Merge release-20210112.0-90-gdaf0d3f6c (automated)

3 files changed, 83 insertions, 90 deletions

diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go
index 05c431e83..8e43fec81 100644
--- a/pkg/tcpip/transport/tcp/endpoint.go
+++ b/pkg/tcpip/transport/tcp/endpoint.go
@@ -557,7 +557,6 @@ type endpoint struct {
 	// When the send side is closed, the protocol goroutine is notified via
 	// sndCloseWaker, and sndClosed is set to true.
 	sndBufMu      sync.Mutex `state:"nosave"`
-	sndBufSize    int
 	sndBufUsed    int
 	sndClosed     bool
 	sndBufInQueue seqnum.Size
@@ -869,7 +868,6 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
 		waiterQueue: waiterQueue,
 		state:       StateInitial,
 		rcvBufSize:  DefaultReceiveBufferSize,
-		sndBufSize:  DefaultSendBufferSize,
 		sndMTU:      int(math.MaxInt32),
 		keepalive: keepalive{
 			// Linux defaults.
@@ -882,13 +880,14 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
 		windowClamp:   DefaultReceiveBufferSize,
 		maxSynRetries: DefaultSynRetries,
 	}
-	e.ops.InitHandler(e)
+	e.ops.InitHandler(e, e.stack)
 	e.ops.SetMulticastLoop(true)
 	e.ops.SetQuickAck(true)
+	e.ops.SetSendBufferSize(DefaultSendBufferSize, false /* notify */, GetTCPSendBufferLimits)
 
 	var ss tcpip.TCPSendBufferSizeRangeOption
 	if err := s.TransportProtocolOption(ProtocolNumber, &ss); err == nil {
-		e.sndBufSize = ss.Default
+		e.ops.SetSendBufferSize(int64(ss.Default), false /* notify */, GetTCPSendBufferLimits)
 	}
 
 	var rs tcpip.TCPReceiveBufferSizeRangeOption
@@ -967,7 +966,8 @@ func (e *endpoint) Readiness(mask waiter.EventMask) waiter.EventMask {
 		// Determine if the endpoint is writable if requested.
 		if (mask & waiter.EventOut) != 0 {
 			e.sndBufMu.Lock()
-			if e.sndClosed || e.sndBufUsed < e.sndBufSize {
+			sndBufSize := e.getSendBufferSize()
+			if e.sndClosed || e.sndBufUsed < sndBufSize {
 				result |= waiter.EventOut
 			}
 			e.sndBufMu.Unlock()
@@ -1499,7 +1499,8 @@ func (e *endpoint) isEndpointWritableLocked() (int, *tcpip.Error) {
 		return 0, tcpip.ErrClosedForSend
 	}
 
-	avail := e.sndBufSize - e.sndBufUsed
+	sndBufSize := e.getSendBufferSize()
+	avail := sndBufSize - e.sndBufUsed
 	if avail <= 0 {
 		return 0, tcpip.ErrWouldBlock
 	}
@@ -1692,6 +1693,14 @@ func (e *endpoint) OnCorkOptionSet(v bool) {
 	}
 }
 
+func (e *endpoint) getSendBufferSize() int {
+	sndBufSize, err := e.ops.GetSendBufferSize()
+	if err != nil {
+		panic(fmt.Sprintf("e.ops.GetSendBufferSize() = %s", err))
+	}
+	return int(sndBufSize)
+}
+
 // SetSockOptInt sets a socket option.
 func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
 	// Lower 2 bits represents ECN bits. RFC 3168, section 23.1
@@ -1785,31 +1794,6 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
 		e.rcvListMu.Unlock()
 		e.UnlockUser()
 
-	case tcpip.SendBufferSizeOption:
-		// Make sure the send buffer size is within the min and max
-		// allowed.
-		var ss tcpip.TCPSendBufferSizeRangeOption
-		if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err != nil {
-			panic(fmt.Sprintf("e.stack.TransportProtocolOption(%d, %#v) = %s", ProtocolNumber, &ss, err))
-		}
-
-		if v > ss.Max {
-			v = ss.Max
-		}
-
-		if v < math.MaxInt32/SegOverheadFactor {
-			v *= SegOverheadFactor
-			if v < ss.Min {
-				v = ss.Min
-			}
-		} else {
-			v = math.MaxInt32
-		}
-
-		e.sndBufMu.Lock()
-		e.sndBufSize = v
-		e.sndBufMu.Unlock()
-
 	case tcpip.TTLOption:
 		e.LockUser()
 		e.ttl = uint8(v)
@@ -1995,12 +1979,6 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
 	case tcpip.ReceiveQueueSizeOption:
 		return e.readyReceiveSize()
 
-	case tcpip.SendBufferSizeOption:
-		e.sndBufMu.Lock()
-		v := e.sndBufSize
-		e.sndBufMu.Unlock()
-		return v, nil
-
 	case tcpip.ReceiveBufferSizeOption:
 		e.rcvListMu.Lock()
 		v := e.rcvBufSize
@@ -2749,13 +2727,14 @@ func (e *endpoint) HandleControlPacket(typ stack.ControlType, extra uint32, pkt
 // updateSndBufferUsage is called by the protocol goroutine when room opens up
 // in the send buffer. The number of newly available bytes is v.
 func (e *endpoint) updateSndBufferUsage(v int) {
+	sendBufferSize := e.getSendBufferSize()
 	e.sndBufMu.Lock()
-	notify := e.sndBufUsed >= e.sndBufSize>>1
+	notify := e.sndBufUsed >= sendBufferSize>>1
 	e.sndBufUsed -= v
-	// We only notify when there is half the sndBufSize available after
+	// We only notify when there is half the sendBufferSize available after
 	// a full buffer event occurs. This ensures that we don't wake up
 	// writers to queue just 1-2 segments and go back to sleep.
-	notify = notify && e.sndBufUsed < e.sndBufSize>>1
+	notify = notify && e.sndBufUsed < int(sendBufferSize)>>1
 	e.sndBufMu.Unlock()
 
 	if notify {
@@ -2967,8 +2946,9 @@ func (e *endpoint) completeState() stack.TCPEndpointState {
 	s.SACK.ReceivedBlocks, s.SACK.MaxSACKED = e.scoreboard.Copy()
 
 	// Copy endpoint send state.
+	sndBufSize := e.getSendBufferSize()
 	e.sndBufMu.Lock()
-	s.SndBufSize = e.sndBufSize
+	s.SndBufSize = sndBufSize
 	s.SndBufUsed = e.sndBufUsed
 	s.SndClosed = e.sndClosed
 	s.SndBufInQueue = e.sndBufInQueue
@@ -3113,3 +3093,17 @@ func (e *endpoint) Wait() {
 func (e *endpoint) SocketOptions() *tcpip.SocketOptions {
 	return &e.ops
 }
+
+// GetTCPSendBufferLimits is used to get send buffer size limits for TCP.
+func GetTCPSendBufferLimits(s tcpip.StackHandler) tcpip.SendBufferSizeOption {
+	var ss tcpip.TCPSendBufferSizeRangeOption
+	if err := s.TransportProtocolOption(header.TCPProtocolNumber, &ss); err != nil {
+		panic(fmt.Sprintf("s.TransportProtocolOption(%d, %#v) = %s", header.TCPProtocolNumber, ss, err))
+	}
+
+	return tcpip.SendBufferSizeOption{
+		Min:     ss.Min,
+		Default: ss.Default,
+		Max:     ss.Max,
+	}
+}
diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go
index ba67176b5..19c1dc67a 100644
--- a/pkg/tcpip/transport/tcp/endpoint_state.go
+++ b/pkg/tcpip/transport/tcp/endpoint_state.go
@@ -179,14 +179,16 @@ func (e *endpoint) afterLoad() {
 // Resume implements tcpip.ResumableEndpoint.Resume.
 func (e *endpoint) Resume(s *stack.Stack) {
 	e.stack = s
+	e.ops.InitHandler(e, e.stack)
 	e.segmentQueue.thaw()
 	epState := e.origEndpointState
 	switch epState {
 	case StateInitial, StateBound, StateListen, StateConnecting, StateEstablished:
 		var ss tcpip.TCPSendBufferSizeRangeOption
 		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))
+			sendBufferSize := e.getSendBufferSize()
+			if sendBufferSize < ss.Min || sendBufferSize > ss.Max {
+				panic(fmt.Sprintf("endpoint sendBufferSize %d is outside the min and max allowed [%d, %d]", sendBufferSize, ss.Min, ss.Max))
 			}
 		}
 
diff --git a/pkg/tcpip/transport/tcp/tcp_state_autogen.go b/pkg/tcpip/transport/tcp/tcp_state_autogen.go
index 6131d151a..c9926f8ca 100644
--- a/pkg/tcpip/transport/tcp/tcp_state_autogen.go
+++ b/pkg/tcpip/transport/tcp/tcp_state_autogen.go
@@ -195,7 +195,6 @@ func (e *endpoint) StateFields() []string {
 		"userMSS",
 		"maxSynRetries",
 		"windowClamp",
-		"sndBufSize",
 		"sndBufUsed",
 		"sndClosed",
 		"sndBufInQueue",
@@ -232,7 +231,7 @@ func (e *endpoint) StateSave(stateSinkObject state.Sink) {
 	var recentTSTimeValue unixTime = e.saveRecentTSTime()
 	stateSinkObject.SaveValue(26, recentTSTimeValue)
 	var acceptedChanValue []*endpoint = e.saveAcceptedChan()
-	stateSinkObject.SaveValue(50, acceptedChanValue)
+	stateSinkObject.SaveValue(49, acceptedChanValue)
 	stateSinkObject.Save(0, &e.EndpointInfo)
 	stateSinkObject.Save(1, &e.DefaultSocketOptionsHandler)
 	stateSinkObject.Save(2, &e.waiterQueue)
@@ -268,28 +267,27 @@ func (e *endpoint) StateSave(stateSinkObject state.Sink) {
 	stateSinkObject.Save(36, &e.userMSS)
 	stateSinkObject.Save(37, &e.maxSynRetries)
 	stateSinkObject.Save(38, &e.windowClamp)
-	stateSinkObject.Save(39, &e.sndBufSize)
-	stateSinkObject.Save(40, &e.sndBufUsed)
-	stateSinkObject.Save(41, &e.sndClosed)
-	stateSinkObject.Save(42, &e.sndBufInQueue)
-	stateSinkObject.Save(43, &e.sndQueue)
-	stateSinkObject.Save(44, &e.cc)
-	stateSinkObject.Save(45, &e.packetTooBigCount)
-	stateSinkObject.Save(46, &e.sndMTU)
-	stateSinkObject.Save(47, &e.keepalive)
-	stateSinkObject.Save(48, &e.userTimeout)
-	stateSinkObject.Save(49, &e.deferAccept)
-	stateSinkObject.Save(51, &e.rcv)
-	stateSinkObject.Save(52, &e.snd)
-	stateSinkObject.Save(53, &e.connectingAddress)
-	stateSinkObject.Save(54, &e.amss)
-	stateSinkObject.Save(55, &e.sendTOS)
-	stateSinkObject.Save(56, &e.gso)
-	stateSinkObject.Save(57, &e.tcpLingerTimeout)
-	stateSinkObject.Save(58, &e.closed)
-	stateSinkObject.Save(59, &e.txHash)
-	stateSinkObject.Save(60, &e.owner)
-	stateSinkObject.Save(61, &e.ops)
+	stateSinkObject.Save(39, &e.sndBufUsed)
+	stateSinkObject.Save(40, &e.sndClosed)
+	stateSinkObject.Save(41, &e.sndBufInQueue)
+	stateSinkObject.Save(42, &e.sndQueue)
+	stateSinkObject.Save(43, &e.cc)
+	stateSinkObject.Save(44, &e.packetTooBigCount)
+	stateSinkObject.Save(45, &e.sndMTU)
+	stateSinkObject.Save(46, &e.keepalive)
+	stateSinkObject.Save(47, &e.userTimeout)
+	stateSinkObject.Save(48, &e.deferAccept)
+	stateSinkObject.Save(50, &e.rcv)
+	stateSinkObject.Save(51, &e.snd)
+	stateSinkObject.Save(52, &e.connectingAddress)
+	stateSinkObject.Save(53, &e.amss)
+	stateSinkObject.Save(54, &e.sendTOS)
+	stateSinkObject.Save(55, &e.gso)
+	stateSinkObject.Save(56, &e.tcpLingerTimeout)
+	stateSinkObject.Save(57, &e.closed)
+	stateSinkObject.Save(58, &e.txHash)
+	stateSinkObject.Save(59, &e.owner)
+	stateSinkObject.Save(60, &e.ops)
 }
 
 func (e *endpoint) StateLoad(stateSourceObject state.Source) {
@@ -328,33 +326,32 @@ func (e *endpoint) StateLoad(stateSourceObject state.Source) {
 	stateSourceObject.Load(36, &e.userMSS)
 	stateSourceObject.Load(37, &e.maxSynRetries)
 	stateSourceObject.Load(38, &e.windowClamp)
-	stateSourceObject.Load(39, &e.sndBufSize)
-	stateSourceObject.Load(40, &e.sndBufUsed)
-	stateSourceObject.Load(41, &e.sndClosed)
-	stateSourceObject.Load(42, &e.sndBufInQueue)
-	stateSourceObject.LoadWait(43, &e.sndQueue)
-	stateSourceObject.Load(44, &e.cc)
-	stateSourceObject.Load(45, &e.packetTooBigCount)
-	stateSourceObject.Load(46, &e.sndMTU)
-	stateSourceObject.Load(47, &e.keepalive)
-	stateSourceObject.Load(48, &e.userTimeout)
-	stateSourceObject.Load(49, &e.deferAccept)
-	stateSourceObject.LoadWait(51, &e.rcv)
-	stateSourceObject.LoadWait(52, &e.snd)
-	stateSourceObject.Load(53, &e.connectingAddress)
-	stateSourceObject.Load(54, &e.amss)
-	stateSourceObject.Load(55, &e.sendTOS)
-	stateSourceObject.Load(56, &e.gso)
-	stateSourceObject.Load(57, &e.tcpLingerTimeout)
-	stateSourceObject.Load(58, &e.closed)
-	stateSourceObject.Load(59, &e.txHash)
-	stateSourceObject.Load(60, &e.owner)
-	stateSourceObject.Load(61, &e.ops)
+	stateSourceObject.Load(39, &e.sndBufUsed)
+	stateSourceObject.Load(40, &e.sndClosed)
+	stateSourceObject.Load(41, &e.sndBufInQueue)
+	stateSourceObject.LoadWait(42, &e.sndQueue)
+	stateSourceObject.Load(43, &e.cc)
+	stateSourceObject.Load(44, &e.packetTooBigCount)
+	stateSourceObject.Load(45, &e.sndMTU)
+	stateSourceObject.Load(46, &e.keepalive)
+	stateSourceObject.Load(47, &e.userTimeout)
+	stateSourceObject.Load(48, &e.deferAccept)
+	stateSourceObject.LoadWait(50, &e.rcv)
+	stateSourceObject.LoadWait(51, &e.snd)
+	stateSourceObject.Load(52, &e.connectingAddress)
+	stateSourceObject.Load(53, &e.amss)
+	stateSourceObject.Load(54, &e.sendTOS)
+	stateSourceObject.Load(55, &e.gso)
+	stateSourceObject.Load(56, &e.tcpLingerTimeout)
+	stateSourceObject.Load(57, &e.closed)
+	stateSourceObject.Load(58, &e.txHash)
+	stateSourceObject.Load(59, &e.owner)
+	stateSourceObject.Load(60, &e.ops)
 	stateSourceObject.LoadValue(4, new(string), func(y interface{}) { e.loadHardError(y.(string)) })
 	stateSourceObject.LoadValue(5, new(string), func(y interface{}) { e.loadLastError(y.(string)) })
 	stateSourceObject.LoadValue(13, new(EndpointState), func(y interface{}) { e.loadState(y.(EndpointState)) })
 	stateSourceObject.LoadValue(26, new(unixTime), func(y interface{}) { e.loadRecentTSTime(y.(unixTime)) })
-	stateSourceObject.LoadValue(50, new([]*endpoint), func(y interface{}) { e.loadAcceptedChan(y.([]*endpoint)) })
+	stateSourceObject.LoadValue(49, new([]*endpoint), func(y interface{}) { e.loadAcceptedChan(y.([]*endpoint)) })
 	stateSourceObject.AfterLoad(e.afterLoad)
 }