channeldb+routing: store full htlc failure reason

This commit extends the htlc fail info with the full failure reason that
was received over the wire. In a later commit, this info will also be
exposed on the rpc interface. Furthermore it serves as a building block
to make SendToRoute reliable across restarts.

channeldb/mp_payment.go

@@ -1,11 +1,14 @@
 package channeldb
 
 import (
+	"bytes"
 	"io"
 	"time"
 
 	"github.com/btcsuite/btcd/btcec"
+	"github.com/btcsuite/btcd/wire"
 	"github.com/lightningnetwork/lnd/lntypes"
+	"github.com/lightningnetwork/lnd/lnwire"
 	"github.com/lightningnetwork/lnd/routing/route"
 )
 
@@ -57,11 +60,45 @@ type HTLCSettleInfo struct {
 	SettleTime time.Time
 }
 
+// HTLCFailReason is the reason an htlc failed.
+type HTLCFailReason byte
+
+const (
+	// HTLCFailUnknown is recorded for htlcs that failed with an unknown
+	// reason.
+	HTLCFailUnknown HTLCFailReason = 0
+
+	// HTLCFailUnknown is recorded for htlcs that had a failure message that
+	// couldn't be decrypted.
+	HTLCFailUnreadable HTLCFailReason = 1
+
+	// HTLCFailInternal is recorded for htlcs that failed because of an
+	// internal error.
+	HTLCFailInternal HTLCFailReason = 2
+
+	// HTLCFailMessage is recorded for htlcs that failed with a network
+	// failure message.
+	HTLCFailMessage HTLCFailReason = 3
+)
+
 // HTLCFailInfo encapsulates the information that augments an HTLCAttempt in the
 // event that the HTLC fails.
 type HTLCFailInfo struct {
 	// FailTime is the time at which this HTLC was failed.
 	FailTime time.Time
+
+	// Message is the wire message that failed this HTLC. This field will be
+	// populated when the failure reason is HTLCFailMessage.
+	Message lnwire.FailureMessage
+
+	// Reason is the failure reason for this HTLC.
+	Reason HTLCFailReason
+
+	// The position in the path of the intermediate or final node that
+	// generated the failure message. Position zero is the sender node. This
+	// field will be populated when the failure reason is either
+	// HTLCFailMessage or HTLCFailUnknown.
+	FailureSourceIndex uint32
 }
 
 // MPPayment is a wrapper around a payment's PaymentCreationInfo and
@@ -130,7 +167,20 @@ func serializeHTLCFailInfo(w io.Writer, f *HTLCFailInfo) error {
 		return err
 	}
 
-	return nil
+	// Write failure. If there is no failure message, write an empty
+	// byte slice.
+	var messageBytes bytes.Buffer
+	if f.Message != nil {
+		err := lnwire.EncodeFailureMessage(&messageBytes, f.Message, 0)
+		if err != nil {
+			return err
+		}
+	}
+	if err := wire.WriteVarBytes(w, 0, messageBytes.Bytes()); err != nil {
+		return err
+	}
+
+	return WriteElements(w, byte(f.Reason), f.FailureSourceIndex)
 }
 
 // deserializeHTLCFailInfo deserializes the details of a failed htlc including
@@ -143,6 +193,29 @@ func deserializeHTLCFailInfo(r io.Reader) (*HTLCFailInfo, error) {
 		return nil, err
 	}
 
+	// Read failure.
+	failureBytes, err := wire.ReadVarBytes(
+		r, 0, lnwire.FailureMessageLength, "failure",
+	)
+	if err != nil {
+		return nil, err
+	}
+	if len(failureBytes) > 0 {
+		f.Message, err = lnwire.DecodeFailureMessage(
+			bytes.NewReader(failureBytes), 0,
+		)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	var reason byte
+	err = ReadElements(r, &reason, &f.FailureSourceIndex)
+	if err != nil {
+		return nil, err
+	}
+	f.Reason = HTLCFailReason(reason)
+
 	return f, nil
 }
 

channeldb/payment_control_test.go

@@ -126,7 +126,9 @@ func TestPaymentControlSwitchFail(t *testing.T) {
 	}
 
 	err = pControl.FailAttempt(
-		info.PaymentHash, 2, &HTLCFailInfo{},
+		info.PaymentHash, 2, &HTLCFailInfo{
+			Reason: HTLCFailUnreadable,
+		},
 	)
 	if err != nil {
 		t.Fatal(err)
@@ -362,7 +364,9 @@ func TestPaymentControlDeleteNonInFligt(t *testing.T) {
 			// Fail the payment attempt.
 			err := pControl.FailAttempt(
 				info.PaymentHash, attempt.AttemptID,
-				&HTLCFailInfo{},
+				&HTLCFailInfo{
+					Reason: HTLCFailUnreadable,
+				},
 			)
 			if err != nil {
 				t.Fatalf("unable to fail htlc: %v", err)

routing/payment_lifecycle.go

@@ -60,7 +60,7 @@ func (p *paymentLifecycle) resumePayment() ([32]byte, *route.Route, error) {
 			if sendErr != nil {
 				// TODO(joostjager): Distinguish unexpected
 				// internal errors from real send errors.
-				err = p.failAttempt()
+				err = p.failAttempt(sendErr)
 				if err != nil {
 					return [32]byte{}, nil, err
 				}
@@ -117,7 +117,7 @@ func (p *paymentLifecycle) resumePayment() ([32]byte, *route.Route, error) {
 				"the Switch, retrying.", p.attempt.AttemptID,
 				p.payment.PaymentHash)
 
-			err = p.failAttempt()
+			err = p.failAttempt(err)
 			if err != nil {
 				return [32]byte{}, nil, err
 			}
@@ -158,7 +158,7 @@ func (p *paymentLifecycle) resumePayment() ([32]byte, *route.Route, error) {
 			log.Errorf("Attempt to send payment %x failed: %v",
 				p.payment.PaymentHash, result.Error)
 
-			err = p.failAttempt()
+			err = p.failAttempt(result.Error)
 			if err != nil {
 				return [32]byte{}, nil, err
 			}
@@ -447,13 +447,60 @@ func (p *paymentLifecycle) handleSendError(sendErr error) error {
 }
 
 // failAttempt calls control tower to fail the current payment attempt.
-func (p *paymentLifecycle) failAttempt() error {
+func (p *paymentLifecycle) failAttempt(sendError error) error {
-	failInfo := &channeldb.HTLCFailInfo{
+	failInfo := marshallError(
-		FailTime: p.router.cfg.Clock.Now(),
+		sendError,
-	}
+		p.router.cfg.Clock.Now(),
+	)
 
 	return p.router.cfg.Control.FailAttempt(
 		p.payment.PaymentHash, p.attempt.AttemptID,
 		failInfo,
 	)
 }
+
+// marshallError marshall an error as received from the switch to a structure
+// that is suitable for database storage.
+func marshallError(sendError error, time time.Time) *channeldb.HTLCFailInfo {
+	response := &channeldb.HTLCFailInfo{
+		FailTime: time,
+	}
+
+	switch sendError {
+
+	case htlcswitch.ErrPaymentIDNotFound:
+		response.Reason = channeldb.HTLCFailInternal
+		return response
+
+	case htlcswitch.ErrUnreadableFailureMessage:
+		response.Reason = channeldb.HTLCFailUnreadable
+		return response
+	}
+
+	rtErr, ok := sendError.(htlcswitch.ClearTextError)
+	if !ok {
+		response.Reason = channeldb.HTLCFailInternal
+		return response
+	}
+
+	message := rtErr.WireMessage()
+	if message != nil {
+		response.Reason = channeldb.HTLCFailMessage
+		response.Message = message
+	} else {
+		response.Reason = channeldb.HTLCFailUnknown
+	}
+
+	// If the ClearTextError received is a ForwardingError, the error
+	// originated from a node along the route, not locally on our outgoing
+	// link. We set failureSourceIdx to the index of the node where the
+	// failure occurred. If the error is not a ForwardingError, the failure
+	// occurred at our node, so we leave the index as 0 to indicate that
+	// we failed locally.
+	fErr, ok := rtErr.(*htlcswitch.ForwardingError)
+	if ok {
+		response.FailureSourceIndex = uint32(fErr.FailureSourceIdx)
+	}
+
+	return response
+}