Merge branch 'master' into ip-forwarding

- Merges aleksej-paschenko's with HEAD
- Adds vfs2 support for ip_forward

3 files changed, 133 insertions, 90 deletions

diff --git a/pkg/tcpip/network/arp/BUILD b/pkg/tcpip/network/arp/BUILD
index df0d3a8c0..eddf7b725 100644
--- a/pkg/tcpip/network/arp/BUILD
+++ b/pkg/tcpip/network/arp/BUILD
@@ -1,15 +1,11 @@
-load("//tools/go_stateify:defs.bzl", "go_library")
-load("@io_bazel_rules_go//go:def.bzl", "go_test")
+load("//tools:defs.bzl", "go_library", "go_test")
 
 package(licenses = ["notice"])
 
 go_library(
     name = "arp",
     srcs = ["arp.go"],
-    importpath = "gvisor.dev/gvisor/pkg/tcpip/network/arp",
-    visibility = [
-        "//visibility:public",
-    ],
+    visibility = ["//visibility:public"],
     deps = [
         "//pkg/tcpip",
         "//pkg/tcpip/buffer",
diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go
index 46178459e..920872c3f 100644
--- a/pkg/tcpip/network/arp/arp.go
+++ b/pkg/tcpip/network/arp/arp.go
@@ -42,7 +42,8 @@ const (
 
 // endpoint implements stack.NetworkEndpoint.
 type endpoint struct {
-	nicid         tcpip.NICID
+	protocol      *protocol
+	nicID         tcpip.NICID
 	linkEP        stack.LinkEndpoint
 	linkAddrCache stack.LinkAddressCache
 }
@@ -58,43 +59,39 @@ func (e *endpoint) MTU() uint32 {
 }
 
 func (e *endpoint) NICID() tcpip.NICID {
-	return e.nicid
+	return e.nicID
 }
 
 func (e *endpoint) Capabilities() stack.LinkEndpointCapabilities {
 	return e.linkEP.Capabilities()
 }
 
-func (e *endpoint) ID() *stack.NetworkEndpointID {
-	return &stack.NetworkEndpointID{ProtocolAddress}
-}
-
-func (e *endpoint) PrefixLen() int {
-	return 0
-}
-
 func (e *endpoint) MaxHeaderLength() uint16 {
 	return e.linkEP.MaxHeaderLength() + header.ARPSize
 }
 
 func (e *endpoint) Close() {}
 
-func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, buffer.Prependable, buffer.VectorisedView, stack.NetworkHeaderParams, stack.PacketLooping) *tcpip.Error {
+func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, stack.NetworkHeaderParams, *stack.PacketBuffer) *tcpip.Error {
 	return tcpip.ErrNotSupported
 }
 
+// NetworkProtocolNumber implements stack.NetworkEndpoint.NetworkProtocolNumber.
+func (e *endpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNumber {
+	return e.protocol.Number()
+}
+
 // WritePackets implements stack.NetworkEndpoint.WritePackets.
-func (e *endpoint) WritePackets(*stack.Route, *stack.GSO, []stack.PacketDescriptor, buffer.VectorisedView, stack.NetworkHeaderParams, stack.PacketLooping) (int, *tcpip.Error) {
+func (e *endpoint) WritePackets(*stack.Route, *stack.GSO, stack.PacketBufferList, stack.NetworkHeaderParams) (int, *tcpip.Error) {
 	return 0, tcpip.ErrNotSupported
 }
 
-func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, payload buffer.VectorisedView, loop stack.PacketLooping) *tcpip.Error {
+func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error {
 	return tcpip.ErrNotSupported
 }
 
-func (e *endpoint) HandlePacket(r *stack.Route, vv buffer.VectorisedView) {
-	v := vv.First()
-	h := header.ARP(v)
+func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) {
+	h := header.ARP(pkt.NetworkHeader().View())
 	if !h.IsValid() {
 		return
 	}
@@ -102,23 +99,25 @@ func (e *endpoint) HandlePacket(r *stack.Route, vv buffer.VectorisedView) {
 	switch h.Op() {
 	case header.ARPRequest:
 		localAddr := tcpip.Address(h.ProtocolAddressTarget())
-		if e.linkAddrCache.CheckLocalAddress(e.nicid, header.IPv4ProtocolNumber, localAddr) == 0 {
+		if e.linkAddrCache.CheckLocalAddress(e.nicID, header.IPv4ProtocolNumber, localAddr) == 0 {
 			return // we have no useful answer, ignore the request
 		}
-		hdr := buffer.NewPrependable(int(e.linkEP.MaxHeaderLength()) + header.ARPSize)
-		pkt := header.ARP(hdr.Prepend(header.ARPSize))
-		pkt.SetIPv4OverEthernet()
-		pkt.SetOp(header.ARPReply)
-		copy(pkt.HardwareAddressSender(), r.LocalLinkAddress[:])
-		copy(pkt.ProtocolAddressSender(), h.ProtocolAddressTarget())
-		copy(pkt.HardwareAddressTarget(), h.HardwareAddressSender())
-		copy(pkt.ProtocolAddressTarget(), h.ProtocolAddressSender())
-		e.linkEP.WritePacket(r, nil /* gso */, hdr, buffer.VectorisedView{}, ProtocolNumber)
+		pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+			ReserveHeaderBytes: int(e.linkEP.MaxHeaderLength()) + header.ARPSize,
+		})
+		packet := header.ARP(pkt.NetworkHeader().Push(header.ARPSize))
+		packet.SetIPv4OverEthernet()
+		packet.SetOp(header.ARPReply)
+		copy(packet.HardwareAddressSender(), r.LocalLinkAddress[:])
+		copy(packet.ProtocolAddressSender(), h.ProtocolAddressTarget())
+		copy(packet.HardwareAddressTarget(), h.HardwareAddressSender())
+		copy(packet.ProtocolAddressTarget(), h.ProtocolAddressSender())
+		_ = e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, pkt)
 		fallthrough // also fill the cache from requests
 	case header.ARPReply:
 		addr := tcpip.Address(h.ProtocolAddressSender())
 		linkAddr := tcpip.LinkAddress(h.HardwareAddressSender())
-		e.linkAddrCache.AddLinkAddress(e.nicid, addr, linkAddr)
+		e.linkAddrCache.AddLinkAddress(e.nicID, addr, linkAddr)
 	}
 }
 
@@ -135,76 +134,77 @@ func (*protocol) ParseAddresses(v buffer.View) (src, dst tcpip.Address) {
 	return tcpip.Address(h.ProtocolAddressSender()), ProtocolAddress
 }
 
-func (p *protocol) NewEndpoint(nicid tcpip.NICID, addrWithPrefix tcpip.AddressWithPrefix, linkAddrCache stack.LinkAddressCache, dispatcher stack.TransportDispatcher, sender stack.LinkEndpoint) (stack.NetworkEndpoint, *tcpip.Error) {
-	if addrWithPrefix.Address != ProtocolAddress {
-		return nil, tcpip.ErrBadLocalAddress
-	}
+func (p *protocol) NewEndpoint(nicID tcpip.NICID, linkAddrCache stack.LinkAddressCache, dispatcher stack.TransportDispatcher, sender stack.LinkEndpoint, st *stack.Stack) stack.NetworkEndpoint {
 	return &endpoint{
-		nicid:         nicid,
+		protocol:      p,
+		nicID:         nicID,
 		linkEP:        sender,
 		linkAddrCache: linkAddrCache,
-	}, nil
+	}
 }
 
-// LinkAddressProtocol implements stack.LinkAddressResolver.
+// LinkAddressProtocol implements stack.LinkAddressResolver.LinkAddressProtocol.
 func (*protocol) LinkAddressProtocol() tcpip.NetworkProtocolNumber {
 	return header.IPv4ProtocolNumber
 }
 
-// LinkAddressRequest implements stack.LinkAddressResolver.
-func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack.LinkEndpoint) *tcpip.Error {
+// LinkAddressRequest implements stack.LinkAddressResolver.LinkAddressRequest.
+func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, remoteLinkAddr tcpip.LinkAddress, linkEP stack.LinkEndpoint) *tcpip.Error {
 	r := &stack.Route{
-		RemoteLinkAddress: broadcastMAC,
+		RemoteLinkAddress: remoteLinkAddr,
+	}
+	if len(r.RemoteLinkAddress) == 0 {
+		r.RemoteLinkAddress = header.EthernetBroadcastAddress
 	}
 
-	hdr := buffer.NewPrependable(int(linkEP.MaxHeaderLength()) + header.ARPSize)
-	h := header.ARP(hdr.Prepend(header.ARPSize))
+	pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+		ReserveHeaderBytes: int(linkEP.MaxHeaderLength()) + header.ARPSize,
+	})
+	h := header.ARP(pkt.NetworkHeader().Push(header.ARPSize))
 	h.SetIPv4OverEthernet()
 	h.SetOp(header.ARPRequest)
 	copy(h.HardwareAddressSender(), linkEP.LinkAddress())
 	copy(h.ProtocolAddressSender(), localAddr)
 	copy(h.ProtocolAddressTarget(), addr)
 
-	return linkEP.WritePacket(r, nil /* gso */, hdr, buffer.VectorisedView{}, ProtocolNumber)
+	return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, pkt)
 }
 
-// ResolveStaticAddress implements stack.LinkAddressResolver.
+// ResolveStaticAddress implements stack.LinkAddressResolver.ResolveStaticAddress.
 func (*protocol) ResolveStaticAddress(addr tcpip.Address) (tcpip.LinkAddress, bool) {
 	if addr == header.IPv4Broadcast {
-		return broadcastMAC, true
+		return header.EthernetBroadcastAddress, true
 	}
 	if header.IsV4MulticastAddress(addr) {
-		// RFC 1112 Host Extensions for IP Multicasting
-		//
-		// 6.4. Extensions to an Ethernet Local Network Module:
-		//
-		// An IP host group address is mapped to an Ethernet multicast
-		// address by placing the low-order 23-bits of the IP address
-		// into the low-order 23 bits of the Ethernet multicast address
-		// 01-00-5E-00-00-00 (hex).
-		return tcpip.LinkAddress([]byte{
-			0x01,
-			0x00,
-			0x5e,
-			addr[header.IPv4AddressSize-3] & 0x7f,
-			addr[header.IPv4AddressSize-2],
-			addr[header.IPv4AddressSize-1],
-		}), true
+		return header.EthernetAddressFromMulticastIPv4Address(addr), true
 	}
-	return "", false
+	return tcpip.LinkAddress([]byte(nil)), false
 }
 
-// SetOption implements NetworkProtocol.
-func (p *protocol) SetOption(option interface{}) *tcpip.Error {
+// SetOption implements stack.NetworkProtocol.SetOption.
+func (*protocol) SetOption(option interface{}) *tcpip.Error {
 	return tcpip.ErrUnknownProtocolOption
 }
 
-// Option implements NetworkProtocol.
-func (p *protocol) Option(option interface{}) *tcpip.Error {
+// Option implements stack.NetworkProtocol.Option.
+func (*protocol) Option(option interface{}) *tcpip.Error {
 	return tcpip.ErrUnknownProtocolOption
 }
 
-var broadcastMAC = tcpip.LinkAddress([]byte{0xff, 0xff, 0xff, 0xff, 0xff, 0xff})
+// Close implements stack.TransportProtocol.Close.
+func (*protocol) Close() {}
+
+// Wait implements stack.TransportProtocol.Wait.
+func (*protocol) Wait() {}
+
+// Parse implements stack.NetworkProtocol.Parse.
+func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNumber, hasTransportHdr bool, ok bool) {
+	_, ok = pkt.NetworkHeader().Consume(header.ARPSize)
+	if !ok {
+		return 0, false, false
+	}
+	return 0, false, true
+}
 
 // NewProtocol returns an ARP network protocol.
 func NewProtocol() stack.NetworkProtocol {
diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go
index 88b57ec03..c2c3e6891 100644
--- a/pkg/tcpip/network/arp/arp_test.go
+++ b/pkg/tcpip/network/arp/arp_test.go
@@ -15,6 +15,7 @@
 package arp_test
 
 import (
+	"context"
 	"strconv"
 	"testing"
 	"time"
@@ -31,10 +32,14 @@ import (
 )
 
 const (
-	stackLinkAddr = tcpip.LinkAddress("\x0a\x0a\x0b\x0b\x0c\x0c")
-	stackAddr1    = tcpip.Address("\x0a\x00\x00\x01")
-	stackAddr2    = tcpip.Address("\x0a\x00\x00\x02")
-	stackAddrBad  = tcpip.Address("\x0a\x00\x00\x03")
+	stackLinkAddr1 = tcpip.LinkAddress("\x0a\x0a\x0b\x0b\x0c\x0c")
+	stackLinkAddr2 = tcpip.LinkAddress("\x0b\x0b\x0c\x0c\x0d\x0d")
+	stackAddr1     = tcpip.Address("\x0a\x00\x00\x01")
+	stackAddr2     = tcpip.Address("\x0a\x00\x00\x02")
+	stackAddrBad   = tcpip.Address("\x0a\x00\x00\x03")
+
+	defaultChannelSize = 1
+	defaultMTU         = 65536
 )
 
 type testContext struct {
@@ -49,8 +54,7 @@ func newTestContext(t *testing.T) *testContext {
 		TransportProtocols: []stack.TransportProtocol{icmp.NewProtocol4()},
 	})
 
-	const defaultMTU = 65536
-	ep := channel.New(256, defaultMTU, stackLinkAddr)
+	ep := channel.New(defaultChannelSize, defaultMTU, stackLinkAddr1)
 	wep := stack.LinkEndpoint(ep)
 
 	if testing.Verbose() {
@@ -83,7 +87,7 @@ func newTestContext(t *testing.T) *testContext {
 }
 
 func (c *testContext) cleanup() {
-	close(c.linkEP.C)
+	c.linkEP.Close()
 }
 
 func TestDirectRequest(t *testing.T) {
@@ -102,21 +106,23 @@ func TestDirectRequest(t *testing.T) {
 
 	inject := func(addr tcpip.Address) {
 		copy(h.ProtocolAddressTarget(), addr)
-		c.linkEP.Inject(arp.ProtocolNumber, v.ToVectorisedView())
+		c.linkEP.InjectInbound(arp.ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{
+			Data: v.ToVectorisedView(),
+		}))
 	}
 
 	for i, address := range []tcpip.Address{stackAddr1, stackAddr2} {
 		t.Run(strconv.Itoa(i), func(t *testing.T) {
 			inject(address)
-			pkt := <-c.linkEP.C
-			if pkt.Proto != arp.ProtocolNumber {
-				t.Fatalf("expected ARP response, got network protocol number %d", pkt.Proto)
+			pi, _ := c.linkEP.ReadContext(context.Background())
+			if pi.Proto != arp.ProtocolNumber {
+				t.Fatalf("expected ARP response, got network protocol number %d", pi.Proto)
 			}
-			rep := header.ARP(pkt.Header)
+			rep := header.ARP(pi.Pkt.NetworkHeader().View())
 			if !rep.IsValid() {
-				t.Fatalf("invalid ARP response len(pkt.Header)=%d", len(pkt.Header))
+				t.Fatalf("invalid ARP response: len = %d; response = %x", len(rep), rep)
 			}
-			if got, want := tcpip.LinkAddress(rep.HardwareAddressSender()), stackLinkAddr; got != want {
+			if got, want := tcpip.LinkAddress(rep.HardwareAddressSender()), stackLinkAddr1; got != want {
 				t.Errorf("got HardwareAddressSender = %s, want = %s", got, want)
 			}
 			if got, want := tcpip.Address(rep.ProtocolAddressSender()), tcpip.Address(h.ProtocolAddressTarget()); got != want {
@@ -132,12 +138,53 @@ func TestDirectRequest(t *testing.T) {
 	}
 
 	inject(stackAddrBad)
-	select {
-	case pkt := <-c.linkEP.C:
+	// Sleep tests are gross, but this will only potentially flake
+	// if there's a bug. If there is no bug this will reliably
+	// succeed.
+	ctx, cancel := context.WithTimeout(context.Background(), 100*time.Millisecond)
+	defer cancel()
+	if pkt, ok := c.linkEP.ReadContext(ctx); ok {
 		t.Errorf("stackAddrBad: unexpected packet sent, Proto=%v", pkt.Proto)
-	case <-time.After(100 * time.Millisecond):
-		// Sleep tests are gross, but this will only potentially flake
-		// if there's a bug. If there is no bug this will reliably
-		// succeed.
+	}
+}
+
+func TestLinkAddressRequest(t *testing.T) {
+	tests := []struct {
+		name           string
+		remoteLinkAddr tcpip.LinkAddress
+		expectLinkAddr tcpip.LinkAddress
+	}{
+		{
+			name:           "Unicast",
+			remoteLinkAddr: stackLinkAddr2,
+			expectLinkAddr: stackLinkAddr2,
+		},
+		{
+			name:           "Multicast",
+			remoteLinkAddr: "",
+			expectLinkAddr: header.EthernetBroadcastAddress,
+		},
+	}
+
+	for _, test := range tests {
+		p := arp.NewProtocol()
+		linkRes, ok := p.(stack.LinkAddressResolver)
+		if !ok {
+			t.Fatal("expected ARP protocol to implement stack.LinkAddressResolver")
+		}
+
+		linkEP := channel.New(defaultChannelSize, defaultMTU, stackLinkAddr1)
+		if err := linkRes.LinkAddressRequest(stackAddr1, stackAddr2, test.remoteLinkAddr, linkEP); err != nil {
+			t.Errorf("got p.LinkAddressRequest(%s, %s, %s, _) = %s", stackAddr1, stackAddr2, test.remoteLinkAddr, err)
+		}
+
+		pkt, ok := linkEP.Read()
+		if !ok {
+			t.Fatal("expected to send a link address request")
+		}
+
+		if got, want := pkt.Route.RemoteLinkAddress, test.expectLinkAddr; got != want {
+			t.Errorf("got pkt.Route.RemoteLinkAddress = %s, want = %s", got, want)
+		}
 	}
 }