diff options
Diffstat (limited to 'pkg/tcpip/link/fdbased')
-rw-r--r-- | pkg/tcpip/link/fdbased/endpoint.go | 143 | ||||
-rw-r--r-- | pkg/tcpip/link/fdbased/endpoint_test.go | 21 | ||||
-rw-r--r-- | pkg/tcpip/link/fdbased/mmap.go | 5 | ||||
-rw-r--r-- | pkg/tcpip/link/fdbased/packet_dispatchers.go | 22 |
4 files changed, 163 insertions, 28 deletions
diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go index f80ac3435..ae4858529 100644 --- a/pkg/tcpip/link/fdbased/endpoint.go +++ b/pkg/tcpip/link/fdbased/endpoint.go @@ -165,6 +165,9 @@ type Options struct { // disabled. GSOMaxSize uint32 + // SoftwareGSOEnabled indicates whether software GSO is enabled or not. + SoftwareGSOEnabled bool + // PacketDispatchMode specifies the type of inbound dispatcher to be // used for this endpoint. PacketDispatchMode PacketDispatchMode @@ -242,7 +245,11 @@ func New(opts *Options) (stack.LinkEndpoint, error) { } if isSocket { if opts.GSOMaxSize != 0 { - e.caps |= stack.CapabilityGSO + if opts.SoftwareGSOEnabled { + e.caps |= stack.CapabilitySoftwareGSO + } else { + e.caps |= stack.CapabilityHardwareGSO + } e.gsoMaxSize = opts.GSOMaxSize } } @@ -397,7 +404,7 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prepen eth.Encode(ethHdr) } - if e.Capabilities()&stack.CapabilityGSO != 0 { + if e.Capabilities()&stack.CapabilityHardwareGSO != 0 { vnetHdr := virtioNetHdr{} vnetHdrBuf := vnetHdrToByteSlice(&vnetHdr) if gso != nil { @@ -430,8 +437,130 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prepen return rawfile.NonBlockingWrite3(e.fds[0], hdr.View(), payload.ToView(), nil) } -// WriteRawPacket writes a raw packet directly to the file descriptor. -func (e *endpoint) WriteRawPacket(dest tcpip.Address, packet []byte) *tcpip.Error { +// WritePackets writes outbound packets to the file descriptor. If it is not +// currently writable, the packet is dropped. +func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) { + var ethHdrBuf []byte + // hdr + data + iovLen := 2 + if e.hdrSize > 0 { + // Add ethernet header if needed. + ethHdrBuf = make([]byte, header.EthernetMinimumSize) + eth := header.Ethernet(ethHdrBuf) + ethHdr := &header.EthernetFields{ + DstAddr: r.RemoteLinkAddress, + Type: protocol, + } + + // Preserve the src address if it's set in the route. + if r.LocalLinkAddress != "" { + ethHdr.SrcAddr = r.LocalLinkAddress + } else { + ethHdr.SrcAddr = e.addr + } + eth.Encode(ethHdr) + iovLen++ + } + + n := len(hdrs) + + views := payload.Views() + /* + * Each bondary in views can add one more iovec. + * + * payload | | | | + * ----------------------------- + * packets | | | | | | | + * ----------------------------- + * iovecs | | | | | | | | | + */ + iovec := make([]syscall.Iovec, n*iovLen+len(views)-1) + mmsgHdrs := make([]rawfile.MMsgHdr, n) + + iovecIdx := 0 + viewIdx := 0 + viewOff := 0 + off := 0 + nextOff := 0 + for i := range hdrs { + prevIovecIdx := iovecIdx + mmsgHdr := &mmsgHdrs[i] + mmsgHdr.Msg.Iov = &iovec[iovecIdx] + packetSize := hdrs[i].Size + hdr := &hdrs[i].Hdr + + off = hdrs[i].Off + if off != nextOff { + // We stop in a different point last time. + size := packetSize + viewIdx = 0 + viewOff = 0 + for size > 0 { + if size >= len(views[viewIdx]) { + viewIdx++ + viewOff = 0 + size -= len(views[viewIdx]) + } else { + viewOff = size + size = 0 + } + } + } + nextOff = off + packetSize + + if ethHdrBuf != nil { + v := &iovec[iovecIdx] + v.Base = ðHdrBuf[0] + v.Len = uint64(len(ethHdrBuf)) + iovecIdx++ + } + + v := &iovec[iovecIdx] + hdrView := hdr.View() + v.Base = &hdrView[0] + v.Len = uint64(len(hdrView)) + iovecIdx++ + + for packetSize > 0 { + vec := &iovec[iovecIdx] + iovecIdx++ + + v := views[viewIdx] + vec.Base = &v[viewOff] + s := len(v) - viewOff + if s <= packetSize { + viewIdx++ + viewOff = 0 + } else { + s = packetSize + viewOff += s + } + vec.Len = uint64(s) + packetSize -= s + } + + mmsgHdr.Msg.Iovlen = uint64(iovecIdx - prevIovecIdx) + } + + packets := 0 + for packets < n { + sent, err := rawfile.NonBlockingSendMMsg(e.fds[0], mmsgHdrs) + if err != nil { + return packets, err + } + packets += sent + mmsgHdrs = mmsgHdrs[sent:] + } + return packets, nil +} + +// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket. +func (e *endpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error { + return rawfile.NonBlockingWrite(e.fds[0], packet.ToView()) +} + +// InjectOutobund implements stack.InjectableEndpoint.InjectOutbound. +func (e *endpoint) InjectOutbound(dest tcpip.Address, packet []byte) *tcpip.Error { return rawfile.NonBlockingWrite(e.fds[0], packet) } @@ -468,9 +597,9 @@ func (e *InjectableEndpoint) Attach(dispatcher stack.NetworkDispatcher) { e.dispatcher = dispatcher } -// Inject injects an inbound packet. -func (e *InjectableEndpoint) Inject(protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView) { - e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, protocol, vv) +// InjectInbound injects an inbound packet. +func (e *InjectableEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView) { + e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, protocol, vv, nil /* linkHeader */) } // NewInjectable creates a new fd-based InjectableEndpoint. diff --git a/pkg/tcpip/link/fdbased/endpoint_test.go b/pkg/tcpip/link/fdbased/endpoint_test.go index 04406bc9a..59378b96c 100644 --- a/pkg/tcpip/link/fdbased/endpoint_test.go +++ b/pkg/tcpip/link/fdbased/endpoint_test.go @@ -43,9 +43,10 @@ const ( ) type packetInfo struct { - raddr tcpip.LinkAddress - proto tcpip.NetworkProtocolNumber - contents buffer.View + raddr tcpip.LinkAddress + proto tcpip.NetworkProtocolNumber + contents buffer.View + linkHeader buffer.View } type context struct { @@ -92,8 +93,8 @@ func (c *context) cleanup() { syscall.Close(c.fds[1]) } -func (c *context) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote tcpip.LinkAddress, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView) { - c.ch <- packetInfo{remote, protocol, vv.ToView()} +func (c *context) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote tcpip.LinkAddress, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView, linkHeader buffer.View) { + c.ch <- packetInfo{remote, protocol, vv.ToView(), linkHeader} } func TestNoEthernetProperties(t *testing.T) { @@ -293,11 +294,12 @@ func TestDeliverPacket(t *testing.T) { b[i] = uint8(rand.Intn(256)) } + var hdr header.Ethernet if !eth { // So that it looks like an IPv4 packet. b[0] = 0x40 } else { - hdr := make(header.Ethernet, header.EthernetMinimumSize) + hdr = make(header.Ethernet, header.EthernetMinimumSize) hdr.Encode(&header.EthernetFields{ SrcAddr: raddr, DstAddr: laddr, @@ -315,9 +317,10 @@ func TestDeliverPacket(t *testing.T) { select { case pi := <-c.ch: want := packetInfo{ - raddr: raddr, - proto: proto, - contents: b, + raddr: raddr, + proto: proto, + contents: b, + linkHeader: buffer.View(hdr), } if !eth { want.proto = header.IPv4ProtocolNumber diff --git a/pkg/tcpip/link/fdbased/mmap.go b/pkg/tcpip/link/fdbased/mmap.go index 8bfeb97e4..554d45715 100644 --- a/pkg/tcpip/link/fdbased/mmap.go +++ b/pkg/tcpip/link/fdbased/mmap.go @@ -169,9 +169,10 @@ func (d *packetMMapDispatcher) dispatch() (bool, *tcpip.Error) { var ( p tcpip.NetworkProtocolNumber remote, local tcpip.LinkAddress + eth header.Ethernet ) if d.e.hdrSize > 0 { - eth := header.Ethernet(pkt) + eth = header.Ethernet(pkt) p = eth.Type() remote = eth.SourceAddress() local = eth.DestinationAddress() @@ -189,6 +190,6 @@ func (d *packetMMapDispatcher) dispatch() (bool, *tcpip.Error) { } pkt = pkt[d.e.hdrSize:] - d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, buffer.NewVectorisedView(len(pkt), []buffer.View{buffer.View(pkt)})) + d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, buffer.NewVectorisedView(len(pkt), []buffer.View{buffer.View(pkt)}), buffer.View(eth)) return true, nil } diff --git a/pkg/tcpip/link/fdbased/packet_dispatchers.go b/pkg/tcpip/link/fdbased/packet_dispatchers.go index 7ca217e5b..12168a1dc 100644 --- a/pkg/tcpip/link/fdbased/packet_dispatchers.go +++ b/pkg/tcpip/link/fdbased/packet_dispatchers.go @@ -53,7 +53,7 @@ func newReadVDispatcher(fd int, e *endpoint) (linkDispatcher, error) { d := &readVDispatcher{fd: fd, e: e} d.views = make([]buffer.View, len(BufConfig)) iovLen := len(BufConfig) - if d.e.Capabilities()&stack.CapabilityGSO != 0 { + if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 { iovLen++ } d.iovecs = make([]syscall.Iovec, iovLen) @@ -63,7 +63,7 @@ func newReadVDispatcher(fd int, e *endpoint) (linkDispatcher, error) { func (d *readVDispatcher) allocateViews(bufConfig []int) { var vnetHdr [virtioNetHdrSize]byte vnetHdrOff := 0 - if d.e.Capabilities()&stack.CapabilityGSO != 0 { + if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 { // The kernel adds virtioNetHdr before each packet, but // we don't use it, so so we allocate a buffer for it, // add it in iovecs but don't add it in a view. @@ -106,7 +106,7 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) { if err != nil { return false, err } - if d.e.Capabilities()&stack.CapabilityGSO != 0 { + if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 { // Skip virtioNetHdr which is added before each packet, it // isn't used and it isn't in a view. n -= virtioNetHdrSize @@ -118,9 +118,10 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) { var ( p tcpip.NetworkProtocolNumber remote, local tcpip.LinkAddress + eth header.Ethernet ) if d.e.hdrSize > 0 { - eth := header.Ethernet(d.views[0]) + eth = header.Ethernet(d.views[0][:header.EthernetMinimumSize]) p = eth.Type() remote = eth.SourceAddress() local = eth.DestinationAddress() @@ -141,7 +142,7 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) { vv := buffer.NewVectorisedView(n, d.views[:used]) vv.TrimFront(d.e.hdrSize) - d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, vv) + d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, vv, buffer.View(eth)) // Prepare e.views for another packet: release used views. for i := 0; i < used; i++ { @@ -194,7 +195,7 @@ func newRecvMMsgDispatcher(fd int, e *endpoint) (linkDispatcher, error) { } d.iovecs = make([][]syscall.Iovec, MaxMsgsPerRecv) iovLen := len(BufConfig) - if d.e.Capabilities()&stack.CapabilityGSO != 0 { + if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 { // virtioNetHdr is prepended before each packet. iovLen++ } @@ -225,7 +226,7 @@ func (d *recvMMsgDispatcher) allocateViews(bufConfig []int) { for k := 0; k < len(d.views); k++ { var vnetHdr [virtioNetHdrSize]byte vnetHdrOff := 0 - if d.e.Capabilities()&stack.CapabilityGSO != 0 { + if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 { // The kernel adds virtioNetHdr before each packet, but // we don't use it, so so we allocate a buffer for it, // add it in iovecs but don't add it in a view. @@ -261,7 +262,7 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) { // Process each of received packets. for k := 0; k < nMsgs; k++ { n := int(d.msgHdrs[k].Len) - if d.e.Capabilities()&stack.CapabilityGSO != 0 { + if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 { n -= virtioNetHdrSize } if n <= d.e.hdrSize { @@ -271,9 +272,10 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) { var ( p tcpip.NetworkProtocolNumber remote, local tcpip.LinkAddress + eth header.Ethernet ) if d.e.hdrSize > 0 { - eth := header.Ethernet(d.views[k][0]) + eth = header.Ethernet(d.views[k][0]) p = eth.Type() remote = eth.SourceAddress() local = eth.DestinationAddress() @@ -293,7 +295,7 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) { used := d.capViews(k, int(n), BufConfig) vv := buffer.NewVectorisedView(int(n), d.views[k][:used]) vv.TrimFront(d.e.hdrSize) - d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, vv) + d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, vv, buffer.View(eth)) // Prepare e.views for another packet: release used views. for i := 0; i < used; i++ { |