summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/header
diff options
context:
space:
mode:
authorGoogler <noreply@google.com>2018-04-27 10:37:02 -0700
committerAdin Scannell <ascannell@google.com>2018-04-28 01:44:26 -0400
commitd02b74a5dcfed4bfc8f2f8e545bca4d2afabb296 (patch)
tree54f95eef73aee6bacbfc736fffc631be2605ed53 /pkg/tcpip/header
parentf70210e742919f40aa2f0934a22f1c9ba6dada62 (diff)
Check in gVisor.
PiperOrigin-RevId: 194583126 Change-Id: Ica1d8821a90f74e7e745962d71801c598c652463
Diffstat (limited to 'pkg/tcpip/header')
-rw-r--r--pkg/tcpip/header/BUILD51
-rw-r--r--pkg/tcpip/header/arp.go90
-rw-r--r--pkg/tcpip/header/checksum.go46
-rw-r--r--pkg/tcpip/header/eth.go64
-rw-r--r--pkg/tcpip/header/gue.go63
-rw-r--r--pkg/tcpip/header/icmpv4.go98
-rw-r--r--pkg/tcpip/header/icmpv6.go111
-rw-r--r--pkg/tcpip/header/interfaces.go82
-rw-r--r--pkg/tcpip/header/ipv4.go251
-rw-r--r--pkg/tcpip/header/ipv6.go191
-rw-r--r--pkg/tcpip/header/ipv6_fragment.go136
-rw-r--r--pkg/tcpip/header/ipversion_test.go57
-rw-r--r--pkg/tcpip/header/tcp.go518
-rw-r--r--pkg/tcpip/header/tcp_test.go134
-rw-r--r--pkg/tcpip/header/udp.go106
15 files changed, 1998 insertions, 0 deletions
diff --git a/pkg/tcpip/header/BUILD b/pkg/tcpip/header/BUILD
new file mode 100644
index 000000000..167ea250d
--- /dev/null
+++ b/pkg/tcpip/header/BUILD
@@ -0,0 +1,51 @@
+package(licenses = ["notice"]) # BSD
+
+load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
+load("//tools/go_stateify:defs.bzl", "go_stateify")
+
+go_stateify(
+ name = "tcp_header_state",
+ srcs = [
+ "tcp.go",
+ ],
+ out = "tcp_header_state.go",
+ package = "header",
+)
+
+go_library(
+ name = "header",
+ srcs = [
+ "arp.go",
+ "checksum.go",
+ "eth.go",
+ "gue.go",
+ "icmpv4.go",
+ "icmpv6.go",
+ "interfaces.go",
+ "ipv4.go",
+ "ipv6.go",
+ "ipv6_fragment.go",
+ "tcp.go",
+ "tcp_header_state.go",
+ "udp.go",
+ ],
+ importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/header",
+ visibility = ["//visibility:public"],
+ deps = [
+ "//pkg/state",
+ "//pkg/tcpip",
+ "//pkg/tcpip/seqnum",
+ ],
+)
+
+go_test(
+ name = "header_test",
+ size = "small",
+ srcs = [
+ "ipversion_test.go",
+ "tcp_test.go",
+ ],
+ deps = [
+ ":header",
+ ],
+)
diff --git a/pkg/tcpip/header/arp.go b/pkg/tcpip/header/arp.go
new file mode 100644
index 000000000..af7f988f3
--- /dev/null
+++ b/pkg/tcpip/header/arp.go
@@ -0,0 +1,90 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import "gvisor.googlesource.com/gvisor/pkg/tcpip"
+
+const (
+ // ARPProtocolNumber is the ARP network protocol number.
+ ARPProtocolNumber tcpip.NetworkProtocolNumber = 0x0806
+
+ // ARPSize is the size of an IPv4-over-Ethernet ARP packet.
+ ARPSize = 2 + 2 + 1 + 1 + 2 + 2*6 + 2*4
+)
+
+// ARPOp is an ARP opcode.
+type ARPOp uint16
+
+// Typical ARP opcodes defined in RFC 826.
+const (
+ ARPRequest ARPOp = 1
+ ARPReply ARPOp = 2
+)
+
+// ARP is an ARP packet stored in a byte array as described in RFC 826.
+type ARP []byte
+
+func (a ARP) hardwareAddressSpace() uint16 { return uint16(a[0])<<8 | uint16(a[1]) }
+func (a ARP) protocolAddressSpace() uint16 { return uint16(a[2])<<8 | uint16(a[3]) }
+func (a ARP) hardwareAddressSize() int { return int(a[4]) }
+func (a ARP) protocolAddressSize() int { return int(a[5]) }
+
+// Op is the ARP opcode.
+func (a ARP) Op() ARPOp { return ARPOp(a[6])<<8 | ARPOp(a[7]) }
+
+// SetOp sets the ARP opcode.
+func (a ARP) SetOp(op ARPOp) {
+ a[6] = uint8(op >> 8)
+ a[7] = uint8(op)
+}
+
+// SetIPv4OverEthernet configures the ARP packet for IPv4-over-Ethernet.
+func (a ARP) SetIPv4OverEthernet() {
+ a[0], a[1] = 0, 1 // htypeEthernet
+ a[2], a[3] = 0x08, 0x00 // IPv4ProtocolNumber
+ a[4] = 6 // macSize
+ a[5] = uint8(IPv4AddressSize)
+}
+
+// HardwareAddressSender is the link address of the sender.
+// It is a view on to the ARP packet so it can be used to set the value.
+func (a ARP) HardwareAddressSender() []byte {
+ const s = 8
+ return a[s : s+6]
+}
+
+// ProtocolAddressSender is the protocol address of the sender.
+// It is a view on to the ARP packet so it can be used to set the value.
+func (a ARP) ProtocolAddressSender() []byte {
+ const s = 8 + 6
+ return a[s : s+4]
+}
+
+// HardwareAddressTarget is the link address of the target.
+// It is a view on to the ARP packet so it can be used to set the value.
+func (a ARP) HardwareAddressTarget() []byte {
+ const s = 8 + 6 + 4
+ return a[s : s+6]
+}
+
+// ProtocolAddressTarget is the protocol address of the target.
+// It is a view on to the ARP packet so it can be used to set the value.
+func (a ARP) ProtocolAddressTarget() []byte {
+ const s = 8 + 6 + 4 + 6
+ return a[s : s+4]
+}
+
+// IsValid reports whether this is an ARP packet for IPv4 over Ethernet.
+func (a ARP) IsValid() bool {
+ if len(a) < ARPSize {
+ return false
+ }
+ const htypeEthernet = 1
+ const macSize = 6
+ return a.hardwareAddressSpace() == htypeEthernet &&
+ a.protocolAddressSpace() == uint16(IPv4ProtocolNumber) &&
+ a.hardwareAddressSize() == macSize &&
+ a.protocolAddressSize() == IPv4AddressSize
+}
diff --git a/pkg/tcpip/header/checksum.go b/pkg/tcpip/header/checksum.go
new file mode 100644
index 000000000..6399b1b95
--- /dev/null
+++ b/pkg/tcpip/header/checksum.go
@@ -0,0 +1,46 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package header provides the implementation of the encoding and decoding of
+// network protocol headers.
+package header
+
+import (
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+// Checksum calculates the checksum (as defined in RFC 1071) of the bytes in the
+// given byte array.
+func Checksum(buf []byte, initial uint16) uint16 {
+ v := uint32(initial)
+
+ l := len(buf)
+ if l&1 != 0 {
+ l--
+ v += uint32(buf[l]) << 8
+ }
+
+ for i := 0; i < l; i += 2 {
+ v += (uint32(buf[i]) << 8) + uint32(buf[i+1])
+ }
+
+ return ChecksumCombine(uint16(v), uint16(v>>16))
+}
+
+// ChecksumCombine combines the two uint16 to form their checksum. This is done
+// by adding them and the carry.
+func ChecksumCombine(a, b uint16) uint16 {
+ v := uint32(a) + uint32(b)
+ return uint16(v + v>>16)
+}
+
+// PseudoHeaderChecksum calculates the pseudo-header checksum for the
+// given destination protocol and network address, ignoring the length
+// field. Pseudo-headers are needed by transport layers when calculating
+// their own checksum.
+func PseudoHeaderChecksum(protocol tcpip.TransportProtocolNumber, srcAddr tcpip.Address, dstAddr tcpip.Address) uint16 {
+ xsum := Checksum([]byte(srcAddr), 0)
+ xsum = Checksum([]byte(dstAddr), xsum)
+ return Checksum([]byte{0, uint8(protocol)}, xsum)
+}
diff --git a/pkg/tcpip/header/eth.go b/pkg/tcpip/header/eth.go
new file mode 100644
index 000000000..23b7efdfc
--- /dev/null
+++ b/pkg/tcpip/header/eth.go
@@ -0,0 +1,64 @@
+// Copyright 2017 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+const (
+ dstMAC = 0
+ srcMAC = 6
+ ethType = 12
+)
+
+// EthernetFields contains the fields of an ethernet frame header. It is used to
+// describe the fields of a frame that needs to be encoded.
+type EthernetFields struct {
+ // SrcAddr is the "MAC source" field of an ethernet frame header.
+ SrcAddr tcpip.LinkAddress
+
+ // DstAddr is the "MAC destination" field of an ethernet frame header.
+ DstAddr tcpip.LinkAddress
+
+ // Type is the "ethertype" field of an ethernet frame header.
+ Type tcpip.NetworkProtocolNumber
+}
+
+// Ethernet represents an ethernet frame header stored in a byte array.
+type Ethernet []byte
+
+const (
+ // EthernetMinimumSize is the minimum size of a valid ethernet frame.
+ EthernetMinimumSize = 14
+
+ // EthernetAddressSize is the size, in bytes, of an ethernet address.
+ EthernetAddressSize = 6
+)
+
+// SourceAddress returns the "MAC source" field of the ethernet frame header.
+func (b Ethernet) SourceAddress() tcpip.LinkAddress {
+ return tcpip.LinkAddress(b[srcMAC:][:EthernetAddressSize])
+}
+
+// DestinationAddress returns the "MAC destination" field of the ethernet frame
+// header.
+func (b Ethernet) DestinationAddress() tcpip.LinkAddress {
+ return tcpip.LinkAddress(b[dstMAC:][:EthernetAddressSize])
+}
+
+// Type returns the "ethertype" field of the ethernet frame header.
+func (b Ethernet) Type() tcpip.NetworkProtocolNumber {
+ return tcpip.NetworkProtocolNumber(binary.BigEndian.Uint16(b[ethType:]))
+}
+
+// Encode encodes all the fields of the ethernet frame header.
+func (b Ethernet) Encode(e *EthernetFields) {
+ binary.BigEndian.PutUint16(b[ethType:], uint16(e.Type))
+ copy(b[srcMAC:][:EthernetAddressSize], e.SrcAddr)
+ copy(b[dstMAC:][:EthernetAddressSize], e.DstAddr)
+}
diff --git a/pkg/tcpip/header/gue.go b/pkg/tcpip/header/gue.go
new file mode 100644
index 000000000..a069fb669
--- /dev/null
+++ b/pkg/tcpip/header/gue.go
@@ -0,0 +1,63 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+const (
+ typeHLen = 0
+ encapProto = 1
+)
+
+// GUEFields contains the fields of a GUE packet. It is used to describe the
+// fields of a packet that needs to be encoded.
+type GUEFields struct {
+ // Type is the "type" field of the GUE header.
+ Type uint8
+
+ // Control is the "control" field of the GUE header.
+ Control bool
+
+ // HeaderLength is the "header length" field of the GUE header. It must
+ // be at least 4 octets, and a multiple of 4 as well.
+ HeaderLength uint8
+
+ // Protocol is the "protocol" field of the GUE header. This is one of
+ // the IPPROTO_* values.
+ Protocol uint8
+}
+
+// GUE represents a Generic UDP Encapsulation header stored in a byte array, the
+// fields are described in https://tools.ietf.org/html/draft-ietf-nvo3-gue-01.
+type GUE []byte
+
+const (
+ // GUEMinimumSize is the minimum size of a valid GUE packet.
+ GUEMinimumSize = 4
+)
+
+// TypeAndControl returns the GUE packet type (top 3 bits of the first byte,
+// which includes the control bit).
+func (b GUE) TypeAndControl() uint8 {
+ return b[typeHLen] >> 5
+}
+
+// HeaderLength returns the total length of the GUE header.
+func (b GUE) HeaderLength() uint8 {
+ return 4 + 4*(b[typeHLen]&0x1f)
+}
+
+// Protocol returns the protocol field of the GUE header.
+func (b GUE) Protocol() uint8 {
+ return b[encapProto]
+}
+
+// Encode encodes all the fields of the GUE header.
+func (b GUE) Encode(i *GUEFields) {
+ ctl := uint8(0)
+ if i.Control {
+ ctl = 1 << 5
+ }
+ b[typeHLen] = ctl | i.Type<<6 | (i.HeaderLength-4)/4
+ b[encapProto] = i.Protocol
+}
diff --git a/pkg/tcpip/header/icmpv4.go b/pkg/tcpip/header/icmpv4.go
new file mode 100644
index 000000000..9f1ad38fc
--- /dev/null
+++ b/pkg/tcpip/header/icmpv4.go
@@ -0,0 +1,98 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+// ICMPv4 represents an ICMPv4 header stored in a byte array.
+type ICMPv4 []byte
+
+const (
+ // ICMPv4MinimumSize is the minimum size of a valid ICMP packet.
+ ICMPv4MinimumSize = 4
+
+ // ICMPv4EchoMinimumSize is the minimum size of a valid ICMP echo packet.
+ ICMPv4EchoMinimumSize = 6
+
+ // ICMPv4DstUnreachableMinimumSize is the minimum size of a valid ICMP
+ // destination unreachable packet.
+ ICMPv4DstUnreachableMinimumSize = ICMPv4MinimumSize + 4
+
+ // ICMPv4ProtocolNumber is the ICMP transport protocol number.
+ ICMPv4ProtocolNumber tcpip.TransportProtocolNumber = 1
+)
+
+// ICMPv4Type is the ICMP type field described in RFC 792.
+type ICMPv4Type byte
+
+// Typical values of ICMPv4Type defined in RFC 792.
+const (
+ ICMPv4EchoReply ICMPv4Type = 0
+ ICMPv4DstUnreachable ICMPv4Type = 3
+ ICMPv4SrcQuench ICMPv4Type = 4
+ ICMPv4Redirect ICMPv4Type = 5
+ ICMPv4Echo ICMPv4Type = 8
+ ICMPv4TimeExceeded ICMPv4Type = 11
+ ICMPv4ParamProblem ICMPv4Type = 12
+ ICMPv4Timestamp ICMPv4Type = 13
+ ICMPv4TimestampReply ICMPv4Type = 14
+ ICMPv4InfoRequest ICMPv4Type = 15
+ ICMPv4InfoReply ICMPv4Type = 16
+)
+
+// Values for ICMP code as defined in RFC 792.
+const (
+ ICMPv4PortUnreachable = 3
+ ICMPv4FragmentationNeeded = 4
+)
+
+// Type is the ICMP type field.
+func (b ICMPv4) Type() ICMPv4Type { return ICMPv4Type(b[0]) }
+
+// SetType sets the ICMP type field.
+func (b ICMPv4) SetType(t ICMPv4Type) { b[0] = byte(t) }
+
+// Code is the ICMP code field. Its meaning depends on the value of Type.
+func (b ICMPv4) Code() byte { return b[1] }
+
+// SetCode sets the ICMP code field.
+func (b ICMPv4) SetCode(c byte) { b[1] = c }
+
+// Checksum is the ICMP checksum field.
+func (b ICMPv4) Checksum() uint16 {
+ return binary.BigEndian.Uint16(b[2:])
+}
+
+// SetChecksum sets the ICMP checksum field.
+func (b ICMPv4) SetChecksum(checksum uint16) {
+ binary.BigEndian.PutUint16(b[2:], checksum)
+}
+
+// SourcePort implements Transport.SourcePort.
+func (ICMPv4) SourcePort() uint16 {
+ return 0
+}
+
+// DestinationPort implements Transport.DestinationPort.
+func (ICMPv4) DestinationPort() uint16 {
+ return 0
+}
+
+// SetSourcePort implements Transport.SetSourcePort.
+func (ICMPv4) SetSourcePort(uint16) {
+}
+
+// SetDestinationPort implements Transport.SetDestinationPort.
+func (ICMPv4) SetDestinationPort(uint16) {
+}
+
+// Payload implements Transport.Payload.
+func (b ICMPv4) Payload() []byte {
+ return b[ICMPv4MinimumSize:]
+}
diff --git a/pkg/tcpip/header/icmpv6.go b/pkg/tcpip/header/icmpv6.go
new file mode 100644
index 000000000..a061cd02b
--- /dev/null
+++ b/pkg/tcpip/header/icmpv6.go
@@ -0,0 +1,111 @@
+// Copyright 2017 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+// ICMPv6 represents an ICMPv6 header stored in a byte array.
+type ICMPv6 []byte
+
+const (
+ // ICMPv6MinimumSize is the minimum size of a valid ICMP packet.
+ ICMPv6MinimumSize = 4
+
+ // ICMPv6ProtocolNumber is the ICMP transport protocol number.
+ ICMPv6ProtocolNumber tcpip.TransportProtocolNumber = 58
+
+ // ICMPv6NeighborSolicitMinimumSize is the minimum size of a
+ // neighbor solicitation packet.
+ ICMPv6NeighborSolicitMinimumSize = ICMPv6MinimumSize + 4 + 16
+
+ // ICMPv6NeighborAdvertSize is size of a neighbor advertisement.
+ ICMPv6NeighborAdvertSize = 32
+
+ // ICMPv6EchoMinimumSize is the minimum size of a valid ICMP echo packet.
+ ICMPv6EchoMinimumSize = 8
+
+ // ICMPv6DstUnreachableMinimumSize is the minimum size of a valid ICMP
+ // destination unreachable packet.
+ ICMPv6DstUnreachableMinimumSize = ICMPv6MinimumSize + 4
+
+ // ICMPv6PacketTooBigMinimumSize is the minimum size of a valid ICMP
+ // packet-too-big packet.
+ ICMPv6PacketTooBigMinimumSize = ICMPv6MinimumSize + 4
+)
+
+// ICMPv6Type is the ICMP type field described in RFC 4443 and friends.
+type ICMPv6Type byte
+
+// Typical values of ICMPv6Type defined in RFC 4443.
+const (
+ ICMPv6DstUnreachable ICMPv6Type = 1
+ ICMPv6PacketTooBig ICMPv6Type = 2
+ ICMPv6TimeExceeded ICMPv6Type = 3
+ ICMPv6ParamProblem ICMPv6Type = 4
+ ICMPv6EchoRequest ICMPv6Type = 128
+ ICMPv6EchoReply ICMPv6Type = 129
+
+ // Neighbor Discovery Protocol (NDP) messages, see RFC 4861.
+
+ ICMPv6RouterSolicit ICMPv6Type = 133
+ ICMPv6RouterAdvert ICMPv6Type = 134
+ ICMPv6NeighborSolicit ICMPv6Type = 135
+ ICMPv6NeighborAdvert ICMPv6Type = 136
+ ICMPv6RedirectMsg ICMPv6Type = 137
+)
+
+// Values for ICMP code as defined in RFC 4443.
+const (
+ ICMPv6PortUnreachable = 4
+)
+
+// Type is the ICMP type field.
+func (b ICMPv6) Type() ICMPv6Type { return ICMPv6Type(b[0]) }
+
+// SetType sets the ICMP type field.
+func (b ICMPv6) SetType(t ICMPv6Type) { b[0] = byte(t) }
+
+// Code is the ICMP code field. Its meaning depends on the value of Type.
+func (b ICMPv6) Code() byte { return b[1] }
+
+// SetCode sets the ICMP code field.
+func (b ICMPv6) SetCode(c byte) { b[1] = c }
+
+// Checksum is the ICMP checksum field.
+func (b ICMPv6) Checksum() uint16 {
+ return binary.BigEndian.Uint16(b[2:])
+}
+
+// SetChecksum calculates and sets the ICMP checksum field.
+func (b ICMPv6) SetChecksum(checksum uint16) {
+ binary.BigEndian.PutUint16(b[2:], checksum)
+}
+
+// SourcePort implements Transport.SourcePort.
+func (ICMPv6) SourcePort() uint16 {
+ return 0
+}
+
+// DestinationPort implements Transport.DestinationPort.
+func (ICMPv6) DestinationPort() uint16 {
+ return 0
+}
+
+// SetSourcePort implements Transport.SetSourcePort.
+func (ICMPv6) SetSourcePort(uint16) {
+}
+
+// SetDestinationPort implements Transport.SetDestinationPort.
+func (ICMPv6) SetDestinationPort(uint16) {
+}
+
+// Payload implements Transport.Payload.
+func (b ICMPv6) Payload() []byte {
+ return b[ICMPv6MinimumSize:]
+}
diff --git a/pkg/tcpip/header/interfaces.go b/pkg/tcpip/header/interfaces.go
new file mode 100644
index 000000000..a92286761
--- /dev/null
+++ b/pkg/tcpip/header/interfaces.go
@@ -0,0 +1,82 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+const (
+ // MaxIPPacketSize is the maximum supported IP packet size, excluding
+ // jumbograms. The maximum IPv4 packet size is 64k-1 (total size must fit
+ // in 16 bits). For IPv6, the payload max size (excluding jumbograms) is
+ // 64k-1 (also needs to fit in 16 bits). So we use 64k - 1 + 2 * m, where
+ // m is the minimum IPv6 header size; we leave room for some potential
+ // IP options.
+ MaxIPPacketSize = 0xffff + 2*IPv6MinimumSize
+)
+
+// Transport offers generic methods to query and/or update the fields of the
+// header of a transport protocol buffer.
+type Transport interface {
+ // SourcePort returns the value of the "source port" field.
+ SourcePort() uint16
+
+ // Destination returns the value of the "destination port" field.
+ DestinationPort() uint16
+
+ // Checksum returns the value of the "checksum" field.
+ Checksum() uint16
+
+ // SetSourcePort sets the value of the "source port" field.
+ SetSourcePort(uint16)
+
+ // SetDestinationPort sets the value of the "destination port" field.
+ SetDestinationPort(uint16)
+
+ // SetChecksum sets the value of the "checksum" field.
+ SetChecksum(uint16)
+
+ // Payload returns the data carried in the transport buffer.
+ Payload() []byte
+}
+
+// Network offers generic methods to query and/or update the fields of the
+// header of a network protocol buffer.
+type Network interface {
+ // SourceAddress returns the value of the "source address" field.
+ SourceAddress() tcpip.Address
+
+ // DestinationAddress returns the value of the "destination address"
+ // field.
+ DestinationAddress() tcpip.Address
+
+ // Checksum returns the value of the "checksum" field.
+ Checksum() uint16
+
+ // SetSourceAddress sets the value of the "source address" field.
+ SetSourceAddress(tcpip.Address)
+
+ // SetDestinationAddress sets the value of the "destination address"
+ // field.
+ SetDestinationAddress(tcpip.Address)
+
+ // SetChecksum sets the value of the "checksum" field.
+ SetChecksum(uint16)
+
+ // TransportProtocol returns the number of the transport protocol
+ // stored in the payload.
+ TransportProtocol() tcpip.TransportProtocolNumber
+
+ // Payload returns a byte slice containing the payload of the network
+ // packet.
+ Payload() []byte
+
+ // TOS returns the values of the "type of service" and "flow label" fields.
+ TOS() (uint8, uint32)
+
+ // SetTOS sets the values of the "type of service" and "flow label" fields.
+ SetTOS(t uint8, l uint32)
+}
diff --git a/pkg/tcpip/header/ipv4.go b/pkg/tcpip/header/ipv4.go
new file mode 100644
index 000000000..cb0d42093
--- /dev/null
+++ b/pkg/tcpip/header/ipv4.go
@@ -0,0 +1,251 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+const (
+ versIHL = 0
+ tos = 1
+ totalLen = 2
+ id = 4
+ flagsFO = 6
+ ttl = 8
+ protocol = 9
+ checksum = 10
+ srcAddr = 12
+ dstAddr = 16
+)
+
+// IPv4Fields contains the fields of an IPv4 packet. It is used to describe the
+// fields of a packet that needs to be encoded.
+type IPv4Fields struct {
+ // IHL is the "internet header length" field of an IPv4 packet.
+ IHL uint8
+
+ // TOS is the "type of service" field of an IPv4 packet.
+ TOS uint8
+
+ // TotalLength is the "total length" field of an IPv4 packet.
+ TotalLength uint16
+
+ // ID is the "identification" field of an IPv4 packet.
+ ID uint16
+
+ // Flags is the "flags" field of an IPv4 packet.
+ Flags uint8
+
+ // FragmentOffset is the "fragment offset" field of an IPv4 packet.
+ FragmentOffset uint16
+
+ // TTL is the "time to live" field of an IPv4 packet.
+ TTL uint8
+
+ // Protocol is the "protocol" field of an IPv4 packet.
+ Protocol uint8
+
+ // Checksum is the "checksum" field of an IPv4 packet.
+ Checksum uint16
+
+ // SrcAddr is the "source ip address" of an IPv4 packet.
+ SrcAddr tcpip.Address
+
+ // DstAddr is the "destination ip address" of an IPv4 packet.
+ DstAddr tcpip.Address
+}
+
+// IPv4 represents an ipv4 header stored in a byte array.
+// Most of the methods of IPv4 access to the underlying slice without
+// checking the boundaries and could panic because of 'index out of range'.
+// Always call IsValid() to validate an instance of IPv4 before using other methods.
+type IPv4 []byte
+
+const (
+ // IPv4MinimumSize is the minimum size of a valid IPv4 packet.
+ IPv4MinimumSize = 20
+
+ // IPv4MaximumHeaderSize is the maximum size of an IPv4 header. Given
+ // that there are only 4 bits to represents the header length in 32-bit
+ // units, the header cannot exceed 15*4 = 60 bytes.
+ IPv4MaximumHeaderSize = 60
+
+ // IPv4AddressSize is the size, in bytes, of an IPv4 address.
+ IPv4AddressSize = 4
+
+ // IPv4ProtocolNumber is IPv4's network protocol number.
+ IPv4ProtocolNumber tcpip.NetworkProtocolNumber = 0x0800
+
+ // IPv4Version is the version of the ipv4 procotol.
+ IPv4Version = 4
+)
+
+// Flags that may be set in an IPv4 packet.
+const (
+ IPv4FlagMoreFragments = 1 << iota
+ IPv4FlagDontFragment
+)
+
+// IPVersion returns the version of IP used in the given packet. It returns -1
+// if the packet is not large enough to contain the version field.
+func IPVersion(b []byte) int {
+ // Length must be at least offset+length of version field.
+ if len(b) < versIHL+1 {
+ return -1
+ }
+ return int(b[versIHL] >> 4)
+}
+
+// HeaderLength returns the value of the "header length" field of the ipv4
+// header.
+func (b IPv4) HeaderLength() uint8 {
+ return (b[versIHL] & 0xf) * 4
+}
+
+// ID returns the value of the identifier field of the ipv4 header.
+func (b IPv4) ID() uint16 {
+ return binary.BigEndian.Uint16(b[id:])
+}
+
+// Protocol returns the value of the protocol field of the ipv4 header.
+func (b IPv4) Protocol() uint8 {
+ return b[protocol]
+}
+
+// Flags returns the "flags" field of the ipv4 header.
+func (b IPv4) Flags() uint8 {
+ return uint8(binary.BigEndian.Uint16(b[flagsFO:]) >> 13)
+}
+
+// TTL returns the "TTL" field of the ipv4 header.
+func (b IPv4) TTL() uint8 {
+ return b[ttl]
+}
+
+// FragmentOffset returns the "fragment offset" field of the ipv4 header.
+func (b IPv4) FragmentOffset() uint16 {
+ return binary.BigEndian.Uint16(b[flagsFO:]) << 3
+}
+
+// TotalLength returns the "total length" field of the ipv4 header.
+func (b IPv4) TotalLength() uint16 {
+ return binary.BigEndian.Uint16(b[totalLen:])
+}
+
+// Checksum returns the checksum field of the ipv4 header.
+func (b IPv4) Checksum() uint16 {
+ return binary.BigEndian.Uint16(b[checksum:])
+}
+
+// SourceAddress returns the "source address" field of the ipv4 header.
+func (b IPv4) SourceAddress() tcpip.Address {
+ return tcpip.Address(b[srcAddr : srcAddr+IPv4AddressSize])
+}
+
+// DestinationAddress returns the "destination address" field of the ipv4
+// header.
+func (b IPv4) DestinationAddress() tcpip.Address {
+ return tcpip.Address(b[dstAddr : dstAddr+IPv4AddressSize])
+}
+
+// TransportProtocol implements Network.TransportProtocol.
+func (b IPv4) TransportProtocol() tcpip.TransportProtocolNumber {
+ return tcpip.TransportProtocolNumber(b.Protocol())
+}
+
+// Payload implements Network.Payload.
+func (b IPv4) Payload() []byte {
+ return b[b.HeaderLength():][:b.PayloadLength()]
+}
+
+// PayloadLength returns the length of the payload portion of the ipv4 packet.
+func (b IPv4) PayloadLength() uint16 {
+ return b.TotalLength() - uint16(b.HeaderLength())
+}
+
+// TOS returns the "type of service" field of the ipv4 header.
+func (b IPv4) TOS() (uint8, uint32) {
+ return b[tos], 0
+}
+
+// SetTOS sets the "type of service" field of the ipv4 header.
+func (b IPv4) SetTOS(v uint8, _ uint32) {
+ b[tos] = v
+}
+
+// SetTotalLength sets the "total length" field of the ipv4 header.
+func (b IPv4) SetTotalLength(totalLength uint16) {
+ binary.BigEndian.PutUint16(b[totalLen:], totalLength)
+}
+
+// SetChecksum sets the checksum field of the ipv4 header.
+func (b IPv4) SetChecksum(v uint16) {
+ binary.BigEndian.PutUint16(b[checksum:], v)
+}
+
+// SetFlagsFragmentOffset sets the "flags" and "fragment offset" fields of the
+// ipv4 header.
+func (b IPv4) SetFlagsFragmentOffset(flags uint8, offset uint16) {
+ v := (uint16(flags) << 13) | (offset >> 3)
+ binary.BigEndian.PutUint16(b[flagsFO:], v)
+}
+
+// SetSourceAddress sets the "source address" field of the ipv4 header.
+func (b IPv4) SetSourceAddress(addr tcpip.Address) {
+ copy(b[srcAddr:srcAddr+IPv4AddressSize], addr)
+}
+
+// SetDestinationAddress sets the "destination address" field of the ipv4
+// header.
+func (b IPv4) SetDestinationAddress(addr tcpip.Address) {
+ copy(b[dstAddr:dstAddr+IPv4AddressSize], addr)
+}
+
+// CalculateChecksum calculates the checksum of the ipv4 header.
+func (b IPv4) CalculateChecksum() uint16 {
+ return Checksum(b[:b.HeaderLength()], 0)
+}
+
+// Encode encodes all the fields of the ipv4 header.
+func (b IPv4) Encode(i *IPv4Fields) {
+ b[versIHL] = (4 << 4) | ((i.IHL / 4) & 0xf)
+ b[tos] = i.TOS
+ b.SetTotalLength(i.TotalLength)
+ binary.BigEndian.PutUint16(b[id:], i.ID)
+ b.SetFlagsFragmentOffset(i.Flags, i.FragmentOffset)
+ b[ttl] = i.TTL
+ b[protocol] = i.Protocol
+ b.SetChecksum(i.Checksum)
+ copy(b[srcAddr:srcAddr+IPv4AddressSize], i.SrcAddr)
+ copy(b[dstAddr:dstAddr+IPv4AddressSize], i.DstAddr)
+}
+
+// EncodePartial updates the total length and checksum fields of ipv4 header,
+// taking in the partial checksum, which is the checksum of the header without
+// the total length and checksum fields. It is useful in cases when similar
+// packets are produced.
+func (b IPv4) EncodePartial(partialChecksum, totalLength uint16) {
+ b.SetTotalLength(totalLength)
+ checksum := Checksum(b[totalLen:totalLen+2], partialChecksum)
+ b.SetChecksum(^checksum)
+}
+
+// IsValid performs basic validation on the packet.
+func (b IPv4) IsValid(pktSize int) bool {
+ if len(b) < IPv4MinimumSize {
+ return false
+ }
+
+ hlen := int(b.HeaderLength())
+ tlen := int(b.TotalLength())
+ if hlen > tlen || tlen > pktSize {
+ return false
+ }
+
+ return true
+}
diff --git a/pkg/tcpip/header/ipv6.go b/pkg/tcpip/header/ipv6.go
new file mode 100644
index 000000000..d8dc138b3
--- /dev/null
+++ b/pkg/tcpip/header/ipv6.go
@@ -0,0 +1,191 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+const (
+ versTCFL = 0
+ payloadLen = 4
+ nextHdr = 6
+ hopLimit = 7
+ v6SrcAddr = 8
+ v6DstAddr = 24
+)
+
+// IPv6Fields contains the fields of an IPv6 packet. It is used to describe the
+// fields of a packet that needs to be encoded.
+type IPv6Fields struct {
+ // TrafficClass is the "traffic class" field of an IPv6 packet.
+ TrafficClass uint8
+
+ // FlowLabel is the "flow label" field of an IPv6 packet.
+ FlowLabel uint32
+
+ // PayloadLength is the "payload length" field of an IPv6 packet.
+ PayloadLength uint16
+
+ // NextHeader is the "next header" field of an IPv6 packet.
+ NextHeader uint8
+
+ // HopLimit is the "hop limit" field of an IPv6 packet.
+ HopLimit uint8
+
+ // SrcAddr is the "source ip address" of an IPv6 packet.
+ SrcAddr tcpip.Address
+
+ // DstAddr is the "destination ip address" of an IPv6 packet.
+ DstAddr tcpip.Address
+}
+
+// IPv6 represents an ipv6 header stored in a byte array.
+// Most of the methods of IPv6 access to the underlying slice without
+// checking the boundaries and could panic because of 'index out of range'.
+// Always call IsValid() to validate an instance of IPv6 before using other methods.
+type IPv6 []byte
+
+const (
+ // IPv6MinimumSize is the minimum size of a valid IPv6 packet.
+ IPv6MinimumSize = 40
+
+ // IPv6AddressSize is the size, in bytes, of an IPv6 address.
+ IPv6AddressSize = 16
+
+ // IPv6ProtocolNumber is IPv6's network protocol number.
+ IPv6ProtocolNumber tcpip.NetworkProtocolNumber = 0x86dd
+
+ // IPv6Version is the version of the ipv6 procotol.
+ IPv6Version = 6
+
+ // IPv6MinimumMTU is the minimum MTU required by IPv6, per RFC 2460,
+ // section 5.
+ IPv6MinimumMTU = 1280
+)
+
+// PayloadLength returns the value of the "payload length" field of the ipv6
+// header.
+func (b IPv6) PayloadLength() uint16 {
+ return binary.BigEndian.Uint16(b[payloadLen:])
+}
+
+// HopLimit returns the value of the "hop limit" field of the ipv6 header.
+func (b IPv6) HopLimit() uint8 {
+ return b[hopLimit]
+}
+
+// NextHeader returns the value of the "next header" field of the ipv6 header.
+func (b IPv6) NextHeader() uint8 {
+ return b[nextHdr]
+}
+
+// TransportProtocol implements Network.TransportProtocol.
+func (b IPv6) TransportProtocol() tcpip.TransportProtocolNumber {
+ return tcpip.TransportProtocolNumber(b.NextHeader())
+}
+
+// Payload implements Network.Payload.
+func (b IPv6) Payload() []byte {
+ return b[IPv6MinimumSize:][:b.PayloadLength()]
+}
+
+// SourceAddress returns the "source address" field of the ipv6 header.
+func (b IPv6) SourceAddress() tcpip.Address {
+ return tcpip.Address(b[v6SrcAddr : v6SrcAddr+IPv6AddressSize])
+}
+
+// DestinationAddress returns the "destination address" field of the ipv6
+// header.
+func (b IPv6) DestinationAddress() tcpip.Address {
+ return tcpip.Address(b[v6DstAddr : v6DstAddr+IPv6AddressSize])
+}
+
+// Checksum implements Network.Checksum. Given that IPv6 doesn't have a
+// checksum, it just returns 0.
+func (IPv6) Checksum() uint16 {
+ return 0
+}
+
+// TOS returns the "traffic class" and "flow label" fields of the ipv6 header.
+func (b IPv6) TOS() (uint8, uint32) {
+ v := binary.BigEndian.Uint32(b[versTCFL:])
+ return uint8(v >> 20), v & 0xfffff
+}
+
+// SetTOS sets the "traffic class" and "flow label" fields of the ipv6 header.
+func (b IPv6) SetTOS(t uint8, l uint32) {
+ vtf := (6 << 28) | (uint32(t) << 20) | (l & 0xfffff)
+ binary.BigEndian.PutUint32(b[versTCFL:], vtf)
+}
+
+// SetPayloadLength sets the "payload length" field of the ipv6 header.
+func (b IPv6) SetPayloadLength(payloadLength uint16) {
+ binary.BigEndian.PutUint16(b[payloadLen:], payloadLength)
+}
+
+// SetSourceAddress sets the "source address" field of the ipv6 header.
+func (b IPv6) SetSourceAddress(addr tcpip.Address) {
+ copy(b[v6SrcAddr:v6SrcAddr+IPv6AddressSize], addr)
+}
+
+// SetDestinationAddress sets the "destination address" field of the ipv6
+// header.
+func (b IPv6) SetDestinationAddress(addr tcpip.Address) {
+ copy(b[v6DstAddr:v6DstAddr+IPv6AddressSize], addr)
+}
+
+// SetNextHeader sets the value of the "next header" field of the ipv6 header.
+func (b IPv6) SetNextHeader(v uint8) {
+ b[nextHdr] = v
+}
+
+// SetChecksum implements Network.SetChecksum. Given that IPv6 doesn't have a
+// checksum, it is empty.
+func (IPv6) SetChecksum(uint16) {
+}
+
+// Encode encodes all the fields of the ipv6 header.
+func (b IPv6) Encode(i *IPv6Fields) {
+ b.SetTOS(i.TrafficClass, i.FlowLabel)
+ b.SetPayloadLength(i.PayloadLength)
+ b[nextHdr] = i.NextHeader
+ b[hopLimit] = i.HopLimit
+ copy(b[v6SrcAddr:v6SrcAddr+IPv6AddressSize], i.SrcAddr)
+ copy(b[v6DstAddr:v6DstAddr+IPv6AddressSize], i.DstAddr)
+}
+
+// IsValid performs basic validation on the packet.
+func (b IPv6) IsValid(pktSize int) bool {
+ if len(b) < IPv6MinimumSize {
+ return false
+ }
+
+ dlen := int(b.PayloadLength())
+ if dlen > pktSize-IPv6MinimumSize {
+ return false
+ }
+
+ return true
+}
+
+// IsV4MappedAddress determines if the provided address is an IPv4 mapped
+// address by checking if its prefix is 0:0:0:0:0:ffff::/96.
+func IsV4MappedAddress(addr tcpip.Address) bool {
+ if len(addr) != IPv6AddressSize {
+ return false
+ }
+
+ const prefix = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xff\xff"
+ for i := 0; i < len(prefix); i++ {
+ if prefix[i] != addr[i] {
+ return false
+ }
+ }
+
+ return true
+}
diff --git a/pkg/tcpip/header/ipv6_fragment.go b/pkg/tcpip/header/ipv6_fragment.go
new file mode 100644
index 000000000..04aa5c7b8
--- /dev/null
+++ b/pkg/tcpip/header/ipv6_fragment.go
@@ -0,0 +1,136 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+const (
+ nextHdrFrag = 0
+ fragOff = 2
+ more = 3
+ idV6 = 4
+)
+
+// IPv6FragmentFields contains the fields of an IPv6 fragment. It is used to describe the
+// fields of a packet that needs to be encoded.
+type IPv6FragmentFields struct {
+ // NextHeader is the "next header" field of an IPv6 fragment.
+ NextHeader uint8
+
+ // FragmentOffset is the "fragment offset" field of an IPv6 fragment.
+ FragmentOffset uint16
+
+ // M is the "more" field of an IPv6 fragment.
+ M bool
+
+ // Identification is the "identification" field of an IPv6 fragment.
+ Identification uint32
+}
+
+// IPv6Fragment represents an ipv6 fragment header stored in a byte array.
+// Most of the methods of IPv6Fragment access to the underlying slice without
+// checking the boundaries and could panic because of 'index out of range'.
+// Always call IsValid() to validate an instance of IPv6Fragment before using other methods.
+type IPv6Fragment []byte
+
+const (
+ // IPv6FragmentHeader header is the number used to specify that the next
+ // header is a fragment header, per RFC 2460.
+ IPv6FragmentHeader = 44
+
+ // IPv6FragmentHeaderSize is the size of the fragment header.
+ IPv6FragmentHeaderSize = 8
+)
+
+// Encode encodes all the fields of the ipv6 fragment.
+func (b IPv6Fragment) Encode(i *IPv6FragmentFields) {
+ b[nextHdrFrag] = i.NextHeader
+ binary.BigEndian.PutUint16(b[fragOff:], i.FragmentOffset<<3)
+ if i.M {
+ b[more] |= 1
+ }
+ binary.BigEndian.PutUint32(b[idV6:], i.Identification)
+}
+
+// IsValid performs basic validation on the fragment header.
+func (b IPv6Fragment) IsValid() bool {
+ return len(b) >= IPv6FragmentHeaderSize
+}
+
+// NextHeader returns the value of the "next header" field of the ipv6 fragment.
+func (b IPv6Fragment) NextHeader() uint8 {
+ return b[nextHdrFrag]
+}
+
+// FragmentOffset returns the "fragment offset" field of the ipv6 fragment.
+func (b IPv6Fragment) FragmentOffset() uint16 {
+ return binary.BigEndian.Uint16(b[fragOff:]) >> 3
+}
+
+// More returns the "more" field of the ipv6 fragment.
+func (b IPv6Fragment) More() bool {
+ return b[more]&1 > 0
+}
+
+// Payload implements Network.Payload.
+func (b IPv6Fragment) Payload() []byte {
+ return b[IPv6FragmentHeaderSize:]
+}
+
+// ID returns the value of the identifier field of the ipv6 fragment.
+func (b IPv6Fragment) ID() uint32 {
+ return binary.BigEndian.Uint32(b[idV6:])
+}
+
+// TransportProtocol implements Network.TransportProtocol.
+func (b IPv6Fragment) TransportProtocol() tcpip.TransportProtocolNumber {
+ return tcpip.TransportProtocolNumber(b.NextHeader())
+}
+
+// The functions below have been added only to satisfy the Network interface.
+
+// Checksum is not supported by IPv6Fragment.
+func (b IPv6Fragment) Checksum() uint16 {
+ panic("not supported")
+}
+
+// SourceAddress is not supported by IPv6Fragment.
+func (b IPv6Fragment) SourceAddress() tcpip.Address {
+ panic("not supported")
+}
+
+// DestinationAddress is not supported by IPv6Fragment.
+func (b IPv6Fragment) DestinationAddress() tcpip.Address {
+ panic("not supported")
+}
+
+// SetSourceAddress is not supported by IPv6Fragment.
+func (b IPv6Fragment) SetSourceAddress(tcpip.Address) {
+ panic("not supported")
+}
+
+// SetDestinationAddress is not supported by IPv6Fragment.
+func (b IPv6Fragment) SetDestinationAddress(tcpip.Address) {
+ panic("not supported")
+}
+
+// SetChecksum is not supported by IPv6Fragment.
+func (b IPv6Fragment) SetChecksum(uint16) {
+ panic("not supported")
+}
+
+// TOS is not supported by IPv6Fragment.
+func (b IPv6Fragment) TOS() (uint8, uint32) {
+ panic("not supported")
+}
+
+// SetTOS is not supported by IPv6Fragment.
+func (b IPv6Fragment) SetTOS(t uint8, l uint32) {
+ panic("not supported")
+}
diff --git a/pkg/tcpip/header/ipversion_test.go b/pkg/tcpip/header/ipversion_test.go
new file mode 100644
index 000000000..5f3956160
--- /dev/null
+++ b/pkg/tcpip/header/ipversion_test.go
@@ -0,0 +1,57 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header_test
+
+import (
+ "testing"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip/header"
+)
+
+func TestIPv4(t *testing.T) {
+ b := header.IPv4(make([]byte, header.IPv4MinimumSize))
+ b.Encode(&header.IPv4Fields{})
+
+ const want = header.IPv4Version
+ if v := header.IPVersion(b); v != want {
+ t.Fatalf("Bad version, want %v, got %v", want, v)
+ }
+}
+
+func TestIPv6(t *testing.T) {
+ b := header.IPv6(make([]byte, header.IPv6MinimumSize))
+ b.Encode(&header.IPv6Fields{})
+
+ const want = header.IPv6Version
+ if v := header.IPVersion(b); v != want {
+ t.Fatalf("Bad version, want %v, got %v", want, v)
+ }
+}
+
+func TestOtherVersion(t *testing.T) {
+ const want = header.IPv4Version + header.IPv6Version
+ b := make([]byte, 1)
+ b[0] = want << 4
+
+ if v := header.IPVersion(b); v != want {
+ t.Fatalf("Bad version, want %v, got %v", want, v)
+ }
+}
+
+func TestTooShort(t *testing.T) {
+ b := make([]byte, 1)
+ b[0] = (header.IPv4Version + header.IPv6Version) << 4
+
+ // Get the version of a zero-length slice.
+ const want = -1
+ if v := header.IPVersion(b[:0]); v != want {
+ t.Fatalf("Bad version, want %v, got %v", want, v)
+ }
+
+ // Get the version of a nil slice.
+ if v := header.IPVersion(nil); v != want {
+ t.Fatalf("Bad version, want %v, got %v", want, v)
+ }
+}
diff --git a/pkg/tcpip/header/tcp.go b/pkg/tcpip/header/tcp.go
new file mode 100644
index 000000000..995df4076
--- /dev/null
+++ b/pkg/tcpip/header/tcp.go
@@ -0,0 +1,518 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+ "gvisor.googlesource.com/gvisor/pkg/tcpip/seqnum"
+)
+
+const (
+ srcPort = 0
+ dstPort = 2
+ seqNum = 4
+ ackNum = 8
+ dataOffset = 12
+ tcpFlags = 13
+ winSize = 14
+ tcpChecksum = 16
+ urgentPtr = 18
+)
+
+const (
+ // MaxWndScale is maximum allowed window scaling, as described in
+ // RFC 1323, section 2.3, page 11.
+ MaxWndScale = 14
+
+ // TCPMaxSACKBlocks is the maximum number of SACK blocks that can
+ // be encoded in a TCP option field.
+ TCPMaxSACKBlocks = 4
+)
+
+// Flags that may be set in a TCP segment.
+const (
+ TCPFlagFin = 1 << iota
+ TCPFlagSyn
+ TCPFlagRst
+ TCPFlagPsh
+ TCPFlagAck
+ TCPFlagUrg
+)
+
+// Options that may be present in a TCP segment.
+const (
+ TCPOptionEOL = 0
+ TCPOptionNOP = 1
+ TCPOptionMSS = 2
+ TCPOptionWS = 3
+ TCPOptionTS = 8
+ TCPOptionSACKPermitted = 4
+ TCPOptionSACK = 5
+)
+
+// TCPFields contains the fields of a TCP packet. It is used to describe the
+// fields of a packet that needs to be encoded.
+type TCPFields struct {
+ // SrcPort is the "source port" field of a TCP packet.
+ SrcPort uint16
+
+ // DstPort is the "destination port" field of a TCP packet.
+ DstPort uint16
+
+ // SeqNum is the "sequence number" field of a TCP packet.
+ SeqNum uint32
+
+ // AckNum is the "acknowledgement number" field of a TCP packet.
+ AckNum uint32
+
+ // DataOffset is the "data offset" field of a TCP packet.
+ DataOffset uint8
+
+ // Flags is the "flags" field of a TCP packet.
+ Flags uint8
+
+ // WindowSize is the "window size" field of a TCP packet.
+ WindowSize uint16
+
+ // Checksum is the "checksum" field of a TCP packet.
+ Checksum uint16
+
+ // UrgentPointer is the "urgent pointer" field of a TCP packet.
+ UrgentPointer uint16
+}
+
+// TCPSynOptions is used to return the parsed TCP Options in a syn
+// segment.
+type TCPSynOptions struct {
+ // MSS is the maximum segment size provided by the peer in the SYN.
+ MSS uint16
+
+ // WS is the window scale option provided by the peer in the SYN.
+ //
+ // Set to -1 if no window scale option was provided.
+ WS int
+
+ // TS is true if the timestamp option was provided in the syn/syn-ack.
+ TS bool
+
+ // TSVal is the value of the TSVal field in the timestamp option.
+ TSVal uint32
+
+ // TSEcr is the value of the TSEcr field in the timestamp option.
+ TSEcr uint32
+
+ // SACKPermitted is true if the SACK option was provided in the SYN/SYN-ACK.
+ SACKPermitted bool
+}
+
+// SACKBlock represents a single contiguous SACK block.
+type SACKBlock struct {
+ // Start indicates the lowest sequence number in the block.
+ Start seqnum.Value
+
+ // End indicates the sequence number immediately following the last
+ // sequence number of this block.
+ End seqnum.Value
+}
+
+// TCPOptions are used to parse and cache the TCP segment options for a non
+// syn/syn-ack segment.
+type TCPOptions struct {
+ // TS is true if the TimeStamp option is enabled.
+ TS bool
+
+ // TSVal is the value in the TSVal field of the segment.
+ TSVal uint32
+
+ // TSEcr is the value in the TSEcr field of the segment.
+ TSEcr uint32
+
+ // SACKBlocks are the SACK blocks specified in the segment.
+ SACKBlocks []SACKBlock
+}
+
+// TCP represents a TCP header stored in a byte array.
+type TCP []byte
+
+const (
+ // TCPMinimumSize is the minimum size of a valid TCP packet.
+ TCPMinimumSize = 20
+
+ // TCPProtocolNumber is TCP's transport protocol number.
+ TCPProtocolNumber tcpip.TransportProtocolNumber = 6
+)
+
+// SourcePort returns the "source port" field of the tcp header.
+func (b TCP) SourcePort() uint16 {
+ return binary.BigEndian.Uint16(b[srcPort:])
+}
+
+// DestinationPort returns the "destination port" field of the tcp header.
+func (b TCP) DestinationPort() uint16 {
+ return binary.BigEndian.Uint16(b[dstPort:])
+}
+
+// SequenceNumber returns the "sequence number" field of the tcp header.
+func (b TCP) SequenceNumber() uint32 {
+ return binary.BigEndian.Uint32(b[seqNum:])
+}
+
+// AckNumber returns the "ack number" field of the tcp header.
+func (b TCP) AckNumber() uint32 {
+ return binary.BigEndian.Uint32(b[ackNum:])
+}
+
+// DataOffset returns the "data offset" field of the tcp header.
+func (b TCP) DataOffset() uint8 {
+ return (b[dataOffset] >> 4) * 4
+}
+
+// Payload returns the data in the tcp packet.
+func (b TCP) Payload() []byte {
+ return b[b.DataOffset():]
+}
+
+// Flags returns the flags field of the tcp header.
+func (b TCP) Flags() uint8 {
+ return b[tcpFlags]
+}
+
+// WindowSize returns the "window size" field of the tcp header.
+func (b TCP) WindowSize() uint16 {
+ return binary.BigEndian.Uint16(b[winSize:])
+}
+
+// Checksum returns the "checksum" field of the tcp header.
+func (b TCP) Checksum() uint16 {
+ return binary.BigEndian.Uint16(b[tcpChecksum:])
+}
+
+// SetSourcePort sets the "source port" field of the tcp header.
+func (b TCP) SetSourcePort(port uint16) {
+ binary.BigEndian.PutUint16(b[srcPort:], port)
+}
+
+// SetDestinationPort sets the "destination port" field of the tcp header.
+func (b TCP) SetDestinationPort(port uint16) {
+ binary.BigEndian.PutUint16(b[dstPort:], port)
+}
+
+// SetChecksum sets the checksum field of the tcp header.
+func (b TCP) SetChecksum(checksum uint16) {
+ binary.BigEndian.PutUint16(b[tcpChecksum:], checksum)
+}
+
+// CalculateChecksum calculates the checksum of the tcp segment given
+// the totalLen and partialChecksum(descriptions below)
+// totalLen is the total length of the segment
+// partialChecksum is the checksum of the network-layer pseudo-header
+// (excluding the total length) and the checksum of the segment data.
+func (b TCP) CalculateChecksum(partialChecksum uint16, totalLen uint16) uint16 {
+ // Add the length portion of the checksum to the pseudo-checksum.
+ tmp := make([]byte, 2)
+ binary.BigEndian.PutUint16(tmp, totalLen)
+ checksum := Checksum(tmp, partialChecksum)
+
+ // Calculate the rest of the checksum.
+ return Checksum(b[:b.DataOffset()], checksum)
+}
+
+// Options returns a slice that holds the unparsed TCP options in the segment.
+func (b TCP) Options() []byte {
+ return b[TCPMinimumSize:b.DataOffset()]
+}
+
+// ParsedOptions returns a TCPOptions structure which parses and caches the TCP
+// option values in the TCP segment. NOTE: Invoking this function repeatedly is
+// expensive as it reparses the options on each invocation.
+func (b TCP) ParsedOptions() TCPOptions {
+ return ParseTCPOptions(b.Options())
+}
+
+func (b TCP) encodeSubset(seq, ack uint32, flags uint8, rcvwnd uint16) {
+ binary.BigEndian.PutUint32(b[seqNum:], seq)
+ binary.BigEndian.PutUint32(b[ackNum:], ack)
+ b[tcpFlags] = flags
+ binary.BigEndian.PutUint16(b[winSize:], rcvwnd)
+}
+
+// Encode encodes all the fields of the tcp header.
+func (b TCP) Encode(t *TCPFields) {
+ b.encodeSubset(t.SeqNum, t.AckNum, t.Flags, t.WindowSize)
+ binary.BigEndian.PutUint16(b[srcPort:], t.SrcPort)
+ binary.BigEndian.PutUint16(b[dstPort:], t.DstPort)
+ b[dataOffset] = (t.DataOffset / 4) << 4
+ binary.BigEndian.PutUint16(b[tcpChecksum:], t.Checksum)
+ binary.BigEndian.PutUint16(b[urgentPtr:], t.UrgentPointer)
+}
+
+// EncodePartial updates a subset of the fields of the tcp header. It is useful
+// in cases when similar segments are produced.
+func (b TCP) EncodePartial(partialChecksum, length uint16, seqnum, acknum uint32, flags byte, rcvwnd uint16) {
+ // Add the total length and "flags" field contributions to the checksum.
+ // We don't use the flags field directly from the header because it's a
+ // one-byte field with an odd offset, so it would be accounted for
+ // incorrectly by the Checksum routine.
+ tmp := make([]byte, 4)
+ binary.BigEndian.PutUint16(tmp, length)
+ binary.BigEndian.PutUint16(tmp[2:], uint16(flags))
+ checksum := Checksum(tmp, partialChecksum)
+
+ // Encode the passed-in fields.
+ b.encodeSubset(seqnum, acknum, flags, rcvwnd)
+
+ // Add the contributions of the passed-in fields to the checksum.
+ checksum = Checksum(b[seqNum:seqNum+8], checksum)
+ checksum = Checksum(b[winSize:winSize+2], checksum)
+
+ // Encode the checksum.
+ b.SetChecksum(^checksum)
+}
+
+// ParseSynOptions parses the options received in a SYN segment and returns the
+// relevant ones. opts should point to the option part of the TCP Header.
+func ParseSynOptions(opts []byte, isAck bool) TCPSynOptions {
+ limit := len(opts)
+
+ synOpts := TCPSynOptions{
+ // Per RFC 1122, page 85: "If an MSS option is not received at
+ // connection setup, TCP MUST assume a default send MSS of 536."
+ MSS: 536,
+ // If no window scale option is specified, WS in options is
+ // returned as -1; this is because the absence of the option
+ // indicates that the we cannot use window scaling on the
+ // receive end either.
+ WS: -1,
+ }
+
+ for i := 0; i < limit; {
+ switch opts[i] {
+ case TCPOptionEOL:
+ i = limit
+ case TCPOptionNOP:
+ i++
+ case TCPOptionMSS:
+ if i+4 > limit || opts[i+1] != 4 {
+ return synOpts
+ }
+ mss := uint16(opts[i+2])<<8 | uint16(opts[i+3])
+ if mss == 0 {
+ return synOpts
+ }
+ synOpts.MSS = mss
+ i += 4
+
+ case TCPOptionWS:
+ if i+3 > limit || opts[i+1] != 3 {
+ return synOpts
+ }
+ ws := int(opts[i+2])
+ if ws > MaxWndScale {
+ ws = MaxWndScale
+ }
+ synOpts.WS = ws
+ i += 3
+
+ case TCPOptionTS:
+ if i+10 > limit || opts[i+1] != 10 {
+ return synOpts
+ }
+ synOpts.TSVal = binary.BigEndian.Uint32(opts[i+2:])
+ if isAck {
+ // If the segment is a SYN-ACK then store the Timestamp Echo Reply
+ // in the segment.
+ synOpts.TSEcr = binary.BigEndian.Uint32(opts[i+6:])
+ }
+ synOpts.TS = true
+ i += 10
+ case TCPOptionSACKPermitted:
+ if i+2 > limit || opts[i+1] != 2 {
+ return synOpts
+ }
+ synOpts.SACKPermitted = true
+ i += 2
+
+ default:
+ // We don't recognize this option, just skip over it.
+ if i+2 > limit {
+ return synOpts
+ }
+ l := int(opts[i+1])
+ // If the length is incorrect or if l+i overflows the
+ // total options length then return false.
+ if l < 2 || i+l > limit {
+ return synOpts
+ }
+ i += l
+ }
+ }
+
+ return synOpts
+}
+
+// ParseTCPOptions extracts and stores all known options in the provided byte
+// slice in a TCPOptions structure.
+func ParseTCPOptions(b []byte) TCPOptions {
+ opts := TCPOptions{}
+ limit := len(b)
+ for i := 0; i < limit; {
+ switch b[i] {
+ case TCPOptionEOL:
+ i = limit
+ case TCPOptionNOP:
+ i++
+ case TCPOptionTS:
+ if i+10 > limit || (b[i+1] != 10) {
+ return opts
+ }
+ opts.TS = true
+ opts.TSVal = binary.BigEndian.Uint32(b[i+2:])
+ opts.TSEcr = binary.BigEndian.Uint32(b[i+6:])
+ i += 10
+ case TCPOptionSACK:
+ if i+2 > limit {
+ // Malformed SACK block, just return and stop parsing.
+ return opts
+ }
+ sackOptionLen := int(b[i+1])
+ if i+sackOptionLen > limit || (sackOptionLen-2)%8 != 0 {
+ // Malformed SACK block, just return and stop parsing.
+ return opts
+ }
+ numBlocks := (sackOptionLen - 2) / 8
+ opts.SACKBlocks = []SACKBlock{}
+ for j := 0; j < numBlocks; j++ {
+ start := binary.BigEndian.Uint32(b[i+2+j*8:])
+ end := binary.BigEndian.Uint32(b[i+2+j*8+4:])
+ opts.SACKBlocks = append(opts.SACKBlocks, SACKBlock{
+ Start: seqnum.Value(start),
+ End: seqnum.Value(end),
+ })
+ }
+ i += sackOptionLen
+ default:
+ // We don't recognize this option, just skip over it.
+ if i+2 > limit {
+ return opts
+ }
+ l := int(b[i+1])
+ // If the length is incorrect or if l+i overflows the
+ // total options length then return false.
+ if l < 2 || i+l > limit {
+ return opts
+ }
+ i += l
+ }
+ }
+ return opts
+}
+
+// EncodeMSSOption encodes the MSS TCP option with the provided MSS values in
+// the supplied buffer. If the provided buffer is not large enough then it just
+// returns without encoding anything. It returns the number of bytes written to
+// the provided buffer.
+func EncodeMSSOption(mss uint32, b []byte) int {
+ // mssOptionSize is the number of bytes in a valid MSS option.
+ const mssOptionSize = 4
+
+ if len(b) < mssOptionSize {
+ return 0
+ }
+ b[0], b[1], b[2], b[3] = TCPOptionMSS, mssOptionSize, byte(mss>>8), byte(mss)
+ return mssOptionSize
+}
+
+// EncodeWSOption encodes the WS TCP option with the WS value in the
+// provided buffer. If the provided buffer is not large enough then it just
+// returns without encoding anything. It returns the number of bytes written to
+// the provided buffer.
+func EncodeWSOption(ws int, b []byte) int {
+ if len(b) < 3 {
+ return 0
+ }
+ b[0], b[1], b[2] = TCPOptionWS, 3, uint8(ws)
+ return int(b[1])
+}
+
+// EncodeTSOption encodes the provided tsVal and tsEcr values as a TCP timestamp
+// option into the provided buffer. If the buffer is smaller than expected it
+// just returns without encoding anything. It returns the number of bytes
+// written to the provided buffer.
+func EncodeTSOption(tsVal, tsEcr uint32, b []byte) int {
+ if len(b) < 10 {
+ return 0
+ }
+ b[0], b[1] = TCPOptionTS, 10
+ binary.BigEndian.PutUint32(b[2:], tsVal)
+ binary.BigEndian.PutUint32(b[6:], tsEcr)
+ return int(b[1])
+}
+
+// EncodeSACKPermittedOption encodes a SACKPermitted option into the provided
+// buffer. If the buffer is smaller than required it just returns without
+// encoding anything. It returns the number of bytes written to the provided
+// buffer.
+func EncodeSACKPermittedOption(b []byte) int {
+ if len(b) < 2 {
+ return 0
+ }
+
+ b[0], b[1] = TCPOptionSACKPermitted, 2
+ return int(b[1])
+}
+
+// EncodeSACKBlocks encodes the provided SACK blocks as a TCP SACK option block
+// in the provided slice. It tries to fit in as many blocks as possible based on
+// number of bytes available in the provided buffer. It returns the number of
+// bytes written to the provided buffer.
+func EncodeSACKBlocks(sackBlocks []SACKBlock, b []byte) int {
+ if len(sackBlocks) == 0 {
+ return 0
+ }
+ l := len(sackBlocks)
+ if l > TCPMaxSACKBlocks {
+ l = TCPMaxSACKBlocks
+ }
+ if ll := (len(b) - 2) / 8; ll < l {
+ l = ll
+ }
+ if l == 0 {
+ // There is not enough space in the provided buffer to add
+ // any SACK blocks.
+ return 0
+ }
+ b[0] = TCPOptionSACK
+ b[1] = byte(l*8 + 2)
+ for i := 0; i < l; i++ {
+ binary.BigEndian.PutUint32(b[i*8+2:], uint32(sackBlocks[i].Start))
+ binary.BigEndian.PutUint32(b[i*8+6:], uint32(sackBlocks[i].End))
+ }
+ return int(b[1])
+}
+
+// EncodeNOP adds an explicit NOP to the option list.
+func EncodeNOP(b []byte) int {
+ if len(b) == 0 {
+ return 0
+ }
+ b[0] = TCPOptionNOP
+ return 1
+}
+
+// AddTCPOptionPadding adds the required number of TCPOptionNOP to quad align
+// the option buffer. It adds padding bytes after the offset specified and
+// returns the number of padding bytes added. The passed in options slice
+// must have space for the padding bytes.
+func AddTCPOptionPadding(options []byte, offset int) int {
+ paddingToAdd := -offset & 3
+ // Now add any padding bytes that might be required to quad align the
+ // options.
+ for i := offset; i < offset+paddingToAdd; i++ {
+ options[i] = TCPOptionNOP
+ }
+ return paddingToAdd
+}
diff --git a/pkg/tcpip/header/tcp_test.go b/pkg/tcpip/header/tcp_test.go
new file mode 100644
index 000000000..27d43479a
--- /dev/null
+++ b/pkg/tcpip/header/tcp_test.go
@@ -0,0 +1,134 @@
+package header_test
+
+import (
+ "reflect"
+ "testing"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip/header"
+)
+
+func TestEncodeSACKBlocks(t *testing.T) {
+ testCases := []struct {
+ sackBlocks []header.SACKBlock
+ want []header.SACKBlock
+ bufSize int
+ }{
+ {
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}, {52, 60}, {62, 70}},
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}},
+ 40,
+ },
+ {
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}, {52, 60}, {62, 70}},
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}},
+ 30,
+ },
+ {
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}, {52, 60}, {62, 70}},
+ []header.SACKBlock{{10, 20}, {22, 30}},
+ 20,
+ },
+ {
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}, {52, 60}, {62, 70}},
+ []header.SACKBlock{{10, 20}},
+ 10,
+ },
+ {
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}, {52, 60}, {62, 70}},
+ nil,
+ 8,
+ },
+ {
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}, {52, 60}, {62, 70}},
+ []header.SACKBlock{{10, 20}, {22, 30}, {32, 40}, {42, 50}},
+ 60,
+ },
+ }
+ for _, tc := range testCases {
+ b := make([]byte, tc.bufSize)
+ t.Logf("testing: %v", tc)
+ header.EncodeSACKBlocks(tc.sackBlocks, b)
+ opts := header.ParseTCPOptions(b)
+ if got, want := opts.SACKBlocks, tc.want; !reflect.DeepEqual(got, want) {
+ t.Errorf("header.EncodeSACKBlocks(%v, %v), encoded blocks got: %v, want: %v", tc.sackBlocks, b, got, want)
+ }
+ }
+}
+
+func TestTCPParseOptions(t *testing.T) {
+ type tsOption struct {
+ tsVal uint32
+ tsEcr uint32
+ }
+
+ generateOptions := func(tsOpt *tsOption, sackBlocks []header.SACKBlock) []byte {
+ l := 0
+ if tsOpt != nil {
+ l += 10
+ }
+ if len(sackBlocks) != 0 {
+ l += len(sackBlocks)*8 + 2
+ }
+ b := make([]byte, l)
+ offset := 0
+ if tsOpt != nil {
+ offset = header.EncodeTSOption(tsOpt.tsVal, tsOpt.tsEcr, b)
+ }
+ header.EncodeSACKBlocks(sackBlocks, b[offset:])
+ return b
+ }
+
+ testCases := []struct {
+ b []byte
+ want header.TCPOptions
+ }{
+ // Trivial cases.
+ {nil, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionNOP}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionNOP, header.TCPOptionNOP}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionEOL}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionNOP, header.TCPOptionEOL, header.TCPOptionTS, 10, 1, 1}, header.TCPOptions{false, 0, 0, nil}},
+
+ // Test timestamp parsing.
+ {[]byte{header.TCPOptionNOP, header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{true, 1, 1, nil}},
+ {[]byte{header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{true, 1, 1, nil}},
+
+ // Test malformed timestamp option.
+ {[]byte{header.TCPOptionTS, 8, 1, 1}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionNOP, header.TCPOptionTS, 8, 1, 1}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionNOP, header.TCPOptionTS, 8, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{false, 0, 0, nil}},
+
+ // Test SACKBlock parsing.
+ {[]byte{header.TCPOptionSACK, 10, 0, 0, 0, 1, 0, 0, 0, 10}, header.TCPOptions{false, 0, 0, []header.SACKBlock{{1, 10}}}},
+ {[]byte{header.TCPOptionSACK, 18, 0, 0, 0, 1, 0, 0, 0, 10, 0, 0, 0, 11, 0, 0, 0, 12}, header.TCPOptions{false, 0, 0, []header.SACKBlock{{1, 10}, {11, 12}}}},
+
+ // Test malformed SACK option.
+ {[]byte{header.TCPOptionSACK, 0}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK, 8, 0, 0, 0, 1, 0, 0, 0, 10}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK, 11, 0, 0, 0, 1, 0, 0, 0, 10, 0, 0, 0, 11, 0, 0, 0, 12}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK, 17, 0, 0, 0, 1, 0, 0, 0, 10, 0, 0, 0, 11, 0, 0, 0, 12}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK, 10}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK, 10, 0, 0, 0, 1, 0, 0, 0}, header.TCPOptions{false, 0, 0, nil}},
+
+ // Test Timestamp + SACK block parsing.
+ {generateOptions(&tsOption{1, 1}, []header.SACKBlock{{1, 10}, {11, 12}}), header.TCPOptions{true, 1, 1, []header.SACKBlock{{1, 10}, {11, 12}}}},
+ {generateOptions(&tsOption{1, 2}, []header.SACKBlock{{1, 10}, {11, 12}}), header.TCPOptions{true, 1, 2, []header.SACKBlock{{1, 10}, {11, 12}}}},
+ {generateOptions(&tsOption{1, 3}, []header.SACKBlock{{1, 10}, {11, 12}, {13, 14}, {14, 15}, {15, 16}}), header.TCPOptions{true, 1, 3, []header.SACKBlock{{1, 10}, {11, 12}, {13, 14}, {14, 15}}}},
+
+ // Test valid timestamp + malformed SACK block parsing.
+ {[]byte{header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1, header.TCPOptionSACK}, header.TCPOptions{true, 1, 1, nil}},
+ {[]byte{header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1, header.TCPOptionSACK, 10}, header.TCPOptions{true, 1, 1, nil}},
+ {[]byte{header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1, header.TCPOptionSACK, 10, 0, 0, 0}, header.TCPOptions{true, 1, 1, nil}},
+ {[]byte{header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1, header.TCPOptionSACK, 11, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{true, 1, 1, nil}},
+ {[]byte{header.TCPOptionSACK, header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{false, 0, 0, nil}},
+ {[]byte{header.TCPOptionSACK, 10, header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{false, 0, 0, []header.SACKBlock{{134873088, 65536}}}},
+ {[]byte{header.TCPOptionSACK, 10, 0, 0, 0, header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{false, 0, 0, []header.SACKBlock{{8, 167772160}}}},
+ {[]byte{header.TCPOptionSACK, 11, 0, 0, 0, 1, 0, 0, 0, 1, header.TCPOptionTS, 10, 0, 0, 0, 1, 0, 0, 0, 1}, header.TCPOptions{false, 0, 0, nil}},
+ }
+ for _, tc := range testCases {
+ if got, want := header.ParseTCPOptions(tc.b), tc.want; !reflect.DeepEqual(got, want) {
+ t.Errorf("ParseTCPOptions(%v) = %v, want: %v", tc.b, got, tc.want)
+ }
+ }
+}
diff --git a/pkg/tcpip/header/udp.go b/pkg/tcpip/header/udp.go
new file mode 100644
index 000000000..7c2548634
--- /dev/null
+++ b/pkg/tcpip/header/udp.go
@@ -0,0 +1,106 @@
+// Copyright 2016 The Netstack Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package header
+
+import (
+ "encoding/binary"
+
+ "gvisor.googlesource.com/gvisor/pkg/tcpip"
+)
+
+const (
+ udpSrcPort = 0
+ udpDstPort = 2
+ udpLength = 4
+ udpChecksum = 6
+)
+
+// UDPFields contains the fields of a UDP packet. It is used to describe the
+// fields of a packet that needs to be encoded.
+type UDPFields struct {
+ // SrcPort is the "source port" field of a UDP packet.
+ SrcPort uint16
+
+ // DstPort is the "destination port" field of a UDP packet.
+ DstPort uint16
+
+ // Length is the "length" field of a UDP packet.
+ Length uint16
+
+ // Checksum is the "checksum" field of a UDP packet.
+ Checksum uint16
+}
+
+// UDP represents a UDP header stored in a byte array.
+type UDP []byte
+
+const (
+ // UDPMinimumSize is the minimum size of a valid UDP packet.
+ UDPMinimumSize = 8
+
+ // UDPProtocolNumber is UDP's transport protocol number.
+ UDPProtocolNumber tcpip.TransportProtocolNumber = 17
+)
+
+// SourcePort returns the "source port" field of the udp header.
+func (b UDP) SourcePort() uint16 {
+ return binary.BigEndian.Uint16(b[udpSrcPort:])
+}
+
+// DestinationPort returns the "destination port" field of the udp header.
+func (b UDP) DestinationPort() uint16 {
+ return binary.BigEndian.Uint16(b[udpDstPort:])
+}
+
+// Length returns the "length" field of the udp header.
+func (b UDP) Length() uint16 {
+ return binary.BigEndian.Uint16(b[udpLength:])
+}
+
+// Payload returns the data contained in the UDP datagram.
+func (b UDP) Payload() []byte {
+ return b[UDPMinimumSize:]
+}
+
+// Checksum returns the "checksum" field of the udp header.
+func (b UDP) Checksum() uint16 {
+ return binary.BigEndian.Uint16(b[udpChecksum:])
+}
+
+// SetSourcePort sets the "source port" field of the udp header.
+func (b UDP) SetSourcePort(port uint16) {
+ binary.BigEndian.PutUint16(b[udpSrcPort:], port)
+}
+
+// SetDestinationPort sets the "destination port" field of the udp header.
+func (b UDP) SetDestinationPort(port uint16) {
+ binary.BigEndian.PutUint16(b[udpDstPort:], port)
+}
+
+// SetChecksum sets the "checksum" field of the udp header.
+func (b UDP) SetChecksum(checksum uint16) {
+ binary.BigEndian.PutUint16(b[udpChecksum:], checksum)
+}
+
+// CalculateChecksum calculates the checksum of the udp packet, given the total
+// length of the packet and the checksum of the network-layer pseudo-header
+// (excluding the total length) and the checksum of the payload.
+func (b UDP) CalculateChecksum(partialChecksum uint16, totalLen uint16) uint16 {
+ // Add the length portion of the checksum to the pseudo-checksum.
+ tmp := make([]byte, 2)
+ binary.BigEndian.PutUint16(tmp, totalLen)
+ checksum := Checksum(tmp, partialChecksum)
+
+ // Calculate the rest of the checksum.
+ return Checksum(b[:UDPMinimumSize], checksum)
+}
+
+// Encode encodes all the fields of the udp header.
+func (b UDP) Encode(u *UDPFields) {
+ binary.BigEndian.PutUint16(b[udpSrcPort:], u.SrcPort)
+ binary.BigEndian.PutUint16(b[udpDstPort:], u.DstPort)
+ binary.BigEndian.PutUint16(b[udpLength:], u.Length)
+ binary.BigEndian.PutUint16(b[udpChecksum:], u.Checksum)
+}