From d6a39734c4934aa98141074c67389e80ed130a94 Mon Sep 17 00:00:00 2001
From: Ghanan Gowripalan <ghanan@google.com>
Date: Thu, 28 Jan 2021 20:30:02 -0800
Subject: [PATCH] Avoid locking when route doesn't require resolution

When a route does not need to resolve a remote link address to send a
packet, avoid having to obtain the pending packets queue's lock.

PiperOrigin-RevId: 354456280
---
 pkg/tcpip/stack/nic.go             | 58 ++++++++++++++++++++----------
 pkg/tcpip/stack/pending_packets.go | 18 ++--------
 2 files changed, 42 insertions(+), 34 deletions(-)

diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go
index 16e55085d..e56a624fe 100644
--- a/pkg/tcpip/stack/nic.go
+++ b/pkg/tcpip/stack/nic.go
@@ -309,25 +309,47 @@ func (n *NIC) WritePacket(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumb
 	return err
 }
 
+func (n *NIC) writePacketBuffer(r RouteInfo, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt pendingPacketBuffer) (int, tcpip.Error) {
+	switch pkt := pkt.(type) {
+	case *PacketBuffer:
+		if err := n.writePacket(r, gso, protocol, pkt); err != nil {
+			return 0, err
+		}
+		return 1, nil
+	case *PacketBufferList:
+		return n.writePackets(r, gso, protocol, *pkt)
+	default:
+		panic(fmt.Sprintf("unrecognized pending packet buffer type = %T", pkt))
+	}
+}
+
 func (n *NIC) enqueuePacketBuffer(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt pendingPacketBuffer) (int, tcpip.Error) {
-	// As per relevant RFCs, we should queue packets while we wait for link
-	// resolution to complete.
-	//
-	// RFC 1122 section 2.3.2.2 (for IPv4):
-	//   The link layer SHOULD save (rather than discard) at least
-	//   one (the latest) packet of each set of packets destined to
-	//   the same unresolved IP address, and transmit the saved
-	//   packet when the address has been resolved.
-	//
-	// RFC 4861 section 7.2.2 (for IPv6):
-	//   While waiting for address resolution to complete, the sender MUST, for
-	//   each neighbor, retain a small queue of packets waiting for address
-	//   resolution to complete. The queue MUST hold at least one packet, and MAY
-	//   contain more. However, the number of queued packets per neighbor SHOULD
-	//   be limited to some small value. When a queue overflows, the new arrival
-	//   SHOULD replace the oldest entry. Once address resolution completes, the
-	//   node transmits any queued packets.
-	return n.linkResQueue.enqueue(r, gso, protocol, pkt)
+	routeInfo, _, err := r.resolvedFields(nil)
+	switch err.(type) {
+	case nil:
+		return n.writePacketBuffer(routeInfo, gso, protocol, pkt)
+	case *tcpip.ErrWouldBlock:
+		// As per relevant RFCs, we should queue packets while we wait for link
+		// resolution to complete.
+		//
+		// RFC 1122 section 2.3.2.2 (for IPv4):
+		//   The link layer SHOULD save (rather than discard) at least
+		//   one (the latest) packet of each set of packets destined to
+		//   the same unresolved IP address, and transmit the saved
+		//   packet when the address has been resolved.
+		//
+		// RFC 4861 section 7.2.2 (for IPv6):
+		//   While waiting for address resolution to complete, the sender MUST, for
+		//   each neighbor, retain a small queue of packets waiting for address
+		//   resolution to complete. The queue MUST hold at least one packet, and
+		//   MAY contain more. However, the number of queued packets per neighbor
+		//   SHOULD be limited to some small value. When a queue overflows, the new
+		//   arrival SHOULD replace the oldest entry. Once address resolution
+		//   completes, the node transmits any queued packets.
+		return n.linkResQueue.enqueue(r, gso, protocol, pkt)
+	default:
+		return 0, err
+	}
 }
 
 // WritePacketToRemote implements NetworkInterface.
diff --git a/pkg/tcpip/stack/pending_packets.go b/pkg/tcpip/stack/pending_packets.go
index c6adfb27b..1c651e216 100644
--- a/pkg/tcpip/stack/pending_packets.go
+++ b/pkg/tcpip/stack/pending_packets.go
@@ -114,20 +114,6 @@ func (f *packetsPendingLinkResolution) dequeue(ch <-chan struct{}, linkAddr tcpi
 	}
 }
 
-func (f *packetsPendingLinkResolution) writePacketBuffer(r RouteInfo, gso *GSO, proto tcpip.NetworkProtocolNumber, pkt pendingPacketBuffer) (int, tcpip.Error) {
-	switch pkt := pkt.(type) {
-	case *PacketBuffer:
-		if err := f.nic.writePacket(r, gso, proto, pkt); err != nil {
-			return 0, err
-		}
-		return 1, nil
-	case *PacketBufferList:
-		return f.nic.writePackets(r, gso, proto, *pkt)
-	default:
-		panic(fmt.Sprintf("unrecognized pending packet buffer type = %T", pkt))
-	}
-}
-
 // enqueue a packet to be sent once link resolution completes.
 //
 // If the maximum number of pending resolutions is reached, the packets
@@ -151,7 +137,7 @@ func (f *packetsPendingLinkResolution) enqueue(r *Route, gso *GSO, proto tcpip.N
 		// The route resolved immediately, so we don't need to wait for link
 		// resolution to send the packet.
 		f.mu.Unlock()
-		return f.writePacketBuffer(routeInfo, gso, proto, pkt)
+		return f.nic.writePacketBuffer(routeInfo, gso, proto, pkt)
 	case *tcpip.ErrWouldBlock:
 		// We need to wait for link resolution to complete.
 	default:
@@ -225,7 +211,7 @@ func (f *packetsPendingLinkResolution) dequeuePackets(packets []pendingPacket, l
 	for _, p := range packets {
 		if success {
 			p.routeInfo.RemoteLinkAddress = linkAddr
-			_, _ = f.writePacketBuffer(p.routeInfo, p.gso, p.proto, p.pkt)
+			_, _ = f.nic.writePacketBuffer(p.routeInfo, p.gso, p.proto, p.pkt)
 		} else {
 			f.incrementOutgoingPacketErrors(p.proto, p.pkt)