summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/tcpip')
-rw-r--r--pkg/tcpip/adapters/gonet/gonet.go5
-rw-r--r--pkg/tcpip/header/tcp.go29
-rw-r--r--pkg/tcpip/link/channel/channel.go2
-rw-r--r--pkg/tcpip/link/fdbased/endpoint.go2
-rw-r--r--pkg/tcpip/link/fdbased/endpoint_test.go2
-rw-r--r--pkg/tcpip/link/fdbased/mmap.go2
-rw-r--r--pkg/tcpip/link/fdbased/packet_dispatchers.go6
-rw-r--r--pkg/tcpip/link/loopback/loopback.go4
-rw-r--r--pkg/tcpip/link/muxed/injectable.go2
-rw-r--r--pkg/tcpip/link/qdisc/fifo/endpoint.go4
-rw-r--r--pkg/tcpip/link/rawfile/blockingpoll_yield_unsafe.go2
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem.go2
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem_test.go2
-rw-r--r--pkg/tcpip/link/sniffer/sniffer.go4
-rw-r--r--pkg/tcpip/link/waitable/waitable.go4
-rw-r--r--pkg/tcpip/link/waitable/waitable_test.go8
-rw-r--r--pkg/tcpip/stack/forwarder_test.go2
-rw-r--r--pkg/tcpip/stack/iptables.go47
-rw-r--r--pkg/tcpip/stack/iptables_types.go62
-rw-r--r--pkg/tcpip/stack/nic.go9
-rw-r--r--pkg/tcpip/stack/nic_test.go2
-rw-r--r--pkg/tcpip/stack/registration.go21
-rw-r--r--pkg/tcpip/stack/route.go15
-rw-r--r--pkg/tcpip/stack/stack.go4
-rw-r--r--pkg/tcpip/tcpip.go90
-rw-r--r--pkg/tcpip/time_unsafe.go2
-rw-r--r--pkg/tcpip/transport/tcp/connect.go1
-rw-r--r--pkg/tcpip/transport/tcp/endpoint.go66
-rw-r--r--pkg/tcpip/transport/tcp/endpoint_state.go68
-rw-r--r--pkg/tcpip/transport/tcp/protocol.go52
-rw-r--r--pkg/tcpip/transport/tcp/snd.go59
-rw-r--r--pkg/tcpip/transport/tcp/tcp_test.go116
-rw-r--r--pkg/tcpip/transport/udp/endpoint.go32
-rw-r--r--pkg/tcpip/transport/udp/endpoint_state.go18
34 files changed, 552 insertions, 194 deletions
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/header/tcp.go b/pkg/tcpip/header/tcp.go
index 29454c4b9..4c6f808e5 100644
--- a/pkg/tcpip/header/tcp.go
+++ b/pkg/tcpip/header/tcp.go
@@ -66,6 +66,14 @@ const (
TCPOptionSACK = 5
)
+// Option Lengths.
+const (
+ TCPOptionMSSLength = 4
+ TCPOptionTSLength = 10
+ TCPOptionWSLength = 3
+ TCPOptionSackPermittedLength = 2
+)
+
// TCPFields contains the fields of a TCP packet. It is used to describe the
// fields of a packet that needs to be encoded.
type TCPFields struct {
@@ -494,14 +502,11 @@ func ParseTCPOptions(b []byte) TCPOptions {
// returns without encoding anything. It returns the number of bytes written to
// the provided buffer.
func EncodeMSSOption(mss uint32, b []byte) int {
- // mssOptionSize is the number of bytes in a valid MSS option.
- const mssOptionSize = 4
-
- if len(b) < mssOptionSize {
+ if len(b) < TCPOptionMSSLength {
return 0
}
- b[0], b[1], b[2], b[3] = TCPOptionMSS, mssOptionSize, byte(mss>>8), byte(mss)
- return mssOptionSize
+ b[0], b[1], b[2], b[3] = TCPOptionMSS, TCPOptionMSSLength, byte(mss>>8), byte(mss)
+ return TCPOptionMSSLength
}
// EncodeWSOption encodes the WS TCP option with the WS value in the
@@ -509,10 +514,10 @@ func EncodeMSSOption(mss uint32, b []byte) int {
// returns without encoding anything. It returns the number of bytes written to
// the provided buffer.
func EncodeWSOption(ws int, b []byte) int {
- if len(b) < 3 {
+ if len(b) < TCPOptionWSLength {
return 0
}
- b[0], b[1], b[2] = TCPOptionWS, 3, uint8(ws)
+ b[0], b[1], b[2] = TCPOptionWS, TCPOptionWSLength, uint8(ws)
return int(b[1])
}
@@ -521,10 +526,10 @@ func EncodeWSOption(ws int, b []byte) int {
// just returns without encoding anything. It returns the number of bytes
// written to the provided buffer.
func EncodeTSOption(tsVal, tsEcr uint32, b []byte) int {
- if len(b) < 10 {
+ if len(b) < TCPOptionTSLength {
return 0
}
- b[0], b[1] = TCPOptionTS, 10
+ b[0], b[1] = TCPOptionTS, TCPOptionTSLength
binary.BigEndian.PutUint32(b[2:], tsVal)
binary.BigEndian.PutUint32(b[6:], tsEcr)
return int(b[1])
@@ -535,11 +540,11 @@ func EncodeTSOption(tsVal, tsEcr uint32, b []byte) int {
// encoding anything. It returns the number of bytes written to the provided
// buffer.
func EncodeSACKPermittedOption(b []byte) int {
- if len(b) < 2 {
+ if len(b) < TCPOptionSackPermittedLength {
return 0
}
- b[0], b[1] = TCPOptionSACKPermitted, 2
+ b[0], b[1] = TCPOptionSACKPermitted, TCPOptionSackPermittedLength
return int(b[1])
}
diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go
index 9bf67686d..5eb78b398 100644
--- a/pkg/tcpip/link/channel/channel.go
+++ b/pkg/tcpip/link/channel/channel.go
@@ -187,7 +187,7 @@ func (e *Endpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber, pkt stack
// 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)
+ e.dispatcher.DeliverNetworkPacket(remote, "" /* local */, protocol, pkt)
}
// Attach saves the stack network-layer dispatcher for use later when packets
diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go
index affa1bbdf..5ee508d48 100644
--- a/pkg/tcpip/link/fdbased/endpoint.go
+++ b/pkg/tcpip/link/fdbased/endpoint.go
@@ -642,7 +642,7 @@ 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)
+ 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..6f41a71a8 100644
--- a/pkg/tcpip/link/fdbased/endpoint_test.go
+++ b/pkg/tcpip/link/fdbased/endpoint_test.go
@@ -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}
}
diff --git a/pkg/tcpip/link/fdbased/mmap.go b/pkg/tcpip/link/fdbased/mmap.go
index fe2bf3b0b..ca4229ed6 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..26c96a655 100644
--- a/pkg/tcpip/link/fdbased/packet_dispatchers.go
+++ b/pkg/tcpip/link/fdbased/packet_dispatchers.go
@@ -145,7 +145,7 @@ func (d *readVDispatcher) dispatch() (bool, *tcpip.Error) {
}
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
@@ -301,7 +301,7 @@ func (d *recvMMsgDispatcher) dispatch() (bool, *tcpip.Error) {
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..20d9e95f6 100644
--- a/pkg/tcpip/link/loopback/loopback.go
+++ b/pkg/tcpip/link/loopback/loopback.go
@@ -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..f0769830a 100644
--- a/pkg/tcpip/link/muxed/injectable.go
+++ b/pkg/tcpip/link/muxed/injectable.go
@@ -81,7 +81,7 @@ 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)
+ m.dispatcher.DeliverNetworkPacket("" /* remote */, "" /* local */, protocol, pkt)
}
// WritePackets writes outbound packets to the appropriate
diff --git a/pkg/tcpip/link/qdisc/fifo/endpoint.go b/pkg/tcpip/link/qdisc/fifo/endpoint.go
index 54432194d..ec5c5048a 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.
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..f5dec0a7f 100644
--- a/pkg/tcpip/link/sharedmem/sharedmem.go
+++ b/pkg/tcpip/link/sharedmem/sharedmem.go
@@ -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..f3fc62607 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,
diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go
index da1c520ae..b060d4627 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) {
+func (e *endpoint) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) {
e.dumpPacket("recv", nil, protocol, &pkt)
- e.dispatcher.DeliverNetworkPacket(e, remote, local, protocol, pkt)
+ e.dispatcher.DeliverNetworkPacket(remote, local, protocol, pkt)
}
// Attach implements the stack.LinkEndpoint interface. It saves the dispatcher
diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go
index 2b3741276..f5a05929f 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()
}
diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go
index 54eb5322b..0a9b99f18 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++
}
@@ -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/stack/forwarder_test.go b/pkg/tcpip/stack/forwarder_test.go
index 8084d50bc..344d60baa 100644
--- a/pkg/tcpip/stack/forwarder_test.go
+++ b/pkg/tcpip/stack/forwarder_test.go
@@ -209,7 +209,7 @@ func (e *fwdTestLinkEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber
// 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)
+ e.dispatcher.DeliverNetworkPacket(remote, "" /* local */, protocol, pkt)
}
// Attach saves the stack network-layer dispatcher for use later when packets
diff --git a/pkg/tcpip/stack/iptables.go b/pkg/tcpip/stack/iptables.go
index 443423b3c..709ede3fa 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"
@@ -314,7 +313,7 @@ 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
}
@@ -335,47 +334,3 @@ 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..a3bd3e700 100644
--- a/pkg/tcpip/stack/iptables_types.go
+++ b/pkg/tcpip/stack/iptables_types.go
@@ -15,7 +15,10 @@
package stack
import (
+ "strings"
+
"gvisor.dev/gvisor/pkg/tcpip"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
)
// A Hook specifies one of the hooks built into the network stack.
@@ -159,6 +162,16 @@ 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 +186,55 @@ 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.
diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go
index 54103fdb3..05646e5e2 100644
--- a/pkg/tcpip/stack/nic.go
+++ b/pkg/tcpip/stack/nic.go
@@ -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.
@@ -1240,7 +1240,7 @@ func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.Link
}
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
}
@@ -1304,13 +1304,16 @@ func (n *NIC) forwardPacket(r *Route, protocol tcpip.NetworkProtocolNumber, pkt
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
diff --git a/pkg/tcpip/stack/nic_test.go b/pkg/tcpip/stack/nic_test.go
index d672fc157..b01b3f476 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/registration.go b/pkg/tcpip/stack/registration.go
index b331427c6..db89234e8 100644
--- a/pkg/tcpip/stack/registration.go
+++ b/pkg/tcpip/stack/registration.go
@@ -240,16 +240,17 @@ 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.
+ // 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.
+ // 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.
@@ -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
@@ -382,9 +383,8 @@ 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
@@ -392,7 +392,8 @@ type LinkEndpoint interface {
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
diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go
index 150297ab9..3d0e5cc6e 100644
--- a/pkg/tcpip/stack/route.go
+++ b/pkg/tcpip/stack/route.go
@@ -158,12 +158,15 @@ func (r *Route) WritePacket(gso *GSO, params NetworkHeaderParams, pkt PacketBuff
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))
@@ -198,12 +204,15 @@ func (r *Route) WriteHeaderIncludedPacket(pkt PacketBuffer) *tcpip.Error {
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 b39ffa9fb..0ab4c3e19 100644
--- a/pkg/tcpip/stack/stack.go
+++ b/pkg/tcpip/stack/stack.go
@@ -235,11 +235,11 @@ type RcvBufAutoTuneParams struct {
// was started.
MeasureTime time.Time
- // CopiedBytes is the number of bytes copied to user space since
+ // CopiedBytes is the number of bytes copied to userspace since
// this measure began.
CopiedBytes int
- // PrevCopiedBytes is the number of bytes copied to user space in
+ // PrevCopiedBytes is the number of bytes copied to userspace in
// the previous RTT period.
PrevCopiedBytes int
diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go
index 1ca4088c9..b7b227328 100644
--- a/pkg/tcpip/tcpip.go
+++ b/pkg/tcpip/tcpip.go
@@ -110,6 +110,71 @@ var (
ErrAddressFamilyNotSupported = &Error{msg: "address family not supported by protocol"}
)
+var messageToError map[string]*Error
+
+var populate sync.Once
+
+// StringToError converts an error message to the error.
+func StringToError(s string) *Error {
+ populate.Do(func() {
+ var errors = []*Error{
+ ErrUnknownProtocol,
+ ErrUnknownNICID,
+ ErrUnknownDevice,
+ ErrUnknownProtocolOption,
+ ErrDuplicateNICID,
+ ErrDuplicateAddress,
+ ErrNoRoute,
+ ErrBadLinkEndpoint,
+ ErrAlreadyBound,
+ ErrInvalidEndpointState,
+ ErrAlreadyConnecting,
+ ErrAlreadyConnected,
+ ErrNoPortAvailable,
+ ErrPortInUse,
+ ErrBadLocalAddress,
+ ErrClosedForSend,
+ ErrClosedForReceive,
+ ErrWouldBlock,
+ ErrConnectionRefused,
+ ErrTimeout,
+ ErrAborted,
+ ErrConnectStarted,
+ ErrDestinationRequired,
+ ErrNotSupported,
+ ErrQueueSizeNotSupported,
+ ErrNotConnected,
+ ErrConnectionReset,
+ ErrConnectionAborted,
+ ErrNoSuchFile,
+ ErrInvalidOptionValue,
+ ErrNoLinkAddress,
+ ErrBadAddress,
+ ErrNetworkUnreachable,
+ ErrMessageTooLong,
+ ErrNoBufferSpace,
+ ErrBroadcastDisabled,
+ ErrNotPermitted,
+ ErrAddressFamilyNotSupported,
+ }
+
+ messageToError = make(map[string]*Error)
+ for _, e := range errors {
+ if messageToError[e.String()] != nil {
+ panic("tcpip errors with duplicated message: " + e.String())
+ }
+ messageToError[e.String()] = e
+ }
+ })
+
+ e, ok := messageToError[s]
+ if !ok {
+ panic("unknown error message: " + s)
+ }
+
+ return e
+}
+
// Errors related to Subnet
var (
errSubnetLengthMismatch = errors.New("subnet length of address and mask differ")
@@ -622,6 +687,19 @@ const (
//
// A zero value indicates the default.
TTLOption
+
+ // TCPSynCountOption is used by SetSockOpt/GetSockOpt to specify the number of
+ // SYN retransmits that TCP should send before aborting the attempt to
+ // connect. It cannot exceed 255.
+ //
+ // NOTE: This option is currently only stubbed out and is no-op.
+ TCPSynCountOption
+
+ // TCPWindowClampOption is used by SetSockOpt/GetSockOpt to bound the size
+ // of the advertised window to this value.
+ //
+ // NOTE: This option is currently only stubed out and is a no-op
+ TCPWindowClampOption
)
// ErrorOption is used in GetSockOpt to specify that the last error reported by
@@ -685,11 +763,23 @@ type TCPDeferAcceptOption time.Duration
// default MinRTO used by the Stack.
type TCPMinRTOOption time.Duration
+// TCPMaxRTOOption is use by SetSockOpt/GetSockOpt to allow overriding
+// default MaxRTO used by the Stack.
+type TCPMaxRTOOption time.Duration
+
+// TCPMaxRetriesOption is used by SetSockOpt/GetSockOpt to set/get the
+// maximum number of retransmits after which we time out the connection.
+type TCPMaxRetriesOption uint64
+
// TCPSynRcvdCountThresholdOption is used by SetSockOpt/GetSockOpt to specify
// the number of endpoints that can be in SYN-RCVD state before the stack
// switches to using SYN cookies.
type TCPSynRcvdCountThresholdOption uint64
+// TCPSynRetriesOption is used by SetSockOpt/GetSockOpt to specify stack-wide
+// default for number of times SYN is retransmitted before aborting a connect.
+type TCPSynRetriesOption uint8
+
// MulticastInterfaceOption is used by SetSockOpt/GetSockOpt to specify a
// default interface for multicast.
type MulticastInterfaceOption struct {
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/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go
index a7e088d4e..e4a06c9e1 100644
--- a/pkg/tcpip/transport/tcp/connect.go
+++ b/pkg/tcpip/transport/tcp/connect.go
@@ -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/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go
index 07d3e64c8..b5ba972f1 100644
--- a/pkg/tcpip/transport/tcp/endpoint.go
+++ b/pkg/tcpip/transport/tcp/endpoint.go
@@ -470,6 +470,17 @@ type endpoint struct {
// for this endpoint using the TCP_MAXSEG setsockopt.
userMSS uint16
+ // maxSynRetries is the maximum number of SYN retransmits that TCP should
+ // send before aborting the attempt to connect. It cannot exceed 255.
+ //
+ // NOTE: This is currently a no-op and does not change the SYN
+ // retransmissions.
+ maxSynRetries uint8
+
+ // windowClamp is used to bound the size of the advertised window to
+ // this value.
+ windowClamp uint32
+
// The following fields are used to manage the send buffer. When
// segments are ready to be sent, they are added to sndQueue and the
// protocol goroutine is signaled via sndWaker.
@@ -795,8 +806,10 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
interval: 75 * time.Second,
count: 9,
},
- uniqueID: s.UniqueID(),
- txHash: s.Rand().Uint32(),
+ uniqueID: s.UniqueID(),
+ txHash: s.Rand().Uint32(),
+ windowClamp: DefaultReceiveBufferSize,
+ maxSynRetries: DefaultSynRetries,
}
var ss SendBufferSizeOption
@@ -829,6 +842,11 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue
e.tcpLingerTimeout = time.Duration(tcpLT)
}
+ var synRetries tcpip.TCPSynRetriesOption
+ if err := s.TransportProtocolOption(ProtocolNumber, &synRetries); err == nil {
+ e.maxSynRetries = uint8(synRetries)
+ }
+
if p := s.GetTCPProbe(); p != nil {
e.probe = p
}
@@ -1079,7 +1097,7 @@ func (e *endpoint) initialReceiveWindow() int {
}
// ModerateRecvBuf adjusts the receive buffer and the advertised window
-// based on the number of bytes copied to user space.
+// based on the number of bytes copied to userspace.
func (e *endpoint) ModerateRecvBuf(copied int) {
e.LockUser()
defer e.UnlockUser()
@@ -1603,6 +1621,36 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
e.ttl = uint8(v)
e.UnlockUser()
+ case tcpip.TCPSynCountOption:
+ if v < 1 || v > 255 {
+ return tcpip.ErrInvalidOptionValue
+ }
+ e.LockUser()
+ e.maxSynRetries = uint8(v)
+ e.UnlockUser()
+
+ case tcpip.TCPWindowClampOption:
+ if v == 0 {
+ e.LockUser()
+ switch e.EndpointState() {
+ case StateClose, StateInitial:
+ e.windowClamp = 0
+ e.UnlockUser()
+ return nil
+ default:
+ e.UnlockUser()
+ return tcpip.ErrInvalidOptionValue
+ }
+ }
+ var rs ReceiveBufferSizeOption
+ if err := e.stack.TransportProtocolOption(ProtocolNumber, &rs); err == nil {
+ if v < rs.Min/2 {
+ v = rs.Min / 2
+ }
+ }
+ e.LockUser()
+ e.windowClamp = uint32(v)
+ e.UnlockUser()
}
return nil
}
@@ -1826,6 +1874,18 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
e.UnlockUser()
return v, nil
+ case tcpip.TCPSynCountOption:
+ e.LockUser()
+ v := int(e.maxSynRetries)
+ e.UnlockUser()
+ return v, nil
+
+ case tcpip.TCPWindowClampOption:
+ e.LockUser()
+ v := int(e.windowClamp)
+ e.UnlockUser()
+ return v, nil
+
default:
return -1, tcpip.ErrUnknownProtocolOption
}
diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go
index 8b7562396..fc43c11e2 100644
--- a/pkg/tcpip/transport/tcp/endpoint_state.go
+++ b/pkg/tcpip/transport/tcp/endpoint_state.go
@@ -314,7 +314,7 @@ func (e *endpoint) loadLastError(s string) {
return
}
- e.lastError = loadError(s)
+ e.lastError = tcpip.StringToError(s)
}
// saveHardError is invoked by stateify.
@@ -332,71 +332,7 @@ func (e *EndpointInfo) loadHardError(s string) {
return
}
- e.HardError = loadError(s)
-}
-
-var messageToError map[string]*tcpip.Error
-
-var populate sync.Once
-
-func loadError(s string) *tcpip.Error {
- populate.Do(func() {
- var errors = []*tcpip.Error{
- tcpip.ErrUnknownProtocol,
- tcpip.ErrUnknownNICID,
- tcpip.ErrUnknownDevice,
- tcpip.ErrUnknownProtocolOption,
- tcpip.ErrDuplicateNICID,
- tcpip.ErrDuplicateAddress,
- tcpip.ErrNoRoute,
- tcpip.ErrBadLinkEndpoint,
- tcpip.ErrAlreadyBound,
- tcpip.ErrInvalidEndpointState,
- tcpip.ErrAlreadyConnecting,
- tcpip.ErrAlreadyConnected,
- tcpip.ErrNoPortAvailable,
- tcpip.ErrPortInUse,
- tcpip.ErrBadLocalAddress,
- tcpip.ErrClosedForSend,
- tcpip.ErrClosedForReceive,
- tcpip.ErrWouldBlock,
- tcpip.ErrConnectionRefused,
- tcpip.ErrTimeout,
- tcpip.ErrAborted,
- tcpip.ErrConnectStarted,
- tcpip.ErrDestinationRequired,
- tcpip.ErrNotSupported,
- tcpip.ErrQueueSizeNotSupported,
- tcpip.ErrNotConnected,
- tcpip.ErrConnectionReset,
- tcpip.ErrConnectionAborted,
- tcpip.ErrNoSuchFile,
- tcpip.ErrInvalidOptionValue,
- tcpip.ErrNoLinkAddress,
- tcpip.ErrBadAddress,
- tcpip.ErrNetworkUnreachable,
- tcpip.ErrMessageTooLong,
- tcpip.ErrNoBufferSpace,
- tcpip.ErrBroadcastDisabled,
- tcpip.ErrNotPermitted,
- tcpip.ErrAddressFamilyNotSupported,
- }
-
- messageToError = make(map[string]*tcpip.Error)
- for _, e := range errors {
- if messageToError[e.String()] != nil {
- panic("tcpip errors with duplicated message: " + e.String())
- }
- messageToError[e.String()] = e
- }
- })
-
- e, ok := messageToError[s]
- if !ok {
- panic("unknown error message: " + s)
- }
-
- return e
+ e.HardError = tcpip.StringToError(s)
}
// saveMeasureTime is invoked by stateify.
diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go
index cfd9a4e8e..2a2a7ddeb 100644
--- a/pkg/tcpip/transport/tcp/protocol.go
+++ b/pkg/tcpip/transport/tcp/protocol.go
@@ -64,6 +64,10 @@ const (
// DefaultTCPTimeWaitTimeout is the amount of time that sockets linger
// in TIME_WAIT state before being marked closed.
DefaultTCPTimeWaitTimeout = 60 * time.Second
+
+ // DefaultSynRetries is the default value for the number of SYN retransmits
+ // before a connect is aborted.
+ DefaultSynRetries = 6
)
// SACKEnabled option can be used to enable SACK support in the TCP
@@ -163,7 +167,10 @@ type protocol struct {
tcpLingerTimeout time.Duration
tcpTimeWaitTimeout time.Duration
minRTO time.Duration
+ maxRTO time.Duration
+ maxRetries uint32
synRcvdCount synRcvdCounter
+ synRetries uint8
dispatcher *dispatcher
}
@@ -340,12 +347,36 @@ func (p *protocol) SetOption(option interface{}) *tcpip.Error {
p.mu.Unlock()
return nil
+ case tcpip.TCPMaxRTOOption:
+ if v < 0 {
+ v = tcpip.TCPMaxRTOOption(MaxRTO)
+ }
+ p.mu.Lock()
+ p.maxRTO = time.Duration(v)
+ p.mu.Unlock()
+ return nil
+
+ case tcpip.TCPMaxRetriesOption:
+ p.mu.Lock()
+ p.maxRetries = uint32(v)
+ p.mu.Unlock()
+ return nil
+
case tcpip.TCPSynRcvdCountThresholdOption:
p.mu.Lock()
p.synRcvdCount.SetThreshold(uint64(v))
p.mu.Unlock()
return nil
+ case tcpip.TCPSynRetriesOption:
+ if v < 1 || v > 255 {
+ return tcpip.ErrInvalidOptionValue
+ }
+ p.mu.Lock()
+ p.synRetries = uint8(v)
+ p.mu.Unlock()
+ return nil
+
default:
return tcpip.ErrUnknownProtocolOption
}
@@ -414,12 +445,30 @@ func (p *protocol) Option(option interface{}) *tcpip.Error {
p.mu.RUnlock()
return nil
+ case *tcpip.TCPMaxRTOOption:
+ p.mu.RLock()
+ *v = tcpip.TCPMaxRTOOption(p.maxRTO)
+ p.mu.RUnlock()
+ return nil
+
+ case *tcpip.TCPMaxRetriesOption:
+ p.mu.RLock()
+ *v = tcpip.TCPMaxRetriesOption(p.maxRetries)
+ p.mu.RUnlock()
+ return nil
+
case *tcpip.TCPSynRcvdCountThresholdOption:
p.mu.RLock()
*v = tcpip.TCPSynRcvdCountThresholdOption(p.synRcvdCount.Threshold())
p.mu.RUnlock()
return nil
+ case *tcpip.TCPSynRetriesOption:
+ p.mu.RLock()
+ *v = tcpip.TCPSynRetriesOption(p.synRetries)
+ p.mu.RUnlock()
+ return nil
+
default:
return tcpip.ErrUnknownProtocolOption
}
@@ -452,6 +501,9 @@ func NewProtocol() stack.TransportProtocol {
tcpTimeWaitTimeout: DefaultTCPTimeWaitTimeout,
synRcvdCount: synRcvdCounter{threshold: SynRcvdCountThreshold},
dispatcher: newDispatcher(runtime.GOMAXPROCS(0)),
+ synRetries: DefaultSynRetries,
minRTO: MinRTO,
+ maxRTO: MaxRTO,
+ maxRetries: MaxRetries,
}
}
diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go
index 9e547a221..06dc9b7d7 100644
--- a/pkg/tcpip/transport/tcp/snd.go
+++ b/pkg/tcpip/transport/tcp/snd.go
@@ -43,7 +43,8 @@ const (
nDupAckThreshold = 3
// MaxRetries is the maximum number of probe retries sender does
- // before timing out the connection, Linux default TCP_RETR2.
+ // before timing out the connection.
+ // Linux default TCP_RETR2, net.ipv4.tcp_retries2.
MaxRetries = 15
)
@@ -165,6 +166,12 @@ type sender struct {
// minRTO is the minimum permitted value for sender.rto.
minRTO time.Duration
+ // maxRTO is the maximum permitted value for sender.rto.
+ maxRTO time.Duration
+
+ // maxRetries is the maximum permitted retransmissions.
+ maxRetries uint32
+
// maxPayloadSize is the maximum size of the payload of a given segment.
// It is initialized on demand.
maxPayloadSize int
@@ -276,12 +283,24 @@ func newSender(ep *endpoint, iss, irs seqnum.Value, sndWnd seqnum.Size, mss uint
// etc.
s.ep.scoreboard = NewSACKScoreboard(uint16(s.maxPayloadSize), iss)
- // Get Stack wide minRTO.
- var v tcpip.TCPMinRTOOption
- if err := ep.stack.TransportProtocolOption(ProtocolNumber, &v); err != nil {
+ // Get Stack wide config.
+ var minRTO tcpip.TCPMinRTOOption
+ if err := ep.stack.TransportProtocolOption(ProtocolNumber, &minRTO); err != nil {
panic(fmt.Sprintf("unable to get minRTO from stack: %s", err))
}
- s.minRTO = time.Duration(v)
+ s.minRTO = time.Duration(minRTO)
+
+ var maxRTO tcpip.TCPMaxRTOOption
+ if err := ep.stack.TransportProtocolOption(ProtocolNumber, &maxRTO); err != nil {
+ panic(fmt.Sprintf("unable to get maxRTO from stack: %s", err))
+ }
+ s.maxRTO = time.Duration(maxRTO)
+
+ var maxRetries tcpip.TCPMaxRetriesOption
+ if err := ep.stack.TransportProtocolOption(ProtocolNumber, &maxRetries); err != nil {
+ panic(fmt.Sprintf("unable to get maxRetries from stack: %s", err))
+ }
+ s.maxRetries = uint32(maxRetries)
return s
}
@@ -485,7 +504,7 @@ func (s *sender) retransmitTimerExpired() bool {
}
elapsed := time.Since(s.firstRetransmittedSegXmitTime)
- remaining := MaxRTO
+ remaining := s.maxRTO
if uto != 0 {
// Cap to the user specified timeout if one is specified.
remaining = uto - elapsed
@@ -494,24 +513,17 @@ func (s *sender) retransmitTimerExpired() bool {
// Always honor the user-timeout irrespective of whether the zero
// window probes were acknowledged.
// net/ipv4/tcp_timer.c::tcp_probe_timer()
- if remaining <= 0 || s.unackZeroWindowProbes >= MaxRetries {
+ if remaining <= 0 || s.unackZeroWindowProbes >= s.maxRetries {
return false
}
- if s.rto >= MaxRTO {
- // RFC 1122 section: 4.2.2.17
- // A TCP MAY keep its offered receive window closed
- // indefinitely. As long as the receiving TCP continues to
- // send acknowledgments in response to the probe segments, the
- // sending TCP MUST allow the connection to stay open.
- if !(s.zeroWindowProbing && s.unackZeroWindowProbes == 0) {
- return false
- }
- }
-
// Set new timeout. The timer will be restarted by the call to sendData
// below.
s.rto *= 2
+ // Cap the RTO as per RFC 1122 4.2.3.1, RFC 6298 5.5
+ if s.rto > s.maxRTO {
+ s.rto = s.maxRTO
+ }
// Cap RTO to remaining time.
if s.rto > remaining {
@@ -565,9 +577,20 @@ func (s *sender) retransmitTimerExpired() bool {
// send.
if s.zeroWindowProbing {
s.sendZeroWindowProbe()
+ // RFC 1122 4.2.2.17: A TCP MAY keep its offered receive window closed
+ // indefinitely. As long as the receiving TCP continues to send
+ // acknowledgments in response to the probe segments, the sending TCP
+ // MUST allow the connection to stay open.
return true
}
+ seg := s.writeNext
+ // RFC 1122 4.2.3.5: Close the connection when the number of
+ // retransmissions for this segment is beyond a limit.
+ if seg != nil && seg.xmitCount > s.maxRetries {
+ return false
+ }
+
s.sendData()
return true
diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go
index d2c90ebd5..6ef32a1b3 100644
--- a/pkg/tcpip/transport/tcp/tcp_test.go
+++ b/pkg/tcpip/transport/tcp/tcp_test.go
@@ -2994,6 +2994,101 @@ func TestSendOnResetConnection(t *testing.T) {
}
}
+// TestMaxRetransmitsTimeout tests if the connection is timed out after
+// a segment has been retransmitted MaxRetries times.
+func TestMaxRetransmitsTimeout(t *testing.T) {
+ c := context.New(t, defaultMTU)
+ defer c.Cleanup()
+
+ const numRetries = 2
+ if err := c.Stack().SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.TCPMaxRetriesOption(numRetries)); err != nil {
+ t.Fatalf("could not set protocol option MaxRetries.\n")
+ }
+
+ c.CreateConnected(789 /* iss */, 30000 /* rcvWnd */, -1 /* epRcvBuf */)
+
+ waitEntry, notifyCh := waiter.NewChannelEntry(nil)
+ c.WQ.EventRegister(&waitEntry, waiter.EventHUp)
+ defer c.WQ.EventUnregister(&waitEntry)
+
+ _, _, err := c.EP.Write(tcpip.SlicePayload(buffer.NewView(1)), tcpip.WriteOptions{})
+ if err != nil {
+ t.Fatalf("Write failed: %v", err)
+ }
+
+ // Expect first transmit and MaxRetries retransmits.
+ for i := 0; i < numRetries+1; i++ {
+ checker.IPv4(t, c.GetPacket(),
+ checker.TCP(
+ checker.DstPort(context.TestPort),
+ checker.TCPFlags(header.TCPFlagAck|header.TCPFlagPsh),
+ ),
+ )
+ }
+ // Wait for the connection to timeout after MaxRetries retransmits.
+ initRTO := 1 * time.Second
+ select {
+ case <-notifyCh:
+ case <-time.After((2 << numRetries) * initRTO):
+ t.Fatalf("connection still alive after maximum retransmits.\n")
+ }
+
+ // Send an ACK and expect a RST as the connection would have been closed.
+ c.SendPacket(nil, &context.Headers{
+ SrcPort: context.TestPort,
+ DstPort: c.Port,
+ Flags: header.TCPFlagAck,
+ })
+
+ checker.IPv4(t, c.GetPacket(),
+ checker.TCP(
+ checker.DstPort(context.TestPort),
+ checker.TCPFlags(header.TCPFlagRst),
+ ),
+ )
+
+ if got := c.Stack().Stats().TCP.EstablishedTimedout.Value(); got != 1 {
+ t.Errorf("got c.Stack().Stats().TCP.EstablishedTimedout.Value() = %v, want = 1", got)
+ }
+}
+
+// TestMaxRTO tests if the retransmit interval caps to MaxRTO.
+func TestMaxRTO(t *testing.T) {
+ c := context.New(t, defaultMTU)
+ defer c.Cleanup()
+
+ rto := 1 * time.Second
+ if err := c.Stack().SetTransportProtocolOption(tcp.ProtocolNumber, tcpip.TCPMaxRTOOption(rto)); err != nil {
+ t.Fatalf("c.stack.SetTransportProtocolOption(tcp, tcpip.TCPMaxRTO(%d) failed: %s", rto, err)
+ }
+
+ c.CreateConnected(789 /* iss */, 30000 /* rcvWnd */, -1 /* epRcvBuf */)
+
+ _, _, err := c.EP.Write(tcpip.SlicePayload(buffer.NewView(1)), tcpip.WriteOptions{})
+ if err != nil {
+ t.Fatalf("Write failed: %v", err)
+ }
+ checker.IPv4(t, c.GetPacket(),
+ checker.TCP(
+ checker.DstPort(context.TestPort),
+ checker.TCPFlagsMatch(header.TCPFlagAck, ^uint8(header.TCPFlagPsh)),
+ ),
+ )
+ const numRetransmits = 2
+ for i := 0; i < numRetransmits; i++ {
+ start := time.Now()
+ checker.IPv4(t, c.GetPacket(),
+ checker.TCP(
+ checker.DstPort(context.TestPort),
+ checker.TCPFlagsMatch(header.TCPFlagAck, ^uint8(header.TCPFlagPsh)),
+ ),
+ )
+ if time.Since(start).Round(time.Second).Seconds() != rto.Seconds() {
+ t.Errorf("Retransmit interval not capped to MaxRTO.\n")
+ }
+ }
+}
+
func TestFinImmediately(t *testing.T) {
c := context.New(t, defaultMTU)
defer c.Cleanup()
@@ -5774,7 +5869,7 @@ func TestReceiveBufferAutoTuning(t *testing.T) {
// Invoke the moderation API. This is required for auto-tuning
// to happen. This method is normally expected to be invoked
// from a higher layer than tcpip.Endpoint. So we simulate
- // copying to user-space by invoking it explicitly here.
+ // copying to userspace by invoking it explicitly here.
c.EP.ModerateRecvBuf(totalCopied)
// Now send a keep-alive packet to trigger an ACK so that we can
@@ -6605,9 +6700,16 @@ func TestTCPUserTimeout(t *testing.T) {
c.CreateConnected(789, 30000, -1 /* epRcvBuf */)
+ waitEntry, notifyCh := waiter.NewChannelEntry(nil)
+ c.WQ.EventRegister(&waitEntry, waiter.EventHUp)
+ defer c.WQ.EventUnregister(&waitEntry)
+
origEstablishedTimedout := c.Stack().Stats().TCP.EstablishedTimedout.Value()
- userTimeout := 50 * time.Millisecond
+ // Ensure that on the next retransmit timer fire, the user timeout has
+ // expired.
+ initRTO := 1 * time.Second
+ userTimeout := initRTO / 2
c.EP.SetSockOpt(tcpip.TCPUserTimeoutOption(userTimeout))
// Send some data and wait before ACKing it.
@@ -6627,9 +6729,13 @@ func TestTCPUserTimeout(t *testing.T) {
),
)
- // Wait for a little over the minimum retransmit timeout of 200ms for
- // the retransmitTimer to fire and close the connection.
- time.Sleep(tcp.MinRTO + 10*time.Millisecond)
+ // Wait for the retransmit timer to be fired and the user timeout to cause
+ // close of the connection.
+ select {
+ case <-notifyCh:
+ case <-time.After(2 * initRTO):
+ t.Fatalf("connection still alive after %s, should have been closed after :%s", 2*initRTO, userTimeout)
+ }
// No packet should be received as the connection should be silently
// closed due to timeout.
diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go
index 756ab913a..647b2067a 100644
--- a/pkg/tcpip/transport/udp/endpoint.go
+++ b/pkg/tcpip/transport/udp/endpoint.go
@@ -106,6 +106,9 @@ type endpoint struct {
bindToDevice tcpip.NICID
broadcast bool
+ lastErrorMu sync.Mutex `state:"nosave"`
+ lastError *tcpip.Error `state:".(string)"`
+
// Values used to reserve a port or register a transport endpoint.
// (which ever happens first).
boundBindToDevice tcpip.NICID
@@ -188,6 +191,15 @@ func (e *endpoint) UniqueID() uint64 {
return e.uniqueID
}
+func (e *endpoint) takeLastError() *tcpip.Error {
+ e.lastErrorMu.Lock()
+ defer e.lastErrorMu.Unlock()
+
+ err := e.lastError
+ e.lastError = nil
+ return err
+}
+
// Abort implements stack.TransportEndpoint.Abort.
func (e *endpoint) Abort() {
e.Close()
@@ -243,6 +255,10 @@ func (e *endpoint) IPTables() (stack.IPTables, error) {
// 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) {
+ if err := e.takeLastError(); err != nil {
+ return buffer.View{}, tcpip.ControlMessages{}, err
+ }
+
e.rcvMu.Lock()
if e.rcvList.Empty() {
@@ -382,6 +398,10 @@ func (e *endpoint) Write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c
}
func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-chan struct{}, *tcpip.Error) {
+ if err := e.takeLastError(); err != nil {
+ return 0, nil, err
+ }
+
// MSG_MORE is unimplemented. (This also means that MSG_EOR is a no-op.)
if opts.More {
return 0, nil, tcpip.ErrInvalidOptionValue
@@ -853,6 +873,7 @@ func (e *endpoint) GetSockOptInt(opt tcpip.SockOptInt) (int, *tcpip.Error) {
func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
switch o := opt.(type) {
case tcpip.ErrorOption:
+ return e.takeLastError()
case *tcpip.MulticastInterfaceOption:
e.mu.Lock()
*o = tcpip.MulticastInterfaceOption{
@@ -1316,6 +1337,17 @@ 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) {
+ if typ == stack.ControlPortUnreachable {
+ e.mu.RLock()
+ defer e.mu.RUnlock()
+
+ if e.state == StateConnected {
+ e.lastErrorMu.Lock()
+ defer e.lastErrorMu.Unlock()
+
+ e.lastError = tcpip.ErrConnectionRefused
+ }
+ }
}
// State implements tcpip.Endpoint.State.
diff --git a/pkg/tcpip/transport/udp/endpoint_state.go b/pkg/tcpip/transport/udp/endpoint_state.go
index 466bd9381..851e6b635 100644
--- a/pkg/tcpip/transport/udp/endpoint_state.go
+++ b/pkg/tcpip/transport/udp/endpoint_state.go
@@ -37,6 +37,24 @@ func (u *udpPacket) loadData(data buffer.VectorisedView) {
u.data = data
}
+// saveLastError is invoked by stateify.
+func (e *endpoint) saveLastError() string {
+ if e.lastError == nil {
+ return ""
+ }
+
+ return e.lastError.String()
+}
+
+// loadLastError is invoked by stateify.
+func (e *endpoint) loadLastError(s string) {
+ if s == "" {
+ return
+ }
+
+ e.lastError = tcpip.StringToError(s)
+}
+
// beforeSave is invoked by stateify.
func (e *endpoint) beforeSave() {
// Stop incoming packets from being handled (and mutate endpoint state).