diff --git a/README.md b/README.md
index 5c20227..6aa34ca 100644
--- a/README.md
+++ b/README.md
@@ -31,10 +31,10 @@ A working library with **real iroh connections** and integration tests:
 - status snapshots, an event stream and honest diagnostics;
 - configuration restored after a restart;
 - correct behaviour when the disposable cache is missing or corrupt;
-- a **WireGuard data plane**, in userspace: its own key per network,
-  deterministic IPv6 overlay addressing, real tunnels carried over iroh, and
-  address ownership enforced rather than believed;
-- a **command line agent**, `tsunagi`.
+- a **WireGuard data plane**, in userspace: its own key per network, a dual
+  stack overlay with deterministically derived addresses, real tunnels carried
+  over iroh, and address ownership enforced rather than believed;
+- a **command line agent**, `tsunagi`, with a local control socket.
 
 ### What it deliberately does **not** do
 
@@ -94,8 +94,41 @@ wireguard: tsunkkcp43lmdje on fd15:1d9e:fa21:f201:…/64 mtu 1280, 1/1 tunnel(s)
   4jO4kx9Z fd15:1d9e:fa21:f201:… handshake 3s ago tx=0 rx=0 dropped=0 path=Direct via Ip(…)
 ```
 
-`1/1 tunnel(s) established` means a real WireGuard handshake completed. Then
-`ping6` the peer's overlay address.
+`1/1 tunnel(s) established` means a real WireGuard handshake completed.
+
+## Checking that it works
+
+From another shell on either machine:
+
+```bash
+tsunagi status
+```
+
+```text
+endpoint  7d76ccbbc21bf30767e14422c0494740a2cecc02aa9f82d5b8d57bdae350e7fc
+hostname  tsunagi-7d76ccbbc2
+bound     0.0.0.0:41641
+
+network lab (z2o4qwrvnj3zb6st2aoqg4abf342j662q2ujttqrsmz22argk2ba)  active
+  peer b345d5271b  tsunagi-b345d5271b  Direct  rtt 24ms
+  overlay tsunz2o4qwrvnj3 fd09:…:c1c6/64 and 100.110.49.177 mtu 1280  1/1 tunnel(s) up
+    SDsEb/WF  fd09:…:c4b / 100.65.243.53  handshake 4s ago  tx 0 rx 0  Direct via Ip(…)
+```
+
+`1/1 tunnel(s) up` and a recent handshake mean the tunnel is live. Then send
+real traffic to the peer's overlay address:
+
+```bash
+ping6 fd09:…:c4b        # or
+ping  100.65.243.53
+```
+
+`tx` and `rx` in the status should start moving.
+
+The overlay is dual stack: every member derives both an IPv6 address, which
+can never collide, and an IPv4 one in `100.64.0.0/10`, which very rarely can —
+see [docs/wireguard.md](docs/wireguard.md#ipv4-alongside-ipv6). `--no-ipv4`
+runs IPv6 only, `--ipv4-range` moves the range.
 
 Notes:
 
@@ -130,6 +163,7 @@ sudo ip tuntap add dev tsunjwc6dcrtmo5 mode tun user ab
 sudo ip link set dev tsunjwc6dcrtmo5 mtu 1280 up
 sudo sysctl -qw net.ipv6.conf.tsunjwc6dcrtmo5.keep_addr_on_down=1
 sudo ip -6 address add fd80:1210:f724:f620:d1bb:f982:3b6e:19bd/64 dev tsunjwc6dcrtmo5 nodad
+sudo ip address add 100.110.49.177/10 dev tsunjwc6dcrtmo5
 ```
 
 The MTU is 1280 because that is the minimum IPv6 requires (RFC 8200). Linux
diff --git a/docs/testing.md b/docs/testing.md
index 20de3da..dd2646c 100644
--- a/docs/testing.md
+++ b/docs/testing.md
@@ -50,6 +50,11 @@ keeping the WireGuard identity, shutdown removing every interface, a forged
 overlay claim being rejected, and the core carrying the payload without
 interpreting it.
 
+`tests/local_control.rs` covers the local control socket end to end: a client
+asking a running agent for status over a real Unix socket, a leftover socket
+file being replaced while a live one is not, and the derived socket path
+staying short enough to bind.
+
 `tests/discovery.rs` covers the discovery contract itself: a static bootstrap
 candidate is enough to join, several backends compose, entries are withdrawn
 when a network stops, and a forgotten network stays forgotten across a restart.
diff --git a/docs/wireguard.md b/docs/wireguard.md
index 04a4cee..999e675 100644
--- a/docs/wireguard.md
+++ b/docs/wireguard.md
@@ -56,6 +56,12 @@ direct path when it can and falls back to a relay when it cannot; the tunnel
 rides on whichever it got. There is no separate STUN, no separate hole punching
 and no second set of NAT problems to solve for WireGuard.
 
+## Checking it from outside
+
+`tsunagi status` asks a running agent over its local control socket and prints
+what it sees, including whether each tunnel has actually handshaken. See
+[../README.md](../README.md#checking-that-it-works).
+
 ## Deterministic overlay addressing
 
 A mesh with no coordinator cannot hand out addresses, so everyone derives their
@@ -77,6 +83,29 @@ Two consequences matter:
 * a member's address is bound to its WireGuard public key, so address
   ownership can be checked locally rather than believed.
 
+## IPv4 alongside IPv6
+
+The overlay is dual stack by default: every member also derives an IPv4
+address, from the same inputs, into `100.64.0.0/10` (RFC 6598 shared address
+space — deliberately not RFC 1918, so it rarely clashes with the network the
+machine is already on). The range is configurable, and IPv4 can be turned off.
+
+**IPv4 is weaker than IPv6 here, and the difference is not cosmetic.** A 64 bit
+interface identifier makes an IPv6 collision impossible in practice. IPv4 has
+nothing like that much room: in a `/10` with 50 members the chance that two
+derive the same address is roughly 0.03%. Small, but not zero, and a mesh with
+no coordinator cannot simply allocate around it.
+
+So a collision is detected and resolved rather than assumed away: the member
+whose WireGuard public key sorts lower keeps the address, a rule every member
+computes identically and therefore agrees on without exchanging anything. The
+other member ends up with **no IPv4 address** and is still fully reachable over
+IPv6. The status output flags it.
+
+That is the honest summary: **IPv6 always works; IPv4 almost always works and
+degrades predictably when it does not.** Allocating IPv4 properly needs the
+agreed state described in [sync-model.md](sync-model.md).
+
 ## Address ownership is enforced, not announced
 
 Kernel WireGuard enforces `AllowedIPs`. In userspace that is our job, and
@@ -87,6 +116,8 @@ Kernel WireGuard enforces `AllowedIPs`. In userspace that is our job, and
 * **inbound**, a decrypted packet is dropped unless its *source* is exactly the
   address derived for the peer whose tunnel decrypted it.
 
+Both apply to IPv4 and IPv6 alike.
+
 So a participant cannot receive traffic addressed to somebody else and cannot
 forge traffic that appears to come from somebody else. A participant who knows
 the network secret can mint many keys and therefore occupy many addresses, but
@@ -192,9 +223,8 @@ async fn main() -> Result<()> {
 
 * **Full mesh only.** Every member runs a tunnel to every other member.
   Routing through an intermediate participant is not implemented.
-* **IPv6 overlay only.** Addressing is IPv6 ULA because it can be derived
-  collision-free. An IPv4 overlay would need an allocator, which needs the
-  agreed state described in [sync-model.md](sync-model.md).
+* **IPv4 addressing can collide.** See above: it is resolved deterministically
+  and the loser keeps IPv6, but a proper allocator needs agreed state.
 * **No routes, DNS or firewall rules.** The plugin creates its interface and
   nothing else. Anything beyond the overlay `/64` is the operator's business.
 * **Membership is session-scoped.** A peer leaves the overlay when its control
diff --git a/src/bin/tsunagi.rs b/src/bin/tsunagi.rs
index f3993b4..2942034 100644
--- a/src/bin/tsunagi.rs
+++ b/src/bin/tsunagi.rs
@@ -98,6 +98,45 @@ struct TunSetupArgs {
     /// Interface MTU, matching `tsunagi up --wg-mtu`. At least 1280.
     #[arg(long)]
     wg_mtu: Option<u32>,
+
+    /// Match `tsunagi up --no-ipv4`.
+    #[arg(long)]
+    no_ipv4: bool,
+
+    /// Match `tsunagi up --ipv4-range`.
+    #[arg(long, value_name = "CIDR", conflicts_with = "no_ipv4")]
+    ipv4_range: Option<String>,
+}
+
+/// Parses `address/prefix` into an IPv4 range.
+fn parse_ipv4_range(text: &str) -> Result<(std::net::Ipv4Addr, u8), String> {
+    let (address, prefix) = text
+        .split_once('/')
+        .ok_or_else(|| format!("`{text}` is not an address with a prefix, e.g. 100.64.0.0/10"))?;
+    let address = address
+        .parse()
+        .map_err(|err| format!("`{address}` is not an IPv4 address: {err}"))?;
+    let prefix: u8 = prefix
+        .parse()
+        .map_err(|err| format!("`{prefix}` is not a prefix length: {err}"))?;
+    if prefix > 30 {
+        return Err(format!("a /{prefix} has no room for hosts"));
+    }
+    Ok((address, prefix))
+}
+
+/// Resolves the IPv4 overlay range from the flags.
+fn resolve_ipv4_range(
+    no_ipv4: bool,
+    range: Option<&String>,
+) -> Result<Option<(std::net::Ipv4Addr, u8)>, Box<dyn std::error::Error>> {
+    if no_ipv4 {
+        return Ok(None);
+    }
+    match range {
+        Some(text) => Ok(Some(parse_ipv4_range(text)?)),
+        None => Ok(Some(tsunagi::dataplane::wireguard::DEFAULT_IPV4_RANGE)),
+    }
 }
 
 #[derive(Debug, Args, Clone)]
@@ -212,6 +251,14 @@ struct UpArgs {
     #[arg(long)]
     wg_mtu: Option<u32>,
 
+    /// Run an IPv6-only overlay instead of dual stack.
+    #[arg(long)]
+    no_ipv4: bool,
+
+    /// IPv4 overlay range, as `address/prefix`. Defaults to 100.64.0.0/10.
+    #[arg(long, value_name = "CIDR", conflicts_with = "no_ipv4")]
+    ipv4_range: Option<String>,
+
     /// How often to print a status summary, in seconds. Zero disables it.
     #[arg(long, default_value_t = 15)]
     status_interval: u64,
@@ -346,6 +393,7 @@ async fn status(args: StatusArgs) -> Result<(), Box<dyn std::error::Error>> {
 async fn tun_setup(args: TunSetupArgs) -> Result<(), Box<dyn std::error::Error>> {
     use tsunagi::dataplane::wireguard::{
         DEFAULT_MTU, OVERLAY_PREFIX_LEN, WgKeyStore, interface_name, overlay_address,
+        overlay_address_v4,
     };
     use tsunagi::identity::NetworkKeys;
 
@@ -364,6 +412,7 @@ async fn tun_setup(args: TunSetupArgs) -> Result<(), Box<dyn std::error::Error>>
 
     let interface = interface_name(&args.wg_prefix, network)?;
     let address = overlay_address(network, &key.public());
+    let ipv4_range = resolve_ipv4_range(args.no_ipv4, args.ipv4_range.as_ref())?;
     let mtu = args.wg_mtu.unwrap_or(DEFAULT_MTU);
     let user = args.user.unwrap_or_else(|| {
         std::env::var("SUDO_USER")
@@ -373,6 +422,11 @@ async fn tun_setup(args: TunSetupArgs) -> Result<(), Box<dyn std::error::Error>>
 
     println!("# Network  {name} ({network})");
     println!("# Interface {interface}, address {address}/{OVERLAY_PREFIX_LEN}, mtu {mtu}");
+    if let Some((base, prefix)) = ipv4_range
+        && let Some(v4) = overlay_address_v4(network, &key.public(), (base, prefix))
+    {
+        println!("# IPv4 overlay address {v4}/{prefix}");
+    }
     println!("# Run once as root; then run `tsunagi up` as {user}.");
     println!(
         "#\n\
@@ -386,6 +440,11 @@ async fn tun_setup(args: TunSetupArgs) -> Result<(), Box<dyn std::error::Error>>
     println!("sudo ip link set dev {interface} mtu {mtu} up");
     println!("sudo sysctl -qw net.ipv6.conf.{interface}.keep_addr_on_down=1");
     println!("sudo ip -6 address add {address}/{OVERLAY_PREFIX_LEN} dev {interface} nodad");
+    if let Some((base, prefix)) = ipv4_range
+        && let Some(v4) = overlay_address_v4(network, &key.public(), (base, prefix))
+    {
+        println!("sudo ip address add {v4}/{prefix} dev {interface}");
+    }
     println!("\n# To check it afterwards:");
     println!("ip -6 addr show dev {interface}");
     println!("\n# To remove it again:");
@@ -516,7 +575,8 @@ async fn up(args: UpArgs) -> Result<(), Box<dyn std::error::Error>> {
             system_tun_factory()?
         };
         let mut wg = WireguardConfig::new(paths.state_dir.join("wireguard"))
-            .with_interface_prefix(args.wg_prefix.clone());
+            .with_interface_prefix(args.wg_prefix.clone())
+            .with_ipv4_range(resolve_ipv4_range(args.no_ipv4, args.ipv4_range.as_ref())?);
         if let Some(mtu) = args.wg_mtu {
             wg = wg.with_mtu(mtu);
         }
@@ -637,6 +697,7 @@ async fn build_report(
                     interface: view.interface.clone(),
                     mtu: view.mtu,
                     address: view.overlay_address.to_string(),
+                    address_v4: view.overlay_address_v4.map(|addr| addr.to_string()),
                     prefix: view.overlay_prefix.to_string(),
                     prefix_len: view.overlay_prefix_len,
                     peers: view
@@ -645,6 +706,7 @@ async fn build_report(
                         .map(|peer| OverlayPeerReport {
                             public_key: peer.public_key.to_string(),
                             address: peer.overlay_address.to_string(),
+                            address_v4: peer.overlay_address_v4.map(|addr| addr.to_string()),
                             handshake_secs_ago: peer
                                 .tunnel
                                 .as_ref()
diff --git a/src/dataplane/wireguard/device.rs b/src/dataplane/wireguard/device.rs
index a51e6fd..e0d9994 100644
--- a/src/dataplane/wireguard/device.rs
+++ b/src/dataplane/wireguard/device.rs
@@ -27,7 +27,7 @@
 //! what it announced.
 
 use std::collections::HashMap;
-use std::net::Ipv6Addr;
+use std::net::{IpAddr, Ipv4Addr, Ipv6Addr};
 use std::sync::atomic::{AtomicU32, AtomicU64, Ordering};
 use std::sync::{Arc, Mutex, RwLock};
 use std::time::Duration;
@@ -42,7 +42,7 @@ use crate::dataplane::transport::{SharedLink, TransportError};
 use crate::identity::NetworkId;
 
 use super::keys::{WgPublicKey, WgSecretKey};
-use super::overlay::overlay_address;
+use super::overlay::{overlay_address, overlay_address_v4};
 use super::packet::IpHeader;
 use super::tun::TunDevice;
 
@@ -109,6 +109,9 @@ struct Peer {
     endpoint_id: EndpointId,
     public_key: WgPublicKey,
     overlay: Ipv6Addr,
+    /// The IPv4 address this peer owns, when the overlay is dual stack and
+    /// nobody else derived the same one.
+    overlay_v4: Mutex<Option<Ipv4Addr>>,
     tunn: Mutex<Tunn>,
     link: SharedLink,
     counters: Arc<PeerCounters>,
@@ -168,6 +171,13 @@ pub struct PeerSummary {
     pub public_key: WgPublicKey,
     /// The overlay address this agent derived for it.
     pub overlay_address: Ipv6Addr,
+    /// Its IPv4 overlay address, when the overlay is dual stack.
+    ///
+    /// `None` with `ipv4_conflict` set means another member derived the same
+    /// address and won it; that peer is still fully reachable over IPv6.
+    pub overlay_address_v4: Option<Ipv4Addr>,
+    /// Whether this peer lost an IPv4 address to a derivation collision.
+    pub ipv4_conflict: bool,
     /// Whether the tunnel has handshaken.
     pub health: PeerHealth,
     /// Traffic counters.
@@ -182,11 +192,15 @@ struct Inner {
     network: NetworkId,
     private_key: WgSecretKey,
     tun: Arc<dyn TunDevice>,
+    /// The IPv4 overlay range, when the overlay is dual stack.
+    ipv4_range: Option<(Ipv4Addr, u8)>,
     peers: RwLock<HashMap<WgPublicKey, Arc<Peer>>>,
-    routes: RwLock<HashMap<Ipv6Addr, WgPublicKey>>,
+    /// Both families, so one lookup routes any packet.
+    routes: RwLock<HashMap<IpAddr, WgPublicKey>>,
     next_index: AtomicU32,
     unroutable: AtomicU64,
     multicast: AtomicU64,
+    ipv4_conflicts: AtomicU64,
 }
 
 impl std::fmt::Debug for Inner {
@@ -207,16 +221,23 @@ pub struct WireguardDevice {
 
 impl WireguardDevice {
     /// Starts a device on top of `tun`.
-    pub fn start(network: NetworkId, private_key: WgSecretKey, tun: Arc<dyn TunDevice>) -> Self {
+    pub fn start(
+        network: NetworkId,
+        private_key: WgSecretKey,
+        tun: Arc<dyn TunDevice>,
+        ipv4_range: Option<(Ipv4Addr, u8)>,
+    ) -> Self {
         let inner = Arc::new(Inner {
             network,
             private_key,
             tun,
+            ipv4_range,
             peers: RwLock::new(HashMap::new()),
             routes: RwLock::new(HashMap::new()),
             next_index: AtomicU32::new(1),
             unroutable: AtomicU64::new(0),
             multicast: AtomicU64::new(0),
+            ipv4_conflicts: AtomicU64::new(0),
         });
 
         let reader = tokio::spawn(read_from_os(Arc::clone(&inner)));
@@ -263,10 +284,12 @@ impl WireguardDevice {
         );
 
         let overlay = overlay_address(self.inner.network, &public_key);
+        let overlay_v4 = self.claim_ipv4(&public_key);
         let peer = Arc::new(Peer {
             endpoint_id,
             public_key,
             overlay,
+            overlay_v4: Mutex::new(overlay_v4),
             tunn: Mutex::new(tunn),
             link,
             counters: Arc::new(PeerCounters::default()),
@@ -279,7 +302,10 @@ impl WireguardDevice {
         }
 
         write_lock(&self.inner.peers).insert(public_key, Arc::clone(&peer));
-        write_lock(&self.inner.routes).insert(overlay, public_key);
+        write_lock(&self.inner.routes).insert(IpAddr::V6(overlay), public_key);
+        if let Some(v4) = overlay_v4 {
+            write_lock(&self.inner.routes).insert(IpAddr::V4(v4), public_key);
+        }
 
         // Start the handshake now instead of waiting for the next timer tick,
         // so the tunnel is usable as soon as the link exists.
@@ -287,13 +313,64 @@ impl WireguardDevice {
         Ok(())
     }
 
+    /// Decides which IPv4 address a new peer gets, if any.
+    ///
+    /// IPv4 has far too little room for a derived address to be collision
+    /// free. When two members derive the same one, the member whose public
+    /// key sorts lower keeps it — a rule every member computes identically,
+    /// so they all agree on the outcome without talking about it. The other
+    /// member simply has no IPv4 address; it is still fully reachable over
+    /// IPv6, which never collides.
+    fn claim_ipv4(&self, public_key: &WgPublicKey) -> Option<Ipv4Addr> {
+        let range = self.inner.ipv4_range?;
+        let wanted = overlay_address_v4(self.inner.network, public_key, range)?;
+
+        let holder = read_lock(&self.inner.routes)
+            .get(&IpAddr::V4(wanted))
+            .copied();
+        let Some(holder) = holder else {
+            return Some(wanted);
+        };
+        if holder == *public_key {
+            return Some(wanted);
+        }
+
+        self.inner.ipv4_conflicts.fetch_add(1, Ordering::Relaxed);
+        if holder.as_bytes() <= public_key.as_bytes() {
+            // The peer already holding it wins.
+            return None;
+        }
+        // The newcomer wins; take the address away from the other peer.
+        if let Some(loser) = read_lock(&self.inner.peers).get(&holder).cloned() {
+            let mut slot = match loser.overlay_v4.lock() {
+                Ok(guard) => guard,
+                Err(poisoned) => poisoned.into_inner(),
+            };
+            *slot = None;
+        }
+        Some(wanted)
+    }
+
     /// Removes a peer and stops its tunnel.
     pub fn remove_peer(&self, public_key: &WgPublicKey) {
         if let Some(peer) = write_lock(&self.inner.peers).remove(public_key) {
-            write_lock(&self.inner.routes).remove(&peer.overlay);
+            let mut routes = write_lock(&self.inner.routes);
+            routes.remove(&IpAddr::V6(peer.overlay));
+            let v4 = match peer.overlay_v4.lock() {
+                Ok(guard) => *guard,
+                Err(poisoned) => *poisoned.into_inner(),
+            };
+            if let Some(v4) = v4 {
+                routes.remove(&IpAddr::V4(v4));
+            }
         }
     }
 
+    /// How many IPv4 derivation collisions have been resolved.
+    pub fn ipv4_conflicts(&self) -> u64 {
+        self.inner.ipv4_conflicts.load(Ordering::Relaxed)
+    }
+
     /// Removes every peer whose key is not in `keep`.
     pub fn retain_peers(&self, keep: &[WgPublicKey]) {
         let stale: Vec<WgPublicKey> = read_lock(&self.inner.peers)
@@ -315,14 +392,22 @@ impl WireguardDevice {
     pub fn peers(&self) -> Vec<PeerSummary> {
         let mut peers: Vec<PeerSummary> = read_lock(&self.inner.peers)
             .values()
-            .map(|peer| PeerSummary {
-                endpoint_id: peer.endpoint_id,
-                public_key: peer.public_key,
-                overlay_address: peer.overlay,
-                health: peer.health(),
-                stats: peer.stats(),
-                path: peer.link.path_description(),
-                max_datagram: peer.link.max_datagram_size(),
+            .map(|peer| {
+                let overlay_v4 = match peer.overlay_v4.lock() {
+                    Ok(guard) => *guard,
+                    Err(poisoned) => *poisoned.into_inner(),
+                };
+                PeerSummary {
+                    endpoint_id: peer.endpoint_id,
+                    public_key: peer.public_key,
+                    overlay_address: peer.overlay,
+                    overlay_address_v4: overlay_v4,
+                    ipv4_conflict: overlay_v4.is_none() && self.inner.ipv4_range.is_some(),
+                    health: peer.health(),
+                    stats: peer.stats(),
+                    path: peer.link.path_description(),
+                    max_datagram: peer.link.max_datagram_size(),
+                }
             })
             .collect();
         peers.sort_by_key(|peer| peer.public_key);
@@ -415,9 +500,8 @@ async fn read_from_os(inner: Arc<Inner>) {
         };
 
         // Route by destination: only the peer that owns that overlay address
-        // may receive it.
-        let Some(destination) = IpHeader::parse(&packet).and_then(|header| header.v6_destination())
-        else {
+        // may receive it. Both families go through the same table.
+        let Some(destination) = IpHeader::parse(&packet).map(|header| header.destination()) else {
             inner.unroutable.fetch_add(1, Ordering::Relaxed);
             continue;
         };
@@ -488,9 +572,11 @@ async fn read_from_link(inner: Arc<Inner>, peer: Arc<Peer>) {
                 match tunn.decapsulate(None, input.unwrap_or(&[]), &mut scratch) {
                     TunnResult::WriteToNetwork(out) => Outcome::ToNetwork(out.len()),
                     TunnResult::WriteToTunnelV6(out, source) => {
-                        Outcome::ToTunnel(out.len(), Some(source))
+                        Outcome::ToTunnel(out.len(), IpAddr::V6(source))
+                    }
+                    TunnResult::WriteToTunnelV4(out, source) => {
+                        Outcome::ToTunnel(out.len(), IpAddr::V4(source))
                     }
-                    TunnResult::WriteToTunnelV4(out, _) => Outcome::ToTunnel(out.len(), None),
                     TunnResult::Done => Outcome::Done,
                     TunnResult::Err(err) => {
                         tracing::trace!(?err, "wireguard decapsulation failed");
@@ -508,9 +594,19 @@ async fn read_from_link(inner: Arc<Inner>, peer: Arc<Peer>) {
                 }
                 Outcome::ToTunnel(len, source) => {
                     let payload = Bytes::copy_from_slice(&scratch[..len]);
-                    // Enforce address ownership: a peer may only send from the
-                    // address derived for its own key.
-                    if source != Some(peer.overlay) {
+                    // Enforce address ownership: a peer may only send from an
+                    // address derived for its own key, in either family.
+                    let owned = match source {
+                        IpAddr::V6(addr) => addr == peer.overlay,
+                        IpAddr::V4(addr) => {
+                            let held = match peer.overlay_v4.lock() {
+                                Ok(guard) => *guard,
+                                Err(poisoned) => *poisoned.into_inner(),
+                            };
+                            held == Some(addr)
+                        }
+                    };
+                    if !owned {
                         peer.counters
                             .dropped_wrong_source
                             .fetch_add(1, Ordering::Relaxed);
@@ -538,7 +634,7 @@ async fn read_from_link(inner: Arc<Inner>, peer: Arc<Peer>) {
 
 enum Outcome {
     ToNetwork(usize),
-    ToTunnel(usize, Option<Ipv6Addr>),
+    ToTunnel(usize, IpAddr),
     Done,
     Failed,
 }
diff --git a/src/dataplane/wireguard/mod.rs b/src/dataplane/wireguard/mod.rs
index ec9b928..4ce9a77 100644
--- a/src/dataplane/wireguard/mod.rs
+++ b/src/dataplane/wireguard/mod.rs
@@ -54,7 +54,9 @@ pub use announcement::{ValidatedAnnouncement, WgAnnouncement};
 pub use config::{Cidr, DEFAULT_INTERFACE_PREFIX, MAX_INTERFACE_NAME_LEN, interface_name};
 pub use device::{PeerHealth, PeerStats, PeerSummary, WireguardDevice};
 pub use keys::{WgPublicKey, WgSecretKey};
-pub use overlay::{OVERLAY_PREFIX_LEN, overlay_address, overlay_prefix};
+pub use overlay::{
+    DEFAULT_IPV4_RANGE, OVERLAY_PREFIX_LEN, overlay_address, overlay_address_v4, overlay_prefix,
+};
 pub use packet::IpHeader;
 pub use plugin::{
     DEFAULT_MTU, MIN_MTU, NetworkOverview, PeerOverview, WIREGUARD_OVERHEAD, WIREGUARD_PROTOCOL,
diff --git a/src/dataplane/wireguard/overlay.rs b/src/dataplane/wireguard/overlay.rs
index 33b05bf..d68a151 100644
--- a/src/dataplane/wireguard/overlay.rs
+++ b/src/dataplane/wireguard/overlay.rs
@@ -20,7 +20,7 @@
 //!   but it cannot choose to collide with an existing member's address without
 //!   finding a hash preimage.
 
-use std::net::Ipv6Addr;
+use std::net::{Ipv4Addr, Ipv6Addr};
 
 use sha2::{Digest, Sha256};
 
@@ -37,6 +37,13 @@ pub const OVERLAY_PREFIX_LEN: u8 = 64;
 /// Prefix length of one member's address inside the overlay.
 pub const OVERLAY_HOST_PREFIX_LEN: u8 = 128;
 
+/// Default IPv4 overlay range: RFC 6598 shared address space.
+///
+/// Deliberately not RFC 1918, so it rarely collides with the home or office
+/// network the machine is already on. It can collide with a carrier-grade NAT
+/// that uses the same range, which is why it is configurable.
+pub const DEFAULT_IPV4_RANGE: (Ipv4Addr, u8) = (Ipv4Addr::new(100, 64, 0, 0), 10);
+
 fn push_lp(out: &mut Vec<u8>, bytes: &[u8]) {
     let len = u32::try_from(bytes.len()).unwrap_or(u32::MAX);
     out.extend_from_slice(&len.to_be_bytes());
@@ -85,6 +92,49 @@ pub fn overlay_address(network: NetworkId, key: &WgPublicKey) -> Ipv6Addr {
     Ipv6Addr::from(octets)
 }
 
+/// The IPv4 address a member with `key` has in `network`.
+///
+/// # Why this is weaker than the IPv6 derivation
+///
+/// A 64 bit interface identifier makes an IPv6 collision impossible in
+/// practice. IPv4 has nothing like that much room, so two members *can* derive
+/// the same address. In a `/10` with 50 members the chance is roughly 0.03%,
+/// which is small but real, so it is detected and resolved rather than
+/// assumed away — see [`super::device`]. IPv6 remains the address that always
+/// works.
+///
+/// Returns `None` when the range has no room for hosts.
+pub fn overlay_address_v4(
+    network: NetworkId,
+    key: &WgPublicKey,
+    range: (Ipv4Addr, u8),
+) -> Option<Ipv4Addr> {
+    let (base, prefix_len) = range;
+    if prefix_len > 32 {
+        return None;
+    }
+    let host_bits = 32 - u32::from(prefix_len);
+    // A usable range needs a network address, a broadcast address and at
+    // least one host between them.
+    if host_bits < 2 {
+        return None;
+    }
+
+    let hash = digest("ipv4", network, Some(key));
+    let raw = u32::from_be_bytes([hash[0], hash[1], hash[2], hash[3]]);
+
+    let usable = (1u64 << host_bits) - 2;
+    let offset = (u64::from(raw) % usable) + 1;
+
+    let mask = if host_bits == 32 {
+        0
+    } else {
+        u32::MAX << host_bits
+    };
+    let network_part = u32::from(base) & mask;
+    Some(Ipv4Addr::from(network_part | offset as u32))
+}
+
 #[cfg(test)]
 mod tests {
     #![allow(clippy::unwrap_used, clippy::expect_used, clippy::panic)]
@@ -132,6 +182,65 @@ mod tests {
         assert_ne!(overlay_prefix(first), overlay_prefix(second));
     }
 
+    #[test]
+    fn ipv4_addresses_land_inside_the_range_and_avoid_its_edges() {
+        let id = network("v4");
+        let range = DEFAULT_IPV4_RANGE;
+        for byte in 0..64u8 {
+            let key = WgPublicKey::from_bytes([byte; 32]);
+            let addr = overlay_address_v4(id, &key, range).unwrap();
+            let raw = u32::from(addr);
+            assert_eq!(
+                raw & 0xffc0_0000,
+                u32::from(range.0),
+                "outside 100.64.0.0/10"
+            );
+            // Never the network address and never the broadcast address.
+            assert_ne!(raw & 0x003f_ffff, 0);
+            assert_ne!(raw & 0x003f_ffff, 0x003f_ffff);
+        }
+    }
+
+    #[test]
+    fn ipv4_derivation_is_deterministic_and_scoped_like_ipv6() {
+        let key = WgPublicKey::from_bytes([9u8; 32]);
+        let first = network("one");
+        let second = network("two");
+        let range = DEFAULT_IPV4_RANGE;
+
+        assert_eq!(
+            overlay_address_v4(first, &key, range),
+            overlay_address_v4(first, &key, range)
+        );
+        assert_ne!(
+            overlay_address_v4(first, &key, range),
+            overlay_address_v4(second, &key, range)
+        );
+        assert_ne!(
+            overlay_address_v4(first, &key, range),
+            overlay_address_v4(first, &WgPublicKey::from_bytes([10u8; 32]), range)
+        );
+        // A different range moves everybody.
+        assert_ne!(
+            overlay_address_v4(first, &key, range),
+            overlay_address_v4(first, &key, (Ipv4Addr::new(10, 0, 0, 0), 8))
+        );
+    }
+
+    #[test]
+    fn a_range_with_no_room_yields_nothing() {
+        let id = network("tiny");
+        let key = WgPublicKey::from_bytes([1u8; 32]);
+        // /31 and /32 have no usable host addresses.
+        assert!(overlay_address_v4(id, &key, (Ipv4Addr::new(10, 0, 0, 0), 31)).is_none());
+        assert!(overlay_address_v4(id, &key, (Ipv4Addr::new(10, 0, 0, 1), 32)).is_none());
+        assert!(overlay_address_v4(id, &key, (Ipv4Addr::new(10, 0, 0, 0), 33)).is_none());
+        // A /30 has two usable addresses.
+        assert!(overlay_address_v4(id, &key, (Ipv4Addr::new(10, 0, 0, 0), 30)).is_some());
+        // A /0 must not overflow.
+        assert!(overlay_address_v4(id, &key, (Ipv4Addr::UNSPECIFIED, 0)).is_some());
+    }
+
     #[test]
     fn addresses_are_never_the_subnet_router_anycast_address() {
         let id = network("anycast");
diff --git a/src/dataplane/wireguard/packet.rs b/src/dataplane/wireguard/packet.rs
index 60ae5ab..e6c36ea 100644
--- a/src/dataplane/wireguard/packet.rs
+++ b/src/dataplane/wireguard/packet.rs
@@ -65,6 +65,22 @@ impl IpHeader {
             IpHeader::V4 { .. } => None,
         }
     }
+
+    /// The destination, whichever family it is.
+    pub fn destination(&self) -> std::net::IpAddr {
+        match self {
+            IpHeader::V4 { destination, .. } => std::net::IpAddr::V4(*destination),
+            IpHeader::V6 { destination, .. } => std::net::IpAddr::V6(*destination),
+        }
+    }
+
+    /// The source, whichever family it is.
+    pub fn source(&self) -> std::net::IpAddr {
+        match self {
+            IpHeader::V4 { source, .. } => std::net::IpAddr::V4(*source),
+            IpHeader::V6 { source, .. } => std::net::IpAddr::V6(*source),
+        }
+    }
 }
 
 #[cfg(test)]
diff --git a/src/dataplane/wireguard/plugin.rs b/src/dataplane/wireguard/plugin.rs
index 735c11e..08211d9 100644
--- a/src/dataplane/wireguard/plugin.rs
+++ b/src/dataplane/wireguard/plugin.rs
@@ -25,7 +25,7 @@
 //! A failure here is reported and retried. It never stops the control plane.
 
 use std::collections::{BTreeSet, HashMap};
-use std::net::IpAddr;
+use std::net::{IpAddr, Ipv4Addr};
 use std::path::PathBuf;
 use std::sync::{Arc, Mutex, OnceLock};
 use std::time::Duration;
@@ -43,7 +43,9 @@ use super::announcement::{ValidatedAnnouncement, WgAnnouncement};
 use super::config::{DEFAULT_INTERFACE_PREFIX, interface_name};
 use super::device::{PeerSummary, WireguardDevice};
 use super::keys::{WgPublicKey, WgSecretKey};
-use super::overlay::{OVERLAY_PREFIX_LEN, overlay_address, overlay_prefix};
+use super::overlay::{
+    DEFAULT_IPV4_RANGE, OVERLAY_PREFIX_LEN, overlay_address, overlay_address_v4, overlay_prefix,
+};
 use super::store::WgKeyStore;
 use super::tun::{TunFactory, TunRequest};
 
@@ -86,6 +88,12 @@ pub struct WireguardConfig {
     pub keepalive: Option<u16>,
     /// Interface MTU. See [`DEFAULT_MTU`].
     pub mtu: u32,
+    /// IPv4 overlay range, or `None` for an IPv6-only overlay.
+    ///
+    /// IPv6 addresses are derived collision-free; IPv4 ones cannot be, so a
+    /// collision is detected and resolved deterministically instead. See
+    /// `docs/wireguard.md`.
+    pub ipv4_range: Option<(Ipv4Addr, u8)>,
     /// How long to coalesce changes before reconciling.
     pub reconcile_debounce: Duration,
     /// How often to reconcile anyway, which is also when a packet interface
@@ -101,6 +109,7 @@ impl WireguardConfig {
             interface_prefix: DEFAULT_INTERFACE_PREFIX.to_string(),
             keepalive: Some(25),
             mtu: DEFAULT_MTU,
+            ipv4_range: Some(DEFAULT_IPV4_RANGE),
             reconcile_debounce: Duration::from_millis(200),
             reconcile_interval: Duration::from_secs(15),
         }
@@ -120,6 +129,12 @@ impl WireguardConfig {
         self
     }
 
+    /// Sets the IPv4 overlay range, or disables IPv4 with `None`.
+    pub fn with_ipv4_range(mut self, range: Option<(Ipv4Addr, u8)>) -> Self {
+        self.ipv4_range = range;
+        self
+    }
+
     /// Sets the reconciliation timings.
     pub fn with_reconcile(mut self, debounce: Duration, interval: Duration) -> Self {
         self.reconcile_debounce = debounce;
@@ -150,6 +165,10 @@ pub struct NetworkOverview {
     pub overlay_prefix: IpAddr,
     /// Prefix length of the overlay subnet.
     pub overlay_prefix_len: u8,
+    /// This agent's IPv4 overlay address, when the overlay is dual stack.
+    pub overlay_address_v4: Option<Ipv4Addr>,
+    /// The IPv4 overlay range in use.
+    pub ipv4_range: Option<(Ipv4Addr, u8)>,
     /// Peers this agent knows about.
     pub peers: Vec<PeerOverview>,
     /// Unicast packets the operating system sent to an address no peer owns.
@@ -174,6 +193,8 @@ pub struct PeerOverview {
     pub public_key: WgPublicKey,
     /// The overlay address derived for it locally.
     pub overlay_address: IpAddr,
+    /// Its IPv4 overlay address, once a tunnel exists and it won the address.
+    pub overlay_address_v4: Option<Ipv4Addr>,
     /// Whether a data plane link to it exists.
     pub has_link: bool,
     /// The running tunnel, once there is a link.
@@ -309,6 +330,9 @@ impl WireguardPlugin {
                 endpoint_id: *endpoint_id,
                 public_key: announcement.public_key,
                 overlay_address: IpAddr::V6(announcement.overlay_address),
+                overlay_address_v4: tunnels
+                    .get(&announcement.public_key)
+                    .and_then(|tunnel| tunnel.overlay_address_v4),
                 has_link: state.links.contains_key(endpoint_id),
                 tunnel: tunnels.get(&announcement.public_key).cloned(),
             })
@@ -323,6 +347,12 @@ impl WireguardPlugin {
             overlay_address: IpAddr::V6(overlay_address(network, &state.key.public())),
             overlay_prefix: IpAddr::V6(overlay_prefix(network)),
             overlay_prefix_len: OVERLAY_PREFIX_LEN,
+            overlay_address_v4: self
+                .worker
+                .config
+                .ipv4_range
+                .and_then(|range| overlay_address_v4(network, &state.key.public(), range)),
+            ipv4_range: self.worker.config.ipv4_range,
             peers,
             unroutable_packets: state
                 .device
@@ -437,10 +467,18 @@ impl Worker {
             name: name.clone(),
             address: overlay_address(network, &key.public()),
             prefix_len: OVERLAY_PREFIX_LEN,
+            address_v4: self.config.ipv4_range.and_then(|range| {
+                overlay_address_v4(network, &key.public(), range).map(|address| (address, range.1))
+            }),
             mtu: self.config.mtu,
         };
         let tun = self.tun_factory.create(request).await?;
-        let device = Arc::new(WireguardDevice::start(network, key, tun));
+        let device = Arc::new(WireguardDevice::start(
+            network,
+            key,
+            tun,
+            self.config.ipv4_range,
+        ));
 
         let mut shared = self.lock_shared();
         if let Some(state) = shared.networks.get_mut(&network) {
diff --git a/src/dataplane/wireguard/tun.rs b/src/dataplane/wireguard/tun.rs
index f2fa481..b78578f 100644
--- a/src/dataplane/wireguard/tun.rs
+++ b/src/dataplane/wireguard/tun.rs
@@ -12,7 +12,7 @@
 //! * `SystemTun`, behind the `tun-device` feature, is a real TUN interface.
 //!   Creating one needs `CAP_NET_ADMIN` on Linux or the equivalent elsewhere.
 
-use std::net::Ipv6Addr;
+use std::net::{Ipv4Addr, Ipv6Addr};
 use std::sync::Arc;
 
 use bytes::Bytes;
@@ -29,6 +29,8 @@ pub struct TunRequest {
     pub address: Ipv6Addr,
     /// Prefix length of the overlay subnet, so the OS routes it here.
     pub prefix_len: u8,
+    /// The IPv4 overlay address and its prefix length, when dual stack.
+    pub address_v4: Option<(Ipv4Addr, u8)>,
     /// Interface MTU.
     pub mtu: u32,
 }
@@ -381,7 +383,7 @@ mod system {
     /// reason: duplicate address detection can never finish with no carrier,
     /// and the address would stay tentative and unusable.
     pub fn setup_commands(request: &TunRequest, user: &str) -> Vec<String> {
-        vec![
+        let mut commands = vec![
             format!(
                 "sudo ip tuntap add dev {} mode tun user {user}",
                 request.name
@@ -398,7 +400,16 @@ mod system {
                 "sudo ip -6 address add {}/{} dev {} nodad",
                 request.address, request.prefix_len, request.name
             ),
-        ]
+        ];
+        if let Some((address, prefix_len)) = request.address_v4 {
+            // IPv4 is not sensitive to carrier the way IPv6 is, so it needs
+            // no extra settings.
+            commands.push(format!(
+                "sudo ip address add {address}/{prefix_len} dev {}",
+                request.name
+            ));
+        }
+        commands
     }
 
     fn current_user() -> String {
@@ -578,7 +589,8 @@ fd559caf9652cb86321feac65c73bd84 05 40 00 08 tsun0
             name: "tsun0".into(),
             address: "fd00::1".parse().unwrap(),
             prefix_len: 64,
-            mtu: 1100,
+            address_v4: Some(("100.64.1.2".parse().unwrap(), 10)),
+            mtu: 1280,
         };
         let commands = setup_commands(&request, "someone");
 
@@ -586,11 +598,24 @@ fd559caf9652cb86321feac65c73bd84 05 40 00 08 tsun0
         // must survive losing carrier, and it must not wait for duplicate
         // address detection that can never complete.
         let joined = commands.join("\n");
-        let up = joined.find("link set dev tsun0 mtu 1100 up").unwrap();
+        let up = joined.find("link set dev tsun0 mtu 1280 up").unwrap();
         let keep = joined.find("keep_addr_on_down=1").unwrap();
         let add = joined.find("address add fd00::1/64").unwrap();
         assert!(up < keep && keep < add, "wrong order:\n{joined}");
         assert!(joined.contains("nodad"));
         assert!(joined.contains("user someone"));
+        // IPv4 needs no carrier tricks, just the address.
+        assert!(joined.contains("ip address add 100.64.1.2/10 dev tsun0"));
+
+        // An IPv6-only overlay says nothing about IPv4.
+        let v6_only = TunRequest {
+            address_v4: None,
+            ..request
+        };
+        assert!(
+            !setup_commands(&v6_only, "someone")
+                .join("\n")
+                .contains("100.64")
+        );
     }
 }
diff --git a/src/ipc/mod.rs b/src/ipc/mod.rs
index ef27bed..196b8c1 100644
--- a/src/ipc/mod.rs
+++ b/src/ipc/mod.rs
@@ -136,6 +136,8 @@ pub struct OverlayReport {
     pub mtu: u32,
     /// This agent's overlay address.
     pub address: String,
+    /// This agent's IPv4 overlay address, when the overlay is dual stack.
+    pub address_v4: Option<String>,
     /// The subnet every member shares.
     pub prefix: String,
     /// Prefix length of that subnet.
@@ -155,6 +157,8 @@ pub struct OverlayPeerReport {
     pub public_key: String,
     /// Its overlay address.
     pub address: String,
+    /// Its IPv4 overlay address, when it has one.
+    pub address_v4: Option<String>,
     /// Seconds since the last WireGuard handshake.
     ///
     /// `None` means the tunnel has never handshaken and cannot carry traffic.
@@ -229,10 +233,14 @@ impl StatusReport {
                 let up = overlay.peers.iter().filter(|peer| peer.is_up()).count();
                 let _ = writeln!(
                     out,
-                    "  overlay {} {}/{} mtu {}  {}/{} tunnel(s) up",
+                    "  overlay {} {}/{}{} mtu {}  {}/{} tunnel(s) up",
                     overlay.interface,
                     overlay.address,
                     overlay.prefix_len,
+                    match &overlay.address_v4 {
+                        Some(v4) => format!(" and {v4}"),
+                        None => String::new(),
+                    },
                     overlay.mtu,
                     up,
                     overlay.peers.len()
@@ -240,9 +248,13 @@ impl StatusReport {
                 for peer in &overlay.peers {
                     let _ = writeln!(
                         out,
-                        "    {}  {}  {}  tx {} rx {}{}  {}",
+                        "    {}  {}{}  {}  tx {} rx {}{}  {}",
                         &peer.public_key[..8.min(peer.public_key.len())],
                         peer.address,
+                        match &peer.address_v4 {
+                            Some(v4) => format!(" / {v4}"),
+                            None => String::new(),
+                        },
                         match peer.handshake_secs_ago {
                             Some(secs) => format!("handshake {secs}s ago"),
                             None => "NOT HANDSHAKEN".to_string(),
diff --git a/tests/wireguard.rs b/tests/wireguard.rs
index 3a81ac0..fc37099 100644
--- a/tests/wireguard.rs
+++ b/tests/wireguard.rs
@@ -10,7 +10,7 @@
 
 mod common;
 
-use std::net::{IpAddr, Ipv6Addr};
+use std::net::{IpAddr, Ipv4Addr, Ipv6Addr};
 use std::sync::Arc;
 use std::time::Duration;
 
@@ -123,6 +123,24 @@ impl WgAgent {
     }
 }
 
+/// Builds a minimal well-formed IPv4 packet.
+fn ipv4_packet(source: Ipv4Addr, destination: Ipv4Addr, payload: &[u8]) -> Bytes {
+    let total = 20 + payload.len();
+    let mut packet = Vec::with_capacity(total);
+    packet.push((4 << 4) | 5); // version 4, header length 5 words
+    packet.push(0); // dscp/ecn
+    packet.extend_from_slice(&(total as u16).to_be_bytes());
+    packet.extend_from_slice(&[0, 0]); // identification
+    packet.extend_from_slice(&[0, 0]); // flags and fragment offset
+    packet.push(64); // ttl
+    packet.push(253); // an experimental protocol number
+    packet.extend_from_slice(&[0, 0]); // checksum, not verified by the overlay
+    packet.extend_from_slice(&source.octets());
+    packet.extend_from_slice(&destination.octets());
+    packet.extend_from_slice(payload);
+    Bytes::from(packet)
+}
+
 /// Builds a minimal well-formed IPv6 packet.
 fn ipv6_packet(source: Ipv6Addr, destination: Ipv6Addr, payload: &[u8]) -> Bytes {
     let mut packet = Vec::with_capacity(40 + payload.len());
@@ -244,6 +262,151 @@ async fn a_peer_cannot_send_from_an_address_it_does_not_own() {
     b.shutdown().await;
 }
 
+#[tokio::test]
+async fn the_overlay_carries_ipv4_alongside_ipv6() {
+    let discovery = SharedMemoryDiscovery::new();
+    let (name, secret) = network("wg-dual-stack");
+
+    let a = WgAgent::spawn(&discovery, "ta").await;
+    let b = WgAgent::spawn(&discovery, "tb").await;
+
+    let network_id = a.agent.join_network(&name, &secret).await.unwrap();
+    b.agent.join_network(&name, &secret).await.unwrap();
+    a.wait_for_tunnels(network_id, 1).await;
+    b.wait_for_tunnels(network_id, 1).await;
+
+    let view_a = a.plugin.overview(network_id).unwrap();
+    let view_b = b.plugin.overview(network_id).unwrap();
+
+    let v4_a = view_a.overlay_address_v4.expect("dual stack by default");
+    let v4_b = view_b.overlay_address_v4.expect("dual stack by default");
+    assert_ne!(v4_a, v4_b);
+    // Both inside the configured range.
+    for addr in [v4_a, v4_b] {
+        assert_eq!(
+            u32::from(addr) & 0xffc0_0000,
+            u32::from(Ipv4Addr::new(100, 64, 0, 0))
+        );
+    }
+    // Each side derived the other's address identically.
+    assert_eq!(view_a.peers[0].overlay_address_v4, Some(v4_b));
+    assert_eq!(view_b.peers[0].overlay_address_v4, Some(v4_a));
+    assert!(!view_a.peers[0].tunnel.as_ref().unwrap().ipv4_conflict);
+
+    let tun_a = a.tun(network_id).await;
+    let tun_b = b.tun(network_id).await;
+
+    // A real IPv4 packet through the same tunnel.
+    tun_a.push_from_os(ipv4_packet(v4_a, v4_b, b"ipv4 over the overlay"));
+    let received = tokio::time::timeout(common::DEADLINE, tun_b.pop_to_os())
+        .await
+        .expect("the IPv4 packet should arrive")
+        .unwrap();
+    assert_eq!(received[0] >> 4, 4, "still an IPv4 packet");
+    assert_eq!(&received[12..16], &v4_a.octets());
+    assert_eq!(&received[16..20], &v4_b.octets());
+    assert_eq!(&received[20..], b"ipv4 over the overlay");
+
+    // IPv6 keeps working on the same tunnel.
+    let addr_a = a.overlay(network_id).await;
+    let addr_b = b.overlay(network_id).await;
+    tun_a.push_from_os(ipv6_packet(addr_a, addr_b, b"and ipv6 too"));
+    let received = tokio::time::timeout(common::DEADLINE, tun_b.pop_to_os())
+        .await
+        .expect("the IPv6 packet should arrive")
+        .unwrap();
+    assert_eq!(received[0] >> 4, 6);
+    assert_eq!(&received[40..], b"and ipv6 too");
+
+    a.shutdown().await;
+    b.shutdown().await;
+}
+
+#[tokio::test]
+async fn an_ipv4_source_a_peer_does_not_own_is_dropped() {
+    let discovery = SharedMemoryDiscovery::new();
+    let (name, secret) = network("wg-v4-spoof");
+
+    let a = WgAgent::spawn(&discovery, "ta").await;
+    let b = WgAgent::spawn(&discovery, "tb").await;
+    let network_id = a.agent.join_network(&name, &secret).await.unwrap();
+    b.agent.join_network(&name, &secret).await.unwrap();
+    a.wait_for_tunnels(network_id, 1).await;
+    b.wait_for_tunnels(network_id, 1).await;
+
+    let v4_a = a
+        .plugin
+        .overview(network_id)
+        .unwrap()
+        .overlay_address_v4
+        .unwrap();
+    let v4_b = b
+        .plugin
+        .overview(network_id)
+        .unwrap()
+        .overlay_address_v4
+        .unwrap();
+    let tun_a = a.tun(network_id).await;
+    let tun_b = b.tun(network_id).await;
+
+    // A claims an IPv4 address that is not the one derived from its key.
+    let forged = Ipv4Addr::from(u32::from(v4_a) ^ 0x0000_00ff);
+    tun_a.push_from_os(ipv4_packet(forged, v4_b, b"spoofed v4"));
+
+    wait_until("the spoofed IPv4 packet is dropped", || async {
+        let view = b.plugin.overview(network_id)?;
+        let tunnel = view.peers.first()?.tunnel.as_ref()?;
+        (tunnel.stats.dropped_wrong_source >= 1).then_some(())
+    })
+    .await;
+
+    // The honest one still gets through.
+    tun_a.push_from_os(ipv4_packet(v4_a, v4_b, b"honest v4"));
+    let received = tokio::time::timeout(common::DEADLINE, tun_b.pop_to_os())
+        .await
+        .expect("the honest packet should arrive")
+        .unwrap();
+    assert_eq!(&received[20..], b"honest v4");
+
+    a.shutdown().await;
+    b.shutdown().await;
+}
+
+#[tokio::test]
+async fn an_ipv6_only_overlay_can_be_asked_for() {
+    let discovery = SharedMemoryDiscovery::new();
+    let (name, secret) = network("wg-v6-only");
+
+    let a = WgAgent::spawn_with(&discovery, "ta", |config| config.with_ipv4_range(None)).await;
+    let b = WgAgent::spawn_with(&discovery, "tb", |config| config.with_ipv4_range(None)).await;
+
+    let network_id = a.agent.join_network(&name, &secret).await.unwrap();
+    b.agent.join_network(&name, &secret).await.unwrap();
+    a.wait_for_tunnels(network_id, 1).await;
+
+    let view = a.plugin.overview(network_id).unwrap();
+    assert_eq!(view.overlay_address_v4, None);
+    assert_eq!(view.ipv4_range, None);
+    assert_eq!(view.peers[0].overlay_address_v4, None);
+    // No IPv4 configured is not a conflict.
+    assert!(!view.peers[0].tunnel.as_ref().unwrap().ipv4_conflict);
+
+    // IPv6 is unaffected.
+    let addr_a = a.overlay(network_id).await;
+    let addr_b = b.overlay(network_id).await;
+    a.tun(network_id)
+        .await
+        .push_from_os(ipv6_packet(addr_a, addr_b, b"v6 only"));
+    let received = tokio::time::timeout(common::DEADLINE, b.tun(network_id).await.pop_to_os())
+        .await
+        .expect("the packet should arrive")
+        .unwrap();
+    assert_eq!(&received[40..], b"v6 only");
+
+    a.shutdown().await;
+    b.shutdown().await;
+}
+
 #[tokio::test]
 async fn packets_for_an_unknown_address_are_counted_not_broadcast() {
     let discovery = SharedMemoryDiscovery::new();