diff --git a/server.go b/server.go
index bcf30cb9e3..4e44122454 100644
--- a/server.go
+++ b/server.go
@@ -1887,6 +1887,8 @@ func (s *Server) Shutdown() error {
 //
 // ShutdownWithContext does not close keepalive connections so it's recommended to set ReadTimeout and IdleTimeout
 // to something else than 0.
+//
+// When ShutdownWithContext returns errors, any operation to the Server is unavailable.
 func (s *Server) ShutdownWithContext(ctx context.Context) (err error) {
 	s.mu.Lock()
 	defer s.mu.Unlock()
@@ -1898,11 +1900,7 @@ func (s *Server) ShutdownWithContext(ctx context.Context) (err error) {
 		return nil
 	}
 
-	for _, ln := range s.ln {
-		if err = ln.Close(); err != nil {
-			return err
-		}
-	}
+	lnerr := s.closeListenersLocked()
 
 	if s.done != nil {
 		close(s.done)
@@ -1913,28 +1911,25 @@ func (s *Server) ShutdownWithContext(ctx context.Context) (err error) {
 	// Now we just have to wait until all workers are done or timeout.
 	ticker := time.NewTicker(time.Millisecond * 100)
 	defer ticker.Stop()
-END:
+
 	for {
 		s.closeIdleConns()
 
 		if open := atomic.LoadInt32(&s.open); open == 0 {
-			break
+			// There may be a pending request to call ctx.Done(). Therefore, we only set it to nil when open == 0.
+			s.done = nil
+			return lnerr
 		}
 		// This is not an optimal solution but using a sync.WaitGroup
 		// here causes data races as it's hard to prevent Add() to be called
 		// while Wait() is waiting.
 		select {
 		case <-ctx.Done():
-			err = ctx.Err()
-			break END
+			return ctx.Err()
 		case <-ticker.C:
 			continue
 		}
 	}
-
-	s.done = nil
-	s.ln = nil
-	return err
 }
 
 func acceptConn(s *Server, ln net.Listener, lastPerIPErrorTime *time.Time) (net.Conn, error) {
@@ -2749,15 +2744,7 @@ func (ctx *RequestCtx) Deadline() (deadline time.Time, ok bool) {
 // Note: Because creating a new channel for every request is just too expensive, so
 // RequestCtx.s.done is only closed when the server is shutting down.
 func (ctx *RequestCtx) Done() <-chan struct{} {
-	// fix use new variables to prevent panic caused by modifying the original done chan to nil.
-	done := ctx.s.done
-
-	if done == nil {
-		done = make(chan struct{}, 1)
-		done <- struct{}{}
-		return done
-	}
-	return done
+	return ctx.s.done
 }
 
 // Err returns a non-nil error value after Done is closed,
@@ -2934,6 +2921,17 @@ func (s *Server) closeIdleConns() {
 	s.idleConnsMu.Unlock()
 }
 
+func (s *Server) closeListenersLocked() error {
+	var err error
+	for _, ln := range s.ln {
+		if cerr := ln.Close(); cerr != nil && err == nil {
+			err = cerr
+		}
+	}
+	s.ln = nil
+	return err
+}
+
 // A ConnState represents the state of a client connection to a server.
 // It's used by the optional Server.ConnState hook.
 type ConnState int
diff --git a/server_race_test.go b/server_race_test.go
new file mode 100644
index 0000000000..a409cd804e
--- /dev/null
+++ b/server_race_test.go
@@ -0,0 +1,46 @@
+//go:build race
+
+package fasthttp
+
+import (
+	"context"
+	"github.com/valyala/fasthttp/fasthttputil"
+	"math"
+	"testing"
+)
+
+func TestServerDoneRace(t *testing.T) {
+	t.Parallel()
+
+	s := &Server{
+		Handler: func(ctx *RequestCtx) {
+			for i := 0; i < math.MaxInt; i++ {
+				ctx.Done()
+			}
+		},
+	}
+
+	ln := fasthttputil.NewInmemoryListener()
+	defer ln.Close()
+
+	go func() {
+		if err := s.Serve(ln); err != nil {
+			t.Errorf("unexpected error: %v", err)
+		}
+	}()
+
+	c, err := ln.Dial()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	defer c.Close()
+	if _, err = c.Write([]byte("POST / HTTP/1.1\r\nHost: go.dev\r\nContent-Length: 3\r\n\r\nABC" +
+		"\r\n\r\n" + // <-- this stuff is bogus, but we'll ignore it
+		"GET / HTTP/1.1\r\nHost: go.dev\r\n\r\n")); err != nil {
+		t.Fatal(err)
+	}
+	ctx, cancelFunc := context.WithCancel(context.Background())
+	cancelFunc()
+
+	s.ShutdownWithContext(ctx)
+}