diff options
Diffstat (limited to 'pkg/tcpip/network/ipv4/ipv4.go')
-rw-r--r-- | pkg/tcpip/network/ipv4/ipv4.go | 135 |
1 files changed, 109 insertions, 26 deletions
diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index b7b07a6c1..ac16c8add 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/pkg/tcpip/network/ipv4/ipv4.go @@ -39,12 +39,15 @@ const ( // TotalLength field of the ipv4 header. MaxTotalSize = 0xffff + // DefaultTTL is the default time-to-live value for this endpoint. + DefaultTTL = 64 + // buckets is the number of identifier buckets. buckets = 2048 ) type endpoint struct { - nicid tcpip.NICID + nicID tcpip.NICID id stack.NetworkEndpointID prefixLen int linkEP stack.LinkEndpoint @@ -54,9 +57,9 @@ type endpoint struct { } // NewEndpoint creates a new ipv4 endpoint. -func (p *protocol) NewEndpoint(nicid tcpip.NICID, addrWithPrefix tcpip.AddressWithPrefix, linkAddrCache stack.LinkAddressCache, dispatcher stack.TransportDispatcher, linkEP stack.LinkEndpoint) (stack.NetworkEndpoint, *tcpip.Error) { +func (p *protocol) NewEndpoint(nicID tcpip.NICID, addrWithPrefix tcpip.AddressWithPrefix, linkAddrCache stack.LinkAddressCache, dispatcher stack.TransportDispatcher, linkEP stack.LinkEndpoint) (stack.NetworkEndpoint, *tcpip.Error) { e := &endpoint{ - nicid: nicid, + nicID: nicID, id: stack.NetworkEndpointID{LocalAddress: addrWithPrefix.Address}, prefixLen: addrWithPrefix.PrefixLen, linkEP: linkEP, @@ -70,7 +73,7 @@ func (p *protocol) NewEndpoint(nicid tcpip.NICID, addrWithPrefix tcpip.AddressWi // DefaultTTL is the default time-to-live value for this endpoint. func (e *endpoint) DefaultTTL() uint8 { - return 255 + return e.protocol.DefaultTTL() } // MTU implements stack.NetworkEndpoint.MTU. It returns the link-layer MTU minus @@ -86,7 +89,7 @@ func (e *endpoint) Capabilities() stack.LinkEndpointCapabilities { // NICID returns the ID of the NIC this endpoint belongs to. func (e *endpoint) NICID() tcpip.NICID { - return e.nicid + return e.nicID } // ID returns the ipv4 endpoint ID. @@ -195,34 +198,44 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, hdr buff return nil } -// WritePacket writes a packet to the given destination address and protocol. -func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, loop stack.PacketLooping) *tcpip.Error { +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() + payload.Size()) + length := uint16(hdr.UsedLength() + payloadSize) id := uint32(0) 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(&e.protocol.ids[hashRoute(r, protocol, e.protocol.hashIV)%buckets], 1) + id = atomic.AddUint32(&e.protocol.ids[hashRoute(r, params.Protocol, e.protocol.hashIV)%buckets], 1) } ip.Encode(&header.IPv4Fields{ IHL: header.IPv4MinimumSize, TotalLength: length, ID: uint16(id), - TTL: ttl, - Protocol: uint8(protocol), + TTL: params.TTL, + TOS: params.TOS, + Protocol: uint8(params.Protocol), SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) ip.SetChecksum(^ip.CalculateChecksum()) + return ip +} + +// WritePacket writes a packet to the given destination address and protocol. +func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, params stack.NetworkHeaderParams, loop stack.PacketLooping) *tcpip.Error { + ip := e.addIPHeader(r, &hdr, payload.Size(), params) if loop&stack.PacketLoop != 0 { views := make([]buffer.View, 1, 1+len(payload.Views())) views[0] = hdr.View() views = append(views, payload.Views()...) - vv := buffer.NewVectorisedView(len(views[0])+payload.Size(), views) loopedR := r.MakeLoopedRoute() - e.HandlePacket(&loopedR, vv) + + e.HandlePacket(&loopedR, tcpip.PacketBuffer{ + Data: buffer.NewVectorisedView(len(views[0])+payload.Size(), views), + NetworkHeader: buffer.View(ip), + }) + loopedR.Release() } if loop&stack.PacketOut == 0 { @@ -238,6 +251,23 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prepen return nil } +// WritePackets implements stack.NetworkEndpoint.WritePackets. +func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, params stack.NetworkHeaderParams, loop stack.PacketLooping) (int, *tcpip.Error) { + if loop&stack.PacketLoop != 0 { + panic("multiple packets in local loop") + } + if loop&stack.PacketOut == 0 { + return len(hdrs), nil + } + + for i := range hdrs { + e.addIPHeader(r, &hdrs[i].Hdr, hdrs[i].Size, params) + } + n, err := e.linkEP.WritePackets(r, gso, hdrs, payload, ProtocolNumber) + r.Stats().IP.PacketsSent.IncrementBy(uint64(n)) + return n, err +} + // WriteHeaderIncludedPacket writes a packet already containing a network // header through the given route. func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, payload buffer.VectorisedView, loop stack.PacketLooping) *tcpip.Error { @@ -276,7 +306,10 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, payload buffer.Vect ip.SetChecksum(^ip.CalculateChecksum()) if loop&stack.PacketLoop != 0 { - e.HandlePacket(r, payload) + e.HandlePacket(r, tcpip.PacketBuffer{ + Data: payload, + NetworkHeader: buffer.View(ip), + }) } if loop&stack.PacketOut == 0 { return nil @@ -289,24 +322,47 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, payload buffer.Vect // HandlePacket is called by the link layer when new ipv4 packets arrive for // this endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, vv buffer.VectorisedView) { - headerView := vv.First() +func (e *endpoint) HandlePacket(r *stack.Route, pkt tcpip.PacketBuffer) { + headerView := pkt.Data.First() h := header.IPv4(headerView) - if !h.IsValid(vv.Size()) { + if !h.IsValid(pkt.Data.Size()) { + r.Stats().IP.MalformedPacketsReceived.Increment() return } + pkt.NetworkHeader = headerView[:h.HeaderLength()] hlen := int(h.HeaderLength()) tlen := int(h.TotalLength()) - vv.TrimFront(hlen) - vv.CapLength(tlen - hlen) + pkt.Data.TrimFront(hlen) + pkt.Data.CapLength(tlen - hlen) more := (h.Flags() & header.IPv4FlagMoreFragments) != 0 if more || h.FragmentOffset() != 0 { + if pkt.Data.Size() == 0 { + // Drop the packet as it's marked as a fragment but has + // no payload. + r.Stats().IP.MalformedPacketsReceived.Increment() + r.Stats().IP.MalformedFragmentsReceived.Increment() + return + } // The packet is a fragment, let's try to reassemble it. - last := h.FragmentOffset() + uint16(vv.Size()) - 1 + last := h.FragmentOffset() + uint16(pkt.Data.Size()) - 1 + // Drop the packet if the fragmentOffset is incorrect. i.e the + // combination of fragmentOffset and pkt.Data.size() causes a + // wrap around resulting in last being less than the offset. + if last < h.FragmentOffset() { + r.Stats().IP.MalformedPacketsReceived.Increment() + r.Stats().IP.MalformedFragmentsReceived.Increment() + return + } var ready bool - vv, ready = e.fragmentation.Process(hash.IPv4FragmentHash(h), h.FragmentOffset(), last, more, vv) + var err error + pkt.Data, ready, err = e.fragmentation.Process(hash.IPv4FragmentHash(h), h.FragmentOffset(), last, more, pkt.Data) + if err != nil { + r.Stats().IP.MalformedPacketsReceived.Increment() + r.Stats().IP.MalformedFragmentsReceived.Increment() + return + } if !ready { return } @@ -314,11 +370,11 @@ func (e *endpoint) HandlePacket(r *stack.Route, vv buffer.VectorisedView) { p := h.TransportProtocol() if p == header.ICMPv4ProtocolNumber { headerView.CapLength(hlen) - e.handleICMP(r, headerView, vv) + e.handleICMP(r, pkt) return } r.Stats().IP.PacketsDelivered.Increment() - e.dispatcher.DeliverTransportPacket(r, p, headerView, vv) + e.dispatcher.DeliverTransportPacket(r, p, pkt) } // Close cleans up resources associated with the endpoint. @@ -327,6 +383,11 @@ func (e *endpoint) Close() {} type protocol struct { ids []uint32 hashIV uint32 + + // defaultTTL is the current default TTL for the protocol. Only the + // uint8 portion of it is meaningful and it must be accessed + // atomically. + defaultTTL uint32 } // Number returns the ipv4 protocol number. @@ -352,12 +413,34 @@ func (*protocol) ParseAddresses(v buffer.View) (src, dst tcpip.Address) { // SetOption implements NetworkProtocol.SetOption. func (p *protocol) SetOption(option interface{}) *tcpip.Error { - return tcpip.ErrUnknownProtocolOption + switch v := option.(type) { + case tcpip.DefaultTTLOption: + p.SetDefaultTTL(uint8(v)) + return nil + default: + return tcpip.ErrUnknownProtocolOption + } } // Option implements NetworkProtocol.Option. func (p *protocol) Option(option interface{}) *tcpip.Error { - return tcpip.ErrUnknownProtocolOption + switch v := option.(type) { + case *tcpip.DefaultTTLOption: + *v = tcpip.DefaultTTLOption(p.DefaultTTL()) + return nil + default: + return tcpip.ErrUnknownProtocolOption + } +} + +// SetDefaultTTL sets the default TTL for endpoints created with this protocol. +func (p *protocol) SetDefaultTTL(ttl uint8) { + atomic.StoreUint32(&p.defaultTTL, uint32(ttl)) +} + +// DefaultTTL returns the default TTL for endpoints created with this protocol. +func (p *protocol) DefaultTTL() uint8 { + return uint8(atomic.LoadUint32(&p.defaultTTL)) } // calculateMTU calculates the network-layer payload MTU based on the link-layer @@ -391,5 +474,5 @@ func NewProtocol() stack.NetworkProtocol { } hashIV := r[buckets] - return &protocol{ids: ids, hashIV: hashIV} + return &protocol{ids: ids, hashIV: hashIV, defaultTTL: DefaultTTL} } |