diff options
Diffstat (limited to 'pkg/tcpip/tcpip.go')
| -rw-r--r-- | pkg/tcpip/tcpip.go | 258 |
1 files changed, 243 insertions, 15 deletions
diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index c021c67ac..bd5eb89ca 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -57,6 +57,9 @@ type Error struct { // String implements fmt.Stringer.String. func (e *Error) String() string { + if e == nil { + return "<nil>" + } return e.msg } @@ -228,6 +231,13 @@ func (s *Subnet) Broadcast() Address { return Address(addr) } +// Equal returns true if s equals o. +// +// Needed to use cmp.Equal on Subnet as its fields are unexported. +func (s Subnet) Equal(o Subnet) bool { + return s == o +} + // NICID is a number that uniquely identifies a NIC. type NICID int32 @@ -252,7 +262,7 @@ type FullAddress struct { // This may not be used by all endpoint types. NIC NICID - // Addr is the network address. + // Addr is the network or link layer address. Addr Address // Port is the transport port. @@ -426,6 +436,26 @@ type Endpoint interface { // IPTables returns the iptables for this endpoint's stack. IPTables() (iptables.IPTables, error) + + // Info returns a copy to the transport endpoint info. + Info() EndpointInfo + + // Stats returns a reference to the endpoint stats. + Stats() EndpointStats +} + +// EndpointInfo is the interface implemented by each endpoint info struct. +type EndpointInfo interface { + // IsEndpointInfo is an empty method to implement the tcpip.EndpointInfo + // marker interface. + IsEndpointInfo() +} + +// EndpointStats is the interface implemented by each endpoint stats struct. +type EndpointStats interface { + // IsEndpointStats is an empty method to implement the tcpip.EndpointStats + // marker interface. + IsEndpointStats() } // WriteOptions contains options for Endpoint.Write. @@ -466,6 +496,11 @@ const ( // number of unread bytes in the output buffer should be returned. SendQueueSizeOption + // DelayOption is used by SetSockOpt/GetSockOpt to specify if data + // should be sent out immediately by the transport protocol. For TCP, + // it determines if the Nagle algorithm is on or off. + DelayOption + // TODO(b/137664753): convert all int socket options to be handled via // GetSockOptInt. ) @@ -478,11 +513,6 @@ type ErrorOption struct{} // socket is to be restricted to sending and receiving IPv6 packets only. type V6OnlyOption int -// DelayOption is used by SetSockOpt/GetSockOpt to specify if data should be -// sent out immediately by the transport protocol. For TCP, it determines if the -// Nagle algorithm is on or off. -type DelayOption int - // CorkOption is used by SetSockOpt/GetSockOpt to specify if data should be // held until segments are full by the TCP transport protocol. type CorkOption int @@ -495,6 +525,10 @@ type ReuseAddressOption int // to be bound to an identical socket address. type ReusePortOption int +// BindToDeviceOption is used by SetSockOpt/GetSockOpt to specify that sockets +// should bind only on a specific NIC. +type BindToDeviceOption string + // QuickAckOption is stubbed out in SetSockOpt/GetSockOpt. type QuickAckOption int @@ -546,6 +580,22 @@ type ModerateReceiveBufferOption bool // Maximum Segment Size(MSS) value as specified using the TCP_MAXSEG option. type MaxSegOption int +// TTLOption is used by SetSockOpt/GetSockOpt to control the default TTL/hop +// limit value for unicast messages. The default is protocol specific. +// +// A zero value indicates the default. +type TTLOption uint8 + +// TCPLingerTimeoutOption is used by SetSockOpt/GetSockOpt to set/get the +// maximum duration for which a socket lingers in the TCP_FIN_WAIT_2 state +// before being marked closed. +type TCPLingerTimeoutOption time.Duration + +// TCPTimeWaitTimeoutOption is used by SetSockOpt/GetSockOpt to set/get the +// maximum duration for which a socket lingers in the TIME_WAIT state +// before being marked closed. +type TCPTimeWaitTimeoutOption time.Duration + // MulticastTTLOption is used by SetSockOpt/GetSockOpt to control the default // TTL value for multicast messages. The default is 1. type MulticastTTLOption uint8 @@ -587,6 +637,18 @@ type OutOfBandInlineOption int // datagram sockets are allowed to send packets to a broadcast address. type BroadcastOption int +// DefaultTTLOption is used by stack.(*Stack).NetworkProtocolOption to specify +// a default TTL. +type DefaultTTLOption uint8 + +// IPv4TOSOption is used by SetSockOpt/GetSockOpt to specify TOS +// for all subsequent outgoing IPv4 packets from the endpoint. +type IPv4TOSOption uint8 + +// IPv6TrafficClassOption is used by SetSockOpt/GetSockOpt to specify TOS +// for all subsequent outgoing IPv6 packets from the endpoint. +type IPv6TrafficClassOption uint8 + // Route is a row in the routing table. It specifies through which NIC (and // gateway) sets of packets should be routed. A row is considered viable if the // masked target address matches the destination address in the row. @@ -628,6 +690,11 @@ func (s *StatCounter) Increment() { s.IncrementBy(1) } +// Decrement minuses one to the counter. +func (s *StatCounter) Decrement() { + s.IncrementBy(^uint64(0)) +} + // Value returns the current value of the counter. func (s *StatCounter) Value() uint64 { return atomic.LoadUint64(&s.count) @@ -816,6 +883,14 @@ type IPStats struct { // OutgoingPacketErrors is the total number of IP packets which failed // to write to a link-layer endpoint. OutgoingPacketErrors *StatCounter + + // MalformedPacketsReceived is the total number of IP Packets that were + // dropped due to the IP packet header failing validation checks. + MalformedPacketsReceived *StatCounter + + // MalformedFragmentsReceived is the total number of IP Fragments that were + // dropped due to the fragment failing validation checks. + MalformedFragmentsReceived *StatCounter } // TCPStats collects TCP-specific stats. @@ -828,6 +903,15 @@ type TCPStats struct { // successfully via Listen. PassiveConnectionOpenings *StatCounter + // CurrentEstablished is the number of TCP connections for which the + // current state is either ESTABLISHED or CLOSE-WAIT. + CurrentEstablished *StatCounter + + // EstablishedResets is the number of times TCP connections have made + // a direct transition to the CLOSED state from either the + // ESTABLISHED state or the CLOSE-WAIT state. + EstablishedResets *StatCounter + // ListenOverflowSynDrop is the number of times the listen queue overflowed // and a SYN was dropped. ListenOverflowSynDrop *StatCounter @@ -862,6 +946,9 @@ type TCPStats struct { // SegmentsSent is the number of TCP segments sent. SegmentsSent *StatCounter + // SegmentSendErrors is the number of TCP segments failed to be sent. + SegmentSendErrors *StatCounter + // ResetsSent is the number of TCP resets sent. ResetsSent *StatCounter @@ -914,6 +1001,9 @@ type UDPStats struct { // PacketsSent is the number of UDP datagrams sent via sendUDP. PacketsSent *StatCounter + + // PacketSendErrors is the number of datagrams failed to be sent. + PacketSendErrors *StatCounter } // Stats holds statistics about the networking stack. @@ -924,7 +1014,7 @@ type Stats struct { // stack that were for an unknown or unsupported protocol. UnknownProtocolRcvdPackets *StatCounter - // MalformedRcvPackets is the number of packets received by the stack + // MalformedRcvdPackets is the number of packets received by the stack // that were deemed malformed. MalformedRcvdPackets *StatCounter @@ -944,18 +1034,95 @@ type Stats struct { UDP UDPStats } +// ReceiveErrors collects packet receive errors within transport endpoint. +type ReceiveErrors struct { + // ReceiveBufferOverflow is the number of received packets dropped + // due to the receive buffer being full. + ReceiveBufferOverflow StatCounter + + // MalformedPacketsReceived is the number of incoming packets + // dropped due to the packet header being in a malformed state. + MalformedPacketsReceived StatCounter + + // ClosedReceiver is the number of received packets dropped because + // of receiving endpoint state being closed. + ClosedReceiver StatCounter +} + +// SendErrors collects packet send errors within the transport layer for +// an endpoint. +type SendErrors struct { + // SendToNetworkFailed is the number of packets failed to be written to + // the network endpoint. + SendToNetworkFailed StatCounter + + // NoRoute is the number of times we failed to resolve IP route. + NoRoute StatCounter + + // NoLinkAddr is the number of times we failed to resolve ARP. + NoLinkAddr StatCounter +} + +// ReadErrors collects segment read errors from an endpoint read call. +type ReadErrors struct { + // ReadClosed is the number of received packet drops because the endpoint + // was shutdown for read. + ReadClosed StatCounter + + // InvalidEndpointState is the number of times we found the endpoint state + // to be unexpected. + InvalidEndpointState StatCounter +} + +// WriteErrors collects packet write errors from an endpoint write call. +type WriteErrors struct { + // WriteClosed is the number of packet drops because the endpoint + // was shutdown for write. + WriteClosed StatCounter + + // InvalidEndpointState is the number of times we found the endpoint state + // to be unexpected. + InvalidEndpointState StatCounter + + // InvalidArgs is the number of times invalid input arguments were + // provided for endpoint Write call. + InvalidArgs StatCounter +} + +// TransportEndpointStats collects statistics about the endpoint. +type TransportEndpointStats struct { + // PacketsReceived is the number of successful packet receives. + PacketsReceived StatCounter + + // PacketsSent is the number of successful packet sends. + PacketsSent StatCounter + + // ReceiveErrors collects packet receive errors within transport layer. + ReceiveErrors ReceiveErrors + + // ReadErrors collects packet read errors from an endpoint read call. + ReadErrors ReadErrors + + // SendErrors collects packet send errors within the transport layer. + SendErrors SendErrors + + // WriteErrors collects packet write errors from an endpoint write call. + WriteErrors WriteErrors +} + +// IsEndpointStats is an empty method to implement the tcpip.EndpointStats +// marker interface. +func (*TransportEndpointStats) IsEndpointStats() {} + func fillIn(v reflect.Value) { for i := 0; i < v.NumField(); i++ { v := v.Field(i) - switch v.Kind() { - case reflect.Ptr: - if s := v.Addr().Interface().(**StatCounter); *s == nil { - *s = &StatCounter{} + if s, ok := v.Addr().Interface().(**StatCounter); ok { + if *s == nil { + *s = new(StatCounter) } - case reflect.Struct: + } else { fillIn(v) - default: - panic(fmt.Sprintf("unexpected type %s", v.Type())) } } } @@ -966,6 +1133,26 @@ func (s Stats) FillIn() Stats { return s } +// Clone returns a copy of the TransportEndpointStats by atomically reading +// each field. +func (src *TransportEndpointStats) Clone() TransportEndpointStats { + var dst TransportEndpointStats + clone(reflect.ValueOf(&dst).Elem(), reflect.ValueOf(src).Elem()) + return dst +} + +func clone(dst reflect.Value, src reflect.Value) { + for i := 0; i < dst.NumField(); i++ { + d := dst.Field(i) + s := src.Field(i) + if c, ok := s.Addr().Interface().(*StatCounter); ok { + d.Addr().Interface().(*StatCounter).IncrementBy(c.Value()) + } else { + clone(d, s) + } + } +} + // String implements the fmt.Stringer interface. func (a Address) String() string { switch len(a) { @@ -1091,6 +1278,47 @@ func (a AddressWithPrefix) String() string { return fmt.Sprintf("%s/%d", a.Address, a.PrefixLen) } +// Subnet converts the address and prefix into a Subnet value and returns it. +func (a AddressWithPrefix) Subnet() Subnet { + addrLen := len(a.Address) + if a.PrefixLen <= 0 { + return Subnet{ + address: Address(strings.Repeat("\x00", addrLen)), + mask: AddressMask(strings.Repeat("\x00", addrLen)), + } + } + if a.PrefixLen >= addrLen*8 { + return Subnet{ + address: a.Address, + mask: AddressMask(strings.Repeat("\xff", addrLen)), + } + } + + sa := make([]byte, addrLen) + sm := make([]byte, addrLen) + n := uint(a.PrefixLen) + for i := 0; i < addrLen; i++ { + if n >= 8 { + sa[i] = a.Address[i] + sm[i] = 0xff + n -= 8 + continue + } + sm[i] = ^byte(0xff >> n) + sa[i] = a.Address[i] & sm[i] + n = 0 + } + + // For extra caution, call NewSubnet rather than directly creating the Subnet + // value. If that fails it indicates a serious bug in this code, so panic is + // in order. + s, err := NewSubnet(Address(sa), AddressMask(sm)) + if err != nil { + panic("invalid subnet: " + err.Error()) + } + return s +} + // ProtocolAddress is an address and the network protocol it is associated // with. type ProtocolAddress struct { @@ -1111,8 +1339,8 @@ var ( // GetDanglingEndpoints returns all dangling endpoints. func GetDanglingEndpoints() []Endpoint { - es := make([]Endpoint, 0, len(danglingEndpoints)) danglingEndpointsMu.Lock() + es := make([]Endpoint, 0, len(danglingEndpoints)) for e := range danglingEndpoints { es = append(es, e) } |