From 47515f475167ffa23267ca0b9d1b39e7907587d6 Mon Sep 17 00:00:00 2001 From: Ting-Yu Wang Date: Thu, 13 Aug 2020 13:07:03 -0700 Subject: Migrate to PacketHeader API for PacketBuffer. Formerly, when a packet is constructed or parsed, all headers are set by the client code. This almost always involved prepending to pk.Header buffer or trimming pk.Data portion. This is known to prone to bugs, due to the complexity and number of the invariants assumed across netstack to maintain. In the new PacketHeader API, client will call Push()/Consume() method to construct/parse an outgoing/incoming packet. All invariants, such as slicing and trimming, are maintained by the API itself. NewPacketBuffer() is introduced to create new PacketBuffer. Zero value is no longer valid. PacketBuffer now assumes the packet is a concatenation of following portions: * LinkHeader * NetworkHeader * TransportHeader * Data Any of them could be empty, or zero-length. PiperOrigin-RevId: 326507688 --- pkg/tcpip/network/arp/arp.go | 26 +++-- pkg/tcpip/network/arp/arp_test.go | 8 +- pkg/tcpip/network/ip_test.go | 80 ++++++++------- pkg/tcpip/network/ipv4/icmp.go | 38 ++++---- pkg/tcpip/network/ipv4/ipv4.go | 188 +++++++++++++++++------------------- pkg/tcpip/network/ipv4/ipv4_test.go | 72 +++++++------- pkg/tcpip/network/ipv6/icmp.go | 55 ++++++----- pkg/tcpip/network/ipv6/icmp_test.go | 46 +++++---- pkg/tcpip/network/ipv6/ipv6.go | 47 ++++----- pkg/tcpip/network/ipv6/ipv6_test.go | 16 +-- pkg/tcpip/network/ipv6/ndp_test.go | 29 +++--- 11 files changed, 301 insertions(+), 304 deletions(-) (limited to 'pkg/tcpip/network') diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go index 31a242482..1ad788a17 100644 --- a/pkg/tcpip/network/arp/arp.go +++ b/pkg/tcpip/network/arp/arp.go @@ -99,7 +99,7 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBu } func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { - h := header.ARP(pkt.NetworkHeader) + h := header.ARP(pkt.NetworkHeader().View()) if !h.IsValid() { return } @@ -110,17 +110,17 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { 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) - packet := header.ARP(hdr.Prepend(header.ARPSize)) + 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, &stack.PacketBuffer{ - Header: hdr, - }) + _ = e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, pkt) fallthrough // also fill the cache from requests case header.ARPReply: addr := tcpip.Address(h.ProtocolAddressSender()) @@ -168,17 +168,17 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, remoteLinkAdd 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 */, ProtocolNumber, &stack.PacketBuffer{ - Header: hdr, - }) + return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, pkt) } // ResolveStaticAddress implements stack.LinkAddressResolver.ResolveStaticAddress. @@ -210,12 +210,10 @@ func (*protocol) Wait() {} // Parse implements stack.NetworkProtocol.Parse. func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNumber, hasTransportHdr bool, ok bool) { - hdr, ok := pkt.Data.PullUp(header.ARPSize) + _, ok = pkt.NetworkHeader().Consume(header.ARPSize) if !ok { return 0, false, false } - pkt.NetworkHeader = hdr - pkt.Data.TrimFront(header.ARPSize) return 0, false, true } diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go index a35a64a0f..c2c3e6891 100644 --- a/pkg/tcpip/network/arp/arp_test.go +++ b/pkg/tcpip/network/arp/arp_test.go @@ -106,9 +106,9 @@ func TestDirectRequest(t *testing.T) { inject := func(addr tcpip.Address) { copy(h.ProtocolAddressTarget(), addr) - c.linkEP.InjectInbound(arp.ProtocolNumber, &stack.PacketBuffer{ + c.linkEP.InjectInbound(arp.ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: v.ToVectorisedView(), - }) + })) } for i, address := range []tcpip.Address{stackAddr1, stackAddr2} { @@ -118,9 +118,9 @@ func TestDirectRequest(t *testing.T) { if pi.Proto != arp.ProtocolNumber { t.Fatalf("expected ARP response, got network protocol number %d", pi.Proto) } - rep := header.ARP(pi.Pkt.Header.View()) + rep := header.ARP(pi.Pkt.NetworkHeader().View()) if !rep.IsValid() { - t.Fatalf("invalid ARP response pi.Pkt.Header.UsedLength()=%d", pi.Pkt.Header.UsedLength()) + t.Fatalf("invalid ARP response: len = %d; response = %x", len(rep), rep) } if got, want := tcpip.LinkAddress(rep.HardwareAddressSender()), stackLinkAddr1; got != want { t.Errorf("got HardwareAddressSender = %s, want = %s", got, want) diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go index 615bae648..e6768258a 100644 --- a/pkg/tcpip/network/ip_test.go +++ b/pkg/tcpip/network/ip_test.go @@ -156,13 +156,13 @@ func (t *testObject) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.Ne var dstAddr tcpip.Address if t.v4 { - h := header.IPv4(pkt.Header.View()) + h := header.IPv4(pkt.NetworkHeader().View()) prot = tcpip.TransportProtocolNumber(h.Protocol()) srcAddr = h.SourceAddress() dstAddr = h.DestinationAddress() } else { - h := header.IPv6(pkt.Header.View()) + h := header.IPv6(pkt.NetworkHeader().View()) prot = tcpip.TransportProtocolNumber(h.NextHeader()) srcAddr = h.SourceAddress() dstAddr = h.DestinationAddress() @@ -243,8 +243,11 @@ func TestIPv4Send(t *testing.T) { payload[i] = uint8(i) } - // Allocate the header buffer. - hdr := buffer.NewPrependable(int(ep.MaxHeaderLength())) + // Setup the packet buffer. + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: int(ep.MaxHeaderLength()), + Data: payload.ToVectorisedView(), + }) // Issue the write. o.protocol = 123 @@ -260,10 +263,7 @@ func TestIPv4Send(t *testing.T) { Protocol: 123, TTL: 123, TOS: stack.DefaultTOS, - }, &stack.PacketBuffer{ - Header: hdr, - Data: payload.ToVectorisedView(), - }); err != nil { + }, pkt); err != nil { t.Fatalf("WritePacket failed: %v", err) } } @@ -303,9 +303,13 @@ func TestIPv4Receive(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - pkt := stack.PacketBuffer{Data: view.ToVectorisedView()} - proto.Parse(&pkt) - ep.HandlePacket(&r, &pkt) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + Data: view.ToVectorisedView(), + }) + if _, _, ok := proto.Parse(pkt); !ok { + t.Fatalf("failed to parse packet: %x", pkt.Data.ToView()) + } + ep.HandlePacket(&r, pkt) if o.dataCalls != 1 { t.Fatalf("Bad number of data calls: got %x, want 1", o.dataCalls) } @@ -455,17 +459,25 @@ func TestIPv4FragmentationReceive(t *testing.T) { } // Send first segment. - pkt := stack.PacketBuffer{Data: frag1.ToVectorisedView()} - proto.Parse(&pkt) - ep.HandlePacket(&r, &pkt) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + Data: frag1.ToVectorisedView(), + }) + if _, _, ok := proto.Parse(pkt); !ok { + t.Fatalf("failed to parse packet: %x", pkt.Data.ToView()) + } + ep.HandlePacket(&r, pkt) if o.dataCalls != 0 { t.Fatalf("Bad number of data calls: got %x, want 0", o.dataCalls) } // Send second segment. - pkt = stack.PacketBuffer{Data: frag2.ToVectorisedView()} - proto.Parse(&pkt) - ep.HandlePacket(&r, &pkt) + pkt = stack.NewPacketBuffer(stack.PacketBufferOptions{ + Data: frag2.ToVectorisedView(), + }) + if _, _, ok := proto.Parse(pkt); !ok { + t.Fatalf("failed to parse packet: %x", pkt.Data.ToView()) + } + ep.HandlePacket(&r, pkt) if o.dataCalls != 1 { t.Fatalf("Bad number of data calls: got %x, want 1", o.dataCalls) } @@ -485,8 +497,11 @@ func TestIPv6Send(t *testing.T) { payload[i] = uint8(i) } - // Allocate the header buffer. - hdr := buffer.NewPrependable(int(ep.MaxHeaderLength())) + // Setup the packet buffer. + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: int(ep.MaxHeaderLength()), + Data: payload.ToVectorisedView(), + }) // Issue the write. o.protocol = 123 @@ -502,10 +517,7 @@ func TestIPv6Send(t *testing.T) { Protocol: 123, TTL: 123, TOS: stack.DefaultTOS, - }, &stack.PacketBuffer{ - Header: hdr, - Data: payload.ToVectorisedView(), - }); err != nil { + }, pkt); err != nil { t.Fatalf("WritePacket failed: %v", err) } } @@ -545,9 +557,13 @@ func TestIPv6Receive(t *testing.T) { t.Fatalf("could not find route: %v", err) } - pkt := stack.PacketBuffer{Data: view.ToVectorisedView()} - proto.Parse(&pkt) - ep.HandlePacket(&r, &pkt) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + Data: view.ToVectorisedView(), + }) + if _, _, ok := proto.Parse(pkt); !ok { + t.Fatalf("failed to parse packet: %x", pkt.Data.ToView()) + } + ep.HandlePacket(&r, pkt) if o.dataCalls != 1 { t.Fatalf("Bad number of data calls: got %x, want 1", o.dataCalls) } @@ -673,11 +689,9 @@ func TestIPv6ReceiveControl(t *testing.T) { // becomes Data. func truncatedPacket(view buffer.View, trunc, netHdrLen int) *stack.PacketBuffer { v := view[:len(view)-trunc] - if len(v) < netHdrLen { - return &stack.PacketBuffer{Data: v.ToVectorisedView()} - } - return &stack.PacketBuffer{ - NetworkHeader: v[:netHdrLen], - Data: v[netHdrLen:].ToVectorisedView(), - } + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + Data: v.ToVectorisedView(), + }) + _, _ = pkt.NetworkHeader().Consume(netHdrLen) + return pkt } diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index 94803a359..067d770f3 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/pkg/tcpip/network/ipv4/icmp.go @@ -89,12 +89,14 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer) { return } + // Make a copy of data before pkt gets sent to raw socket. + // DeliverTransportPacket will take ownership of pkt. + replyData := pkt.Data.Clone(nil) + replyData.TrimFront(header.ICMPv4MinimumSize) + // It's possible that a raw socket expects to receive this. h.SetChecksum(wantChecksum) - e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, &stack.PacketBuffer{ - Data: pkt.Data.Clone(nil), - NetworkHeader: append(buffer.View(nil), pkt.NetworkHeader...), - }) + e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, pkt) remoteLinkAddr := r.RemoteLinkAddress @@ -116,24 +118,26 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer) { // Use the remote link address from the incoming packet. r.ResolveWith(remoteLinkAddr) - vv := pkt.Data.Clone(nil) - vv.TrimFront(header.ICMPv4MinimumSize) - hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.ICMPv4MinimumSize) - pkt := header.ICMPv4(hdr.Prepend(header.ICMPv4MinimumSize)) - copy(pkt, h) - pkt.SetType(header.ICMPv4EchoReply) - pkt.SetChecksum(0) - pkt.SetChecksum(^header.Checksum(pkt, header.ChecksumVV(vv, 0))) + // Prepare a reply packet. + icmpHdr := make(header.ICMPv4, header.ICMPv4MinimumSize) + copy(icmpHdr, h) + icmpHdr.SetType(header.ICMPv4EchoReply) + icmpHdr.SetChecksum(0) + icmpHdr.SetChecksum(^header.Checksum(icmpHdr, header.ChecksumVV(replyData, 0))) + dataVV := buffer.View(icmpHdr).ToVectorisedView() + dataVV.Append(replyData) + replyPkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: int(r.MaxHeaderLength()), + Data: dataVV, + }) + + // Send out the reply packet. sent := stats.ICMP.V4PacketsSent if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{ Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS, - }, &stack.PacketBuffer{ - Header: hdr, - Data: vv, - TransportHeader: buffer.View(pkt), - }); err != nil { + }, replyPkt); err != nil { sent.Dropped.Increment() return } diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index 9ff27a363..3cd48ceb3 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/pkg/tcpip/network/ipv4/ipv4.go @@ -21,7 +21,6 @@ package ipv4 import ( - "fmt" "sync/atomic" "gvisor.dev/gvisor/pkg/tcpip" @@ -127,14 +126,12 @@ func (e *endpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNumber { } // writePacketFragments calls e.linkEP.WritePacket with each packet fragment to -// write. It assumes that the IP header is entirely in pkt.Header but does not -// assume that only the IP header is in pkt.Header. It assumes that the input -// packet's stated length matches the length of the header+payload. mtu -// includes the IP header and options. This does not support the DontFragment -// IP flag. +// write. It assumes that the IP header is already present in pkt.NetworkHeader. +// pkt.TransportHeader may be set. mtu includes the IP header and options. This +// does not support the DontFragment IP flag. func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, pkt *stack.PacketBuffer) *tcpip.Error { // This packet is too big, it needs to be fragmented. - ip := header.IPv4(pkt.Header.View()) + ip := header.IPv4(pkt.NetworkHeader().View()) flags := ip.Flags() // Update mtu to take into account the header, which will exist in all @@ -148,88 +145,84 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, outerMTU := innerMTU + int(ip.HeaderLength()) offset := ip.FragmentOffset() - originalAvailableLength := pkt.Header.AvailableLength() + + // Keep the length reserved for link-layer, we need to create fragments with + // the same reserved length. + reservedForLink := pkt.AvailableHeaderBytes() + + // Destroy the packet, pull all payloads out for fragmentation. + transHeader, data := pkt.TransportHeader().View(), pkt.Data + + // Where possible, the first fragment that is sent has the same + // number of bytes reserved for header as the input packet. The link-layer + // endpoint may depend on this for looking at, eg, L4 headers. + transFitsFirst := len(transHeader) <= innerMTU + for i := 0; i < n; i++ { - // Where possible, the first fragment that is sent has the same - // pkt.Header.UsedLength() as the input packet. The link-layer - // endpoint may depend on this for looking at, eg, L4 headers. - h := ip - if i > 0 { - pkt.Header = buffer.NewPrependable(int(ip.HeaderLength()) + originalAvailableLength) - h = header.IPv4(pkt.Header.Prepend(int(ip.HeaderLength()))) - copy(h, ip[:ip.HeaderLength()]) + reserve := reservedForLink + int(ip.HeaderLength()) + if i == 0 && transFitsFirst { + // Reserve for transport header if it's going to be put in the first + // fragment. + reserve += len(transHeader) + } + fragPkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: reserve, + }) + fragPkt.NetworkProtocolNumber = header.IPv4ProtocolNumber + + // Copy data for the fragment. + avail := innerMTU + + if n := len(transHeader); n > 0 { + if n > avail { + n = avail + } + if i == 0 && transFitsFirst { + copy(fragPkt.TransportHeader().Push(n), transHeader) + } else { + fragPkt.Data.AppendView(transHeader[:n:n]) + } + transHeader = transHeader[n:] + avail -= n } + + if avail > 0 { + n := data.Size() + if n > avail { + n = avail + } + data.ReadToVV(&fragPkt.Data, n) + avail -= n + } + + copied := uint16(innerMTU - avail) + + // Set lengths in header and calculate checksum. + h := header.IPv4(fragPkt.NetworkHeader().Push(len(ip))) + copy(h, ip) if i != n-1 { h.SetTotalLength(uint16(outerMTU)) h.SetFlagsFragmentOffset(flags|header.IPv4FlagMoreFragments, offset) } else { - h.SetTotalLength(uint16(h.HeaderLength()) + uint16(pkt.Data.Size())) + h.SetTotalLength(uint16(h.HeaderLength()) + copied) h.SetFlagsFragmentOffset(flags, offset) } h.SetChecksum(0) h.SetChecksum(^h.CalculateChecksum()) - offset += uint16(innerMTU) - if i > 0 { - newPayload := pkt.Data.Clone(nil) - newPayload.CapLength(innerMTU) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ - Header: pkt.Header, - Data: newPayload, - NetworkHeader: buffer.View(h), - NetworkProtocolNumber: header.IPv4ProtocolNumber, - }); err != nil { - return err - } - r.Stats().IP.PacketsSent.Increment() - pkt.Data.TrimFront(newPayload.Size()) - continue - } - // Special handling for the first fragment because it comes - // from the header. - if outerMTU >= pkt.Header.UsedLength() { - // This fragment can fit all of pkt.Header and possibly - // some of pkt.Data, too. - newPayload := pkt.Data.Clone(nil) - newPayloadLength := outerMTU - pkt.Header.UsedLength() - newPayload.CapLength(newPayloadLength) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ - Header: pkt.Header, - Data: newPayload, - NetworkHeader: buffer.View(h), - NetworkProtocolNumber: header.IPv4ProtocolNumber, - }); err != nil { - return err - } - r.Stats().IP.PacketsSent.Increment() - pkt.Data.TrimFront(newPayloadLength) - } else { - // The fragment is too small to fit all of pkt.Header. - startOfHdr := pkt.Header - startOfHdr.TrimBack(pkt.Header.UsedLength() - outerMTU) - emptyVV := buffer.NewVectorisedView(0, []buffer.View{}) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ - Header: startOfHdr, - Data: emptyVV, - NetworkHeader: buffer.View(h), - NetworkProtocolNumber: header.IPv4ProtocolNumber, - }); err != nil { - return err - } - r.Stats().IP.PacketsSent.Increment() - // Add the unused bytes of pkt.Header into the pkt.Data - // that remains to be sent. - restOfHdr := pkt.Header.View()[outerMTU:] - tmp := buffer.NewVectorisedView(len(restOfHdr), []buffer.View{buffer.NewViewFromBytes(restOfHdr)}) - tmp.Append(pkt.Data) - pkt.Data = tmp + offset += copied + + // Send out the fragment. + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, fragPkt); err != nil { + return err } + r.Stats().IP.PacketsSent.Increment() } return nil } -func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadSize int, params stack.NetworkHeaderParams) header.IPv4 { - ip := header.IPv4(hdr.Prepend(header.IPv4MinimumSize)) - length := uint16(hdr.UsedLength() + payloadSize) +func (e *endpoint) addIPHeader(r *stack.Route, pkt *stack.PacketBuffer, params stack.NetworkHeaderParams) { + ip := header.IPv4(pkt.NetworkHeader().Push(header.IPv4MinimumSize)) + length := uint16(pkt.Size()) // RFC 6864 section 4.3 mandates uniqueness of ID values for non-atomic // datagrams. Since the DF bit is never being set here, all datagrams // are non-atomic and need an ID. @@ -245,14 +238,12 @@ func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadS DstAddr: r.RemoteAddress, }) ip.SetChecksum(^ip.CalculateChecksum()) - return ip + pkt.NetworkProtocolNumber = header.IPv4ProtocolNumber } // WritePacket writes a packet to the given destination address and protocol. func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { - ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) - pkt.NetworkHeader = buffer.View(ip) - pkt.NetworkProtocolNumber = header.IPv4ProtocolNumber + e.addIPHeader(r, pkt, params) // iptables filtering. All packets that reach here are locally // generated. @@ -269,7 +260,7 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw // only NATted packets, but removing this check short circuits broadcasts // before they are sent out to other hosts. if pkt.NatDone { - netHeader := header.IPv4(pkt.NetworkHeader) + netHeader := header.IPv4(pkt.NetworkHeader().View()) ep, err := e.stack.FindNetworkEndpoint(header.IPv4ProtocolNumber, netHeader.DestinationAddress()) if err == nil { route := r.ReverseRoute(netHeader.SourceAddress(), netHeader.DestinationAddress()) @@ -286,7 +277,7 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw if r.Loop&stack.PacketOut == 0 { return nil } - if pkt.Header.UsedLength()+pkt.Data.Size() > int(e.linkEP.MTU()) && (gso == nil || gso.Type == stack.GSONone) { + if pkt.Size() > int(e.linkEP.MTU()) && (gso == nil || gso.Type == stack.GSONone) { return e.writePacketFragments(r, gso, int(e.linkEP.MTU()), pkt) } if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, pkt); err != nil { @@ -306,9 +297,7 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe } for pkt := pkts.Front(); pkt != nil; { - ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) - pkt.NetworkHeader = buffer.View(ip) - pkt.NetworkProtocolNumber = header.IPv4ProtocolNumber + e.addIPHeader(r, pkt, params) pkt = pkt.Next() } @@ -333,7 +322,7 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe continue } if _, ok := natPkts[pkt]; ok { - netHeader := header.IPv4(pkt.NetworkHeader) + netHeader := header.IPv4(pkt.NetworkHeader().View()) if ep, err := e.stack.FindNetworkEndpoint(header.IPv4ProtocolNumber, netHeader.DestinationAddress()); err == nil { src := netHeader.SourceAddress() dst := netHeader.DestinationAddress() @@ -402,17 +391,14 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBu r.Stats().IP.PacketsSent.Increment() - ip = ip[:ip.HeaderLength()] - pkt.Header = buffer.NewPrependableFromView(buffer.View(ip)) - pkt.Data.TrimFront(int(ip.HeaderLength())) return e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, pkt) } // HandlePacket is called by the link layer when new ipv4 packets arrive for // this endpoint. func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { - h := header.IPv4(pkt.NetworkHeader) - if !h.IsValid(pkt.Data.Size() + len(pkt.NetworkHeader) + len(pkt.TransportHeader)) { + h := header.IPv4(pkt.NetworkHeader().View()) + if !h.IsValid(pkt.Data.Size() + pkt.NetworkHeader().View().Size() + pkt.TransportHeader().View().Size()) { r.Stats().IP.MalformedPacketsReceived.Increment() return } @@ -426,7 +412,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { } if h.More() || h.FragmentOffset() != 0 { - if pkt.Data.Size()+len(pkt.TransportHeader) == 0 { + if pkt.Data.Size()+pkt.TransportHeader().View().Size() == 0 { // Drop the packet as it's marked as a fragment but has // no payload. r.Stats().IP.MalformedPacketsReceived.Increment() @@ -470,7 +456,6 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { } p := h.TransportProtocol() if p == header.ICMPv4ProtocolNumber { - pkt.NetworkHeader.CapLength(int(h.HeaderLength())) e.handleICMP(r, pkt) return } @@ -560,14 +545,19 @@ func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNu } ipHdr := header.IPv4(hdr) - // If there are options, pull those into hdr as well. - if headerLen := int(ipHdr.HeaderLength()); headerLen > header.IPv4MinimumSize && headerLen <= pkt.Data.Size() { - hdr, ok = pkt.Data.PullUp(headerLen) - if !ok { - panic(fmt.Sprintf("There are only %d bytes in pkt.Data, but there should be at least %d", pkt.Data.Size(), headerLen)) - } - ipHdr = header.IPv4(hdr) + // Header may have options, determine the true header length. + headerLen := int(ipHdr.HeaderLength()) + if headerLen < header.IPv4MinimumSize { + // TODO(gvisor.dev/issue/2404): Per RFC 791, IHL needs to be at least 5 in + // order for the packet to be valid. Figure out if we want to reject this + // case. + headerLen = header.IPv4MinimumSize + } + hdr, ok = pkt.NetworkHeader().Consume(headerLen) + if !ok { + return 0, false, false } + ipHdr = header.IPv4(hdr) // If this is a fragment, don't bother parsing the transport header. parseTransportHeader := true @@ -576,8 +566,6 @@ func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNu } pkt.NetworkProtocolNumber = header.IPv4ProtocolNumber - pkt.NetworkHeader = hdr - pkt.Data.TrimFront(len(hdr)) pkt.Data.CapLength(int(ipHdr.TotalLength()) - len(hdr)) return ipHdr.TransportProtocol(), parseTransportHeader, true } diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go index 63e2c36c2..afd3ac06d 100644 --- a/pkg/tcpip/network/ipv4/ipv4_test.go +++ b/pkg/tcpip/network/ipv4/ipv4_test.go @@ -17,6 +17,7 @@ package ipv4_test import ( "bytes" "encoding/hex" + "fmt" "math/rand" "testing" @@ -91,15 +92,11 @@ func TestExcludeBroadcast(t *testing.T) { }) } -// makeHdrAndPayload generates a randomize packet. hdrLength indicates how much +// makeRandPkt generates a randomize packet. hdrLength indicates how much // data should already be in the header before WritePacket. extraLength // indicates how much extra space should be in the header. The payload is made // from many Views of the sizes listed in viewSizes. -func makeHdrAndPayload(hdrLength int, extraLength int, viewSizes []int) (buffer.Prependable, buffer.VectorisedView) { - hdr := buffer.NewPrependable(hdrLength + extraLength) - hdr.Prepend(hdrLength) - rand.Read(hdr.View()) - +func makeRandPkt(hdrLength int, extraLength int, viewSizes []int) *stack.PacketBuffer { var views []buffer.View totalLength := 0 for _, s := range viewSizes { @@ -108,8 +105,16 @@ func makeHdrAndPayload(hdrLength int, extraLength int, viewSizes []int) (buffer. views = append(views, newView) totalLength += s } - payload := buffer.NewVectorisedView(totalLength, views) - return hdr, payload + + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: hdrLength + extraLength, + Data: buffer.NewVectorisedView(totalLength, views), + }) + pkt.NetworkProtocolNumber = header.IPv4ProtocolNumber + if _, err := rand.Read(pkt.TransportHeader().Push(hdrLength)); err != nil { + panic(fmt.Sprintf("rand.Read: %s", err)) + } + return pkt } // comparePayloads compared the contents of all the packets against the contents @@ -117,9 +122,9 @@ func makeHdrAndPayload(hdrLength int, extraLength int, viewSizes []int) (buffer. func compareFragments(t *testing.T, packets []*stack.PacketBuffer, sourcePacketInfo *stack.PacketBuffer, mtu uint32) { t.Helper() // Make a complete array of the sourcePacketInfo packet. - source := header.IPv4(packets[0].Header.View()[:header.IPv4MinimumSize]) - source = append(source, sourcePacketInfo.Header.View()...) - source = append(source, sourcePacketInfo.Data.ToView()...) + source := header.IPv4(packets[0].NetworkHeader().View()[:header.IPv4MinimumSize]) + vv := buffer.NewVectorisedView(sourcePacketInfo.Size(), sourcePacketInfo.Views()) + source = append(source, vv.ToView()...) // Make a copy of the IP header, which will be modified in some fields to make // an expected header. @@ -132,8 +137,7 @@ func compareFragments(t *testing.T, packets []*stack.PacketBuffer, sourcePacketI var reassembledPayload []byte for i, packet := range packets { // Confirm that the packet is valid. - allBytes := packet.Header.View().ToVectorisedView() - allBytes.Append(packet.Data) + allBytes := buffer.NewVectorisedView(packet.Size(), packet.Views()) ip := header.IPv4(allBytes.ToView()) if !ip.IsValid(len(ip)) { t.Errorf("IP packet is invalid:\n%s", hex.Dump(ip)) @@ -144,10 +148,17 @@ func compareFragments(t *testing.T, packets []*stack.PacketBuffer, sourcePacketI if got, want := len(ip), int(mtu); got > want { t.Errorf("fragment is too large, got %d want %d", got, want) } - if got, want := packet.Header.UsedLength(), sourcePacketInfo.Header.UsedLength()+header.IPv4MinimumSize; i == 0 && want < int(mtu) && got != want { - t.Errorf("first fragment hdr parts should have unmodified length if possible: got %d, want %d", got, want) + if i == 0 { + got := packet.NetworkHeader().View().Size() + packet.TransportHeader().View().Size() + // sourcePacketInfo does not have NetworkHeader added, simulate one. + want := header.IPv4MinimumSize + sourcePacketInfo.TransportHeader().View().Size() + // Check that it kept the transport header in packet.TransportHeader if + // it fits in the first fragment. + if want < int(mtu) && got != want { + t.Errorf("first fragment hdr parts should have unmodified length if possible: got %d, want %d", got, want) + } } - if got, want := packet.Header.AvailableLength(), sourcePacketInfo.Header.AvailableLength()-header.IPv4MinimumSize; got != want { + if got, want := packet.AvailableHeaderBytes(), sourcePacketInfo.AvailableHeaderBytes()-header.IPv4MinimumSize; got != want { t.Errorf("fragment #%d should have the same available space for prepending as source: got %d, want %d", i, got, want) } if got, want := packet.NetworkProtocolNumber, sourcePacketInfo.NetworkProtocolNumber; got != want { @@ -284,22 +295,14 @@ func TestFragmentation(t *testing.T) { for _, ft := range fragTests { t.Run(ft.description, func(t *testing.T) { - hdr, payload := makeHdrAndPayload(ft.hdrLength, ft.extraLength, ft.payloadViewsSizes) - source := &stack.PacketBuffer{ - Header: hdr, - // Save the source payload because WritePacket will modify it. - Data: payload.Clone(nil), - NetworkProtocolNumber: header.IPv4ProtocolNumber, - } + pkt := makeRandPkt(ft.hdrLength, ft.extraLength, ft.payloadViewsSizes) + source := pkt.Clone() c := buildContext(t, nil, ft.mtu) err := c.Route.WritePacket(ft.gso, stack.NetworkHeaderParams{ Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS, - }, &stack.PacketBuffer{ - Header: hdr, - Data: payload, - }) + }, pkt) if err != nil { t.Errorf("err got %v, want %v", err, nil) } @@ -344,16 +347,13 @@ func TestFragmentationErrors(t *testing.T) { for _, ft := range fragTests { t.Run(ft.description, func(t *testing.T) { - hdr, payload := makeHdrAndPayload(ft.hdrLength, header.IPv4MinimumSize, ft.payloadViewsSizes) + pkt := makeRandPkt(ft.hdrLength, header.IPv4MinimumSize, ft.payloadViewsSizes) c := buildContext(t, ft.packetCollectorErrors, ft.mtu) err := c.Route.WritePacket(&stack.GSO{}, stack.NetworkHeaderParams{ Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS, - }, &stack.PacketBuffer{ - Header: hdr, - Data: payload, - }) + }, pkt) for i := 0; i < len(ft.packetCollectorErrors)-1; i++ { if got, want := ft.packetCollectorErrors[i], (*tcpip.Error)(nil); got != want { t.Errorf("ft.packetCollectorErrors[%d] got %v, want %v", i, got, want) @@ -472,9 +472,9 @@ func TestInvalidFragments(t *testing.T) { s.CreateNIC(nicID, sniffer.New(ep)) for _, pkt := range tc.packets { - ep.InjectLinkAddr(header.IPv4ProtocolNumber, remoteLinkAddr, &stack.PacketBuffer{ + ep.InjectLinkAddr(header.IPv4ProtocolNumber, remoteLinkAddr, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: buffer.NewVectorisedView(len(pkt), []buffer.View{pkt}), - }) + })) } if got, want := s.Stats().IP.MalformedPacketsReceived.Value(), tc.wantMalformedIPPackets; got != want { @@ -859,9 +859,9 @@ func TestReceiveFragments(t *testing.T) { vv := hdr.View().ToVectorisedView() vv.AppendView(frag.payload) - e.InjectInbound(header.IPv4ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(header.IPv4ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: vv, - }) + })) } if got, want := s.Stats().UDP.PacketsReceived.Value(), uint64(len(test.expectedPayloads)); got != want { diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index ded91d83a..39ae19295 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/pkg/tcpip/network/ipv6/icmp.go @@ -83,7 +83,7 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer, hasFragme return } h := header.ICMPv6(v) - iph := header.IPv6(pkt.NetworkHeader) + iph := header.IPv6(pkt.NetworkHeader().View()) // Validate ICMPv6 checksum before processing the packet. // @@ -276,8 +276,10 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer, hasFragme optsSerializer := header.NDPOptionsSerializer{ header.NDPTargetLinkLayerAddressOption(r.LocalLinkAddress), } - hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.ICMPv6NeighborAdvertMinimumSize + int(optsSerializer.Length())) - packet := header.ICMPv6(hdr.Prepend(header.ICMPv6NeighborAdvertSize)) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: int(r.MaxHeaderLength()) + header.ICMPv6NeighborAdvertMinimumSize + int(optsSerializer.Length()), + }) + packet := header.ICMPv6(pkt.TransportHeader().Push(header.ICMPv6NeighborAdvertSize)) packet.SetType(header.ICMPv6NeighborAdvert) na := header.NDPNeighborAdvert(packet.NDPPayload()) na.SetSolicitedFlag(solicited) @@ -293,9 +295,7 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer, hasFragme // // The IP Hop Limit field has a value of 255, i.e., the packet // could not possibly have been forwarded by a router. - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ - Header: hdr, - }); err != nil { + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: stack.DefaultTOS}, pkt); err != nil { sent.Dropped.Increment() return } @@ -384,7 +384,7 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer, hasFragme case header.ICMPv6EchoRequest: received.EchoRequest.Increment() - icmpHdr, ok := pkt.Data.PullUp(header.ICMPv6EchoMinimumSize) + icmpHdr, ok := pkt.TransportHeader().Consume(header.ICMPv6EchoMinimumSize) if !ok { received.Invalid.Increment() return @@ -409,16 +409,15 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer, hasFragme // Use the link address from the source of the original packet. r.ResolveWith(remoteLinkAddr) - pkt.Data.TrimFront(header.ICMPv6EchoMinimumSize) - hdr := buffer.NewPrependable(int(r.MaxHeaderLength()) + header.ICMPv6EchoMinimumSize) - packet := header.ICMPv6(hdr.Prepend(header.ICMPv6EchoMinimumSize)) + replyPkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: int(r.MaxHeaderLength()) + header.ICMPv6EchoMinimumSize, + Data: pkt.Data, + }) + packet := header.ICMPv6(replyPkt.TransportHeader().Push(header.ICMPv6EchoMinimumSize)) copy(packet, icmpHdr) packet.SetType(header.ICMPv6EchoReply) packet.SetChecksum(header.ICMPv6Checksum(packet, r.LocalAddress, r.RemoteAddress, pkt.Data)) - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, &stack.PacketBuffer{ - Header: hdr, - Data: pkt.Data, - }); err != nil { + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, replyPkt); err != nil { sent.Dropped.Increment() return } @@ -539,17 +538,19 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, remoteLinkAdd r.RemoteLinkAddress = header.EthernetAddressFromMulticastIPv6Address(snaddr) } - 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(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, buffer.VectorisedView{})) - - length := uint16(hdr.UsedLength()) - ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: int(linkEP.MaxHeaderLength()) + header.IPv6MinimumSize + header.ICMPv6NeighborAdvertSize, + }) + icmpHdr := header.ICMPv6(pkt.TransportHeader().Push(header.ICMPv6NeighborAdvertSize)) + icmpHdr.SetType(header.ICMPv6NeighborSolicit) + copy(icmpHdr[icmpV6OptOffset-len(addr):], addr) + icmpHdr[icmpV6OptOffset] = ndpOptSrcLinkAddr + icmpHdr[icmpV6LengthOffset] = 1 + copy(icmpHdr[icmpV6LengthOffset+1:], linkEP.LinkAddress()) + icmpHdr.SetChecksum(header.ICMPv6Checksum(icmpHdr, r.LocalAddress, r.RemoteAddress, buffer.VectorisedView{})) + + length := uint16(pkt.Size()) + ip := header.IPv6(pkt.NetworkHeader().Push(header.IPv6MinimumSize)) ip.Encode(&header.IPv6Fields{ PayloadLength: length, NextHeader: uint8(header.ICMPv6ProtocolNumber), @@ -559,9 +560,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, remoteLinkAdd }) // TODO(stijlist): count this in ICMP stats. - return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ - Header: hdr, - }) + return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, pkt) } // ResolveStaticAddress implements stack.LinkAddressResolver. diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go index f86aaed1d..2a2f7de01 100644 --- a/pkg/tcpip/network/ipv6/icmp_test.go +++ b/pkg/tcpip/network/ipv6/icmp_test.go @@ -183,7 +183,11 @@ func TestICMPCounts(t *testing.T) { } handleIPv6Payload := func(icmp header.ICMPv6) { - ip := header.IPv6(buffer.NewView(header.IPv6MinimumSize)) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: header.IPv6MinimumSize, + Data: buffer.View(icmp).ToVectorisedView(), + }) + ip := header.IPv6(pkt.NetworkHeader().Push(header.IPv6MinimumSize)) ip.Encode(&header.IPv6Fields{ PayloadLength: uint16(len(icmp)), NextHeader: uint8(header.ICMPv6ProtocolNumber), @@ -191,10 +195,7 @@ func TestICMPCounts(t *testing.T) { SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) - ep.HandlePacket(&r, &stack.PacketBuffer{ - NetworkHeader: buffer.View(ip), - Data: buffer.View(icmp).ToVectorisedView(), - }) + ep.HandlePacket(&r, pkt) } for _, typ := range types { @@ -323,12 +324,10 @@ func routeICMPv6Packet(t *testing.T, args routeArgs, fn func(*testing.T, header. pi, _ := args.src.ReadContext(context.Background()) { - views := []buffer.View{pi.Pkt.Header.View(), pi.Pkt.Data.ToView()} - size := pi.Pkt.Header.UsedLength() + pi.Pkt.Data.Size() - vv := buffer.NewVectorisedView(size, views) - args.dst.InjectLinkAddr(pi.Proto, args.dst.LinkAddress(), &stack.PacketBuffer{ - Data: vv, + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + Data: buffer.NewVectorisedView(pi.Pkt.Size(), pi.Pkt.Views()), }) + args.dst.InjectLinkAddr(pi.Proto, args.dst.LinkAddress(), pkt) } if pi.Proto != ProtocolNumber { @@ -340,7 +339,9 @@ func routeICMPv6Packet(t *testing.T, args routeArgs, fn func(*testing.T, header. t.Errorf("got remote link address = %s, want = %s", pi.Route.RemoteLinkAddress, args.remoteLinkAddr) } - ipv6 := header.IPv6(pi.Pkt.Header.View()) + // Pull the full payload since network header. Needed for header.IPv6 to + // extract its payload. + ipv6 := header.IPv6(stack.PayloadSince(pi.Pkt.NetworkHeader())) transProto := tcpip.TransportProtocolNumber(ipv6.NextHeader()) if transProto != header.ICMPv6ProtocolNumber { t.Errorf("unexpected transport protocol number %d", transProto) @@ -558,9 +559,10 @@ func TestICMPChecksumValidationSimple(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: buffer.NewVectorisedView(len(ip)+len(icmp), []buffer.View{buffer.View(ip), buffer.View(icmp)}), }) + e.InjectInbound(ProtocolNumber, pkt) } stats := s.Stats().ICMP.V6PacketsReceived @@ -719,12 +721,12 @@ func TestICMPChecksumValidationWithPayload(t *testing.T) { handleIPv6Payload := func(typ header.ICMPv6Type, size, payloadSize int, payloadFn func(buffer.View), checksum bool) { icmpSize := size + payloadSize hdr := buffer.NewPrependable(header.IPv6MinimumSize + icmpSize) - pkt := header.ICMPv6(hdr.Prepend(icmpSize)) - pkt.SetType(typ) - payloadFn(pkt.Payload()) + icmpHdr := header.ICMPv6(hdr.Prepend(icmpSize)) + icmpHdr.SetType(typ) + payloadFn(icmpHdr.Payload()) if checksum { - pkt.SetChecksum(header.ICMPv6Checksum(pkt, lladdr1, lladdr0, buffer.VectorisedView{})) + icmpHdr.SetChecksum(header.ICMPv6Checksum(icmpHdr, lladdr1, lladdr0, buffer.VectorisedView{})) } ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) @@ -735,9 +737,10 @@ func TestICMPChecksumValidationWithPayload(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), }) + e.InjectInbound(ProtocolNumber, pkt) } stats := s.Stats().ICMP.V6PacketsReceived @@ -895,14 +898,14 @@ func TestICMPChecksumValidationWithPayloadMultipleViews(t *testing.T) { handleIPv6Payload := func(typ header.ICMPv6Type, size, payloadSize int, payloadFn func(buffer.View), checksum bool) { hdr := buffer.NewPrependable(header.IPv6MinimumSize + size) - pkt := header.ICMPv6(hdr.Prepend(size)) - pkt.SetType(typ) + icmpHdr := header.ICMPv6(hdr.Prepend(size)) + icmpHdr.SetType(typ) payload := buffer.NewView(payloadSize) payloadFn(payload) if checksum { - pkt.SetChecksum(header.ICMPv6Checksum(pkt, lladdr1, lladdr0, payload.ToVectorisedView())) + icmpHdr.SetChecksum(header.ICMPv6Checksum(icmpHdr, lladdr1, lladdr0, payload.ToVectorisedView())) } ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) @@ -913,9 +916,10 @@ func TestICMPChecksumValidationWithPayloadMultipleViews(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: buffer.NewVectorisedView(header.IPv6MinimumSize+size+payloadSize, []buffer.View{hdr.View(), payload}), }) + e.InjectInbound(ProtocolNumber, pkt) } stats := s.Stats().ICMP.V6PacketsReceived diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go index d7d7fc611..0ade655b2 100644 --- a/pkg/tcpip/network/ipv6/ipv6.go +++ b/pkg/tcpip/network/ipv6/ipv6.go @@ -99,9 +99,9 @@ func (e *endpoint) GSOMaxSize() uint32 { return 0 } -func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadSize int, params stack.NetworkHeaderParams) header.IPv6 { - length := uint16(hdr.UsedLength() + payloadSize) - ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) +func (e *endpoint) addIPHeader(r *stack.Route, pkt *stack.PacketBuffer, params stack.NetworkHeaderParams) { + length := uint16(pkt.Size()) + ip := header.IPv6(pkt.NetworkHeader().Push(header.IPv6MinimumSize)) ip.Encode(&header.IPv6Fields{ PayloadLength: length, NextHeader: uint8(params.Protocol), @@ -110,26 +110,20 @@ func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadS SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) - return ip + pkt.NetworkProtocolNumber = header.IPv6ProtocolNumber } // WritePacket writes a packet to the given destination address and protocol. func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { - ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) - pkt.NetworkHeader = buffer.View(ip) - pkt.NetworkProtocolNumber = header.IPv6ProtocolNumber + e.addIPHeader(r, pkt, params) if r.Loop&stack.PacketLoop != 0 { - // The inbound path expects the network header to still be in - // the PacketBuffer's Data field. - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) loopedR := r.MakeLoopedRoute() - e.HandlePacket(&loopedR, &stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), - }) + e.HandlePacket(&loopedR, stack.NewPacketBuffer(stack.PacketBufferOptions{ + // The inbound path expects an unparsed packet. + Data: buffer.NewVectorisedView(pkt.Size(), pkt.Views()), + })) loopedR.Release() } @@ -151,9 +145,7 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe } for pb := pkts.Front(); pb != nil; pb = pb.Next() { - ip := e.addIPHeader(r, &pb.Header, pb.Data.Size(), params) - pb.NetworkHeader = buffer.View(ip) - pb.NetworkProtocolNumber = header.IPv6ProtocolNumber + e.addIPHeader(r, pb, params) } n, err := e.linkEP.WritePackets(r, gso, pkts, ProtocolNumber) @@ -171,8 +163,8 @@ func (*endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuff // HandlePacket is called by the link layer when new ipv6 packets arrive for // this endpoint. func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { - h := header.IPv6(pkt.NetworkHeader) - if !h.IsValid(pkt.Data.Size() + len(pkt.NetworkHeader) + len(pkt.TransportHeader)) { + h := header.IPv6(pkt.NetworkHeader().View()) + if !h.IsValid(pkt.Data.Size() + pkt.NetworkHeader().View().Size() + pkt.TransportHeader().View().Size()) { r.Stats().IP.MalformedPacketsReceived.Increment() return } @@ -181,8 +173,8 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { // - Any IPv6 header bytes after the first 40 (i.e. extensions). // - The transport header, if present. // - Any other payload data. - vv := pkt.NetworkHeader[header.IPv6MinimumSize:].ToVectorisedView() - vv.AppendView(pkt.TransportHeader) + vv := pkt.NetworkHeader().View()[header.IPv6MinimumSize:].ToVectorisedView() + vv.AppendView(pkt.TransportHeader().View()) vv.Append(pkt.Data) it := header.MakeIPv6PayloadIterator(header.IPv6ExtensionHeaderIdentifier(h.NextHeader()), vv) hasFragmentHeader := false @@ -410,7 +402,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { // // For reassembled fragments, pkt.TransportHeader is unset, so this is a // no-op and pkt.Data begins with the transport header. - extHdr.Buf.TrimFront(len(pkt.TransportHeader)) + extHdr.Buf.TrimFront(pkt.TransportHeader().View().Size()) pkt.Data = extHdr.Buf if p := tcpip.TransportProtocolNumber(extHdr.Identifier); p == header.ICMPv6ProtocolNumber { @@ -581,17 +573,14 @@ traverseExtensions: } } - // Put the IPv6 header with extensions in pkt.NetworkHeader. - hdr, ok = pkt.Data.PullUp(header.IPv6MinimumSize + extensionsSize) + // Put the IPv6 header with extensions in pkt.NetworkHeader(). + hdr, ok = pkt.NetworkHeader().Consume(header.IPv6MinimumSize + extensionsSize) if !ok { panic(fmt.Sprintf("pkt.Data should have at least %d bytes, but only has %d.", header.IPv6MinimumSize+extensionsSize, pkt.Data.Size())) } ipHdr = header.IPv6(hdr) - - pkt.NetworkProtocolNumber = header.IPv6ProtocolNumber - pkt.NetworkHeader = hdr - pkt.Data.TrimFront(len(hdr)) pkt.Data.CapLength(int(ipHdr.PayloadLength())) + pkt.NetworkProtocolNumber = header.IPv6ProtocolNumber return nextHdr, foundNext, true } diff --git a/pkg/tcpip/network/ipv6/ipv6_test.go b/pkg/tcpip/network/ipv6/ipv6_test.go index 3d65814de..081afb051 100644 --- a/pkg/tcpip/network/ipv6/ipv6_test.go +++ b/pkg/tcpip/network/ipv6/ipv6_test.go @@ -65,9 +65,9 @@ func testReceiveICMP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst DstAddr: dst, }) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) stats := s.Stats().ICMP.V6PacketsReceived @@ -123,9 +123,9 @@ func testReceiveUDP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst DstAddr: dst, }) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) stat := s.Stats().UDP.PacketsReceived @@ -637,9 +637,9 @@ func TestReceiveIPv6ExtHdrs(t *testing.T) { DstAddr: addr2, }) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) stats := s.Stats().UDP.PacketsReceived @@ -1469,9 +1469,9 @@ func TestReceiveIPv6Fragments(t *testing.T) { vv := hdr.View().ToVectorisedView() vv.Append(f.data) - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: vv, - }) + })) } if got, want := s.Stats().UDP.PacketsReceived.Value(), uint64(len(test.expectedPayloads)); got != want { diff --git a/pkg/tcpip/network/ipv6/ndp_test.go b/pkg/tcpip/network/ipv6/ndp_test.go index 64239ce9a..fe159b24f 100644 --- a/pkg/tcpip/network/ipv6/ndp_test.go +++ b/pkg/tcpip/network/ipv6/ndp_test.go @@ -136,9 +136,9 @@ func TestNeighorSolicitationWithSourceLinkLayerOption(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) linkAddr, c, err := s.GetLinkAddress(nicID, lladdr1, lladdr0, ProtocolNumber, nil) if linkAddr != test.expectedLinkAddr { @@ -380,9 +380,9 @@ func TestNeighorSolicitationResponse(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, &stack.PacketBuffer{ + e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) if test.nsInvalid { if got := invalid.Value(); got != 1 { @@ -410,7 +410,7 @@ func TestNeighorSolicitationResponse(t *testing.T) { t.Errorf("got p.Route.RemoteLinkAddress = %s, want = %s", p.Route.RemoteLinkAddress, test.naDstLinkAddr) } - checker.IPv6(t, p.Pkt.Header.View(), + checker.IPv6(t, stack.PayloadSince(p.Pkt.NetworkHeader()), checker.SrcAddr(test.naSrc), checker.DstAddr(test.naDst), checker.TTL(header.NDPHopLimit), @@ -497,9 +497,9 @@ func TestNeighorAdvertisementWithTargetLinkLayerOption(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) linkAddr, c, err := s.GetLinkAddress(nicID, lladdr1, lladdr0, ProtocolNumber, nil) if linkAddr != test.expectedLinkAddr { @@ -560,7 +560,11 @@ func TestNDPValidation(t *testing.T) { nextHdr = uint8(header.IPv6FragmentExtHdrIdentifier) } - ip := header.IPv6(buffer.NewView(header.IPv6MinimumSize + len(extensions))) + pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{ + ReserveHeaderBytes: header.IPv6MinimumSize + len(extensions), + Data: payload.ToVectorisedView(), + }) + ip := header.IPv6(pkt.NetworkHeader().Push(header.IPv6MinimumSize + len(extensions))) ip.Encode(&header.IPv6Fields{ PayloadLength: uint16(len(payload) + len(extensions)), NextHeader: nextHdr, @@ -571,10 +575,7 @@ func TestNDPValidation(t *testing.T) { if n := copy(ip[header.IPv6MinimumSize:], extensions); n != len(extensions) { t.Fatalf("expected to write %d bytes of extensions, but wrote %d", len(extensions), n) } - ep.HandlePacket(r, &stack.PacketBuffer{ - NetworkHeader: buffer.View(ip), - Data: payload.ToVectorisedView(), - }) + ep.HandlePacket(r, pkt) } var tllData [header.NDPLinkLayerAddressSize]byte @@ -885,9 +886,9 @@ func TestRouterAdvertValidation(t *testing.T) { t.Fatalf("got rxRA = %d, want = 0", got) } - e.InjectInbound(header.IPv6ProtocolNumber, &stack.PacketBuffer{ + e.InjectInbound(header.IPv6ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{ Data: hdr.View().ToVectorisedView(), - }) + })) if got := rxRA.Value(); got != 1 { t.Fatalf("got rxRA = %d, want = 1", got) -- cgit v1.2.3