ICMP实现 ip数据报分片机制还未实现

go.mod

@@ -1,3 +1,3 @@
 module netstack
 
-go 1.18
+go 1.19

tcpip/header/ipv4.go

@@ -2,6 +2,7 @@ package header
 
 import (
 	"encoding/binary"
+	"fmt"
 	"netstack/tcpip"
 )
 
@@ -106,6 +107,12 @@ const (
 	IPv4Any tcpip.Address = "\x00\x00\x00\x00"
 )
 
+// Flags that may be set in an IPv4 packet.
+const (
+	IPv4FlagMoreFragments = 1 << iota
+	IPv4FlagDontFragment
+)
+
 func IPVersion(b []byte) int {
 	if len(b) < versIHL+1 {
 		return -1
@@ -216,6 +223,11 @@ 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)
@@ -264,3 +276,28 @@ func IsV4MulticastAddress(addr tcpip.Address) bool {
 	}
 	return (addr[0] & 0xf0) == 0xe0
 }
+
+var ipv4Fmt string = `
+|% 4s|% 4s|% 8s| % 16s|
+|  % 16s|%s|%s|%s|% 11s|
+| % 8s|% 8s|% 16s |
+|% 32s    |
+|% 32s    |
+|        Options       |   Padding   |
+%v
+`
+
+type Types [] struct {}
+
+func atoi[T int | int8 | int16 | int32 | int64 | uint | uint8 |uint16 | uint32](i T) string {
+	return fmt.Sprintf("%d", i)
+}
+
+func (b IPv4) String() string {
+	return fmt.Sprintf(ipv4Fmt, atoi(IPVersion(b)), atoi(b.HeaderLength()), atoi(0), atoi(b.TotalLength()),
+		atoi(b.ID()), atoi(b.Flags()>>2), atoi((b.Flags()&2)>>1), atoi(b.Flags()&1), atoi(b.FragmentOffset()),
+		atoi(b.TTL()), atoi(b.Protocol()), atoi(b.Checksum()),
+		b.SourceAddress().String(),
+		b.DestinationAddress().String(),
+		b.Payload())
+}

tcpip/link/fdbased/endpoint.go

@@ -1,6 +1,7 @@
 package fdbased
 
 import (
+	"log"
 	"netstack/tcpip"
 	"netstack/tcpip/buffer"
 	"netstack/tcpip/header"
@@ -126,6 +127,7 @@ func (e *endpoint) WritePacket(r *stack.Route, hdr buffer.Prependable,
 		ethHdr.SrcAddr = e.addr // 说明这是一个原始报文
 	}
 	eth.Encode(ethHdr) // 将以太帧信息作为报文头编入
+	log.Println("链路层写回报文")
 	// 写入网卡中
 	if payload.Size() == 0 {
 		return rawfile.NonBlockingWrite(e.fd, hdr.View())

tcpip/network/ipv4/icmp.go

@@ -0,0 +1,128 @@
+package ipv4
+
+import (
+	"encoding/binary"
+	"log"
+	"netstack/tcpip"
+	"netstack/tcpip/buffer"
+	"netstack/tcpip/header"
+	"netstack/tcpip/stack"
+)
+
+/*
+ ICMP 的全称是 Internet Control Message Protocol 。与 IP 协议一样同属 TCP/IP 模型中的网络层，并且 ICMP 数据包是包裹在 IP 数据包中的
+
+ 0                   1                   2                   3
+ 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
++-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+|     Type      |     Code      |          Checksum             |
++-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+|                                                               |
+|                   不同的Type和Code有不同的内容            	 |
+|                                                               |
++-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+*/
+
+type echoRequest struct {
+	r stack.Route
+	v buffer.View
+}
+
+// handleControl处理ICMP数据包包含导致ICMP发送的原始数据包的标头的情况。
+// 此信息用于确定必须通知哪个传输端点有关ICMP数据包。
+func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, vv buffer.VectorisedView) {
+	h := header.IPv4(vv.First())
+
+	// We don't use IsValid() here because ICMP only requires that the IP
+	// header plus 8 bytes of the transport header be included. So it's
+	// likely that it is truncated, which would cause IsValid to return
+	// false.
+	//
+	// Drop packet if it doesn't have the basic IPv4 header or if the
+	// original source address doesn't match the endpoint's address.
+	if len(h) < header.IPv4MinimumSize || h.SourceAddress() != e.id.LocalAddress {
+		return
+	}
+
+	hlen := int(h.HeaderLength())
+	if vv.Size() < hlen || h.FragmentOffset() != 0 {
+		// We won't be able to handle this if it doesn't contain the
+		// full IPv4 header, or if it's a fragment not at offset 0
+		// (because it won't have the transport header).
+		return
+	}
+
+	// Skip the ip header, then deliver control message.
+	vv.TrimFront(hlen)
+	p := h.TransportProtocol()
+	e.dispatcher.DeliverTransportControlPacket(e.id.LocalAddress, h.DestinationAddress(), ProtocolNumber, p, typ, extra, vv)
+}
+
+// 处理ICMP报文
+func (e *endpoint) handleICMP(r *stack.Route, vv buffer.VectorisedView) {
+	v := vv.First()
+	if len(v) < header.ICMPv4MinimumSize {
+		return
+	}
+	h := header.ICMPv4(v)
+
+	// 更具icmp的类型来进行相应的处理
+	switch h.Type() {
+	case header.ICMPv4Echo: // icmp echo请求
+		if len(v) < header.ICMPv4EchoMinimumSize {
+			return
+		}
+		log.Printf("ICMP echo")
+		vv.TrimFront(header.ICMPv4MinimumSize) // 去掉头部
+		req := echoRequest{r: r.Clone(), v: vv.ToView()}
+		select {
+		case e.echoRequests <- req: // 发送给echoReplier处理 在那里会重新组一个头部
+		default:
+			req.r.Release()
+		}
+
+	case header.ICMPv4EchoReply: // icmp echo响应
+		if len(v) < header.ICMPv4EchoMinimumSize {
+			return
+		}
+		e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, vv)
+
+	case header.ICMPv4DstUnreachable: // 目标不可达
+		if len(v) < header.ICMPv4DstUnreachableMinimumSize {
+			return
+		}
+		vv.TrimFront(header.ICMPv4DstUnreachableMinimumSize)
+		switch h.Code() {
+		case header.ICMPv4PortUnreachable: // 端口不可达
+			e.handleControl(stack.ControlPortUnreachable, 0, vv)
+
+		case header.ICMPv4FragmentationNeeded: // 需要进行分片但设置不分片标志
+			mtu := uint32(binary.BigEndian.Uint16(v[header.ICMPv4DstUnreachableMinimumSize-2:]))
+			e.handleControl(stack.ControlPacketTooBig, calculateMTU(mtu), vv)
+		}
+	}
+}
+
+// 处理icmp echo请求的goroutine
+func (e *endpoint) echoReplier() {
+	for req := range e.echoRequests {
+		sendPing4(&req.r, 0, req.v)
+		req.r.Release()
+	}
+}
+
+// 根据icmp echo请求，封装icmp echo响应报文，并传给ip层处理
+func sendPing4(r *stack.Route, code byte, data buffer.View) *tcpip.Error {
+	hdr := buffer.NewPrependable(header.ICMPv4EchoMinimumSize + int(r.MaxHeaderLength()))
+
+	icmpv4 := header.ICMPv4(hdr.Prepend(header.ICMPv4EchoMinimumSize))
+	icmpv4.SetType(header.ICMPv4EchoReply)
+	icmpv4.SetCode(code)
+	copy(icmpv4[header.ICMPv4MinimumSize:], data)
+	data = data[header.ICMPv4EchoMinimumSize-header.ICMPv4MinimumSize:]
+	icmpv4.SetChecksum(^header.Checksum(icmpv4, header.Checksum(data, 0)))
+
+	log.Printf("ICMP 回应报文组完 再次包装到IP报文")
+	// 传给ip层处理
+	return r.WritePacket(hdr, data.ToVectorisedView(), header.ICMPv4ProtocolNumber, r.DefaultTTL())
+}

tcpip/network/ipv4/ipv4.go

@@ -1,6 +1,7 @@
 package ipv4
 
 import (
+	"log"
 	"netstack/tcpip"
 	"netstack/tcpip/buffer"
 	"netstack/tcpip/header"
@@ -30,7 +31,12 @@ type endpoint struct {
 	id stack.NetworkEndpointID
 	// 链路端的表示
 	linkEP stack.LinkEndpoint
-	// TODO 需要添加
+	// 报文分发器
+	dispatcher stack.TransportDispatcher
+	// ping请求报文接收队列
+	echoRequests chan echoRequest
+	// ip报文分片处理器
+	//fragmentation *fragmentation.Fragmentation
 }
 
 // DefaultTTL is the default time-to-live value for this endpoint.
@@ -73,13 +79,82 @@ func (e *endpoint) MaxHeaderLength() uint16 {
 // 将传输层的数据封装加上IP头，并调用网卡的写入接口，写入IP报文
 func (e *endpoint) WritePacket(r *stack.Route, hdr buffer.Prependable, payload buffer.VectorisedView,
 	protocol tcpip.TransportProtocolNumber, ttl uint8) *tcpip.Error {
-	return nil
+	// 预留ip报文的空间
+	ip := header.IPv4(hdr.Prepend(header.IPv4MinimumSize))
+	length := uint16(hdr.UsedLength() + payload.Size())
+	id := uint32(0)
+	// 如果报文长度大于68
+	if length > header.IPv4MaximumHeaderSize+8 {
+		// Packets of 68 bytes or less are required by RFC 791 to not be
+		// fragmented, so we only assign ids to larger packets.
+		//id = atomic.AddUint32(&ids[hashRoute(r, protocol)%buckets], 1)
+	}
+	// ip首部编码
+	ip.Encode(&header.IPv4Fields{
+		IHL:         header.IPv4MinimumSize,
+		TotalLength: length,
+		ID:          uint16(id),
+		TTL:         ttl,
+		Protocol:    uint8(protocol),
+		SrcAddr:     r.LocalAddress,
+		DstAddr:     r.RemoteAddress,
+	})
+	// 计算校验和和设置校验和
+	ip.SetChecksum(^ip.CalculateChecksum())
+	r.Stats().IP.PacketsSent.Increment()
+
+	// 写入网卡接口
+	if protocol == header.ICMPv4ProtocolNumber {
+		log.Printf("IP 写回ICMP报文 长度: %d\n", hdr.UsedLength()+payload.Size())
+	} else {
+		log.Printf("send ipv4 packet %d bytes, proto: 0x%x", hdr.UsedLength()+payload.Size(), protocol)
+	}
+	return e.linkEP.WritePacket(r, hdr, payload, ProtocolNumber)
 }
 
 // HandlePacket is called by the link layer when new ipv4 packets arrive for
 // this endpoint.
 // 收到ip包的处理
 func (e *endpoint) HandlePacket(r *stack.Route, vv buffer.VectorisedView) {
+	// 得到ip报文
+	h := header.IPv4(vv.First())
+	// 检查报文是否有效
+	if !h.IsValid(vv.Size()) {
+		return
+	}
+	log.Println(h)
+
+	hlen := int(h.HeaderLength())
+	tlen := int(h.TotalLength())
+	vv.TrimFront(hlen)
+	vv.CapLength(tlen - hlen)
+
+	// 报文重组 TODO
+	more := (h.Flags() & header.IPv4FlagMoreFragments) != 0
+	// 是否需要ip重组
+	if more || h.FragmentOffset() != 0 {
+		// The packet is a fragment, let's try to reassemble it.
+		last := h.FragmentOffset() + uint16(vv.Size()) - 1
+		var ready bool
+		log.Println(last)
+		// ip分片重组
+		//vv, ready = e.fragmentation.Process(hash.IPv4FragmentHash(h), h.FragmentOffset(), last, more, vv)
+		if !ready {
+			return
+		}
+	}
+
+	// 得到传输层的协议
+	p := h.TransportProtocol()
+	// 如果时ICMP协议，则进入ICMP处理函数
+	if p == header.ICMPv4ProtocolNumber {
+		e.handleICMP(r, vv)
+		return
+	}
+	r.Stats().IP.PacketsDelivered.Increment()
+	// 根据协议分发到不同处理函数，比如协议时TCP，会进入tcp.HandlePacket
+	log.Printf("recv ipv4 packet %d bytes, proto: 0x%x", tlen, p)
+	e.dispatcher.DeliverTransportPacket(r, p, vv)
 }
 
 // Close cleans up resources associated with the endpoint.
@@ -94,11 +169,15 @@ type protocol struct{}
 func (p *protocol) NewEndpoint(nicid tcpip.NICID, addr tcpip.Address, linkAddrCache stack.LinkAddressCache,
 	dispatcher stack.TransportDispatcher, linkEP stack.LinkEndpoint) (stack.NetworkEndpoint, *tcpip.Error) {
 	e := &endpoint{
-		nicid:  nicid,
+		nicid:        nicid,
-		id:     stack.NetworkEndpointID{LocalAddress: addr},
+		id:           stack.NetworkEndpointID{LocalAddress: addr},
-		linkEP: linkEP,
+		linkEP:       linkEP,
+		dispatcher:   dispatcher,
+		echoRequests: make(chan echoRequest, 10),
 	}
 
+	go e.echoReplier()
+
 	return e, nil
 }
 
@@ -122,9 +201,8 @@ func (p *protocol) MinimumPacketSize() int {
 
 // ParseAddresses implements NetworkProtocol.ParseAddresses.
 func (*protocol) ParseAddresses(v buffer.View) (src, dst tcpip.Address) {
-	//h := header.IPv4(v)
+	h := header.IPv4(v)
-	//return h.SourceAddress(), h.DestinationAddress()
+	return h.SourceAddress(), h.DestinationAddress()
-	return "", ""
 }
 
 // SetOption implements NetworkProtocol.SetOption.

tcpip/network/ipv4/ipv4_test.go

@@ -0,0 +1,7 @@
+package ipv4_test
+
+import "testing"
+
+func TestIPv4Base(t *testing.T) {
+
+}

tcpip/stack/stack.go

@@ -238,10 +238,10 @@ func (s *Stack) FindRoute(id tcpip.NICID, localAddr, remoteAddr tcpip.Address,
 	defer s.mu.RUnlock()
 
 	for i := range s.routeTable {
-		//if (id != 0 && id != s.routeTable[i].NIC) ||
+		if (id != 0 && id != s.routeTable[i].NIC) ||
-		//	(len(remoteAddr) != 0 && !s.routeTable[i].Match(remoteAddr)) {
+			(len(remoteAddr) != 0 && !s.routeTable[i].Match(remoteAddr)) {
-		//	continue
+			continue
-		//}
+		}
 
 		nic := s.nics[s.routeTable[i].NIC]
 		if nic == nil {

tcpip/tcpip.go

@@ -183,6 +183,21 @@ type Route struct {
 	NIC         NICID       // 使用的网卡设备
 }
 
+// Match determines if r is viable for the given destination address.
+func (r *Route) Match(addr Address) bool {
+	if len(addr) != len(r.Destination) {
+		return false
+	}
+
+	for i := 0; i < len(r.Destination); i++ {
+		if (addr[i] & r.Mask[i]) != r.Destination[i] {
+			return false
+		}
+	}
+
+	return true
+}
+
 // Stats 包含了网络栈的统计信息
 type Stats struct {
 	// TODO 需要解读