receive: drop handshakes if queue lock is contended

If we're being delivered packets from multiple CPUs so quickly that the
ring lock is contended for CPU tries, then it's safe to assume that the
queue is near capacity anyway, so just drop the packet rather than
spinning. This helps deal with multicore DoS that can interfere with
data path performance. It _still_ does not completely fix the issue, but
it again chips away at it.

Reported-by: Streun Fabio <fstreun@student.ethz.ch>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Author: zx2c4

src/receive.c

@@ -563,9 +563,19 @@ void wg_packet_receive(struct wg_device *wg, struct sk_buff *skb)
 	case cpu_to_le32(MESSAGE_HANDSHAKE_INITIATION):
 	case cpu_to_le32(MESSAGE_HANDSHAKE_RESPONSE):
 	case cpu_to_le32(MESSAGE_HANDSHAKE_COOKIE): {
-		int cpu;
-		if (unlikely(!rng_is_initialized() ||
-			     ptr_ring_produce_bh(&wg->handshake_queue.ring, skb))) {
+		int cpu, ret = -EBUSY;
+
+		if (unlikely(!rng_is_initialized()))
+			goto drop;
+		if (atomic_read(&wg->handshake_queue_len) > MAX_QUEUED_INCOMING_HANDSHAKES / 2) {
+			if (spin_trylock_bh(&wg->handshake_queue.ring.producer_lock)) {
+				ret = __ptr_ring_produce(&wg->handshake_queue.ring, skb);
+				spin_unlock_bh(&wg->handshake_queue.ring.producer_lock);
+			}
+		} else
+			ret = ptr_ring_produce_bh(&wg->handshake_queue.ring, skb);
+		if (ret) {
+	drop:
 			net_dbg_skb_ratelimited("%s: Dropping handshake packet from %pISpfsc\n",
 						wg->dev->name, skb);
 			goto err;