summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/link/fdbased
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/tcpip/link/fdbased')
-rw-r--r--pkg/tcpip/link/fdbased/BUILD4
-rw-r--r--pkg/tcpip/link/fdbased/endpoint.go171
-rw-r--r--pkg/tcpip/link/fdbased/endpoint_test.go32
-rw-r--r--pkg/tcpip/link/fdbased/mmap.go8
-rw-r--r--pkg/tcpip/link/fdbased/packet_dispatchers.go36
5 files changed, 208 insertions, 43 deletions
diff --git a/pkg/tcpip/link/fdbased/BUILD b/pkg/tcpip/link/fdbased/BUILD
index 8fa9e3984..897c94821 100644
--- a/pkg/tcpip/link/fdbased/BUILD
+++ b/pkg/tcpip/link/fdbased/BUILD
@@ -14,9 +14,7 @@ go_library(
"packet_dispatchers.go",
],
importpath = "gvisor.dev/gvisor/pkg/tcpip/link/fdbased",
- visibility = [
- "//visibility:public",
- ],
+ visibility = ["//visibility:public"],
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go
index f80ac3435..fa8a703d9 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
}
}
@@ -379,10 +386,11 @@ const (
// 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, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.NetworkProtocolNumber) *tcpip.Error {
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt tcpip.PacketBuffer) *tcpip.Error {
if e.hdrSize > 0 {
// Add ethernet header if needed.
- eth := header.Ethernet(hdr.Prepend(header.EthernetMinimumSize))
+ eth := header.Ethernet(pkt.Header.Prepend(header.EthernetMinimumSize))
+ pkt.LinkHeader = buffer.View(eth)
ethHdr := &header.EthernetFields{
DstAddr: r.RemoteLinkAddress,
Type: protocol,
@@ -397,17 +405,17 @@ 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 {
- vnetHdr.hdrLen = uint16(hdr.UsedLength())
+ vnetHdr.hdrLen = uint16(pkt.Header.UsedLength())
if gso.NeedsCsum {
vnetHdr.flags = _VIRTIO_NET_HDR_F_NEEDS_CSUM
vnetHdr.csumStart = header.EthernetMinimumSize + gso.L3HdrLen
vnetHdr.csumOffset = gso.CsumOffset
}
- if gso.Type != stack.GSONone && uint16(payload.Size()) > gso.MSS {
+ if gso.Type != stack.GSONone && uint16(pkt.Data.Size()) > gso.MSS {
switch gso.Type {
case stack.GSOTCPv4:
vnetHdr.gsoType = _VIRTIO_NET_HDR_GSO_TCPV4
@@ -420,18 +428,151 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prepen
}
}
- return rawfile.NonBlockingWrite3(e.fds[0], vnetHdrBuf, hdr.View(), payload.ToView())
+ return rawfile.NonBlockingWrite3(e.fds[0], vnetHdrBuf, pkt.Header.View(), pkt.Data.ToView())
+ }
+
+ if pkt.Data.Size() == 0 {
+ return rawfile.NonBlockingWrite(e.fds[0], pkt.Header.View())
}
- if payload.Size() == 0 {
- return rawfile.NonBlockingWrite(e.fds[0], hdr.View())
+ return rawfile.NonBlockingWrite3(e.fds[0], pkt.Header.View(), pkt.Data.ToView(), nil)
+}
+
+// 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, pkts []tcpip.PacketBuffer, 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++
}
- return rawfile.NonBlockingWrite3(e.fds[0], hdr.View(), payload.ToView(), nil)
+ n := len(pkts)
+
+ views := pkts[0].Data.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 pkts {
+ // TODO(b/134618279): Different packets may have different data
+ // in the future. We should handle this.
+ if !viewsEqual(pkts[i].Data.Views(), views) {
+ panic("All packets in pkts should have the same Data.")
+ }
+
+ prevIovecIdx := iovecIdx
+ mmsgHdr := &mmsgHdrs[i]
+ mmsgHdr.Msg.Iov = &iovec[iovecIdx]
+ packetSize := pkts[i].DataSize
+ hdr := &pkts[i].Header
+
+ off = pkts[i].DataOffset
+ 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
+}
+
+// viewsEqual tests whether v1 and v2 refer to the same backing bytes.
+func viewsEqual(vs1, vs2 []buffer.View) bool {
+ return len(vs1) == len(vs2) && (len(vs1) == 0 || &vs1[0] == &vs2[0])
+}
+
+// WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket.
+func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
+ return rawfile.NonBlockingWrite(e.fds[0], vv.ToView())
}
-// WriteRawPacket writes a raw packet directly to the file descriptor.
-func (e *endpoint) WriteRawPacket(dest tcpip.Address, packet []byte) *tcpip.Error {
+// 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 +609,9 @@ 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, pkt tcpip.PacketBuffer) {
+ e.dispatcher.DeliverNetworkPacket(e, "" /* 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 04406bc9a..2066987eb 100644
--- a/pkg/tcpip/link/fdbased/endpoint_test.go
+++ b/pkg/tcpip/link/fdbased/endpoint_test.go
@@ -45,7 +45,7 @@ const (
type packetInfo struct {
raddr tcpip.LinkAddress
proto tcpip.NetworkProtocolNumber
- contents buffer.View
+ contents tcpip.PacketBuffer
}
type context struct {
@@ -92,8 +92,8 @@ 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, pkt tcpip.PacketBuffer) {
+ c.ch <- packetInfo{remote, protocol, pkt}
}
func TestNoEthernetProperties(t *testing.T) {
@@ -168,7 +168,10 @@ func testWritePacket(t *testing.T, plen int, eth bool, gsoMaxSize uint32) {
L3HdrLen: header.IPv4MaximumHeaderSize,
}
}
- if err := c.ep.WritePacket(r, gso, hdr, payload.ToVectorisedView(), proto); err != nil {
+ if err := c.ep.WritePacket(r, gso, proto, tcpip.PacketBuffer{
+ Header: hdr,
+ Data: payload.ToVectorisedView(),
+ }); err != nil {
t.Fatalf("WritePacket failed: %v", err)
}
@@ -258,7 +261,10 @@ func TestPreserveSrcAddress(t *testing.T) {
// 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 */, hdr, buffer.VectorisedView{}, proto); err != nil {
+ if err := c.ep.WritePacket(r, nil /* gso */, proto, tcpip.PacketBuffer{
+ Header: hdr,
+ Data: buffer.VectorisedView{},
+ }); err != nil {
t.Fatalf("WritePacket failed: %v", err)
}
@@ -293,11 +299,12 @@ 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,
@@ -315,14 +322,21 @@ func TestDeliverPacket(t *testing.T) {
select {
case pi := <-c.ch:
want := packetInfo{
- raddr: raddr,
- proto: proto,
- contents: b,
+ raddr: raddr,
+ proto: proto,
+ contents: tcpip.PacketBuffer{
+ Data: buffer.View(b).ToVectorisedView(),
+ LinkHeader: buffer.View(hdr),
+ },
}
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)
}
diff --git a/pkg/tcpip/link/fdbased/mmap.go b/pkg/tcpip/link/fdbased/mmap.go
index 8bfeb97e4..62ed1e569 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 +190,9 @@ 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, tcpip.PacketBuffer{
+ Data: buffer.View(pkt).ToVectorisedView(),
+ LinkHeader: 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..c67d684ce 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()
@@ -138,10 +139,13 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) {
}
used := d.capViews(n, BufConfig)
- vv := buffer.NewVectorisedView(n, d.views[:used])
- vv.TrimFront(d.e.hdrSize)
+ pkt := tcpip.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, vv)
+ d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, pkt)
// Prepare e.views for another packet: release used views.
for i := 0; i < used; i++ {
@@ -194,7 +198,7 @@ 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 +229,7 @@ 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 +265,7 @@ 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 +275,10 @@ 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()
@@ -291,9 +296,12 @@ 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)
+ pkt := tcpip.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)
// Prepare e.views for another packet: release used views.
for i := 0; i < used; i++ {
generated by cgit v1.2.3 (git 2.39.1) at 2025-03-24 11:44:57 +0000