Add support for TIME_WAIT timeout.

This change adds explicit support for honoring the 2MSL timeout
for sockets in TIME_WAIT state. It also adds support for the
TCP_LINGER2 option that allows modification of the FIN_WAIT2
state timeout duration for a given socket.

It also adds an option to modify the Stack wide TIME_WAIT timeout
but this is only for testing. On Linux this is fixed at 60s.

Further, we also now correctly process RST's in CLOSE_WAIT and
close the socket similar to linux without moving it to error
state.

We also now handle SYN in ESTABLISHED state as per
RFC5961#section-4.1. Earlier we would just drop these SYNs.
Which can result in some tests that pass on linux to fail on
gVisor.

Netstack now honors TIME_WAIT correctly as well as handles the
following cases correctly.

- TCP RSTs in TIME_WAIT are ignored.
- A duplicate TCP FIN during TIME_WAIT extends the TIME_WAIT
  and a dup ACK is sent in response to the FIN as the dup FIN
  indicates potential loss of the original final ACK.
- An out of order segment during TIME_WAIT generates a dup ACK.
- A new SYN w/ a sequence number > the highest sequence number
  in the previous connection closes the TIME_WAIT early and
  opens a new connection.

Further to make the SYN case work correctly the ISN (Initial
Sequence Number) generation for Netstack has been updated to
be as per RFC. Its not a pure random number anymore and follows
the recommendation in https://tools.ietf.org/html/rfc6528#page-3.

The current hash used is not a cryptographically secure hash
function. A separate change will update the hash function used
to Siphash similar to what is used in Linux.

PiperOrigin-RevId: 279106406

1 files changed, 159 insertions, 8 deletions

diff --git a/pkg/tcpip/transport/tcp/rcv.go b/pkg/tcpip/transport/tcp/rcv.go
index e90f9a7d9..068b90fb6 100644
--- a/pkg/tcpip/transport/tcp/rcv.go
+++ b/pkg/tcpip/transport/tcp/rcv.go
@@ -18,6 +18,7 @@ import (
 	"container/heap"
 	"time"
 
+	"gvisor.dev/gvisor/pkg/tcpip"
 	"gvisor.dev/gvisor/pkg/tcpip/header"
 	"gvisor.dev/gvisor/pkg/tcpip/seqnum"
 )
@@ -209,6 +210,11 @@ func (r *receiver) consumeSegment(s *segment, segSeq seqnum.Value, segLen seqnum
 		switch r.ep.state {
 		case StateFinWait1:
 			r.ep.state = StateFinWait2
+			// Notify protocol goroutine that we have received an
+			// ACK to our FIN so that it can start the FIN_WAIT2
+			// timer to abort connection if the other side does
+			// not close within 2MSL.
+			r.ep.notifyProtocolGoroutine(notifyClose)
 		case StateClosing:
 			r.ep.state = StateTimeWait
 		case StateLastAck:
@@ -253,23 +259,105 @@ func (r *receiver) updateRTT() {
 	r.ep.rcvListMu.Unlock()
 }
 
-// handleRcvdSegment handles TCP segments directed at the connection managed by
-// r as they arrive. It is called by the protocol main loop.
-func (r *receiver) handleRcvdSegment(s *segment) {
+func (r *receiver) handleRcvdSegmentClosing(s *segment, state EndpointState, closed bool) (drop bool, err *tcpip.Error) {
+	r.ep.rcvListMu.Lock()
+	rcvClosed := r.ep.rcvClosed || r.closed
+	r.ep.rcvListMu.Unlock()
+
+	// If we are in one of the shutdown states then we need to do
+	// additional checks before we try and process the segment.
+	switch state {
+	case StateCloseWait, StateClosing, StateLastAck:
+		if !s.sequenceNumber.LessThanEq(r.rcvNxt) {
+			s.decRef()
+			// Just drop the segment as we have
+			// already received a FIN and this
+			// segment is after the sequence number
+			// for the FIN.
+			return true, nil
+		}
+		fallthrough
+	case StateFinWait1:
+		fallthrough
+	case StateFinWait2:
+		// If we are closed for reads (either due to an
+		// incoming FIN or the user calling shutdown(..,
+		// SHUT_RD) then any data past the rcvNxt should
+		// trigger a RST.
+		endDataSeq := s.sequenceNumber.Add(seqnum.Size(s.data.Size()))
+		if rcvClosed && r.rcvNxt.LessThan(endDataSeq) {
+			s.decRef()
+			return true, tcpip.ErrConnectionAborted
+		}
+		if state == StateFinWait1 {
+			break
+		}
+
+		// If it's a retransmission of an old data segment
+		// or a pure ACK then allow it.
+		if s.sequenceNumber.Add(s.logicalLen()).LessThanEq(r.rcvNxt) ||
+			s.logicalLen() == 0 {
+			break
+		}
+
+		// In FIN-WAIT2 if the socket is fully
+		// closed(not owned by application on our end
+		// then the only acceptable segment is a
+		// FIN. Since FIN can technically also carry
+		// data we verify that the segment carrying a
+		// FIN ends at exactly e.rcvNxt+1.
+		//
+		// From RFC793 page 25.
+		//
+		// For sequence number purposes, the SYN is
+		// considered to occur before the first actual
+		// data octet of the segment in which it occurs,
+		// while the FIN is considered to occur after
+		// the last actual data octet in a segment in
+		// which it occurs.
+		if closed && (!s.flagIsSet(header.TCPFlagFin) || s.sequenceNumber.Add(s.logicalLen()) != r.rcvNxt+1) {
+			s.decRef()
+			return true, tcpip.ErrConnectionAborted
+		}
+	}
+
 	// We don't care about receive processing anymore if the receive side
 	// is closed.
-	if r.closed {
-		return
+	//
+	// NOTE: We still want to permit a FIN as it's possible only our
+	// end has closed and the peer is yet to send a FIN. Hence we
+	// compare only the payload.
+	segEnd := s.sequenceNumber.Add(seqnum.Size(s.data.Size()))
+	if rcvClosed && !segEnd.LessThanEq(r.rcvNxt) {
+		return true, nil
+	}
+	return false, nil
+}
+
+// handleRcvdSegment handles TCP segments directed at the connection managed by
+// r as they arrive. It is called by the protocol main loop.
+func (r *receiver) handleRcvdSegment(s *segment) (drop bool, err *tcpip.Error) {
+	r.ep.mu.RLock()
+	state := r.ep.state
+	closed := r.ep.closed
+	r.ep.mu.RUnlock()
+
+	if state != StateEstablished {
+		drop, err := r.handleRcvdSegmentClosing(s, state, closed)
+		if drop || err != nil {
+			return drop, err
+		}
 	}
 
 	segLen := seqnum.Size(s.data.Size())
 	segSeq := s.sequenceNumber
 
 	// If the sequence number range is outside the acceptable range, just
-	// send an ACK. This is according to RFC 793, page 37.
+	// send an ACK and stop further processing of the segment.
+	// This is according to RFC 793, page 68.
 	if !r.acceptable(segSeq, segLen) {
 		r.ep.snd.sendAck()
-		return
+		return true, nil
 	}
 
 	// Defer segment processing if it can't be consumed now.
@@ -288,7 +376,7 @@ func (r *receiver) handleRcvdSegment(s *segment) {
 			// have to retransmit.
 			r.ep.snd.sendAck()
 		}
-		return
+		return false, nil
 	}
 
 	// Since we consumed a segment update the receiver's RTT estimate
@@ -315,4 +403,67 @@ func (r *receiver) handleRcvdSegment(s *segment) {
 		r.pendingBufUsed -= s.logicalLen()
 		s.decRef()
 	}
+	return false, nil
+}
+
+// handleTimeWaitSegment handles inbound segments received when the endpoint
+// has entered the TIME_WAIT state.
+func (r *receiver) handleTimeWaitSegment(s *segment) (resetTimeWait bool, newSyn bool) {
+	segSeq := s.sequenceNumber
+	segLen := seqnum.Size(s.data.Size())
+
+	// Just silently drop any RST packets in TIME_WAIT. We do not support
+	// TIME_WAIT assasination as a result we confirm w/ fix 1 as described
+	// in https://tools.ietf.org/html/rfc1337#section-3.
+	if s.flagIsSet(header.TCPFlagRst) {
+		return false, false
+	}
+
+	// If it's a SYN and the sequence number is higher than any seen before
+	// for this connection then try and redirect it to a listening endpoint
+	// if available.
+	//
+	// RFC 1122:
+	//   "When a connection is [...] on TIME-WAIT state [...]
+	//   [a TCP] MAY accept a new SYN from the remote TCP to
+	//   reopen the connection directly, if it:
+
+	//    (1) assigns its initial sequence number for the new
+	//     connection to be larger than the largest sequence
+	//     number it used on the previous connection incarnation,
+	//     and
+
+	//    (2) returns to TIME-WAIT state if the SYN turns out
+	//      to be an old duplicate".
+	if s.flagIsSet(header.TCPFlagSyn) && r.rcvNxt.LessThan(segSeq) {
+
+		return false, true
+	}
+
+	// Drop the segment if it does not contain an ACK.
+	if !s.flagIsSet(header.TCPFlagAck) {
+		return false, false
+	}
+
+	// Update Timestamp if required. See RFC7323, section-4.3.
+	if r.ep.sendTSOk && s.parsedOptions.TS {
+		r.ep.updateRecentTimestamp(s.parsedOptions.TSVal, r.ep.snd.maxSentAck, segSeq)
+	}
+
+	if segSeq.Add(1) == r.rcvNxt && s.flagIsSet(header.TCPFlagFin) {
+		// If it's a FIN-ACK then resetTimeWait and send an ACK, as it
+		// indicates our final ACK could have been lost.
+		r.ep.snd.sendAck()
+		return true, false
+	}
+
+	// If the sequence number range is outside the acceptable range or
+	// carries data then just send an ACK. This is according to RFC 793,
+	// page 37.
+	//
+	// NOTE: In TIME_WAIT the only acceptable sequence number is rcvNxt.
+	if segSeq != r.rcvNxt || segLen != 0 {
+		r.ep.snd.sendAck()
+	}
+	return false, false
 }