Log cases where an onion failure cannot be attributed or interpreted

Create more visibility into these edge cases. The non-attributable
failure in particular can be used to disrupt sender operation and it is
therefore good to at least log these cases clearly.

Author: joostjager

lightning/src/ln/onion_utils.rs

@@ -1044,22 +1044,43 @@ where
 		let amt_to_forward = htlc_msat - route_hop.fee_msat;
 		htlc_msat = amt_to_forward;
 
-		let err_packet = match decrypt_onion_error_packet(&mut encrypted_packet, shared_secret) {
-			Ok(p) => p,
-			Err(_) => return,
-		};
+		let decrypt_result = decrypt_onion_error_packet(&mut encrypted_packet, shared_secret);
+
 		let um = gen_um_from_shared_secret(shared_secret.as_ref());
 		let mut hmac = HmacEngine::<Sha256>::new(&um);
-		hmac.input(&err_packet.encode()[32..]);
+		hmac.input(&encrypted_packet[32..]);
 
-		if !fixed_time_eq(&Hmac::from_engine(hmac).to_byte_array(), &err_packet.hmac) {
+		if !fixed_time_eq(&Hmac::from_engine(hmac).to_byte_array(), &encrypted_packet[..32]) {
 			return;
 		}
+
+		let err_packet = match decrypt_result {
+			Ok(p) => p,
+			Err(_) => {
+				log_warn!(logger, "Unreadable failure from {}", route_hop.pubkey);
+
+				let network_update = Some(NetworkUpdate::NodeFailure {
+					node_id: route_hop.pubkey,
+					is_permanent: true,
+				});
+				let short_channel_id = Some(route_hop.short_channel_id);
+				res = Some(FailureLearnings {
+					network_update,
+					short_channel_id,
+					payment_failed_permanently: is_from_final_node,
+					failed_within_blinded_path: false,
+				});
+				return;
+			},
+		};
+
 		let error_code_slice = match err_packet.failuremsg.get(0..2) {
 			Some(s) => s,
 			None => {
 				// Useless packet that we can't use but it passed HMAC, so it definitely came from the peer
 				// in question
+				log_warn!(logger, "Missing error code in failure from {}", route_hop.pubkey);
+
 				let network_update = Some(NetworkUpdate::NodeFailure {
 					node_id: route_hop.pubkey,
 					is_permanent: true,
@@ -1219,6 +1240,12 @@ where
 	} else {
 		// only not set either packet unparseable or hmac does not match with any
 		// payment not retryable only when garbage is from the final node
+		log_warn!(
+			logger,
+			"Non-attributable failure encountered on route {}",
+			path.hops.iter().map(|h| h.pubkey.to_string()).collect::<Vec<_>>().join("->")
+		);
+
 		DecodedOnionFailure {
 			network_update: None,
 			short_channel_id: None,
@@ -1768,7 +1795,7 @@ mod tests {
 
 	use crate::io;
 	use crate::ln::channelmanager::PaymentId;
-	use crate::ln::msgs;
+	use crate::ln::msgs::{self, OnionErrorPacket};
 	use crate::routing::router::{Path, PaymentParameters, Route, RouteHop};
 	use crate::types::features::{ChannelFeatures, NodeFeatures};
 	use crate::types::payment::PaymentHash;
@@ -2104,6 +2131,81 @@ mod tests {
 		assert_eq!(decrypted_failure.onion_error_code, Some(0x2002));
 	}
 
+	#[test]
+	fn test_non_attributable_failure_packet_onion() {
+		let corrupt_failure_packet = vec![1u8; 292];
+
+		let logger: Arc<TestLogger> = Arc::new(TestLogger::new());
+		let ctx_full = Secp256k1::new();
+		let path = build_test_path();
+		let htlc_source = HTLCSource::OutboundRoute {
+			path: path,
+			session_priv: get_test_session_key(),
+			first_hop_htlc_msat: 0,
+			payment_id: PaymentId([1; 32])
+,		};
+
+		// For a corrupt failure message, the failing channel cannot be identified.
+		let decrypted_failure = process_onion_failure(&ctx_full, &logger, &htlc_source, corrupt_failure_packet);
+		assert_eq!(decrypted_failure.short_channel_id, None);
+	}
+
+	#[test]
+	fn test_unreadable_failure_packet_onion() {
+		// Create a failure packet with a valid hmac but unreadable failure message.
+		let onion_keys: Vec<OnionKeys> = build_test_onion_keys();
+		let shared_secret = onion_keys[0].shared_secret.as_ref();
+		let um = gen_um_from_shared_secret(&shared_secret);
+
+		// The failure message is a single 0 byte.
+		let mut packet = [0u8; 33];
+
+		let mut hmac = HmacEngine::<Sha256>::new(&um);
+		hmac.input(&packet[32..]);
+		let hmac = Hmac::from_engine(hmac).to_byte_array();
+		packet[..32].copy_from_slice(&hmac);
+
+		let packet = encrypt_failure_packet(shared_secret, &packet);
+
+		test_failure_attribution(packet);
+	}
+
+	#[test]
+	fn test_missing_error_code() {
+		// Create a failure packet with a valid hmac and structure, but no error code.
+		let onion_keys: Vec<OnionKeys> = build_test_onion_keys();
+		let shared_secret = onion_keys[0].shared_secret.as_ref();
+		let um = gen_um_from_shared_secret(&shared_secret);
+
+		let failuremsg = vec![1];
+		let pad = Vec::new();
+		let mut packet = msgs::DecodedOnionErrorPacket { hmac: [0; 32], failuremsg, pad };
+
+		let mut hmac = HmacEngine::<Sha256>::new(&um);
+		hmac.input(&packet.encode()[32..]);
+		packet.hmac = Hmac::from_engine(hmac).to_byte_array();
+
+		let packet = encrypt_failure_packet(shared_secret, &packet.encode()[..]);
+
+		test_failure_attribution(packet);
+	}
+
+	fn test_failure_attribution(packet: OnionErrorPacket) {
+		let logger: Arc<TestLogger> = Arc::new(TestLogger::new());
+		let ctx_full = Secp256k1::new();
+		let path = build_test_path();
+		let htlc_source = HTLCSource::OutboundRoute {
+			path: path,
+			session_priv: get_test_session_key(),
+			first_hop_htlc_msat: 0,
+			payment_id: PaymentId([1; 32])
+,		};
+
+		// For the unreadable failure, it is still expected that the failing channel can be identified.
+		let decrypted_failure = process_onion_failure(&ctx_full, &logger, &htlc_source, packet.data);
+		assert_eq!(decrypted_failure.short_channel_id, Some(0));
+	}
+
 	struct RawOnionHopData {
 		data: Vec<u8>,
 	}