Add support for Stack level options.

Linux controls socket send/receive buffers using a few sysctl variables
  - net.core.rmem_default
  - net.core.rmem_max
  - net.core.wmem_max
  - net.core.wmem_default
  - net.ipv4.tcp_rmem
  - net.ipv4.tcp_wmem

The first 4 control the default socket buffer sizes for all sockets
raw/packet/tcp/udp and also the maximum permitted socket buffer that can be
specified in setsockopt(SOL_SOCKET, SO_(RCV|SND)BUF,...).

The last two control the TCP auto-tuning limits and override the default
specified in rmem_default/wmem_default as well as the max limits.

Netstack today only implements tcp_rmem/tcp_wmem and incorrectly uses it
to limit the maximum size in setsockopt() as well as uses it for raw/udp
sockets.

This changelist introduces the other 4 and updates the udp/raw sockets to use
the newly introduced variables. The values for min/max match the current
tcp_rmem/wmem values and the default value buffers for UDP/RAW sockets is
updated to match the linux value of 212KiB up from the really low current value
of 32 KiB.

Updates #3043
Fixes #3043

PiperOrigin-RevId: 318089805

7 files changed, 69 insertions, 40 deletions

diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go
index 377643b82..9d4dce826 100644
--- a/pkg/tcpip/transport/tcp/connect.go
+++ b/pkg/tcpip/transport/tcp/connect.go
@@ -521,7 +521,7 @@ func (h *handshake) execute() *tcpip.Error {
 	s.AddWaker(&h.ep.newSegmentWaker, wakerForNewSegment)
 	defer s.Done()
 
-	var sackEnabled tcpip.StackSACKEnabled
+	var sackEnabled SACKEnabled
 	if err := h.ep.stack.TransportProtocolOption(ProtocolNumber, &sackEnabled); err != nil {
 		// If stack returned an error when checking for SACKEnabled
 		// status then just default to switching off SACK negotiation.
diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go
index 1e4c2f507..99a691815 100644
--- a/pkg/tcpip/transport/tcp/endpoint.go
+++ b/pkg/tcpip/transport/tcp/endpoint.go
@@ -847,12 +847,12 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
 		maxSynRetries: DefaultSynRetries,
 	}
 
-	var ss tcpip.StackSendBufferSizeOption
+	var ss SendBufferSizeOption
 	if err := s.TransportProtocolOption(ProtocolNumber, &ss); err == nil {
 		e.sndBufSize = ss.Default
 	}
 
-	var rs tcpip.StackReceiveBufferSizeOption
+	var rs ReceiveBufferSizeOption
 	if err := s.TransportProtocolOption(ProtocolNumber, &rs); err == nil {
 		e.rcvBufSize = rs.Default
 	}
@@ -867,7 +867,7 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
 		e.rcvAutoParams.disabled = !bool(mrb)
 	}
 
-	var de tcpip.StackDelayEnabled
+	var de DelayEnabled
 	if err := s.TransportProtocolOption(ProtocolNumber, &de); err == nil && de {
 		e.SetSockOptBool(tcpip.DelayOption, true)
 	}
@@ -1584,7 +1584,7 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
 	case tcpip.ReceiveBufferSizeOption:
 		// Make sure the receive buffer size is within the min and max
 		// allowed.
-		var rs tcpip.StackReceiveBufferSizeOption
+		var rs ReceiveBufferSizeOption
 		if err := e.stack.TransportProtocolOption(ProtocolNumber, &rs); err == nil {
 			if v < rs.Min {
 				v = rs.Min
@@ -1634,7 +1634,7 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
 	case tcpip.SendBufferSizeOption:
 		// Make sure the send buffer size is within the min and max
 		// allowed.
-		var ss tcpip.StackSendBufferSizeOption
+		var ss SendBufferSizeOption
 		if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err == nil {
 			if v < ss.Min {
 				v = ss.Min
@@ -1674,7 +1674,7 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
 				return tcpip.ErrInvalidOptionValue
 			}
 		}
-		var rs tcpip.StackReceiveBufferSizeOption
+		var rs ReceiveBufferSizeOption
 		if err := e.stack.TransportProtocolOption(ProtocolNumber, &rs); err == nil {
 			if v < rs.Min/2 {
 				v = rs.Min / 2
@@ -2595,7 +2595,7 @@ func (e *endpoint) receiveBufferSize() int {
 }
 
 func (e *endpoint) maxReceiveBufferSize() int {
-	var rs tcpip.StackReceiveBufferSizeOption
+	var rs ReceiveBufferSizeOption
 	if err := e.stack.TransportProtocolOption(ProtocolNumber, &rs); err != nil {
 		// As a fallback return the hardcoded max buffer size.
 		return MaxBufferSize
@@ -2676,7 +2676,7 @@ func timeStampOffset() uint32 {
 // if the SYN options indicate that the SACK option was negotiated and the TCP
 // stack is configured to enable TCP SACK option.
 func (e *endpoint) maybeEnableSACKPermitted(synOpts *header.TCPSynOptions) {
-	var v tcpip.StackSACKEnabled
+	var v SACKEnabled
 	if err := e.stack.TransportProtocolOption(ProtocolNumber, &v); err != nil {
 		// Stack doesn't support SACK. So just return.
 		return
diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go
index 8258c0ecc..abf1ac5c9 100644
--- a/pkg/tcpip/transport/tcp/endpoint_state.go
+++ b/pkg/tcpip/transport/tcp/endpoint_state.go
@@ -182,13 +182,17 @@ func (e *endpoint) Resume(s *stack.Stack) {
 	epState := e.origEndpointState
 	switch epState {
 	case StateInitial, StateBound, StateListen, StateConnecting, StateEstablished:
-		var ss tcpip.StackSendBufferSizeOption
+		var ss SendBufferSizeOption
 		if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err == nil {
 			if e.sndBufSize < ss.Min || e.sndBufSize > ss.Max {
 				panic(fmt.Sprintf("endpoint.sndBufSize %d is outside the min and max allowed [%d, %d]", e.sndBufSize, ss.Min, ss.Max))
 			}
-			if e.rcvBufSize < ss.Min || e.rcvBufSize > ss.Max {
-				panic(fmt.Sprintf("endpoint.rcvBufSize %d is outside the min and max allowed [%d, %d]", e.rcvBufSize, ss.Min, ss.Max))
+		}
+
+		var rs ReceiveBufferSizeOption
+		if err := e.stack.TransportProtocolOption(ProtocolNumber, &rs); err == nil {
+			if e.rcvBufSize < rs.Min || e.rcvBufSize > rs.Max {
+				panic(fmt.Sprintf("endpoint.rcvBufSize %d is outside the min and max allowed [%d, %d]", e.rcvBufSize, rs.Min, rs.Max))
 			}
 		}
 	}
diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go
index 3cff55afa..f2ae6ce50 100644
--- a/pkg/tcpip/transport/tcp/protocol.go
+++ b/pkg/tcpip/transport/tcp/protocol.go
@@ -76,6 +76,31 @@ const (
 	ccCubic = "cubic"
 )
 
+// SACKEnabled is used by stack.(*Stack).TransportProtocolOption to
+// enable/disable SACK support in TCP. See: https://tools.ietf.org/html/rfc2018.
+type SACKEnabled bool
+
+// DelayEnabled is used by stack.(Stack*).TransportProtocolOption to
+// enable/disable Nagle's algorithm in TCP.
+type DelayEnabled bool
+
+// SendBufferSizeOption is used by stack.(Stack*).TransportProtocolOption
+// to get/set the default, min and max TCP send buffer sizes.
+type SendBufferSizeOption struct {
+	Min     int
+	Default int
+	Max     int
+}
+
+// ReceiveBufferSizeOption is used by
+// stack.(Stack*).TransportProtocolOption to get/set the default, min and max
+// TCP receive buffer sizes.
+type ReceiveBufferSizeOption struct {
+	Min     int
+	Default int
+	Max     int
+}
+
 // syncRcvdCounter tracks the number of endpoints in the SYN-RCVD state. The
 // value is protected by a mutex so that we can increment only when it's
 // guaranteed not to go above a threshold.
@@ -137,8 +162,8 @@ type protocol struct {
 	mu                         sync.RWMutex
 	sackEnabled                bool
 	delayEnabled               bool
-	sendBufferSize             tcpip.StackSendBufferSizeOption
-	recvBufferSize             tcpip.StackReceiveBufferSizeOption
+	sendBufferSize             SendBufferSizeOption
+	recvBufferSize             ReceiveBufferSizeOption
 	congestionControl          string
 	availableCongestionControl []string
 	moderateReceiveBuffer      bool
@@ -249,19 +274,19 @@ func replyWithReset(s *segment, tos, ttl uint8) {
 // SetOption implements stack.TransportProtocol.SetOption.
 func (p *protocol) SetOption(option interface{}) *tcpip.Error {
 	switch v := option.(type) {
-	case tcpip.StackSACKEnabled:
+	case SACKEnabled:
 		p.mu.Lock()
 		p.sackEnabled = bool(v)
 		p.mu.Unlock()
 		return nil
 
-	case tcpip.StackDelayEnabled:
+	case DelayEnabled:
 		p.mu.Lock()
 		p.delayEnabled = bool(v)
 		p.mu.Unlock()
 		return nil
 
-	case tcpip.StackSendBufferSizeOption:
+	case SendBufferSizeOption:
 		if v.Min <= 0 || v.Default < v.Min || v.Default > v.Max {
 			return tcpip.ErrInvalidOptionValue
 		}
@@ -270,7 +295,7 @@ func (p *protocol) SetOption(option interface{}) *tcpip.Error {
 		p.mu.Unlock()
 		return nil
 
-	case tcpip.StackReceiveBufferSizeOption:
+	case ReceiveBufferSizeOption:
 		if v.Min <= 0 || v.Default < v.Min || v.Default > v.Max {
 			return tcpip.ErrInvalidOptionValue
 		}
@@ -363,25 +388,25 @@ func (p *protocol) SetOption(option interface{}) *tcpip.Error {
 // Option implements stack.TransportProtocol.Option.
 func (p *protocol) Option(option interface{}) *tcpip.Error {
 	switch v := option.(type) {
-	case *tcpip.StackSACKEnabled:
+	case *SACKEnabled:
 		p.mu.RLock()
-		*v = tcpip.StackSACKEnabled(p.sackEnabled)
+		*v = SACKEnabled(p.sackEnabled)
 		p.mu.RUnlock()
 		return nil
 
-	case *tcpip.StackDelayEnabled:
+	case *DelayEnabled:
 		p.mu.RLock()
-		*v = tcpip.StackDelayEnabled(p.delayEnabled)
+		*v = DelayEnabled(p.delayEnabled)
 		p.mu.RUnlock()
 		return nil
 
-	case *tcpip.StackSendBufferSizeOption:
+	case *SendBufferSizeOption:
 		p.mu.RLock()
 		*v = p.sendBufferSize
 		p.mu.RUnlock()
 		return nil
 
-	case *tcpip.StackReceiveBufferSizeOption:
+	case *ReceiveBufferSizeOption:
 		p.mu.RLock()
 		*v = p.recvBufferSize
 		p.mu.RUnlock()
@@ -491,12 +516,12 @@ func (*protocol) Parse(pkt *stack.PacketBuffer) bool {
 // NewProtocol returns a TCP transport protocol.
 func NewProtocol() stack.TransportProtocol {
 	return &protocol{
-		sendBufferSize: tcpip.StackSendBufferSizeOption{
+		sendBufferSize: SendBufferSizeOption{
 			Min:     MinBufferSize,
 			Default: DefaultSendBufferSize,
 			Max:     MaxBufferSize,
 		},
-		recvBufferSize: tcpip.StackReceiveBufferSizeOption{
+		recvBufferSize: ReceiveBufferSizeOption{
 			Min:     MinBufferSize,
 			Default: DefaultReceiveBufferSize,
 			Max:     MaxBufferSize,
diff --git a/pkg/tcpip/transport/tcp/tcp_sack_test.go b/pkg/tcpip/transport/tcp/tcp_sack_test.go
index 812e503bc..99521f0c1 100644
--- a/pkg/tcpip/transport/tcp/tcp_sack_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_sack_test.go
@@ -46,8 +46,8 @@ func createConnectedWithSACKAndTS(c *context.Context) *context.RawEndpoint {
 
 func setStackSACKPermitted(t *testing.T, c *context.Context, enable bool) {
 	t.Helper()
-	if err := c.Stack().SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackSACKEnabled(enable)); err != nil {
-		t.Fatalf("c.s.SetTransportProtocolOption(tcp.ProtocolNumber, StackSACKEnabled(%t) = %s", enable, err)
+	if err := c.Stack().SetTransportProtocolOption(tcp.ProtocolNumber, tcp.SACKEnabled(enable)); err != nil {
+		t.Fatalf("c.s.SetTransportProtocolOption(tcp.ProtocolNumber, SACKEnabled(%t) = %s", enable, err)
 	}
 }
 
diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go
index 2632a3c67..169adb16b 100644
--- a/pkg/tcpip/transport/tcp/tcp_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_test.go
@@ -4005,7 +4005,7 @@ func TestDefaultBufferSizes(t *testing.T) {
 	checkRecvBufferSize(t, ep, tcp.DefaultReceiveBufferSize)
 
 	// Change the default send buffer size.
-	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackSendBufferSizeOption{
+	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.SendBufferSizeOption{
 		Min:     1,
 		Default: tcp.DefaultSendBufferSize * 2,
 		Max:     tcp.DefaultSendBufferSize * 20}); err != nil {
@@ -4022,7 +4022,7 @@ func TestDefaultBufferSizes(t *testing.T) {
 	checkRecvBufferSize(t, ep, tcp.DefaultReceiveBufferSize)
 
 	// Change the default receive buffer size.
-	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackReceiveBufferSizeOption{
+	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.ReceiveBufferSizeOption{
 		Min:     1,
 		Default: tcp.DefaultReceiveBufferSize * 3,
 		Max:     tcp.DefaultReceiveBufferSize * 30}); err != nil {
@@ -4053,11 +4053,11 @@ func TestMinMaxBufferSizes(t *testing.T) {
 	defer ep.Close()
 
 	// Change the min/max values for send/receive
-	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackReceiveBufferSizeOption{Min: 200, Default: tcp.DefaultReceiveBufferSize * 2, Max: tcp.DefaultReceiveBufferSize * 20}); err != nil {
+	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.ReceiveBufferSizeOption{Min: 200, Default: tcp.DefaultReceiveBufferSize * 2, Max: tcp.DefaultReceiveBufferSize * 20}); err != nil {
 		t.Fatalf("SetTransportProtocolOption failed: %s", err)
 	}
 
-	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackSendBufferSizeOption{Min: 300, Default: tcp.DefaultSendBufferSize * 3, Max: tcp.DefaultSendBufferSize * 30}); err != nil {
+	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.SendBufferSizeOption{Min: 300, Default: tcp.DefaultSendBufferSize * 3, Max: tcp.DefaultSendBufferSize * 30}); err != nil {
 		t.Fatalf("SetTransportProtocolOption failed: %s", err)
 	}
 
@@ -5696,7 +5696,7 @@ func TestReceiveBufferAutoTuningApplicationLimited(t *testing.T) {
 	// the segment queue holding unprocessed packets is limited to 500.
 	const receiveBufferSize = 80 << 10 // 80KB.
 	const maxReceiveBufferSize = receiveBufferSize * 10
-	if err := stk.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackReceiveBufferSizeOption{Min: 1, Default: receiveBufferSize, Max: maxReceiveBufferSize}); err != nil {
+	if err := stk.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.ReceiveBufferSizeOption{Min: 1, Default: receiveBufferSize, Max: maxReceiveBufferSize}); err != nil {
 		t.Fatalf("SetTransportProtocolOption failed: %s", err)
 	}
 
@@ -5817,7 +5817,7 @@ func TestReceiveBufferAutoTuning(t *testing.T) {
 	// the segment queue holding unprocessed packets is limited to 300.
 	const receiveBufferSize = 80 << 10 // 80KB.
 	const maxReceiveBufferSize = receiveBufferSize * 10
-	if err := stk.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackReceiveBufferSizeOption{Min: 1, Default: receiveBufferSize, Max: maxReceiveBufferSize}); err != nil {
+	if err := stk.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.ReceiveBufferSizeOption{Min: 1, Default: receiveBufferSize, Max: maxReceiveBufferSize}); err != nil {
 		t.Fatalf("SetTransportProtocolOption failed: %s", err)
 	}
 
@@ -5959,7 +5959,7 @@ func TestDelayEnabled(t *testing.T) {
 	checkDelayOption(t, c, false, false) // Delay is disabled by default.
 
 	for _, v := range []struct {
-		delayEnabled    tcpip.StackDelayEnabled
+		delayEnabled    tcp.DelayEnabled
 		wantDelayOption bool
 	}{
 		{delayEnabled: false, wantDelayOption: false},
@@ -5974,10 +5974,10 @@ func TestDelayEnabled(t *testing.T) {
 	}
 }
 
-func checkDelayOption(t *testing.T, c *context.Context, wantDelayEnabled tcpip.StackDelayEnabled, wantDelayOption bool) {
+func checkDelayOption(t *testing.T, c *context.Context, wantDelayEnabled tcp.DelayEnabled, wantDelayOption bool) {
 	t.Helper()
 
-	var gotDelayEnabled tcpip.StackDelayEnabled
+	var gotDelayEnabled tcp.DelayEnabled
 	if err := c.Stack().TransportProtocolOption(tcp.ProtocolNumber, &gotDelayEnabled); err != nil {
 		t.Fatalf("TransportProtocolOption(tcp, &gotDelayEnabled) failed: %s", err)
 	}
diff --git a/pkg/tcpip/transport/tcp/testing/context/context.go b/pkg/tcpip/transport/tcp/testing/context/context.go
index 9e262c272..06fde2a79 100644
--- a/pkg/tcpip/transport/tcp/testing/context/context.go
+++ b/pkg/tcpip/transport/tcp/testing/context/context.go
@@ -144,11 +144,11 @@ func New(t *testing.T, mtu uint32) *Context {
 	})
 
 	// Allow minimum send/receive buffer sizes to be 1 during tests.
-	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackSendBufferSizeOption{Min: 1, Default: tcp.DefaultSendBufferSize, Max: 10 * tcp.DefaultSendBufferSize}); err != nil {
+	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.SendBufferSizeOption{Min: 1, Default: tcp.DefaultSendBufferSize, Max: 10 * tcp.DefaultSendBufferSize}); err != nil {
 		t.Fatalf("SetTransportProtocolOption failed: %s", err)
 	}
 
-	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.StackReceiveBufferSizeOption{Min: 1, Default: tcp.DefaultReceiveBufferSize, Max: 10 * tcp.DefaultReceiveBufferSize}); err != nil {
+	if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.ReceiveBufferSizeOption{Min: 1, Default: tcp.DefaultReceiveBufferSize, Max: 10 * tcp.DefaultReceiveBufferSize}); err != nil {
 		t.Fatalf("SetTransportProtocolOption failed: %s", err)
 	}
 
@@ -1091,7 +1091,7 @@ func (c *Context) PassiveConnectWithOptions(maxPayload, wndScale int, synOptions
 // SACKEnabled returns true if the TCP Protocol option SACKEnabled is set to true
 // for the Stack in the context.
 func (c *Context) SACKEnabled() bool {
-	var v tcpip.StackSACKEnabled
+	var v tcp.SACKEnabled
 	if err := c.Stack().TransportProtocolOption(tcp.ProtocolNumber, &v); err != nil {
 		// Stack doesn't support SACK. So just return.
 		return false