cmd/derper,derp: make TCP write timeout configurable

The timeout still defaults to 2 seconds, but can now be changed via command-line flag.

Updates tailscale/corp#26045

Signed-off-by: Percy Wegmann <percy@tailscale.com>

cmd/derper/derper.go

@@ -77,6 +77,8 @@ var (
 	tcpKeepAlive = flag.Duration("tcp-keepalive-time", 10*time.Minute, "TCP keepalive time")
 	// tcpUserTimeout is intentionally short, so that hung connections are cleaned up promptly. DERPs should be nearby users.
 	tcpUserTimeout = flag.Duration("tcp-user-timeout", 15*time.Second, "TCP user timeout")
+	// tcpWriteTimeout is the timeout for writing to client TCP connections. It does not apply to mesh connections.
+	tcpWriteTimeout = flag.Duration("tcp-write-timeout", derp.DefaultTCPWiteTimeout, "TCP write timeout; 0 results in no timeout being set on writes")
 )
 
 var (
@@ -173,6 +175,7 @@ func main() {
 	s.SetVerifyClient(*verifyClients)
 	s.SetVerifyClientURL(*verifyClientURL)
 	s.SetVerifyClientURLFailOpen(*verifyFailOpen)
+	s.SetTCPWriteTimeout(*tcpWriteTimeout)
 
 	if *meshPSKFile != "" {
 		b, err := os.ReadFile(*meshPSKFile)

derp/derp_server.go

@@ -84,7 +84,7 @@ func init() {
 
 const (
 	defaultPerClientSendQueueDepth = 32 // default packets buffered for sending
-	writeTimeout                   = 2 * time.Second
+	DefaultTCPWiteTimeout          = 2 * time.Second
 	privilegedWriteTimeout         = 30 * time.Second // for clients with the mesh key
 )
 
@@ -201,6 +201,8 @@ type Server struct {
 	// Sets the client send queue depth for the server.
 	perClientSendQueueDepth int
 
+	tcpWriteTimeout time.Duration
+
 	clock tstime.Clock
 }
 
@@ -377,6 +379,7 @@ func NewServer(privateKey key.NodePrivate, logf logger.Logf) *Server {
 		bufferedWriteFrames: metrics.NewHistogram([]float64{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 50, 100}),
 		keyOfAddr:           map[netip.AddrPort]key.NodePublic{},
 		clock:               tstime.StdClock{},
+		tcpWriteTimeout:     DefaultTCPWiteTimeout,
 	}
 	s.initMetacert()
 	s.packetsRecvDisco = s.packetsRecvByKind.Get(string(packetKindDisco))
@@ -481,6 +484,13 @@ func (s *Server) SetVerifyClientURLFailOpen(v bool) {
 	s.verifyClientsURLFailOpen = v
 }
 
+// SetTCPWriteTimeout sets the timeout for writing to connected clients.
+// This timeout does not apply to mesh connections.
+// Defaults to 2 seconds.
+func (s *Server) SetTCPWriteTimeout(d time.Duration) {
+	s.tcpWriteTimeout = d
+}
+
 // HasMeshKey reports whether the server is configured with a mesh key.
 func (s *Server) HasMeshKey() bool { return s.meshKey != "" }
 
@@ -1805,7 +1815,7 @@ func (c *sclient) sendLoop(ctx context.Context) error {
 }
 
 func (c *sclient) setWriteDeadline() {
-	d := writeTimeout
+	d := c.s.tcpWriteTimeout
 	if c.canMesh {
 		// Trusted peers get more tolerance.
 		//
@@ -1817,7 +1827,10 @@ func (c *sclient) setWriteDeadline() {
 		// of connected peers.
 		d = privilegedWriteTimeout
 	}
-	c.nc.SetWriteDeadline(time.Now().Add(d))
+	// Ignore the error from setting the write deadline. In practice,
+	// setting the deadline will only fail if the connection is closed
+	// or closing, so the subsequent Write() will fail anyway.
+	_ = c.nc.SetWriteDeadline(time.Now().Add(d))
 }
 
 // sendKeepAlive sends a keep-alive frame, without flushing.