Merge pull request #10330 from Roasbeef/fix-error-handling-gossiper

discovery: fix potential infinite loop bug re context cancel error handling in gossip syncer

discovery/syncer.go

@@ -492,15 +492,19 @@ func (g *GossipSyncer) Stop() {
 
 // handleSyncingChans handles the state syncingChans for the GossipSyncer. When
 // in this state, we will send a QueryChannelRange msg to our peer and advance
-// the syncer's state to waitingQueryRangeReply.
+// the syncer's state to waitingQueryRangeReply. Returns an error if a fatal
-func (g *GossipSyncer) handleSyncingChans(ctx context.Context) {
+// error occurs that should cause the goroutine to exit.
+func (g *GossipSyncer) handleSyncingChans(ctx context.Context) error {
 	// Prepare the query msg.
 	queryRangeMsg, err := g.genChanRangeQuery(
 		ctx, g.genHistoricalChanRangeQuery,
 	)
 	if err != nil {
 		log.Errorf("Unable to gen chan range query: %v", err)
-		return
+
+		// Any error here is likely fatal (context cancelled, db error,
+		// etc.), so return it to exit the goroutine cleanly.
+		return err
 	}
 
 	// Acquire a lock so the following state transition is atomic.
@@ -517,12 +521,18 @@ func (g *GossipSyncer) handleSyncingChans(ctx context.Context) {
 	err = g.sendToPeer(ctx, queryRangeMsg)
 	if err != nil {
 		log.Errorf("Unable to send chan range query: %v", err)
-		return
+
+		// Any send error (peer exiting, connection closed, rate
+		// limiter signaling exit, etc.) is fatal, so return it to
+		// exit the goroutine cleanly.
+		return err
 	}
 
 	// With the message sent successfully, we'll transition into the next
 	// state where we wait for their reply.
 	g.setSyncState(waitingQueryRangeReply)
+
+	return nil
 }
 
 // channelGraphSyncer is the main goroutine responsible for ensuring that we
@@ -545,7 +555,14 @@ func (g *GossipSyncer) channelGraphSyncer(ctx context.Context) {
 		// understand, as we'll as responding to any other queries by
 		// them.
 		case syncingChans:
-			g.handleSyncingChans(ctx)
+			err := g.handleSyncingChans(ctx)
+			if err != nil {
+				log.Debugf("GossipSyncer(%x): exiting due to "+
+					"error in syncingChans: %v",
+					g.cfg.peerPub[:], err)
+
+				return
+			}
 
 		// In this state, we've sent out our initial channel range
 		// query and are waiting for the final response from the remote
@@ -593,7 +610,14 @@ func (g *GossipSyncer) channelGraphSyncer(ctx context.Context) {
 			// First, we'll attempt to continue our channel
 			// synchronization by continuing to send off another
 			// query chunk.
-			done := g.synchronizeChanIDs(ctx)
+			done, err := g.synchronizeChanIDs(ctx)
+			if err != nil {
+				log.Debugf("GossipSyncer(%x): exiting due to "+
+					"error in queryNewChannels: %v",
+					g.cfg.peerPub[:], err)
+
+				return
+			}
 
 			// If this wasn't our last query, then we'll need to
 			// transition to our waiting state.
@@ -819,8 +843,10 @@ func (g *GossipSyncer) sendGossipTimestampRange(ctx context.Context,
 // range. This method will be called continually until the entire range has
 // been queried for with a response received. We'll chunk our requests as
 // required to ensure they fit into a single message. We may re-renter this
-// state in the case that chunking is required.
+// state in the case that chunking is required. Returns true if synchronization
-func (g *GossipSyncer) synchronizeChanIDs(ctx context.Context) bool {
+// is complete, and an error if a fatal error occurs that should cause the
+// goroutine to exit.
+func (g *GossipSyncer) synchronizeChanIDs(ctx context.Context) (bool, error) {
 	// If we're in this state yet there are no more new channels to query
 	// for, then we'll transition to our final synced state and return true
 	// to signal that we're fully synchronized.
@@ -828,7 +854,7 @@ func (g *GossipSyncer) synchronizeChanIDs(ctx context.Context) bool {
 		log.Infof("GossipSyncer(%x): no more chans to query",
 			g.cfg.peerPub[:])
 
-		return true
+		return true, nil
 	}
 
 	// Otherwise, we'll issue our next chunked query to receive replies
@@ -864,9 +890,14 @@ func (g *GossipSyncer) synchronizeChanIDs(ctx context.Context) bool {
 	})
 	if err != nil {
 		log.Errorf("Unable to sync chan IDs: %v", err)
+
+		// Any send error (peer exiting, connection closed, rate
+		// limiter signaling exit, etc.) is fatal, so return it to
+		// exit the goroutine cleanly.
+		return false, err
 	}
 
-	return false
+	return false, nil
 }
 
 // isLegacyReplyChannelRange determines where a ReplyChannelRange message is

discovery/syncer_test.go

@@ -16,6 +16,7 @@ import (
 	"github.com/btcsuite/btcd/chaincfg/chainhash"
 	"github.com/davecgh/go-spew/spew"
 	graphdb "github.com/lightningnetwork/lnd/graph/db"
+	"github.com/lightningnetwork/lnd/lnpeer"
 	"github.com/lightningnetwork/lnd/lnwire"
 	"github.com/stretchr/testify/require"
 )
@@ -229,9 +230,111 @@ func newTestSyncer(hID lnwire.ShortChannelID,
 
 	syncer := newGossipSyncer(cfg, syncerSema)
 
+	//nolint:forcetypeassert
 	return msgChan, syncer, cfg.channelSeries.(*mockChannelGraphTimeSeries)
 }
 
+// errorInjector provides thread-safe error injection for test syncers and
+// tracks the number of send attempts to detect endless loops.
+type errorInjector struct {
+	mu           sync.Mutex
+	err          error
+	attemptCount int
+}
+
+// setError sets the error that will be returned by sendMsg calls.
+func (ei *errorInjector) setError(err error) {
+	ei.mu.Lock()
+	defer ei.mu.Unlock()
+	ei.err = err
+}
+
+// getError retrieves the current error in a thread-safe manner and increments
+// the attempt counter.
+func (ei *errorInjector) getError() error {
+	ei.mu.Lock()
+	defer ei.mu.Unlock()
+	ei.attemptCount++
+
+	return ei.err
+}
+
+// getAttemptCount returns the number of times sendMsg was called.
+func (ei *errorInjector) getAttemptCount() int {
+	ei.mu.Lock()
+	defer ei.mu.Unlock()
+	return ei.attemptCount
+}
+
+// newErrorInjectingSyncer creates a GossipSyncer with controllable error
+// injection for testing error handling. The returned errorInjector can be used
+// to inject errors into sendMsg calls.
+func newErrorInjectingSyncer(hID lnwire.ShortChannelID, chunkSize int32) (
+	*GossipSyncer, *errorInjector, chan []lnwire.Message) {
+
+	ei := &errorInjector{}
+	msgChan := make(chan []lnwire.Message, 20)
+
+	cfg := gossipSyncerCfg{
+		channelSeries:          newMockChannelGraphTimeSeries(hID),
+		encodingType:           defaultEncoding,
+		chunkSize:              chunkSize,
+		batchSize:              chunkSize,
+		noSyncChannels:         false,
+		noReplyQueries:         true,
+		noTimestampQueryOption: false,
+		sendMsg: func(_ context.Context, _ bool,
+			msgs ...lnwire.Message) error {
+
+			// Check if we should inject an error.
+			if err := ei.getError(); err != nil {
+				return err
+			}
+
+			msgChan <- msgs
+			return nil
+		},
+		bestHeight: func() uint32 {
+			return latestKnownHeight
+		},
+		markGraphSynced:          func() {},
+		maxQueryChanRangeReplies: maxQueryChanRangeReplies,
+		timestampQueueSize:       10,
+	}
+
+	syncerSema := make(chan struct{}, 1)
+	syncerSema <- struct{}{}
+
+	syncer := newGossipSyncer(cfg, syncerSema)
+
+	return syncer, ei, msgChan
+}
+
+// assertSyncerExitsCleanly verifies that a syncer stops cleanly within the
+// given timeout. This is used to ensure error handling doesn't cause endless
+// loops.
+func assertSyncerExitsCleanly(t *testing.T, syncer *GossipSyncer,
+	timeout time.Duration) {
+
+	t.Helper()
+
+	stopChan := make(chan struct{})
+	go func() {
+		syncer.Stop()
+		close(stopChan)
+	}()
+
+	select {
+	case <-stopChan:
+		// Success - syncer stopped cleanly.
+	case <-time.After(timeout):
+		t.Fatal(
+			"syncer did not stop within timeout - possible " +
+				"endless loop",
+		)
+	}
+}
+
 // TestGossipSyncerFilterGossipMsgsNoHorizon tests that if the remote peer
 // doesn't have a horizon set, then we won't send any incoming messages to it.
 func TestGossipSyncerFilterGossipMsgsNoHorizon(t *testing.T) {
@@ -1518,7 +1621,8 @@ func TestGossipSyncerSynchronizeChanIDs(t *testing.T) {
 
 	for i := 0; i < chunkSize*2; i += 2 {
 		// With our set up complete, we'll request a sync of chan ID's.
-		done := syncer.synchronizeChanIDs(t.Context())
+		done, err := syncer.synchronizeChanIDs(t.Context())
+		require.NoError(t, err)
 
 		// At this point, we shouldn't yet be done as only 2 items
 		// should have been queried for.
@@ -1565,7 +1669,8 @@ func TestGossipSyncerSynchronizeChanIDs(t *testing.T) {
 	}
 
 	// If we issue another query, the syncer should tell us that it's done.
-	done := syncer.synchronizeChanIDs(t.Context())
+	done, err := syncer.synchronizeChanIDs(t.Context())
+	require.NoError(t, err)
 	if done {
 		t.Fatalf("syncer should be finished!")
 	}
@@ -2409,3 +2514,107 @@ func TestGossipSyncerMaxChannelRangeReplies(t *testing.T) {
 		},
 	}, nil))
 }
+
+// TestGossipSyncerStateHandlerErrors tests that errors in state handlers cause
+// the channelGraphSyncer goroutine to exit cleanly without endless retry loops.
+// This is a table-driven test covering various error types and states.
+func TestGossipSyncerStateHandlerErrors(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name        string
+		state       syncerState
+		setupState  func(*GossipSyncer)
+		chunkSize   int32
+		injectedErr error
+	}{
+		{
+			name:        "context cancel during syncingChans",
+			state:       syncingChans,
+			chunkSize:   defaultChunkSize,
+			injectedErr: context.Canceled,
+			setupState:  func(s *GossipSyncer) {},
+		},
+		{
+			name:        "peer exit during syncingChans",
+			state:       syncingChans,
+			chunkSize:   defaultChunkSize,
+			injectedErr: lnpeer.ErrPeerExiting,
+			setupState:  func(s *GossipSyncer) {},
+		},
+		{
+			name:        "context cancel during queryNewChannels",
+			state:       queryNewChannels,
+			chunkSize:   2,
+			injectedErr: context.Canceled,
+			setupState: func(s *GossipSyncer) {
+				s.newChansToQuery = []lnwire.ShortChannelID{
+					lnwire.NewShortChanIDFromInt(1),
+					lnwire.NewShortChanIDFromInt(2),
+					lnwire.NewShortChanIDFromInt(3),
+				}
+			},
+		},
+		{
+			name:        "network error during queryNewChannels",
+			state:       queryNewChannels,
+			chunkSize:   2,
+			injectedErr: errors.New("connection closed"),
+			setupState: func(s *GossipSyncer) {
+				s.newChansToQuery = []lnwire.ShortChannelID{
+					lnwire.NewShortChanIDFromInt(1),
+					lnwire.NewShortChanIDFromInt(2),
+				}
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		tt := tt
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			// Create syncer with error injection capability.
+			hID := lnwire.NewShortChanIDFromInt(10)
+			syncer, errInj, _ := newErrorInjectingSyncer(
+				hID, tt.chunkSize,
+			)
+
+			// Set up the initial state and any required state data.
+			syncer.setSyncState(tt.state)
+			tt.setupState(syncer)
+
+			// Inject the error that should cause the goroutine to
+			// exit.
+			errInj.setError(tt.injectedErr)
+
+			// Start the syncer which spawns the channelGraphSyncer
+			// goroutine.
+			syncer.Start()
+
+			// Wait long enough that an endless loop would
+			// accumulate many attempts. With the fix, we should
+			// only see 1-3 attempts. Without the fix, we'd see
+			// 50-100+ attempts.
+			time.Sleep(500 * time.Millisecond)
+
+			// Check how many send attempts were made. This verifies
+			// that the state handler doesn't loop endlessly.
+			attemptCount := errInj.getAttemptCount()
+			require.GreaterOrEqual(
+				t, attemptCount, 1,
+				"state handler was not called - test "+
+					"setup issue",
+			)
+			require.LessOrEqual(
+				t, attemptCount, 5,
+				"too many attempts (%d) - endless loop "+
+					"not fixed",
+				attemptCount,
+			)
+
+			// Verify the syncer exits cleanly without hanging.
+			assertSyncerExitsCleanly(t, syncer, 2*time.Second)
+		})
+	}
+}

docs/release-notes/release-notes-0.20.0.md

@@ -56,6 +56,9 @@
   utxonursery (the legacy sweeper) where htlcs with a locktime of 0 would not
   be swept.
 
+- [Fixed a bug](https://github.com/lightningnetwork/lnd/pull/10330) to ensure that goroutine resources are properly freed in the case
+  of a disconnection or other failure event.
+
 # New Features
  
 * Use persisted [nodeannouncement](https://github.com/lightningnetwork/lnd/pull/8825)