working websocket implementation

2025-12-01 10:45:58 +01:00
parent 675e114278
commit 667db6be83
11 changed files with 386 additions and 23 deletions
--- a/ws/websocket.go
+++ b/ws/websocket.go
@@ -0,0 +1,186 @@
+package ws
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"homestead/homestead_gateway/util/config"
+	"log/slog"
+	"net"
+	"net/http"
+	"strings"
+	"time"
+
+	"github.com/gorilla/websocket"
+)
+
+func (m *GatewayMessageIn) Validate() error {
+	if strings.TrimSpace(m.MsgID) == "" {
+		return errors.New("msg_id missing")
+	}
+	if strings.TrimSpace(m.Server) == "" {
+		return errors.New("server missing")
+	}
+	if strings.TrimSpace(m.User.ID) == "" {
+		return errors.New("user.mod_uid missing")
+	}
+	if strings.TrimSpace(m.Content) == "" {
+		return errors.New("content missing")
+	}
+	return nil
+}
+
+func NewWsGateway(cfg config.GatewayConfig, logger *slog.Logger, closefn func() error) *WebsocketGateway {
+	return &WebsocketGateway{
+		logger:  logger,
+		closefn: closefn,
+		apiKey:  cfg.Websocket,
+		upgrader: websocket.Upgrader{
+			ReadBufferSize:  1024,
+			WriteBufferSize: 1024,
+			CheckOrigin: func(r *http.Request) bool {
+				return true // local by default; change for production
+			},
+		},
+		outgoingCh:    make(chan GatewayMessageOut, cfg.QueueSize),
+		modConns:      make(map[*websocket.Conn]struct{}),
+		bodySizeBytes: int64(cfg.BodySize) * 1024 * 1024,
+		port:          cfg.HttpPort,
+	}
+}
+
+// Serve starts the HTTP server and /ws endpoint and blocks until ctx cancelled or server fails.
+func (g *WebsocketGateway) Serve(ctx context.Context, listenAddr string) error {
+	mux := http.NewServeMux()
+	mux.HandleFunc("/ws", g.handleWS)
+	mux.HandleFunc("/health", func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(200)
+		_, _ = w.Write([]byte("ok"))
+	})
+
+	srv := &http.Server{
+		Addr:        listenAddr,
+		Handler:     loggingMiddleware(g.logger, mux),
+		BaseContext: func(l net.Listener) context.Context { return ctx },
+	}
+
+	errCh := make(chan error, 1)
+	go func() {
+		g.logger.Info("ws gateway listening", "addr", listenAddr)
+		if err := srv.ListenAndServe(); err != nil && err != http.ErrServerClosed {
+			errCh <- err
+		}
+		close(errCh)
+	}()
+
+	select {
+	case <-ctx.Done():
+		shutdownCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+		defer cancel()
+		g.logger.Info("shutting down http server")
+		_ = srv.Shutdown(shutdownCtx)
+		g.closeAll()
+		return nil
+	case err := <-errCh:
+		return err
+	}
+}
+
+func (g *WebsocketGateway) handleWS(w http.ResponseWriter, r *http.Request) {
+	if !g.validateApiKey(r) {
+		http.Error(w, "unauthorized", http.StatusUnauthorized)
+		g.logger.Warn("ws auth failed", "remote", r.RemoteAddr)
+		return
+	}
+
+	conn, err := g.upgrader.Upgrade(w, r, nil)
+	if err != nil {
+		g.logger.Error("ws upgrade error", err, "remote", r.RemoteAddr)
+		return
+	}
+
+	g.registerConn(conn)
+	g.logger.Info("ws connected", "remote", conn.RemoteAddr().String())
+
+	// Configure read limits & pong handler
+	if g.bodySizeBytes > 0 {
+		conn.SetReadLimit(g.bodySizeBytes)
+	} else {
+		conn.SetReadLimit(1 << 20) // sensible default 1MiB
+	}
+
+	_ = conn.SetReadDeadline(time.Now().Add(60 * time.Second))
+	conn.SetPongHandler(func(appData string) error {
+		_ = conn.SetReadDeadline(time.Now().Add(60 * time.Second))
+		return nil
+	})
+
+	go g.readLoop(conn)
+}
+
+func (g *WebsocketGateway) readLoop(c *websocket.Conn) {
+	defer func() {
+		g.unregisterConn(c)
+		_ = c.Close()
+		g.logger.Info("ws disconnected", "remote", c.RemoteAddr().String())
+	}()
+
+	pingTicker := time.NewTicker(30 * time.Second)
+	defer pingTicker.Stop()
+
+	for {
+		// Read one message (blocks until message arrives)
+		typ, data, err := c.ReadMessage()
+		if err != nil {
+			if websocket.IsUnexpectedCloseError(err, websocket.CloseGoingAway, websocket.CloseNormalClosure) {
+				g.logger.Warn("unexpected ws close", "err", err)
+			} else {
+				g.logger.Debug("ws read error", "err", err)
+			}
+			return
+		}
+		if typ != websocket.TextMessage && typ != websocket.BinaryMessage {
+			continue
+		}
+
+		var in GatewayMessageIn
+		if err := json.Unmarshal(data, &in); err != nil {
+			_ = writeJSONSafe(c, map[string]string{"error": "invalid json: " + err.Error()})
+			g.logger.Warn("invalid json from client", "remote", c.RemoteAddr().String(), "err", err)
+			continue
+		}
+		in.ReceivedAt = time.Now().UTC()
+		if err := in.Validate(); err != nil {
+			_ = writeJSONSafe(c, map[string]string{"error": err.Error()})
+			g.logger.Warn("message validation failed", "remote", c.RemoteAddr().String(), "err", err)
+			continue
+		}
+
+		out := GatewayMessageOut{
+			Type:        "message",
+			Payload:     in,
+			ForwardedAt: time.Now().UTC(),
+		}
+
+		// Non-blocking enqueue with backpressure
+		select {
+		case g.outgoingCh <- out:
+			_ = writeJSONSafe(c, map[string]string{"status": "queued"})
+			g.logger.Debug("enqueued message", "msg_id", in.MsgID, "server", in.Server)
+		default:
+			_ = writeJSONSafe(c, map[string]string{"error": "gateway busy"})
+			g.logger.Warn("outgoing queue full", "msg_id", in.MsgID, "remote", c.RemoteAddr().String())
+		}
+
+		// also handle pings periodically (so client sees ping frequently)
+		select {
+		case <-pingTicker.C:
+			_ = c.SetWriteDeadline(time.Now().Add(5 * time.Second))
+			if err := c.WriteMessage(websocket.PingMessage, nil); err != nil {
+				g.logger.Debug("write ping failed", "err", err)
+				return
+			}
+		default:
+		}
+	}
+}