From 8db147b55423d7dbe5f9af4e6154eab2d19025e1 Mon Sep 17 00:00:00 2001
From: Sam Balana <sbalana@google.com>
Date: Fri, 23 Oct 2020 12:31:11 -0700
Subject: Wait before transitioning NUD entries from Probe to Failed

Wait an additional RetransmitTimer duration after the last probe before
transitioning to Failed. The previous implementation transitions immediately to
Failed after sending the last probe, which is erroneous behavior.

PiperOrigin-RevId: 338723794
---
 pkg/tcpip/stack/neighbor_entry.go      |  6 ---
 pkg/tcpip/stack/neighbor_entry_test.go | 82 ++++++++++++++++++++--------------
 2 files changed, 49 insertions(+), 39 deletions(-)

(limited to 'pkg/tcpip')

diff --git a/pkg/tcpip/stack/neighbor_entry.go b/pkg/tcpip/stack/neighbor_entry.go
index 0f16feaa5..1f0365460 100644
--- a/pkg/tcpip/stack/neighbor_entry.go
+++ b/pkg/tcpip/stack/neighbor_entry.go
@@ -238,12 +238,6 @@ func (e *neighborEntry) setStateLocked(next NeighborState) {
 			}
 
 			retryCounter++
-			if retryCounter == config.MaxUnicastProbes {
-				e.dispatchRemoveEventLocked()
-				e.setStateLocked(Failed)
-				return
-			}
-
 			e.job = e.nic.stack.newJob(&e.mu, sendUnicastProbe)
 			e.job.Schedule(config.RetransmitTimer)
 		}
diff --git a/pkg/tcpip/stack/neighbor_entry_test.go b/pkg/tcpip/stack/neighbor_entry_test.go
index bb148e96a..9de9f88a1 100644
--- a/pkg/tcpip/stack/neighbor_entry_test.go
+++ b/pkg/tcpip/stack/neighbor_entry_test.go
@@ -2938,10 +2938,31 @@ func TestEntryProbeToFailed(t *testing.T) {
 	c := DefaultNUDConfigurations()
 	c.MaxMulticastProbes = 3
 	c.MaxUnicastProbes = 3
+	c.DelayFirstProbeTime = c.RetransmitTimer
 	e, nudDisp, linkRes, clock := entryTestSetup(c)
 
 	e.mu.Lock()
 	e.handlePacketQueuedLocked(entryTestAddr2)
+	e.mu.Unlock()
+
+	{
+		wantProbes := []entryTestProbeInfo{
+			// Caused by the Unknown-to-Incomplete transition.
+			{
+				RemoteAddress: entryTestAddr1,
+				LocalAddress:  entryTestAddr2,
+			},
+		}
+		linkRes.mu.Lock()
+		diff := cmp.Diff(linkRes.probes, wantProbes)
+		linkRes.probes = nil
+		linkRes.mu.Unlock()
+		if diff != "" {
+			t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+		}
+	}
+
+	e.mu.Lock()
 	e.handleConfirmationLocked(entryTestLinkAddr1, ReachabilityConfirmationFlags{
 		Solicited: false,
 		Override:  false,
@@ -2950,36 +2971,37 @@ func TestEntryProbeToFailed(t *testing.T) {
 	e.handlePacketQueuedLocked(entryTestAddr2)
 	e.mu.Unlock()
 
-	waitFor := c.DelayFirstProbeTime + c.RetransmitTimer*time.Duration(c.MaxUnicastProbes)
-	clock.Advance(waitFor)
+	// Observe each probe sent while in the Probe state.
+	for i := uint32(0); i < c.MaxUnicastProbes; i++ {
+		clock.Advance(c.RetransmitTimer)
+		wantProbes := []entryTestProbeInfo{
+			{
+				RemoteAddress:     entryTestAddr1,
+				RemoteLinkAddress: entryTestLinkAddr1,
+			},
+		}
+		linkRes.mu.Lock()
+		diff := cmp.Diff(linkRes.probes, wantProbes)
+		linkRes.probes = nil
+		linkRes.mu.Unlock()
+		if diff != "" {
+			t.Fatalf("link address resolver probe #%d mismatch (-got, +want):\n%s", i+1, diff)
+		}
 
-	wantProbes := []entryTestProbeInfo{
-		// The first probe is caused by the Unknown-to-Incomplete transition.
-		{
-			RemoteAddress:     entryTestAddr1,
-			RemoteLinkAddress: tcpip.LinkAddress(""),
-			LocalAddress:      entryTestAddr2,
-		},
-		// The next three probe are caused by the Delay-to-Probe transition.
-		{
-			RemoteAddress:     entryTestAddr1,
-			RemoteLinkAddress: entryTestLinkAddr1,
-		},
-		{
-			RemoteAddress:     entryTestAddr1,
-			RemoteLinkAddress: entryTestLinkAddr1,
-		},
-		{
-			RemoteAddress:     entryTestAddr1,
-			RemoteLinkAddress: entryTestLinkAddr1,
-		},
+		e.mu.Lock()
+		if e.neigh.State != Probe {
+			t.Errorf("got e.neigh.State = %q, want = %q", e.neigh.State, Probe)
+		}
+		e.mu.Unlock()
 	}
-	linkRes.mu.Lock()
-	diff := cmp.Diff(linkRes.probes, wantProbes)
-	linkRes.mu.Unlock()
-	if diff != "" {
-		t.Fatalf("link address resolver probes mismatch (-got, +want):\n%s", diff)
+
+	// Wait for the last probe to expire, causing a transition to Failed.
+	clock.Advance(c.RetransmitTimer)
+	e.mu.Lock()
+	if e.neigh.State != Failed {
+		t.Errorf("got e.neigh.State = %q, want = %q", e.neigh.State, Failed)
 	}
+	e.mu.Unlock()
 
 	wantEvents := []testEntryEventInfo{
 		{
@@ -3023,12 +3045,6 @@ func TestEntryProbeToFailed(t *testing.T) {
 		t.Errorf("nud dispatcher events mismatch (-got, +want):\n%s", diff)
 	}
 	nudDisp.mu.Unlock()
-
-	e.mu.Lock()
-	if got, want := e.neigh.State, Failed; got != want {
-		t.Errorf("got e.neigh.State = %q, want = %q", got, want)
-	}
-	e.mu.Unlock()
 }
 
 func TestEntryFailedGetsDeleted(t *testing.T) {
-- 
cgit v1.2.3