Block for link address resolution

Previously, if address resolution for UDP or Ping sockets required sending
packets using Write in Transport layer, Resolve would return ErrWouldBlock
and Write would return ErrNoLinkAddress. Meanwhile startAddressResolution
would run in background. Further calls to Write using same address would also
return ErrNoLinkAddress until resolution has been completed successfully.

Since Write is not allowed to block and System Calls need to be
interruptible in System Call layer, the caller to Write is responsible for
blocking upon return of ErrWouldBlock.

Now, when startAddressResolution is called a notification channel for
the completion of the address resolution is returned.
The channel will traverse up to the calling function of Write as well as
ErrNoLinkAddress. Once address resolution is complete (success or not) the
channel is closed. The caller would call Write again to send packets and
check if address resolution was compeleted successfully or not.

Fixes google/gvisor#5

Change-Id: Idafaf31982bee1915ca084da39ae7bd468cebd93
PiperOrigin-RevId: 214962200

4 files changed, 35 insertions, 35 deletions

diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go
index 68c0d4472..27dbcace2 100644
--- a/pkg/tcpip/transport/tcp/connect.go
+++ b/pkg/tcpip/transport/tcp/connect.go
@@ -365,7 +365,7 @@ func (h *handshake) resolveRoute() *tcpip.Error {
 	for {
 		switch index {
 		case wakerForResolution:
-			if err := h.ep.route.Resolve(resolutionWaker); err != tcpip.ErrWouldBlock {
+			if _, err := h.ep.route.Resolve(resolutionWaker); err != tcpip.ErrWouldBlock {
 				// Either success (err == nil) or failure.
 				return err
 			}
diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go
index e82e25233..707d6be96 100644
--- a/pkg/tcpip/transport/tcp/endpoint.go
+++ b/pkg/tcpip/transport/tcp/endpoint.go
@@ -492,7 +492,7 @@ func (e *endpoint) readLocked() (buffer.View, *tcpip.Error) {
 }
 
 // Write writes data to the endpoint's peer.
-func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tcpip.Error) {
+func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, <-chan struct{}, *tcpip.Error) {
 	// Linux completely ignores any address passed to sendto(2) for TCP sockets
 	// (without the MSG_FASTOPEN flag). Corking is unimplemented, so opts.More
 	// and opts.EndOfRecord are also ignored.
@@ -504,15 +504,15 @@ func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tc
 	if e.state != stateConnected {
 		switch e.state {
 		case stateError:
-			return 0, e.hardError
+			return 0, nil, e.hardError
 		default:
-			return 0, tcpip.ErrClosedForSend
+			return 0, nil, tcpip.ErrClosedForSend
 		}
 	}
 
 	// Nothing to do if the buffer is empty.
 	if p.Size() == 0 {
-		return 0, nil
+		return 0, nil, nil
 	}
 
 	e.sndBufMu.Lock()
@@ -520,20 +520,20 @@ func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tc
 	// Check if the connection has already been closed for sends.
 	if e.sndClosed {
 		e.sndBufMu.Unlock()
-		return 0, tcpip.ErrClosedForSend
+		return 0, nil, tcpip.ErrClosedForSend
 	}
 
 	// Check against the limit.
 	avail := e.sndBufSize - e.sndBufUsed
 	if avail <= 0 {
 		e.sndBufMu.Unlock()
-		return 0, tcpip.ErrWouldBlock
+		return 0, nil, tcpip.ErrWouldBlock
 	}
 
 	v, perr := p.Get(avail)
 	if perr != nil {
 		e.sndBufMu.Unlock()
-		return 0, perr
+		return 0, nil, perr
 	}
 
 	var err *tcpip.Error
@@ -558,7 +558,7 @@ func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tc
 		// Let the protocol goroutine do the work.
 		e.sndWaker.Assert()
 	}
-	return uintptr(l), err
+	return uintptr(l), nil, err
 }
 
 // Peek reads data without consuming it from the endpoint.
diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go
index ac21e565b..48852ea47 100644
--- a/pkg/tcpip/transport/tcp/tcp_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_test.go
@@ -869,7 +869,7 @@ func TestSimpleSend(t *testing.T) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -910,7 +910,7 @@ func TestZeroWindowSend(t *testing.T) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	_, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{})
+	_, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{})
 	if err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
@@ -971,7 +971,7 @@ func TestScaledWindowConnect(t *testing.T) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1004,7 +1004,7 @@ func TestNonScaledWindowConnect(t *testing.T) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1077,7 +1077,7 @@ func TestScaledWindowAccept(t *testing.T) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1150,7 +1150,7 @@ func TestNonScaledWindowAccept(t *testing.T) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1265,7 +1265,7 @@ func testBrokenUpWrite(t *testing.T, c *context.Context, maxPayload int) {
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1653,7 +1653,7 @@ func TestSendOnResetConnection(t *testing.T) {
 
 	// Try to write.
 	view := buffer.NewView(10)
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != tcpip.ErrConnectionReset {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != tcpip.ErrConnectionReset {
 		t.Fatalf("got c.EP.Write(...) = %v, want = %v", err, tcpip.ErrConnectionReset)
 	}
 }
@@ -1763,7 +1763,7 @@ func TestFinWithNoPendingData(t *testing.T) {
 
 	// Write something out, and have it acknowledged.
 	view := buffer.NewView(10)
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1836,7 +1836,7 @@ func TestFinWithPendingDataCwndFull(t *testing.T) {
 	// any of them.
 	view := buffer.NewView(10)
 	for i := tcp.InitialCwnd; i > 0; i-- {
-		if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+		if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 			t.Fatalf("Write failed: %v", err)
 		}
 	}
@@ -1922,7 +1922,7 @@ func TestFinWithPendingData(t *testing.T) {
 
 	// Write something out, and acknowledge it to get cwnd to 2.
 	view := buffer.NewView(10)
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -1948,7 +1948,7 @@ func TestFinWithPendingData(t *testing.T) {
 	})
 
 	// Write new data, but don't acknowledge it.
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2009,7 +2009,7 @@ func TestFinWithPartialAck(t *testing.T) {
 	// Write something out, and acknowledge it to get cwnd to 2. Also send
 	// FIN from the test side.
 	view := buffer.NewView(10)
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2046,7 +2046,7 @@ func TestFinWithPartialAck(t *testing.T) {
 	)
 
 	// Write new data, but don't acknowledge it.
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2116,7 +2116,7 @@ func TestExponentialIncreaseDuringSlowStart(t *testing.T) {
 
 	// Write all the data in one shot. Packets will only be written at the
 	// MTU size though.
-	if _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2158,7 +2158,7 @@ func TestCongestionAvoidance(t *testing.T) {
 
 	// Write all the data in one shot. Packets will only be written at the
 	// MTU size though.
-	if _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2263,7 +2263,7 @@ func TestCubicCongestionAvoidance(t *testing.T) {
 
 	// Write all the data in one shot. Packets will only be written at the
 	// MTU size though.
-	if _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2371,7 +2371,7 @@ func TestFastRecovery(t *testing.T) {
 
 	// Write all the data in one shot. Packets will only be written at the
 	// MTU size though.
-	if _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2503,11 +2503,11 @@ func TestRetransmit(t *testing.T) {
 	// Write all the data in two shots. Packets will only be written at the
 	// MTU size though.
 	half := data[:len(data)/2]
-	if _, err := c.EP.Write(tcpip.SlicePayload(half), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(half), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 	half = data[len(data)/2:]
-	if _, err := c.EP.Write(tcpip.SlicePayload(half), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(half), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -2605,7 +2605,7 @@ func scaledSendWindow(t *testing.T, scale uint8) {
 
 	// Send some data. Check that it's capped by the window size.
 	view := buffer.NewView(65535)
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -3099,7 +3099,7 @@ func TestSelfConnect(t *testing.T) {
 	data := []byte{1, 2, 3}
 	view := buffer.NewView(len(data))
 	copy(view, data)
-	if _, err := ep.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := ep.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -3290,7 +3290,7 @@ func TestPathMTUDiscovery(t *testing.T) {
 		data[i] = byte(i)
 	}
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(data), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
@@ -3495,7 +3495,7 @@ func TestKeepalive(t *testing.T) {
 	// Send some data and wait before ACKing it. Keepalives should be disabled
 	// during this period.
 	view := buffer.NewView(3)
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Write failed: %v", err)
 	}
 
diff --git a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go
index 894ead507..ca16fc8fa 100644
--- a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go
@@ -147,7 +147,7 @@ func timeStampEnabledAccept(t *testing.T, cookieEnabled bool, wndScale int, wndS
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Unexpected error from Write: %v", err)
 	}
 
@@ -210,7 +210,7 @@ func timeStampDisabledAccept(t *testing.T, cookieEnabled bool, wndScale int, wnd
 	view := buffer.NewView(len(data))
 	copy(view, data)
 
-	if _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
+	if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil {
 		t.Fatalf("Unexpected error from Write: %v", err)
 	}