summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/link
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/tcpip/link')
-rw-r--r--pkg/tcpip/link/channel/channel.go48
-rw-r--r--pkg/tcpip/link/fdbased/endpoint.go143
-rw-r--r--pkg/tcpip/link/fdbased/endpoint_test.go21
-rw-r--r--pkg/tcpip/link/fdbased/mmap.go5
-rw-r--r--pkg/tcpip/link/fdbased/packet_dispatchers.go22
-rw-r--r--pkg/tcpip/link/loopback/BUILD1
-rw-r--r--pkg/tcpip/link/loopback/loopback.go27
-rw-r--r--pkg/tcpip/link/muxed/injectable.go34
-rw-r--r--pkg/tcpip/link/muxed/injectable_test.go2
-rw-r--r--pkg/tcpip/link/rawfile/BUILD5
-rw-r--r--pkg/tcpip/link/rawfile/rawfile_unsafe.go11
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem.go22
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem_test.go16
-rw-r--r--pkg/tcpip/link/sniffer/sniffer.go85
-rw-r--r--pkg/tcpip/link/waitable/waitable.go28
-rw-r--r--pkg/tcpip/link/waitable/waitable_test.go19
16 files changed, 411 insertions, 78 deletions
diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go
index 18adb2085..14f197a77 100644
--- a/pkg/tcpip/link/channel/channel.go
+++ b/pkg/tcpip/link/channel/channel.go
@@ -72,7 +72,7 @@ func (e *Endpoint) Inject(protocol tcpip.NetworkProtocolNumber, vv buffer.Vector
// InjectLinkAddr injects an inbound packet with a remote link address.
func (e *Endpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remote tcpip.LinkAddress, vv buffer.VectorisedView) {
- e.dispatcher.DeliverNetworkPacket(e, remote, "" /* local */, protocol, vv.Clone(nil))
+ e.dispatcher.DeliverNetworkPacket(e, remote, "" /* local */, protocol, vv.Clone(nil), nil /* linkHeader */)
}
// Attach saves the stack network-layer dispatcher for use later when packets
@@ -96,7 +96,7 @@ func (e *Endpoint) MTU() uint32 {
func (e *Endpoint) Capabilities() stack.LinkEndpointCapabilities {
caps := stack.LinkEndpointCapabilities(0)
if e.GSO {
- caps |= stack.CapabilityGSO
+ caps |= stack.CapabilityHardwareGSO
}
return caps
}
@@ -134,5 +134,49 @@ func (e *Endpoint) WritePacket(_ *stack.Route, gso *stack.GSO, hdr buffer.Prepen
return nil
}
+// WritePackets stores outbound packets into the channel.
+func (e *Endpoint) WritePackets(_ *stack.Route, gso *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ payloadView := payload.ToView()
+ n := 0
+packetLoop:
+ for i := range hdrs {
+ hdr := &hdrs[i].Hdr
+ off := hdrs[i].Off
+ size := hdrs[i].Size
+ p := PacketInfo{
+ Header: hdr.View(),
+ Proto: protocol,
+ Payload: buffer.NewViewFromBytes(payloadView[off : off+size]),
+ GSO: gso,
+ }
+
+ select {
+ case e.C <- p:
+ n++
+ default:
+ break packetLoop
+ }
+ }
+
+ return n, nil
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (e *Endpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ p := PacketInfo{
+ Header: packet.ToView(),
+ Proto: 0,
+ Payload: buffer.View{},
+ GSO: nil,
+ }
+
+ select {
+ case e.C <- p:
+ default:
+ }
+
+ return nil
+}
+
// Wait implements stack.LinkEndpoint.Wait.
func (*Endpoint) Wait() {}
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 = &ethHdrBuf[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++ {
diff --git a/pkg/tcpip/link/loopback/BUILD b/pkg/tcpip/link/loopback/BUILD
index 47a54845c..23e4d1418 100644
--- a/pkg/tcpip/link/loopback/BUILD
+++ b/pkg/tcpip/link/loopback/BUILD
@@ -10,6 +10,7 @@ go_library(
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/loopback/loopback.go b/pkg/tcpip/link/loopback/loopback.go
index b36629d2c..a3b48fa73 100644
--- a/pkg/tcpip/link/loopback/loopback.go
+++ b/pkg/tcpip/link/loopback/loopback.go
@@ -23,6 +23,7 @@ package loopback
import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -70,6 +71,9 @@ func (*endpoint) LinkAddress() tcpip.LinkAddress {
return ""
}
+// Wait implements stack.LinkEndpoint.Wait.
+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, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) *tcpip.Error {
@@ -81,10 +85,27 @@ func (e *endpoint) WritePacket(_ *stack.Route, _ *stack.GSO, hdr buffer.Prependa
// 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, vv)
+ e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, protocol, vv, nil /* linkHeader */)
return nil
}
-// Wait implements stack.LinkEndpoint.Wait.
-func (*endpoint) Wait() {}
+// WritePackets implements stack.LinkEndpoint.WritePackets.
+func (e *endpoint) WritePackets(_ *stack.Route, _ *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ panic("not implemented")
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (e *endpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ // Reject the packet if it's shorter than an ethernet header.
+ if packet.Size() < header.EthernetMinimumSize {
+ return tcpip.ErrBadAddress
+ }
+
+ // There should be an ethernet header at the beginning of packet.
+ linkHeader := header.Ethernet(packet.First()[:header.EthernetMinimumSize])
+ packet.TrimFront(len(linkHeader))
+ e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, linkHeader.Type(), packet, buffer.View(linkHeader))
+
+ return nil
+}
diff --git a/pkg/tcpip/link/muxed/injectable.go b/pkg/tcpip/link/muxed/injectable.go
index 7c946101d..682b60291 100644
--- a/pkg/tcpip/link/muxed/injectable.go
+++ b/pkg/tcpip/link/muxed/injectable.go
@@ -79,29 +79,47 @@ func (m *InjectableEndpoint) IsAttached() bool {
return m.dispatcher != nil
}
-// Inject implements stack.InjectableLinkEndpoint.
-func (m *InjectableEndpoint) Inject(protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView) {
- m.dispatcher.DeliverNetworkPacket(m, "" /* remote */, "" /* local */, protocol, vv)
+// InjectInbound implements stack.InjectableLinkEndpoint.
+func (m *InjectableEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView) {
+ m.dispatcher.DeliverNetworkPacket(m, "" /* remote */, "" /* local */, protocol, vv, nil /* linkHeader */)
+}
+
+// WritePackets 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) WritePackets(r *stack.Route, gso *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ endpoint, ok := m.routes[r.RemoteAddress]
+ if !ok {
+ return 0, tcpip.ErrNoRoute
+ }
+ return endpoint.WritePackets(r, gso, hdrs, payload, protocol)
}
// 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, _ *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) *tcpip.Error {
+func (m *InjectableEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) *tcpip.Error {
if endpoint, ok := m.routes[r.RemoteAddress]; ok {
- return endpoint.WritePacket(r, nil /* gso */, hdr, payload, protocol)
+ return endpoint.WritePacket(r, gso, hdr, payload, protocol)
}
return tcpip.ErrNoRoute
}
-// WriteRawPacket writes outbound packets to the appropriate
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (m *InjectableEndpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ // WriteRawPacket doesn't get a route or network address, so there's
+ // nowhere to write this.
+ return tcpip.ErrNoRoute
+}
+
+// InjectOutbound writes outbound packets to the appropriate
// LinkInjectableEndpoint based on the dest address.
-func (m *InjectableEndpoint) WriteRawPacket(dest tcpip.Address, packet []byte) *tcpip.Error {
+func (m *InjectableEndpoint) InjectOutbound(dest tcpip.Address, packet []byte) *tcpip.Error {
endpoint, ok := m.routes[dest]
if !ok {
return tcpip.ErrNoRoute
}
- return endpoint.WriteRawPacket(dest, packet)
+ return endpoint.InjectOutbound(dest, packet)
}
// Wait implements stack.LinkEndpoint.Wait.
diff --git a/pkg/tcpip/link/muxed/injectable_test.go b/pkg/tcpip/link/muxed/injectable_test.go
index 3086fec00..9cd300af8 100644
--- a/pkg/tcpip/link/muxed/injectable_test.go
+++ b/pkg/tcpip/link/muxed/injectable_test.go
@@ -31,7 +31,7 @@ import (
func TestInjectableEndpointRawDispatch(t *testing.T) {
endpoint, sock, dstIP := makeTestInjectableEndpoint(t)
- endpoint.WriteRawPacket(dstIP, []byte{0xFA})
+ endpoint.InjectOutbound(dstIP, []byte{0xFA})
buf := make([]byte, ipv4.MaxTotalSize)
bytesRead, err := sock.Read(buf)
diff --git a/pkg/tcpip/link/rawfile/BUILD b/pkg/tcpip/link/rawfile/BUILD
index 2e8bc772a..05c7b8024 100644
--- a/pkg/tcpip/link/rawfile/BUILD
+++ b/pkg/tcpip/link/rawfile/BUILD
@@ -16,5 +16,8 @@ go_library(
visibility = [
"//visibility:public",
],
- deps = ["//pkg/tcpip"],
+ deps = [
+ "//pkg/tcpip",
+ "@org_golang_x_sys//unix:go_default_library",
+ ],
)
diff --git a/pkg/tcpip/link/rawfile/rawfile_unsafe.go b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
index 7e286a3a6..44e25d475 100644
--- a/pkg/tcpip/link/rawfile/rawfile_unsafe.go
+++ b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
@@ -22,6 +22,7 @@ import (
"syscall"
"unsafe"
+ "golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/tcpip"
)
@@ -101,6 +102,16 @@ func NonBlockingWrite3(fd int, b1, b2, b3 []byte) *tcpip.Error {
return nil
}
+// NonBlockingSendMMsg sends multiple messages on a socket.
+func NonBlockingSendMMsg(fd int, msgHdrs []MMsgHdr) (int, *tcpip.Error) {
+ n, _, e := syscall.RawSyscall6(unix.SYS_SENDMMSG, uintptr(fd), uintptr(unsafe.Pointer(&msgHdrs[0])), uintptr(len(msgHdrs)), syscall.MSG_DONTWAIT, 0, 0)
+ if e != 0 {
+ return 0, TranslateErrno(e)
+ }
+
+ return int(n), nil
+}
+
// PollEvent represents the pollfd structure passed to a poll() system call.
type PollEvent struct {
FD int32
diff --git a/pkg/tcpip/link/sharedmem/sharedmem.go b/pkg/tcpip/link/sharedmem/sharedmem.go
index 9e71d4edf..279e2b457 100644
--- a/pkg/tcpip/link/sharedmem/sharedmem.go
+++ b/pkg/tcpip/link/sharedmem/sharedmem.go
@@ -212,6 +212,26 @@ func (e *endpoint) WritePacket(r *stack.Route, _ *stack.GSO, hdr buffer.Prependa
return nil
}
+// WritePackets implements stack.LinkEndpoint.WritePackets.
+func (e *endpoint) WritePackets(r *stack.Route, _ *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ panic("not implemented")
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (e *endpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ v := packet.ToView()
+ // Transmit the packet.
+ e.mu.Lock()
+ ok := e.tx.transmit(v, buffer.View{})
+ e.mu.Unlock()
+
+ if !ok {
+ return tcpip.ErrWouldBlock
+ }
+
+ return nil
+}
+
// dispatchLoop reads packets from the rx queue in a loop and dispatches them
// to the network stack.
func (e *endpoint) dispatchLoop(d stack.NetworkDispatcher) {
@@ -254,7 +274,7 @@ func (e *endpoint) dispatchLoop(d stack.NetworkDispatcher) {
// Send packet up the stack.
eth := header.Ethernet(b)
- d.DeliverNetworkPacket(e, eth.SourceAddress(), eth.DestinationAddress(), eth.Type(), buffer.View(b[header.EthernetMinimumSize:]).ToVectorisedView())
+ d.DeliverNetworkPacket(e, eth.SourceAddress(), eth.DestinationAddress(), eth.Type(), buffer.View(b[header.EthernetMinimumSize:]).ToVectorisedView(), buffer.View(eth))
}
// Clean state.
diff --git a/pkg/tcpip/link/sharedmem/sharedmem_test.go b/pkg/tcpip/link/sharedmem/sharedmem_test.go
index 0e9ba0846..f3e9705c9 100644
--- a/pkg/tcpip/link/sharedmem/sharedmem_test.go
+++ b/pkg/tcpip/link/sharedmem/sharedmem_test.go
@@ -78,9 +78,10 @@ func (q *queueBuffers) cleanup() {
}
type packetInfo struct {
- addr tcpip.LinkAddress
- proto tcpip.NetworkProtocolNumber
- vv buffer.VectorisedView
+ addr tcpip.LinkAddress
+ proto tcpip.NetworkProtocolNumber
+ vv buffer.VectorisedView
+ linkHeader buffer.View
}
type testContext struct {
@@ -130,12 +131,13 @@ 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, vv buffer.VectorisedView) {
+func (c *testContext) DeliverNetworkPacket(_ stack.LinkEndpoint, remoteLinkAddr, localLinkAddr tcpip.LinkAddress, proto tcpip.NetworkProtocolNumber, vv buffer.VectorisedView, linkHeader buffer.View) {
c.mu.Lock()
c.packets = append(c.packets, packetInfo{
- addr: remoteLinkAddr,
- proto: proto,
- vv: vv.Clone(nil),
+ addr: remoteLinkAddr,
+ proto: proto,
+ vv: vv.Clone(nil),
+ linkHeader: linkHeader,
})
c.mu.Unlock()
diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go
index e401dce44..39757ea2a 100644
--- a/pkg/tcpip/link/sniffer/sniffer.go
+++ b/pkg/tcpip/link/sniffer/sniffer.go
@@ -116,7 +116,7 @@ func NewWithFile(lower stack.LinkEndpoint, file *os.File, snapLen uint32) (stack
// 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, vv buffer.VectorisedView) {
+func (e *endpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView, linkHeader buffer.View) {
if atomic.LoadUint32(&LogPackets) == 1 && e.file == nil {
logPacket("recv", protocol, vv.First(), nil)
}
@@ -147,7 +147,7 @@ func (e *endpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local
panic(err)
}
}
- e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, vv)
+ e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, vv, linkHeader)
}
// Attach implements the stack.LinkEndpoint interface. It saves the dispatcher
@@ -193,10 +193,7 @@ func (e *endpoint) GSOMaxSize() uint32 {
return 0
}
-// 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, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) *tcpip.Error {
+func (e *endpoint) dumpPacket(gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) {
if atomic.LoadUint32(&LogPackets) == 1 && e.file == nil {
logPacket("send", protocol, hdr.View(), gso)
}
@@ -218,28 +215,74 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prepen
panic(err)
}
length -= len(hdrBuf)
- if length > 0 {
- for _, v := range payload.Views() {
- if len(v) > length {
- v = v[:length]
- }
- n, err := buf.Write(v)
- if err != nil {
- panic(err)
- }
- length -= n
- if length == 0 {
- break
- }
- }
- }
+ logVectorisedView(payload, length, buf)
if _, err := e.file.Write(buf.Bytes()); err != nil {
panic(err)
}
}
+}
+
+// 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, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) *tcpip.Error {
+ e.dumpPacket(gso, hdr, payload, protocol)
return e.lower.WritePacket(r, gso, hdr, payload, protocol)
}
+// WritePackets 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) WritePackets(r *stack.Route, gso *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ view := payload.ToView()
+ for _, d := range hdrs {
+ e.dumpPacket(gso, d.Hdr, buffer.NewVectorisedView(d.Size, []buffer.View{view[d.Off:][:d.Size]}), protocol)
+ }
+ return e.lower.WritePackets(r, gso, hdrs, payload, protocol)
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (e *endpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ if atomic.LoadUint32(&LogPackets) == 1 && e.file == nil {
+ logPacket("send", 0, buffer.View("[raw packet, no header available]"), nil /* gso */)
+ }
+ if e.file != nil && atomic.LoadUint32(&LogPacketsToFile) == 1 {
+ length := packet.Size()
+ if length > int(e.maxPCAPLen) {
+ length = int(e.maxPCAPLen)
+ }
+
+ buf := bytes.NewBuffer(make([]byte, 0, pcapPacketHeaderLen+length))
+ if err := binary.Write(buf, binary.BigEndian, newPCAPPacketHeader(uint32(length), uint32(packet.Size()))); err != nil {
+ panic(err)
+ }
+ logVectorisedView(packet, length, buf)
+ if _, err := e.file.Write(buf.Bytes()); err != nil {
+ panic(err)
+ }
+ }
+ return e.lower.WriteRawPacket(packet)
+}
+
+func logVectorisedView(vv buffer.VectorisedView, length int, buf *bytes.Buffer) {
+ if length <= 0 {
+ return
+ }
+ for _, v := range vv.Views() {
+ if len(v) > length {
+ v = v[:length]
+ }
+ n, err := buf.Write(v)
+ if err != nil {
+ panic(err)
+ }
+ length -= n
+ if length == 0 {
+ return
+ }
+ }
+}
+
// Wait implements stack.LinkEndpoint.Wait.
func (*endpoint) Wait() {}
diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go
index 5a1791cb5..a04fc1062 100644
--- a/pkg/tcpip/link/waitable/waitable.go
+++ b/pkg/tcpip/link/waitable/waitable.go
@@ -50,12 +50,12 @@ 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, vv buffer.VectorisedView) {
+func (e *Endpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView, linkHeader buffer.View) {
if !e.dispatchGate.Enter() {
return
}
- e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, vv)
+ e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, vv, linkHeader)
e.dispatchGate.Leave()
}
@@ -109,6 +109,30 @@ func (e *Endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prepen
return err
}
+// WritePackets implements stack.LinkEndpoint.WritePackets. 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) WritePackets(r *stack.Route, gso *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ if !e.writeGate.Enter() {
+ return len(hdrs), nil
+ }
+
+ n, err := e.lower.WritePackets(r, gso, hdrs, payload, protocol)
+ e.writeGate.Leave()
+ return n, err
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (e *Endpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ if !e.writeGate.Enter() {
+ return nil
+ }
+
+ err := e.lower.WriteRawPacket(packet)
+ e.writeGate.Leave()
+ return err
+}
+
// WaitWrite prevents new calls to WritePacket from reaching the lower endpoint,
// and waits for inflight ones to finish before returning.
func (e *Endpoint) WaitWrite() {
diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go
index ae23c96b7..5f0f8fa2d 100644
--- a/pkg/tcpip/link/waitable/waitable_test.go
+++ b/pkg/tcpip/link/waitable/waitable_test.go
@@ -35,7 +35,7 @@ type countedEndpoint struct {
dispatcher stack.NetworkDispatcher
}
-func (e *countedEndpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView) {
+func (e *countedEndpoint) DeliverNetworkPacket(linkEP stack.LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, vv buffer.VectorisedView, linkHeader buffer.View) {
e.dispatchCount++
}
@@ -70,6 +70,17 @@ func (e *countedEndpoint) WritePacket(r *stack.Route, _ *stack.GSO, hdr buffer.P
return nil
}
+// WritePackets implements stack.LinkEndpoint.WritePackets.
+func (e *countedEndpoint) WritePackets(r *stack.Route, _ *stack.GSO, hdrs []stack.PacketDescriptor, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) {
+ e.writeCount += len(hdrs)
+ return len(hdrs), nil
+}
+
+func (e *countedEndpoint) WriteRawPacket(packet buffer.VectorisedView) *tcpip.Error {
+ e.writeCount++
+ return nil
+}
+
// Wait implements stack.LinkEndpoint.Wait.
func (*countedEndpoint) Wait() {}
@@ -109,21 +120,21 @@ func TestWaitDispatch(t *testing.T) {
}
// Dispatch and check that it goes through.
- ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, buffer.VectorisedView{})
+ ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, buffer.VectorisedView{}, buffer.View{})
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, buffer.VectorisedView{})
+ ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, buffer.VectorisedView{}, buffer.View{})
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, buffer.VectorisedView{})
+ ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, buffer.VectorisedView{}, buffer.View{})
if want := 2; ep.dispatchCount != want {
t.Fatalf("Unexpected dispatchCount: got=%v, want=%v", ep.dispatchCount, want)
}