Send NUD probes in a separate gorountine

Send NUD probes in another gorountine to free the thread of execution for
finishing the state transition. This is necessary to avoid deadlock where
sending and processing probes are done in the same call stack, such as loopback
and integration tests.

Fixes #4701

PiperOrigin-RevId: 340362481

1 files changed, 19 insertions, 2 deletions

diff --git a/pkg/tcpip/stack/neighbor_entry.go b/pkg/tcpip/stack/neighbor_entry.go
index aec77610d..493e48031 100644
--- a/pkg/tcpip/stack/neighbor_entry.go
+++ b/pkg/tcpip/stack/neighbor_entry.go
@@ -17,12 +17,19 @@ package stack
 import (
 	"fmt"
 	"sync"
+	"time"
 
 	"gvisor.dev/gvisor/pkg/sleep"
 	"gvisor.dev/gvisor/pkg/tcpip"
 	"gvisor.dev/gvisor/pkg/tcpip/header"
 )
 
+const (
+	// immediateDuration is a duration of zero for scheduling work that needs to
+	// be done immediately but asynchronously to avoid deadlock.
+	immediateDuration time.Duration = 0
+)
+
 // NeighborEntry describes a neighboring device in the local network.
 type NeighborEntry struct {
 	Addr           tcpip.Address
@@ -242,7 +249,12 @@ func (e *neighborEntry) setStateLocked(next NeighborState) {
 			e.job.Schedule(config.RetransmitTimer)
 		}
 
-		sendUnicastProbe()
+		// Send a probe in another gorountine to free this thread of execution
+		// for finishing the state transition. This is necessary to avoid
+		// deadlock where sending and processing probes are done synchronously,
+		// such as loopback and integration tests.
+		e.job = e.nic.stack.newJob(&e.mu, sendUnicastProbe)
+		e.job.Schedule(immediateDuration)
 
 	case Failed:
 		e.notifyWakersLocked()
@@ -324,7 +336,12 @@ func (e *neighborEntry) handlePacketQueuedLocked(localAddr tcpip.Address) {
 			e.job.Schedule(config.RetransmitTimer)
 		}
 
-		sendMulticastProbe()
+		// Send a probe in another gorountine to free this thread of execution
+		// for finishing the state transition. This is necessary to avoid
+		// deadlock where sending and processing probes are done synchronously,
+		// such as loopback and integration tests.
+		e.job = e.nic.stack.newJob(&e.mu, sendMulticastProbe)
+		e.job.Schedule(immediateDuration)
 
 	case Stale:
 		e.setStateLocked(Delay)