ipv6: ICMP support

This CL does NDP link-address discovery for IPv6.

It includes several small changes necessary to get linux to talk to
this implementation. In particular, a hop limit of 255 is necessary
for ICMPv6.

PiperOrigin-RevId: 211103930
Change-Id: If25370ab84c6b1decfb15de917f3b0020f2c4e0e

10 files changed, 490 insertions, 27 deletions

diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go
index 9d69f1d45..dd3fe7c87 100644
--- a/pkg/tcpip/link/channel/channel.go
+++ b/pkg/tcpip/link/channel/channel.go
@@ -71,6 +71,12 @@ func (e *Endpoint) Inject(protocol tcpip.NetworkProtocolNumber, vv *buffer.Vecto
 	e.dispatcher.DeliverNetworkPacket(e, "", protocol, &uu)
 }
 
+// InjectLinkAddr injects an inbound packet with a remote link address.
+func (e *Endpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remoteLinkAddr tcpip.LinkAddress, vv *buffer.VectorisedView) {
+	uu := vv.Clone(nil)
+	e.dispatcher.DeliverNetworkPacket(e, remoteLinkAddr, protocol, &uu)
+}
+
 // Attach saves the stack network-layer dispatcher for use later when packets
 // are injected.
 func (e *Endpoint) Attach(dispatcher stack.NetworkDispatcher) {
diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go
index 414a7edb4..1e302f557 100644
--- a/pkg/tcpip/link/sniffer/sniffer.go
+++ b/pkg/tcpip/link/sniffer/sniffer.go
@@ -303,6 +303,37 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, b, plb []byt
 		log.Infof("%s %s %v -> %v %s len:%d id:%04x code:%d", prefix, transName, src, dst, icmpType, size, id, icmp.Code())
 		return
 
+	case header.ICMPv6ProtocolNumber:
+		transName = "icmp"
+		icmp := header.ICMPv6(b)
+		icmpType := "unknown"
+		switch icmp.Type() {
+		case header.ICMPv6DstUnreachable:
+			icmpType = "destination unreachable"
+		case header.ICMPv6PacketTooBig:
+			icmpType = "packet too big"
+		case header.ICMPv6TimeExceeded:
+			icmpType = "time exceeded"
+		case header.ICMPv6ParamProblem:
+			icmpType = "param problem"
+		case header.ICMPv6EchoRequest:
+			icmpType = "echo request"
+		case header.ICMPv6EchoReply:
+			icmpType = "echo reply"
+		case header.ICMPv6RouterSolicit:
+			icmpType = "router solicit"
+		case header.ICMPv6RouterAdvert:
+			icmpType = "router advert"
+		case header.ICMPv6NeighborSolicit:
+			icmpType = "neighbor solicit"
+		case header.ICMPv6NeighborAdvert:
+			icmpType = "neighbor advert"
+		case header.ICMPv6RedirectMsg:
+			icmpType = "redirect message"
+		}
+		log.Infof("%s %s %v -> %v %s len:%d id:%04x code:%d", prefix, transName, src, dst, icmpType, size, id, icmp.Code())
+		return
+
 	case header.UDPProtocolNumber:
 		transName = "udp"
 		udp := header.UDP(b)
diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go
index e6b1f5128..512ef0d60 100644
--- a/pkg/tcpip/network/ipv4/ipv4.go
+++ b/pkg/tcpip/network/ipv4/ipv4.go
@@ -67,7 +67,7 @@ func newEndpoint(nicid tcpip.NICID, addr tcpip.Address, dispatcher stack.Transpo
 		fragmentation: fragmentation.NewFragmentation(fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout),
 	}
 	copy(e.address[:], addr)
-	e.id = stack.NetworkEndpointID{tcpip.Address(e.address[:])}
+	e.id = stack.NetworkEndpointID{LocalAddress: tcpip.Address(e.address[:])}
 
 	go e.echoReplier()
 
diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD
index 1c3eccae0..2f19a659e 100644
--- a/pkg/tcpip/network/ipv6/BUILD
+++ b/pkg/tcpip/network/ipv6/BUILD
@@ -1,6 +1,6 @@
 package(licenses = ["notice"])  # Apache 2.0
 
-load("//tools/go_stateify:defs.bzl", "go_library")
+load("//tools/go_stateify:defs.bzl", "go_library", "go_test")
 
 go_library(
     name = "ipv6",
@@ -19,3 +19,20 @@ go_library(
         "//pkg/tcpip/stack",
     ],
 )
+
+go_test(
+    name = "ipv6_test",
+    size = "small",
+    srcs = ["icmp_test.go"],
+    embed = [":ipv6"],
+    deps = [
+        "//pkg/tcpip",
+        "//pkg/tcpip/buffer",
+        "//pkg/tcpip/header",
+        "//pkg/tcpip/link/channel",
+        "//pkg/tcpip/link/sniffer",
+        "//pkg/tcpip/stack",
+        "//pkg/tcpip/transport/ping",
+        "//pkg/waiter",
+    ],
+)
diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go
index 8b8539def..6db004baa 100644
--- a/pkg/tcpip/network/ipv6/icmp.go
+++ b/pkg/tcpip/network/ipv6/icmp.go
@@ -17,6 +17,7 @@ package ipv6
 import (
 	"encoding/binary"
 
+	"gvisor.googlesource.com/gvisor/pkg/tcpip"
 	"gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"
 	"gvisor.googlesource.com/gvisor/pkg/tcpip/header"
 	"gvisor.googlesource.com/gvisor/pkg/tcpip/stack"
@@ -86,5 +87,141 @@ func (e *endpoint) handleICMP(r *stack.Route, vv *buffer.VectorisedView) {
 		case header.ICMPv6PortUnreachable:
 			e.handleControl(stack.ControlPortUnreachable, 0, vv)
 		}
+
+	case header.ICMPv6NeighborSolicit:
+		if len(v) < header.ICMPv6NeighborSolicitMinimumSize {
+			return
+		}
+		targetAddr := tcpip.Address(v[8 : 8+16])
+		if e.linkAddrCache.CheckLocalAddress(e.nicid, ProtocolNumber, targetAddr) == 0 {
+			// We don't have a useful answer; the best we can do is ignore the request.
+			return
+		}
+		hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.IPv6MinimumSize + header.ICMPv6NeighborAdvertSize)
+		pkt := header.ICMPv6(hdr.Prepend(header.ICMPv6NeighborAdvertSize))
+		pkt.SetType(header.ICMPv6NeighborAdvert)
+		pkt[icmpV6FlagOffset] = ndpSolicitedFlag | ndpOverrideFlag
+		copy(pkt[icmpV6OptOffset-len(targetAddr):], targetAddr)
+		pkt[icmpV6OptOffset] = ndpOptDstLinkAddr
+		pkt[icmpV6LengthOffset] = 1
+		copy(pkt[icmpV6LengthOffset+1:], r.LocalLinkAddress[:])
+		pkt.SetChecksum(icmpChecksum(pkt, r.LocalAddress, r.RemoteAddress, nil))
+		r.WritePacket(&hdr, nil, header.ICMPv6ProtocolNumber)
+
+		e.linkAddrCache.AddLinkAddress(e.nicid, r.RemoteAddress, r.RemoteLinkAddress)
+
+	case header.ICMPv6NeighborAdvert:
+		if len(v) < header.ICMPv6NeighborAdvertSize {
+			return
+		}
+		targetAddr := tcpip.Address(v[8 : 8+16])
+		e.linkAddrCache.AddLinkAddress(e.nicid, targetAddr, r.RemoteLinkAddress)
+		if targetAddr != r.RemoteAddress {
+			e.linkAddrCache.AddLinkAddress(e.nicid, r.RemoteAddress, r.RemoteLinkAddress)
+		}
+
+	case header.ICMPv6EchoRequest:
+		if len(v) < header.ICMPv6EchoMinimumSize {
+			return
+		}
+		vv.TrimFront(header.ICMPv6EchoMinimumSize)
+		data := vv.ToView()
+		hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.IPv6MinimumSize + header.ICMPv6EchoMinimumSize)
+		pkt := header.ICMPv6(hdr.Prepend(header.ICMPv6EchoMinimumSize))
+		copy(pkt, h)
+		pkt.SetType(header.ICMPv6EchoReply)
+		pkt.SetChecksum(icmpChecksum(pkt, r.LocalAddress, r.RemoteAddress, data))
+		r.WritePacket(&hdr, data, header.ICMPv6ProtocolNumber)
+
+	case header.ICMPv6EchoReply:
+		if len(v) < header.ICMPv6EchoMinimumSize {
+			return
+		}
+		e.dispatcher.DeliverTransportPacket(r, header.ICMPv6ProtocolNumber, vv)
+
 	}
 }
+
+const (
+	ndpSolicitedFlag = 1 << 6
+	ndpOverrideFlag  = 1 << 5
+
+	ndpOptSrcLinkAddr = 1
+	ndpOptDstLinkAddr = 2
+
+	icmpV6FlagOffset   = 4
+	icmpV6OptOffset    = 24
+	icmpV6LengthOffset = 25
+
+	icmpV6HopLimit = 255
+)
+
+// solicitedNodeAddr computes the solicited-node multicast address.
+// This is used for NDP. Described in RFC 4291.
+func solicitedNodeAddr(addr tcpip.Address) tcpip.Address {
+	const solicitedNodeMulticastPrefix = "\xff\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x01\xff"
+	return solicitedNodeMulticastPrefix + addr[len(addr)-3:]
+}
+
+var broadcastMAC = tcpip.LinkAddress([]byte{0xff, 0xff, 0xff, 0xff, 0xff, 0xff})
+
+var _ stack.LinkAddressResolver = (*protocol)(nil)
+
+// LinkAddressProtocol implements stack.LinkAddressResolver.
+func (*protocol) LinkAddressProtocol() tcpip.NetworkProtocolNumber {
+	return header.IPv6ProtocolNumber
+}
+
+// LinkAddressRequest implements stack.LinkAddressResolver.
+func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack.LinkEndpoint) *tcpip.Error {
+	snaddr := solicitedNodeAddr(addr)
+	r := &stack.Route{
+		LocalAddress:      localAddr,
+		RemoteAddress:     snaddr,
+		RemoteLinkAddress: broadcastMAC,
+	}
+	hdr := buffer.NewPrependable(int(linkEP.MaxHeaderLength()) + header.IPv6MinimumSize + header.ICMPv6NeighborAdvertSize)
+	pkt := header.ICMPv6(hdr.Prepend(header.ICMPv6NeighborAdvertSize))
+	pkt.SetType(header.ICMPv6NeighborSolicit)
+	copy(pkt[icmpV6OptOffset-len(addr):], addr)
+	pkt[icmpV6OptOffset] = ndpOptSrcLinkAddr
+	pkt[icmpV6LengthOffset] = 1
+	copy(pkt[icmpV6LengthOffset+1:], linkEP.LinkAddress())
+	pkt.SetChecksum(icmpChecksum(pkt, r.LocalAddress, r.RemoteAddress, nil))
+
+	length := uint16(hdr.UsedLength())
+	ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize))
+	ip.Encode(&header.IPv6Fields{
+		PayloadLength: length,
+		NextHeader:    uint8(header.ICMPv6ProtocolNumber),
+		HopLimit:      icmpV6HopLimit,
+		SrcAddr:       r.LocalAddress,
+		DstAddr:       r.RemoteAddress,
+	})
+
+	return linkEP.WritePacket(r, &hdr, nil, ProtocolNumber)
+}
+
+// ResolveStaticAddress implements stack.LinkAddressResolver.
+func (*protocol) ResolveStaticAddress(addr tcpip.Address) (tcpip.LinkAddress, bool) {
+	return "", false
+}
+
+func icmpChecksum(h header.ICMPv6, src, dst tcpip.Address, data []byte) uint16 {
+	// Calculate the IPv6 pseudo-header upper-layer checksum.
+	xsum := header.Checksum([]byte(src), 0)
+	xsum = header.Checksum([]byte(dst), xsum)
+	var upperLayerLength [4]byte
+	binary.BigEndian.PutUint32(upperLayerLength[:], uint32(len(h)+len(data)))
+	xsum = header.Checksum(upperLayerLength[:], xsum)
+	xsum = header.Checksum([]byte{0, 0, 0, uint8(header.ICMPv6ProtocolNumber)}, xsum)
+	xsum = header.Checksum(data, xsum)
+
+	// h[2:4] is the checksum itself, set it aside to avoid checksumming the checksum.
+	h2, h3 := h[2], h[3]
+	h[2], h[3] = 0, 0
+	xsum = ^header.Checksum(h, xsum)
+	h[2], h[3] = h2, h3
+
+	return xsum
+}
diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go
new file mode 100644
index 000000000..582bbb40e
--- /dev/null
+++ b/pkg/tcpip/network/ipv6/icmp_test.go
@@ -0,0 +1,236 @@
+// Copyright 2018 Google Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ipv6
+
+import (
+	"context"
+	"runtime"
+	"strings"
+	"testing"
+	"time"
+
+	"gvisor.googlesource.com/gvisor/pkg/tcpip"
+	"gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"
+	"gvisor.googlesource.com/gvisor/pkg/tcpip/header"
+	"gvisor.googlesource.com/gvisor/pkg/tcpip/link/channel"
+	"gvisor.googlesource.com/gvisor/pkg/tcpip/link/sniffer"
+	"gvisor.googlesource.com/gvisor/pkg/tcpip/stack"
+	"gvisor.googlesource.com/gvisor/pkg/tcpip/transport/ping"
+	"gvisor.googlesource.com/gvisor/pkg/waiter"
+)
+
+const (
+	linkAddr0 = tcpip.LinkAddress("\x01\x02\x03\x04\x05\x06")
+	linkAddr1 = tcpip.LinkAddress("\x0a\x0b\x0c\x0d\x0e\x0f")
+)
+
+// linkLocalAddr computes the default IPv6 link-local address from
+// a link-layer (MAC) address.
+func linkLocalAddr(linkAddr tcpip.LinkAddress) tcpip.Address {
+	// Convert a 48-bit MAC to an EUI-64 and then prepend the
+	// link-local header, FE80::.
+	//
+	// The conversion is very nearly:
+	//	aa:bb:cc:dd:ee:ff => FE80::Aabb:ccFF:FEdd:eeff
+	// Note the capital A. The conversion aa->Aa involves a bit flip.
+	lladdrb := [16]byte{
+		0:  0xFE,
+		1:  0x80,
+		8:  linkAddr[0] ^ 2,
+		9:  linkAddr[1],
+		10: linkAddr[2],
+		11: 0xFF,
+		12: 0xFE,
+		13: linkAddr[3],
+		14: linkAddr[4],
+		15: linkAddr[5],
+	}
+	return tcpip.Address(lladdrb[:])
+}
+
+var (
+	lladdr0 = linkLocalAddr(linkAddr0)
+	lladdr1 = linkLocalAddr(linkAddr1)
+)
+
+type testContext struct {
+	t  *testing.T
+	s0 *stack.Stack
+	s1 *stack.Stack
+
+	linkEP0 *channel.Endpoint
+	linkEP1 *channel.Endpoint
+
+	icmpCh chan header.ICMPv6Type
+}
+
+type endpointWithResolutionCapability struct {
+	stack.LinkEndpoint
+}
+
+func (e endpointWithResolutionCapability) Capabilities() stack.LinkEndpointCapabilities {
+	return e.LinkEndpoint.Capabilities() | stack.CapabilityResolutionRequired
+}
+
+func newTestContext(t *testing.T) *testContext {
+	c := &testContext{
+		t:      t,
+		s0:     stack.New([]string{ProtocolName}, []string{ping.ProtocolName6}, stack.Options{}),
+		s1:     stack.New([]string{ProtocolName}, []string{ping.ProtocolName6}, stack.Options{}),
+		icmpCh: make(chan header.ICMPv6Type, 10),
+	}
+
+	const defaultMTU = 65536
+	_, linkEP0 := channel.New(256, defaultMTU, linkAddr0)
+	c.linkEP0 = linkEP0
+	wrappedEP0 := endpointWithResolutionCapability{LinkEndpoint: linkEP0}
+	id0 := stack.RegisterLinkEndpoint(wrappedEP0)
+	if testing.Verbose() {
+		id0 = sniffer.New(id0)
+	}
+	if err := c.s0.CreateNIC(1, id0); err != nil {
+		t.Fatalf("CreateNIC s0: %v", err)
+	}
+	if err := c.s0.AddAddress(1, ProtocolNumber, lladdr0); err != nil {
+		t.Fatalf("AddAddress lladdr0: %v", err)
+	}
+	if err := c.s0.AddAddress(1, ProtocolNumber, solicitedNodeAddr(lladdr0)); err != nil {
+		t.Fatalf("AddAddress sn lladdr0: %v", err)
+	}
+
+	_, linkEP1 := channel.New(256, defaultMTU, linkAddr1)
+	c.linkEP1 = linkEP1
+	wrappedEP1 := endpointWithResolutionCapability{LinkEndpoint: linkEP1}
+	id1 := stack.RegisterLinkEndpoint(wrappedEP1)
+	if err := c.s1.CreateNIC(1, id1); err != nil {
+		t.Fatalf("CreateNIC failed: %v", err)
+	}
+	if err := c.s1.AddAddress(1, ProtocolNumber, lladdr1); err != nil {
+		t.Fatalf("AddAddress lladdr1: %v", err)
+	}
+	if err := c.s1.AddAddress(1, ProtocolNumber, solicitedNodeAddr(lladdr1)); err != nil {
+		t.Fatalf("AddAddress sn lladdr1: %v", err)
+	}
+
+	c.s0.SetRouteTable(
+		[]tcpip.Route{{
+			Destination: lladdr1,
+			Mask:        tcpip.Address(strings.Repeat("\xff", 16)),
+			NIC:         1,
+		}},
+	)
+	c.s1.SetRouteTable(
+		[]tcpip.Route{{
+			Destination: lladdr0,
+			Mask:        tcpip.Address(strings.Repeat("\xff", 16)),
+			NIC:         1,
+		}},
+	)
+
+	go c.routePackets(linkEP0.C, linkEP1)
+	go c.routePackets(linkEP1.C, linkEP0)
+
+	return c
+}
+
+func (c *testContext) countPacket(pkt channel.PacketInfo) {
+	if pkt.Proto != ProtocolNumber {
+		return
+	}
+	ipv6 := header.IPv6(pkt.Header)
+	transProto := tcpip.TransportProtocolNumber(ipv6.NextHeader())
+	if transProto != header.ICMPv6ProtocolNumber {
+		return
+	}
+	b := pkt.Header[header.IPv6MinimumSize:]
+	icmp := header.ICMPv6(b)
+	c.icmpCh <- icmp.Type()
+}
+
+func (c *testContext) routePackets(ch <-chan channel.PacketInfo, ep *channel.Endpoint) {
+	for pkt := range ch {
+		c.countPacket(pkt)
+		views := []buffer.View{pkt.Header, pkt.Payload}
+		size := len(pkt.Header) + len(pkt.Payload)
+		vv := buffer.NewVectorisedView(size, views)
+		ep.InjectLinkAddr(pkt.Proto, ep.LinkAddress(), &vv)
+	}
+}
+
+func (c *testContext) cleanup() {
+	close(c.linkEP0.C)
+	close(c.linkEP1.C)
+}
+
+func TestLinkResolution(t *testing.T) {
+	c := newTestContext(t)
+	defer c.cleanup()
+	r, err := c.s0.FindRoute(1, lladdr0, lladdr1, ProtocolNumber)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer r.Release()
+
+	hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.IPv6MinimumSize + header.ICMPv6EchoMinimumSize)
+	pkt := header.ICMPv6(hdr.Prepend(header.ICMPv6EchoMinimumSize))
+	pkt.SetType(header.ICMPv6EchoRequest)
+	pkt.SetChecksum(icmpChecksum(pkt, r.LocalAddress, r.RemoteAddress, nil))
+	payload := tcpip.SlicePayload(hdr.UsedBytes())
+
+	// We can't send our payload directly over the route because that
+	// doesn't provoke NDP discovery.
+	var wq waiter.Queue
+	ep, err := c.s0.NewEndpoint(header.ICMPv6ProtocolNumber, ProtocolNumber, &wq)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// This actually takes about 10 milliseconds, so no need to wait for
+	// a multi-minute go test timeout if something is broken.
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+	defer cancel()
+
+	for {
+		if ctx.Err() != nil {
+			break
+		}
+		if _, err := ep.Write(payload, tcpip.WriteOptions{To: &tcpip.FullAddress{NIC: 1, Addr: lladdr1}}); err == tcpip.ErrNoLinkAddress {
+			// There's something asynchronous going on; yield to let it do its thing.
+			runtime.Gosched()
+		} else if err == nil {
+			break
+		} else {
+			t.Fatal(err)
+		}
+	}
+
+	stats := make(map[header.ICMPv6Type]int)
+	for {
+		select {
+		case <-ctx.Done():
+			t.Errorf("timeout waiting for ICMP, got: %#+v", stats)
+			return
+		case typ := <-c.icmpCh:
+			stats[typ]++
+
+			if stats[header.ICMPv6NeighborSolicit] > 0 &&
+				stats[header.ICMPv6NeighborAdvert] > 0 &&
+				stats[header.ICMPv6EchoRequest] > 0 &&
+				stats[header.ICMPv6EchoReply] > 0 {
+				return
+			}
+		}
+	}
+}
diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go
index f48d120c5..51d34b7ec 100644
--- a/pkg/tcpip/network/ipv6/ipv6.go
+++ b/pkg/tcpip/network/ipv6/ipv6.go
@@ -42,18 +42,12 @@ const (
 type address [header.IPv6AddressSize]byte
 
 type endpoint struct {
-	nicid      tcpip.NICID
-	id         stack.NetworkEndpointID
-	address    address
-	linkEP     stack.LinkEndpoint
-	dispatcher stack.TransportDispatcher
-}
-
-func newEndpoint(nicid tcpip.NICID, addr tcpip.Address, dispatcher stack.TransportDispatcher, linkEP stack.LinkEndpoint) *endpoint {
-	e := &endpoint{nicid: nicid, linkEP: linkEP, dispatcher: dispatcher}
-	copy(e.address[:], addr)
-	e.id = stack.NetworkEndpointID{tcpip.Address(e.address[:])}
-	return e
+	nicid         tcpip.NICID
+	id            stack.NetworkEndpointID
+	address       address
+	linkEP        stack.LinkEndpoint
+	linkAddrCache stack.LinkAddressCache
+	dispatcher    stack.TransportDispatcher
 }
 
 // MTU implements stack.NetworkEndpoint.MTU. It returns the link-layer MTU minus
@@ -93,7 +87,7 @@ func (e *endpoint) WritePacket(r *stack.Route, hdr *buffer.Prependable, payload
 	ip.Encode(&header.IPv6Fields{
 		PayloadLength: length,
 		NextHeader:    uint8(protocol),
-		HopLimit:      65,
+		HopLimit:      icmpV6HopLimit,
 		SrcAddr:       tcpip.Address(e.address[:]),
 		DstAddr:       r.RemoteAddress,
 	})
@@ -154,7 +148,15 @@ func (*protocol) ParseAddresses(v buffer.View) (src, dst tcpip.Address) {
 
 // NewEndpoint creates a new ipv6 endpoint.
 func (p *protocol) NewEndpoint(nicid tcpip.NICID, addr tcpip.Address, linkAddrCache stack.LinkAddressCache, dispatcher stack.TransportDispatcher, linkEP stack.LinkEndpoint) (stack.NetworkEndpoint, *tcpip.Error) {
-	return newEndpoint(nicid, addr, dispatcher, linkEP), nil
+	e := &endpoint{
+		nicid:         nicid,
+		linkEP:        linkEP,
+		linkAddrCache: linkAddrCache,
+		dispatcher:    dispatcher,
+	}
+	copy(e.address[:], addr)
+	e.id = stack.NetworkEndpointID{LocalAddress: tcpip.Address(e.address[:])}
+	return e, nil
 }
 
 // SetOption implements NetworkProtocol.SetOption.
diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go
index c158e2005..77134c42a 100644
--- a/pkg/tcpip/stack/nic.go
+++ b/pkg/tcpip/stack/nic.go
@@ -165,7 +165,7 @@ func (n *NIC) addAddressLocked(protocol tcpip.NetworkProtocolNumber, addr tcpip.
 
 	// Set up cache if link address resolution exists for this protocol.
 	if n.linkEP.Capabilities()&CapabilityResolutionRequired != 0 {
-		if linkRes := n.stack.linkAddrResolvers[protocol]; linkRes != nil {
+		if _, ok := n.stack.linkAddrResolvers[protocol]; ok {
 			ref.linkCache = n.stack
 		}
 	}
diff --git a/pkg/tcpip/transport/ping/endpoint.go b/pkg/tcpip/transport/ping/endpoint.go
index f097ac057..7e10c0aae 100644
--- a/pkg/tcpip/transport/ping/endpoint.go
+++ b/pkg/tcpip/transport/ping/endpoint.go
@@ -56,6 +56,7 @@ type endpoint struct {
 	// change throughout the lifetime of the endpoint.
 	stack       *stack.Stack `state:"manual"`
 	netProto    tcpip.NetworkProtocolNumber
+	transProto  tcpip.TransportProtocolNumber
 	waiterQueue *waiter.Queue
 
 	// The following fields are used to manage the receive queue, and are
@@ -81,10 +82,11 @@ type endpoint struct {
 	route         stack.Route `state:"manual"`
 }
 
-func newEndpoint(stack *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQueue *waiter.Queue) *endpoint {
+func newEndpoint(stack *stack.Stack, netProto tcpip.NetworkProtocolNumber, transProto tcpip.TransportProtocolNumber, waiterQueue *waiter.Queue) *endpoint {
 	return &endpoint{
 		stack:         stack,
 		netProto:      netProto,
+		transProto:    transProto,
 		waiterQueue:   waiterQueue,
 		rcvBufSizeMax: 32 * 1024,
 		sndBufSize:    32 * 1024,
@@ -98,7 +100,7 @@ func (e *endpoint) Close() {
 	e.shutdownFlags = tcpip.ShutdownRead | tcpip.ShutdownWrite
 	switch e.state {
 	case stateBound, stateConnected:
-		e.stack.UnregisterTransportEndpoint(e.regNICID, []tcpip.NetworkProtocolNumber{e.netProto}, ProtocolNumber4, e.id)
+		e.stack.UnregisterTransportEndpoint(e.regNICID, []tcpip.NetworkProtocolNumber{e.netProto}, e.transProto, e.id)
 	}
 
 	// Close the receive list and drain it.
@@ -297,7 +299,7 @@ func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tc
 		err = sendPing4(route, e.id.LocalPort, v)
 
 	case header.IPv6ProtocolNumber:
-		// TODO: Support IPv6.
+		err = sendPing6(route, e.id.LocalPort, v)
 	}
 
 	return uintptr(len(v)), err
@@ -385,6 +387,30 @@ func sendPing4(r *stack.Route, ident uint16, data buffer.View) *tcpip.Error {
 	return r.WritePacket(&hdr, data, header.ICMPv4ProtocolNumber)
 }
 
+func sendPing6(r *stack.Route, ident uint16, data buffer.View) *tcpip.Error {
+	if len(data) < header.ICMPv6EchoMinimumSize {
+		return tcpip.ErrInvalidEndpointState
+	}
+
+	// Set the ident. Sequence number is provided by the user.
+	binary.BigEndian.PutUint16(data[header.ICMPv6MinimumSize:], ident)
+
+	hdr := buffer.NewPrependable(header.ICMPv6EchoMinimumSize + int(r.MaxHeaderLength()))
+
+	icmpv6 := header.ICMPv6(hdr.Prepend(header.ICMPv6EchoMinimumSize))
+	copy(icmpv6, data)
+	data = data[header.ICMPv6EchoMinimumSize:]
+
+	if icmpv6.Type() != header.ICMPv6EchoRequest || icmpv6.Code() != 0 {
+		return tcpip.ErrInvalidEndpointState
+	}
+
+	icmpv6.SetChecksum(0)
+	icmpv6.SetChecksum(^header.Checksum(icmpv6, header.Checksum(data, 0)))
+
+	return r.WritePacket(&hdr, data, header.ICMPv6ProtocolNumber)
+}
+
 func (e *endpoint) checkV4Mapped(addr *tcpip.FullAddress, allowMismatch bool) (tcpip.NetworkProtocolNumber, *tcpip.Error) {
 	netProto := e.netProto
 	if header.IsV4MappedAddress(addr.Addr) {
@@ -508,14 +534,14 @@ func (e *endpoint) registerWithStack(nicid tcpip.NICID, netProtos []tcpip.Networ
 	if id.LocalPort != 0 {
 		// The endpoint already has a local port, just attempt to
 		// register it.
-		err := e.stack.RegisterTransportEndpoint(nicid, netProtos, ProtocolNumber4, id, e)
+		err := e.stack.RegisterTransportEndpoint(nicid, netProtos, e.transProto, id, e)
 		return id, err
 	}
 
 	// We need to find a port for the endpoint.
 	_, err := e.stack.PickEphemeralPort(func(p uint16) (bool, *tcpip.Error) {
 		id.LocalPort = p
-		err := e.stack.RegisterTransportEndpoint(nicid, netProtos, ProtocolNumber4, id, e)
+		err := e.stack.RegisterTransportEndpoint(nicid, netProtos, e.transProto, id, e)
 		switch err {
 		case nil:
 			return true, nil
@@ -564,7 +590,7 @@ func (e *endpoint) bindLocked(addr tcpip.FullAddress, commit func() *tcpip.Error
 	if commit != nil {
 		if err := commit(); err != nil {
 			// Unregister, the commit failed.
-			e.stack.UnregisterTransportEndpoint(addr.NIC, netProtos, ProtocolNumber4, id)
+			e.stack.UnregisterTransportEndpoint(addr.NIC, netProtos, e.transProto, id)
 			return err
 		}
 	}
diff --git a/pkg/tcpip/transport/ping/protocol.go b/pkg/tcpip/transport/ping/protocol.go
index 8a8192064..b885f3627 100644
--- a/pkg/tcpip/transport/ping/protocol.go
+++ b/pkg/tcpip/transport/ping/protocol.go
@@ -71,7 +71,7 @@ func (p *protocol) NewEndpoint(stack *stack.Stack, netProto tcpip.NetworkProtoco
 	if netProto != p.netProto() {
 		return nil, tcpip.ErrUnknownProtocol
 	}
-	return newEndpoint(stack, netProto, waiterQueue), nil
+	return newEndpoint(stack, netProto, p.number, waiterQueue), nil
 }
 
 // MinimumPacketSize returns the minimum valid ping packet size.
@@ -87,8 +87,14 @@ func (p *protocol) MinimumPacketSize() int {
 
 // ParsePorts returns the source and destination ports stored in the given ping
 // packet.
-func (*protocol) ParsePorts(v buffer.View) (src, dst uint16, err *tcpip.Error) {
-	return 0, binary.BigEndian.Uint16(v[header.ICMPv4MinimumSize:]), nil
+func (p *protocol) ParsePorts(v buffer.View) (src, dst uint16, err *tcpip.Error) {
+	switch p.number {
+	case ProtocolNumber4:
+		return 0, binary.BigEndian.Uint16(v[header.ICMPv4MinimumSize:]), nil
+	case ProtocolNumber6:
+		return 0, binary.BigEndian.Uint16(v[header.ICMPv6MinimumSize:]), nil
+	}
+	panic(fmt.Sprint("unknown protocol number: ", p.number))
 }
 
 // HandleUnknownDestinationPacket handles packets targeted at this protocol but
@@ -112,5 +118,7 @@ func init() {
 		return &protocol{ProtocolNumber4}
 	})
 
-	// TODO: Support IPv6.
+	stack.RegisterTransportProtocolFactory(ProtocolName6, func() stack.TransportProtocol {
+		return &protocol{ProtocolNumber6}
+	})
 }