
                 WgDeviceAttrs::Peers(nlas) => {
                     for nla in nlas {
                         let peer = Peer::from_nlas(nla);
-                        self.peers.insert(peer.public_key.clone(), peer);
+                        if self.peers.contains_key(&peer.public_key) {
+                            if let Some(existing_peer) = self.peers.get_mut(&peer.public_key) {
+                                if peer.tx_bytes != 0 {
+                                    existing_peer.tx_bytes = peer.tx_bytes;
+                                }
+                                if peer.rx_bytes != 0 {
+                                    existing_peer.rx_bytes = peer.rx_bytes;
+                                }
+                                if peer.last_handshake.is_some() {
+                                    existing_peer.last_handshake = peer.last_handshake;
+                                }
+                                if peer.endpoint.is_some() {
+                                    existing_peer.endpoint = peer.endpoint;
+                                }
+                                if peer.persistent_keepalive_interval.is_some() {
+                                    existing_peer.persistent_keepalive_interval =
+                                        peer.persistent_keepalive_interval;
+                                }
+                                let updated_peer = existing_peer.clone();
+                                // End mutable borrow before inserting
+                                println!(
+                                    "Updated existing peer: {:#?} with {:#?}",
+                                    existing_peer, peer
+                                );
+                                self.peers.insert(peer.public_key.clone(), updated_peer);
+                            }
+                        } else {
+                            self.peers.insert(peer.public_key.clone(), peer);
+                        }
                     }
                 }
                 _ => (),

That seems to fix the output of read_interface_data.

[wojcik91]

@dupondje Could you please provide some more context about the OS and kernel version you are running?

I haven't been able to reproduce it on 6.17.1, so I'd like to do some more digging.

[twouters]

Not the original reporter, but I experience this issue on 6.17.1 (Arch Linux)

$ uname -a                                  
Linux britannic 6.17.1-arch1-1 #1 SMP PREEMPT_DYNAMIC Mon, 06 Oct 2025 18:48:29 +0000 x86_64 GNU/Linux

[10/13/2025, 10:37:24 AM][DEBUG][Service] Gathering network usage statistics for client's network activity on wg0
[10/13/2025, 10:37:24 AM][DEBUG][Service] Reading host info for interface wg0
[10/13/2025, 10:37:24 AM][DEBUG][Service] Reading Netlink data for interface wg0
[10/13/2025, 10:37:24 AM][DEBUG][Service] Host info read for interface wg0
[10/13/2025, 10:37:24 AM][DEBUG][Service] Found 1 peers configured on WireGuard interface
[10/13/2025, 10:37:24 AM][DEBUG][Service] Peer XXXX statistics didn't change; ignoring it.
[10/13/2025, 10:37:24 AM][DEBUG][Service] Network activity statistics for interface wg0 sent to the client
[10/13/2025, 10:37:24 AM][INFO][Client] Emitted config-changed event for instance MyInstance(1)
[10/13/2025, 10:37:24 AM][INFO][Client] Automatically retrieved the newest instance configuration from core for 1 instances, sleeping for 30s

I've set the session timeout to one year to prevent disconnects due to timeouts.

[j-chmielewski]

Can you show me the output of these commands when the interface is up?

sudo wg show

sudo ip a show dev wg0