diff options
Diffstat (limited to 'pkg/tcpip')
65 files changed, 586 insertions, 436 deletions
diff --git a/pkg/tcpip/adapters/gonet/BUILD b/pkg/tcpip/adapters/gonet/BUILD index e57d45f2a..a984f1712 100644 --- a/pkg/tcpip/adapters/gonet/BUILD +++ b/pkg/tcpip/adapters/gonet/BUILD @@ -22,7 +22,6 @@ go_test( size = "small", srcs = ["gonet_test.go"], library = ":gonet", - tags = ["flaky"], deps = [ "//pkg/tcpip", "//pkg/tcpip/header", diff --git a/pkg/tcpip/adapters/gonet/gonet.go b/pkg/tcpip/adapters/gonet/gonet.go index 6e0db2741..d82ed5205 100644 --- a/pkg/tcpip/adapters/gonet/gonet.go +++ b/pkg/tcpip/adapters/gonet/gonet.go @@ -335,6 +335,11 @@ func (c *TCPConn) Read(b []byte) (int, error) { deadline := c.readCancel() numRead := 0 + defer func() { + if numRead != 0 { + c.ep.ModerateRecvBuf(numRead) + } + }() for numRead != len(b) { if len(c.read) == 0 { var err error diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go index 9bf67686d..20b183da0 100644 --- a/pkg/tcpip/link/channel/channel.go +++ b/pkg/tcpip/link/channel/channel.go @@ -181,13 +181,13 @@ 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 +229,13 @@ 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, diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go index affa1bbdf..f34082e1a 100644 --- a/pkg/tcpip/link/fdbased/endpoint.go +++ b/pkg/tcpip/link/fdbased/endpoint.go @@ -387,7 +387,7 @@ 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, 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.hdrSize > 0 { // Add ethernet header if needed. eth := header.Ethernet(pkt.Header.Prepend(header.EthernetMinimumSize)) @@ -641,8 +641,8 @@ 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..eaee7e5d7 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 stack.PacketBuffer + contents *stack.PacketBuffer } type context struct { @@ -103,7 +103,7 @@ 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} } @@ -179,7 +179,7 @@ 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{ + if err := c.ep.WritePacket(r, gso, proto, &stack.PacketBuffer{ Header: hdr, Data: payload.ToVectorisedView(), Hash: hash, @@ -295,7 +295,7 @@ 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 */, proto, stack.PacketBuffer{ + if err := c.ep.WritePacket(r, nil /* gso */, proto, &stack.PacketBuffer{ Header: hdr, Data: buffer.VectorisedView{}, }); err != nil { @@ -358,7 +358,7 @@ func TestDeliverPacket(t *testing.T) { want := packetInfo{ raddr: raddr, proto: proto, - contents: stack.PacketBuffer{ + contents: &stack.PacketBuffer{ Data: buffer.View(b).ToVectorisedView(), LinkHeader: buffer.View(hdr), }, diff --git a/pkg/tcpip/link/fdbased/mmap.go b/pkg/tcpip/link/fdbased/mmap.go index fe2bf3b0b..2dfd29aa9 100644 --- a/pkg/tcpip/link/fdbased/mmap.go +++ b/pkg/tcpip/link/fdbased/mmap.go @@ -191,7 +191,7 @@ func (d *packetMMapDispatcher) dispatch() (bool, *tcpip.Error) { } pkt = pkt[d.e.hdrSize:] - d.e.dispatcher.DeliverNetworkPacket(d.e, remote, local, p, stack.PacketBuffer{ + d.e.dispatcher.DeliverNetworkPacket(remote, local, p, &stack.PacketBuffer{ Data: buffer.View(pkt).ToVectorisedView(), LinkHeader: buffer.View(eth), }) diff --git a/pkg/tcpip/link/fdbased/packet_dispatchers.go b/pkg/tcpip/link/fdbased/packet_dispatchers.go index cb4cbea69..f04738cfb 100644 --- a/pkg/tcpip/link/fdbased/packet_dispatchers.go +++ b/pkg/tcpip/link/fdbased/packet_dispatchers.go @@ -139,13 +139,13 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) { } used := d.capViews(n, BufConfig) - pkt := stack.PacketBuffer{ + 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 +169,7 @@ 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 @@ -296,12 +296,12 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) { } used := d.capViews(k, int(n), BufConfig) - pkt := stack.PacketBuffer{ + 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..568c6874f 100644 --- a/pkg/tcpip/link/loopback/loopback.go +++ b/pkg/tcpip/link/loopback/loopback.go @@ -76,7 +76,7 @@ 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 { +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()...) @@ -84,7 +84,7 @@ func (e *endpoint) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.Netw // 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{ + e.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, protocol, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), }) @@ -106,7 +106,7 @@ func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error { } linkHeader := header.Ethernet(hdr) vv.TrimFront(len(linkHeader)) - e.dispatcher.DeliverNetworkPacket(e, "" /* remote */, "" /* local */, linkHeader.Type(), stack.PacketBuffer{ + e.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, linkHeader.Type(), &stack.PacketBuffer{ Data: vv, LinkHeader: buffer.View(linkHeader), }) diff --git a/pkg/tcpip/link/muxed/injectable.go b/pkg/tcpip/link/muxed/injectable.go index a5478ce17..c69d6b7e9 100644 --- a/pkg/tcpip/link/muxed/injectable.go +++ b/pkg/tcpip/link/muxed/injectable.go @@ -80,8 +80,8 @@ 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 +98,7 @@ 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) } diff --git a/pkg/tcpip/link/muxed/injectable_test.go b/pkg/tcpip/link/muxed/injectable_test.go index 87c734c1f..0744f66d6 100644 --- a/pkg/tcpip/link/muxed/injectable_test.go +++ b/pkg/tcpip/link/muxed/injectable_test.go @@ -50,7 +50,7 @@ func TestInjectableEndpointDispatch(t *testing.T) { hdr.Prepend(1)[0] = 0xFA packetRoute := stack.Route{RemoteAddress: dstIP} - endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, stack.PacketBuffer{ + endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buffer.NewViewFromBytes([]byte{0xFB}).ToVectorisedView(), }) @@ -70,7 +70,7 @@ func TestInjectableEndpointDispatchHdrOnly(t *testing.T) { hdr := buffer.NewPrependable(1) hdr.Prepend(1)[0] = 0xFA packetRoute := stack.Route{RemoteAddress: dstIP} - endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, stack.PacketBuffer{ + endpoint.WritePacket(&packetRoute, nil /* gso */, ipv4.ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buffer.NewView(0).ToVectorisedView(), }) diff --git a/pkg/tcpip/link/qdisc/fifo/endpoint.go b/pkg/tcpip/link/qdisc/fifo/endpoint.go index 54432194d..b5dfb7850 100644 --- a/pkg/tcpip/link/qdisc/fifo/endpoint.go +++ b/pkg/tcpip/link/qdisc/fifo/endpoint.go @@ -102,8 +102,8 @@ 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) } // Attach implements stack.LinkEndpoint.Attach. @@ -146,7 +146,7 @@ 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 +154,7 @@ 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() 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/sharedmem/sharedmem.go b/pkg/tcpip/link/sharedmem/sharedmem.go index 0796d717e..0374a2441 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem.go +++ b/pkg/tcpip/link/sharedmem/sharedmem.go @@ -185,7 +185,7 @@ func (e *endpoint) LinkAddress() tcpip.LinkAddress { // 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 { +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) @@ -275,7 +275,7 @@ func (e *endpoint) dispatchLoop(d stack.NetworkDispatcher) { // Send packet up the stack. eth := header.Ethernet(b[:header.EthernetMinimumSize]) - d.DeliverNetworkPacket(e, eth.SourceAddress(), eth.DestinationAddress(), eth.Type(), stack.PacketBuffer{ + d.DeliverNetworkPacket(eth.SourceAddress(), eth.DestinationAddress(), eth.Type(), &stack.PacketBuffer{ Data: buffer.View(b[header.EthernetMinimumSize:]).ToVectorisedView(), LinkHeader: buffer.View(eth), }) diff --git a/pkg/tcpip/link/sharedmem/sharedmem_test.go b/pkg/tcpip/link/sharedmem/sharedmem_test.go index 33f640b85..28a2e88ba 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, @@ -273,7 +273,7 @@ func TestSimpleSend(t *testing.T) { randomFill(buf) proto := tcpip.NetworkProtocolNumber(rand.Intn(0x10000)) - if err := c.ep.WritePacket(&r, nil /* gso */, proto, stack.PacketBuffer{ + if err := c.ep.WritePacket(&r, nil /* gso */, proto, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { @@ -345,7 +345,7 @@ func TestPreserveSrcAddressInSend(t *testing.T) { hdr := buffer.NewPrependable(header.EthernetMinimumSize) proto := tcpip.NetworkProtocolNumber(rand.Intn(0x10000)) - if err := c.ep.WritePacket(&r, nil /* gso */, proto, stack.PacketBuffer{ + if err := c.ep.WritePacket(&r, nil /* gso */, proto, &stack.PacketBuffer{ Header: hdr, }); err != nil { t.Fatalf("WritePacket failed: %v", err) @@ -401,7 +401,7 @@ func TestFillTxQueue(t *testing.T) { 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{ + if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { @@ -419,7 +419,7 @@ 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{ + if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != want { @@ -447,7 +447,7 @@ 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{ + if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { @@ -470,7 +470,7 @@ func TestFillTxQueueAfterBadCompletion(t *testing.T) { 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{ + if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { @@ -488,7 +488,7 @@ 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{ + if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != want { @@ -514,7 +514,7 @@ func TestFillTxMemory(t *testing.T) { 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{ + if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { @@ -533,7 +533,7 @@ 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{ + err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }) @@ -561,7 +561,7 @@ func TestFillTxMemoryWithMultiBuffer(t *testing.T) { // 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{ + if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { @@ -577,7 +577,7 @@ func TestFillTxMemoryWithMultiBuffer(t *testing.T) { { 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{ + if want, err := tcpip.ErrWouldBlock, c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: uu, }); err != want { @@ -588,7 +588,7 @@ func TestFillTxMemoryWithMultiBuffer(t *testing.T) { // 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{ + if err := c.ep.WritePacket(&r, nil /* gso */, header.IPv4ProtocolNumber, &stack.PacketBuffer{ Header: hdr, Data: buf.ToVectorisedView(), }); err != nil { diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go index da1c520ae..ae3186314 100644 --- a/pkg/tcpip/link/sniffer/sniffer.go +++ b/pkg/tcpip/link/sniffer/sniffer.go @@ -120,9 +120,9 @@ func NewWithWriter(lower stack.LinkEndpoint, writer io.Writer, snapLen uint32) ( // 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) +func (e *endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) { + e.dumpPacket("recv", nil, protocol, pkt) + e.dispatcher.DeliverNetworkPacket(remote, local, protocol, pkt) } // Attach implements the stack.LinkEndpoint interface. It saves the dispatcher @@ -208,8 +208,8 @@ 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) +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) } diff --git a/pkg/tcpip/link/tun/device.go b/pkg/tcpip/link/tun/device.go index 617446ea2..6bc9033d0 100644 --- a/pkg/tcpip/link/tun/device.go +++ b/pkg/tcpip/link/tun/device.go @@ -213,7 +213,7 @@ func (d *Device) Write(data []byte) (int64, error) { remote = tcpip.LinkAddress(zeroMAC[:]) } - pkt := stack.PacketBuffer{ + pkt := &stack.PacketBuffer{ Data: buffer.View(data).ToVectorisedView(), } if ethHdr != nil { diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go index 2b3741276..949b3f2b2 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, 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() } @@ -99,7 +99,7 @@ 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 } diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go index 54eb5322b..63bf40562 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, pkt stack.PacketBuffer) { +func (e *countedEndpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) { e.dispatchCount++ } @@ -65,7 +65,7 @@ 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 } @@ -89,21 +89,21 @@ func TestWaitWrite(t *testing.T) { 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.PacketBuffer{}) 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.PacketBuffer{}) 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.PacketBuffer{}) if want := 2; ep.writeCount != want { t.Fatalf("Unexpected writeCount: got=%v, want=%v", ep.writeCount, want) } @@ -120,21 +120,21 @@ func TestWaitDispatch(t *testing.T) { } // Dispatch and check that it goes through. - ep.dispatcher.DeliverNetworkPacket(ep, "", "", 0, stack.PacketBuffer{}) + ep.dispatcher.DeliverNetworkPacket("", "", 0, &stack.PacketBuffer{}) 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.PacketBuffer{}) 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.PacketBuffer{}) if want := 2; ep.dispatchCount != want { t.Fatalf("Unexpected dispatchCount: got=%v, want=%v", ep.dispatchCount, want) } diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go index 9d0797af7..ea1acba83 100644 --- a/pkg/tcpip/network/arp/arp.go +++ b/pkg/tcpip/network/arp/arp.go @@ -80,7 +80,7 @@ func (e *endpoint) MaxHeaderLength() uint16 { func (e *endpoint) Close() {} -func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, stack.NetworkHeaderParams, stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, stack.NetworkHeaderParams, *stack.PacketBuffer) *tcpip.Error { return tcpip.ErrNotSupported } @@ -94,11 +94,11 @@ func (e *endpoint) WritePackets(*stack.Route, *stack.GSO, stack.PacketBufferList return 0, tcpip.ErrNotSupported } -func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { return tcpip.ErrNotSupported } -func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { v, ok := pkt.Data.PullUp(header.ARPSize) if !ok { return @@ -122,7 +122,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { copy(packet.ProtocolAddressSender(), h.ProtocolAddressTarget()) copy(packet.HardwareAddressTarget(), h.HardwareAddressSender()) copy(packet.ProtocolAddressTarget(), h.ProtocolAddressSender()) - e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, stack.PacketBuffer{ + e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ Header: hdr, }) fallthrough // also fill the cache from requests @@ -177,7 +177,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. copy(h.ProtocolAddressSender(), localAddr) copy(h.ProtocolAddressTarget(), addr) - return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, stack.PacketBuffer{ + return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ Header: hdr, }) } diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go index 1646d9cde..66e67429c 100644 --- a/pkg/tcpip/network/arp/arp_test.go +++ b/pkg/tcpip/network/arp/arp_test.go @@ -103,7 +103,7 @@ func TestDirectRequest(t *testing.T) { inject := func(addr tcpip.Address) { copy(h.ProtocolAddressTarget(), addr) - c.linkEP.InjectInbound(arp.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(arp.ProtocolNumber, &stack.PacketBuffer{ Data: v.ToVectorisedView(), }) } diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go index 4c20301c6..d9b62f2db 100644 --- a/pkg/tcpip/network/ip_test.go +++ b/pkg/tcpip/network/ip_test.go @@ -96,7 +96,7 @@ func (t *testObject) checkValues(protocol tcpip.TransportProtocolNumber, vv buff // DeliverTransportPacket is called by network endpoints after parsing incoming // packets. This is used by the test object to verify that the results of the // parsing are expected. -func (t *testObject) DeliverTransportPacket(r *stack.Route, protocol tcpip.TransportProtocolNumber, pkt stack.PacketBuffer) { +func (t *testObject) DeliverTransportPacket(r *stack.Route, protocol tcpip.TransportProtocolNumber, pkt *stack.PacketBuffer) { t.checkValues(protocol, pkt.Data, r.RemoteAddress, r.LocalAddress) t.dataCalls++ } @@ -104,7 +104,7 @@ func (t *testObject) DeliverTransportPacket(r *stack.Route, protocol tcpip.Trans // DeliverTransportControlPacket is called by network endpoints after parsing // incoming control (ICMP) packets. This is used by the test object to verify // that the results of the parsing are expected. -func (t *testObject) DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (t *testObject) DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { t.checkValues(trans, pkt.Data, remote, local) if typ != t.typ { t.t.Errorf("typ = %v, want %v", typ, t.typ) @@ -150,7 +150,7 @@ func (*testObject) Wait() {} // WritePacket is called by network endpoints after producing a packet and // writing it to the link endpoint. This is used by the test object to verify // that the produced packet is as expected. -func (t *testObject) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error { +func (t *testObject) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error { var prot tcpip.TransportProtocolNumber var srcAddr tcpip.Address var dstAddr tcpip.Address @@ -246,7 +246,11 @@ func TestIPv4Send(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{Protocol: 123, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{ + Protocol: 123, + TTL: 123, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload.ToVectorisedView(), }); err != nil { @@ -289,7 +293,7 @@ func TestIPv4Receive(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: view.ToVectorisedView(), }) if o.dataCalls != 1 { @@ -379,7 +383,7 @@ func TestIPv4ReceiveControl(t *testing.T) { o.extra = c.expectedExtra vv := view[:len(view)-c.trunc].ToVectorisedView() - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: vv, }) if want := c.expectedCount; o.controlCalls != want { @@ -444,7 +448,7 @@ func TestIPv4FragmentationReceive(t *testing.T) { } // Send first segment. - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: frag1.ToVectorisedView(), }) if o.dataCalls != 0 { @@ -452,7 +456,7 @@ func TestIPv4FragmentationReceive(t *testing.T) { } // Send second segment. - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: frag2.ToVectorisedView(), }) if o.dataCalls != 1 { @@ -487,7 +491,11 @@ func TestIPv6Send(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{Protocol: 123, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{ + Protocol: 123, + TTL: 123, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload.ToVectorisedView(), }); err != nil { @@ -530,7 +538,7 @@ func TestIPv6Receive(t *testing.T) { t.Fatalf("could not find route: %v", err) } - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: view.ToVectorisedView(), }) if o.dataCalls != 1 { @@ -644,7 +652,7 @@ func TestIPv6ReceiveControl(t *testing.T) { // Set ICMPv6 checksum. icmp.SetChecksum(header.ICMPv6Checksum(icmp, outerSrcAddr, localIpv6Addr, buffer.VectorisedView{})) - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: view[:len(view)-c.trunc].ToVectorisedView(), }) if want := c.expectedCount; o.controlCalls != want { diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index 4cbefe5ab..d1c3ae835 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/pkg/tcpip/network/ipv4/icmp.go @@ -24,7 +24,7 @@ import ( // the original packet that caused the ICMP one to be sent. This information is // used to find out which transport endpoint must be notified about the ICMP // packet. -func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { h, ok := pkt.Data.PullUp(header.IPv4MinimumSize) if !ok { return @@ -56,7 +56,7 @@ func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack. e.dispatcher.DeliverTransportControlPacket(e.id.LocalAddress, hdr.DestinationAddress(), ProtocolNumber, p, typ, extra, pkt) } -func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { +func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer) { stats := r.Stats() received := stats.ICMP.V4PacketsReceived v, ok := pkt.Data.PullUp(header.ICMPv4MinimumSize) @@ -88,7 +88,7 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { // It's possible that a raw socket expects to receive this. h.SetChecksum(wantChecksum) - e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, stack.PacketBuffer{ + e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, &stack.PacketBuffer{ Data: pkt.Data.Clone(nil), NetworkHeader: append(buffer.View(nil), pkt.NetworkHeader...), }) @@ -102,7 +102,11 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { pkt.SetChecksum(0) pkt.SetChecksum(^header.Checksum(pkt, header.ChecksumVV(vv, 0))) sent := stats.ICMP.V4PacketsSent - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{ + Protocol: header.ICMPv4ProtocolNumber, + TTL: r.DefaultTTL(), + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: vv, TransportHeader: buffer.View(pkt), diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index 64046cbbf..959f7e007 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/pkg/tcpip/network/ipv4/ipv4.go @@ -129,7 +129,7 @@ func (e *endpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNumber { // packet's stated length matches the length of the header+payload. mtu // includes the IP header and options. This does not support the DontFragment // IP flag. -func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, pkt *stack.PacketBuffer) *tcpip.Error { // This packet is too big, it needs to be fragmented. ip := header.IPv4(pkt.Header.View()) flags := ip.Flags() @@ -169,7 +169,7 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, if i > 0 { newPayload := pkt.Data.Clone(nil) newPayload.CapLength(innerMTU) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, stack.PacketBuffer{ + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ Header: pkt.Header, Data: newPayload, NetworkHeader: buffer.View(h), @@ -188,7 +188,7 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, newPayload := pkt.Data.Clone(nil) newPayloadLength := outerMTU - pkt.Header.UsedLength() newPayload.CapLength(newPayloadLength) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, stack.PacketBuffer{ + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ Header: pkt.Header, Data: newPayload, NetworkHeader: buffer.View(h), @@ -202,7 +202,7 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, startOfHdr := pkt.Header startOfHdr.TrimBack(pkt.Header.UsedLength() - outerMTU) emptyVV := buffer.NewVectorisedView(0, []buffer.View{}) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, stack.PacketBuffer{ + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ Header: startOfHdr, Data: emptyVV, NetworkHeader: buffer.View(h), @@ -245,7 +245,7 @@ func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadS } // WritePacket writes a packet to the given destination address and protocol. -func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) pkt.NetworkHeader = buffer.View(ip) @@ -253,43 +253,29 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw // iptables filtering. All packets that reach here are locally // generated. ipt := e.stack.IPTables() - if ok := ipt.Check(stack.Output, &pkt, gso, r, "", nicName); !ok { + if ok := ipt.Check(stack.Output, pkt, gso, r, "", nicName); !ok { // iptables is telling us to drop the packet. return nil } + // If the packet is manipulated as per NAT Ouput rules, handle packet + // based on destination address and do not send the packet to link layer. + // TODO(gvisor.dev/issue/170): We should do this for every packet, rather than + // only NATted packets, but removing this check short circuits broadcasts + // before they are sent out to other hosts. if pkt.NatDone { - // If the packet is manipulated as per NAT Ouput rules, handle packet - // based on destination address and do not send the packet to link layer. netHeader := header.IPv4(pkt.NetworkHeader) ep, err := e.stack.FindNetworkEndpoint(header.IPv4ProtocolNumber, netHeader.DestinationAddress()) if err == nil { - src := netHeader.SourceAddress() - dst := netHeader.DestinationAddress() - route := r.ReverseRoute(src, dst) - - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) - packet := stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views)} - ep.HandlePacket(&route, packet) + route := r.ReverseRoute(netHeader.SourceAddress(), netHeader.DestinationAddress()) + handleLoopback(&route, pkt, ep) return nil } } if r.Loop&stack.PacketLoop != 0 { - // The inbound path expects the network header to still be in - // the PacketBuffer's Data field. - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) loopedR := r.MakeLoopedRoute() - - e.HandlePacket(&loopedR, stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), - }) - + handleLoopback(&loopedR, pkt, e) loopedR.Release() } if r.Loop&stack.PacketOut == 0 { @@ -305,6 +291,17 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw return nil } +func handleLoopback(route *stack.Route, pkt *stack.PacketBuffer, ep stack.NetworkEndpoint) { + // The inbound path expects the network header to still be in + // the PacketBuffer's Data field. + views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) + views[0] = pkt.Header.View() + views = append(views, pkt.Data.Views()...) + ep.HandlePacket(route, &stack.PacketBuffer{ + Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), + }) +} + // WritePackets implements stack.NetworkEndpoint.WritePackets. func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.PacketBufferList, params stack.NetworkHeaderParams) (int, *tcpip.Error) { if r.Loop&stack.PacketLoop != 0 { @@ -347,18 +344,12 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe src := netHeader.SourceAddress() dst := netHeader.DestinationAddress() route := r.ReverseRoute(src, dst) - - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) - packet := stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views)} - ep.HandlePacket(&route, packet) + handleLoopback(&route, pkt, ep) n++ continue } } - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, *pkt); err != nil { + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, pkt); err != nil { r.Stats().IP.PacketsSent.IncrementBy(uint64(n)) return n, err } @@ -370,7 +361,7 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe // WriteHeaderIncludedPacket writes a packet already containing a network // header through the given route. -func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { // The packet already has an IP header, but there are a few required // checks. h, ok := pkt.Data.PullUp(header.IPv4MinimumSize) @@ -426,7 +417,7 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuf // HandlePacket is called by the link layer when new ipv4 packets arrive for // this endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { headerView, ok := pkt.Data.PullUp(header.IPv4MinimumSize) if !ok { r.Stats().IP.MalformedPacketsReceived.Increment() @@ -447,7 +438,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { // iptables filtering. All packets that reach here are intended for // this machine and will not be forwarded. ipt := e.stack.IPTables() - if ok := ipt.Check(stack.Input, &pkt, nil, nil, "", ""); !ok { + if ok := ipt.Check(stack.Input, pkt, nil, nil, "", ""); !ok { // iptables is telling us to drop the packet. return } diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go index 36035c820..c208ebd99 100644 --- a/pkg/tcpip/network/ipv4/ipv4_test.go +++ b/pkg/tcpip/network/ipv4/ipv4_test.go @@ -114,7 +114,7 @@ func makeHdrAndPayload(hdrLength int, extraLength int, viewSizes []int) (buffer. // comparePayloads compared the contents of all the packets against the contents // of the source packet. -func compareFragments(t *testing.T, packets []stack.PacketBuffer, sourcePacketInfo stack.PacketBuffer, mtu uint32) { +func compareFragments(t *testing.T, packets []*stack.PacketBuffer, sourcePacketInfo *stack.PacketBuffer, mtu uint32) { t.Helper() // Make a complete array of the sourcePacketInfo packet. source := header.IPv4(packets[0].Header.View()[:header.IPv4MinimumSize]) @@ -174,7 +174,7 @@ func compareFragments(t *testing.T, packets []stack.PacketBuffer, sourcePacketIn type errorChannel struct { *channel.Endpoint - Ch chan stack.PacketBuffer + Ch chan *stack.PacketBuffer packetCollectorErrors []*tcpip.Error } @@ -184,7 +184,7 @@ type errorChannel struct { func newErrorChannel(size int, mtu uint32, linkAddr tcpip.LinkAddress, packetCollectorErrors []*tcpip.Error) *errorChannel { return &errorChannel{ Endpoint: channel.New(size, mtu, linkAddr), - Ch: make(chan stack.PacketBuffer, size), + Ch: make(chan *stack.PacketBuffer, size), packetCollectorErrors: packetCollectorErrors, } } @@ -203,7 +203,7 @@ func (e *errorChannel) Drain() int { } // WritePacket stores outbound packets into the channel. -func (e *errorChannel) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error { +func (e *errorChannel) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error { select { case e.Ch <- pkt: default: @@ -282,13 +282,17 @@ func TestFragmentation(t *testing.T) { for _, ft := range fragTests { t.Run(ft.description, func(t *testing.T) { hdr, payload := makeHdrAndPayload(ft.hdrLength, ft.extraLength, ft.payloadViewsSizes) - source := stack.PacketBuffer{ + source := &stack.PacketBuffer{ Header: hdr, // Save the source payload because WritePacket will modify it. Data: payload.Clone(nil), } c := buildContext(t, nil, ft.mtu) - err := c.Route.WritePacket(ft.gso, stack.NetworkHeaderParams{Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + err := c.Route.WritePacket(ft.gso, stack.NetworkHeaderParams{ + Protocol: tcp.ProtocolNumber, + TTL: 42, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload, }) @@ -296,7 +300,7 @@ func TestFragmentation(t *testing.T) { t.Errorf("err got %v, want %v", err, nil) } - var results []stack.PacketBuffer + var results []*stack.PacketBuffer L: for { select { @@ -338,7 +342,11 @@ func TestFragmentationErrors(t *testing.T) { t.Run(ft.description, func(t *testing.T) { hdr, payload := makeHdrAndPayload(ft.hdrLength, header.IPv4MinimumSize, ft.payloadViewsSizes) c := buildContext(t, ft.packetCollectorErrors, ft.mtu) - err := c.Route.WritePacket(&stack.GSO{}, stack.NetworkHeaderParams{Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + err := c.Route.WritePacket(&stack.GSO{}, stack.NetworkHeaderParams{ + Protocol: tcp.ProtocolNumber, + TTL: 42, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload, }) @@ -460,7 +468,7 @@ func TestInvalidFragments(t *testing.T) { s.CreateNIC(nicID, sniffer.New(ep)) for _, pkt := range tc.packets { - ep.InjectLinkAddr(header.IPv4ProtocolNumber, remoteLinkAddr, stack.PacketBuffer{ + ep.InjectLinkAddr(header.IPv4ProtocolNumber, remoteLinkAddr, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(len(pkt), []buffer.View{pkt}), }) } @@ -698,7 +706,7 @@ func TestReceiveFragments(t *testing.T) { vv := hdr.View().ToVectorisedView() vv.AppendView(frag.payload) - e.InjectInbound(header.IPv4ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(header.IPv4ProtocolNumber, &stack.PacketBuffer{ Data: vv, }) } diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index bdf3a0d25..b62fb1de6 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/pkg/tcpip/network/ipv6/icmp.go @@ -27,7 +27,7 @@ import ( // the original packet that caused the ICMP one to be sent. This information is // used to find out which transport endpoint must be notified about the ICMP // packet. -func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { h, ok := pkt.Data.PullUp(header.IPv6MinimumSize) if !ok { return @@ -70,7 +70,7 @@ func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack. e.dispatcher.DeliverTransportControlPacket(e.id.LocalAddress, hdr.DestinationAddress(), ProtocolNumber, p, typ, extra, pkt) } -func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.PacketBuffer, hasFragmentHeader bool) { +func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt *stack.PacketBuffer, hasFragmentHeader bool) { stats := r.Stats().ICMP sent := stats.V6PacketsSent received := stats.V6PacketsReceived @@ -288,7 +288,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P // // The IP Hop Limit field has a value of 255, i.e., the packet // could not possibly have been forwarded by a router. - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, }); err != nil { sent.Dropped.Increment() @@ -390,7 +390,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P copy(packet, icmpHdr) packet.SetType(header.ICMPv6EchoReply) packet.SetChecksum(header.ICMPv6Checksum(packet, r.LocalAddress, r.RemoteAddress, pkt.Data)) - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: pkt.Data, }); err != nil { @@ -532,7 +532,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. }) // TODO(stijlist): count this in ICMP stats. - return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, stack.PacketBuffer{ + return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ Header: hdr, }) } diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go index d412ff688..a720f626f 100644 --- a/pkg/tcpip/network/ipv6/icmp_test.go +++ b/pkg/tcpip/network/ipv6/icmp_test.go @@ -57,7 +57,7 @@ func (*stubLinkEndpoint) LinkAddress() tcpip.LinkAddress { return "" } -func (*stubLinkEndpoint) WritePacket(*stack.Route, *stack.GSO, tcpip.NetworkProtocolNumber, stack.PacketBuffer) *tcpip.Error { +func (*stubLinkEndpoint) WritePacket(*stack.Route, *stack.GSO, tcpip.NetworkProtocolNumber, *stack.PacketBuffer) *tcpip.Error { return nil } @@ -67,7 +67,7 @@ type stubDispatcher struct { stack.TransportDispatcher } -func (*stubDispatcher) DeliverTransportPacket(*stack.Route, tcpip.TransportProtocolNumber, stack.PacketBuffer) { +func (*stubDispatcher) DeliverTransportPacket(*stack.Route, tcpip.TransportProtocolNumber, *stack.PacketBuffer) { } type stubLinkAddressCache struct { @@ -189,7 +189,7 @@ func TestICMPCounts(t *testing.T) { SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) - ep.HandlePacket(&r, stack.PacketBuffer{ + ep.HandlePacket(&r, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) } @@ -328,7 +328,7 @@ func routeICMPv6Packet(t *testing.T, args routeArgs, fn func(*testing.T, header. views := []buffer.View{pi.Pkt.Header.View(), pi.Pkt.Data.ToView()} size := pi.Pkt.Header.UsedLength() + pi.Pkt.Data.Size() vv := buffer.NewVectorisedView(size, views) - args.dst.InjectLinkAddr(pi.Proto, args.dst.LinkAddress(), stack.PacketBuffer{ + args.dst.InjectLinkAddr(pi.Proto, args.dst.LinkAddress(), &stack.PacketBuffer{ Data: vv, }) } @@ -563,7 +563,7 @@ func TestICMPChecksumValidationSimple(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) } @@ -740,7 +740,7 @@ func TestICMPChecksumValidationWithPayload(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) } @@ -918,7 +918,7 @@ func TestICMPChecksumValidationWithPayloadMultipleViews(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(header.IPv6MinimumSize+size+payloadSize, []buffer.View{hdr.View(), payload}), }) } diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go index daf1fcbc6..0d94ad122 100644 --- a/pkg/tcpip/network/ipv6/ipv6.go +++ b/pkg/tcpip/network/ipv6/ipv6.go @@ -116,7 +116,7 @@ func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadS } // WritePacket writes a packet to the given destination address and protocol. -func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) pkt.NetworkHeader = buffer.View(ip) @@ -128,7 +128,7 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw views = append(views, pkt.Data.Views()...) loopedR := r.MakeLoopedRoute() - e.HandlePacket(&loopedR, stack.PacketBuffer{ + e.HandlePacket(&loopedR, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), }) @@ -163,14 +163,14 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe // WriteHeaderIncludedPacker implements stack.NetworkEndpoint. It is not yet // supported by IPv6. -func (*endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (*endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { // TODO(b/146666412): Support IPv6 header-included packets. return tcpip.ErrNotSupported } // HandlePacket is called by the link layer when new ipv6 packets arrive for // this endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { headerView, ok := pkt.Data.PullUp(header.IPv6MinimumSize) if !ok { r.Stats().IP.MalformedPacketsReceived.Increment() diff --git a/pkg/tcpip/network/ipv6/ipv6_test.go b/pkg/tcpip/network/ipv6/ipv6_test.go index 841a0cb7a..213ff64f2 100644 --- a/pkg/tcpip/network/ipv6/ipv6_test.go +++ b/pkg/tcpip/network/ipv6/ipv6_test.go @@ -65,7 +65,7 @@ func testReceiveICMP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst DstAddr: dst, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -123,7 +123,7 @@ func testReceiveUDP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst DstAddr: dst, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -637,7 +637,7 @@ func TestReceiveIPv6ExtHdrs(t *testing.T) { DstAddr: addr2, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -1238,7 +1238,7 @@ func TestReceiveIPv6Fragments(t *testing.T) { vv := hdr.View().ToVectorisedView() vv.Append(f.data) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: vv, }) } diff --git a/pkg/tcpip/network/ipv6/ndp_test.go b/pkg/tcpip/network/ipv6/ndp_test.go index 12b70f7e9..3c141b91b 100644 --- a/pkg/tcpip/network/ipv6/ndp_test.go +++ b/pkg/tcpip/network/ipv6/ndp_test.go @@ -136,7 +136,7 @@ func TestNeighorSolicitationWithSourceLinkLayerOption(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -380,7 +380,7 @@ func TestNeighorSolicitationResponse(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, stack.PacketBuffer{ + e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -497,7 +497,7 @@ func TestNeighorAdvertisementWithTargetLinkLayerOption(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -568,7 +568,7 @@ func TestNDPValidation(t *testing.T) { SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) - ep.HandlePacket(r, stack.PacketBuffer{ + ep.HandlePacket(r, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) } @@ -884,7 +884,7 @@ func TestRouterAdvertValidation(t *testing.T) { t.Fatalf("got rxRA = %d, want = 0", got) } - e.InjectInbound(header.IPv6ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(header.IPv6ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) diff --git a/pkg/tcpip/stack/conntrack.go b/pkg/tcpip/stack/conntrack.go index 7d1ede1f2..d4053be08 100644 --- a/pkg/tcpip/stack/conntrack.go +++ b/pkg/tcpip/stack/conntrack.go @@ -186,7 +186,7 @@ func parseHeaders(pkt *PacketBuffer) { } // packetToTuple converts packet to a tuple in original direction. -func packetToTuple(pkt PacketBuffer, hook Hook) (connTrackTuple, *tcpip.Error) { +func packetToTuple(pkt *PacketBuffer, hook Hook) (connTrackTuple, *tcpip.Error) { var tuple connTrackTuple netHeader := header.IPv4(pkt.NetworkHeader) @@ -265,7 +265,7 @@ func (ct *ConnTrackTable) connTrackForPacket(pkt *PacketBuffer, hook Hook, creat } var dir ctDirection - tuple, err := packetToTuple(*pkt, hook) + tuple, err := packetToTuple(pkt, hook) if err != nil { return nil, dir } diff --git a/pkg/tcpip/stack/forwarder.go b/pkg/tcpip/stack/forwarder.go index 6b64cd37f..3eff141e6 100644 --- a/pkg/tcpip/stack/forwarder.go +++ b/pkg/tcpip/stack/forwarder.go @@ -32,7 +32,7 @@ type pendingPacket struct { nic *NIC route *Route proto tcpip.NetworkProtocolNumber - pkt PacketBuffer + pkt *PacketBuffer } type forwardQueue struct { @@ -50,7 +50,7 @@ func newForwardQueue() *forwardQueue { return &forwardQueue{packets: make(map[<-chan struct{}][]*pendingPacket)} } -func (f *forwardQueue) enqueue(ch <-chan struct{}, n *NIC, r *Route, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) { +func (f *forwardQueue) enqueue(ch <-chan struct{}, n *NIC, r *Route, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) { shouldWait := false f.Lock() diff --git a/pkg/tcpip/stack/forwarder_test.go b/pkg/tcpip/stack/forwarder_test.go index 8084d50bc..63537aaad 100644 --- a/pkg/tcpip/stack/forwarder_test.go +++ b/pkg/tcpip/stack/forwarder_test.go @@ -68,7 +68,7 @@ func (f *fwdTestNetworkEndpoint) ID() *NetworkEndpointID { return &f.id } -func (f *fwdTestNetworkEndpoint) HandlePacket(r *Route, pkt PacketBuffer) { +func (f *fwdTestNetworkEndpoint) HandlePacket(r *Route, pkt *PacketBuffer) { // Consume the network header. b, ok := pkt.Data.PullUp(fwdTestNetHeaderLen) if !ok { @@ -96,7 +96,7 @@ func (f *fwdTestNetworkEndpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNu return f.proto.Number() } -func (f *fwdTestNetworkEndpoint) WritePacket(r *Route, gso *GSO, params NetworkHeaderParams, pkt PacketBuffer) *tcpip.Error { +func (f *fwdTestNetworkEndpoint) WritePacket(r *Route, gso *GSO, params NetworkHeaderParams, pkt *PacketBuffer) *tcpip.Error { // Add the protocol's header to the packet and send it to the link // endpoint. b := pkt.Header.Prepend(fwdTestNetHeaderLen) @@ -112,7 +112,7 @@ func (f *fwdTestNetworkEndpoint) WritePackets(r *Route, gso *GSO, pkts PacketBuf panic("not implemented") } -func (*fwdTestNetworkEndpoint) WriteHeaderIncludedPacket(r *Route, pkt PacketBuffer) *tcpip.Error { +func (*fwdTestNetworkEndpoint) WriteHeaderIncludedPacket(r *Route, pkt *PacketBuffer) *tcpip.Error { return tcpip.ErrNotSupported } @@ -190,7 +190,7 @@ func (f *fwdTestNetworkProtocol) LinkAddressProtocol() tcpip.NetworkProtocolNumb type fwdTestPacketInfo struct { RemoteLinkAddress tcpip.LinkAddress LocalLinkAddress tcpip.LinkAddress - Pkt PacketBuffer + Pkt *PacketBuffer } type fwdTestLinkEndpoint struct { @@ -203,13 +203,13 @@ type fwdTestLinkEndpoint struct { } // InjectInbound injects an inbound packet. -func (e *fwdTestLinkEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) { +func (e *fwdTestLinkEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) { e.InjectLinkAddr(protocol, "", pkt) } // InjectLinkAddr injects an inbound packet with a remote link address. -func (e *fwdTestLinkEndpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remote tcpip.LinkAddress, pkt PacketBuffer) { - e.dispatcher.DeliverNetworkPacket(e, remote, "" /* local */, protocol, pkt) +func (e *fwdTestLinkEndpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remote tcpip.LinkAddress, pkt *PacketBuffer) { + e.dispatcher.DeliverNetworkPacket(remote, "" /* local */, protocol, pkt) } // Attach saves the stack network-layer dispatcher for use later when packets @@ -251,7 +251,7 @@ func (e *fwdTestLinkEndpoint) LinkAddress() tcpip.LinkAddress { return e.linkAddr } -func (e fwdTestLinkEndpoint) WritePacket(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) *tcpip.Error { +func (e fwdTestLinkEndpoint) WritePacket(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) *tcpip.Error { p := fwdTestPacketInfo{ RemoteLinkAddress: r.RemoteLinkAddress, LocalLinkAddress: r.LocalLinkAddress, @@ -270,7 +270,7 @@ func (e fwdTestLinkEndpoint) WritePacket(r *Route, gso *GSO, protocol tcpip.Netw func (e *fwdTestLinkEndpoint) WritePackets(r *Route, gso *GSO, pkts PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) { n := 0 for pkt := pkts.Front(); pkt != nil; pkt = pkt.Next() { - e.WritePacket(r, gso, protocol, *pkt) + e.WritePacket(r, gso, protocol, pkt) n++ } @@ -280,7 +280,7 @@ func (e *fwdTestLinkEndpoint) WritePackets(r *Route, gso *GSO, pkts PacketBuffer // WriteRawPacket implements stack.LinkEndpoint.WriteRawPacket. func (e *fwdTestLinkEndpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error { p := fwdTestPacketInfo{ - Pkt: PacketBuffer{Data: vv}, + Pkt: &PacketBuffer{Data: vv}, } select { @@ -362,7 +362,7 @@ func TestForwardingWithStaticResolver(t *testing.T) { // forwarded to NIC 2. buf := buffer.NewView(30) buf[0] = 3 - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) @@ -399,7 +399,7 @@ func TestForwardingWithFakeResolver(t *testing.T) { // forwarded to NIC 2. buf := buffer.NewView(30) buf[0] = 3 - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) @@ -430,7 +430,7 @@ func TestForwardingWithNoResolver(t *testing.T) { // forwarded to NIC 2. buf := buffer.NewView(30) buf[0] = 3 - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) @@ -460,7 +460,7 @@ func TestForwardingWithFakeResolverPartialTimeout(t *testing.T) { // not be forwarded. buf := buffer.NewView(30) buf[0] = 4 - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) @@ -468,7 +468,7 @@ func TestForwardingWithFakeResolverPartialTimeout(t *testing.T) { // forwarded to NIC 2. buf = buffer.NewView(30) buf[0] = 3 - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) @@ -510,7 +510,7 @@ func TestForwardingWithFakeResolverTwoPackets(t *testing.T) { for i := 0; i < 2; i++ { buf := buffer.NewView(30) buf[0] = 3 - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) } @@ -557,7 +557,7 @@ func TestForwardingWithFakeResolverManyPackets(t *testing.T) { buf[0] = 3 // Set the packet sequence number. binary.BigEndian.PutUint16(buf[fwdTestNetHeaderLen:], uint16(i)) - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) } @@ -610,7 +610,7 @@ func TestForwardingWithFakeResolverManyResolutions(t *testing.T) { // maxPendingResolutions + 7). buf := buffer.NewView(30) buf[0] = byte(3 + i) - ep1.InjectInbound(fwdTestNetNumber, PacketBuffer{ + ep1.InjectInbound(fwdTestNetNumber, &PacketBuffer{ Data: buf.ToVectorisedView(), }) } diff --git a/pkg/tcpip/stack/iptables.go b/pkg/tcpip/stack/iptables.go index 443423b3c..4e9b404c8 100644 --- a/pkg/tcpip/stack/iptables.go +++ b/pkg/tcpip/stack/iptables.go @@ -16,7 +16,6 @@ package stack import ( "fmt" - "strings" "gvisor.dev/gvisor/pkg/tcpip" "gvisor.dev/gvisor/pkg/tcpip/header" @@ -44,11 +43,11 @@ const HookUnset = -1 // DefaultTables returns a default set of tables. Each chain is set to accept // all packets. -func DefaultTables() IPTables { +func DefaultTables() *IPTables { // TODO(gvisor.dev/issue/170): We may be able to swap out some strings for // iotas. - return IPTables{ - Tables: map[string]Table{ + return &IPTables{ + tables: map[string]Table{ TablenameNat: Table{ Rules: []Rule{ Rule{Target: AcceptTarget{}}, @@ -107,7 +106,7 @@ func DefaultTables() IPTables { UserChains: map[string]int{}, }, }, - Priorities: map[Hook][]string{ + priorities: map[Hook][]string{ Input: []string{TablenameNat, TablenameFilter}, Prerouting: []string{TablenameMangle, TablenameNat}, Output: []string{TablenameMangle, TablenameNat, TablenameFilter}, @@ -159,6 +158,36 @@ func EmptyNatTable() Table { } } +// GetTable returns table by name. +func (it *IPTables) GetTable(name string) (Table, bool) { + it.mu.RLock() + defer it.mu.RUnlock() + t, ok := it.tables[name] + return t, ok +} + +// ReplaceTable replaces or inserts table by name. +func (it *IPTables) ReplaceTable(name string, table Table) { + it.mu.Lock() + defer it.mu.Unlock() + it.tables[name] = table +} + +// ModifyTables acquires write-lock and calls fn with internal name-to-table +// map. This function can be used to update multiple tables atomically. +func (it *IPTables) ModifyTables(fn func(map[string]Table)) { + it.mu.Lock() + defer it.mu.Unlock() + fn(it.tables) +} + +// GetPriorities returns slice of priorities associated with hook. +func (it *IPTables) GetPriorities(hook Hook) []string { + it.mu.RLock() + defer it.mu.RUnlock() + return it.priorities[hook] +} + // A chainVerdict is what a table decides should be done with a packet. type chainVerdict int @@ -185,8 +214,8 @@ func (it *IPTables) Check(hook Hook, pkt *PacketBuffer, gso *GSO, r *Route, addr it.connections.HandlePacket(pkt, hook, gso, r) // Go through each table containing the hook. - for _, tablename := range it.Priorities[hook] { - table := it.Tables[tablename] + for _, tablename := range it.GetPriorities(hook) { + table, _ := it.GetTable(tablename) ruleIdx := table.BuiltinChains[hook] switch verdict := it.checkChain(hook, pkt, table, ruleIdx, gso, r, address, nicName); verdict { // If the table returns Accept, move on to the next table. @@ -314,7 +343,7 @@ func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx } // Check whether the packet matches the IP header filter. - if !filterMatch(rule.Filter, header.IPv4(pkt.NetworkHeader), hook, nicName) { + if !rule.Filter.match(header.IPv4(pkt.NetworkHeader), hook, nicName) { // Continue on to the next rule. return RuleJump, ruleIdx + 1 } @@ -322,7 +351,7 @@ func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx // Go through each rule matcher. If they all match, run // the rule target. for _, matcher := range rule.Matchers { - matches, hotdrop := matcher.Match(hook, *pkt, "") + matches, hotdrop := matcher.Match(hook, pkt, "") if hotdrop { return RuleDrop, 0 } @@ -335,47 +364,3 @@ func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx // All the matchers matched, so run the target. return rule.Target.Action(pkt, &it.connections, hook, gso, r, address) } - -func filterMatch(filter IPHeaderFilter, hdr header.IPv4, hook Hook, nicName string) bool { - // TODO(gvisor.dev/issue/170): Support other fields of the filter. - // Check the transport protocol. - if filter.Protocol != 0 && filter.Protocol != hdr.TransportProtocol() { - return false - } - - // Check the destination IP. - dest := hdr.DestinationAddress() - matches := true - for i := range filter.Dst { - if dest[i]&filter.DstMask[i] != filter.Dst[i] { - matches = false - break - } - } - if matches == filter.DstInvert { - return false - } - - // Check the output interface. - // TODO(gvisor.dev/issue/170): Add the check for FORWARD and POSTROUTING - // hooks after supported. - if hook == Output { - n := len(filter.OutputInterface) - if n == 0 { - return true - } - - // If the interface name ends with '+', any interface which begins - // with the name should be matched. - ifName := filter.OutputInterface - matches = true - if strings.HasSuffix(ifName, "+") { - matches = strings.HasPrefix(nicName, ifName[:n-1]) - } else { - matches = nicName == ifName - } - return filter.OutputInterfaceInvert != matches - } - - return true -} diff --git a/pkg/tcpip/stack/iptables_types.go b/pkg/tcpip/stack/iptables_types.go index fe06007ae..4a6a5c6f1 100644 --- a/pkg/tcpip/stack/iptables_types.go +++ b/pkg/tcpip/stack/iptables_types.go @@ -15,7 +15,11 @@ package stack import ( + "strings" + "sync" + "gvisor.dev/gvisor/pkg/tcpip" + "gvisor.dev/gvisor/pkg/tcpip/header" ) // A Hook specifies one of the hooks built into the network stack. @@ -75,13 +79,17 @@ const ( // IPTables holds all the tables for a netstack. type IPTables struct { - // Tables maps table names to tables. User tables have arbitrary names. - Tables map[string]Table + // mu protects tables and priorities. + mu sync.RWMutex + + // tables maps table names to tables. User tables have arbitrary names. mu + // needs to be locked for accessing. + tables map[string]Table - // Priorities maps each hook to a list of table names. The order of the + // priorities maps each hook to a list of table names. The order of the // list is the order in which each table should be visited for that - // hook. - Priorities map[Hook][]string + // hook. mu needs to be locked for accessing. + priorities map[Hook][]string connections ConnTrackTable } @@ -159,6 +167,16 @@ type IPHeaderFilter struct { // comparison. DstInvert bool + // Src matches the source IP address. + Src tcpip.Address + + // SrcMask masks bits of the source IP address when comparing with Src. + SrcMask tcpip.Address + + // SrcInvert inverts the meaning of the source IP check, i.e. when true the + // filter will match packets that fail the source comparison. + SrcInvert bool + // OutputInterface matches the name of the outgoing interface for the // packet. OutputInterface string @@ -173,6 +191,55 @@ type IPHeaderFilter struct { OutputInterfaceInvert bool } +// match returns whether hdr matches the filter. +func (fl IPHeaderFilter) match(hdr header.IPv4, hook Hook, nicName string) bool { + // TODO(gvisor.dev/issue/170): Support other fields of the filter. + // Check the transport protocol. + if fl.Protocol != 0 && fl.Protocol != hdr.TransportProtocol() { + return false + } + + // Check the source and destination IPs. + if !filterAddress(hdr.DestinationAddress(), fl.DstMask, fl.Dst, fl.DstInvert) || !filterAddress(hdr.SourceAddress(), fl.SrcMask, fl.Src, fl.SrcInvert) { + return false + } + + // Check the output interface. + // TODO(gvisor.dev/issue/170): Add the check for FORWARD and POSTROUTING + // hooks after supported. + if hook == Output { + n := len(fl.OutputInterface) + if n == 0 { + return true + } + + // If the interface name ends with '+', any interface which begins + // with the name should be matched. + ifName := fl.OutputInterface + matches := true + if strings.HasSuffix(ifName, "+") { + matches = strings.HasPrefix(nicName, ifName[:n-1]) + } else { + matches = nicName == ifName + } + return fl.OutputInterfaceInvert != matches + } + + return true +} + +// filterAddress returns whether addr matches the filter. +func filterAddress(addr, mask, filterAddr tcpip.Address, invert bool) bool { + matches := true + for i := range filterAddr { + if addr[i]&mask[i] != filterAddr[i] { + matches = false + break + } + } + return matches != invert +} + // A Matcher is the interface for matching packets. type Matcher interface { // Name returns the name of the Matcher. @@ -183,7 +250,7 @@ type Matcher interface { // used for suspicious packets. // // Precondition: packet.NetworkHeader is set. - Match(hook Hook, packet PacketBuffer, interfaceName string) (matches bool, hotdrop bool) + Match(hook Hook, packet *PacketBuffer, interfaceName string) (matches bool, hotdrop bool) } // A Target is the interface for taking an action for a packet. diff --git a/pkg/tcpip/stack/ndp.go b/pkg/tcpip/stack/ndp.go index 526c7d6ff..ae7a8f740 100644 --- a/pkg/tcpip/stack/ndp.go +++ b/pkg/tcpip/stack/ndp.go @@ -750,7 +750,7 @@ func (ndp *ndpState) sendDADPacket(addr tcpip.Address) *tcpip.Error { Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: DefaultTOS, - }, PacketBuffer{Header: hdr}, + }, &PacketBuffer{Header: hdr}, ); err != nil { sent.Dropped.Increment() return err @@ -1881,7 +1881,7 @@ func (ndp *ndpState) startSolicitingRouters() { Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: DefaultTOS, - }, PacketBuffer{Header: hdr}, + }, &PacketBuffer{Header: hdr}, ); err != nil { sent.Dropped.Increment() log.Printf("startSolicitingRouters: error writing NDP router solicit message on NIC(%d); err = %s", ndp.nic.ID(), err) diff --git a/pkg/tcpip/stack/ndp_test.go b/pkg/tcpip/stack/ndp_test.go index b3d174cdd..58f1ebf60 100644 --- a/pkg/tcpip/stack/ndp_test.go +++ b/pkg/tcpip/stack/ndp_test.go @@ -613,7 +613,7 @@ func TestDADFail(t *testing.T) { // Receive a packet to simulate multiple nodes owning or // attempting to own the same address. hdr := test.makeBuf(addr1) - e.InjectInbound(header.IPv6ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(header.IPv6ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -935,7 +935,7 @@ func TestSetNDPConfigurations(t *testing.T) { // raBufWithOptsAndDHCPv6 returns a valid NDP Router Advertisement with options // and DHCPv6 configurations specified. -func raBufWithOptsAndDHCPv6(ip tcpip.Address, rl uint16, managedAddress, otherConfigurations bool, optSer header.NDPOptionsSerializer) stack.PacketBuffer { +func raBufWithOptsAndDHCPv6(ip tcpip.Address, rl uint16, managedAddress, otherConfigurations bool, optSer header.NDPOptionsSerializer) *stack.PacketBuffer { icmpSize := header.ICMPv6HeaderSize + header.NDPRAMinimumSize + int(optSer.Length()) hdr := buffer.NewPrependable(header.IPv6MinimumSize + icmpSize) pkt := header.ICMPv6(hdr.Prepend(icmpSize)) @@ -970,14 +970,14 @@ func raBufWithOptsAndDHCPv6(ip tcpip.Address, rl uint16, managedAddress, otherCo DstAddr: header.IPv6AllNodesMulticastAddress, }) - return stack.PacketBuffer{Data: hdr.View().ToVectorisedView()} + return &stack.PacketBuffer{Data: hdr.View().ToVectorisedView()} } // raBufWithOpts returns a valid NDP Router Advertisement with options. // // Note, raBufWithOpts does not populate any of the RA fields other than the // Router Lifetime. -func raBufWithOpts(ip tcpip.Address, rl uint16, optSer header.NDPOptionsSerializer) stack.PacketBuffer { +func raBufWithOpts(ip tcpip.Address, rl uint16, optSer header.NDPOptionsSerializer) *stack.PacketBuffer { return raBufWithOptsAndDHCPv6(ip, rl, false, false, optSer) } @@ -986,7 +986,7 @@ func raBufWithOpts(ip tcpip.Address, rl uint16, optSer header.NDPOptionsSerializ // // Note, raBufWithDHCPv6 does not populate any of the RA fields other than the // DHCPv6 related ones. -func raBufWithDHCPv6(ip tcpip.Address, managedAddresses, otherConfiguratiosns bool) stack.PacketBuffer { +func raBufWithDHCPv6(ip tcpip.Address, managedAddresses, otherConfiguratiosns bool) *stack.PacketBuffer { return raBufWithOptsAndDHCPv6(ip, 0, managedAddresses, otherConfiguratiosns, header.NDPOptionsSerializer{}) } @@ -994,7 +994,7 @@ func raBufWithDHCPv6(ip tcpip.Address, managedAddresses, otherConfiguratiosns bo // // Note, raBuf does not populate any of the RA fields other than the // Router Lifetime. -func raBuf(ip tcpip.Address, rl uint16) stack.PacketBuffer { +func raBuf(ip tcpip.Address, rl uint16) *stack.PacketBuffer { return raBufWithOpts(ip, rl, header.NDPOptionsSerializer{}) } @@ -1003,7 +1003,7 @@ func raBuf(ip tcpip.Address, rl uint16) stack.PacketBuffer { // // Note, raBufWithPI does not populate any of the RA fields other than the // Router Lifetime. -func raBufWithPI(ip tcpip.Address, rl uint16, prefix tcpip.AddressWithPrefix, onLink, auto bool, vl, pl uint32) stack.PacketBuffer { +func raBufWithPI(ip tcpip.Address, rl uint16, prefix tcpip.AddressWithPrefix, onLink, auto bool, vl, pl uint32) *stack.PacketBuffer { flags := uint8(0) if onLink { // The OnLink flag is the 7th bit in the flags byte. diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go index 54103fdb3..6664aea06 100644 --- a/pkg/tcpip/stack/nic.go +++ b/pkg/tcpip/stack/nic.go @@ -1153,7 +1153,7 @@ func (n *NIC) isInGroup(addr tcpip.Address) bool { return joins != 0 } -func handlePacket(protocol tcpip.NetworkProtocolNumber, dst, src tcpip.Address, localLinkAddr, remotelinkAddr tcpip.LinkAddress, ref *referencedNetworkEndpoint, pkt PacketBuffer) { +func handlePacket(protocol tcpip.NetworkProtocolNumber, dst, src tcpip.Address, localLinkAddr, remotelinkAddr tcpip.LinkAddress, ref *referencedNetworkEndpoint, pkt *PacketBuffer) { r := makeRoute(protocol, dst, src, localLinkAddr, ref, false /* handleLocal */, false /* multicastLoop */) r.RemoteLinkAddress = remotelinkAddr @@ -1167,7 +1167,7 @@ func handlePacket(protocol tcpip.NetworkProtocolNumber, dst, src tcpip.Address, // Note that the ownership of the slice backing vv is retained by the caller. // This rule applies only to the slice itself, not to the items of the slice; // the ownership of the items is not retained by the caller. -func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) { +func (n *NIC) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) { n.mu.RLock() enabled := n.mu.enabled // If the NIC is not yet enabled, don't receive any packets. @@ -1229,18 +1229,19 @@ func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.Link } // TODO(gvisor.dev/issue/170): Not supporting iptables for IPv6 yet. - if protocol == header.IPv4ProtocolNumber { + // Loopback traffic skips the prerouting chain. + if protocol == header.IPv4ProtocolNumber && !n.isLoopback() { // iptables filtering. ipt := n.stack.IPTables() address := n.primaryAddress(protocol) - if ok := ipt.Check(Prerouting, &pkt, nil, nil, address.Address, ""); !ok { + if ok := ipt.Check(Prerouting, pkt, nil, nil, address.Address, ""); !ok { // iptables is telling us to drop the packet. return } } if ref := n.getRef(protocol, dst); ref != nil { - handlePacket(protocol, dst, src, linkEP.LinkAddress(), remote, ref, pkt) + handlePacket(protocol, dst, src, n.linkEP.LinkAddress(), remote, ref, pkt) return } @@ -1298,24 +1299,27 @@ func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.Link } } -func (n *NIC) forwardPacket(r *Route, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) { +func (n *NIC) forwardPacket(r *Route, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) { // TODO(b/143425874) Decrease the TTL field in forwarded packets. if linkHeaderLen := int(n.linkEP.MaxHeaderLength()); linkHeaderLen != 0 { pkt.Header = buffer.NewPrependable(linkHeaderLen) } + // WritePacket takes ownership of pkt, calculate numBytes first. + numBytes := pkt.Header.UsedLength() + pkt.Data.Size() + if err := n.linkEP.WritePacket(r, nil /* gso */, protocol, pkt); err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() return } n.stats.Tx.Packets.Increment() - n.stats.Tx.Bytes.IncrementBy(uint64(pkt.Header.UsedLength() + pkt.Data.Size())) + n.stats.Tx.Bytes.IncrementBy(uint64(numBytes)) } // DeliverTransportPacket delivers the packets to the appropriate transport // protocol endpoint. -func (n *NIC) DeliverTransportPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt PacketBuffer) { +func (n *NIC) DeliverTransportPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt *PacketBuffer) { state, ok := n.stack.transportProtocols[protocol] if !ok { n.stack.stats.UnknownProtocolRcvdPackets.Increment() @@ -1362,7 +1366,7 @@ func (n *NIC) DeliverTransportPacket(r *Route, protocol tcpip.TransportProtocolN // DeliverTransportControlPacket delivers control packets to the appropriate // transport protocol endpoint. -func (n *NIC) DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ ControlType, extra uint32, pkt PacketBuffer) { +func (n *NIC) DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ ControlType, extra uint32, pkt *PacketBuffer) { state, ok := n.stack.transportProtocols[trans] if !ok { return diff --git a/pkg/tcpip/stack/nic_test.go b/pkg/tcpip/stack/nic_test.go index d672fc157..fea46158c 100644 --- a/pkg/tcpip/stack/nic_test.go +++ b/pkg/tcpip/stack/nic_test.go @@ -44,7 +44,7 @@ func TestDisabledRxStatsWhenNICDisabled(t *testing.T) { t.FailNow() } - nic.DeliverNetworkPacket(nil, "", "", 0, PacketBuffer{Data: buffer.View([]byte{1, 2, 3, 4}).ToVectorisedView()}) + nic.DeliverNetworkPacket("", "", 0, &PacketBuffer{Data: buffer.View([]byte{1, 2, 3, 4}).ToVectorisedView()}) if got := nic.stats.DisabledRx.Packets.Value(); got != 1 { t.Errorf("got DisabledRx.Packets = %d, want = 1", got) diff --git a/pkg/tcpip/stack/packet_buffer.go b/pkg/tcpip/stack/packet_buffer.go index 926df4d7b..1b5da6017 100644 --- a/pkg/tcpip/stack/packet_buffer.go +++ b/pkg/tcpip/stack/packet_buffer.go @@ -24,6 +24,8 @@ import ( // multiple endpoints. Clone() should be called in such cases so that // modifications to the Data field do not affect other copies. type PacketBuffer struct { + _ noCopy + // PacketBufferEntry is used to build an intrusive list of // PacketBuffers. PacketBufferEntry @@ -82,7 +84,32 @@ type PacketBuffer struct { // VectorisedView but does not deep copy the underlying bytes. // // Clone also does not deep copy any of its other fields. -func (pk PacketBuffer) Clone() PacketBuffer { - pk.Data = pk.Data.Clone(nil) - return pk +// +// FIXME(b/153685824): Data gets copied but not other header references. +func (pk *PacketBuffer) Clone() *PacketBuffer { + return &PacketBuffer{ + PacketBufferEntry: pk.PacketBufferEntry, + Data: pk.Data.Clone(nil), + Header: pk.Header, + LinkHeader: pk.LinkHeader, + NetworkHeader: pk.NetworkHeader, + TransportHeader: pk.TransportHeader, + Hash: pk.Hash, + Owner: pk.Owner, + EgressRoute: pk.EgressRoute, + GSOOptions: pk.GSOOptions, + NetworkProtocolNumber: pk.NetworkProtocolNumber, + NatDone: pk.NatDone, + } } + +// noCopy may be embedded into structs which must not be copied +// after the first use. +// +// See https://golang.org/issues/8005#issuecomment-190753527 +// for details. +type noCopy struct{} + +// Lock is a no-op used by -copylocks checker from `go vet`. +func (*noCopy) Lock() {} +func (*noCopy) Unlock() {} diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go index b331427c6..94f177841 100644 --- a/pkg/tcpip/stack/registration.go +++ b/pkg/tcpip/stack/registration.go @@ -67,12 +67,12 @@ type TransportEndpoint interface { // this transport endpoint. It sets pkt.TransportHeader. // // HandlePacket takes ownership of pkt. - HandlePacket(r *Route, id TransportEndpointID, pkt PacketBuffer) + HandlePacket(r *Route, id TransportEndpointID, pkt *PacketBuffer) // HandleControlPacket is called by the stack when new control (e.g. // ICMP) packets arrive to this transport endpoint. // HandleControlPacket takes ownership of pkt. - HandleControlPacket(id TransportEndpointID, typ ControlType, extra uint32, pkt PacketBuffer) + HandleControlPacket(id TransportEndpointID, typ ControlType, extra uint32, pkt *PacketBuffer) // Abort initiates an expedited endpoint teardown. It puts the endpoint // in a closed state and frees all resources associated with it. This @@ -100,7 +100,7 @@ type RawTransportEndpoint interface { // layer up. // // HandlePacket takes ownership of pkt. - HandlePacket(r *Route, pkt PacketBuffer) + HandlePacket(r *Route, pkt *PacketBuffer) } // PacketEndpoint is the interface that needs to be implemented by packet @@ -118,7 +118,7 @@ type PacketEndpoint interface { // should construct its own ethernet header for applications. // // HandlePacket takes ownership of pkt. - HandlePacket(nicID tcpip.NICID, addr tcpip.LinkAddress, netProto tcpip.NetworkProtocolNumber, pkt PacketBuffer) + HandlePacket(nicID tcpip.NICID, addr tcpip.LinkAddress, netProto tcpip.NetworkProtocolNumber, pkt *PacketBuffer) } // TransportProtocol is the interface that needs to be implemented by transport @@ -150,7 +150,7 @@ type TransportProtocol interface { // stats purposes only). // // HandleUnknownDestinationPacket takes ownership of pkt. - HandleUnknownDestinationPacket(r *Route, id TransportEndpointID, pkt PacketBuffer) bool + HandleUnknownDestinationPacket(r *Route, id TransportEndpointID, pkt *PacketBuffer) bool // SetOption allows enabling/disabling protocol specific features. // SetOption returns an error if the option is not supported or the @@ -180,7 +180,7 @@ type TransportDispatcher interface { // pkt.NetworkHeader must be set before calling DeliverTransportPacket. // // DeliverTransportPacket takes ownership of pkt. - DeliverTransportPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt PacketBuffer) + DeliverTransportPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt *PacketBuffer) // DeliverTransportControlPacket delivers control packets to the // appropriate transport protocol endpoint. @@ -189,7 +189,7 @@ type TransportDispatcher interface { // DeliverTransportControlPacket. // // DeliverTransportControlPacket takes ownership of pkt. - DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ ControlType, extra uint32, pkt PacketBuffer) + DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ ControlType, extra uint32, pkt *PacketBuffer) } // PacketLooping specifies where an outbound packet should be sent. @@ -240,17 +240,18 @@ type NetworkEndpoint interface { MaxHeaderLength() uint16 // WritePacket writes a packet to the given destination address and - // protocol. It sets pkt.NetworkHeader. pkt.TransportHeader must have - // already been set. - WritePacket(r *Route, gso *GSO, params NetworkHeaderParams, pkt PacketBuffer) *tcpip.Error + // protocol. It takes ownership of pkt. pkt.TransportHeader must have already + // been set. + WritePacket(r *Route, gso *GSO, params NetworkHeaderParams, pkt *PacketBuffer) *tcpip.Error // WritePackets writes packets to the given destination address and - // protocol. pkts must not be zero length. + // protocol. pkts must not be zero length. It takes ownership of pkts and + // underlying packets. WritePackets(r *Route, gso *GSO, pkts PacketBufferList, params NetworkHeaderParams) (int, *tcpip.Error) // WriteHeaderIncludedPacket writes a packet that includes a network - // header to the given destination address. - WriteHeaderIncludedPacket(r *Route, pkt PacketBuffer) *tcpip.Error + // header to the given destination address. It takes ownership of pkt. + WriteHeaderIncludedPacket(r *Route, pkt *PacketBuffer) *tcpip.Error // ID returns the network protocol endpoint ID. ID() *NetworkEndpointID @@ -265,7 +266,7 @@ type NetworkEndpoint interface { // this network endpoint. It sets pkt.NetworkHeader. // // HandlePacket takes ownership of pkt. - HandlePacket(r *Route, pkt PacketBuffer) + HandlePacket(r *Route, pkt *PacketBuffer) // Close is called when the endpoint is reomved from a stack. Close() @@ -326,7 +327,7 @@ type NetworkDispatcher interface { // packets sent via loopback), and won't have the field set. // // DeliverNetworkPacket takes ownership of pkt. - DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) + DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) } // LinkEndpointCapabilities is the type associated with the capabilities @@ -382,17 +383,17 @@ type LinkEndpoint interface { LinkAddress() tcpip.LinkAddress // WritePacket writes a packet with the given protocol through the - // given route. It sets pkt.LinkHeader if a link layer header exists. - // pkt.NetworkHeader and pkt.TransportHeader must have already been - // set. + // given route. It takes ownership of pkt. pkt.NetworkHeader and + // pkt.TransportHeader must have already been set. // // To participate in transparent bridging, a LinkEndpoint implementation // should call eth.Encode with header.EthernetFields.SrcAddr set to // r.LocalLinkAddress if it is provided. - WritePacket(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) *tcpip.Error + WritePacket(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) *tcpip.Error // WritePackets writes packets with the given protocol through the - // given route. pkts must not be zero length. + // given route. pkts must not be zero length. It takes ownership of pkts and + // underlying packets. // // Right now, WritePackets is used only when the software segmentation // offload is enabled. If it will be used for something else, it may @@ -400,7 +401,7 @@ type LinkEndpoint interface { WritePackets(r *Route, gso *GSO, pkts PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) // WriteRawPacket writes a packet directly to the link. The packet - // should already have an ethernet header. + // should already have an ethernet header. It takes ownership of vv. WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error // Attach attaches the data link layer endpoint to the network-layer @@ -430,7 +431,7 @@ type InjectableLinkEndpoint interface { LinkEndpoint // InjectInbound injects an inbound packet. - InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) + InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) // InjectOutbound writes a fully formed outbound packet directly to the // link. diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go index 150297ab9..f5b6ca0b9 100644 --- a/pkg/tcpip/stack/route.go +++ b/pkg/tcpip/stack/route.go @@ -153,17 +153,20 @@ func (r *Route) IsResolutionRequired() bool { } // WritePacket writes the packet through the given route. -func (r *Route) WritePacket(gso *GSO, params NetworkHeaderParams, pkt PacketBuffer) *tcpip.Error { +func (r *Route) WritePacket(gso *GSO, params NetworkHeaderParams, pkt *PacketBuffer) *tcpip.Error { if !r.ref.isValidForOutgoing() { return tcpip.ErrInvalidEndpointState } + // WritePacket takes ownership of pkt, calculate numBytes first. + numBytes := pkt.Header.UsedLength() + pkt.Data.Size() + err := r.ref.ep.WritePacket(r, gso, params, pkt) if err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() } else { r.ref.nic.stats.Tx.Packets.Increment() - r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(pkt.Header.UsedLength() + pkt.Data.Size())) + r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(numBytes)) } return err } @@ -175,9 +178,12 @@ func (r *Route) WritePackets(gso *GSO, pkts PacketBufferList, params NetworkHead return 0, tcpip.ErrInvalidEndpointState } + // WritePackets takes ownership of pkt, calculate length first. + numPkts := pkts.Len() + n, err := r.ref.ep.WritePackets(r, gso, pkts, params) if err != nil { - r.Stats().IP.OutgoingPacketErrors.IncrementBy(uint64(pkts.Len() - n)) + r.Stats().IP.OutgoingPacketErrors.IncrementBy(uint64(numPkts - n)) } r.ref.nic.stats.Tx.Packets.IncrementBy(uint64(n)) @@ -193,17 +199,20 @@ func (r *Route) WritePackets(gso *GSO, pkts PacketBufferList, params NetworkHead // WriteHeaderIncludedPacket writes a packet already containing a network // header through the given route. -func (r *Route) WriteHeaderIncludedPacket(pkt PacketBuffer) *tcpip.Error { +func (r *Route) WriteHeaderIncludedPacket(pkt *PacketBuffer) *tcpip.Error { if !r.ref.isValidForOutgoing() { return tcpip.ErrInvalidEndpointState } + // WriteHeaderIncludedPacket takes ownership of pkt, calculate numBytes first. + numBytes := pkt.Data.Size() + if err := r.ref.ep.WriteHeaderIncludedPacket(r, pkt); err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() return err } r.ref.nic.stats.Tx.Packets.Increment() - r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(pkt.Data.Size())) + r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(numBytes)) return nil } diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index 0ab4c3e19..294ce8775 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -52,7 +52,7 @@ const ( type transportProtocolState struct { proto TransportProtocol - defaultHandler func(r *Route, id TransportEndpointID, pkt PacketBuffer) bool + defaultHandler func(r *Route, id TransportEndpointID, pkt *PacketBuffer) bool } // TCPProbeFunc is the expected function type for a TCP probe function to be @@ -424,12 +424,8 @@ type Stack struct { // handleLocal allows non-loopback interfaces to loop packets. handleLocal bool - // tablesMu protects iptables. - tablesMu sync.RWMutex - - // tables are the iptables packet filtering and manipulation rules. The are - // protected by tablesMu.` - tables IPTables + // tables are the iptables packet filtering and manipulation rules. + tables *IPTables // resumableEndpoints is a list of endpoints that need to be resumed if the // stack is being restored. @@ -676,6 +672,7 @@ func New(opts Options) *Stack { clock: clock, stats: opts.Stats.FillIn(), handleLocal: opts.HandleLocal, + tables: DefaultTables(), icmpRateLimiter: NewICMPRateLimiter(), seed: generateRandUint32(), ndpConfigs: opts.NDPConfigs, @@ -778,7 +775,7 @@ func (s *Stack) TransportProtocolOption(transport tcpip.TransportProtocolNumber, // // It must be called only during initialization of the stack. Changing it as the // stack is operating is not supported. -func (s *Stack) SetTransportProtocolHandler(p tcpip.TransportProtocolNumber, h func(*Route, TransportEndpointID, PacketBuffer) bool) { +func (s *Stack) SetTransportProtocolHandler(p tcpip.TransportProtocolNumber, h func(*Route, TransportEndpointID, *PacketBuffer) bool) { state := s.transportProtocols[p] if state != nil { state.defaultHandler = h @@ -1741,18 +1738,8 @@ func (s *Stack) IsInGroup(nicID tcpip.NICID, multicastAddr tcpip.Address) (bool, } // IPTables returns the stack's iptables. -func (s *Stack) IPTables() IPTables { - s.tablesMu.RLock() - t := s.tables - s.tablesMu.RUnlock() - return t -} - -// SetIPTables sets the stack's iptables. -func (s *Stack) SetIPTables(ipt IPTables) { - s.tablesMu.Lock() - s.tables = ipt - s.tablesMu.Unlock() +func (s *Stack) IPTables() *IPTables { + return s.tables } // ICMPLimit returns the maximum number of ICMP messages that can be sent diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go index 1a2cf007c..f6ddc3ced 100644 --- a/pkg/tcpip/stack/stack_test.go +++ b/pkg/tcpip/stack/stack_test.go @@ -90,7 +90,7 @@ func (f *fakeNetworkEndpoint) ID() *stack.NetworkEndpointID { return &f.id } -func (f *fakeNetworkEndpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { +func (f *fakeNetworkEndpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { // Increment the received packet count in the protocol descriptor. f.proto.packetCount[int(f.id.LocalAddress[0])%len(f.proto.packetCount)]++ @@ -132,7 +132,7 @@ func (f *fakeNetworkEndpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNumbe return f.proto.Number() } -func (f *fakeNetworkEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt stack.PacketBuffer) *tcpip.Error { +func (f *fakeNetworkEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { // Increment the sent packet count in the protocol descriptor. f.proto.sendPacketCount[int(r.RemoteAddress[0])%len(f.proto.sendPacketCount)]++ @@ -147,7 +147,7 @@ func (f *fakeNetworkEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, params views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) views[0] = pkt.Header.View() views = append(views, pkt.Data.Views()...) - f.HandlePacket(r, stack.PacketBuffer{ + f.HandlePacket(r, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), }) } @@ -163,7 +163,7 @@ func (f *fakeNetworkEndpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts panic("not implemented") } -func (*fakeNetworkEndpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (*fakeNetworkEndpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { return tcpip.ErrNotSupported } @@ -293,7 +293,7 @@ func TestNetworkReceive(t *testing.T) { // Make sure packet with wrong address is not delivered. buf[0] = 3 - ep.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeNet.packetCount[1] != 0 { @@ -305,7 +305,7 @@ func TestNetworkReceive(t *testing.T) { // Make sure packet is delivered to first endpoint. buf[0] = 1 - ep.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeNet.packetCount[1] != 1 { @@ -317,7 +317,7 @@ func TestNetworkReceive(t *testing.T) { // Make sure packet is delivered to second endpoint. buf[0] = 2 - ep.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeNet.packetCount[1] != 1 { @@ -328,7 +328,7 @@ func TestNetworkReceive(t *testing.T) { } // Make sure packet is not delivered if protocol number is wrong. - ep.InjectInbound(fakeNetNumber-1, stack.PacketBuffer{ + ep.InjectInbound(fakeNetNumber-1, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeNet.packetCount[1] != 1 { @@ -340,7 +340,7 @@ func TestNetworkReceive(t *testing.T) { // Make sure packet that is too small is dropped. buf.CapLength(2) - ep.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeNet.packetCount[1] != 1 { @@ -362,7 +362,7 @@ func sendTo(s *stack.Stack, addr tcpip.Address, payload buffer.View) *tcpip.Erro func send(r stack.Route, payload buffer.View) *tcpip.Error { hdr := buffer.NewPrependable(int(r.MaxHeaderLength())) - return r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: fakeTransNumber, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + return r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: fakeTransNumber, TTL: 123, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: payload.ToVectorisedView(), }) @@ -420,7 +420,7 @@ func testFailingRecv(t *testing.T, fakeNet *fakeNetworkProtocol, localAddrByte b func testRecvInternal(t *testing.T, fakeNet *fakeNetworkProtocol, localAddrByte byte, ep *channel.Endpoint, buf buffer.View, want int) { t.Helper() - ep.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if got := fakeNet.PacketCount(localAddrByte); got != want { @@ -2263,7 +2263,7 @@ func TestNICStats(t *testing.T) { // Send a packet to address 1. buf := buffer.NewView(30) - ep1.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep1.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if got, want := s.NICInfo()[1].Stats.Rx.Packets.Value(), uint64(1); got != want { @@ -2345,7 +2345,7 @@ func TestNICForwarding(t *testing.T) { // Send a packet to dstAddr. buf := buffer.NewView(30) buf[0] = dstAddr[0] - ep1.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep1.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) diff --git a/pkg/tcpip/stack/transport_demuxer.go b/pkg/tcpip/stack/transport_demuxer.go index 9a33ed375..e09866405 100644 --- a/pkg/tcpip/stack/transport_demuxer.go +++ b/pkg/tcpip/stack/transport_demuxer.go @@ -152,7 +152,7 @@ func (epsByNIC *endpointsByNIC) transportEndpoints() []TransportEndpoint { // HandlePacket is called by the stack when new packets arrive to this transport // endpoint. -func (epsByNIC *endpointsByNIC) handlePacket(r *Route, id TransportEndpointID, pkt PacketBuffer) { +func (epsByNIC *endpointsByNIC) handlePacket(r *Route, id TransportEndpointID, pkt *PacketBuffer) { epsByNIC.mu.RLock() mpep, ok := epsByNIC.endpoints[r.ref.nic.ID()] @@ -183,7 +183,7 @@ func (epsByNIC *endpointsByNIC) handlePacket(r *Route, id TransportEndpointID, p } // HandleControlPacket implements stack.TransportEndpoint.HandleControlPacket. -func (epsByNIC *endpointsByNIC) handleControlPacket(n *NIC, id TransportEndpointID, typ ControlType, extra uint32, pkt PacketBuffer) { +func (epsByNIC *endpointsByNIC) handleControlPacket(n *NIC, id TransportEndpointID, typ ControlType, extra uint32, pkt *PacketBuffer) { epsByNIC.mu.RLock() defer epsByNIC.mu.RUnlock() @@ -251,7 +251,7 @@ type transportDemuxer struct { // the dispatcher to delivery packets to the QueuePacket method instead of // calling HandlePacket directly on the endpoint. type queuedTransportProtocol interface { - QueuePacket(r *Route, ep TransportEndpoint, id TransportEndpointID, pkt PacketBuffer) + QueuePacket(r *Route, ep TransportEndpoint, id TransportEndpointID, pkt *PacketBuffer) } func newTransportDemuxer(stack *Stack) *transportDemuxer { @@ -379,7 +379,7 @@ func selectEndpoint(id TransportEndpointID, mpep *multiPortEndpoint, seed uint32 return mpep.endpoints[idx] } -func (ep *multiPortEndpoint) handlePacketAll(r *Route, id TransportEndpointID, pkt PacketBuffer) { +func (ep *multiPortEndpoint) handlePacketAll(r *Route, id TransportEndpointID, pkt *PacketBuffer) { ep.mu.RLock() queuedProtocol, mustQueue := ep.demux.queuedProtocols[protocolIDs{ep.netProto, ep.transProto}] // HandlePacket takes ownership of pkt, so each endpoint needs @@ -470,7 +470,7 @@ func (d *transportDemuxer) unregisterEndpoint(netProtos []tcpip.NetworkProtocolN // deliverPacket attempts to find one or more matching transport endpoints, and // then, if matches are found, delivers the packet to them. Returns true if // the packet no longer needs to be handled. -func (d *transportDemuxer) deliverPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt PacketBuffer, id TransportEndpointID) bool { +func (d *transportDemuxer) deliverPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt *PacketBuffer, id TransportEndpointID) bool { eps, ok := d.protocol[protocolIDs{r.NetProto, protocol}] if !ok { return false @@ -520,7 +520,7 @@ func (d *transportDemuxer) deliverPacket(r *Route, protocol tcpip.TransportProto // deliverRawPacket attempts to deliver the given packet and returns whether it // was delivered successfully. -func (d *transportDemuxer) deliverRawPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt PacketBuffer) bool { +func (d *transportDemuxer) deliverRawPacket(r *Route, protocol tcpip.TransportProtocolNumber, pkt *PacketBuffer) bool { eps, ok := d.protocol[protocolIDs{r.NetProto, protocol}] if !ok { return false @@ -544,7 +544,7 @@ func (d *transportDemuxer) deliverRawPacket(r *Route, protocol tcpip.TransportPr // deliverControlPacket attempts to deliver the given control packet. Returns // true if it found an endpoint, false otherwise. -func (d *transportDemuxer) deliverControlPacket(n *NIC, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ ControlType, extra uint32, pkt PacketBuffer, id TransportEndpointID) bool { +func (d *transportDemuxer) deliverControlPacket(n *NIC, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ ControlType, extra uint32, pkt *PacketBuffer, id TransportEndpointID) bool { eps, ok := d.protocol[protocolIDs{net, trans}] if !ok { return false diff --git a/pkg/tcpip/stack/transport_demuxer_test.go b/pkg/tcpip/stack/transport_demuxer_test.go index 2474a7db3..67d778137 100644 --- a/pkg/tcpip/stack/transport_demuxer_test.go +++ b/pkg/tcpip/stack/transport_demuxer_test.go @@ -127,7 +127,7 @@ func (c *testContext) sendV4Packet(payload []byte, h *headers, linkEpID tcpip.NI u.SetChecksum(^u.CalculateChecksum(xsum)) // Inject packet. - c.linkEps[linkEpID].InjectInbound(ipv4.ProtocolNumber, stack.PacketBuffer{ + c.linkEps[linkEpID].InjectInbound(ipv4.ProtocolNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), NetworkHeader: buffer.View(ip), TransportHeader: buffer.View(u), @@ -165,7 +165,7 @@ func (c *testContext) sendV6Packet(payload []byte, h *headers, linkEpID tcpip.NI u.SetChecksum(^u.CalculateChecksum(xsum)) // Inject packet. - c.linkEps[linkEpID].InjectInbound(ipv6.ProtocolNumber, stack.PacketBuffer{ + c.linkEps[linkEpID].InjectInbound(ipv6.ProtocolNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), NetworkHeader: buffer.View(ip), TransportHeader: buffer.View(u), diff --git a/pkg/tcpip/stack/transport_test.go b/pkg/tcpip/stack/transport_test.go index a611e44ab..cb350ead3 100644 --- a/pkg/tcpip/stack/transport_test.go +++ b/pkg/tcpip/stack/transport_test.go @@ -88,7 +88,7 @@ func (f *fakeTransportEndpoint) Write(p tcpip.Payloader, opts tcpip.WriteOptions if err != nil { return 0, nil, err } - if err := f.route.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: fakeTransNumber, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := f.route.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: fakeTransNumber, TTL: 123, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: buffer.View(v).ToVectorisedView(), }); err != nil { @@ -215,7 +215,7 @@ func (*fakeTransportEndpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Erro return tcpip.FullAddress{}, nil } -func (f *fakeTransportEndpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, _ stack.PacketBuffer) { +func (f *fakeTransportEndpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, _ *stack.PacketBuffer) { // Increment the number of received packets. f.proto.packetCount++ if f.acceptQueue != nil { @@ -232,7 +232,7 @@ func (f *fakeTransportEndpoint) HandlePacket(r *stack.Route, id stack.TransportE } } -func (f *fakeTransportEndpoint) HandleControlPacket(stack.TransportEndpointID, stack.ControlType, uint32, stack.PacketBuffer) { +func (f *fakeTransportEndpoint) HandleControlPacket(stack.TransportEndpointID, stack.ControlType, uint32, *stack.PacketBuffer) { // Increment the number of received control packets. f.proto.controlCount++ } @@ -289,7 +289,7 @@ func (*fakeTransportProtocol) ParsePorts(buffer.View) (src, dst uint16, err *tcp return 0, 0, nil } -func (*fakeTransportProtocol) HandleUnknownDestinationPacket(*stack.Route, stack.TransportEndpointID, stack.PacketBuffer) bool { +func (*fakeTransportProtocol) HandleUnknownDestinationPacket(*stack.Route, stack.TransportEndpointID, *stack.PacketBuffer) bool { return true } @@ -369,7 +369,7 @@ func TestTransportReceive(t *testing.T) { // Make sure packet with wrong protocol is not delivered. buf[0] = 1 buf[2] = 0 - linkEP.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + linkEP.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeTrans.packetCount != 0 { @@ -380,7 +380,7 @@ func TestTransportReceive(t *testing.T) { buf[0] = 1 buf[1] = 3 buf[2] = byte(fakeTransNumber) - linkEP.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + linkEP.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeTrans.packetCount != 0 { @@ -391,7 +391,7 @@ func TestTransportReceive(t *testing.T) { buf[0] = 1 buf[1] = 2 buf[2] = byte(fakeTransNumber) - linkEP.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + linkEP.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeTrans.packetCount != 1 { @@ -446,7 +446,7 @@ func TestTransportControlReceive(t *testing.T) { buf[fakeNetHeaderLen+0] = 0 buf[fakeNetHeaderLen+1] = 1 buf[fakeNetHeaderLen+2] = 0 - linkEP.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + linkEP.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeTrans.controlCount != 0 { @@ -457,7 +457,7 @@ func TestTransportControlReceive(t *testing.T) { buf[fakeNetHeaderLen+0] = 3 buf[fakeNetHeaderLen+1] = 1 buf[fakeNetHeaderLen+2] = byte(fakeTransNumber) - linkEP.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + linkEP.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeTrans.controlCount != 0 { @@ -468,7 +468,7 @@ func TestTransportControlReceive(t *testing.T) { buf[fakeNetHeaderLen+0] = 2 buf[fakeNetHeaderLen+1] = 1 buf[fakeNetHeaderLen+2] = byte(fakeTransNumber) - linkEP.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + linkEP.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) if fakeTrans.controlCount != 1 { @@ -623,7 +623,7 @@ func TestTransportForwarding(t *testing.T) { req[0] = 1 req[1] = 3 req[2] = byte(fakeTransNumber) - ep2.InjectInbound(fakeNetNumber, stack.PacketBuffer{ + ep2.InjectInbound(fakeNetNumber, &stack.PacketBuffer{ Data: req.ToVectorisedView(), }) diff --git a/pkg/tcpip/time_unsafe.go b/pkg/tcpip/time_unsafe.go index 2f98a996f..7f172f978 100644 --- a/pkg/tcpip/time_unsafe.go +++ b/pkg/tcpip/time_unsafe.go @@ -13,7 +13,7 @@ // limitations under the License. // +build go1.9 -// +build !go1.15 +// +build !go1.16 // Check go:linkname function signatures when updating Go version. diff --git a/pkg/tcpip/transport/icmp/endpoint.go b/pkg/tcpip/transport/icmp/endpoint.go index b1d820372..57e0a069b 100644 --- a/pkg/tcpip/transport/icmp/endpoint.go +++ b/pkg/tcpip/transport/icmp/endpoint.go @@ -140,11 +140,6 @@ func (e *endpoint) SetOwner(owner tcpip.PacketOwner) { e.owner = owner } -// IPTables implements tcpip.Endpoint.IPTables. -func (e *endpoint) IPTables() (stack.IPTables, error) { - return e.stack.IPTables(), nil -} - // Read reads data from the endpoint. This method does not block if // there is no data pending. func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { @@ -450,7 +445,7 @@ func send4(r *stack.Route, ident uint16, data buffer.View, ttl uint8, owner tcpi if ttl == 0 { ttl = r.DefaultTTL() } - return r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: ttl, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + return r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: ttl, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: data.ToVectorisedView(), TransportHeader: buffer.View(icmpv4), @@ -481,7 +476,7 @@ func send6(r *stack.Route, ident uint16, data buffer.View, ttl uint8) *tcpip.Err if ttl == 0 { ttl = r.DefaultTTL() } - return r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: ttl, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + return r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: ttl, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: dataVV, TransportHeader: buffer.View(icmpv6), @@ -511,6 +506,7 @@ func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { nicID := addr.NIC localPort := uint16(0) switch e.state { + case stateInitial: case stateBound, stateConnected: localPort = e.ID.LocalPort if e.BindNICID == 0 { @@ -743,7 +739,7 @@ func (e *endpoint) Readiness(mask waiter.EventMask) waiter.EventMask { // HandlePacket is called by the stack when new packets arrive to this transport // endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) { // Only accept echo replies. switch e.NetProto { case header.IPv4ProtocolNumber: @@ -805,7 +801,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pk } // HandleControlPacket implements stack.TransportEndpoint.HandleControlPacket. -func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { } // State implements tcpip.Endpoint.State. The ICMP endpoint currently doesn't diff --git a/pkg/tcpip/transport/icmp/protocol.go b/pkg/tcpip/transport/icmp/protocol.go index 3c47692b2..2ec6749c7 100644 --- a/pkg/tcpip/transport/icmp/protocol.go +++ b/pkg/tcpip/transport/icmp/protocol.go @@ -104,7 +104,7 @@ func (p *protocol) ParsePorts(v buffer.View) (src, dst uint16, err *tcpip.Error) // HandleUnknownDestinationPacket handles packets targeted at this protocol but // that don't match any existing endpoint. -func (*protocol) HandleUnknownDestinationPacket(*stack.Route, stack.TransportEndpointID, stack.PacketBuffer) bool { +func (*protocol) HandleUnknownDestinationPacket(*stack.Route, stack.TransportEndpointID, *stack.PacketBuffer) bool { return true } diff --git a/pkg/tcpip/transport/packet/endpoint.go b/pkg/tcpip/transport/packet/endpoint.go index 23158173d..baf08eda6 100644 --- a/pkg/tcpip/transport/packet/endpoint.go +++ b/pkg/tcpip/transport/packet/endpoint.go @@ -132,11 +132,6 @@ func (ep *endpoint) Close() { // ModerateRecvBuf implements tcpip.Endpoint.ModerateRecvBuf. func (ep *endpoint) ModerateRecvBuf(copied int) {} -// IPTables implements tcpip.Endpoint.IPTables. -func (ep *endpoint) IPTables() (stack.IPTables, error) { - return ep.stack.IPTables(), nil -} - // Read implements tcpip.Endpoint.Read. func (ep *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { ep.rcvMu.Lock() @@ -298,7 +293,7 @@ func (ep *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) { } // HandlePacket implements stack.PacketEndpoint.HandlePacket. -func (ep *endpoint) HandlePacket(nicID tcpip.NICID, localAddr tcpip.LinkAddress, netProto tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) { +func (ep *endpoint) HandlePacket(nicID tcpip.NICID, localAddr tcpip.LinkAddress, netProto tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) { ep.rcvMu.Lock() // Drop the packet if our buffer is currently full. diff --git a/pkg/tcpip/transport/raw/endpoint.go b/pkg/tcpip/transport/raw/endpoint.go index eee754a5a..21c34fac2 100644 --- a/pkg/tcpip/transport/raw/endpoint.go +++ b/pkg/tcpip/transport/raw/endpoint.go @@ -166,11 +166,6 @@ func (e *endpoint) SetOwner(owner tcpip.PacketOwner) { e.owner = owner } -// IPTables implements tcpip.Endpoint.IPTables. -func (e *endpoint) IPTables() (stack.IPTables, error) { - return e.stack.IPTables(), nil -} - // Read implements tcpip.Endpoint.Read. func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { if !e.associated { @@ -348,7 +343,7 @@ func (e *endpoint) finishWrite(payloadBytes []byte, route *stack.Route) (int64, switch e.NetProto { case header.IPv4ProtocolNumber: if !e.associated { - if err := route.WriteHeaderIncludedPacket(stack.PacketBuffer{ + if err := route.WriteHeaderIncludedPacket(&stack.PacketBuffer{ Data: buffer.View(payloadBytes).ToVectorisedView(), }); err != nil { return 0, nil, err @@ -357,7 +352,7 @@ func (e *endpoint) finishWrite(payloadBytes []byte, route *stack.Route) (int64, } hdr := buffer.NewPrependable(len(payloadBytes) + int(route.MaxHeaderLength())) - if err := route.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: e.TransProto, TTL: route.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := route.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: e.TransProto, TTL: route.DefaultTTL(), TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: buffer.View(payloadBytes).ToVectorisedView(), Owner: e.owner, @@ -584,7 +579,7 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) { } // HandlePacket implements stack.RawTransportEndpoint.HandlePacket. -func (e *endpoint) HandlePacket(route *stack.Route, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(route *stack.Route, pkt *stack.PacketBuffer) { e.rcvMu.Lock() // Drop the packet if our buffer is currently full. diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index f38eb6833..e26f01fae 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -86,10 +86,6 @@ go_test( "tcp_test.go", "tcp_timestamp_test.go", ], - # FIXME(b/68809571) - tags = [ - "flaky", - ], deps = [ ":tcp", "//pkg/sync", diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index a7e088d4e..7da93dcc4 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/pkg/tcpip/transport/tcp/connect.go @@ -833,13 +833,13 @@ func sendTCP(r *stack.Route, tf tcpFields, data buffer.VectorisedView, gso *stac return sendTCPBatch(r, tf, data, gso, owner) } - pkt := stack.PacketBuffer{ + pkt := &stack.PacketBuffer{ Header: buffer.NewPrependable(header.TCPMinimumSize + int(r.MaxHeaderLength()) + optLen), Data: data, Hash: tf.txHash, Owner: owner, } - buildTCPHdr(r, tf, &pkt, gso) + buildTCPHdr(r, tf, pkt, gso) if tf.ttl == 0 { tf.ttl = r.DefaultTTL() @@ -1347,6 +1347,7 @@ func (e *endpoint) protocolMainLoop(handshake bool, wakerInitDone chan<- struct{ e.setEndpointState(StateError) e.HardError = err + e.workerCleanup = true // Lock released below. epilogue() return err diff --git a/pkg/tcpip/transport/tcp/dispatcher.go b/pkg/tcpip/transport/tcp/dispatcher.go index 6062ca916..047704c80 100644 --- a/pkg/tcpip/transport/tcp/dispatcher.go +++ b/pkg/tcpip/transport/tcp/dispatcher.go @@ -186,7 +186,7 @@ func (d *dispatcher) wait() { } } -func (d *dispatcher) queuePacket(r *stack.Route, stackEP stack.TransportEndpoint, id stack.TransportEndpointID, pkt stack.PacketBuffer) { +func (d *dispatcher) queuePacket(r *stack.Route, stackEP stack.TransportEndpoint, id stack.TransportEndpointID, pkt *stack.PacketBuffer) { ep := stackEP.(*endpoint) s := newSegment(r, id, pkt) if !s.parse() { diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index b5ba972f1..19f7bf449 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -63,7 +63,8 @@ const ( StateClosing ) -// connected is the set of states where an endpoint is connected to a peer. +// connected returns true when s is one of the states representing an +// endpoint connected to a peer. func (s EndpointState) connected() bool { switch s { case StateEstablished, StateFinWait1, StateFinWait2, StateTimeWait, StateCloseWait, StateLastAck, StateClosing: @@ -73,6 +74,40 @@ func (s EndpointState) connected() bool { } } +// connecting returns true when s is one of the states representing a +// connection in progress, but not yet fully established. +func (s EndpointState) connecting() bool { + switch s { + case StateConnecting, StateSynSent, StateSynRecv: + return true + default: + return false + } +} + +// handshake returns true when s is one of the states representing an endpoint +// in the middle of a TCP handshake. +func (s EndpointState) handshake() bool { + switch s { + case StateSynSent, StateSynRecv: + return true + default: + return false + } +} + +// closed returns true when s is one of the states an endpoint transitions to +// when closed or when it encounters an error. This is distinct from a newly +// initialized endpoint that was never connected. +func (s EndpointState) closed() bool { + switch s { + case StateClose, StateError: + return true + default: + return false + } +} + // String implements fmt.Stringer.String. func (s EndpointState) String() string { switch s { @@ -1172,11 +1207,6 @@ func (e *endpoint) SetOwner(owner tcpip.PacketOwner) { e.owner = owner } -// IPTables implements tcpip.Endpoint.IPTables. -func (e *endpoint) IPTables() (stack.IPTables, error) { - return e.stack.IPTables(), nil -} - // Read reads data from the endpoint. func (e *endpoint) Read(*tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { e.LockUser() @@ -2462,7 +2492,7 @@ func (e *endpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) { }, nil } -func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) { // TCP HandlePacket is not required anymore as inbound packets first // land at the Dispatcher which then can either delivery using the // worker go routine or directly do the invoke the tcp processing inline @@ -2481,7 +2511,7 @@ func (e *endpoint) enqueueSegment(s *segment) bool { } // HandleControlPacket implements stack.TransportEndpoint.HandleControlPacket. -func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { switch typ { case stack.ControlPacketTooBig: e.sndBufMu.Lock() diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go index fc43c11e2..cbb779666 100644 --- a/pkg/tcpip/transport/tcp/endpoint_state.go +++ b/pkg/tcpip/transport/tcp/endpoint_state.go @@ -49,11 +49,10 @@ func (e *endpoint) beforeSave() { e.mu.Lock() defer e.mu.Unlock() - switch e.EndpointState() { - case StateInitial, StateBound: - // TODO(b/138137272): this enumeration duplicates - // EndpointState.connected. remove it. - case StateEstablished, StateSynSent, StateSynRecv, StateFinWait1, StateFinWait2, StateTimeWait, StateCloseWait, StateLastAck, StateClosing: + epState := e.EndpointState() + switch { + case epState == StateInitial || epState == StateBound: + case epState.connected() || epState.handshake(): if e.route.Capabilities()&stack.CapabilitySaveRestore == 0 { if e.route.Capabilities()&stack.CapabilityDisconnectOk == 0 { panic(tcpip.ErrSaveRejection{fmt.Errorf("endpoint cannot be saved in connected state: local %v:%d, remote %v:%d", e.ID.LocalAddress, e.ID.LocalPort, e.ID.RemoteAddress, e.ID.RemotePort)}) @@ -69,15 +68,16 @@ func (e *endpoint) beforeSave() { break } fallthrough - case StateListen, StateConnecting: + case epState == StateListen || epState == StateConnecting: e.drainSegmentLocked() - if e.EndpointState() != StateClose && e.EndpointState() != StateError { + // Refresh epState, since drainSegmentLocked may have changed it. + epState = e.EndpointState() + if !epState.closed() { if !e.workerRunning { panic("endpoint has no worker running in listen, connecting, or connected state") } - break } - case StateError, StateClose: + case epState.closed(): for e.workerRunning { e.mu.Unlock() time.Sleep(100 * time.Millisecond) @@ -148,23 +148,23 @@ var connectingLoading sync.WaitGroup // Bound endpoint loading happens last. // loadState is invoked by stateify. -func (e *endpoint) loadState(state EndpointState) { +func (e *endpoint) loadState(epState EndpointState) { // This is to ensure that the loading wait groups include all applicable // endpoints before any asynchronous calls to the Wait() methods. // For restore purposes we treat TimeWait like a connected endpoint. - if state.connected() || state == StateTimeWait { + if epState.connected() || epState == StateTimeWait { connectedLoading.Add(1) } - switch state { - case StateListen: + switch { + case epState == StateListen: listenLoading.Add(1) - case StateConnecting, StateSynSent, StateSynRecv: + case epState.connecting(): connectingLoading.Add(1) } // Directly update the state here rather than using e.setEndpointState // as the endpoint is still being loaded and the stack reference is not // yet initialized. - atomic.StoreUint32((*uint32)(&e.state), uint32(state)) + atomic.StoreUint32((*uint32)(&e.state), uint32(epState)) } // afterLoad is invoked by stateify. @@ -183,8 +183,8 @@ func (e *endpoint) afterLoad() { func (e *endpoint) Resume(s *stack.Stack) { e.stack = s e.segmentQueue.setLimit(MaxUnprocessedSegments) - state := e.origEndpointState - switch state { + epState := e.origEndpointState + switch epState { case StateInitial, StateBound, StateListen, StateConnecting, StateEstablished: var ss SendBufferSizeOption if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err == nil { @@ -208,8 +208,8 @@ func (e *endpoint) Resume(s *stack.Stack) { } } - switch state { - case StateEstablished, StateFinWait1, StateFinWait2, StateTimeWait, StateCloseWait, StateLastAck, StateClosing: + switch { + case epState.connected(): bind() if len(e.connectingAddress) == 0 { e.connectingAddress = e.ID.RemoteAddress @@ -232,13 +232,13 @@ func (e *endpoint) Resume(s *stack.Stack) { closed := e.closed e.mu.Unlock() e.notifyProtocolGoroutine(notifyTickleWorker) - if state == StateFinWait2 && closed { + if epState == StateFinWait2 && closed { // If the endpoint has been closed then make sure we notify so // that the FIN_WAIT2 timer is started after a restore. e.notifyProtocolGoroutine(notifyClose) } connectedLoading.Done() - case StateListen: + case epState == StateListen: tcpip.AsyncLoading.Add(1) go func() { connectedLoading.Wait() @@ -255,7 +255,7 @@ func (e *endpoint) Resume(s *stack.Stack) { listenLoading.Done() tcpip.AsyncLoading.Done() }() - case StateConnecting, StateSynSent, StateSynRecv: + case epState.connecting(): tcpip.AsyncLoading.Add(1) go func() { connectedLoading.Wait() @@ -267,7 +267,7 @@ func (e *endpoint) Resume(s *stack.Stack) { connectingLoading.Done() tcpip.AsyncLoading.Done() }() - case StateBound: + case epState == StateBound: tcpip.AsyncLoading.Add(1) go func() { connectedLoading.Wait() @@ -276,7 +276,7 @@ func (e *endpoint) Resume(s *stack.Stack) { bind() tcpip.AsyncLoading.Done() }() - case StateClose: + case epState == StateClose: if e.isPortReserved { tcpip.AsyncLoading.Add(1) go func() { @@ -291,12 +291,11 @@ func (e *endpoint) Resume(s *stack.Stack) { e.state = StateClose e.stack.CompleteTransportEndpointCleanup(e) tcpip.DeleteDanglingEndpoint(e) - case StateError: + case epState == StateError: e.state = StateError e.stack.CompleteTransportEndpointCleanup(e) tcpip.DeleteDanglingEndpoint(e) } - } // saveLastError is invoked by stateify. diff --git a/pkg/tcpip/transport/tcp/forwarder.go b/pkg/tcpip/transport/tcp/forwarder.go index 704d01c64..070b634b4 100644 --- a/pkg/tcpip/transport/tcp/forwarder.go +++ b/pkg/tcpip/transport/tcp/forwarder.go @@ -61,7 +61,7 @@ func NewForwarder(s *stack.Stack, rcvWnd, maxInFlight int, handler func(*Forward // // This function is expected to be passed as an argument to the // stack.SetTransportProtocolHandler function. -func (f *Forwarder) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) bool { +func (f *Forwarder) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) bool { s := newSegment(r, id, pkt) defer s.decRef() diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go index 2a2a7ddeb..c827d0277 100644 --- a/pkg/tcpip/transport/tcp/protocol.go +++ b/pkg/tcpip/transport/tcp/protocol.go @@ -206,7 +206,7 @@ func (*protocol) ParsePorts(v buffer.View) (src, dst uint16, err *tcpip.Error) { // to a specific processing queue. Each queue is serviced by its own processor // goroutine which is responsible for dequeuing and doing full TCP dispatch of // the packet. -func (p *protocol) QueuePacket(r *stack.Route, ep stack.TransportEndpoint, id stack.TransportEndpointID, pkt stack.PacketBuffer) { +func (p *protocol) QueuePacket(r *stack.Route, ep stack.TransportEndpoint, id stack.TransportEndpointID, pkt *stack.PacketBuffer) { p.dispatcher.queuePacket(r, ep, id, pkt) } @@ -217,7 +217,7 @@ func (p *protocol) QueuePacket(r *stack.Route, ep stack.TransportEndpoint, id st // a reset is sent in response to any incoming segment except another reset. In // particular, SYNs addressed to a non-existent connection are rejected by this // means." -func (*protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) bool { +func (*protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) bool { s := newSegment(r, id, pkt) defer s.decRef() diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index 074edded6..0c099e2fd 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -60,7 +60,7 @@ type segment struct { xmitCount uint32 } -func newSegment(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) *segment { +func newSegment(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) *segment { s := &segment{ refCnt: 1, id: id, diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index 06dc9b7d7..acacb42e4 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -618,6 +618,20 @@ func (s *sender) splitSeg(seg *segment, size int) { nSeg.data.TrimFront(size) nSeg.sequenceNumber.UpdateForward(seqnum.Size(size)) s.writeList.InsertAfter(seg, nSeg) + + // The segment being split does not carry PUSH flag because it is + // followed by the newly split segment. + // RFC1122 section 4.2.2.2: MUST set the PSH bit in the last buffered + // segment (i.e., when there is no more queued data to be sent). + // Linux removes PSH flag only when the segment is being split over MSS + // and retains it when we are splitting the segment over lack of sender + // window space. + // ref: net/ipv4/tcp_output.c::tcp_write_xmit(), tcp_mss_split_point() + // ref: net/ipv4/tcp_output.c::tcp_write_wakeup(), tcp_snd_wnd_test() + if seg.data.Size() > s.maxPayloadSize { + seg.flags ^= header.TCPFlagPsh + } + seg.data.CapLength(size) } @@ -739,7 +753,7 @@ func (s *sender) maybeSendSegment(seg *segment, limit int, end seqnum.Value) (se if !s.isAssignedSequenceNumber(seg) { // Merge segments if allowed. if seg.data.Size() != 0 { - available := int(seg.sequenceNumber.Size(end)) + available := int(s.sndNxt.Size(end)) if available > limit { available = limit } @@ -782,8 +796,11 @@ func (s *sender) maybeSendSegment(seg *segment, limit int, end seqnum.Value) (se // sent all at once. return false } - if atomic.LoadUint32(&s.ep.cork) != 0 { - // Hold back the segment until full. + // With TCP_CORK, hold back until minimum of the available + // send space and MSS. + // TODO(gvisor.dev/issue/2833): Drain the held segments after a + // timeout. + if seg.data.Size() < s.maxPayloadSize && atomic.LoadUint32(&s.ep.cork) != 0 { return false } } @@ -816,6 +833,25 @@ func (s *sender) maybeSendSegment(seg *segment, limit int, end seqnum.Value) (se panic("Netstack queues FIN segments without data.") } + segEnd = seg.sequenceNumber.Add(seqnum.Size(seg.data.Size())) + // If the entire segment cannot be accomodated in the receiver + // advertized window, skip splitting and sending of the segment. + // ref: net/ipv4/tcp_output.c::tcp_snd_wnd_test() + // + // Linux checks this for all segment transmits not triggered + // by a probe timer. On this condition, it defers the segment + // split and transmit to a short probe timer. + // ref: include/net/tcp.h::tcp_check_probe_timer() + // ref: net/ipv4/tcp_output.c::tcp_write_wakeup() + // + // Instead of defining a new transmit timer, we attempt to split the + // segment right here if there are no pending segments. + // If there are pending segments, segment transmits are deferred + // to the retransmit timer handler. + if s.sndUna != s.sndNxt && !segEnd.LessThan(end) { + return false + } + if !seg.sequenceNumber.LessThan(end) { return false } @@ -824,9 +860,17 @@ func (s *sender) maybeSendSegment(seg *segment, limit int, end seqnum.Value) (se if available == 0 { return false } + + // The segment size limit is computed as a function of sender congestion + // window and MSS. When sender congestion window is > 1, this limit can + // be larger than MSS. Ensure that the currently available send space + // is not greater than minimum of this limit and MSS. if available > limit { available = limit } + if available > s.maxPayloadSize { + available = s.maxPayloadSize + } if seg.data.Size() > available { s.splitSeg(seg, available) diff --git a/pkg/tcpip/transport/tcp/testing/context/context.go b/pkg/tcpip/transport/tcp/testing/context/context.go index 7b1d72cf4..9721f6caf 100644 --- a/pkg/tcpip/transport/tcp/testing/context/context.go +++ b/pkg/tcpip/transport/tcp/testing/context/context.go @@ -316,7 +316,7 @@ func (c *Context) SendICMPPacket(typ header.ICMPv4Type, code uint8, p1, p2 []byt copy(icmp[header.ICMPv4PayloadOffset:], p2) // Inject packet. - c.linkEP.InjectInbound(ipv4.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv4.ProtocolNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) } @@ -372,7 +372,7 @@ func (c *Context) BuildSegmentWithAddrs(payload []byte, h *Headers, src, dst tcp // SendSegment sends a TCP segment that has already been built and written to a // buffer.VectorisedView. func (c *Context) SendSegment(s buffer.VectorisedView) { - c.linkEP.InjectInbound(ipv4.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv4.ProtocolNumber, &stack.PacketBuffer{ Data: s, }) } @@ -380,7 +380,7 @@ func (c *Context) SendSegment(s buffer.VectorisedView) { // SendPacket builds and sends a TCP segment(with the provided payload & TCP // headers) in an IPv4 packet via the link layer endpoint. func (c *Context) SendPacket(payload []byte, h *Headers) { - c.linkEP.InjectInbound(ipv4.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv4.ProtocolNumber, &stack.PacketBuffer{ Data: c.BuildSegment(payload, h), }) } @@ -389,7 +389,7 @@ func (c *Context) SendPacket(payload []byte, h *Headers) { // & TCPheaders) in an IPv4 packet via the link layer endpoint using the // provided source and destination IPv4 addresses. func (c *Context) SendPacketWithAddrs(payload []byte, h *Headers, src, dst tcpip.Address) { - c.linkEP.InjectInbound(ipv4.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv4.ProtocolNumber, &stack.PacketBuffer{ Data: c.BuildSegmentWithAddrs(payload, h, src, dst), }) } @@ -564,7 +564,7 @@ func (c *Context) SendV6PacketWithAddrs(payload []byte, h *Headers, src, dst tcp t.SetChecksum(^t.CalculateChecksum(xsum)) // Inject packet. - c.linkEP.InjectInbound(ipv6.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv6.ProtocolNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), }) } diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 647b2067a..663af8fec 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -247,11 +247,6 @@ func (e *endpoint) Close() { // ModerateRecvBuf implements tcpip.Endpoint.ModerateRecvBuf. func (e *endpoint) ModerateRecvBuf(copied int) {} -// IPTables implements tcpip.Endpoint.IPTables. -func (e *endpoint) IPTables() (stack.IPTables, error) { - return e.stack.IPTables(), nil -} - // Read reads data from the endpoint. This method does not block if // there is no data pending. func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { @@ -921,7 +916,11 @@ func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort u if useDefaultTTL { ttl = r.DefaultTTL() } - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: ProtocolNumber, TTL: ttl, TOS: tos}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{ + Protocol: ProtocolNumber, + TTL: ttl, + TOS: tos, + }, &stack.PacketBuffer{ Header: hdr, Data: data, TransportHeader: buffer.View(udp), @@ -1269,7 +1268,7 @@ func (e *endpoint) Readiness(mask waiter.EventMask) waiter.EventMask { // HandlePacket is called by the stack when new packets arrive to this transport // endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) { +func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) { // Get the header then trim it from the view. hdr, ok := pkt.Data.PullUp(header.UDPMinimumSize) if !ok || int(header.UDP(hdr).Length()) > pkt.Data.Size() { @@ -1336,7 +1335,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pk } // HandleControlPacket implements stack.TransportEndpoint.HandleControlPacket. -func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { if typ == stack.ControlPortUnreachable { e.mu.RLock() defer e.mu.RUnlock() diff --git a/pkg/tcpip/transport/udp/forwarder.go b/pkg/tcpip/transport/udp/forwarder.go index a674ceb68..7abfa0ed2 100644 --- a/pkg/tcpip/transport/udp/forwarder.go +++ b/pkg/tcpip/transport/udp/forwarder.go @@ -43,7 +43,7 @@ func NewForwarder(s *stack.Stack, handler func(*ForwarderRequest)) *Forwarder { // // This function is expected to be passed as an argument to the // stack.SetTransportProtocolHandler function. -func (f *Forwarder) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) bool { +func (f *Forwarder) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) bool { f.handler(&ForwarderRequest{ stack: f.stack, route: r, @@ -61,7 +61,7 @@ type ForwarderRequest struct { stack *stack.Stack route *stack.Route id stack.TransportEndpointID - pkt stack.PacketBuffer + pkt *stack.PacketBuffer } // ID returns the 4-tuple (src address, src port, dst address, dst port) that diff --git a/pkg/tcpip/transport/udp/protocol.go b/pkg/tcpip/transport/udp/protocol.go index 52af6de22..e320c5758 100644 --- a/pkg/tcpip/transport/udp/protocol.go +++ b/pkg/tcpip/transport/udp/protocol.go @@ -66,7 +66,7 @@ func (*protocol) ParsePorts(v buffer.View) (src, dst uint16, err *tcpip.Error) { // HandleUnknownDestinationPacket handles packets targeted at this protocol but // that don't match any existing endpoint. -func (p *protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.TransportEndpointID, pkt stack.PacketBuffer) bool { +func (p *protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.TransportEndpointID, pkt *stack.PacketBuffer) bool { // Get the header then trim it from the view. h, ok := pkt.Data.PullUp(header.UDPMinimumSize) if !ok { @@ -140,7 +140,7 @@ func (p *protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.Trans pkt.SetType(header.ICMPv4DstUnreachable) pkt.SetCode(header.ICMPv4PortUnreachable) pkt.SetChecksum(header.ICMPv4Checksum(pkt, payload)) - r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: payload, }) @@ -177,7 +177,7 @@ func (p *protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.Trans pkt.SetType(header.ICMPv6DstUnreachable) pkt.SetCode(header.ICMPv6PortUnreachable) pkt.SetChecksum(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, payload)) - r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: payload, }) diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go index 8acaa607a..e8ade882b 100644 --- a/pkg/tcpip/transport/udp/udp_test.go +++ b/pkg/tcpip/transport/udp/udp_test.go @@ -440,7 +440,7 @@ func (c *testContext) injectV6Packet(payload []byte, h *header4Tuple, valid bool u.SetChecksum(^u.CalculateChecksum(xsum)) // Inject packet. - c.linkEP.InjectInbound(ipv6.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv6.ProtocolNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), NetworkHeader: buffer.View(ip), TransportHeader: buffer.View(u), @@ -487,7 +487,7 @@ func (c *testContext) injectV4Packet(payload []byte, h *header4Tuple, valid bool // Inject packet. - c.linkEP.InjectInbound(ipv4.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(ipv4.ProtocolNumber, &stack.PacketBuffer{ Data: buf.ToVectorisedView(), NetworkHeader: buffer.View(ip), TransportHeader: buffer.View(u), |