Merge release-20210208.0-52-g845d0a65f (automated)

5 files changed, 40 insertions, 14 deletions

diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index cee8120ab..a632b8bcd 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -222,6 +222,7 @@ var Metrics = tcpip.Stats{
 		Retransmits:                        mustCreateMetric("/netstack/tcp/retransmits", "Number of TCP segments retransmitted."),
 		FastRecovery:                       mustCreateMetric("/netstack/tcp/fast_recovery", "Number of times fast recovery was used to recover from packet loss."),
 		SACKRecovery:                       mustCreateMetric("/netstack/tcp/sack_recovery", "Number of times SACK recovery was used to recover from packet loss."),
+		TLPRecovery:                        mustCreateMetric("/netstack/tcp/tlp_recovery", "Number of times tail loss probe triggers recovery from tail loss."),
 		SlowStartRetransmits:               mustCreateMetric("/netstack/tcp/slow_start_retransmits", "Number of segments retransmitted in slow start mode."),
 		FastRetransmit:                     mustCreateMetric("/netstack/tcp/fast_retransmit", "Number of TCP segments which were fast retransmitted."),
 		Timeouts:                           mustCreateMetric("/netstack/tcp/timeouts", "Number of times RTO expired."),
diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go
index d5c2b0487..ba063dc26 100644
--- a/pkg/tcpip/tcpip.go
+++ b/pkg/tcpip/tcpip.go
@@ -1731,6 +1731,10 @@ type TCPStats struct {
 	// recover from packet loss.
 	SACKRecovery *StatCounter
 
+	// TLPRecovery is the number of times recovery was accomplished by the tail
+	// loss probe.
+	TLPRecovery *StatCounter
+
 	// SlowStartRetransmits is the number of segments retransmitted in slow
 	// start.
 	SlowStartRetransmits *StatCounter
diff --git a/pkg/tcpip/transport/tcp/rack.go b/pkg/tcpip/transport/tcp/rack.go
index 9959b60b8..ecabeceb4 100644
--- a/pkg/tcpip/transport/tcp/rack.go
+++ b/pkg/tcpip/transport/tcp/rack.go
@@ -263,7 +263,10 @@ func (s *sender) probeTimerExpired() tcpip.Error {
 		}
 	}
 
-	s.postXmit(dataSent)
+	// Whether or not the probe was sent, the sender must arm the resend timer,
+	// not the probe timer. This ensures that the sender does not send repeated,
+	// back-to-back tail loss probes.
+	s.postXmit(dataSent, false /* shouldScheduleProbe */)
 	return nil
 }
 
@@ -477,7 +480,5 @@ func (rc *rackControl) DoRecovery(_ *segment, fastRetransmit bool) {
 		snd.sendSegment(seg)
 	}
 
-	// Rearm the RTO.
-	snd.resendTimer.enable(snd.rto)
-	snd.postXmit(dataSent)
+	snd.postXmit(dataSent, true /* shouldScheduleProbe */)
 }
diff --git a/pkg/tcpip/transport/tcp/sack_recovery.go b/pkg/tcpip/transport/tcp/sack_recovery.go
index 7e813fa96..9d406b0bc 100644
--- a/pkg/tcpip/transport/tcp/sack_recovery.go
+++ b/pkg/tcpip/transport/tcp/sack_recovery.go
@@ -116,5 +116,5 @@ func (sr *sackRecovery) DoRecovery(rcvdSeg *segment, fastRetransmit bool) {
 	// RFC 6675 recovery algorithm step C 1-5.
 	end := snd.sndUna.Add(snd.sndWnd)
 	dataSent := sr.handleSACKRecovery(snd.maxPayloadSize, end)
-	snd.postXmit(dataSent)
+	snd.postXmit(dataSent, true /* shouldScheduleProbe */)
 }
diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go
index 7911e6b85..83c8deb0e 100644
--- a/pkg/tcpip/transport/tcp/snd.go
+++ b/pkg/tcpip/transport/tcp/snd.go
@@ -966,7 +966,7 @@ func (s *sender) disableZeroWindowProbing() {
 	s.resendTimer.disable()
 }
 
-func (s *sender) postXmit(dataSent bool) {
+func (s *sender) postXmit(dataSent bool, shouldScheduleProbe bool) {
 	if dataSent {
 		// We sent data, so we should stop the keepalive timer to ensure
 		// that no keepalives are sent while there is pending data.
@@ -980,13 +980,22 @@ func (s *sender) postXmit(dataSent bool) {
 		s.enableZeroWindowProbing()
 	}
 
-	// Enable the timer if we have pending data and it's not enabled yet.
-	if !s.resendTimer.enabled() && s.sndUna != s.sndNxt {
-		s.resendTimer.enable(s.rto)
-	}
 	// If we have no more pending data, start the keepalive timer.
 	if s.sndUna == s.sndNxt {
 		s.ep.resetKeepaliveTimer(false)
+	} else {
+		// Enable timers if we have pending data.
+		if shouldScheduleProbe && s.shouldSchedulePTO() {
+			// Schedule PTO after transmitting new data that wasn't itself a TLP probe.
+			s.schedulePTO()
+		} else if !s.resendTimer.enabled() {
+			s.probeTimer.disable()
+			if s.outstanding > 0 {
+				// Enable the resend timer if it's not enabled yet and there is
+				// outstanding data.
+				s.resendTimer.enable(s.rto)
+			}
+		}
 	}
 }
 
@@ -1029,7 +1038,7 @@ func (s *sender) sendData() {
 		s.writeNext = seg.Next()
 	}
 
-	s.postXmit(dataSent)
+	s.postXmit(dataSent, true /* shouldScheduleProbe */)
 }
 
 func (s *sender) enterRecovery() {
@@ -1052,6 +1061,10 @@ func (s *sender) enterRecovery() {
 		s.ep.stack.Stats().TCP.SACKRecovery.Increment()
 		// Set TLPRxtOut to false according to
 		// https://tools.ietf.org/html/draft-ietf-tcpm-rack-08#section-7.6.1.
+		if s.rc.tlpRxtOut {
+			// The tail loss probe triggered recovery.
+			s.ep.stack.Stats().TCP.TLPRecovery.Increment()
+		}
 		s.rc.tlpRxtOut = false
 		return
 	}
@@ -1415,9 +1428,16 @@ func (s *sender) handleRcvdSegment(rcvdSeg *segment) {
 			s.updateRTO(elapsed)
 		}
 
-		// When an ack is received we must rearm the timer.
-		// RFC 6298 5.3
-		s.resendTimer.enable(s.rto)
+		if s.shouldSchedulePTO() {
+			// Schedule PTO upon receiving an ACK that cumulatively acknowledges data.
+			// See https://tools.ietf.org/html/draft-ietf-tcpm-rack-08#section-7.5.1.
+			s.schedulePTO()
+		} else {
+			// When an ack is received we must rearm the timer.
+			// RFC 6298 5.3
+			s.probeTimer.disable()
+			s.resendTimer.enable(s.rto)
+		}
 
 		// Remove all acknowledged data from the write list.
 		acked := s.sndUna.Size(ack)