fix(xpc): prevent startup hangs with proper timeout handling (#561)

Sources/ContainerXPC/XPCClient.swift

@@ -65,14 +65,17 @@ extension XPCClient {
     /// Send the provided message to the service.
     @discardableResult
     public func send(_ message: XPCMessage, responseTimeout: Duration? = nil) async throws -> XPCMessage {
-        try await withThrowingTaskGroup(of: XPCMessage.self, returning: XPCMessage.self) { group in
+        // Use a more robust timeout that cancels the XPC connection if it hangs
+        return try await withThrowingTaskGroup(of: XPCMessage.self, returning: XPCMessage.self) { group in
             if let responseTimeout {
                 group.addTask {
                     try await Task.sleep(for: responseTimeout)
+                    // Cancel the connection to unblock the XPC call
+                    xpc_connection_cancel(self.connection)
                     let route = message.string(key: XPCMessage.routeKey) ?? "nil"
                     throw ContainerizationError(
-                        .internalError,
-                        message: "XPC timeout for request to \(self.service)/\(route)"
+                        .timeout,
+                        message: "XPC timeout after \(Int(responseTimeout.components.seconds))s for \(self.service)/\(route). Is the apiserver running? Try: container system start"
                     )
                 }
             }

Sources/ContainerXPC/XPCServer.swift

@@ -52,7 +52,7 @@ public struct XPCServer: Sendable {
 
     public func listen() async throws {
         let connections = AsyncStream<xpc_connection_t> { cont in
-            lock.withLock {
+            self.lock.withLock {
                 xpc_connection_set_event_handler(self.connection) { object in
                     switch xpc_get_type(object) {
                     case XPC_TYPE_CONNECTION:

Sources/Helpers/APIServer/APIServer+Start.swift

@@ -240,12 +240,13 @@ extension APIServer {
                 log: log
             )
 
+            // Check for default network - don't block on creation during startup
+            // to avoid blocking if vmnet plugin fails. Users can create networks on-demand.
             let defaultNetwork = try await service.list()
                 .filter { $0.id == ClientNetwork.defaultNetworkName }
                 .first
             if defaultNetwork == nil {
-                let config = try NetworkConfiguration(id: ClientNetwork.defaultNetworkName, mode: .nat)
-                _ = try await service.create(configuration: config)
+                log.info("default network not present, can be created with 'container network create'")
             }
 
             let harness = NetworksHarness(service: service, log: log)

Sources/Services/ContainerAPIService/Networks/NetworksService.swift

@@ -95,7 +95,7 @@ public actor NetworksService {
                         "id": "\(configuration.id)",
                         "state": "\(networkState.state)",
                     ])
-                return
+                continue
             }
         }
     }

Sources/Services/ContainerNetworkService/NetworkClient.swift

@@ -37,11 +37,14 @@ public struct NetworkClient: Sendable {
 
 // Runtime Methods
 extension NetworkClient {
+    /// Default timeout for network XPC calls.
+    private static let xpcTimeout: Duration = .seconds(5)
+
     public func state() async throws -> NetworkState {
         let request = XPCMessage(route: NetworkRoutes.state.rawValue)
         let client = createClient()
 
-        let response = try await client.send(request)
+        let response = try await client.send(request, responseTimeout: Self.xpcTimeout)
         let state = try response.state()
         return state
     }
@@ -55,7 +58,7 @@ extension NetworkClient {
 
         let client = createClient()
 
-        let response = try await client.send(request)
+        let response = try await client.send(request, responseTimeout: Self.xpcTimeout)
         let attachment = try response.attachment()
         let additionalData = response.additionalData()
         return (attachment, additionalData)
@@ -66,7 +69,7 @@ extension NetworkClient {
         request.set(key: NetworkKeys.hostname.rawValue, value: hostname)
 
         let client = createClient()
-        try await client.send(request)
+        try await client.send(request, responseTimeout: Self.xpcTimeout)
     }
 
     public func lookup(hostname: String) async throws -> Attachment? {
@@ -75,7 +78,7 @@ extension NetworkClient {
 
         let client = createClient()
 
-        let response = try await client.send(request)
+        let response = try await client.send(request, responseTimeout: Self.xpcTimeout)
         return try response.dataNoCopy(key: NetworkKeys.attachment.rawValue).map {
             try JSONDecoder().decode(Attachment.self, from: $0)
         }
@@ -86,7 +89,7 @@ extension NetworkClient {
 
         let client = createClient()
 
-        let response = try await client.send(request)
+        let response = try await client.send(request, responseTimeout: Self.xpcTimeout)
         return try response.allocatorDisabled()
     }