v1.19.2

changelog.md

@@ -1,5 +1,58 @@
 # Changelog
 
+## 2026-04-06 - 1.19.2 - fix(server)
+clean up bridge and hybrid shutdown handling
+
+- persist bridge teardown metadata so stop() can restore host IP configuration and remove the bridge in bridge and hybrid modes
+- use separate shutdown channels for hybrid socket and bridge engines to stop both forwarding paths correctly
+- avoid IP pool leaks when client registration fails and ignore unspecified IPv4 addresses when selecting WireGuard peer addresses
+- make daemon bridge stop await nftables cleanup and process exit, and cap effective tunnel MTU to the link MTU
+
+## 2026-04-01 - 1.19.1 - fix(rust)
+clean up unused Rust warnings in bridge, network, and server modules
+
+- remove the unused error import from the bridge module
+- mark IpPool.prefix_len as intentionally unused to suppress dead code warnings
+- rename the unused socket shutdown sender binding in the server to an underscore-prefixed variable
+
+## 2026-04-01 - 1.19.0 - feat(forwarding)
+add hybrid forwarding mode with per-client bridge and VLAN settings
+
+- introduces a new hybrid forwarding mode that routes each client through either userspace NAT or bridge mode based on per-client configuration
+- adds per-client bridge options including useHostIp, DHCP, static LAN IP, and VLAN assignment fields to the server and TypeScript interfaces
+- adds Linux bridge VLAN helper functions and updates documentation to cover hybrid mode and VLAN-capable bridge clients
+
+## 2026-03-31 - 1.18.0 - feat(server)
+add bridge forwarding mode and per-client destination policy overrides
+
+- introduces Linux bridge-based forwarding so VPN clients can receive IPs from a LAN subnet via TAP/bridge integration
+- adds bridge server configuration options for LAN subnet, physical interface, and client IP allocation range
+- adds per-client destinationPolicy overrides in the client registry and applies them in the userspace NAT engine based on assigned tunnel IP
+- extends IP pool allocation to support constrained address ranges needed for bridge mode
+- updates TypeScript interfaces and documentation to cover bridge mode and per-client destination policy behavior
+
+## 2026-03-31 - 1.17.1 - fix(readme)
+document per-transport metrics and handshake-driven WireGuard connection state
+
+- Add README examples for getStatistics() per-transport active client and total connection counters
+- Clarify that WireGuard peers are marked connected only after a successful handshake and disconnect after idle timeout
+- Refresh API and project structure documentation to reflect newly documented stats fields and source files
+
+## 2026-03-31 - 1.17.0 - feat(wireguard)
+track per-transport server statistics and make WireGuard clients active only after handshake
+
+- add websocket, quic, and wireguard active-client and total-connection counters to server statistics
+- register WireGuard peers without marking them active until handshake/data is received, and remove them from active clients on expiration or idle timeout
+- sync WireGuard byte counters into aggregate server stats independently of active client presence and expose new statistics fields in TypeScript interfaces
+
+## 2026-03-31 - 1.16.5 - fix(rust-userspace-nat)
+improve TCP session backpressure, buffering, and idle cleanup in userspace NAT
+
+- apply proper bridge-channel backpressure by reserving channel capacity before consuming smoltcp TCP data
+- defer bridge sender initialization until the bridge task starts and track TCP session activity timestamps
+- cap per-session pending TCP send buffers at 512KB and abort stalled sessions when clients cannot keep up
+- add idle TCP session cleanup and switch NAT polling to a dynamic smoltcp-driven delay
+
 ## 2026-03-31 - 1.16.4 - fix(server)
 register preloaded WireGuard clients as peers on server startup
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@push.rocks/smartvpn",
-  "version": "1.16.4",
+  "version": "1.19.2",
   "private": false,
   "description": "A VPN solution with TypeScript control plane and Rust data plane daemon",
   "type": "module",

readme.md

@@ -2,15 +2,19 @@
 
 A high-performance VPN solution with a **TypeScript control plane** and a **Rust data plane daemon**. Enterprise-ready client authentication, triple transport support (WebSocket + QUIC + WireGuard), and a typed hub API for managing clients from code.
 
-🔐 **Noise IK** mutual authentication — per-client X25519 keypairs, server-side registry
+- 🔐 **Noise IK** mutual authentication — per-client X25519 keypairs, server-side registry
-🚀 **Triple transport**: WebSocket (Cloudflare-friendly), raw **QUIC** (datagrams), and **WireGuard** (standard protocol)
+- 🚀 **Triple transport**: WebSocket (Cloudflare-friendly), raw **QUIC** (datagrams), and **WireGuard** (standard protocol)
-🛡️ **ACL engine** — deny-overrides-allow IP filtering, aligned with SmartProxy conventions
+- 🛡️ **ACL engine** — deny-overrides-allow IP filtering, aligned with SmartProxy conventions
-🔀 **PROXY protocol v2** — real client IPs behind reverse proxies (HAProxy, SmartProxy, Cloudflare Spectrum)
+- 🔀 **PROXY protocol v2** — real client IPs behind reverse proxies (HAProxy, SmartProxy, Cloudflare Spectrum)
-📊 **Adaptive QoS**: per-client rate limiting, priority queues, connection quality tracking
+- 📊 **Per-transport metrics**: active clients and total connections broken down by websocket, QUIC, and WireGuard
-🔄 **Hub API**: one `createClient()` call generates keys, assigns IP, returns both SmartVPN + WireGuard configs
+- 🔄 **Hub API**: one `createClient()` call generates keys, assigns IP, returns both SmartVPN + WireGuard configs
-📡 **Real-time telemetry**: RTT, jitter, loss ratio, link health — all via typed APIs
+- 📡 **Real-time telemetry**: RTT, jitter, loss ratio, link health — all via typed APIs
-🌐 **Unified forwarding pipeline**: all transports share the same engine — TUN (kernel), userspace NAT (no root), or testing mode
+- 🌐 **Unified forwarding pipeline**: all transports share the same engine — TUN (kernel), userspace NAT (no root), L2 bridge, hybrid, or testing mode
-🎯 **Destination routing policy**: force-target, block, or allow traffic per destination with nftables integration
+- 🏠 **Bridge mode**: VPN clients get IPs from your LAN subnet — seamlessly bridge remote clients onto a physical network
+- 🔀 **Hybrid mode**: per-client routing — some clients bridge to the LAN, others use userspace NAT, all on the same server
+- 🏷️ **VLAN support**: assign individual clients to 802.1Q VLANs on the bridge
+- 🎯 **Destination routing policy**: force-target, block, or allow traffic per destination with nftables integration
+- ⚡ **Handshake-driven WireGuard state**: peers appear as "connected" only after a successful WireGuard handshake, and auto-disconnect on idle timeout
 
 ## Issue Reporting and Security
 
@@ -83,7 +87,7 @@ await server.start({
   publicKey: '<server-noise-public-key-base64>',
   subnet: '10.8.0.0/24',
   transportMode: 'all',       // WebSocket + QUIC + WireGuard simultaneously (default)
-  forwardingMode: 'tun',      // 'tun' (kernel), 'socket' (userspace NAT), or 'testing'
+  forwardingMode: 'tun',      // 'tun' | 'socket' | 'bridge' | 'hybrid' | 'testing'
   wgPrivateKey: '<server-wg-private-key-base64>', // required for WireGuard transport
   enableNat: true,
   dns: ['1.1.1.1', '8.8.8.8'],
@@ -140,6 +144,30 @@ Every client authenticates with a **Noise IK handshake** (`Noise_IK_25519_ChaCha
 
 The server runs **all three simultaneously** by default with `transportMode: 'all'`. All transports share the same unified forwarding pipeline (`ForwardingEngine`), IP pool, client registry, and stats — so WireGuard peers get the same userspace NAT, rate limiting, and monitoring as WS/QUIC clients. Clients auto-negotiate with `transport: 'auto'` (tries QUIC first, falls back to WS).
 
+### 📊 Per-Transport Metrics
+
+Server statistics include per-transport breakdowns so you can see exactly how many clients use each protocol:
+
+```typescript
+const stats = await server.getStatistics();
+
+// Aggregate
+console.log(stats.activeClients);       // total connected clients
+console.log(stats.totalConnections);    // total connections since start
+
+// Per-transport active clients
+console.log(stats.activeClientsWebsocket);   // currently connected via WS
+console.log(stats.activeClientsQuic);        // currently connected via QUIC
+console.log(stats.activeClientsWireguard);   // currently connected via WireGuard
+
+// Per-transport total connections
+console.log(stats.totalConnectionsWebsocket);
+console.log(stats.totalConnectionsQuic);
+console.log(stats.totalConnectionsWireguard);
+```
+
+**WireGuard connection state is handshake-driven** — registered WireGuard peers do NOT appear as "connected" until their first successful WireGuard handshake completes. They automatically disconnect after 180 seconds of inactivity or when boringtun reports `ConnectionExpired`. This matches how WebSocket/QUIC clients behave: they appear on connection and disappear on disconnect.
+
 ### 🛡️ ACL Engine (SmartProxy-Aligned)
 
 Security policies per client, using the same `ipAllowList` / `ipBlockList` naming convention as `@push.rocks/smartproxy`:
@@ -212,6 +240,21 @@ In **TUN mode**, destination policies are enforced via **nftables** rules (using
 
 In **socket mode**, the policy is evaluated in the userspace NAT engine before per-client ACLs.
 
+**Per-client override** — individual clients can have their own destination policy that overrides the server-level default:
+
+```typescript
+await server.createClient({
+  clientId: 'restricted-client',
+  security: {
+    destinationPolicy: {
+      default: 'block',                         // block everything by default
+      allowList: ['10.0.0.0/8'],                 // except internal network
+    },
+    // ... other security settings
+  },
+});
+```
+
 ### 🔗 Socket Forward Proxy Protocol
 
 When using `forwardingMode: 'socket'` (userspace NAT), you can prepend **PROXY protocol v2 headers** on outbound TCP connections. This conveys the VPN client's tunnel IP as the source address to downstream services (e.g., SmartProxy):
@@ -226,12 +269,14 @@ await server.start({
 
 ### 📦 Packet Forwarding Modes
 
-SmartVPN supports three forwarding modes, configurable per-server and per-client:
+SmartVPN supports five forwarding modes, configurable per-server:
 
 | Mode | Flag | Description | Root Required |
 |------|------|-------------|---------------|
 | **TUN** | `'tun'` | Kernel TUN device — real packet forwarding with system routing | ✅ Yes |
 | **Userspace NAT** | `'socket'` | Userspace TCP/UDP proxy via `connect(2)` — no TUN, no root needed | ❌ No |
+| **Bridge** | `'bridge'` | L2 bridge — VPN clients get IPs from a physical LAN subnet | ✅ Yes |
+| **Hybrid** | `'hybrid'` | Per-client routing: some clients use socket NAT, others use bridge — both engines run simultaneously | ✅ Yes |
 | **Testing** | `'testing'` | Monitoring only — packets are counted but not forwarded | ❌ No |
 
 ```typescript
@@ -242,6 +287,23 @@ await server.start({
   enableNat: true,
 });
 
+// Server with bridge mode — VPN clients appear on the LAN
+await server.start({
+  // ...
+  forwardingMode: 'bridge',
+  bridgeLanSubnet: '192.168.1.0/24',      // LAN subnet to bridge into
+  bridgePhysicalInterface: 'eth0',         // auto-detected if omitted
+  bridgeIpRangeStart: 200,                 // clients get .200–.250 (defaults)
+  bridgeIpRangeEnd: 250,
+});
+
+// Server with hybrid mode — per-client routing
+await server.start({
+  // ...
+  forwardingMode: 'hybrid',
+  bridgePhysicalInterface: 'eth0',         // for bridge clients
+});
+
 // Client with TUN device
 const { assignedIp } = await client.connect({
   // ...
@@ -249,15 +311,64 @@ const { assignedIp } = await client.connect({
 });
 ```
 
-The userspace NAT mode extracts destination IP/port from IP packets, opens a real socket to the destination, and relays data — supporting both TCP streams and UDP datagrams without requiring `CAP_NET_ADMIN` or root privileges.
+The **userspace NAT** mode extracts destination IP/port from IP packets, opens a real socket to the destination, and relays data — supporting both TCP streams and UDP datagrams without requiring `CAP_NET_ADMIN` or root privileges.
+
+The **bridge** mode assigns VPN clients IPs from a real LAN subnet instead of a virtual VPN subnet. Clients appear as if they're directly on the physical network — perfect for remote access to home labs, office networks, or IoT devices.
+
+The **hybrid** mode runs both engines simultaneously with a **per-client routing table**. Each client's `useHostIp` flag determines whether its packets go through the bridge (L2, LAN IP) or socket NAT (userspace, VPN IP). This is ideal when most clients need internet NAT but some need direct LAN access.
+
+### 🏠 Per-Client Bridge & VLAN Settings
+
+When using `bridge` or `hybrid` mode, each client can be individually configured for LAN bridging, static IPs, DHCP, and 802.1Q VLAN assignment:
+
+```typescript
+// Client that bridges to the LAN with a static IP
+await server.createClient({
+  clientId: 'office-printer',
+  useHostIp: true,           // bridge to LAN instead of VPN subnet
+  staticIp: '192.168.1.210', // fixed LAN IP
+});
+
+// Client that gets a LAN IP via DHCP
+await server.createClient({
+  clientId: 'roaming-laptop',
+  useHostIp: true,
+  useDhcp: true,             // obtain IP from LAN DHCP server
+});
+
+// Client on a specific VLAN
+await server.createClient({
+  clientId: 'iot-sensor',
+  useHostIp: true,
+  forceVlan: true,
+  vlanId: 100,               // 802.1Q VLAN ID (1-4094)
+});
+
+// Regular NAT client (default, no bridge)
+await server.createClient({
+  clientId: 'remote-worker',
+  // useHostIp defaults to false → uses socket NAT
+});
+```
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `useHostIp` | `boolean` | `true` = bridge to LAN (host IP), `false` = VPN subnet via NAT (default) |
+| `useDhcp` | `boolean` | When `useHostIp` is true, obtain IP via DHCP relay instead of static/auto-assign |
+| `staticIp` | `string` | Fixed LAN IP when `useHostIp` is true and `useDhcp` is false |
+| `forceVlan` | `boolean` | Assign this client to a specific 802.1Q VLAN on the bridge |
+| `vlanId` | `number` | VLAN ID (1-4094), required when `forceVlan` is true |
+
+VLAN support uses Linux bridge VLAN filtering — each client's TAP port gets tagged with the specified VLAN ID, isolating traffic at Layer 2.
 
 ### 📊 Telemetry & QoS
 
 - **Connection quality**: Smoothed RTT, jitter, min/max RTT, loss ratio, link health (`healthy` / `degraded` / `critical`)
 - **Adaptive keepalives**: Interval adjusts based on link health (60s → 30s → 10s)
 - **Per-client rate limiting**: Token bucket with configurable bytes/sec and burst
-- **Dead-peer detection**: 180s inactivity timeout
+- **Dead-peer detection**: 180s inactivity timeout (all transports)
 - **MTU management**: Automatic overhead calculation (IP+TCP+WS+Noise = 79 bytes)
+- **Per-transport stats**: Active client and total connection counts broken down by websocket, QUIC, and WireGuard
 
 ### 🏷️ Client Tags (Trusted vs Informational)
 
@@ -418,13 +529,14 @@ server.on('reconnected', () => { /* socket transport reconnected */ });
 
 | Interface | Purpose |
 |-----------|---------|
-| `IVpnServerConfig` | Server configuration (listen addr, keys, subnet, transport mode, forwarding mode, clients, proxy protocol, destination policy) |
+| `IVpnServerConfig` | Server configuration (listen addr, keys, subnet, transport mode, forwarding mode incl. bridge/hybrid, clients, proxy protocol, destination policy) |
 | `IVpnClientConfig` | Client configuration (server URL, keys, transport, forwarding mode, WG options, client-defined tags) |
-| `IClientEntry` | Server-side client definition (ID, keys, security, priority, server/client tags, expiry) |
+| `IClientEntry` | Server-side client definition (ID, keys, security, priority, server/client tags, expiry, bridge/VLAN settings) |
-| `IClientSecurity` | Per-client ACLs and rate limits (SmartProxy-aligned naming) |
+| `IClientSecurity` | Per-client ACLs, rate limits, and destination policy override (SmartProxy-aligned naming) |
 | `IClientRateLimit` | Rate limiting config (bytesPerSec, burstBytes) |
 | `IClientConfigBundle` | Full config bundle returned by `createClient()` — includes SmartVPN config, WireGuard .conf, and secrets |
 | `IVpnClientInfo` | Connected client info (IP, stats, authenticated key, remote addr, transport type) |
+| `IVpnServerStatistics` | Server stats with per-transport breakdowns (activeClientsWebsocket/Quic/Wireguard, totalConnections*) |
 | `IVpnConnectionQuality` | RTT, jitter, loss ratio, link health |
 | `IVpnMtuInfo` | TUN MTU, effective MTU, overhead bytes, oversized packet stats |
 | `IVpnKeypair` | Base64-encoded public/private key pair |
@@ -443,7 +555,7 @@ server.on('reconnected', () => { /* socket transport reconnected */ });
 | `exportClientConfig` | Re-export as SmartVPN config or WireGuard `.conf` |
 | `listClients` / `disconnectClient` | Manage live connections |
 | `setClientRateLimit` / `removeClientRateLimit` | Runtime rate limit adjustments |
-| `getStatus` / `getStatistics` / `getClientTelemetry` | Monitoring |
+| `getStatus` / `getStatistics` / `getClientTelemetry` | Monitoring (stats include per-transport breakdowns) |
 | `generateKeypair` / `generateWgKeypair` / `generateClientKeypair` | Key generation |
 | `addWgPeer` / `removeWgPeer` / `listWgPeers` | WireGuard peer management |
 
@@ -541,6 +653,7 @@ smartvpn/
 │   ├── index.ts                # All exports
 │   ├── smartvpn.interfaces.ts  # Interfaces, types, IPC command maps
 │   ├── smartvpn.plugins.ts     # Dependency imports
+│   ├── smartvpn.paths.ts       # Binary path resolution
 │   ├── smartvpn.classes.vpnserver.ts
 │   ├── smartvpn.classes.vpnclient.ts
 │   ├── smartvpn.classes.vpnbridge.ts
@@ -558,13 +671,19 @@ smartvpn/
 │       ├── proxy_protocol.rs   # PROXY protocol v2 parser
 │       ├── management.rs       # JSON-lines IPC
 │       ├── transport.rs        # WebSocket transport
+│       ├── transport_trait.rs  # Transport abstraction (Sink/Stream)
 │       ├── quic_transport.rs   # QUIC transport
 │       ├── wireguard.rs        # WireGuard (boringtun)
 │       ├── codec.rs            # Binary frame protocol
 │       ├── keepalive.rs        # Adaptive keepalives
 │       ├── ratelimit.rs        # Token bucket
 │       ├── userspace_nat.rs    # Userspace TCP/UDP NAT proxy
-│       └── ...                 # tunnel, network, telemetry, qos, mtu, reconnect
+│       ├── tunnel.rs           # TUN device management
+│       ├── network.rs          # IP pool + networking
+│       ├── telemetry.rs        # RTT/jitter/loss tracking
+│       ├── qos.rs              # Priority queues + smart dropping
+│       ├── mtu.rs              # MTU + ICMP too-big
+│       └── reconnect.rs        # Exponential backoff + session tokens
 ├── test/                       # Test files
 ├── dist_ts/                    # Compiled TypeScript
 └── dist_rust/                  # Cross-compiled binaries (linux amd64 + arm64)
@@ -572,7 +691,7 @@ smartvpn/
 
 ## License and Legal Information
 
-This repository contains open-source code licensed under the MIT License. A copy of the license can be found in the [license](./license.md) file.
+This repository contains open-source code licensed under the MIT License. A copy of the license can be found in the [LICENSE](./LICENSE) file.
 
 **Please note:** The MIT License does not grant permission to use the trade names, trademarks, service marks, or product names of the project, except as required for reasonable and customary use in describing the origin of the work and reproducing the content of the NOTICE file.
 
@@ -584,7 +703,7 @@ Use of these trademarks must comply with Task Venture Capital GmbH's Trademark G
 
 ### Company Information
 
-Task Venture Capital GmbH
+Task Venture Capital GmbH  
 Registered at District Court Bremen HRB 35230 HB, Germany
 
 For any legal inquiries or further information, please contact us via email at hello@task.vc.

rust/src/acl.rs

@@ -164,6 +164,7 @@ mod tests {
             destination_block_list: dst_block.map(|v| v.into_iter().map(String::from).collect()),
             max_connections: None,
             rate_limit: None,
+            destination_policy: None,
         }
     }
 

rust/src/bridge.rs

@@ -0,0 +1,396 @@
+//! L2 Bridge forwarding engine.
+//!
+//! Provides server-side bridging: receives L3 IP packets from VPN clients,
+//! wraps them in Ethernet frames, and injects them into a Linux bridge
+//! connected to the host's physical network interface.
+//!
+//! Return traffic from the bridge is stripped of its Ethernet header and
+//! routed back to VPN clients via `tun_routes`.
+
+use anyhow::Result;
+use std::collections::HashMap;
+use std::net::Ipv4Addr;
+use std::sync::Arc;
+use tokio::io::{AsyncReadExt, AsyncWriteExt};
+use tokio::sync::mpsc;
+use tracing::{debug, info, warn};
+
+use crate::server::ServerState;
+
+/// Configuration for the bridge forwarding engine.
+pub struct BridgeConfig {
+    /// TAP device name (e.g., "svpn_tap0")
+    pub tap_name: String,
+    /// Linux bridge name (e.g., "svpn_br0")
+    pub bridge_name: String,
+    /// Physical interface to bridge (e.g., "eth0")
+    pub physical_interface: String,
+    /// Gateway IP on the bridge (host's LAN IP)
+    pub gateway_ip: Ipv4Addr,
+    /// Subnet prefix length (e.g., 24)
+    pub prefix_len: u8,
+    /// MTU for the TAP device
+    pub mtu: u16,
+}
+
+/// Ethernet frame constants
+const ETH_HEADER_LEN: usize = 14;
+const ETH_TYPE_IPV4: [u8; 2] = [0x08, 0x00];
+const ETH_TYPE_ARP: [u8; 2] = [0x08, 0x06];
+const BROADCAST_MAC: [u8; 6] = [0xff; 6];
+
+/// Generate a deterministic locally-administered MAC from an IPv4 address.
+/// Uses prefix 02:53:56 (locally administered, "SVP" in hex-ish).
+fn mac_from_ip(ip: Ipv4Addr) -> [u8; 6] {
+    let octets = ip.octets();
+    [0x02, 0x53, 0x56, octets[1], octets[2], octets[3]]
+}
+
+/// Wrap an IP packet in an Ethernet frame.
+fn wrap_in_ethernet(ip_packet: &[u8], src_mac: [u8; 6], dst_mac: [u8; 6]) -> Vec<u8> {
+    let mut frame = Vec::with_capacity(ETH_HEADER_LEN + ip_packet.len());
+    frame.extend_from_slice(&dst_mac);
+    frame.extend_from_slice(&src_mac);
+    frame.extend_from_slice(&ETH_TYPE_IPV4);
+    frame.extend_from_slice(ip_packet);
+    frame
+}
+
+/// Extract the EtherType and payload from an Ethernet frame.
+fn unwrap_ethernet(frame: &[u8]) -> Option<([u8; 2], &[u8])> {
+    if frame.len() < ETH_HEADER_LEN {
+        return None;
+    }
+    let ether_type = [frame[12], frame[13]];
+    Some((ether_type, &frame[ETH_HEADER_LEN..]))
+}
+
+/// Extract destination IPv4 from a raw IP packet header.
+fn dst_ip_from_packet(packet: &[u8]) -> Option<Ipv4Addr> {
+    if packet.len() < 20 {
+        return None;
+    }
+    // Version must be 4
+    if (packet[0] >> 4) != 4 {
+        return None;
+    }
+    Some(Ipv4Addr::new(packet[16], packet[17], packet[18], packet[19]))
+}
+
+/// Extract source IPv4 from a raw IP packet header.
+fn src_ip_from_packet(packet: &[u8]) -> Option<Ipv4Addr> {
+    if packet.len() < 20 {
+        return None;
+    }
+    if (packet[0] >> 4) != 4 {
+        return None;
+    }
+    Some(Ipv4Addr::new(packet[12], packet[13], packet[14], packet[15]))
+}
+
+/// Build a gratuitous ARP announcement frame.
+fn build_garp(ip: Ipv4Addr, mac: [u8; 6]) -> Vec<u8> {
+    let ip_bytes = ip.octets();
+    let mut frame = Vec::with_capacity(42); // 14 eth + 28 ARP
+    // Ethernet header
+    frame.extend_from_slice(&BROADCAST_MAC); // dst: broadcast
+    frame.extend_from_slice(&mac);           // src: our MAC
+    frame.extend_from_slice(&ETH_TYPE_ARP);  // EtherType: ARP
+    // ARP payload
+    frame.extend_from_slice(&[0x00, 0x01]); // Hardware type: Ethernet
+    frame.extend_from_slice(&[0x08, 0x00]); // Protocol type: IPv4
+    frame.push(6);                          // Hardware addr len
+    frame.push(4);                          // Protocol addr len
+    frame.extend_from_slice(&[0x00, 0x01]); // Operation: ARP Request (GARP uses request)
+    frame.extend_from_slice(&mac);          // Sender hardware addr
+    frame.extend_from_slice(&ip_bytes);     // Sender protocol addr
+    frame.extend_from_slice(&[0x00; 6]);    // Target hardware addr (ignored in GARP)
+    frame.extend_from_slice(&ip_bytes);     // Target protocol addr (same as sender for GARP)
+    frame
+}
+
+// ============================================================================
+// Linux bridge management (ip commands)
+// ============================================================================
+
+async fn run_ip_cmd(args: &[&str]) -> Result<String> {
+    let output = tokio::process::Command::new("ip")
+        .args(args)
+        .output()
+        .await?;
+    if !output.status.success() {
+        let stderr = String::from_utf8_lossy(&output.stderr);
+        anyhow::bail!("ip {} failed: {}", args.join(" "), stderr.trim());
+    }
+    Ok(String::from_utf8_lossy(&output.stdout).to_string())
+}
+
+/// Create a Linux bridge interface.
+pub async fn create_bridge(name: &str) -> Result<()> {
+    run_ip_cmd(&["link", "add", name, "type", "bridge"]).await?;
+    info!("Created bridge {}", name);
+    Ok(())
+}
+
+/// Add an interface to a bridge.
+pub async fn bridge_add_interface(bridge: &str, iface: &str) -> Result<()> {
+    run_ip_cmd(&["link", "set", iface, "master", bridge]).await?;
+    info!("Added {} to bridge {}", iface, bridge);
+    Ok(())
+}
+
+/// Bring an interface up.
+pub async fn set_interface_up(iface: &str) -> Result<()> {
+    run_ip_cmd(&["link", "set", iface, "up"]).await?;
+    Ok(())
+}
+
+/// Remove a bridge interface.
+pub async fn remove_bridge(name: &str) -> Result<()> {
+    // First bring it down, ignore errors
+    let _ = run_ip_cmd(&["link", "set", name, "down"]).await;
+    run_ip_cmd(&["link", "del", name]).await?;
+    info!("Removed bridge {}", name);
+    Ok(())
+}
+
+/// Detect the default network interface from the routing table.
+pub async fn detect_default_interface() -> Result<String> {
+    let output = run_ip_cmd(&["route", "show", "default"]).await?;
+    // Format: "default via X.X.X.X dev IFACE ..."
+    let parts: Vec<&str> = output.split_whitespace().collect();
+    if let Some(idx) = parts.iter().position(|&s| s == "dev") {
+        if let Some(iface) = parts.get(idx + 1) {
+            return Ok(iface.to_string());
+        }
+    }
+    anyhow::bail!("Could not detect default network interface from route table");
+}
+
+/// Get the IP address and prefix length of a network interface.
+pub async fn get_interface_ip(iface: &str) -> Result<(Ipv4Addr, u8)> {
+    let output = run_ip_cmd(&["-4", "addr", "show", "dev", iface]).await?;
+    // Parse "inet X.X.X.X/NN" from output
+    for line in output.lines() {
+        let trimmed = line.trim();
+        if let Some(rest) = trimmed.strip_prefix("inet ") {
+            let addr_cidr = rest.split_whitespace().next().unwrap_or("");
+            let parts: Vec<&str> = addr_cidr.split('/').collect();
+            if parts.len() == 2 {
+                let ip: Ipv4Addr = parts[0].parse()?;
+                let prefix: u8 = parts[1].parse()?;
+                return Ok((ip, prefix));
+            }
+        }
+    }
+    anyhow::bail!("Could not find IPv4 address on interface {}", iface);
+}
+
+/// Migrate the host's IP from a physical interface to a bridge.
+/// This is the most delicate operation — briefly interrupts connectivity.
+pub async fn migrate_host_ip_to_bridge(
+    physical_iface: &str,
+    bridge: &str,
+    ip: Ipv4Addr,
+    prefix: u8,
+) -> Result<()> {
+    let cidr = format!("{}/{}", ip, prefix);
+    // Remove IP from physical interface
+    let _ = run_ip_cmd(&["addr", "del", &cidr, "dev", physical_iface]).await;
+    // Add IP to bridge
+    run_ip_cmd(&["addr", "add", &cidr, "dev", bridge]).await?;
+    info!("Migrated IP {} from {} to {}", cidr, physical_iface, bridge);
+    Ok(())
+}
+
+/// Restore the host's IP from bridge back to the physical interface.
+pub async fn restore_host_ip(
+    physical_iface: &str,
+    bridge: &str,
+    ip: Ipv4Addr,
+    prefix: u8,
+) -> Result<()> {
+    let cidr = format!("{}/{}", ip, prefix);
+    let _ = run_ip_cmd(&["addr", "del", &cidr, "dev", bridge]).await;
+    run_ip_cmd(&["addr", "add", &cidr, "dev", physical_iface]).await?;
+    info!("Restored IP {} to {}", cidr, physical_iface);
+    Ok(())
+}
+
+/// Enable proxy ARP on an interface via sysctl.
+pub async fn enable_proxy_arp(iface: &str) -> Result<()> {
+    let path = format!("/proc/sys/net/ipv4/conf/{}/proxy_arp", iface);
+    tokio::fs::write(&path, "1").await?;
+    info!("Enabled proxy_arp on {}", iface);
+    Ok(())
+}
+
+// ============================================================================
+// VLAN support (802.1Q via Linux bridge VLAN filtering)
+// ============================================================================
+
+async fn run_bridge_cmd(args: &[&str]) -> Result<String> {
+    let output = tokio::process::Command::new("bridge")
+        .args(args)
+        .output()
+        .await?;
+    if !output.status.success() {
+        let stderr = String::from_utf8_lossy(&output.stderr);
+        anyhow::bail!("bridge {} failed: {}", args.join(" "), stderr.trim());
+    }
+    Ok(String::from_utf8_lossy(&output.stdout).to_string())
+}
+
+/// Enable VLAN filtering on a bridge.
+pub async fn enable_vlan_filtering(bridge: &str) -> Result<()> {
+    run_ip_cmd(&["link", "set", bridge, "type", "bridge", "vlan_filtering", "1"]).await?;
+    info!("Enabled VLAN filtering on bridge {}", bridge);
+    Ok(())
+}
+
+/// Add a VLAN ID to a bridge port (TAP or physical interface).
+/// `pvid` = set as port VLAN ID (untagged ingress), `untagged` = strip tag on egress.
+pub async fn add_vlan_to_port(port: &str, vlan_id: u16, pvid: bool, untagged: bool) -> Result<()> {
+    let mut args = vec!["vlan", "add", "dev", port, "vid"];
+    let vid_str = vlan_id.to_string();
+    args.push(&vid_str);
+    if pvid { args.push("pvid"); }
+    if untagged { args.push("untagged"); }
+    run_bridge_cmd(&args).await?;
+    info!("Added VLAN {} to port {} (pvid={}, untagged={})", vlan_id, port, pvid, untagged);
+    Ok(())
+}
+
+/// Remove a VLAN ID from a bridge port.
+pub async fn remove_vlan_from_port(port: &str, vlan_id: u16) -> Result<()> {
+    let vid_str = vlan_id.to_string();
+    run_bridge_cmd(&["vlan", "del", "dev", port, "vid", &vid_str]).await?;
+    info!("Removed VLAN {} from port {}", vlan_id, port);
+    Ok(())
+}
+
+/// Create a TAP device (L2) using the tun crate.
+pub fn create_tap(name: &str, mtu: u16) -> Result<tun::AsyncDevice> {
+    let mut config = tun::Configuration::default();
+    config
+        .tun_name(name)
+        .layer(tun::Layer::L2)
+        .mtu(mtu)
+        .up();
+
+    #[cfg(target_os = "linux")]
+    config.platform_config(|p| {
+        p.ensure_root_privileges(true);
+    });
+
+    let device = tun::create_as_async(&config)?;
+    info!("TAP device {} created (L2, mtu={})", name, mtu);
+    Ok(device)
+}
+
+// ============================================================================
+// BridgeEngine — main event loop
+// ============================================================================
+
+/// The BridgeEngine wraps/unwraps Ethernet frames and bridges VPN traffic
+/// to the host's physical LAN via a Linux bridge + TAP device.
+pub struct BridgeEngine {
+    state: Arc<ServerState>,
+    /// Learned MAC addresses for LAN peers (dst IP → MAC).
+    /// Populated from ARP replies and Ethernet frame src MACs.
+    arp_cache: HashMap<Ipv4Addr, [u8; 6]>,
+}
+
+impl BridgeEngine {
+    pub fn new(state: Arc<ServerState>) -> Self {
+        Self {
+            state,
+            arp_cache: HashMap::new(),
+        }
+    }
+
+    /// Run the bridge engine event loop.
+    /// Receives L3 IP packets from VPN clients, wraps in Ethernet, writes to TAP.
+    /// Reads Ethernet frames from TAP, strips header, routes back to VPN clients.
+    pub async fn run(
+        mut self,
+        mut tap_device: tun::AsyncDevice,
+        mut packet_rx: mpsc::Receiver<Vec<u8>>,
+        mut shutdown_rx: mpsc::Receiver<()>,
+    ) -> Result<()> {
+        let mut buf = vec![0u8; 2048];
+
+        info!("BridgeEngine started");
+
+        loop {
+            tokio::select! {
+                // Packet from VPN client → wrap in Ethernet → write to TAP
+                Some(ip_packet) = packet_rx.recv() => {
+                    if let Some(dst_ip) = dst_ip_from_packet(&ip_packet) {
+                        let src_ip = src_ip_from_packet(&ip_packet).unwrap_or(Ipv4Addr::UNSPECIFIED);
+                        let src_mac = mac_from_ip(src_ip);
+                        let dst_mac = self.arp_cache.get(&dst_ip)
+                            .copied()
+                            .unwrap_or(BROADCAST_MAC);
+                        let frame = wrap_in_ethernet(&ip_packet, src_mac, dst_mac);
+                        if let Err(e) = tap_device.write_all(&frame).await {
+                            warn!("TAP write error: {}", e);
+                        }
+                    }
+                }
+
+                // Frame from TAP (LAN) → strip Ethernet → route to VPN client
+                result = tap_device.read(&mut buf) => {
+                    match result {
+                        Ok(len) if len >= ETH_HEADER_LEN => {
+                            let frame = &buf[..len];
+
+                            // Learn src MAC from incoming frames
+                            if let Some((ether_type, payload)) = unwrap_ethernet(frame) {
+                                // Learn ARP cache from src MAC + src IP
+                                let src_mac: [u8; 6] = frame[6..12].try_into().unwrap_or([0; 6]);
+                                if ether_type == ETH_TYPE_IPV4 {
+                                    if let Some(src_ip) = src_ip_from_packet(payload) {
+                                        self.arp_cache.insert(src_ip, src_mac);
+                                    }
+                                }
+
+                                // Only forward IPv4 packets to VPN clients
+                                if ether_type == ETH_TYPE_IPV4 {
+                                    if let Some(dst_ip) = dst_ip_from_packet(payload) {
+                                        // Look up VPN client by dst IP in tun_routes
+                                        let routes = self.state.tun_routes.read().await;
+                                        if let Some(sender) = routes.get(&dst_ip) {
+                                            let _ = sender.try_send(payload.to_vec());
+                                        }
+                                    }
+                                }
+                            }
+                        }
+                        Ok(_) => {} // Frame too short, ignore
+                        Err(e) => {
+                            warn!("TAP read error: {}", e);
+                        }
+                    }
+                }
+
+                _ = shutdown_rx.recv() => {
+                    info!("BridgeEngine shutting down");
+                    break;
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    /// Send a gratuitous ARP for a VPN client IP.
+    pub async fn announce_client(tap: &mut tun::AsyncDevice, ip: Ipv4Addr) -> Result<()> {
+        let mac = mac_from_ip(ip);
+        let garp = build_garp(ip, mac);
+        tap.write_all(&garp).await?;
+        debug!("Sent GARP for {} (MAC {:02x}:{:02x}:{:02x}:{:02x}:{:02x}:{:02x})",
+            ip, mac[0], mac[1], mac[2], mac[3], mac[4], mac[5]);
+        Ok(())
+    }
+}

rust/src/client_registry.rs

@@ -26,6 +26,9 @@ pub struct ClientSecurity {
     pub max_connections: Option<u32>,
     /// Per-client rate limiting.
     pub rate_limit: Option<ClientRateLimit>,
+    /// Per-client destination routing policy override.
+    /// When set, overrides the server-level DestinationPolicy for this client's traffic.
+    pub destination_policy: Option<crate::server::DestinationPolicyConfig>,
 }
 
 /// A registered client entry — the server-side source of truth.
@@ -57,6 +60,19 @@ pub struct ClientEntry {
     pub expires_at: Option<String>,
     /// Assigned VPN IP address.
     pub assigned_ip: Option<String>,
+
+    // Per-client bridge/host-IP settings
+
+    /// If true, client gets a host network IP via bridge mode.
+    pub use_host_ip: Option<bool>,
+    /// If true and use_host_ip is true, obtain IP via DHCP relay.
+    pub use_dhcp: Option<bool>,
+    /// Static LAN IP when use_host_ip is true and use_dhcp is false.
+    pub static_ip: Option<String>,
+    /// If true, assign this client to a specific 802.1Q VLAN.
+    pub force_vlan: Option<bool>,
+    /// 802.1Q VLAN ID (1-4094).
+    pub vlan_id: Option<u16>,
 }
 
 impl ClientEntry {
@@ -76,12 +92,14 @@ impl ClientEntry {
     }
 }
 
-/// In-memory client registry with dual-key indexing.
+/// In-memory client registry with triple-key indexing.
 pub struct ClientRegistry {
     /// Primary index: clientId → ClientEntry
     entries: HashMap<String, ClientEntry>,
     /// Secondary index: publicKey (base64) → clientId (fast lookup during handshake)
     key_index: HashMap<String, String>,
+    /// Tertiary index: assignedIp → clientId (fast lookup during NAT destination policy)
+    ip_index: HashMap<String, String>,
 }
 
 impl ClientRegistry {
@@ -89,6 +107,7 @@ impl ClientRegistry {
         Self {
             entries: HashMap::new(),
             key_index: HashMap::new(),
+            ip_index: HashMap::new(),
         }
     }
 
@@ -114,6 +133,9 @@ impl ClientRegistry {
             anyhow::bail!("Public key already registered to another client");
         }
         self.key_index.insert(entry.public_key.clone(), entry.client_id.clone());
+        if let Some(ref ip) = entry.assigned_ip {
+            self.ip_index.insert(ip.clone(), entry.client_id.clone());
+        }
         self.entries.insert(entry.client_id.clone(), entry);
         Ok(())
     }
@@ -123,6 +145,9 @@ impl ClientRegistry {
         let entry = self.entries.remove(client_id)
             .ok_or_else(|| anyhow::anyhow!("Client '{}' not found", client_id))?;
         self.key_index.remove(&entry.public_key);
+        if let Some(ref ip) = entry.assigned_ip {
+            self.ip_index.remove(ip);
+        }
         Ok(entry)
     }
 
@@ -137,6 +162,12 @@ impl ClientRegistry {
         self.entries.get(client_id)
     }
 
+    /// Get a client by assigned IP (used for per-client destination policy in NAT engine).
+    pub fn get_by_assigned_ip(&self, ip: &str) -> Option<&ClientEntry> {
+        let client_id = self.ip_index.get(ip)?;
+        self.entries.get(client_id)
+    }
+
     /// Check if a public key is authorized (exists, enabled, not expired).
     pub fn is_authorized(&self, public_key: &str) -> bool {
         match self.get_by_key(public_key) {
@@ -153,12 +184,22 @@ impl ClientRegistry {
         let entry = self.entries.get_mut(client_id)
             .ok_or_else(|| anyhow::anyhow!("Client '{}' not found", client_id))?;
         let old_key = entry.public_key.clone();
+        let old_ip = entry.assigned_ip.clone();
         updater(entry);
-        // If public key changed, update the index
+        // If public key changed, update the key index
         if entry.public_key != old_key {
             self.key_index.remove(&old_key);
             self.key_index.insert(entry.public_key.clone(), client_id.to_string());
         }
+        // If assigned IP changed, update the IP index
+        if entry.assigned_ip != old_ip {
+            if let Some(ref old) = old_ip {
+                self.ip_index.remove(old);
+            }
+            if let Some(ref new_ip) = entry.assigned_ip {
+                self.ip_index.insert(new_ip.clone(), client_id.to_string());
+            }
+        }
         Ok(())
     }
 
@@ -208,6 +249,11 @@ mod tests {
             description: None,
             expires_at: None,
             assigned_ip: None,
+            use_host_ip: None,
+            use_dhcp: None,
+            static_ip: None,
+            force_vlan: None,
+            vlan_id: None,
         }
     }
 
@@ -362,6 +408,7 @@ mod tests {
                 bytes_per_sec: 1_000_000,
                 burst_bytes: 2_000_000,
             }),
+            destination_policy: None,
         });
         let mut reg = ClientRegistry::new();
         reg.add(entry).unwrap();

rust/src/lib.rs

@@ -22,3 +22,4 @@ pub mod client_registry;
 pub mod acl;
 pub mod proxy_protocol;
 pub mod userspace_nat;
+pub mod bridge;

rust/src/network.rs

@@ -8,11 +8,16 @@ pub struct IpPool {
     /// Network address (e.g., 10.8.0.0)
     network: Ipv4Addr,
     /// Prefix length (e.g., 24)
+    #[allow(dead_code)]
     prefix_len: u8,
     /// Allocated IPs: IP -> client_id
     allocated: HashMap<Ipv4Addr, String>,
     /// Next candidate offset (skipping .0 network and .1 gateway)
     next_offset: u32,
+    /// Minimum allocation offset (inclusive). Default: 2 (skip .0 network and .1 gateway).
+    min_offset: u32,
+    /// Maximum allocation offset (exclusive). Default: broadcast offset.
+    max_offset: u32,
 }
 
 impl IpPool {
@@ -28,11 +33,47 @@ impl IpPool {
             anyhow::bail!("Prefix too long for VPN pool: /{}", prefix_len);
         }
 
+        let host_bits = 32 - prefix_len as u32;
+        let max_offset = (1u32 << host_bits) - 1; // broadcast offset
+
         Ok(Self {
             network,
             prefix_len,
             allocated: HashMap::new(),
             next_offset: 2, // Skip .0 (network) and .1 (server/gateway)
+            min_offset: 2,
+            max_offset,
+        })
+    }
+
+    /// Create a new IP pool with a restricted allocation range within the subnet.
+    /// `range_start` and `range_end` are host offsets (e.g., 200 and 250 for .200-.250).
+    pub fn new_with_range(subnet: &str, range_start: u32, range_end: u32) -> Result<Self> {
+        let parts: Vec<&str> = subnet.split('/').collect();
+        if parts.len() != 2 {
+            anyhow::bail!("Invalid subnet format: {}", subnet);
+        }
+        let network: Ipv4Addr = parts[0].parse()?;
+        let prefix_len: u8 = parts[1].parse()?;
+        if prefix_len > 30 {
+            anyhow::bail!("Prefix too long for VPN pool: /{}", prefix_len);
+        }
+        if range_start >= range_end {
+            anyhow::bail!("Invalid IP range: start ({}) must be less than end ({})", range_start, range_end);
+        }
+        let host_bits = 32 - prefix_len as u32;
+        let broadcast_offset = (1u32 << host_bits) - 1;
+        if range_end > broadcast_offset {
+            anyhow::bail!("IP range end ({}) exceeds subnet broadcast ({})", range_end, broadcast_offset);
+        }
+
+        Ok(Self {
+            network,
+            prefix_len,
+            allocated: HashMap::new(),
+            next_offset: range_start,
+            min_offset: range_start,
+            max_offset: range_end + 1, // exclusive
         })
     }
 
@@ -44,22 +85,17 @@ impl IpPool {
 
     /// Total number of usable client addresses in the pool.
     pub fn capacity(&self) -> u32 {
-        let host_bits = 32 - self.prefix_len as u32;
+        self.max_offset.saturating_sub(self.min_offset)
-        let total = 1u32 << host_bits;
-        total.saturating_sub(3) // minus network, gateway, broadcast
     }
 
     /// Allocate an IP for a client. Returns the assigned IP.
     pub fn allocate(&mut self, client_id: &str) -> Result<Ipv4Addr> {
-        let host_bits = 32 - self.prefix_len as u32;
-        let max_offset = (1u32 << host_bits) - 1; // broadcast offset
-
         // Try to find a free IP starting from next_offset
         let start = self.next_offset;
         let mut offset = start;
         loop {
-            if offset >= max_offset {
+            if offset >= self.max_offset {
-                offset = 2; // wrap around
+                offset = self.min_offset; // wrap around
             }
 
             let ip = Ipv4Addr::from(u32::from(self.network) + offset);

rust/src/server.rs

@@ -25,7 +25,7 @@ use crate::tunnel::{self, TunConfig};
 const DEAD_PEER_TIMEOUT: Duration = Duration::from_secs(180);
 
 /// Destination routing policy for VPN client traffic.
-#[derive(Debug, Clone, Deserialize)]
+#[derive(Debug, Clone, Deserialize, Serialize)]
 #[serde(rename_all = "camelCase")]
 pub struct DestinationPolicyConfig {
     /// Default action: "forceTarget", "block", or "allow".
@@ -92,6 +92,17 @@ pub struct ServerConfig {
     /// Defaults to ["0.0.0.0/0"] (full tunnel).
     #[serde(alias = "clientAllowedIPs")]
     pub client_allowed_ips: Option<Vec<String>>,
+
+    // Bridge mode configuration (forwarding_mode: "bridge")
+
+    /// LAN subnet CIDR for bridge mode (e.g. "192.168.1.0/24").
+    pub bridge_lan_subnet: Option<String>,
+    /// Physical network interface to bridge (e.g. "eth0"). Auto-detected if omitted.
+    pub bridge_physical_interface: Option<String>,
+    /// Start of VPN client IP range within the LAN subnet (host offset, e.g. 200).
+    pub bridge_ip_range_start: Option<u32>,
+    /// End of VPN client IP range within the LAN subnet (host offset, e.g. 250).
+    pub bridge_ip_range_end: Option<u32>,
 }
 
 /// Information about a connected client.
@@ -132,6 +143,14 @@ pub struct ServerStatistics {
     pub uptime_seconds: u64,
     pub active_clients: u64,
     pub total_connections: u64,
+    /// Per-transport active client counts.
+    pub active_clients_websocket: u64,
+    pub active_clients_quic: u64,
+    pub active_clients_wireguard: u64,
+    /// Per-transport total connection counts.
+    pub total_connections_websocket: u64,
+    pub total_connections_quic: u64,
+    pub total_connections_wireguard: u64,
 }
 
 /// The forwarding engine determines how decrypted IP packets are routed.
@@ -140,10 +159,28 @@ pub enum ForwardingEngine {
     Tun(tokio::io::WriteHalf<tun::AsyncDevice>),
     /// Userspace NAT — packets sent to smoltcp-based NAT engine via channel.
     Socket(mpsc::Sender<Vec<u8>>),
+    /// L2 Bridge — packets sent to BridgeEngine via channel, bridged to host LAN.
+    Bridge(mpsc::Sender<Vec<u8>>),
+    /// Hybrid — both socket NAT and bridge engines running simultaneously.
+    /// Per-client routing: look up src_ip in routing_table to decide socket vs bridge.
+    Hybrid {
+        socket_tx: mpsc::Sender<Vec<u8>>,
+        bridge_tx: mpsc::Sender<Vec<u8>>,
+        /// Fast lookup: VPN IP → true if client uses bridge (host IP), false for socket.
+        routing_table: Arc<RwLock<HashMap<Ipv4Addr, bool>>>,
+    },
     /// Testing/monitoring — packets are counted but not forwarded.
     Testing,
 }
 
+/// Info needed to tear down bridge infrastructure on stop().
+pub struct BridgeCleanupInfo {
+    pub physical_iface: String,
+    pub bridge_name: String,
+    pub host_ip: Ipv4Addr,
+    pub host_prefix: u8,
+}
+
 /// Shared server state.
 pub struct ServerState {
     pub config: ServerConfig,
@@ -160,6 +197,10 @@ pub struct ServerState {
     pub tun_routes: RwLock<HashMap<Ipv4Addr, mpsc::Sender<Vec<u8>>>>,
     /// Shutdown signal for the forwarding background task (TUN reader or NAT engine).
     pub tun_shutdown: mpsc::Sender<()>,
+    /// Shutdown signal for the bridge engine (bridge/hybrid modes only).
+    pub bridge_shutdown: Option<mpsc::Sender<()>>,
+    /// Bridge teardown info (bridge/hybrid modes only).
+    pub bridge_cleanup: Option<BridgeCleanupInfo>,
 }
 
 /// The VPN server.
@@ -183,7 +224,15 @@ impl VpnServer {
             anyhow::bail!("Server is already running");
         }
 
-        let ip_pool = IpPool::new(&config.subnet)?;
+        let mode = config.forwarding_mode.as_deref().unwrap_or("testing");
+        let ip_pool = if mode == "bridge" {
+            let lan_subnet = config.bridge_lan_subnet.as_deref().unwrap_or(&config.subnet);
+            let range_start = config.bridge_ip_range_start.unwrap_or(200);
+            let range_end = config.bridge_ip_range_end.unwrap_or(250);
+            IpPool::new_with_range(lan_subnet, range_start, range_end)?
+        } else {
+            IpPool::new(&config.subnet)?
+        };
 
         if config.enable_nat.unwrap_or(false) {
             if let Err(e) = crate::network::enable_ip_forwarding() {
@@ -197,7 +246,6 @@ impl VpnServer {
         }
 
         let link_mtu = config.mtu.unwrap_or(1420);
-        let mode = config.forwarding_mode.as_deref().unwrap_or("testing");
         let gateway_ip = ip_pool.gateway_addr();
 
         // Create forwarding engine based on mode
@@ -212,9 +260,28 @@ impl VpnServer {
                 packet_rx: mpsc::Receiver<Vec<u8>>,
                 shutdown_rx: mpsc::Receiver<()>,
             },
+            Bridge {
+                packet_tx: mpsc::Sender<Vec<u8>>,
+                packet_rx: mpsc::Receiver<Vec<u8>>,
+                tap_device: tun::AsyncDevice,
+                shutdown_rx: mpsc::Receiver<()>,
+            },
+            Hybrid {
+                socket_tx: mpsc::Sender<Vec<u8>>,
+                socket_rx: mpsc::Receiver<Vec<u8>>,
+                socket_shutdown_rx: mpsc::Receiver<()>,
+                bridge_tx: mpsc::Sender<Vec<u8>>,
+                bridge_rx: mpsc::Receiver<Vec<u8>>,
+                bridge_shutdown_rx: mpsc::Receiver<()>,
+                tap_device: tun::AsyncDevice,
+                routing_table: Arc<RwLock<HashMap<Ipv4Addr, bool>>>,
+            },
             Testing,
         }
 
+        let mut bridge_cleanup_info: Option<BridgeCleanupInfo> = None;
+        let mut bridge_shut_tx: Option<mpsc::Sender<()>> = None;
+
         let (setup, fwd_shutdown_tx) = match mode {
             "tun" => {
                 let tun_config = TunConfig {
@@ -235,6 +302,88 @@ impl VpnServer {
                 let (tx, rx) = mpsc::channel::<()>(1);
                 (ForwardingSetup::Socket { packet_tx, packet_rx, shutdown_rx: rx }, tx)
             }
+            "bridge" => {
+                info!("Starting L2 bridge forwarding (requires CAP_NET_ADMIN)");
+                let phys_iface = match &config.bridge_physical_interface {
+                    Some(i) => i.clone(),
+                    None => crate::bridge::detect_default_interface().await?,
+                };
+                let (host_ip, host_prefix) = crate::bridge::get_interface_ip(&phys_iface).await?;
+
+                let bridge_name = "svpn_br0";
+                let tap_name = "svpn_tap0";
+
+                // Create TAP + bridge infrastructure
+                let tap_device = crate::bridge::create_tap(tap_name, link_mtu)?;
+                crate::bridge::create_bridge(bridge_name).await?;
+                crate::bridge::set_interface_up(bridge_name).await?;
+                crate::bridge::bridge_add_interface(bridge_name, tap_name).await?;
+                crate::bridge::set_interface_up(tap_name).await?;
+                crate::bridge::bridge_add_interface(bridge_name, &phys_iface).await?;
+                crate::bridge::migrate_host_ip_to_bridge(&phys_iface, bridge_name, host_ip, host_prefix).await?;
+                crate::bridge::enable_proxy_arp(bridge_name).await?;
+
+                info!("Bridge {} created: TAP={}, physical={}, IP={}/{}", bridge_name, tap_name, phys_iface, host_ip, host_prefix);
+
+                bridge_cleanup_info = Some(BridgeCleanupInfo {
+                    physical_iface: phys_iface,
+                    bridge_name: bridge_name.to_string(),
+                    host_ip,
+                    host_prefix,
+                });
+
+                let (packet_tx, packet_rx) = mpsc::channel::<Vec<u8>>(4096);
+                let (tx, rx) = mpsc::channel::<()>(1);
+                (ForwardingSetup::Bridge { packet_tx, packet_rx, tap_device, shutdown_rx: rx }, tx)
+            }
+            "hybrid" => {
+                info!("Starting hybrid forwarding (socket + bridge, per-client routing)");
+
+                // Socket engine setup
+                let (s_tx, s_rx) = mpsc::channel::<Vec<u8>>(4096);
+                let (s_shut_tx, s_shut_rx) = mpsc::channel::<()>(1);
+
+                // Bridge engine setup
+                let phys_iface = match &config.bridge_physical_interface {
+                    Some(i) => i.clone(),
+                    None => crate::bridge::detect_default_interface().await?,
+                };
+                let (host_ip, host_prefix) = crate::bridge::get_interface_ip(&phys_iface).await?;
+                let bridge_name = "svpn_br0";
+                let tap_name = "svpn_tap0";
+
+                let tap_device = crate::bridge::create_tap(tap_name, link_mtu)?;
+                crate::bridge::create_bridge(bridge_name).await?;
+                crate::bridge::set_interface_up(bridge_name).await?;
+                crate::bridge::bridge_add_interface(bridge_name, tap_name).await?;
+                crate::bridge::set_interface_up(tap_name).await?;
+                crate::bridge::bridge_add_interface(bridge_name, &phys_iface).await?;
+                crate::bridge::migrate_host_ip_to_bridge(&phys_iface, bridge_name, host_ip, host_prefix).await?;
+                crate::bridge::enable_proxy_arp(bridge_name).await?;
+
+                let (b_tx, b_rx) = mpsc::channel::<Vec<u8>>(4096);
+                let (b_shut_tx, b_shut_rx) = mpsc::channel::<()>(1);
+
+                // Build routing table from registered clients
+                let routing_table = Arc::new(RwLock::new(HashMap::<Ipv4Addr, bool>::new()));
+
+                info!("Hybrid mode: socket + bridge (TAP={}, physical={}, IP={}/{})", tap_name, phys_iface, host_ip, host_prefix);
+
+                bridge_cleanup_info = Some(BridgeCleanupInfo {
+                    physical_iface: phys_iface,
+                    bridge_name: bridge_name.to_string(),
+                    host_ip,
+                    host_prefix,
+                });
+                bridge_shut_tx = Some(b_shut_tx);
+
+                // Socket engine uses fwd_shutdown_tx (stored in state.tun_shutdown)
+                (ForwardingSetup::Hybrid {
+                    socket_tx: s_tx, socket_rx: s_rx, socket_shutdown_rx: s_shut_rx,
+                    bridge_tx: b_tx, bridge_rx: b_rx, bridge_shutdown_rx: b_shut_rx,
+                    tap_device, routing_table,
+                }, s_shut_tx)
+            }
             _ => {
                 info!("Forwarding disabled (testing/monitoring mode)");
                 let (tx, _rx) = mpsc::channel::<()>(1);
@@ -244,7 +393,7 @@ impl VpnServer {
 
         // Compute effective MTU from overhead
         let overhead = TunnelOverhead::default_overhead();
-        let mtu_config = MtuConfig::new(overhead.effective_tun_mtu(1500).max(link_mtu));
+        let mtu_config = MtuConfig::new(overhead.effective_tun_mtu(1500).min(link_mtu));
 
         // Build client registry from config
         let registry = ClientRegistry::from_entries(
@@ -264,6 +413,8 @@ impl VpnServer {
             forwarding_engine: Mutex::new(ForwardingEngine::Testing),
             tun_routes: RwLock::new(HashMap::new()),
             tun_shutdown: fwd_shutdown_tx,
+            bridge_shutdown: bridge_shut_tx,
+            bridge_cleanup: bridge_cleanup_info,
         });
 
         // Spawn the forwarding background task and set the engine
@@ -293,6 +444,60 @@ impl VpnServer {
                     }
                 });
             }
+            ForwardingSetup::Bridge { packet_tx, packet_rx, tap_device, shutdown_rx } => {
+                *state.forwarding_engine.lock().await = ForwardingEngine::Bridge(packet_tx);
+                let bridge_engine = crate::bridge::BridgeEngine::new(state.clone());
+                tokio::spawn(async move {
+                    if let Err(e) = bridge_engine.run(tap_device, packet_rx, shutdown_rx).await {
+                        error!("Bridge engine error: {}", e);
+                    }
+                });
+            }
+            ForwardingSetup::Hybrid {
+                socket_tx, socket_rx, socket_shutdown_rx,
+                bridge_tx, bridge_rx, bridge_shutdown_rx,
+                tap_device, routing_table,
+            } => {
+                // Populate routing table from registered clients
+                {
+                    let registry = state.client_registry.read().await;
+                    let mut rt = routing_table.write().await;
+                    for entry in registry.list() {
+                        if let Some(ref ip_str) = entry.assigned_ip {
+                            if let Ok(ip) = ip_str.parse::<Ipv4Addr>() {
+                                rt.insert(ip, entry.use_host_ip.unwrap_or(false));
+                            }
+                        }
+                    }
+                }
+
+                // Start socket (NAT) engine
+                let proxy_protocol = config.socket_forward_proxy_protocol.unwrap_or(false);
+                let nat_engine = crate::userspace_nat::NatEngine::new(
+                    gateway_ip,
+                    link_mtu as usize,
+                    state.clone(),
+                    proxy_protocol,
+                    config.destination_policy.clone(),
+                );
+                tokio::spawn(async move {
+                    if let Err(e) = nat_engine.run(socket_rx, socket_shutdown_rx).await {
+                        error!("NAT engine error (hybrid): {}", e);
+                    }
+                });
+
+                // Start bridge engine
+                let bridge_engine = crate::bridge::BridgeEngine::new(state.clone());
+                tokio::spawn(async move {
+                    if let Err(e) = bridge_engine.run(tap_device, bridge_rx, bridge_shutdown_rx).await {
+                        error!("Bridge engine error (hybrid): {}", e);
+                    }
+                });
+
+                *state.forwarding_engine.lock().await = ForwardingEngine::Hybrid {
+                    socket_tx, bridge_tx, routing_table,
+                };
+            }
             ForwardingSetup::Testing => {}
         }
 
@@ -413,6 +618,43 @@ impl VpnServer {
                     let _ = state.tun_shutdown.send(()).await;
                     *state.forwarding_engine.lock().await = ForwardingEngine::Testing;
                 }
+                "bridge" => {
+                    let _ = state.tun_shutdown.send(()).await;
+                    *state.forwarding_engine.lock().await = ForwardingEngine::Testing;
+                    // Restore host networking: move IP back and remove bridge
+                    if let Some(ref cleanup) = state.bridge_cleanup {
+                        if let Err(e) = crate::bridge::restore_host_ip(
+                            &cleanup.physical_iface, &cleanup.bridge_name,
+                            cleanup.host_ip, cleanup.host_prefix,
+                        ).await {
+                            warn!("Failed to restore host IP: {}", e);
+                        }
+                        if let Err(e) = crate::bridge::remove_bridge(&cleanup.bridge_name).await {
+                            warn!("Failed to remove bridge: {}", e);
+                        }
+                    }
+                }
+                "hybrid" => {
+                    // Shut down socket (NAT) engine
+                    let _ = state.tun_shutdown.send(()).await;
+                    // Shut down bridge engine
+                    if let Some(ref bridge_shut) = state.bridge_shutdown {
+                        let _ = bridge_shut.send(()).await;
+                    }
+                    *state.forwarding_engine.lock().await = ForwardingEngine::Testing;
+                    // Restore host networking: move IP back and remove bridge
+                    if let Some(ref cleanup) = state.bridge_cleanup {
+                        if let Err(e) = crate::bridge::restore_host_ip(
+                            &cleanup.physical_iface, &cleanup.bridge_name,
+                            cleanup.host_ip, cleanup.host_prefix,
+                        ).await {
+                            warn!("Failed to restore host IP: {}", e);
+                        }
+                        if let Err(e) = crate::bridge::remove_bridge(&cleanup.bridge_name).await {
+                            warn!("Failed to remove bridge: {}", e);
+                        }
+                    }
+                }
                 _ => {}
             }
 
@@ -450,7 +692,21 @@ impl VpnServer {
         if let Some(ref state) = self.state {
             let mut stats = state.stats.read().await.clone();
             stats.uptime_seconds = state.started_at.elapsed().as_secs();
-            stats.active_clients = state.clients.read().await.len() as u64;
+            let clients = state.clients.read().await;
+            stats.active_clients = clients.len() as u64;
+            // Compute per-transport active counts
+            stats.active_clients_websocket = 0;
+            stats.active_clients_quic = 0;
+            stats.active_clients_wireguard = 0;
+            for info in clients.values() {
+                match info.transport_type.as_str() {
+                    "websocket" => stats.active_clients_websocket += 1,
+                    "quic" => stats.active_clients_quic += 1,
+                    "wireguard" => stats.active_clients_wireguard += 1,
+                    _ => {}
+                }
+            }
+            drop(clients);
             stats
         } else {
             ServerStatistics::default()
@@ -611,10 +867,18 @@ impl VpnServer {
             description: partial.get("description").and_then(|v| v.as_str()).map(String::from),
             expires_at: partial.get("expiresAt").and_then(|v| v.as_str()).map(String::from),
             assigned_ip: Some(assigned_ip.to_string()),
+            use_host_ip: partial.get("useHostIp").and_then(|v| v.as_bool()),
+            use_dhcp: partial.get("useDhcp").and_then(|v| v.as_bool()),
+            static_ip: partial.get("staticIp").and_then(|v| v.as_str()).map(String::from),
+            force_vlan: partial.get("forceVlan").and_then(|v| v.as_bool()),
+            vlan_id: partial.get("vlanId").and_then(|v| v.as_u64()).map(|v| v as u16),
         };
 
-        // Add to registry
+        // Add to registry — release IP on failure to avoid pool leak
-        state.client_registry.write().await.add(entry.clone())?;
+        if let Err(e) = state.client_registry.write().await.add(entry.clone()) {
+            state.ip_pool.lock().await.release(&assigned_ip);
+            return Err(e);
+        }
 
         // Register WG peer with the running WG listener (if active)
         if self.wg_command_tx.is_some() {
@@ -1303,6 +1567,11 @@ async fn handle_client_connection(
     {
         let mut stats = state.stats.write().await;
         stats.total_connections += 1;
+        match transport_type {
+            "websocket" => stats.total_connections_websocket += 1,
+            "quic" => stats.total_connections_quic += 1,
+            _ => {}
+        }
     }
 
     // Send assigned IP info (encrypted), include effective MTU
@@ -1403,6 +1672,20 @@ async fn handle_client_connection(
                                                     ForwardingEngine::Socket(sender) => {
                                                         let _ = sender.try_send(buf[..len].to_vec());
                                                     }
+                                                    ForwardingEngine::Bridge(sender) => {
+                                                        let _ = sender.try_send(buf[..len].to_vec());
+                                                    }
+                                                    ForwardingEngine::Hybrid { socket_tx, bridge_tx, routing_table } => {
+                                                        if len >= 20 {
+                                                            let src_ip = Ipv4Addr::new(buf[12], buf[13], buf[14], buf[15]);
+                                                            let use_bridge = routing_table.read().await.get(&src_ip).copied().unwrap_or(false);
+                                                            if use_bridge {
+                                                                let _ = bridge_tx.try_send(buf[..len].to_vec());
+                                                            } else {
+                                                                let _ = socket_tx.try_send(buf[..len].to_vec());
+                                                            }
+                                                        }
+                                                    }
                                                     ForwardingEngine::Testing => {}
                                                 }
                                             }

rust/src/userspace_nat.rs

@@ -17,6 +17,10 @@ use crate::acl;
 use crate::server::{DestinationPolicyConfig, ServerState};
 use crate::tunnel;
 
+/// Maximum size of per-session pending send buffer (512KB = 8x socket buffer).
+/// Sessions exceeding this are aborted — the client cannot keep up.
+const TCP_PENDING_SEND_MAX: usize = 512 * 1024;
+
 // ============================================================================
 // Virtual IP device for smoltcp
 // ============================================================================
@@ -121,7 +125,8 @@ struct SessionKey {
 
 struct TcpSession {
     smoltcp_handle: SocketHandle,
-    bridge_data_tx: mpsc::Sender<Vec<u8>>,
+    /// Channel to send data to the bridge task. None until bridge starts.
+    bridge_data_tx: Option<mpsc::Sender<Vec<u8>>>,
     #[allow(dead_code)]
     client_ip: Ipv4Addr,
     /// Bridge task has been spawned (deferred until handshake completes)
@@ -132,6 +137,8 @@ struct TcpSession {
     pending_send: Vec<u8>,
     /// Session is closing (FIN in progress), don't accept new SYNs
     closing: bool,
+    /// Last time data flowed through this session (for idle timeout)
+    last_activity: tokio::time::Instant,
 }
 
 struct UdpSession {
@@ -260,8 +267,19 @@ impl NatEngine {
     }
 
     /// Evaluate destination policy for a packet's destination IP.
-    fn evaluate_destination(&self, dst_ip: Ipv4Addr, dst_port: u16) -> DestinationAction {
+    /// Checks per-client policy first (via src_ip → client registry lookup),
-        let policy = match &self.destination_policy {
+    /// falls back to server-wide policy.
+    fn evaluate_destination(&self, src_ip: Ipv4Addr, dst_ip: Ipv4Addr, dst_port: u16) -> DestinationAction {
+        // Try per-client destination policy (lookup by tunnel IP)
+        let client_policy = if let Ok(registry) = self.state.client_registry.try_read() {
+            registry.get_by_assigned_ip(&src_ip.to_string())
+                .and_then(|e| e.security.as_ref())
+                .and_then(|s| s.destination_policy.clone())
+        } else {
+            None
+        };
+
+        let policy = match client_policy.as_ref().or(self.destination_policy.as_ref()) {
             Some(p) => p,
             None => return DestinationAction::PassThrough(SocketAddr::new(dst_ip.into(), dst_port)),
         };
@@ -319,7 +337,7 @@ impl NatEngine {
                 // Skip if session exists (including closing sessions — let FIN complete)
                 let session_exists = self.tcp_sessions.contains_key(&key);
                 if is_syn && !session_exists {
-                    match self.evaluate_destination(dst_ip, dst_port) {
+                    match self.evaluate_destination(src_ip, dst_ip, dst_port) {
                         DestinationAction::Drop => {
                             debug!("NAT: destination policy blocked TCP {}:{} -> {}:{}", src_ip, src_port, dst_ip, dst_port);
                             return;
@@ -343,7 +361,7 @@ impl NatEngine {
                 };
 
                 if !self.udp_sessions.contains_key(&key) {
-                    match self.evaluate_destination(dst_ip, dst_port) {
+                    match self.evaluate_destination(src_ip, dst_ip, dst_port) {
                         DestinationAction::Drop => {
                             debug!("NAT: destination policy blocked UDP {}:{} -> {}:{}", src_ip, src_port, dst_ip, dst_port);
                             return;
@@ -385,17 +403,15 @@ impl NatEngine {
 
         let handle = self.sockets.add(socket);
 
-        // Channel for sending data from NAT engine to bridge task
-        let (data_tx, _data_rx) = mpsc::channel::<Vec<u8>>(256);
-
         let session = TcpSession {
             smoltcp_handle: handle,
-            bridge_data_tx: data_tx,
+            bridge_data_tx: None,
             client_ip: key.src_ip,
             bridge_started: false,
             connect_addr,
             pending_send: Vec::new(),
             closing: false,
+            last_activity: tokio::time::Instant::now(),
         };
         self.tcp_sessions.insert(key.clone(), session);
 
@@ -470,9 +486,8 @@ impl NatEngine {
                 let socket = self.sockets.get_mut::<tcp::Socket>(session.smoltcp_handle);
                 if socket.is_active() {
                     session.bridge_started = true;
-                    // Recreate the data channel — the old receiver was dropped
                     let (data_tx, data_rx) = mpsc::channel::<Vec<u8>>(256);
-                    session.bridge_data_tx = data_tx;
+                    session.bridge_data_tx = Some(data_tx);
                     let btx = bridge_tx_clone.clone();
                     let k = key.clone();
                     let addr = session.connect_addr;
@@ -503,15 +518,29 @@ impl NatEngine {
 
         // Bridge: read data from smoltcp TCP sockets → send to bridge tasks
         let mut closed_tcp: Vec<SessionKey> = Vec::new();
-        let mut tcp_outbound: Vec<(mpsc::Sender<Vec<u8>>, Vec<u8>)> = Vec::new();
+        let mut active_tcp: Vec<SessionKey> = Vec::new();
         for (key, session) in &self.tcp_sessions {
             let socket = self.sockets.get_mut::<tcp::Socket>(session.smoltcp_handle);
             if session.bridge_started && socket.can_recv() {
-                let sender = session.bridge_data_tx.clone();
+                if let Some(ref sender) = session.bridge_data_tx {
-                let _ = socket.recv(|data| {
+                    // Reserve channel slot BEFORE consuming from smoltcp.
-                    tcp_outbound.push((sender.clone(), data.to_vec()));
+                    // If the channel is full, we don't consume — smoltcp's RX buffer
-                    (data.len(), ())
+                    // fills up, it stops advertising TCP window space, and the VPN
-                });
+                    // client's TCP stack backs off. Proper end-to-end backpressure.
+                    match sender.try_reserve() {
+                        Ok(permit) => {
+                            let _ = socket.recv(|data| {
+                                permit.send(data.to_vec());
+                                (data.len(), ())
+                            });
+                            active_tcp.push(key.clone());
+                        }
+                        Err(_) => {
+                            debug!("NAT: bridge channel full for {}:{} -> {}:{}, applying backpressure",
+                                key.src_ip, key.src_port, key.dst_ip, key.dst_port);
+                        }
+                    }
+                }
             }
             // Detect closed connections
             if !socket.is_open() && !socket.is_listening() {
@@ -519,10 +548,11 @@ impl NatEngine {
             }
         }
 
-        // Send TCP data to bridge tasks (outside borrow of self.tcp_sessions)
+        // Update last_activity for sessions that had data flow
-        for (sender, data) in tcp_outbound {
+        let now = tokio::time::Instant::now();
-            if sender.try_send(data).is_err() {
+        for key in active_tcp {
-                debug!("NAT: bridge channel full, TCP data dropped");
+            if let Some(session) = self.tcp_sessions.get_mut(&key) {
+                session.last_activity = now;
             }
         }
 
@@ -561,39 +591,31 @@ impl NatEngine {
         match msg {
             BridgeMessage::TcpData { key, data } => {
                 if let Some(session) = self.tcp_sessions.get_mut(&key) {
+                    session.last_activity = tokio::time::Instant::now();
+                    // Append to pending buffer, then flush as much as possible
+                    session.pending_send.extend_from_slice(&data);
                     let socket =
                         self.sockets.get_mut::<tcp::Socket>(session.smoltcp_handle);
-                    if socket.can_send() {
+                    if socket.can_send() && !session.pending_send.is_empty() {
-                        // Try to write directly first
+                        match socket.send_slice(&session.pending_send) {
-                        let all_data = if session.pending_send.is_empty() {
+                            Ok(written) if written > 0 => {
-                            &data
+                                session.pending_send.drain(..written);
-                        } else {
-                            session.pending_send.extend_from_slice(&data);
-                            &session.pending_send.clone()
-                        };
-                        match socket.send_slice(all_data) {
-                            Ok(written) if written < all_data.len() => {
-                                // Partial write — buffer the rest
-                                if session.pending_send.is_empty() {
-                                    session.pending_send = data[written..].to_vec();
-                                } else {
-                                    session.pending_send.drain(..written);
-                                }
-                            }
-                            Ok(_) => {
-                                // Full write — clear any pending data
-                                session.pending_send.clear();
-                            }
-                            Err(_) => {
-                                // Write failed — buffer everything
-                                if session.pending_send.is_empty() {
-                                    session.pending_send = data;
-                                }
                             }
+                            _ => {}
                         }
-                    } else {
+                    }
-                        // Can't send yet — buffer for later
+                    // Cap check — abort session if client can't keep up
-                        session.pending_send.extend_from_slice(&data);
+                    if session.pending_send.len() > TCP_PENDING_SEND_MAX {
+                        warn!(
+                            "NAT: TCP session {}:{} -> {}:{} pending buffer exceeded {}KB, aborting",
+                            key.src_ip, key.src_port, key.dst_ip, key.dst_port,
+                            TCP_PENDING_SEND_MAX / 1024
+                        );
+                        let socket =
+                            self.sockets.get_mut::<tcp::Socket>(session.smoltcp_handle);
+                        socket.abort();
+                        session.pending_send.clear();
+                        session.closing = true;
                     }
                 }
             }
@@ -644,6 +666,29 @@ impl NatEngine {
         }
     }
 
+    fn cleanup_idle_tcp_sessions(&mut self) {
+        let timeout = Duration::from_secs(300); // 5 minutes
+        let now = tokio::time::Instant::now();
+        let expired: Vec<SessionKey> = self
+            .tcp_sessions
+            .iter()
+            .filter(|(_, s)| now.duration_since(s.last_activity) > timeout)
+            .map(|(k, _)| k.clone())
+            .collect();
+
+        for key in expired {
+            if let Some(session) = self.tcp_sessions.remove(&key) {
+                let socket = self.sockets.get_mut::<tcp::Socket>(session.smoltcp_handle);
+                socket.abort();
+                self.sockets.remove(session.smoltcp_handle);
+                warn!(
+                    "NAT: TCP session timed out {}:{} -> {}:{}",
+                    key.src_ip, key.src_port, key.dst_ip, key.dst_port
+                );
+            }
+        }
+    }
+
     /// Main async event loop for the NAT engine.
     pub async fn run(
         mut self,
@@ -651,9 +696,13 @@ impl NatEngine {
         mut shutdown_rx: mpsc::Receiver<()>,
     ) -> Result<()> {
         info!("Userspace NAT engine started");
-        let mut timer = tokio::time::interval(Duration::from_millis(50));
+        let default_poll_delay = Duration::from_millis(50);
         let mut cleanup_timer = tokio::time::interval(Duration::from_secs(10));
 
+        // Dynamic poll timer — reset after each event using smoltcp's poll_delay()
+        let poll_sleep = tokio::time::sleep(default_poll_delay);
+        tokio::pin!(poll_sleep);
+
         loop {
             tokio::select! {
                 Some(packet) = packet_rx.recv() => {
@@ -664,18 +713,26 @@ impl NatEngine {
                     self.handle_bridge_message(msg);
                     self.process().await;
                 }
-                _ = timer.tick() => {
+                () = &mut poll_sleep => {
                     // Periodic poll for smoltcp maintenance (TCP retransmit, etc.)
                     self.process().await;
                 }
                 _ = cleanup_timer.tick() => {
                     self.cleanup_idle_udp_sessions();
+                    self.cleanup_idle_tcp_sessions();
                 }
                 _ = shutdown_rx.recv() => {
                     info!("Userspace NAT engine shutting down");
                     break;
                 }
             }
+
+            // Reset poll delay based on smoltcp's actual timer needs
+            let now = self.smoltcp_now();
+            let delay = self.iface.poll_delay(now, &self.sockets)
+                .map(|d| Duration::from_millis(d.total_millis()))
+                .unwrap_or(default_poll_delay);
+            poll_sleep.as_mut().reset(tokio::time::Instant::now() + delay);
         }
 
         Ok(())

rust/src/wireguard.rs

@@ -220,6 +220,15 @@ struct PeerState {
     #[allow(dead_code)]
     persistent_keepalive: Option<u16>,
     stats: WgPeerStats,
+    /// Whether this peer has completed a WireGuard handshake and is in state.clients.
+    is_connected: bool,
+    /// Last time we received data or handshake activity from this peer.
+    last_activity_at: Option<tokio::time::Instant>,
+    /// VPN IP assigned during registration (used for connect/disconnect).
+    vpn_ip: Option<Ipv4Addr>,
+    /// Previous synced byte counts for aggregate stats delta tracking.
+    prev_synced_bytes_sent: u64,
+    prev_synced_bytes_received: u64,
 }
 
 impl PeerState {
@@ -276,6 +285,11 @@ fn add_peer_to_loop(
         endpoint,
         persistent_keepalive: config.persistent_keepalive,
         stats: WgPeerStats::default(),
+        is_connected: false,
+        last_activity_at: None,
+        vpn_ip: None,
+        prev_synced_bytes_sent: 0,
+        prev_synced_bytes_received: 0,
     });
 
     info!("Added WireGuard peer: {}", config.public_key);
@@ -305,10 +319,12 @@ fn extract_peer_vpn_ip(allowed_ips: &[AllowedIp]) -> Option<Ipv4Addr> {
             }
         }
     }
-    // Fallback: use the first IPv4 address from any prefix length
+    // Fallback: use the first non-unspecified IPv4 address from any prefix length
     for aip in allowed_ips {
         if let IpAddr::V4(v4) = aip.addr {
-            return Some(v4);
+            if !v4.is_unspecified() {
+                return Some(v4);
+            }
         }
     }
     None
@@ -323,8 +339,9 @@ fn wg_timestamp_now() -> String {
     format!("{}", duration.as_secs())
 }
 
-/// Register a WG peer in ServerState (tun_routes, clients, ip_pool).
+/// Register a WG peer in ServerState (tun_routes + ip_pool only).
-/// Returns the VPN IP and the per-peer return-packet receiver.
+/// Does NOT add to state.clients — peers appear there only after handshake.
+/// Returns the VPN IP.
 async fn register_wg_peer(
     state: &Arc<ServerState>,
     peer: &PeerState,
@@ -366,13 +383,23 @@ async fn register_wg_peer(
         });
     }
 
-    // Insert ClientInfo
+    info!("WG peer {} registered with IP {} (not yet connected)", peer.public_key_b64, vpn_ip);
+    Ok(Some(vpn_ip))
+}
+
+/// Add a WG peer to state.clients on first successful handshake (data received).
+async fn connect_wg_peer(
+    state: &Arc<ServerState>,
+    peer: &PeerState,
+    vpn_ip: Ipv4Addr,
+) {
+    let client_id = format!("wg-{}", &peer.public_key_b64[..8.min(peer.public_key_b64.len())]);
     let client_info = ClientInfo {
         client_id: client_id.clone(),
         assigned_ip: vpn_ip.to_string(),
         connected_since: wg_timestamp_now(),
-        bytes_sent: 0,
+        bytes_sent: peer.stats.bytes_sent,
-        bytes_received: 0,
+        bytes_received: peer.stats.bytes_received,
         packets_dropped: 0,
         bytes_dropped: 0,
         last_keepalive_at: None,
@@ -380,13 +407,31 @@ async fn register_wg_peer(
         rate_limit_bytes_per_sec: None,
         burst_bytes: None,
         authenticated_key: peer.public_key_b64.clone(),
-        registered_client_id: client_id,
+        registered_client_id: client_id.clone(),
         remote_addr: peer.endpoint.map(|e| e.to_string()),
         transport_type: "wireguard".to_string(),
     };
     state.clients.write().await.insert(client_info.client_id.clone(), client_info);
 
-    Ok(Some(vpn_ip))
+    // Increment total_connections
+    {
+        let mut stats = state.stats.write().await;
+        stats.total_connections += 1;
+        stats.total_connections_wireguard += 1;
+    }
+
+    info!("WG peer {} connected (IP: {})", peer.public_key_b64, vpn_ip);
+}
+
+/// Remove a WG peer from state.clients (disconnect without unregistering).
+async fn disconnect_wg_peer(
+    state: &Arc<ServerState>,
+    pubkey: &str,
+) {
+    let client_id = format!("wg-{}", &pubkey[..8.min(pubkey.len())]);
+    if state.clients.write().await.remove(&client_id).is_some() {
+        info!("WG peer {} disconnected (removed from active clients)", pubkey);
+    }
 }
 
 /// Unregister a WG peer from ServerState.
@@ -460,6 +505,11 @@ pub async fn run_wg_listener(
             endpoint,
             persistent_keepalive: peer_config.persistent_keepalive,
             stats: WgPeerStats::default(),
+            is_connected: false,
+            last_activity_at: None,
+            vpn_ip: None,
+            prev_synced_bytes_sent: 0,
+            prev_synced_bytes_received: 0,
         });
     }
 
@@ -470,11 +520,12 @@ pub async fn run_wg_listener(
     // Merged return-packet channel: all per-peer channels feed into this
     let (wg_return_tx, mut wg_return_rx) = mpsc::channel::<(String, Vec<u8>)>(1024);
 
-    // Register initial peers in ServerState and track their VPN IPs
+    // Register initial peers in ServerState (IP reservation + tun_routes only, NOT state.clients)
     let mut peer_vpn_ips: HashMap<String, Ipv4Addr> = HashMap::new();
-    for peer in &peers {
+    for peer in peers.iter_mut() {
         if let Ok(Some(ip)) = register_wg_peer(&state, peer, &wg_return_tx).await {
             peer_vpn_ips.insert(peer.public_key_b64.clone(), ip);
+            peer.vpn_ip = Some(ip);
         }
     }
 
@@ -483,6 +534,7 @@ pub async fn run_wg_listener(
     let mut dst_buf = vec![0u8; WG_BUFFER_SIZE];
     let mut timer = tokio::time::interval(std::time::Duration::from_millis(TIMER_TICK_MS));
     let mut stats_timer = tokio::time::interval(std::time::Duration::from_secs(1));
+    let mut idle_check_timer = tokio::time::interval(std::time::Duration::from_secs(10));
 
     loop {
         tokio::select! {
@@ -506,6 +558,8 @@ pub async fn run_wg_listener(
                                 }
                             }
                             peer.endpoint = Some(src_addr);
+                            // Handshake response counts as activity
+                            peer.last_activity_at = Some(tokio::time::Instant::now());
                             handled = true;
                             break;
                         }
@@ -524,12 +578,35 @@ pub async fn run_wg_listener(
                                     ForwardingEngine::Socket(sender) => {
                                         let _ = sender.try_send(packet.to_vec());
                                     }
+                                    ForwardingEngine::Bridge(sender) => {
+                                        let _ = sender.try_send(packet.to_vec());
+                                    }
+                                    ForwardingEngine::Hybrid { socket_tx, bridge_tx, routing_table } => {
+                                        if packet.len() >= 20 {
+                                            let src_ip = Ipv4Addr::new(packet[12], packet[13], packet[14], packet[15]);
+                                            let use_bridge = routing_table.read().await.get(&src_ip).copied().unwrap_or(false);
+                                            if use_bridge {
+                                                let _ = bridge_tx.try_send(packet.to_vec());
+                                            } else {
+                                                let _ = socket_tx.try_send(packet.to_vec());
+                                            }
+                                        }
+                                    }
                                     ForwardingEngine::Testing => {}
                                 }
                                 peer.stats.bytes_received += pkt_len;
                                 peer.stats.packets_received += 1;
                             }
                             peer.endpoint = Some(src_addr);
+                            // Track activity and detect handshake completion
+                            peer.last_activity_at = Some(tokio::time::Instant::now());
+                            if !peer.is_connected {
+                                peer.is_connected = true;
+                                peer.stats.last_handshake_time = Some(wg_timestamp_now());
+                                if let Some(vpn_ip) = peer.vpn_ip {
+                                    connect_wg_peer(&state, peer, vpn_ip).await;
+                                }
+                            }
                             handled = true;
                             break;
                         }
@@ -547,12 +624,35 @@ pub async fn run_wg_listener(
                                     ForwardingEngine::Socket(sender) => {
                                         let _ = sender.try_send(packet.to_vec());
                                     }
+                                    ForwardingEngine::Bridge(sender) => {
+                                        let _ = sender.try_send(packet.to_vec());
+                                    }
+                                    ForwardingEngine::Hybrid { socket_tx, bridge_tx, routing_table } => {
+                                        if packet.len() >= 20 {
+                                            let src_ip = Ipv4Addr::new(packet[12], packet[13], packet[14], packet[15]);
+                                            let use_bridge = routing_table.read().await.get(&src_ip).copied().unwrap_or(false);
+                                            if use_bridge {
+                                                let _ = bridge_tx.try_send(packet.to_vec());
+                                            } else {
+                                                let _ = socket_tx.try_send(packet.to_vec());
+                                            }
+                                        }
+                                    }
                                     ForwardingEngine::Testing => {}
                                 }
                                 peer.stats.bytes_received += pkt_len;
                                 peer.stats.packets_received += 1;
                             }
                             peer.endpoint = Some(src_addr);
+                            // Track activity and detect handshake completion
+                            peer.last_activity_at = Some(tokio::time::Instant::now());
+                            if !peer.is_connected {
+                                peer.is_connected = true;
+                                peer.stats.last_handshake_time = Some(wg_timestamp_now());
+                                if let Some(vpn_ip) = peer.vpn_ip {
+                                    connect_wg_peer(&state, peer, vpn_ip).await;
+                                }
+                            }
                             handled = true;
                             break;
                         }
@@ -603,6 +703,10 @@ pub async fn run_wg_listener(
                         }
                         TunnResult::Err(WireGuardError::ConnectionExpired) => {
                             warn!("WG peer {} connection expired", peer.public_key_b64);
+                            if peer.is_connected {
+                                peer.is_connected = false;
+                                disconnect_wg_peer(&state, &peer.public_key_b64).await;
+                            }
                         }
                         TunnResult::Err(e) => {
                             debug!("Timer error for WG peer {}: {:?}",
@@ -617,19 +721,39 @@ pub async fn run_wg_listener(
             _ = stats_timer.tick() => {
                 let mut clients = state.clients.write().await;
                 let mut stats = state.stats.write().await;
-                for peer in peers.iter() {
+                for peer in peers.iter_mut() {
+                    // Always update aggregate stats (regardless of connection state)
+                    let delta_sent = peer.stats.bytes_sent.saturating_sub(peer.prev_synced_bytes_sent);
+                    let delta_recv = peer.stats.bytes_received.saturating_sub(peer.prev_synced_bytes_received);
+                    if delta_sent > 0 || delta_recv > 0 {
+                        stats.bytes_sent += delta_sent;
+                        stats.bytes_received += delta_recv;
+                        peer.prev_synced_bytes_sent = peer.stats.bytes_sent;
+                        peer.prev_synced_bytes_received = peer.stats.bytes_received;
+                    }
+
+                    // Only update ClientInfo if peer is connected (in state.clients)
                     let client_id = format!("wg-{}", &peer.public_key_b64[..8.min(peer.public_key_b64.len())]);
                     if let Some(info) = clients.get_mut(&client_id) {
-                        // Update stats delta
-                        let prev_sent = info.bytes_sent;
-                        let prev_recv = info.bytes_received;
                         info.bytes_sent = peer.stats.bytes_sent;
                         info.bytes_received = peer.stats.bytes_received;
                         info.remote_addr = peer.endpoint.map(|e| e.to_string());
+                    }
+                }
+            }
 
-                        // Update aggregate stats
+            // --- Idle timeout check (every 10s) ---
-                        stats.bytes_sent += peer.stats.bytes_sent.saturating_sub(prev_sent);
+            _ = idle_check_timer.tick() => {
-                        stats.bytes_received += peer.stats.bytes_received.saturating_sub(prev_recv);
+                let now = tokio::time::Instant::now();
+                for peer in peers.iter_mut() {
+                    if peer.is_connected {
+                        if let Some(last) = peer.last_activity_at {
+                            if now.duration_since(last) > std::time::Duration::from_secs(180) {
+                                info!("WG peer {} idle timeout (180s), disconnecting", peer.public_key_b64);
+                                peer.is_connected = false;
+                                disconnect_wg_peer(&state, &peer.public_key_b64).await;
+                            }
+                        }
                     }
                 }
             }
@@ -646,11 +770,12 @@ pub async fn run_wg_listener(
                             &config.private_key,
                         );
                         if result.is_ok() {
-                            // Register new peer in ServerState
+                            // Register new peer in ServerState (IP + tun_routes only)
-                            let peer = peers.last().unwrap();
+                            let peer = peers.last_mut().unwrap();
                             match register_wg_peer(&state, peer, &wg_return_tx).await {
                                 Ok(Some(ip)) => {
                                     peer_vpn_ips.insert(peer_config.public_key.clone(), ip);
+                                    peer.vpn_ip = Some(ip);
                                 }
                                 Ok(None) => {}
                                 Err(e) => {
@@ -1239,7 +1364,7 @@ mod tests {
                 let _ = server_tunn.decapsulate(None, &pkt_copy, &mut buf_b);
             }
             TunnResult::Done => {}
-            other => {
+            _other => {
                 // Drain
                 loop {
                     match client_tunn.decapsulate(None, &[], &mut buf_a) {

ts/00_commitinfo_data.ts

@@ -3,6 +3,6 @@
  */
 export const commitinfo = {
   name: '@push.rocks/smartvpn',
-  version: '1.16.4',
+  version: '1.19.2',
   description: 'A VPN solution with TypeScript control plane and Rust data plane daemon'
 }

ts/smartvpn.classes.vpnserver.ts

@@ -333,17 +333,35 @@ export class VpnServer extends plugins.events.EventEmitter {
   /**
    * Stop the daemon bridge.
    */
-  public stop(): void {
+  public async stop(): Promise<void> {
     // Clean up nftables rules
     if (this.nftHealthInterval) {
       clearInterval(this.nftHealthInterval);
       this.nftHealthInterval = undefined;
     }
     if (this.nft) {
-      this.nft.cleanup().catch(() => {}); // best-effort cleanup
+      try {
+        await this.nft.cleanup();
+      } catch (e) {
+        console.warn(`[smartvpn] nftables cleanup failed: ${e}`);
+      }
       this.nft = undefined;
     }
+
+    // Wait for bridge process to exit (with timeout)
+    const exitPromise = new Promise<void>((resolve) => {
+      if (!this.bridge.running) {
+        resolve();
+        return;
+      }
+      const timeout = setTimeout(() => resolve(), 5000);
+      this.bridge.once('exit', () => {
+        clearTimeout(timeout);
+        resolve();
+      });
+    });
     this.bridge.stop();
+    await exitPromise;
   }
 
   /**

ts/smartvpn.interfaces.ts

@@ -92,8 +92,9 @@ export interface IVpnServerConfig {
   /** Enable NAT/masquerade for client traffic */
   enableNat?: boolean;
   /** Forwarding mode: 'tun' (kernel TUN, requires root), 'socket' (userspace NAT),
+   *  'bridge' (L2 bridge to host LAN), 'hybrid' (per-client socket+bridge),
    *  or 'testing' (monitoring only). Default: 'testing'. */
-  forwardingMode?: 'tun' | 'socket' | 'testing';
+  forwardingMode?: 'tun' | 'socket' | 'bridge' | 'hybrid' | 'testing';
   /** Default rate limit for new clients (bytes/sec). Omit for unlimited. */
   defaultRateLimitBytesPerSec?: number;
   /** Default burst size for new clients (bytes). Omit for unlimited. */
@@ -137,6 +138,22 @@ export interface IVpnServerConfig {
    *  Controls what traffic the client routes through the VPN tunnel.
    *  Defaults to ['0.0.0.0/0'] (full tunnel). Set to e.g. ['10.8.0.0/24'] for split tunnel. */
   clientAllowedIPs?: string[];
+
+  // Bridge mode configuration (forwardingMode: 'bridge')
+
+  /** LAN subnet CIDR for bridge mode (e.g. '192.168.1.0/24').
+   *  VPN clients get IPs from this subnet instead of the VPN subnet.
+   *  Required when forwardingMode is 'bridge'. */
+  bridgeLanSubnet?: string;
+  /** Physical network interface to bridge (e.g. 'eth0').
+   *  Auto-detected from the default route if omitted. */
+  bridgePhysicalInterface?: string;
+  /** Start of VPN client IP range within the LAN subnet (host offset, e.g. 200 for .200).
+   *  Default: 200. */
+  bridgeIpRangeStart?: number;
+  /** End of VPN client IP range within the LAN subnet (host offset, e.g. 250 for .250).
+   *  Default: 250. */
+  bridgeIpRangeEnd?: number;
 }
 
 /**
@@ -217,6 +234,14 @@ export interface IVpnClientInfo {
 export interface IVpnServerStatistics extends IVpnStatistics {
   activeClients: number;
   totalConnections: number;
+  /** Per-transport active client counts. */
+  activeClientsWebsocket: number;
+  activeClientsQuic: number;
+  activeClientsWireguard: number;
+  /** Per-transport total connection counts. */
+  totalConnectionsWebsocket: number;
+  totalConnectionsQuic: number;
+  totalConnectionsWireguard: number;
 }
 
 export interface IVpnKeypair {
@@ -302,6 +327,10 @@ export interface IClientSecurity {
   maxConnections?: number;
   /** Per-client rate limiting. */
   rateLimit?: IClientRateLimit;
+  /** Per-client destination routing policy override.
+   *  When set, overrides the server-level destinationPolicy for this client's traffic.
+   *  Supports the same options: forceTarget, block, allow with allow/block lists. */
+  destinationPolicy?: IDestinationPolicy;
 }
 
 /**
@@ -333,6 +362,21 @@ export interface IClientEntry {
   expiresAt?: string;
   /** Assigned VPN IP address (set by server) */
   assignedIp?: string;
+
+  // Per-client bridge/host-IP settings
+
+  /** If true, client gets a host network IP via bridge mode (L2 to LAN).
+   *  If false (default), client gets a VPN subnet IP via socket/NAT mode. */
+  useHostIp?: boolean;
+  /** If true and useHostIp is true, obtain IP via DHCP relay.
+   *  If false or omitted, use staticIp or auto-assign from bridge IP range. */
+  useDhcp?: boolean;
+  /** Static LAN IP when useHostIp is true and useDhcp is false. */
+  staticIp?: string;
+  /** If true, assign this client to a specific 802.1Q VLAN on the bridge. */
+  forceVlan?: boolean;
+  /** 802.1Q VLAN ID (1-4094). Required when forceVlan is true. */
+  vlanId?: number;
 }
 
 /**