summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/link
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/tcpip/link')
-rw-r--r--pkg/tcpip/link/channel/BUILD1
-rw-r--r--pkg/tcpip/link/channel/channel.go24
-rw-r--r--pkg/tcpip/link/fdbased/BUILD2
-rw-r--r--pkg/tcpip/link/fdbased/endpoint.go117
-rw-r--r--pkg/tcpip/link/fdbased/endpoint_test.go213
-rw-r--r--pkg/tcpip/link/fdbased/mmap.go16
-rw-r--r--pkg/tcpip/link/fdbased/packet_dispatchers.go51
-rw-r--r--pkg/tcpip/link/loopback/loopback.go31
-rw-r--r--pkg/tcpip/link/muxed/BUILD1
-rw-r--r--pkg/tcpip/link/muxed/injectable.go16
-rw-r--r--pkg/tcpip/link/muxed/injectable_test.go25
-rw-r--r--pkg/tcpip/link/nested/BUILD32
-rw-r--r--pkg/tcpip/link/nested/nested.go152
-rw-r--r--pkg/tcpip/link/nested/nested_test.go109
-rw-r--r--pkg/tcpip/link/packetsocket/BUILD14
-rw-r--r--pkg/tcpip/link/packetsocket/endpoint.go50
-rw-r--r--pkg/tcpip/link/qdisc/fifo/BUILD1
-rw-r--r--pkg/tcpip/link/qdisc/fifo/endpoint.go26
-rw-r--r--pkg/tcpip/link/rawfile/blockingpoll_yield_unsafe.go2
-rw-r--r--pkg/tcpip/link/rawfile/rawfile_unsafe.go33
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem.go51
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem_test.go141
-rw-r--r--pkg/tcpip/link/sharedmem/tx.go14
-rw-r--r--pkg/tcpip/link/sniffer/BUILD1
-rw-r--r--pkg/tcpip/link/sniffer/sniffer.go129
-rw-r--r--pkg/tcpip/link/tun/BUILD1
-rw-r--r--pkg/tcpip/link/tun/device.go79
-rw-r--r--pkg/tcpip/link/waitable/BUILD2
-rw-r--r--pkg/tcpip/link/waitable/waitable.go26
-rw-r--r--pkg/tcpip/link/waitable/waitable_test.go31
30 files changed, 942 insertions, 449 deletions
diff --git a/pkg/tcpip/link/channel/BUILD b/pkg/tcpip/link/channel/BUILD
index b8b93e78e..39ca774ef 100644
--- a/pkg/tcpip/link/channel/BUILD
+++ b/pkg/tcpip/link/channel/BUILD
@@ -10,6 +10,7 @@ go_library(
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go
index 9bf67686d..c95aef63c 100644
--- a/pkg/tcpip/link/channel/channel.go
+++ b/pkg/tcpip/link/channel/channel.go
@@ -23,6 +23,7 @@ import (
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -181,13 +182,13 @@ func (e *Endpoint) NumQueued() int {
}
// InjectInbound injects an inbound packet.
-func (e *Endpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
+func (e *Endpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
e.InjectLinkAddr(protocol, "", pkt)
}
// InjectLinkAddr injects an inbound packet with a remote link address.
-func (e *Endpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remote tcpip.LinkAddress, pkt stack.PacketBuffer) {
- e.dispatcher.DeliverNetworkPacket(e, remote, "" /* local */, protocol, pkt)
+func (e *Endpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remote tcpip.LinkAddress, pkt *stack.PacketBuffer) {
+ e.dispatcher.DeliverNetworkPacket(remote, "" /* local */, protocol, pkt)
}
// Attach saves the stack network-layer dispatcher for use later when packets
@@ -229,13 +230,13 @@ func (e *Endpoint) LinkAddress() tcpip.LinkAddress {
}
// WritePacket stores outbound packets into the channel.
-func (e *Endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
+func (e *Endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
// Clone r then release its resource so we only get the relevant fields from
// stack.Route without holding a reference to a NIC's endpoint.
route := r.Clone()
route.Release()
p := PacketInfo{
- Pkt: &pkt,
+ Pkt: pkt,
Proto: protocol,
GSO: gso,
Route: route,
@@ -273,7 +274,9 @@ func (e *Endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe
// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
func (e *Endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
p := PacketInfo{
- Pkt: &stack.PacketBuffer{Data: vv},
+ Pkt: stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: vv,
+ }),
Proto: 0,
GSO: nil,
}
@@ -296,3 +299,12 @@ func (e *Endpoint) AddNotify(notify Notification) *NotificationHandle {
func (e *Endpoint) RemoveNotify(handle *NotificationHandle) {
e.q.RemoveNotify(handle)
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*Endpoint) ARPHardwareType() header.ARPHardwareType {
+ return header.ARPHardwareNone
+}
+
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *Endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+}
diff --git a/pkg/tcpip/link/fdbased/BUILD b/pkg/tcpip/link/fdbased/BUILD
index aa6db9aea..10072eac1 100644
--- a/pkg/tcpip/link/fdbased/BUILD
+++ b/pkg/tcpip/link/fdbased/BUILD
@@ -15,6 +15,7 @@ go_library(
visibility = ["//visibility:public"],
deps = [
"//pkg/binary",
+ "//pkg/iovec",
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
@@ -36,5 +37,6 @@ go_test(
"//pkg/tcpip/header",
"//pkg/tcpip/link/rawfile",
"//pkg/tcpip/stack",
+ "@com_github_google_go_cmp//cmp:go_default_library",
],
)
diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go
index affa1bbdf..975309fc8 100644
--- a/pkg/tcpip/link/fdbased/endpoint.go
+++ b/pkg/tcpip/link/fdbased/endpoint.go
@@ -45,6 +45,7 @@ import (
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/binary"
+ "gvisor.dev/gvisor/pkg/iovec"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
@@ -385,32 +386,40 @@ const (
_VIRTIO_NET_HDR_GSO_TCPV6 = 4
)
-// WritePacket writes outbound packets to the file descriptor. If it is not
-// currently writable, the packet is dropped.
-func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
if e.hdrSize > 0 {
// Add ethernet header if needed.
- eth := header.Ethernet(pkt.Header.Prepend(header.EthernetMinimumSize))
- pkt.LinkHeader = buffer.View(eth)
+ eth := header.Ethernet(pkt.LinkHeader().Push(header.EthernetMinimumSize))
ethHdr := &header.EthernetFields{
- DstAddr: r.RemoteLinkAddress,
+ DstAddr: remote,
Type: protocol,
}
// Preserve the src address if it's set in the route.
- if r.LocalLinkAddress != "" {
- ethHdr.SrcAddr = r.LocalLinkAddress
+ if local != "" {
+ ethHdr.SrcAddr = local
} else {
ethHdr.SrcAddr = e.addr
}
eth.Encode(ethHdr)
}
+}
+
+// WritePacket writes outbound packets to the file descriptor. If it is not
+// currently writable, the packet is dropped.
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
+ if e.hdrSize > 0 {
+ e.AddHeader(r.LocalLinkAddress, r.RemoteLinkAddress, protocol, pkt)
+ }
+
+ var builder iovec.Builder
fd := e.fds[pkt.Hash%uint32(len(e.fds))]
if e.Capabilities()&stack.CapabilityHardwareGSO != 0 {
vnetHdr := virtioNetHdr{}
if gso != nil {
- vnetHdr.hdrLen = uint16(pkt.Header.UsedLength())
+ vnetHdr.hdrLen = uint16(pkt.HeaderSize())
if gso.NeedsCsum {
vnetHdr.flags = _VIRTIO_NET_HDR_F_NEEDS_CSUM
vnetHdr.csumStart = header.EthernetMinimumSize + gso.L3HdrLen
@@ -430,49 +439,28 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.Ne
}
vnetHdrBuf := binary.Marshal(make([]byte, 0, virtioNetHdrSize), binary.LittleEndian, vnetHdr)
- return rawfile.NonBlockingWrite3(fd, vnetHdrBuf, pkt.Header.View(), pkt.Data.ToView())
+ builder.Add(vnetHdrBuf)
}
- if pkt.Data.Size() == 0 {
- return rawfile.NonBlockingWrite(fd, pkt.Header.View())
+ for _, v := range pkt.Views() {
+ builder.Add(v)
}
- if pkt.Header.UsedLength() == 0 {
- return rawfile.NonBlockingWrite(fd, pkt.Data.ToView())
- }
-
- return rawfile.NonBlockingWrite3(fd, pkt.Header.View(), pkt.Data.ToView(), nil)
+ return rawfile.NonBlockingWriteIovec(fd, builder.Build())
}
func (e *endpoint) sendBatch(batchFD int, batch []*stack.PacketBuffer) (int, *tcpip.Error) {
// Send a batch of packets through batchFD.
mmsgHdrs := make([]rawfile.MMsgHdr, 0, len(batch))
for _, pkt := range batch {
- var ethHdrBuf []byte
- iovLen := 0
if e.hdrSize > 0 {
- // Add ethernet header if needed.
- ethHdrBuf = make([]byte, header.EthernetMinimumSize)
- eth := header.Ethernet(ethHdrBuf)
- ethHdr := &header.EthernetFields{
- DstAddr: pkt.EgressRoute.RemoteLinkAddress,
- Type: pkt.NetworkProtocolNumber,
- }
-
- // Preserve the src address if it's set in the route.
- if pkt.EgressRoute.LocalLinkAddress != "" {
- ethHdr.SrcAddr = pkt.EgressRoute.LocalLinkAddress
- } else {
- ethHdr.SrcAddr = e.addr
- }
- eth.Encode(ethHdr)
- iovLen++
+ e.AddHeader(pkt.EgressRoute.LocalLinkAddress, pkt.EgressRoute.RemoteLinkAddress, pkt.NetworkProtocolNumber, pkt)
}
- vnetHdr := virtioNetHdr{}
var vnetHdrBuf []byte
if e.Capabilities()&stack.CapabilityHardwareGSO != 0 {
+ vnetHdr := virtioNetHdr{}
if pkt.GSOOptions != nil {
- vnetHdr.hdrLen = uint16(pkt.Header.UsedLength())
+ vnetHdr.hdrLen = uint16(pkt.HeaderSize())
if pkt.GSOOptions.NeedsCsum {
vnetHdr.flags = _VIRTIO_NET_HDR_F_NEEDS_CSUM
vnetHdr.csumStart = header.EthernetMinimumSize + pkt.GSOOptions.L3HdrLen
@@ -491,45 +479,18 @@ func (e *endpoint) sendBatch(batchFD int, batch []*stack.PacketBuffer) (int, *tc
}
}
vnetHdrBuf = binary.Marshal(make([]byte, 0, virtioNetHdrSize), binary.LittleEndian, vnetHdr)
- iovLen++
}
- iovecs := make([]syscall.Iovec, iovLen+1+len(pkt.Data.Views()))
+ var builder iovec.Builder
+ builder.Add(vnetHdrBuf)
+ for _, v := range pkt.Views() {
+ builder.Add(v)
+ }
+ iovecs := builder.Build()
+
var mmsgHdr rawfile.MMsgHdr
mmsgHdr.Msg.Iov = &iovecs[0]
- iovecIdx := 0
- if vnetHdrBuf != nil {
- v := &iovecs[iovecIdx]
- v.Base = &vnetHdrBuf[0]
- v.Len = uint64(len(vnetHdrBuf))
- iovecIdx++
- }
- if ethHdrBuf != nil {
- v := &iovecs[iovecIdx]
- v.Base = &ethHdrBuf[0]
- v.Len = uint64(len(ethHdrBuf))
- iovecIdx++
- }
- pktSize := uint64(0)
- // Encode L3 Header
- v := &iovecs[iovecIdx]
- hdr := &pkt.Header
- hdrView := hdr.View()
- v.Base = &hdrView[0]
- v.Len = uint64(len(hdrView))
- pktSize += v.Len
- iovecIdx++
-
- // Now encode the Transport Payload.
- pktViews := pkt.Data.Views()
- for i := range pktViews {
- vec := &iovecs[iovecIdx]
- iovecIdx++
- vec.Base = &pktViews[i][0]
- vec.Len = uint64(len(pktViews[i]))
- pktSize += vec.Len
- }
- mmsgHdr.Msg.Iovlen = uint64(iovecIdx)
+ mmsgHdr.Msg.Iovlen = uint64(len(iovecs))
mmsgHdrs = append(mmsgHdrs, mmsgHdr)
}
@@ -626,6 +587,14 @@ func (e *endpoint) GSOMaxSize() uint32 {
return e.gsoMaxSize
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (e *endpoint) ARPHardwareType() header.ARPHardwareType {
+ if e.hdrSize > 0 {
+ return header.ARPHardwareEther
+ }
+ return header.ARPHardwareNone
+}
+
// InjectableEndpoint is an injectable fd-based endpoint. The endpoint writes
// to the FD, but does not read from it. All reads come from injected packets.
type InjectableEndpoint struct {
@@ -641,8 +610,8 @@ func (e *InjectableEndpoint) Attach(dispatcher stack.NetworkDispatcher) {
}
// InjectInbound injects an inbound packet.
-func (e *InjectableEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
- e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, protocol, pkt)
+func (e *InjectableEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, protocol, pkt)
}
// 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 3bfb15a8e..709f829c8 100644
--- a/pkg/tcpip/link/fdbased/endpoint_test.go
+++ b/pkg/tcpip/link/fdbased/endpoint_test.go
@@ -26,6 +26,7 @@ import (
"time"
"unsafe"
+ "github.com/google/go-cmp/cmp"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
"gvisor.dev/gvisor/pkg/tcpip/header"
@@ -43,9 +44,36 @@ const (
)
type packetInfo struct {
- raddr tcpip.LinkAddress
- proto tcpip.NetworkProtocolNumber
- contents stack.PacketBuffer
+ Raddr tcpip.LinkAddress
+ Proto tcpip.NetworkProtocolNumber
+ Contents *stack.PacketBuffer
+}
+
+type packetContents struct {
+ LinkHeader buffer.View
+ NetworkHeader buffer.View
+ TransportHeader buffer.View
+ Data buffer.View
+}
+
+func checkPacketInfoEqual(t *testing.T, got, want packetInfo) {
+ t.Helper()
+ if diff := cmp.Diff(
+ want, got,
+ cmp.Transformer("ExtractPacketBuffer", func(pk *stack.PacketBuffer) *packetContents {
+ if pk == nil {
+ return nil
+ }
+ return &packetContents{
+ LinkHeader: pk.LinkHeader().View(),
+ NetworkHeader: pk.NetworkHeader().View(),
+ TransportHeader: pk.TransportHeader().View(),
+ Data: pk.Data.ToView(),
+ }
+ }),
+ ); diff != "" {
+ t.Errorf("unexpected packetInfo (-want +got):\n%s", diff)
+ }
}
type context struct {
@@ -103,10 +131,14 @@ func (c *context) cleanup() {
}
}
-func (c *context) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote tcpip.LinkAddress, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
+func (c *context) DeliverNetworkPacket(remote tcpip.LinkAddress, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
c.ch <- packetInfo{remote, protocol, pkt}
}
+func (c *context) DeliverOutboundPacket(remote tcpip.LinkAddress, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ panic("unimplemented")
+}
+
func TestNoEthernetProperties(t *testing.T) {
c := newContext(t, &Options{MTU: mtu})
defer c.cleanup()
@@ -155,19 +187,28 @@ func testWritePacket(t *testing.T, plen int, eth bool, gsoMaxSize uint32, hash u
RemoteLinkAddress: raddr,
}
- // Build header.
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()) + 100)
- b := hdr.Prepend(100)
- for i := range b {
- b[i] = uint8(rand.Intn(256))
+ // Build payload.
+ payload := buffer.NewView(plen)
+ if _, err := rand.Read(payload); err != nil {
+ t.Fatalf("rand.Read(payload): %s", err)
}
- // Build payload and write.
- payload := make(buffer.View, plen)
- for i := range payload {
- payload[i] = uint8(rand.Intn(256))
+ // Build packet buffer.
+ const netHdrLen = 100
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()) + netHdrLen,
+ Data: payload.ToVectorisedView(),
+ })
+ pkt.Hash = hash
+
+ // Build header.
+ b := pkt.NetworkHeader().Push(netHdrLen)
+ if _, err := rand.Read(b); err != nil {
+ t.Fatalf("rand.Read(b): %s", err)
}
- want := append(hdr.View(), payload...)
+
+ // Write.
+ want := append(append(buffer.View(nil), b...), payload...)
var gso *stack.GSO
if gsoMaxSize != 0 {
gso = &stack.GSO{
@@ -179,11 +220,7 @@ func testWritePacket(t *testing.T, plen int, eth bool, gsoMaxSize uint32, hash u
L3HdrLen: header.IPv4MaximumHeaderSize,
}
}
- if err := c.ep.WritePacket(r, gso, proto, stack.PacketBuffer{
- Header: hdr,
- Data: payload.ToVectorisedView(),
- Hash: hash,
- }); err != nil {
+ if err := c.ep.WritePacket(r, gso, proto, pkt); err != nil {
t.Fatalf("WritePacket failed: %v", err)
}
@@ -292,13 +329,14 @@ func TestPreserveSrcAddress(t *testing.T) {
LocalLinkAddress: baddr,
}
- // WritePacket panics given a prependable with anything less than
- // the minimum size of the ethernet header.
- hdr := buffer.NewPrependable(header.EthernetMinimumSize)
- if err := c.ep.WritePacket(r, nil /* gso */, proto, stack.PacketBuffer{
- Header: hdr,
- Data: buffer.VectorisedView{},
- }); err != nil {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ // WritePacket panics given a prependable with anything less than
+ // the minimum size of the ethernet header.
+ // TODO(b/153685824): Figure out if this should use c.ep.MaxHeaderLength().
+ ReserveHeaderBytes: header.EthernetMinimumSize,
+ Data: buffer.VectorisedView{},
+ })
+ if err := c.ep.WritePacket(r, nil /* gso */, proto, pkt); err != nil {
t.Fatalf("WritePacket failed: %v", err)
}
@@ -327,24 +365,25 @@ func TestDeliverPacket(t *testing.T) {
defer c.cleanup()
// Build packet.
- b := make([]byte, plen)
- all := b
- for i := range b {
- b[i] = uint8(rand.Intn(256))
+ all := make([]byte, plen)
+ if _, err := rand.Read(all); err != nil {
+ t.Fatalf("rand.Read(all): %s", err)
}
-
- var hdr header.Ethernet
- if !eth {
- // So that it looks like an IPv4 packet.
- b[0] = 0x40
- } else {
- hdr = make(header.Ethernet, header.EthernetMinimumSize)
+ // Make it look like an IPv4 packet.
+ all[0] = 0x40
+
+ wantPkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: header.EthernetMinimumSize,
+ Data: buffer.NewViewFromBytes(all).ToVectorisedView(),
+ })
+ if eth {
+ hdr := header.Ethernet(wantPkt.LinkHeader().Push(header.EthernetMinimumSize))
hdr.Encode(&header.EthernetFields{
SrcAddr: raddr,
DstAddr: laddr,
Type: proto,
})
- all = append(hdr, b...)
+ all = append(hdr, all...)
}
// Write packet via the file descriptor.
@@ -356,24 +395,15 @@ func TestDeliverPacket(t *testing.T) {
select {
case pi := <-c.ch:
want := packetInfo{
- raddr: raddr,
- proto: proto,
- contents: stack.PacketBuffer{
- Data: buffer.View(b).ToVectorisedView(),
- LinkHeader: buffer.View(hdr),
- },
+ Raddr: raddr,
+ Proto: proto,
+ Contents: wantPkt,
}
if !eth {
- want.proto = header.IPv4ProtocolNumber
- want.raddr = ""
- }
- // want.contents.Data will be a single
- // view, so make pi do the same for the
- // DeepEqual check.
- pi.contents.Data = pi.contents.Data.ToView().ToVectorisedView()
- if !reflect.DeepEqual(want, pi) {
- t.Fatalf("Unexpected received packet: %+v, want %+v", pi, want)
+ want.Proto = header.IPv4ProtocolNumber
+ want.Raddr = ""
}
+ checkPacketInfoEqual(t, pi, want)
case <-time.After(10 * time.Second):
t.Fatalf("Timed out waiting for packet")
}
@@ -500,3 +530,80 @@ func TestRecvMMsgDispatcherCapLength(t *testing.T) {
}
}
+
+// fakeNetworkDispatcher delivers packets to pkts.
+type fakeNetworkDispatcher struct {
+ pkts []*stack.PacketBuffer
+}
+
+func (d *fakeNetworkDispatcher) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ d.pkts = append(d.pkts, pkt)
+}
+
+func (d *fakeNetworkDispatcher) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ panic("unimplemented")
+}
+
+func TestDispatchPacketFormat(t *testing.T) {
+ for _, test := range []struct {
+ name string
+ newDispatcher func(fd int, e *endpoint) (linkDispatcher, error)
+ }{
+ {
+ name: "readVDispatcher",
+ newDispatcher: newReadVDispatcher,
+ },
+ {
+ name: "recvMMsgDispatcher",
+ newDispatcher: newRecvMMsgDispatcher,
+ },
+ } {
+ t.Run(test.name, func(t *testing.T) {
+ // Create a socket pair to send/recv.
+ fds, err := syscall.Socketpair(syscall.AF_UNIX, syscall.SOCK_DGRAM, 0)
+ if err != nil {
+ t.Fatal(err)
+ }
+ defer syscall.Close(fds[0])
+ defer syscall.Close(fds[1])
+
+ data := []byte{
+ // Ethernet header.
+ 1, 2, 3, 4, 5, 60,
+ 1, 2, 3, 4, 5, 61,
+ 8, 0,
+ // Mock network header.
+ 40, 41, 42, 43,
+ }
+ err = syscall.Sendmsg(fds[1], data, nil, nil, 0)
+ if err != nil {
+ t.Fatal(err)
+ }
+
+ // Create and run dispatcher once.
+ sink := &fakeNetworkDispatcher{}
+ d, err := test.newDispatcher(fds[0], &endpoint{
+ hdrSize: header.EthernetMinimumSize,
+ dispatcher: sink,
+ })
+ if err != nil {
+ t.Fatal(err)
+ }
+ if ok, err := d.dispatch(); !ok || err != nil {
+ t.Fatalf("d.dispatch() = %v, %v", ok, err)
+ }
+
+ // Verify packet.
+ if got, want := len(sink.pkts), 1; got != want {
+ t.Fatalf("len(sink.pkts) = %d, want %d", got, want)
+ }
+ pkt := sink.pkts[0]
+ if got, want := pkt.LinkHeader().View().Size(), header.EthernetMinimumSize; got != want {
+ t.Errorf("pkt.LinkHeader().View().Size() = %d, want %d", got, want)
+ }
+ if got, want := pkt.Data.Size(), 4; got != want {
+ t.Errorf("pkt.Data.Size() = %d, want %d", got, want)
+ }
+ })
+ }
+}
diff --git a/pkg/tcpip/link/fdbased/mmap.go b/pkg/tcpip/link/fdbased/mmap.go
index fe2bf3b0b..c475dda20 100644
--- a/pkg/tcpip/link/fdbased/mmap.go
+++ b/pkg/tcpip/link/fdbased/mmap.go
@@ -18,6 +18,7 @@ package fdbased
import (
"encoding/binary"
+ "fmt"
"syscall"
"golang.org/x/sys/unix"
@@ -170,10 +171,9 @@ 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()
@@ -190,10 +190,14 @@ func (d *packetMMapDispatcher) dispatch() (bool, *tcpip.Error) {
}
}
- pkt = pkt[d.e.hdrSize:]
- d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, stack.PacketBuffer{
- Data: buffer.View(pkt).ToVectorisedView(),
- LinkHeader: buffer.View(eth),
+ pbuf := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: buffer.View(pkt).ToVectorisedView(),
})
+ if d.e.hdrSize > 0 {
+ if _, ok := pbuf.LinkHeader().Consume(d.e.hdrSize); !ok {
+ panic(fmt.Sprintf("LinkHeader().Consume(%d) must succeed", d.e.hdrSize))
+ }
+ }
+ d.e.dispatcher.DeliverNetworkPacket(remote, local, p, pbuf)
return true, nil
}
diff --git a/pkg/tcpip/link/fdbased/packet_dispatchers.go b/pkg/tcpip/link/fdbased/packet_dispatchers.go
index cb4cbea69..8c3ca86d6 100644
--- a/pkg/tcpip/link/fdbased/packet_dispatchers.go
+++ b/pkg/tcpip/link/fdbased/packet_dispatchers.go
@@ -103,7 +103,7 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) {
d.allocateViews(BufConfig)
n, err := rawfile.BlockingReadv(d.fd, d.iovecs)
- if err != nil {
+ if n == 0 || err != nil {
return false, err
}
if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 {
@@ -111,17 +111,22 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) {
// isn't used and it isn't in a view.
n -= virtioNetHdrSize
}
- if n <= d.e.hdrSize {
- return false, nil
- }
+
+ used := d.capViews(n, BufConfig)
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: buffer.NewVectorisedView(n, append([]buffer.View(nil), d.views[:used]...)),
+ })
var (
p tcpip.NetworkProtocolNumber
remote, local tcpip.LinkAddress
- eth header.Ethernet
)
if d.e.hdrSize > 0 {
- eth = header.Ethernet(d.views[0][:header.EthernetMinimumSize])
+ hdr, ok := pkt.LinkHeader().Consume(d.e.hdrSize)
+ if !ok {
+ return false, nil
+ }
+ eth := header.Ethernet(hdr)
p = eth.Type()
remote = eth.SourceAddress()
local = eth.DestinationAddress()
@@ -138,14 +143,7 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) {
}
}
- used := d.capViews(n, BufConfig)
- pkt := stack.PacketBuffer{
- Data: buffer.NewVectorisedView(n, append([]buffer.View(nil), d.views[:used]...)),
- LinkHeader: buffer.View(eth),
- }
- pkt.Data.TrimFront(d.e.hdrSize)
-
- d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, pkt)
+ d.e.dispatcher.DeliverNetworkPacket(remote, local, p, pkt)
// Prepare e.views for another packet: release used views.
for i := 0; i < used; i++ {
@@ -169,7 +167,7 @@ type recvMMsgDispatcher struct {
// iovecs is an array of array of iovec records where each iovec base
// pointer and length are initialzed to the corresponding view above,
- // except when GSO is neabled then the first iovec in each array of
+ // except when GSO is enabled then the first iovec in each array of
// iovecs points to a buffer for the vnet header which is stripped
// before the views are passed up the stack for further processing.
iovecs [][]syscall.Iovec
@@ -268,17 +266,22 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) {
if d.e.Capabilities()&stack.CapabilityHardwareGSO != 0 {
n -= virtioNetHdrSize
}
- if n <= d.e.hdrSize {
- return false, nil
- }
+
+ used := d.capViews(k, int(n), BufConfig)
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: buffer.NewVectorisedView(int(n), append([]buffer.View(nil), d.views[k][:used]...)),
+ })
var (
p tcpip.NetworkProtocolNumber
remote, local tcpip.LinkAddress
- eth header.Ethernet
)
if d.e.hdrSize > 0 {
- eth = header.Ethernet(d.views[k][0])
+ hdr, ok := pkt.LinkHeader().Consume(d.e.hdrSize)
+ if !ok {
+ return false, nil
+ }
+ eth := header.Ethernet(hdr)
p = eth.Type()
remote = eth.SourceAddress()
local = eth.DestinationAddress()
@@ -295,13 +298,7 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) {
}
}
- used := d.capViews(k, int(n), BufConfig)
- pkt := stack.PacketBuffer{
- Data: buffer.NewVectorisedView(int(n), append([]buffer.View(nil), d.views[k][:used]...)),
- LinkHeader: buffer.View(eth),
- }
- pkt.Data.TrimFront(d.e.hdrSize)
- d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, pkt)
+ d.e.dispatcher.DeliverNetworkPacket(remote, local, p, pkt)
// Prepare e.views for another packet: release used views.
for i := 0; i < used; i++ {
diff --git a/pkg/tcpip/link/loopback/loopback.go b/pkg/tcpip/link/loopback/loopback.go
index 073c84ef9..38aa694e4 100644
--- a/pkg/tcpip/link/loopback/loopback.go
+++ b/pkg/tcpip/link/loopback/loopback.go
@@ -76,17 +76,17 @@ func (*endpoint) Wait() {}
// WritePacket implements stack.LinkEndpoint.WritePacket. It delivers outbound
// packets to the network-layer dispatcher.
-func (e *endpoint) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
- views := make([]buffer.View, 1, 1+len(pkt.Data.Views()))
- views[0] = pkt.Header.View()
- views = append(views, pkt.Data.Views()...)
+func (e *endpoint) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
+ // Construct data as the unparsed portion for the loopback packet.
+ data := buffer.NewVectorisedView(pkt.Size(), pkt.Views())
// Because we're immediately turning around and writing the packet back
// to the rx path, we intentionally don't preserve the remote and local
// link addresses from the stack.Route we're passed.
- e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, protocol, stack.PacketBuffer{
- Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views),
+ newPkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: data,
})
+ e.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, protocol, newPkt)
return nil
}
@@ -98,18 +98,25 @@ func (e *endpoint) WritePackets(*stack.Route, *stack.GSO, stack.PacketBufferList
// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: vv,
+ })
// There should be an ethernet header at the beginning of vv.
- hdr, ok := vv.PullUp(header.EthernetMinimumSize)
+ hdr, ok := pkt.LinkHeader().Consume(header.EthernetMinimumSize)
if !ok {
// Reject the packet if it's shorter than an ethernet header.
return tcpip.ErrBadAddress
}
linkHeader := header.Ethernet(hdr)
- vv.TrimFront(len(linkHeader))
- e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, linkHeader.Type(), stack.PacketBuffer{
- Data: vv,
- LinkHeader: buffer.View(linkHeader),
- })
+ e.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, linkHeader.Type(), pkt)
return nil
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*endpoint) ARPHardwareType() header.ARPHardwareType {
+ return header.ARPHardwareLoopback
+}
+
+func (e *endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+}
diff --git a/pkg/tcpip/link/muxed/BUILD b/pkg/tcpip/link/muxed/BUILD
index 82b441b79..e7493e5c5 100644
--- a/pkg/tcpip/link/muxed/BUILD
+++ b/pkg/tcpip/link/muxed/BUILD
@@ -9,6 +9,7 @@ go_library(
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/muxed/injectable.go b/pkg/tcpip/link/muxed/injectable.go
index a5478ce17..56a611825 100644
--- a/pkg/tcpip/link/muxed/injectable.go
+++ b/pkg/tcpip/link/muxed/injectable.go
@@ -18,6 +18,7 @@ package muxed
import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -80,8 +81,8 @@ func (m *InjectableEndpoint) IsAttached() bool {
}
// InjectInbound implements stack.InjectableLinkEndpoint.
-func (m *InjectableEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
- m.dispatcher.DeliverNetworkPacket(m, "" /* remote */, "" /* local */, protocol, pkt)
+func (m *InjectableEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ m.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, protocol, pkt)
}
// WritePackets writes outbound packets to the appropriate
@@ -98,7 +99,7 @@ func (m *InjectableEndpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts s
// WritePacket writes outbound packets to the appropriate LinkInjectableEndpoint
// based on the RemoteAddress. HandleLocal only works if r.RemoteAddress has a
// route registered in this endpoint.
-func (m *InjectableEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
+func (m *InjectableEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
if endpoint, ok := m.routes[r.RemoteAddress]; ok {
return endpoint.WritePacket(r, gso, protocol, pkt)
}
@@ -129,6 +130,15 @@ func (m *InjectableEndpoint) Wait() {
}
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*InjectableEndpoint) ARPHardwareType() header.ARPHardwareType {
+ panic("unsupported operation")
+}
+
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (*InjectableEndpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+}
+
// NewInjectableEndpoint creates a new multi-endpoint injectable endpoint.
func NewInjectableEndpoint(routes map[tcpip.Address]stack.InjectableLinkEndpoint) *InjectableEndpoint {
return &InjectableEndpoint{
diff --git a/pkg/tcpip/link/muxed/injectable_test.go b/pkg/tcpip/link/muxed/injectable_test.go
index 87c734c1f..3e4afcdad 100644
--- a/pkg/tcpip/link/muxed/injectable_test.go
+++ b/pkg/tcpip/link/muxed/injectable_test.go
@@ -46,14 +46,14 @@ func TestInjectableEndpointRawDispatch(t *testing.T) {
func TestInjectableEndpointDispatch(t *testing.T) {
endpoint, sock, dstIP := makeTestInjectableEndpoint(t)
- hdr := buffer.NewPrependable(1)
- hdr.Prepend(1)[0] = 0xFA
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: 1,
+ Data: buffer.NewViewFromBytes([]byte{0xFB}).ToVectorisedView(),
+ })
+ pkt.TransportHeader().Push(1)[0] = 0xFA
packetRoute := stack.Route{RemoteAddress: dstIP}
- endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buffer.NewViewFromBytes([]byte{0xFB}).ToVectorisedView(),
- })
+ endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, pkt)
buf := make([]byte, 6500)
bytesRead, err := sock.Read(buf)
@@ -67,13 +67,14 @@ func TestInjectableEndpointDispatch(t *testing.T) {
func TestInjectableEndpointDispatchHdrOnly(t *testing.T) {
endpoint, sock, dstIP := makeTestInjectableEndpoint(t)
- hdr := buffer.NewPrependable(1)
- hdr.Prepend(1)[0] = 0xFA
- packetRoute := stack.Route{RemoteAddress: dstIP}
- endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buffer.NewView(0).ToVectorisedView(),
+
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: 1,
+ Data: buffer.NewView(0).ToVectorisedView(),
})
+ pkt.TransportHeader().Push(1)[0] = 0xFA
+ packetRoute := stack.Route{RemoteAddress: dstIP}
+ endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, pkt)
buf := make([]byte, 6500)
bytesRead, err := sock.Read(buf)
if err != nil {
diff --git a/pkg/tcpip/link/nested/BUILD b/pkg/tcpip/link/nested/BUILD
new file mode 100644
index 000000000..2cdb23475
--- /dev/null
+++ b/pkg/tcpip/link/nested/BUILD
@@ -0,0 +1,32 @@
+load("//tools:defs.bzl", "go_library", "go_test")
+
+package(licenses = ["notice"])
+
+go_library(
+ name = "nested",
+ srcs = [
+ "nested.go",
+ ],
+ visibility = ["//visibility:public"],
+ deps = [
+ "//pkg/sync",
+ "//pkg/tcpip",
+ "//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
+ "//pkg/tcpip/stack",
+ ],
+)
+
+go_test(
+ name = "nested_test",
+ size = "small",
+ srcs = [
+ "nested_test.go",
+ ],
+ deps = [
+ "//pkg/tcpip",
+ "//pkg/tcpip/header",
+ "//pkg/tcpip/link/nested",
+ "//pkg/tcpip/stack",
+ ],
+)
diff --git a/pkg/tcpip/link/nested/nested.go b/pkg/tcpip/link/nested/nested.go
new file mode 100644
index 000000000..d40de54df
--- /dev/null
+++ b/pkg/tcpip/link/nested/nested.go
@@ -0,0 +1,152 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package nested provides helpers to implement the pattern of nested
+// stack.LinkEndpoints.
+package nested
+
+import (
+ "gvisor.dev/gvisor/pkg/sync"
+ "gvisor.dev/gvisor/pkg/tcpip"
+ "gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
+ "gvisor.dev/gvisor/pkg/tcpip/stack"
+)
+
+// Endpoint is a wrapper around stack.LinkEndpoint and stack.NetworkDispatcher
+// that can be used to implement nesting safely by providing lifecycle
+// concurrency guards.
+//
+// See the tests in this package for example usage.
+type Endpoint struct {
+ child stack.LinkEndpoint
+ embedder stack.NetworkDispatcher
+
+ // mu protects dispatcher.
+ mu sync.RWMutex
+ dispatcher stack.NetworkDispatcher
+}
+
+var _ stack.GSOEndpoint = (*Endpoint)(nil)
+var _ stack.LinkEndpoint = (*Endpoint)(nil)
+var _ stack.NetworkDispatcher = (*Endpoint)(nil)
+
+// Init initializes a nested.Endpoint that uses embedder as the dispatcher for
+// child on Attach.
+//
+// See the tests in this package for example usage.
+func (e *Endpoint) Init(child stack.LinkEndpoint, embedder stack.NetworkDispatcher) {
+ e.child = child
+ e.embedder = embedder
+}
+
+// DeliverNetworkPacket implements stack.NetworkDispatcher.
+func (e *Endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.mu.RLock()
+ d := e.dispatcher
+ e.mu.RUnlock()
+ if d != nil {
+ d.DeliverNetworkPacket(remote, local, protocol, pkt)
+ }
+}
+
+// DeliverOutboundPacket implements stack.NetworkDispatcher.DeliverOutboundPacket.
+func (e *Endpoint) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.mu.RLock()
+ d := e.dispatcher
+ e.mu.RUnlock()
+ if d != nil {
+ d.DeliverOutboundPacket(remote, local, protocol, pkt)
+ }
+}
+
+// Attach implements stack.LinkEndpoint.
+func (e *Endpoint) Attach(dispatcher stack.NetworkDispatcher) {
+ e.mu.Lock()
+ e.dispatcher = dispatcher
+ e.mu.Unlock()
+ // If we're attaching to a valid dispatcher, pass embedder as the dispatcher
+ // to our child, otherwise detach the child by giving it a nil dispatcher.
+ var pass stack.NetworkDispatcher
+ if dispatcher != nil {
+ pass = e.embedder
+ }
+ e.child.Attach(pass)
+}
+
+// IsAttached implements stack.LinkEndpoint.
+func (e *Endpoint) IsAttached() bool {
+ e.mu.RLock()
+ isAttached := e.dispatcher != nil
+ e.mu.RUnlock()
+ return isAttached
+}
+
+// MTU implements stack.LinkEndpoint.
+func (e *Endpoint) MTU() uint32 {
+ return e.child.MTU()
+}
+
+// Capabilities implements stack.LinkEndpoint.
+func (e *Endpoint) Capabilities() stack.LinkEndpointCapabilities {
+ return e.child.Capabilities()
+}
+
+// MaxHeaderLength implements stack.LinkEndpoint.
+func (e *Endpoint) MaxHeaderLength() uint16 {
+ return e.child.MaxHeaderLength()
+}
+
+// LinkAddress implements stack.LinkEndpoint.
+func (e *Endpoint) LinkAddress() tcpip.LinkAddress {
+ return e.child.LinkAddress()
+}
+
+// WritePacket implements stack.LinkEndpoint.
+func (e *Endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
+ return e.child.WritePacket(r, gso, protocol, pkt)
+}
+
+// WritePackets implements stack.LinkEndpoint.
+func (e *Endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ return e.child.WritePackets(r, gso, pkts, protocol)
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.
+func (e *Endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
+ return e.child.WriteRawPacket(vv)
+}
+
+// Wait implements stack.LinkEndpoint.
+func (e *Endpoint) Wait() {
+ e.child.Wait()
+}
+
+// GSOMaxSize implements stack.GSOEndpoint.
+func (e *Endpoint) GSOMaxSize() uint32 {
+ if e, ok := e.child.(stack.GSOEndpoint); ok {
+ return e.GSOMaxSize()
+ }
+ return 0
+}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
+func (e *Endpoint) ARPHardwareType() header.ARPHardwareType {
+ return e.child.ARPHardwareType()
+}
+
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *Endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.child.AddHeader(local, remote, protocol, pkt)
+}
diff --git a/pkg/tcpip/link/nested/nested_test.go b/pkg/tcpip/link/nested/nested_test.go
new file mode 100644
index 000000000..c1f9d308c
--- /dev/null
+++ b/pkg/tcpip/link/nested/nested_test.go
@@ -0,0 +1,109 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package nested_test
+
+import (
+ "testing"
+
+ "gvisor.dev/gvisor/pkg/tcpip"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
+ "gvisor.dev/gvisor/pkg/tcpip/link/nested"
+ "gvisor.dev/gvisor/pkg/tcpip/stack"
+)
+
+type parentEndpoint struct {
+ nested.Endpoint
+}
+
+var _ stack.LinkEndpoint = (*parentEndpoint)(nil)
+var _ stack.NetworkDispatcher = (*parentEndpoint)(nil)
+
+type childEndpoint struct {
+ stack.LinkEndpoint
+ dispatcher stack.NetworkDispatcher
+}
+
+var _ stack.LinkEndpoint = (*childEndpoint)(nil)
+
+func (c *childEndpoint) Attach(dispatcher stack.NetworkDispatcher) {
+ c.dispatcher = dispatcher
+}
+
+func (c *childEndpoint) IsAttached() bool {
+ return c.dispatcher != nil
+}
+
+type counterDispatcher struct {
+ count int
+}
+
+var _ stack.NetworkDispatcher = (*counterDispatcher)(nil)
+
+func (d *counterDispatcher) DeliverNetworkPacket(tcpip.LinkAddress, tcpip.LinkAddress, tcpip.NetworkProtocolNumber, *stack.PacketBuffer) {
+ d.count++
+}
+
+func (d *counterDispatcher) DeliverOutboundPacket(tcpip.LinkAddress, tcpip.LinkAddress, tcpip.NetworkProtocolNumber, *stack.PacketBuffer) {
+ panic("unimplemented")
+}
+
+func TestNestedLinkEndpoint(t *testing.T) {
+ const emptyAddress = tcpip.LinkAddress("")
+
+ var (
+ childEP childEndpoint
+ nestedEP parentEndpoint
+ disp counterDispatcher
+ )
+ nestedEP.Endpoint.Init(&childEP, &nestedEP)
+
+ if childEP.IsAttached() {
+ t.Error("On init, childEP.IsAttached() = true, want = false")
+ }
+ if nestedEP.IsAttached() {
+ t.Error("On init, nestedEP.IsAttached() = true, want = false")
+ }
+
+ nestedEP.Attach(&disp)
+ if disp.count != 0 {
+ t.Fatalf("After attach, got disp.count = %d, want = 0", disp.count)
+ }
+ if !childEP.IsAttached() {
+ t.Error("After attach, childEP.IsAttached() = false, want = true")
+ }
+ if !nestedEP.IsAttached() {
+ t.Error("After attach, nestedEP.IsAttached() = false, want = true")
+ }
+
+ nestedEP.DeliverNetworkPacket(emptyAddress, emptyAddress, header.IPv4ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
+ if disp.count != 1 {
+ t.Errorf("After first packet with dispatcher attached, got disp.count = %d, want = 1", disp.count)
+ }
+
+ nestedEP.Attach(nil)
+ if childEP.IsAttached() {
+ t.Error("After detach, childEP.IsAttached() = true, want = false")
+ }
+ if nestedEP.IsAttached() {
+ t.Error("After detach, nestedEP.IsAttached() = true, want = false")
+ }
+
+ disp.count = 0
+ nestedEP.DeliverNetworkPacket(emptyAddress, emptyAddress, header.IPv4ProtocolNumber, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
+ if disp.count != 0 {
+ t.Errorf("After second packet with dispatcher detached, got disp.count = %d, want = 0", disp.count)
+ }
+
+}
diff --git a/pkg/tcpip/link/packetsocket/BUILD b/pkg/tcpip/link/packetsocket/BUILD
new file mode 100644
index 000000000..6fff160ce
--- /dev/null
+++ b/pkg/tcpip/link/packetsocket/BUILD
@@ -0,0 +1,14 @@
+load("//tools:defs.bzl", "go_library")
+
+package(licenses = ["notice"])
+
+go_library(
+ name = "packetsocket",
+ srcs = ["endpoint.go"],
+ visibility = ["//visibility:public"],
+ deps = [
+ "//pkg/tcpip",
+ "//pkg/tcpip/link/nested",
+ "//pkg/tcpip/stack",
+ ],
+)
diff --git a/pkg/tcpip/link/packetsocket/endpoint.go b/pkg/tcpip/link/packetsocket/endpoint.go
new file mode 100644
index 000000000..3922c2a04
--- /dev/null
+++ b/pkg/tcpip/link/packetsocket/endpoint.go
@@ -0,0 +1,50 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package packetsocket provides a link layer endpoint that provides the ability
+// to loop outbound packets to any AF_PACKET sockets that may be interested in
+// the outgoing packet.
+package packetsocket
+
+import (
+ "gvisor.dev/gvisor/pkg/tcpip"
+ "gvisor.dev/gvisor/pkg/tcpip/link/nested"
+ "gvisor.dev/gvisor/pkg/tcpip/stack"
+)
+
+type endpoint struct {
+ nested.Endpoint
+}
+
+// New creates a new packetsocket LinkEndpoint.
+func New(lower stack.LinkEndpoint) stack.LinkEndpoint {
+ e := &endpoint{}
+ e.Endpoint.Init(lower, e)
+ return e
+}
+
+// WritePacket implements stack.LinkEndpoint.WritePacket.
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
+ e.Endpoint.DeliverOutboundPacket(r.RemoteLinkAddress, r.LocalLinkAddress, protocol, pkt)
+ return e.Endpoint.WritePacket(r, gso, protocol, pkt)
+}
+
+// WritePackets implements stack.LinkEndpoint.WritePackets.
+func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.PacketBufferList, proto tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ for pkt := pkts.Front(); pkt != nil; pkt = pkt.Next() {
+ e.Endpoint.DeliverOutboundPacket(pkt.EgressRoute.RemoteLinkAddress, pkt.EgressRoute.LocalLinkAddress, pkt.NetworkProtocolNumber, pkt)
+ }
+
+ return e.Endpoint.WritePackets(r, gso, pkts, proto)
+}
diff --git a/pkg/tcpip/link/qdisc/fifo/BUILD b/pkg/tcpip/link/qdisc/fifo/BUILD
index 054c213bc..1d0079bd6 100644
--- a/pkg/tcpip/link/qdisc/fifo/BUILD
+++ b/pkg/tcpip/link/qdisc/fifo/BUILD
@@ -14,6 +14,7 @@ go_library(
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/qdisc/fifo/endpoint.go b/pkg/tcpip/link/qdisc/fifo/endpoint.go
index 54432194d..fc1e34fc7 100644
--- a/pkg/tcpip/link/qdisc/fifo/endpoint.go
+++ b/pkg/tcpip/link/qdisc/fifo/endpoint.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -102,8 +103,13 @@ func (q *queueDispatcher) dispatchLoop() {
}
// DeliverNetworkPacket implements stack.NetworkDispatcher.DeliverNetworkPacket.
-func (e *endpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
- e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, pkt)
+func (e *endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.dispatcher.DeliverNetworkPacket(remote, local, protocol, pkt)
+}
+
+// DeliverOutboundPacket implements stack.NetworkDispatcher.DeliverOutboundPacket.
+func (e *endpoint) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.dispatcher.DeliverOutboundPacket(remote, local, protocol, pkt)
}
// Attach implements stack.LinkEndpoint.Attach.
@@ -146,7 +152,7 @@ func (e *endpoint) GSOMaxSize() uint32 {
}
// WritePacket implements stack.LinkEndpoint.WritePacket.
-func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
// WritePacket caller's do not set the following fields in PacketBuffer
// so we populate them here.
newRoute := r.Clone()
@@ -154,7 +160,7 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.Ne
pkt.GSOOptions = gso
pkt.NetworkProtocolNumber = protocol
d := e.dispatchers[int(pkt.Hash)%len(e.dispatchers)]
- if !d.q.enqueue(&pkt) {
+ if !d.q.enqueue(pkt) {
return tcpip.ErrNoBufferSpace
}
d.newPacketWaker.Assert()
@@ -193,6 +199,8 @@ func (e *endpoint) WritePackets(_ *stack.Route, _ *stack.GSO, pkts stack.PacketB
// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
+ // TODO(gvisor.dev/issue/3267): Queue these packets as well once
+ // WriteRawPacket takes PacketBuffer instead of VectorisedView.
return e.lower.WriteRawPacket(vv)
}
@@ -207,3 +215,13 @@ func (e *endpoint) Wait() {
e.wg.Wait()
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
+func (e *endpoint) ARPHardwareType() header.ARPHardwareType {
+ return e.lower.ARPHardwareType()
+}
+
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.lower.AddHeader(local, remote, protocol, pkt)
+}
diff --git a/pkg/tcpip/link/rawfile/blockingpoll_yield_unsafe.go b/pkg/tcpip/link/rawfile/blockingpoll_yield_unsafe.go
index 0b5a6cf49..99313ee25 100644
--- a/pkg/tcpip/link/rawfile/blockingpoll_yield_unsafe.go
+++ b/pkg/tcpip/link/rawfile/blockingpoll_yield_unsafe.go
@@ -14,7 +14,7 @@
// +build linux,amd64 linux,arm64
// +build go1.12
-// +build !go1.15
+// +build !go1.16
// Check go:linkname function signatures when updating Go version.
diff --git a/pkg/tcpip/link/rawfile/rawfile_unsafe.go b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
index 44e25d475..f4c32c2da 100644
--- a/pkg/tcpip/link/rawfile/rawfile_unsafe.go
+++ b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
@@ -66,39 +66,14 @@ func NonBlockingWrite(fd int, buf []byte) *tcpip.Error {
return nil
}
-// NonBlockingWrite3 writes up to three byte slices to a file descriptor in a
-// single syscall. It fails if partial data is written.
-func NonBlockingWrite3(fd int, b1, b2, b3 []byte) *tcpip.Error {
- // If the is no second buffer, issue a regular write.
- if len(b2) == 0 {
- return NonBlockingWrite(fd, b1)
- }
-
- // We have two buffers. Build the iovec that represents them and issue
- // a writev syscall.
- iovec := [3]syscall.Iovec{
- {
- Base: &b1[0],
- Len: uint64(len(b1)),
- },
- {
- Base: &b2[0],
- Len: uint64(len(b2)),
- },
- }
- iovecLen := uintptr(2)
-
- if len(b3) > 0 {
- iovecLen++
- iovec[2].Base = &b3[0]
- iovec[2].Len = uint64(len(b3))
- }
-
+// NonBlockingWriteIovec writes iovec to a file descriptor in a single syscall.
+// It fails if partial data is written.
+func NonBlockingWriteIovec(fd int, iovec []syscall.Iovec) *tcpip.Error {
+ iovecLen := uintptr(len(iovec))
_, _, e := syscall.RawSyscall(syscall.SYS_WRITEV, uintptr(fd), uintptr(unsafe.Pointer(&iovec[0])), iovecLen)
if e != 0 {
return TranslateErrno(e)
}
-
return nil
}
diff --git a/pkg/tcpip/link/sharedmem/sharedmem.go b/pkg/tcpip/link/sharedmem/sharedmem.go
index 0796d717e..7fb8a6c49 100644
--- a/pkg/tcpip/link/sharedmem/sharedmem.go
+++ b/pkg/tcpip/link/sharedmem/sharedmem.go
@@ -183,27 +183,33 @@ func (e *endpoint) LinkAddress() tcpip.LinkAddress {
return e.addr
}
-// WritePacket writes outbound packets to the file descriptor. If it is not
-// currently writable, the packet is dropped.
-func (e *endpoint) WritePacket(r *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
- // Add the ethernet header here.
- eth := header.Ethernet(pkt.Header.Prepend(header.EthernetMinimumSize))
- pkt.LinkHeader = buffer.View(eth)
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ // Add ethernet header if needed.
+ eth := header.Ethernet(pkt.LinkHeader().Push(header.EthernetMinimumSize))
ethHdr := &header.EthernetFields{
- DstAddr: r.RemoteLinkAddress,
+ DstAddr: remote,
Type: protocol,
}
- if r.LocalLinkAddress != "" {
- ethHdr.SrcAddr = r.LocalLinkAddress
+
+ // Preserve the src address if it's set in the route.
+ if local != "" {
+ ethHdr.SrcAddr = local
} else {
ethHdr.SrcAddr = e.addr
}
eth.Encode(ethHdr)
+}
+
+// WritePacket writes outbound packets to the file descriptor. If it is not
+// currently writable, the packet is dropped.
+func (e *endpoint) WritePacket(r *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
+ e.AddHeader(r.LocalLinkAddress, r.RemoteLinkAddress, protocol, pkt)
- v := pkt.Data.ToView()
+ views := pkt.Views()
// Transmit the packet.
e.mu.Lock()
- ok := e.tx.transmit(pkt.Header.View(), v)
+ ok := e.tx.transmit(views...)
e.mu.Unlock()
if !ok {
@@ -220,10 +226,10 @@ func (e *endpoint) WritePackets(r *stack.Route, _ *stack.GSO, pkts stack.PacketB
// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
- v := vv.ToView()
+ views := vv.Views()
// Transmit the packet.
e.mu.Lock()
- ok := e.tx.transmit(v, buffer.View{})
+ ok := e.tx.transmit(views...)
e.mu.Unlock()
if !ok {
@@ -269,16 +275,18 @@ func (e *endpoint) dispatchLoop(d stack.NetworkDispatcher) {
rxb[i].Size = e.bufferSize
}
- if n < header.EthernetMinimumSize {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ Data: buffer.View(b).ToVectorisedView(),
+ })
+
+ hdr, ok := pkt.LinkHeader().Consume(header.EthernetMinimumSize)
+ if !ok {
continue
}
+ eth := header.Ethernet(hdr)
// Send packet up the stack.
- eth := header.Ethernet(b[:header.EthernetMinimumSize])
- d.DeliverNetworkPacket(e, eth.SourceAddress(), eth.DestinationAddress(), eth.Type(), stack.PacketBuffer{
- Data: buffer.View(b[header.EthernetMinimumSize:]).ToVectorisedView(),
- LinkHeader: buffer.View(eth),
- })
+ d.DeliverNetworkPacket(eth.SourceAddress(), eth.DestinationAddress(), eth.Type(), pkt)
}
// Clean state.
@@ -287,3 +295,8 @@ func (e *endpoint) dispatchLoop(d stack.NetworkDispatcher) {
e.completed.Done()
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
+func (*endpoint) ARPHardwareType() header.ARPHardwareType {
+ return header.ARPHardwareEther
+}
diff --git a/pkg/tcpip/link/sharedmem/sharedmem_test.go b/pkg/tcpip/link/sharedmem/sharedmem_test.go
index 33f640b85..22d5c97f1 100644
--- a/pkg/tcpip/link/sharedmem/sharedmem_test.go
+++ b/pkg/tcpip/link/sharedmem/sharedmem_test.go
@@ -131,7 +131,7 @@ func newTestContext(t *testing.T, mtu, bufferSize uint32, addr tcpip.LinkAddress
return c
}
-func (c *testContext) DeliverNetworkPacket(_ stack.LinkEndpoint, remoteLinkAddr, localLinkAddr tcpip.LinkAddress, proto tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
+func (c *testContext) DeliverNetworkPacket(remoteLinkAddr, localLinkAddr tcpip.LinkAddress, proto tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
c.mu.Lock()
c.packets = append(c.packets, packetInfo{
addr: remoteLinkAddr,
@@ -143,6 +143,10 @@ func (c *testContext) DeliverNetworkPacket(_ stack.LinkEndpoint, remoteLinkAddr,
c.packetCh <- struct{}{}
}
+func (c *testContext) DeliverOutboundPacket(remoteLinkAddr, localLinkAddr tcpip.LinkAddress, proto tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ panic("unimplemented")
+}
+
func (c *testContext) cleanup() {
c.ep.Close()
closeFDs(&c.txCfg)
@@ -262,21 +266,23 @@ func TestSimpleSend(t *testing.T) {
for iters := 1000; iters > 0; iters-- {
func() {
+ hdrLen, dataLen := rand.Intn(10000), rand.Intn(10000)
+
// Prepare and send packet.
- n := rand.Intn(10000)
- hdr := buffer.NewPrependable(n + int(c.ep.MaxHeaderLength()))
- hdrBuf := hdr.Prepend(n)
+ hdrBuf := buffer.NewView(hdrLen)
randomFill(hdrBuf)
- n = rand.Intn(10000)
- buf := buffer.NewView(n)
- randomFill(buf)
+ data := buffer.NewView(dataLen)
+ randomFill(data)
+
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: hdrLen + int(c.ep.MaxHeaderLength()),
+ Data: data.ToVectorisedView(),
+ })
+ copy(pkt.NetworkHeader().Push(hdrLen), hdrBuf)
proto := tcpip.NetworkProtocolNumber(rand.Intn(0x10000))
- if err := c.ep.WritePacket(&r, nil /* gso */, proto, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ if err := c.ep.WritePacket(&r, nil /* gso */, proto, pkt); err != nil {
t.Fatalf("WritePacket failed: %v", err)
}
@@ -313,7 +319,7 @@ func TestSimpleSend(t *testing.T) {
// Compare contents skipping the ethernet header added by the
// endpoint.
- merged := append(hdrBuf, buf...)
+ merged := append(hdrBuf, data...)
if uint32(len(contents)) < pi.Size {
t.Fatalf("Sum of buffers is less than packet size: %v < %v", len(contents), pi.Size)
}
@@ -340,14 +346,14 @@ func TestPreserveSrcAddressInSend(t *testing.T) {
LocalLinkAddress: newLocalLinkAddress,
}
- // WritePacket panics given a prependable with anything less than
- // the minimum size of the ethernet header.
- hdr := buffer.NewPrependable(header.EthernetMinimumSize)
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ // WritePacket panics given a prependable with anything less than
+ // the minimum size of the ethernet header.
+ ReserveHeaderBytes: header.EthernetMinimumSize,
+ })
proto := tcpip.NetworkProtocolNumber(rand.Intn(0x10000))
- if err := c.ep.WritePacket(&r, nil /* gso */, proto, stack.PacketBuffer{
- Header: hdr,
- }); err != nil {
+ if err := c.ep.WritePacket(&r, nil /* gso */, proto, pkt); err != nil {
t.Fatalf("WritePacket failed: %v", err)
}
@@ -399,12 +405,12 @@ func TestFillTxQueue(t *testing.T) {
// until the tx queue if full.
ids := make(map[uint64]struct{})
for i := queuePipeSize / 40; i > 0; i-- {
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
- if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != nil {
t.Fatalf("WritePacket failed unexpectedly: %v", err)
}
@@ -418,11 +424,11 @@ func TestFillTxQueue(t *testing.T) {
}
// Next attempt to write must fail.
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != want {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != want {
t.Fatalf("WritePacket return unexpected result: got %v, want %v", err, want)
}
}
@@ -446,11 +452,11 @@ func TestFillTxQueueAfterBadCompletion(t *testing.T) {
// Send two packets so that the id slice has at least two slots.
for i := 2; i > 0; i-- {
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != nil {
t.Fatalf("WritePacket failed unexpectedly: %v", err)
}
}
@@ -469,11 +475,11 @@ func TestFillTxQueueAfterBadCompletion(t *testing.T) {
// until the tx queue if full.
ids := make(map[uint64]struct{})
for i := queuePipeSize / 40; i > 0; i-- {
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != nil {
t.Fatalf("WritePacket failed unexpectedly: %v", err)
}
@@ -487,11 +493,11 @@ func TestFillTxQueueAfterBadCompletion(t *testing.T) {
}
// Next attempt to write must fail.
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != want {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != want {
t.Fatalf("WritePacket return unexpected result: got %v, want %v", err, want)
}
}
@@ -513,11 +519,11 @@ func TestFillTxMemory(t *testing.T) {
// we fill the memory.
ids := make(map[uint64]struct{})
for i := queueDataSize / bufferSize; i > 0; i-- {
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != nil {
t.Fatalf("WritePacket failed unexpectedly: %v", err)
}
@@ -532,11 +538,11 @@ func TestFillTxMemory(t *testing.T) {
}
// Next attempt to write must fail.
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
})
+ err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt)
if want := tcpip.ErrWouldBlock; err != want {
t.Fatalf("WritePacket return unexpected result: got %v, want %v", err, want)
}
@@ -560,11 +566,11 @@ func TestFillTxMemoryWithMultiBuffer(t *testing.T) {
// Each packet is uses up one buffer, so write as many as possible
// until there is only one buffer left.
for i := queueDataSize/bufferSize - 1; i > 0; i-- {
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != nil {
t.Fatalf("WritePacket failed unexpectedly: %v", err)
}
@@ -575,23 +581,22 @@ func TestFillTxMemoryWithMultiBuffer(t *testing.T) {
// Attempt to write a two-buffer packet. It must fail.
{
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- uu := buffer.NewView(bufferSize).ToVectorisedView()
- if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: uu,
- }); err != want {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buffer.NewView(bufferSize).ToVectorisedView(),
+ })
+ if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != want {
t.Fatalf("WritePacket return unexpected result: got %v, want %v", err, want)
}
}
// Attempt to write the one-buffer packet again. It must succeed.
{
- hdr := buffer.NewPrependable(int(c.ep.MaxHeaderLength()))
- if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, stack.PacketBuffer{
- Header: hdr,
- Data: buf.ToVectorisedView(),
- }); err != nil {
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: int(c.ep.MaxHeaderLength()),
+ Data: buf.ToVectorisedView(),
+ })
+ if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, pkt); err != nil {
t.Fatalf("WritePacket failed unexpectedly: %v", err)
}
}
diff --git a/pkg/tcpip/link/sharedmem/tx.go b/pkg/tcpip/link/sharedmem/tx.go
index 6b8d7859d..44f421c2d 100644
--- a/pkg/tcpip/link/sharedmem/tx.go
+++ b/pkg/tcpip/link/sharedmem/tx.go
@@ -18,6 +18,7 @@ import (
"math"
"syscall"
+ "gvisor.dev/gvisor/pkg/tcpip/buffer"
"gvisor.dev/gvisor/pkg/tcpip/link/sharedmem/queue"
)
@@ -76,9 +77,9 @@ func (t *tx) cleanup() {
syscall.Munmap(t.data)
}
-// transmit sends a packet made up of up to two buffers. Returns a boolean that
-// specifies whether the packet was successfully transmitted.
-func (t *tx) transmit(a, b []byte) bool {
+// transmit sends a packet made of bufs. Returns a boolean that specifies
+// whether the packet was successfully transmitted.
+func (t *tx) transmit(bufs ...buffer.View) bool {
// Pull completions from the tx queue and add their buffers back to the
// pool so that we can reuse them.
for {
@@ -93,7 +94,10 @@ func (t *tx) transmit(a, b []byte) bool {
}
bSize := t.bufs.entrySize
- total := uint32(len(a) + len(b))
+ total := uint32(0)
+ for _, data := range bufs {
+ total += uint32(len(data))
+ }
bufCount := (total + bSize - 1) / bSize
// Allocate enough buffers to hold all the data.
@@ -115,7 +119,7 @@ func (t *tx) transmit(a, b []byte) bool {
// Copy data into allocated buffers.
nBuf := buf
var dBuf []byte
- for _, data := range [][]byte{a, b} {
+ for _, data := range bufs {
for len(data) > 0 {
if len(dBuf) == 0 {
dBuf = t.data[nBuf.Offset:][:nBuf.Size]
diff --git a/pkg/tcpip/link/sniffer/BUILD b/pkg/tcpip/link/sniffer/BUILD
index 230a8d53a..7cbc305e7 100644
--- a/pkg/tcpip/link/sniffer/BUILD
+++ b/pkg/tcpip/link/sniffer/BUILD
@@ -14,6 +14,7 @@ go_library(
"//pkg/tcpip",
"//pkg/tcpip/buffer",
"//pkg/tcpip/header",
+ "//pkg/tcpip/link/nested",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go
index da1c520ae..4fb127978 100644
--- a/pkg/tcpip/link/sniffer/sniffer.go
+++ b/pkg/tcpip/link/sniffer/sniffer.go
@@ -31,6 +31,7 @@ import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
"gvisor.dev/gvisor/pkg/tcpip/header"
+ "gvisor.dev/gvisor/pkg/tcpip/link/nested"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -47,26 +48,22 @@ var LogPackets uint32 = 1
// LogPacketsToPCAP must be accessed atomically.
var LogPacketsToPCAP uint32 = 1
-var transportProtocolMinSizes map[tcpip.TransportProtocolNumber]int = map[tcpip.TransportProtocolNumber]int{
- header.ICMPv4ProtocolNumber: header.IPv4MinimumSize,
- header.ICMPv6ProtocolNumber: header.IPv6MinimumSize,
- header.UDPProtocolNumber: header.UDPMinimumSize,
- header.TCPProtocolNumber: header.TCPMinimumSize,
-}
-
type endpoint struct {
- dispatcher stack.NetworkDispatcher
- lower stack.LinkEndpoint
+ nested.Endpoint
writer io.Writer
maxPCAPLen uint32
}
+var _ stack.GSOEndpoint = (*endpoint)(nil)
+var _ stack.LinkEndpoint = (*endpoint)(nil)
+var _ stack.NetworkDispatcher = (*endpoint)(nil)
+
// New creates a new sniffer link-layer endpoint. It wraps around another
// endpoint and logs packets and they traverse the endpoint.
func New(lower stack.LinkEndpoint) stack.LinkEndpoint {
- return &endpoint{
- lower: lower,
- }
+ sniffer := &endpoint{}
+ sniffer.Endpoint.Init(lower, sniffer)
+ return sniffer
}
func zoneOffset() (int32, error) {
@@ -110,62 +107,25 @@ func NewWithWriter(lower stack.LinkEndpoint, writer io.Writer, snapLen uint32) (
if err := writePCAPHeader(writer, snapLen); err != nil {
return nil, err
}
- return &endpoint{
- lower: lower,
+ sniffer := &endpoint{
writer: writer,
maxPCAPLen: snapLen,
- }, nil
+ }
+ sniffer.Endpoint.Init(lower, sniffer)
+ return sniffer, nil
}
// DeliverNetworkPacket implements the stack.NetworkDispatcher interface. It is
// called by the link-layer endpoint being wrapped when a packet arrives, and
// logs the packet before forwarding to the actual dispatcher.
-func (e *endpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
- e.dumpPacket("recv", nil, protocol, &pkt)
- e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, pkt)
-}
-
-// Attach implements the stack.LinkEndpoint interface. It saves the dispatcher
-// and registers with the lower endpoint as its dispatcher so that "e" is called
-// for inbound packets.
-func (e *endpoint) Attach(dispatcher stack.NetworkDispatcher) {
- e.dispatcher = dispatcher
- e.lower.Attach(e)
-}
-
-// IsAttached implements stack.LinkEndpoint.IsAttached.
-func (e *endpoint) IsAttached() bool {
- return e.dispatcher != nil
+func (e *endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.dumpPacket("recv", nil, protocol, pkt)
+ e.Endpoint.DeliverNetworkPacket(remote, local, protocol, pkt)
}
-// MTU implements stack.LinkEndpoint.MTU. It just forwards the request to the
-// lower endpoint.
-func (e *endpoint) MTU() uint32 {
- return e.lower.MTU()
-}
-
-// Capabilities implements stack.LinkEndpoint.Capabilities. It just forwards the
-// request to the lower endpoint.
-func (e *endpoint) Capabilities() stack.LinkEndpointCapabilities {
- return e.lower.Capabilities()
-}
-
-// MaxHeaderLength implements the stack.LinkEndpoint interface. It just forwards
-// the request to the lower endpoint.
-func (e *endpoint) MaxHeaderLength() uint16 {
- return e.lower.MaxHeaderLength()
-}
-
-func (e *endpoint) LinkAddress() tcpip.LinkAddress {
- return e.lower.LinkAddress()
-}
-
-// GSOMaxSize returns the maximum GSO packet size.
-func (e *endpoint) GSOMaxSize() uint32 {
- if gso, ok := e.lower.(stack.GSOEndpoint); ok {
- return gso.GSOMaxSize()
- }
- return 0
+// DeliverOutboundPacket implements stack.NetworkDispatcher.DeliverOutboundPacket.
+func (e *endpoint) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.Endpoint.DeliverOutboundPacket(remote, local, protocol, pkt)
}
func (e *endpoint) dumpPacket(prefix string, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
@@ -174,7 +134,7 @@ func (e *endpoint) dumpPacket(prefix string, gso *stack.GSO, protocol tcpip.Netw
logPacket(prefix, protocol, pkt, gso)
}
if writer != nil && atomic.LoadUint32(&LogPacketsToPCAP) == 1 {
- totalLength := pkt.Header.UsedLength() + pkt.Data.Size()
+ totalLength := pkt.Size()
length := totalLength
if max := int(e.maxPCAPLen); length > max {
length = max
@@ -195,12 +155,11 @@ func (e *endpoint) dumpPacket(prefix string, gso *stack.GSO, protocol tcpip.Netw
length -= n
}
}
- write(pkt.Header.View())
- for _, view := range pkt.Data.Views() {
+ for _, v := range pkt.Views() {
if length == 0 {
break
}
- write(view)
+ write(v)
}
}
}
@@ -208,9 +167,9 @@ func (e *endpoint) dumpPacket(prefix string, gso *stack.GSO, protocol tcpip.Netw
// WritePacket implements the stack.LinkEndpoint interface. It is called by
// higher-level protocols to write packets; it just logs the packet and
// forwards the request to the lower endpoint.
-func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
- e.dumpPacket("send", gso, protocol, &pkt)
- return e.lower.WritePacket(r, gso, protocol, pkt)
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
+ e.dumpPacket("send", gso, protocol, pkt)
+ return e.Endpoint.WritePacket(r, gso, protocol, pkt)
}
// WritePackets implements the stack.LinkEndpoint interface. It is called by
@@ -220,20 +179,17 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe
for pkt := pkts.Front(); pkt != nil; pkt = pkt.Next() {
e.dumpPacket("send", gso, protocol, pkt)
}
- return e.lower.WritePackets(r, gso, pkts, protocol)
+ return e.Endpoint.WritePackets(r, gso, pkts, protocol)
}
// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
- e.dumpPacket("send", nil, 0, &stack.PacketBuffer{
+ e.dumpPacket("send", nil, 0, stack.NewPacketBuffer(stack.PacketBufferOptions{
Data: vv,
- })
- return e.lower.WriteRawPacket(vv)
+ }))
+ return e.Endpoint.WriteRawPacket(vv)
}
-// Wait implements stack.LinkEndpoint.Wait.
-func (e *endpoint) Wait() { e.lower.Wait() }
-
func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer, gso *stack.GSO) {
// Figure out the network layer info.
var transProto uint8
@@ -244,12 +200,8 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
var fragmentOffset uint16
var moreFragments bool
- // Create a clone of pkt, including any headers if present. Avoid allocating
- // backing memory for the clone.
- views := [8]buffer.View{}
- vv := buffer.NewVectorisedView(0, views[:0])
- vv.AppendView(pkt.Header.View())
- vv.Append(pkt.Data)
+ // Examine the packet using a new VV. Backing storage must not be written.
+ vv := buffer.NewVectorisedView(pkt.Size(), pkt.Views())
switch protocol {
case header.IPv4ProtocolNumber:
@@ -287,7 +239,7 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
vv.TrimFront(header.ARPSize)
arp := header.ARP(hdr)
log.Infof(
- "%s arp %v (%v) -> %v (%v) valid:%v",
+ "%s arp %s (%s) -> %s (%s) valid:%t",
prefix,
tcpip.Address(arp.ProtocolAddressSender()), tcpip.LinkAddress(arp.HardwareAddressSender()),
tcpip.Address(arp.ProtocolAddressTarget()), tcpip.LinkAddress(arp.HardwareAddressTarget()),
@@ -299,13 +251,6 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
return
}
- // We aren't guaranteed to have a transport header - it's possible for
- // writes via raw endpoints to contain only network headers.
- if minSize, ok := transportProtocolMinSizes[tcpip.TransportProtocolNumber(transProto)]; ok && vv.Size() < minSize {
- log.Infof("%s %v -> %v transport protocol: %d, but no transport header found (possible raw packet)", prefix, src, dst, transProto)
- return
- }
-
// Figure out the transport layer info.
transName := "unknown"
srcPort := uint16(0)
@@ -346,7 +291,7 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
icmpType = "info reply"
}
}
- log.Infof("%s %s %v -> %v %s len:%d id:%04x code:%d", prefix, transName, src, dst, icmpType, size, id, icmp.Code())
+ log.Infof("%s %s %s -> %s %s len:%d id:%04x code:%d", prefix, transName, src, dst, icmpType, size, id, icmp.Code())
return
case header.ICMPv6ProtocolNumber:
@@ -381,7 +326,7 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
case header.ICMPv6RedirectMsg:
icmpType = "redirect message"
}
- log.Infof("%s %s %v -> %v %s len:%d id:%04x code:%d", prefix, transName, src, dst, icmpType, size, id, icmp.Code())
+ log.Infof("%s %s %s -> %s %s len:%d id:%04x code:%d", prefix, transName, src, dst, icmpType, size, id, icmp.Code())
return
case header.UDPProtocolNumber:
@@ -428,7 +373,7 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
flagsStr[i] = ' '
}
}
- details = fmt.Sprintf("flags:0x%02x (%v) seqnum: %v ack: %v win: %v xsum:0x%x", flags, string(flagsStr), tcp.SequenceNumber(), tcp.AckNumber(), tcp.WindowSize(), tcp.Checksum())
+ details = fmt.Sprintf("flags:0x%02x (%s) seqnum: %d ack: %d win: %d xsum:0x%x", flags, string(flagsStr), tcp.SequenceNumber(), tcp.AckNumber(), tcp.WindowSize(), tcp.Checksum())
if flags&header.TCPFlagSyn != 0 {
details += fmt.Sprintf(" options: %+v", header.ParseSynOptions(tcp.Options(), flags&header.TCPFlagAck != 0))
} else {
@@ -437,7 +382,7 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
}
default:
- log.Infof("%s %v -> %v unknown transport protocol: %d", prefix, src, dst, transProto)
+ log.Infof("%s %s -> %s unknown transport protocol: %d", prefix, src, dst, transProto)
return
}
@@ -445,5 +390,5 @@ func logPacket(prefix string, protocol tcpip.NetworkProtocolNumber, pkt *stack.P
details += fmt.Sprintf(" gso: %+v", gso)
}
- log.Infof("%s %s %v:%v -> %v:%v len:%d id:%04x %s", prefix, transName, src, srcPort, dst, dstPort, size, id, details)
+ log.Infof("%s %s %s:%d -> %s:%d len:%d id:%04x %s", prefix, transName, src, srcPort, dst, dstPort, size, id, details)
}
diff --git a/pkg/tcpip/link/tun/BUILD b/pkg/tcpip/link/tun/BUILD
index e0db6cf54..6c137f693 100644
--- a/pkg/tcpip/link/tun/BUILD
+++ b/pkg/tcpip/link/tun/BUILD
@@ -12,6 +12,7 @@ go_library(
visibility = ["//visibility:public"],
deps = [
"//pkg/abi/linux",
+ "//pkg/context",
"//pkg/refs",
"//pkg/sync",
"//pkg/syserror",
diff --git a/pkg/tcpip/link/tun/device.go b/pkg/tcpip/link/tun/device.go
index 617446ea2..3b1510a33 100644
--- a/pkg/tcpip/link/tun/device.go
+++ b/pkg/tcpip/link/tun/device.go
@@ -18,6 +18,7 @@ import (
"fmt"
"gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/refs"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
@@ -64,14 +65,14 @@ func (d *Device) beforeSave() {
}
// Release implements fs.FileOperations.Release.
-func (d *Device) Release() {
+func (d *Device) Release(ctx context.Context) {
d.mu.Lock()
defer d.mu.Unlock()
// Decrease refcount if there is an endpoint associated with this file.
if d.endpoint != nil {
d.endpoint.RemoveNotify(d.notifyHandle)
- d.endpoint.DecRef()
+ d.endpoint.DecRef(ctx)
d.endpoint = nil
}
}
@@ -139,6 +140,7 @@ func attachOrCreateNIC(s *stack.Stack, name, prefix string, linkCaps stack.LinkE
stack: s,
nicID: id,
name: name,
+ isTap: prefix == "tap",
}
endpoint.Endpoint.LinkEPCapabilities = linkCaps
if endpoint.name == "" {
@@ -213,12 +215,11 @@ func (d *Device) Write(data []byte) (int64, error) {
remote = tcpip.LinkAddress(zeroMAC[:])
}
- pkt := stack.PacketBuffer{
- Data: buffer.View(data).ToVectorisedView(),
- }
- if ethHdr != nil {
- pkt.LinkHeader = buffer.View(ethHdr)
- }
+ pkt := stack.NewPacketBuffer(stack.PacketBufferOptions{
+ ReserveHeaderBytes: len(ethHdr),
+ Data: buffer.View(data).ToVectorisedView(),
+ })
+ copy(pkt.LinkHeader().Push(len(ethHdr)), ethHdr)
endpoint.InjectLinkAddr(protocol, remote, pkt)
return dataLen, nil
}
@@ -263,33 +264,22 @@ func (d *Device) encodePkt(info *channel.PacketInfo) (buffer.View, bool) {
// If the packet does not already have link layer header, and the route
// does not exist, we can't compute it. This is possibly a raw packet, tun
// device doesn't support this at the moment.
- if info.Pkt.LinkHeader == nil && info.Route.RemoteLinkAddress == "" {
+ if info.Pkt.LinkHeader().View().IsEmpty() && info.Route.RemoteLinkAddress == "" {
return nil, false
}
// Ethernet header (TAP only).
if d.hasFlags(linux.IFF_TAP) {
// Add ethernet header if not provided.
- if info.Pkt.LinkHeader == nil {
- hdr := &header.EthernetFields{
- SrcAddr: info.Route.LocalLinkAddress,
- DstAddr: info.Route.RemoteLinkAddress,
- Type: info.Proto,
- }
- if hdr.SrcAddr == "" {
- hdr.SrcAddr = d.endpoint.LinkAddress()
- }
-
- eth := make(header.Ethernet, header.EthernetMinimumSize)
- eth.Encode(hdr)
- vv.AppendView(buffer.View(eth))
- } else {
- vv.AppendView(info.Pkt.LinkHeader)
+ if info.Pkt.LinkHeader().View().IsEmpty() {
+ d.endpoint.AddHeader(info.Route.LocalLinkAddress, info.Route.RemoteLinkAddress, info.Proto, info.Pkt)
}
+ vv.AppendView(info.Pkt.LinkHeader().View())
}
// Append upper headers.
- vv.AppendView(buffer.View(info.Pkt.Header.View()[len(info.Pkt.LinkHeader):]))
+ vv.AppendView(info.Pkt.NetworkHeader().View())
+ vv.AppendView(info.Pkt.TransportHeader().View())
// Append data payload.
vv.Append(info.Pkt.Data)
@@ -348,11 +338,46 @@ type tunEndpoint struct {
stack *stack.Stack
nicID tcpip.NICID
name string
+ isTap bool
}
// DecRef decrements refcount of e, removes NIC if refcount goes to 0.
-func (e *tunEndpoint) DecRef() {
- e.DecRefWithDestructor(func() {
+func (e *tunEndpoint) DecRef(ctx context.Context) {
+ e.DecRefWithDestructor(ctx, func(context.Context) {
e.stack.RemoveNIC(e.nicID)
})
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (e *tunEndpoint) ARPHardwareType() header.ARPHardwareType {
+ if e.isTap {
+ return header.ARPHardwareEther
+ }
+ return header.ARPHardwareNone
+}
+
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *tunEndpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ if !e.isTap {
+ return
+ }
+ eth := header.Ethernet(pkt.LinkHeader().Push(header.EthernetMinimumSize))
+ hdr := &header.EthernetFields{
+ SrcAddr: local,
+ DstAddr: remote,
+ Type: protocol,
+ }
+ if hdr.SrcAddr == "" {
+ hdr.SrcAddr = e.LinkAddress()
+ }
+
+ eth.Encode(hdr)
+}
+
+// MaxHeaderLength returns the maximum size of the link layer header.
+func (e *tunEndpoint) MaxHeaderLength() uint16 {
+ if e.isTap {
+ return header.EthernetMinimumSize
+ }
+ return 0
+}
diff --git a/pkg/tcpip/link/waitable/BUILD b/pkg/tcpip/link/waitable/BUILD
index 0956d2c65..ee84c3d96 100644
--- a/pkg/tcpip/link/waitable/BUILD
+++ b/pkg/tcpip/link/waitable/BUILD
@@ -12,6 +12,7 @@ go_library(
"//pkg/gate",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
@@ -25,6 +26,7 @@ go_test(
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go
index 2b3741276..b152a0f26 100644
--- a/pkg/tcpip/link/waitable/waitable.go
+++ b/pkg/tcpip/link/waitable/waitable.go
@@ -25,6 +25,7 @@ import (
"gvisor.dev/gvisor/pkg/gate"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -50,12 +51,21 @@ func New(lower stack.LinkEndpoint) *Endpoint {
// It is called by the link-layer endpoint being wrapped when a packet arrives,
// and only forwards to the actual dispatcher if Wait or WaitDispatch haven't
// been called.
-func (e *Endpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
+func (e *Endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
if !e.dispatchGate.Enter() {
return
}
- e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, pkt)
+ e.dispatcher.DeliverNetworkPacket(remote, local, protocol, pkt)
+ e.dispatchGate.Leave()
+}
+
+// DeliverOutboundPacket implements stack.NetworkDispatcher.DeliverOutboundPacket.
+func (e *Endpoint) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ if !e.dispatchGate.Enter() {
+ return
+ }
+ e.dispatcher.DeliverOutboundPacket(remote, local, protocol, pkt)
e.dispatchGate.Leave()
}
@@ -99,7 +109,7 @@ func (e *Endpoint) LinkAddress() tcpip.LinkAddress {
// WritePacket implements stack.LinkEndpoint.WritePacket. It is called by
// higher-level protocols to write packets. It only forwards packets to the
// lower endpoint if Wait or WaitWrite haven't been called.
-func (e *Endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
+func (e *Endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
if !e.writeGate.Enter() {
return nil
}
@@ -147,3 +157,13 @@ func (e *Endpoint) WaitDispatch() {
// Wait implements stack.LinkEndpoint.Wait.
func (e *Endpoint) Wait() {}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (e *Endpoint) ARPHardwareType() header.ARPHardwareType {
+ return e.lower.ARPHardwareType()
+}
+
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *Endpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ e.lower.AddHeader(local, remote, protocol, pkt)
+}
diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go
index 54eb5322b..94827fc56 100644
--- a/pkg/tcpip/link/waitable/waitable_test.go
+++ b/pkg/tcpip/link/waitable/waitable_test.go
@@ -19,6 +19,7 @@ import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -35,10 +36,14 @@ type countedEndpoint struct {
dispatcher stack.NetworkDispatcher
}
-func (e *countedEndpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
+func (e *countedEndpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
e.dispatchCount++
}
+func (e *countedEndpoint) DeliverOutboundPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ panic("unimplemented")
+}
+
func (e *countedEndpoint) Attach(dispatcher stack.NetworkDispatcher) {
e.attachCount++
e.dispatcher = dispatcher
@@ -65,7 +70,7 @@ func (e *countedEndpoint) LinkAddress() tcpip.LinkAddress {
return e.linkAddr
}
-func (e *countedEndpoint) WritePacket(r *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error {
+func (e *countedEndpoint) WritePacket(r *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error {
e.writeCount++
return nil
}
@@ -81,29 +86,39 @@ func (e *countedEndpoint) WriteRawPacket(buffer.VectorisedView) *tcpip.Error {
return nil
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*countedEndpoint) ARPHardwareType() header.ARPHardwareType {
+ panic("unimplemented")
+}
+
// Wait implements stack.LinkEndpoint.Wait.
func (*countedEndpoint) Wait() {}
+// AddHeader implements stack.LinkEndpoint.AddHeader.
+func (e *countedEndpoint) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) {
+ panic("unimplemented")
+}
+
func TestWaitWrite(t *testing.T) {
ep := &countedEndpoint{}
wep := New(ep)
// Write and check that it goes through.
- wep.WritePacket(nil, nil /* gso */, 0, stack.PacketBuffer{})
+ wep.WritePacket(nil, nil /* gso */, 0, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
if want := 1; ep.writeCount != want {
t.Fatalf("Unexpected writeCount: got=%v, want=%v", ep.writeCount, want)
}
// Wait on dispatches, then try to write. It must go through.
wep.WaitDispatch()
- wep.WritePacket(nil, nil /* gso */, 0, stack.PacketBuffer{})
+ wep.WritePacket(nil, nil /* gso */, 0, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
if want := 2; ep.writeCount != want {
t.Fatalf("Unexpected writeCount: got=%v, want=%v", ep.writeCount, want)
}
// Wait on writes, then try to write. It must not go through.
wep.WaitWrite()
- wep.WritePacket(nil, nil /* gso */, 0, stack.PacketBuffer{})
+ wep.WritePacket(nil, nil /* gso */, 0, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
if want := 2; ep.writeCount != want {
t.Fatalf("Unexpected writeCount: got=%v, want=%v", ep.writeCount, want)
}
@@ -120,21 +135,21 @@ func TestWaitDispatch(t *testing.T) {
}
// Dispatch and check that it goes through.
- ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, stack.PacketBuffer{})
+ ep.dispatcher.DeliverNetworkPacket("", "", 0, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
if want := 1; ep.dispatchCount != want {
t.Fatalf("Unexpected dispatchCount: got=%v, want=%v", ep.dispatchCount, want)
}
// Wait on writes, then try to dispatch. It must go through.
wep.WaitWrite()
- ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, stack.PacketBuffer{})
+ ep.dispatcher.DeliverNetworkPacket("", "", 0, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
if want := 2; ep.dispatchCount != want {
t.Fatalf("Unexpected dispatchCount: got=%v, want=%v", ep.dispatchCount, want)
}
// Wait on dispatches, then try to dispatch. It must not go through.
wep.WaitDispatch()
- ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, stack.PacketBuffer{})
+ ep.dispatcher.DeliverNetworkPacket("", "", 0, stack.NewPacketBuffer(stack.PacketBufferOptions{}))
if want := 2; ep.dispatchCount != want {
t.Fatalf("Unexpected dispatchCount: got=%v, want=%v", ep.dispatchCount, want)
}