Merge pull request #424 from Serph91P/fix/compat-registration-chain-race

fix: preserve the pending WireGuard registration chain

exitnode/exitnode.go

@@ -1,176 +0,0 @@
-// Package exitnode implements the exit-node ping dance run before
-// registering with the server: request the candidate exit nodes, ping each
-// one over HTTP, and report the results so the server can pick the best one.
-// It is shared between newt and olm, which both register the same way.
-package exitnode
-
-import (
-	"net/http"
-	"strings"
-	"time"
-
-	"github.com/fosrl/newt/logger"
-)
-
-// ExitNodeData is the payload the server sends in response to a
-// "*/ping/request" message.
-type ExitNodeData struct {
-	ExitNodes []ExitNode `json:"exitNodes"`
-	ChainId   string     `json:"chainId"`
-}
-
-// ExitNode is a candidate exit node offered by the server for ping selection.
-type ExitNode struct {
-	ID                     int     `json:"exitNodeId"`
-	Name                   string  `json:"exitNodeName"`
-	Endpoint               string  `json:"endpoint"`
-	Weight                 float64 `json:"weight"`
-	WasPreviouslyConnected bool    `json:"wasPreviouslyConnected"`
-}
-
-// ExitNodePingResult is the measured latency (or error) for one exit node,
-// sent back to the server in the "*/wg/register" message's pingResults field.
-type ExitNodePingResult struct {
-	ExitNodeID             int     `json:"exitNodeId"`
-	LatencyMs              int64   `json:"latencyMs"`
-	Weight                 float64 `json:"weight"`
-	Error                  string  `json:"error,omitempty"`
-	Name                   string  `json:"exitNodeName"`
-	Endpoint               string  `json:"endpoint"`
-	WasPreviouslyConnected bool    `json:"wasPreviouslyConnected"`
-}
-
-// PingExitNodes pings the given exit nodes over HTTP and returns a per-node
-// ExitNodePingResult suitable for inclusion in a wg/register message's
-// pingResults field, so the server can select the best exit node.
-//
-// If there's only one exit node, or preferEndpoint names one of them, the
-// matching node is returned immediately with LatencyMs 0 and no pinging is
-// done. Otherwise every node is pinged pingAttempts times over HTTP GET
-// <endpoint>/ping and the average latency of successful attempts is used.
-//
-// When alreadyConnected is true, a node flagged WasPreviouslyConnected is
-// excluded from the results as long as at least one other healthy node is
-// available, biasing reconnects toward switching away from a possibly
-// degraded node.
-func PingExitNodes(exitNodes []ExitNode, preferEndpoint string, alreadyConnected bool) []ExitNodePingResult {
-	if len(exitNodes) == 0 {
-		return nil
-	}
-
-	if len(exitNodes) == 1 || preferEndpoint != "" {
-		selected := exitNodes[0]
-		if preferEndpoint != "" {
-			for _, node := range exitNodes {
-				if node.Endpoint == preferEndpoint {
-					selected = node
-					break
-				}
-			}
-		}
-
-		logger.Debug("Only one exit node available, using it directly: %s", selected.Endpoint)
-
-		return []ExitNodePingResult{
-			{
-				ExitNodeID:             selected.ID,
-				LatencyMs:              0,
-				Weight:                 selected.Weight,
-				Error:                  "",
-				Name:                   selected.Name,
-				Endpoint:               selected.Endpoint,
-				WasPreviouslyConnected: selected.WasPreviouslyConnected,
-			},
-		}
-	}
-
-	type nodeResult struct {
-		Node    ExitNode
-		Latency time.Duration
-		Err     error
-	}
-
-	results := make([]nodeResult, len(exitNodes))
-	const pingAttempts = 3
-	for i, node := range exitNodes {
-		var totalLatency time.Duration
-		var lastErr error
-		successes := 0
-		httpClient := &http.Client{
-			Timeout: 5 * time.Second,
-		}
-		url := node.Endpoint
-		if !strings.HasPrefix(url, "http://") && !strings.HasPrefix(url, "https://") {
-			url = "http://" + url
-		}
-		if !strings.HasSuffix(url, "/ping") {
-			url = strings.TrimRight(url, "/") + "/ping"
-		}
-		for j := 0; j < pingAttempts; j++ {
-			start := time.Now()
-			resp, err := httpClient.Get(url)
-			latency := time.Since(start)
-			if err != nil {
-				lastErr = err
-				logger.Warn("Failed to ping exit node %d (%s) attempt %d: %v", node.ID, url, j+1, err)
-				continue
-			}
-			resp.Body.Close()
-			totalLatency += latency
-			successes++
-		}
-		var avgLatency time.Duration
-		if successes > 0 {
-			avgLatency = totalLatency / time.Duration(successes)
-		}
-		if successes == 0 {
-			results[i] = nodeResult{Node: node, Latency: 0, Err: lastErr}
-		} else {
-			results[i] = nodeResult{Node: node, Latency: avgLatency, Err: nil}
-		}
-	}
-
-	var pingResults []ExitNodePingResult
-	for _, res := range results {
-		errMsg := ""
-		if res.Err != nil {
-			errMsg = res.Err.Error()
-		}
-		pingResults = append(pingResults, ExitNodePingResult{
-			ExitNodeID:             res.Node.ID,
-			LatencyMs:              res.Latency.Milliseconds(),
-			Weight:                 res.Node.Weight,
-			Error:                  errMsg,
-			Name:                   res.Node.Name,
-			Endpoint:               res.Node.Endpoint,
-			WasPreviouslyConnected: res.Node.WasPreviouslyConnected,
-		})
-	}
-
-	if alreadyConnected {
-		var filteredPingResults []ExitNodePingResult
-		previouslyConnectedNodeIdx := -1
-		for i, res := range pingResults {
-			if res.WasPreviouslyConnected {
-				previouslyConnectedNodeIdx = i
-			}
-		}
-		goodNodeCount := 0
-		for i, res := range pingResults {
-			if i != previouslyConnectedNodeIdx && res.LatencyMs > 0 && res.Error == "" {
-				goodNodeCount++
-			}
-		}
-		if previouslyConnectedNodeIdx != -1 && goodNodeCount > 0 {
-			for i, res := range pingResults {
-				if i != previouslyConnectedNodeIdx {
-					filteredPingResults = append(filteredPingResults, res)
-				}
-			}
-			pingResults = filteredPingResults
-			logger.Info("Excluding previously connected exit node from ping results due to other available nodes")
-		}
-	}
-
-	return pingResults
-}

network/interface.go

@@ -115,10 +115,6 @@ func FindUnusedUTUN() (string, error) {
 }
 
 func configureDarwin(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
-	if NativeConfigDisabled {
-		return nil
-	}
-
 	logger.Info("Configuring darwin interface: %s", interfaceName)
 
 	prefix, _ := ipNet.Mask.Size()
@@ -171,99 +167,3 @@ func configureLinux(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
 
 	return nil
 }
-
-// AddSecondaryAddress adds an additional IP address (given as CIDR, e.g. "10.10.0.5/32")
-// to an already-configured interface. It also records the address in the shared
-// NetworkSettings (see AddIPv4Address) so mobile (iOS/Android) packet-tunnel providers
-// pick it up on their next settings poll - those platforms have no OS-level interface
-// to configure directly, so this is the only way they learn about the address.
-func AddSecondaryAddress(interfaceName string, addr string) error {
-	ip, ipNet, err := net.ParseCIDR(addr)
-	if err != nil {
-		return fmt.Errorf("invalid IP address: %v", err)
-	}
-
-	mask := net.IP(ipNet.Mask).String()
-	AddIPv4Address(ip.String(), mask)
-
-	if interfaceName == "" {
-		return nil
-	}
-
-	switch runtime.GOOS {
-	case "linux":
-		return configureLinux(interfaceName, ip, ipNet)
-	case "darwin":
-		return configureDarwin(interfaceName, ip, ipNet)
-	case "windows":
-		return configureWindows(interfaceName, ip, ipNet)
-	default:
-		return nil
-	}
-}
-
-// RemoveSecondaryAddress removes an IP address (given as CIDR) previously added with
-// AddSecondaryAddress, including from the shared NetworkSettings used by mobile
-// packet-tunnel providers.
-func RemoveSecondaryAddress(interfaceName string, addr string) error {
-	ip, ipNet, err := net.ParseCIDR(addr)
-	if err != nil {
-		return fmt.Errorf("invalid IP address: %v", err)
-	}
-
-	RemoveIPv4Address(ip.String())
-
-	if interfaceName == "" {
-		return nil
-	}
-
-	switch runtime.GOOS {
-	case "linux":
-		return removeLinuxAddress(interfaceName, ip, ipNet)
-	case "darwin":
-		return removeDarwinAddress(interfaceName, ip, ipNet)
-	case "windows":
-		return removeWindowsAddress(interfaceName, ip, ipNet)
-	default:
-		return nil
-	}
-}
-
-func removeLinuxAddress(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
-	link, err := netlink.LinkByName(interfaceName)
-	if err != nil {
-		return fmt.Errorf("failed to get interface %s: %v", interfaceName, err)
-	}
-
-	addr := &netlink.Addr{
-		IPNet: &net.IPNet{
-			IP:   ip,
-			Mask: ipNet.Mask,
-		},
-	}
-
-	if err := netlink.AddrDel(link, addr); err != nil {
-		return fmt.Errorf("failed to remove IP address: %v", err)
-	}
-
-	return nil
-}
-
-func removeDarwinAddress(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
-	if NativeConfigDisabled {
-		return nil
-	}
-
-	prefix, _ := ipNet.Mask.Size()
-	ipStr := fmt.Sprintf("%s/%d", ip.String(), prefix)
-
-	cmd := exec.Command("/sbin/ifconfig", interfaceName, "inet", ipStr, "-alias")
-	logger.Info("Running command: %v", cmd)
-
-	out, err := cmd.CombinedOutput()
-	if err != nil {
-		return fmt.Errorf("ifconfig command failed: %v, output: %s", err, out)
-	}
-
-	return nil
-}

network/interface_notwindows.go

@@ -10,7 +10,3 @@ import (
 func configureWindows(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
 	return fmt.Errorf("configureWindows called on non-Windows platform")
 }
-
-func removeWindowsAddress(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
-	return fmt.Errorf("removeWindowsAddress called on non-Windows platform")
-}

network/interface_windows.go

@@ -61,35 +61,3 @@ func configureWindows(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
 
 	return nil
 }
-
-func removeWindowsAddress(interfaceName string, ip net.IP, ipNet *net.IPNet) error {
-	iface, err := net.InterfaceByName(interfaceName)
-	if err != nil {
-		return fmt.Errorf("failed to get interface %s: %v", interfaceName, err)
-	}
-
-	luid, err := winipcfg.LUIDFromIndex(uint32(iface.Index))
-	if err != nil {
-		return fmt.Errorf("failed to get LUID for interface %s: %v", interfaceName, err)
-	}
-
-	maskBits, _ := ipNet.Mask.Size()
-
-	var addr netip.Addr
-	if ip4 := ip.To4(); ip4 != nil {
-		addr, _ = netip.AddrFromSlice(ip4)
-	} else {
-		addr, _ = netip.AddrFromSlice(ip)
-	}
-	if !addr.IsValid() {
-		return fmt.Errorf("failed to convert IP address")
-	}
-	prefix := netip.PrefixFrom(addr, maskBits)
-
-	logger.Info("Removing IP address %s from interface %s", prefix.String(), interfaceName)
-	if err := luid.DeleteIPAddress(prefix); err != nil {
-		return fmt.Errorf("failed to remove IP address: %v", err)
-	}
-
-	return nil
-}

network/route.go

@@ -32,20 +32,6 @@ const VPNRouteMetric = 9999
 // this to true (e.g. from a config value) before routes are added.
 var PreferLocalRoutes = false
 
-// NativeConfigDisabled, when true, skips the raw `ifconfig`/`route` subprocess
-// calls this package otherwise makes on darwin (configureDarwin,
-// removeDarwinAddress, DarwinAddRouteWithSource, DarwinRemoveRoute) while still
-// populating the JSON-facing NetworkSettings state. This must be set when the
-// TUN device's addresses/routes are instead owned by an external mechanism
-// that reconciles them independently - namely Apple's NetworkExtension
-// (NEPacketTunnelProvider.setTunnelNetworkSettings), which is the sole
-// sanctioned way to configure that virtual interface. Running our own
-// ifconfig/route commands in addition to NE applying its own settings was
-// observed to install two competing routes to the same destination (one via
-// NE's gatewayAddress-based route, one via our own `-ifa` route), so the two
-// mechanisms must be mutually exclusive rather than layered.
-var NativeConfigDisabled = false
-
 // DarwinAddRoute adds a route via the BSD routing table. Unlike Linux/Windows,
 // BSD's routing table has no per-route metric - preference between an
 // overlapping local route and this VPN route is instead resolved by
@@ -53,43 +39,22 @@ var NativeConfigDisabled = false
 // rather than replacing an existing route to the same destination, so a local
 // route is never displaced by one we add here.
 func DarwinAddRoute(destination string, gateway string, interfaceName string) error {
-	return DarwinAddRouteWithSource(destination, gateway, interfaceName, "")
-}
-
-// DarwinAddRouteWithSource is DarwinAddRoute with an explicit source address
-// (route(8) `-ifa`). This is required when the interface carries more than
-// one address (e.g. an exit node's secondary tunnel address alongside the
-// site tunnel's primary address): without `-ifa`, BSD picks a source address
-// for the route on its own - typically the interface's primary address - and
-// WireGuard's own reverse-path filtering on the remote end will silently drop
-// packets whose source doesn't match the peer's configured AllowedIPs, even
-// though the tunnel/handshake itself stays up.
-func DarwinAddRouteWithSource(destination string, gateway string, interfaceName string, sourceIP string) error {
 	if runtime.GOOS != "darwin" {
 		return nil
 	}
-	if NativeConfigDisabled {
-		return nil
-	}
 
-	var args []string
+	var cmd *exec.Cmd
 
 	if gateway != "" {
 		// Route with specific gateway
-		args = []string{"-q", "-n", "add", "-inet", destination, "-gateway", gateway}
+		cmd = exec.Command("route", "-q", "-n", "add", "-inet", destination, "-gateway", gateway)
 	} else if interfaceName != "" {
 		// Route via interface
-		args = []string{"-q", "-n", "add", "-inet", destination, "-interface", interfaceName}
+		cmd = exec.Command("route", "-q", "-n", "add", "-inet", destination, "-interface", interfaceName)
 	} else {
 		return fmt.Errorf("either gateway or interface must be specified")
 	}
 
-	if sourceIP != "" {
-		args = append(args, "-ifa", sourceIP)
-	}
-
-	cmd := exec.Command("route", args...)
-
 	logger.Info("Running command: %v", cmd)
 
 	out, err := cmd.CombinedOutput()
@@ -104,9 +69,6 @@ func DarwinRemoveRoute(destination string) error {
 	if runtime.GOOS != "darwin" {
 		return nil
 	}
-	if NativeConfigDisabled {
-		return nil
-	}
 
 	cmd := exec.Command("route", "-q", "-n", "delete", "-inet", destination)
 	logger.Info("Running command: %v", cmd)
@@ -211,30 +173,15 @@ func LinuxRemoveRoute(destination string, interfaceName string) error {
 
 // addRouteForServerIP adds an OS-specific route for the server IP
 func AddRouteForServerIP(serverIP, interfaceName string) error {
-	return AddRouteForServerIPWithSource(serverIP, interfaceName, "")
-}
-
-// AddRouteForServerIPWithSource is AddRouteForServerIP with an explicit source
-// address for the darwin route (see DarwinAddRouteWithSource) - needed when
-// the interface carries more than one address, e.g. an exit node connection
-// where the interface's primary address belongs to the site tunnel rather
-// than the exit node.
-func AddRouteForServerIPWithSource(serverIP, interfaceName string, sourceIP string) error {
 	if interfaceName == "" {
 		return nil
 	}
-
-	// Populate the NetworkSettings entry (and its gatewayAddress, for the
-	// NetworkExtension source-pinning trick above) unconditionally, same as
-	// AddRoutesWithSource does for remote subnets - mobile packet-tunnel
-	// providers rely on this regardless of GOOS.
-	if err := AddRouteForNetworkConfigWithGateway(serverIP, sourceIP); err != nil {
-		return err
-	}
-
 	// TODO: does this also need to be ios?
 	if runtime.GOOS == "darwin" { // macos requires routes for each peer to be added but this messes with other platforms
-		return DarwinAddRouteWithSource(serverIP, "", interfaceName, sourceIP)
+		if err := AddRouteForNetworkConfig(serverIP); err != nil {
+			return err
+		}
+		return DarwinAddRoute(serverIP, "", interfaceName)
 	}
 	// else if runtime.GOOS == "windows" {
 	//	return WindowsAddRoute(serverIP, "", interfaceName)
@@ -246,24 +193,14 @@ func AddRouteForServerIPWithSource(serverIP, interfaceName string, sourceIP stri
 
 // removeRouteForServerIP removes an OS-specific route for the server IP
 func RemoveRouteForServerIP(serverIP string, interfaceName string) error {
-	return RemoveRouteForServerIPWithSource(serverIP, interfaceName, "")
-}
-
-// RemoveRouteForServerIPWithSource is RemoveRouteForServerIP with an explicit
-// source/gateway address - must match whatever was passed to
-// AddRouteForServerIPWithSource when the route was added (see
-// RemoveRouteForNetworkConfigWithGateway).
-func RemoveRouteForServerIPWithSource(serverIP string, interfaceName string, sourceIP string) error {
 	if interfaceName == "" {
 		return nil
 	}
-
-	if err := RemoveRouteForNetworkConfigWithGateway(serverIP, sourceIP); err != nil {
-		return err
-	}
-
 	// TODO: does this also need to be ios?
 	if runtime.GOOS == "darwin" { // macos requires routes for each peer to be added but this messes with other platforms
+		if err := RemoveRouteForNetworkConfig(serverIP); err != nil {
+			return err
+		}
 		return DarwinRemoveRoute(serverIP)
 	}
 	// else if runtime.GOOS == "windows" {
@@ -275,20 +212,6 @@ func RemoveRouteForServerIPWithSource(serverIP string, interfaceName string, sou
 }
 
 func AddRouteForNetworkConfig(destination string) error {
-	return AddRouteForNetworkConfigWithGateway(destination, "")
-}
-
-// AddRouteForNetworkConfigWithGateway is AddRouteForNetworkConfig with an
-// explicit gateway address for the route entry surfaced via NetworkSettings.
-// This is consumed by mobile (iOS/macOS NetworkExtension) packet-tunnel
-// providers as NEIPv4Route.gatewayAddress. NetworkExtension gives us no
-// direct way to pin a route's source address (no equivalent of BSD's `route
-// -ifa`) - but setting gatewayAddress to one of the tunnel interface's own
-// addresses makes the OS resolve "how do I reach this gateway" recursively
-// to that address/interface pairing, which is what determines the source
-// address used for packets matching the route. This is the same underlying
-// mechanism as `route add -gateway` (see DarwinAddRoute's gateway branch).
-func AddRouteForNetworkConfigWithGateway(destination string, gateway string) error {
 	// Parse the subnet to extract IP and mask
 	_, ipNet, err := net.ParseCIDR(destination)
 	if err != nil {
@@ -299,21 +222,12 @@ func AddRouteForNetworkConfigWithGateway(destination string, gateway string) err
 	mask := net.IP(ipNet.Mask).String()
 	destinationAddress := ipNet.IP.String()
 
-	AddIPv4IncludedRoute(IPv4Route{DestinationAddress: destinationAddress, SubnetMask: mask, GatewayAddress: gateway})
+	AddIPv4IncludedRoute(IPv4Route{DestinationAddress: destinationAddress, SubnetMask: mask})
 
 	return nil
 }
 
 func RemoveRouteForNetworkConfig(destination string) error {
-	return RemoveRouteForNetworkConfigWithGateway(destination, "")
-}
-
-// RemoveRouteForNetworkConfigWithGateway is RemoveRouteForNetworkConfig with
-// an explicit gateway address. This must match whatever gateway the route was
-// added with (see AddRouteForNetworkConfigWithGateway) - RemoveIPv4IncludedRoute
-// matches by full struct equality, so a mismatched gateway means the entry is
-// silently never found/removed.
-func RemoveRouteForNetworkConfigWithGateway(destination string, gateway string) error {
 	// Parse the subnet to extract IP and mask
 	_, ipNet, err := net.ParseCIDR(destination)
 	if err != nil {
@@ -324,23 +238,13 @@ func RemoveRouteForNetworkConfigWithGateway(destination string, gateway string)
 	mask := net.IP(ipNet.Mask).String()
 	destinationAddress := ipNet.IP.String()
 
-	RemoveIPv4IncludedRoute(IPv4Route{DestinationAddress: destinationAddress, SubnetMask: mask, GatewayAddress: gateway})
+	RemoveIPv4IncludedRoute(IPv4Route{DestinationAddress: destinationAddress, SubnetMask: mask})
 
 	return nil
 }
 
 // addRoutes adds routes for each subnet in RemoteSubnets
 func AddRoutes(remoteSubnets []string, interfaceName string) error {
-	return AddRoutesWithSource(remoteSubnets, interfaceName, "")
-}
-
-// AddRoutesWithSource is AddRoutes with an explicit source address for the
-// darwin routes (see DarwinAddRouteWithSource) - needed when the interface
-// carries more than one address (e.g. a site tunnel address alongside an
-// exit node's secondary address), so the routes for these subnets are pinned
-// to the address they actually belong to rather than whichever address
-// darwin would otherwise default to.
-func AddRoutesWithSource(remoteSubnets []string, interfaceName string, sourceIP string) error {
 	if len(remoteSubnets) == 0 {
 		return nil
 	}
@@ -364,7 +268,7 @@ func AddRoutesWithSource(remoteSubnets []string, interfaceName string, sourceIP
 
 		switch runtime.GOOS {
 		case "darwin":
-			if err := DarwinAddRouteWithSource(subnet, "", interfaceName, sourceIP); err != nil {
+			if err := DarwinAddRoute(subnet, "", interfaceName); err != nil {
 				logger.Error("Failed to add Darwin route for subnet %s: %v", subnet, err)
 			}
 		case "windows":

network/settings.go

@@ -81,45 +81,6 @@ func SetIPv4Settings(addresses []string, subnetMasks []string) {
 	logger.Info("Set IPv4 addresses: %v, subnet masks: %v", addresses, subnetMasks)
 }
 
-// AddIPv4Address appends an additional IPv4 address/subnet mask pair to the
-// tunnel's network settings. This is how a secondary interface address gets
-// exposed to mobile (iOS/Android) packet-tunnel providers, which read the
-// full IPv4Addresses/IPv4SubnetMasks arrays (not just the first entry) and
-// re-apply them on every settings poll.
-func AddIPv4Address(address string, subnetMask string) {
-	networkSettingsMutex.Lock()
-	defer networkSettingsMutex.Unlock()
-
-	for _, a := range networkSettings.IPv4Addresses {
-		if a == address {
-			logger.Info("IPv4 address already exists: %s", address)
-			return
-		}
-	}
-
-	networkSettings.IPv4Addresses = append(networkSettings.IPv4Addresses, address)
-	networkSettings.IPv4SubnetMasks = append(networkSettings.IPv4SubnetMasks, subnetMask)
-	incrementor++
-	logger.Info("Added IPv4 address: %s/%s", address, subnetMask)
-}
-
-// RemoveIPv4Address removes a previously added secondary IPv4 address.
-func RemoveIPv4Address(address string) {
-	networkSettingsMutex.Lock()
-	defer networkSettingsMutex.Unlock()
-
-	for i, a := range networkSettings.IPv4Addresses {
-		if a == address {
-			networkSettings.IPv4Addresses = append(networkSettings.IPv4Addresses[:i], networkSettings.IPv4Addresses[i+1:]...)
-			networkSettings.IPv4SubnetMasks = append(networkSettings.IPv4SubnetMasks[:i], networkSettings.IPv4SubnetMasks[i+1:]...)
-			incrementor++
-			logger.Info("Removed IPv4 address: %s", address)
-			return
-		}
-	}
-	logger.Info("IPv4 address not found for removal: %s", address)
-}
-
 // SetIPv4IncludedRoutes sets the included IPv4 routes
 func SetIPv4IncludedRoutes(routes []IPv4Route) {
 	networkSettingsMutex.Lock()

newt/handlers.go

@@ -10,13 +10,13 @@ import (
 	"net/http"
 	"os"
 	"os/signal"
+	"strings"
 	"syscall"
 	"time"
 
 	"github.com/fosrl/newt/authdaemon"
 	"github.com/fosrl/newt/browsergateway"
 	"github.com/fosrl/newt/docker"
-	"github.com/fosrl/newt/exitnode"
 	"github.com/fosrl/newt/healthcheck"
 	"github.com/fosrl/newt/internal/state"
 	"github.com/fosrl/newt/internal/telemetry"
@@ -140,7 +140,129 @@ func (n *Newt) registerHandlers(ctx context.Context) {
 			return
 		}
 
-		pingResults := exitnode.PingExitNodes(exitNodes, n.config.PreferEndpoint, n.connected)
+		if len(exitNodes) == 1 || n.config.PreferEndpoint != "" {
+			logger.Debug("Only one exit node available, using it directly: %s", exitNodes[0].Endpoint)
+
+			if n.config.PreferEndpoint != "" {
+				for _, node := range exitNodes {
+					if node.Endpoint == n.config.PreferEndpoint {
+						exitNodes[0] = node
+						break
+					}
+				}
+			}
+
+			pingResults := []ExitNodePingResult{
+				{
+					ExitNodeID:             exitNodes[0].ID,
+					LatencyMs:              0,
+					Weight:                 exitNodes[0].Weight,
+					Error:                  "",
+					Name:                   exitNodes[0].Name,
+					Endpoint:               exitNodes[0].Endpoint,
+					WasPreviouslyConnected: exitNodes[0].WasPreviouslyConnected,
+				},
+			}
+
+			chainId := generateChainId()
+			n.pendingRegisterChainId = chainId
+			n.stopFunc = n.client.SendMessageInterval(topicWGRegister, map[string]interface{}{
+				"publicKey":   n.publicKey.String(),
+				"pingResults": pingResults,
+				"newtVersion": n.config.Version,
+				"chainId":     chainId,
+			}, 2*time.Second)
+
+			return
+		}
+
+		type nodeResult struct {
+			Node    ExitNode
+			Latency time.Duration
+			Err     error
+		}
+
+		results := make([]nodeResult, len(exitNodes))
+		const pingAttempts = 3
+		for i, node := range exitNodes {
+			var totalLatency time.Duration
+			var lastErr error
+			successes := 0
+			httpClient := &http.Client{
+				Timeout: 5 * time.Second,
+			}
+			url := node.Endpoint
+			if !strings.HasPrefix(url, "http://") && !strings.HasPrefix(url, "https://") {
+				url = "http://" + url
+			}
+			if !strings.HasSuffix(url, "/ping") {
+				url = strings.TrimRight(url, "/") + "/ping"
+			}
+			for j := 0; j < pingAttempts; j++ {
+				start := time.Now()
+				resp, err := httpClient.Get(url)
+				latency := time.Since(start)
+				if err != nil {
+					lastErr = err
+					logger.Warn("Failed to ping exit node %d (%s) attempt %d: %v", node.ID, url, j+1, err)
+					continue
+				}
+				resp.Body.Close()
+				totalLatency += latency
+				successes++
+			}
+			var avgLatency time.Duration
+			if successes > 0 {
+				avgLatency = totalLatency / time.Duration(successes)
+			}
+			if successes == 0 {
+				results[i] = nodeResult{Node: node, Latency: 0, Err: lastErr}
+			} else {
+				results[i] = nodeResult{Node: node, Latency: avgLatency, Err: nil}
+			}
+		}
+
+		var pingResults []ExitNodePingResult
+		for _, res := range results {
+			errMsg := ""
+			if res.Err != nil {
+				errMsg = res.Err.Error()
+			}
+			pingResults = append(pingResults, ExitNodePingResult{
+				ExitNodeID:             res.Node.ID,
+				LatencyMs:              res.Latency.Milliseconds(),
+				Weight:                 res.Node.Weight,
+				Error:                  errMsg,
+				Name:                   res.Node.Name,
+				Endpoint:               res.Node.Endpoint,
+				WasPreviouslyConnected: res.Node.WasPreviouslyConnected,
+			})
+		}
+
+		if n.connected {
+			var filteredPingResults []ExitNodePingResult
+			previouslyConnectedNodeIdx := -1
+			for i, res := range pingResults {
+				if res.WasPreviouslyConnected {
+					previouslyConnectedNodeIdx = i
+				}
+			}
+			goodNodeCount := 0
+			for i, res := range pingResults {
+				if i != previouslyConnectedNodeIdx && res.LatencyMs > 0 && res.Error == "" {
+					goodNodeCount++
+				}
+			}
+			if previouslyConnectedNodeIdx != -1 && goodNodeCount > 0 {
+				for i, res := range pingResults {
+					if i != previouslyConnectedNodeIdx {
+						filteredPingResults = append(filteredPingResults, res)
+					}
+				}
+				pingResults = filteredPingResults
+				logger.Info("Excluding previously connected exit node from ping results due to other available nodes")
+			}
+		}
 
 		chainId := generateChainId()
 		n.pendingRegisterChainId = chainId
@@ -888,7 +1010,9 @@ func (n *Newt) registerHandlers(ctx context.Context) {
 		}
 
 		bcChainId := generateChainId()
-		n.pendingRegisterChainId = bcChainId
+		// Pangolin intentionally does not answer backwards-compatible
+		// registrations with newt/wg/connect. Do not replace the chain ID of
+		// the real registration while its response may already be in flight.
 		if err := n.client.SendMessage(topicWGRegister, map[string]interface{}{
 			"publicKey":           n.publicKey.String(),
 			"newtVersion":         n.config.Version,

newt/ping.go

@@ -280,7 +280,8 @@ func (n *Newt) startPingCheck(fn pingFunc, serverIP, tunnelID string) chan struc
 							"chainId": pingChainId,
 						}, 3*time.Second)
 						bcChainId := generateChainId()
-						n.pendingRegisterChainId = bcChainId
+						// This compatibility message has no wg/connect response and must
+						// not supersede the pending real registration chain.
 						if err := n.client.SendMessage("newt/wg/register", map[string]interface{}{
 							"publicKey":           n.publicKey.String(),
 							"backwardsCompatible": true,

newt/types.go

@@ -2,7 +2,6 @@ package newt
 
 import (
 	wgclients "github.com/fosrl/newt/clients"
-	"github.com/fosrl/newt/exitnode"
 	"github.com/fosrl/newt/healthcheck"
 )
 
@@ -36,14 +35,28 @@ type TargetData struct {
 	Targets []string `json:"targets"`
 }
 
-// ExitNodeData, ExitNode and ExitNodePingResult are aliases for the shared
+type ExitNodeData struct {
-// exit-node ping dance types in package exitnode, kept here so existing code
+	ExitNodes []ExitNode `json:"exitNodes"`
-// in this package can keep referring to them unqualified.
+	ChainId   string     `json:"chainId"`
-type (
+}
-	ExitNodeData       = exitnode.ExitNodeData
+
-	ExitNode           = exitnode.ExitNode
+type ExitNode struct {
-	ExitNodePingResult = exitnode.ExitNodePingResult
+	ID                     int     `json:"exitNodeId"`
-)
+	Name                   string  `json:"exitNodeName"`
+	Endpoint               string  `json:"endpoint"`
+	Weight                 float64 `json:"weight"`
+	WasPreviouslyConnected bool    `json:"wasPreviouslyConnected"`
+}
+
+type ExitNodePingResult struct {
+	ExitNodeID             int     `json:"exitNodeId"`
+	LatencyMs              int64   `json:"latencyMs"`
+	Weight                 float64 `json:"weight"`
+	Error                  string  `json:"error,omitempty"`
+	Name                   string  `json:"exitNodeName"`
+	Endpoint               string  `json:"endpoint"`
+	WasPreviouslyConnected bool    `json:"wasPreviouslyConnected"`
+}
 
 type BlueprintResult struct {
 	Success bool   `json:"success"`

updates/selfupdate.go

@@ -52,8 +52,8 @@ type versionResponse struct {
 }
 
 // ErrAutoUpdateUnsupportedInOfficialContainer indicates auto-update is not
-// available when running inside container images.
+// available when running inside official Fossorial container images.
-var ErrAutoUpdateUnsupportedInOfficialContainer = errors.New("auto-update unsupported in container images")
+var ErrAutoUpdateUnsupportedInOfficialContainer = errors.New("auto-update unsupported in official Fossorial container images")
 
 // isOfficialContainer returns true when the process is running inside an
 // official Fossorial-built container image.  The image sets

websocket/client.go

@@ -28,15 +28,6 @@ import (
 	"go.opentelemetry.io/otel"
 )
 
-// writeDeadline bounds how long a websocket write may block before it is
-// treated as a failure. Without this, a write to a TCP connection whose
-// underlying network interface has disappeared (e.g. laptop sleep/resume,
-// Wi-Fi roam) can sit buffered in the kernel for minutes without erroring.
-// This matters even with the read-deadline/pong machinery below: if the
-// WriteJSON call in sendPing blocks, execution never reaches the
-// WriteControl ping that would otherwise trigger that read-side detection.
-const writeDeadline = 10 * time.Second
-
 type Client struct {
 	conn              *websocket.Conn
 	config            *Config
@@ -266,9 +257,6 @@ func (c *Client) SendMessage(messageType string, data interface{}) error {
 
 	c.writeMux.Lock()
 	defer c.writeMux.Unlock()
-	if err := c.conn.SetWriteDeadline(time.Now().Add(writeDeadline)); err != nil {
-		return err
-	}
 	if err := c.conn.WriteJSON(msg); err != nil {
 		return err
 	}
@@ -289,9 +277,6 @@ func (c *Client) SendMessageNoLog(messageType string, data interface{}) error {
 
 	c.writeMux.Lock()
 	defer c.writeMux.Unlock()
-	if err := c.conn.SetWriteDeadline(time.Now().Add(writeDeadline)); err != nil {
-		return err
-	}
 	if err := c.conn.WriteJSON(msg); err != nil {
 		return err
 	}
@@ -775,17 +760,14 @@ func (c *Client) sendPing() {
 		c.writeMux.Unlock()
 		return
 	}
-	err := c.conn.SetWriteDeadline(time.Now().Add(writeDeadline))
+	err := c.conn.WriteJSON(pingMsg)
-	if err == nil {
-		err = c.conn.WriteJSON(pingMsg)
-	}
 	if err == nil {
 		telemetry.IncWSMessage(c.metricsContext(), "out", "ping")
 		// Protocol-level ping: a standards-compliant server replies with a PONG,
 		// which refreshes the read deadline. This is what lets us notice a
 		// half-open connection where writes still succeed (buffered) but the
 		// peer is gone.
-		_ = c.conn.WriteControl(websocket.PingMessage, nil, time.Now().Add(writeDeadline))
+		_ = c.conn.WriteControl(websocket.PingMessage, nil, time.Now().Add(10*time.Second))
 	}
 	c.writeMux.Unlock()