Pass PacketBuffer as pointer.

Historically we've been passing PacketBuffer by shallow copying through out
the stack. Right now, this is only correct as the caller would not use
PacketBuffer after passing into the next layer in netstack.

With new buffer management effort in gVisor/netstack, PacketBuffer will
own a Buffer (to be added). Internally, both PacketBuffer and Buffer may
have pointers and shallow copying shouldn't be used.

Updates #2404.

PiperOrigin-RevId: 314610879

1 files changed, 2 insertions, 2 deletions

diff --git a/pkg/tcpip/stack/forwarder.go b/pkg/tcpip/stack/forwarder.go
index 6b64cd37f..3eff141e6 100644
--- a/pkg/tcpip/stack/forwarder.go
+++ b/pkg/tcpip/stack/forwarder.go
@@ -32,7 +32,7 @@ type pendingPacket struct {
 	nic   *NIC
 	route *Route
 	proto tcpip.NetworkProtocolNumber
-	pkt   PacketBuffer
+	pkt   *PacketBuffer
 }
 
 type forwardQueue struct {
@@ -50,7 +50,7 @@ func newForwardQueue() *forwardQueue {
 	return &forwardQueue{packets: make(map[<-chan struct{}][]*pendingPacket)}
 }
 
-func (f *forwardQueue) enqueue(ch <-chan struct{}, n *NIC, r *Route, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) {
+func (f *forwardQueue) enqueue(ch <-chan struct{}, n *NIC, r *Route, protocol tcpip.NetworkProtocolNumber, pkt *PacketBuffer) {
 	shouldWait := false
 
 	f.Lock()