feat: fix deadlock in splitstore-mpool interaction

blockstore/splitstore/splitstore.go

@@ -164,7 +164,7 @@ type SplitStore struct {
 	path string
 
 	mx          sync.Mutex
-	warmupEpoch abi.ChainEpoch // protected by mx
+	warmupEpoch atomic.Int64
 	baseEpoch   abi.ChainEpoch // protected by compaction lock
 	pruneEpoch  abi.ChainEpoch // protected by compaction lock
 
@@ -684,9 +684,7 @@ func (s *SplitStore) View(ctx context.Context, cid cid.Cid, cb func([]byte) erro
 }
 
 func (s *SplitStore) isWarm() bool {
-	s.mx.Lock()
-	defer s.mx.Unlock()
-	return s.warmupEpoch > 0
+	return s.warmupEpoch.Load() > 0
 }
 
 // State tracking
@@ -757,7 +755,7 @@ func (s *SplitStore) Start(chain ChainAccessor, us stmgr.UpgradeSchedule) error
 	bs, err = s.ds.Get(s.ctx, warmupEpochKey)
 	switch err {
 	case nil:
-		s.warmupEpoch = bytesToEpoch(bs)
+		s.warmupEpoch.Store(bytesToInt64(bs))
 
 	case dstore.ErrNotFound:
 		warmup = true
@@ -791,7 +789,7 @@ func (s *SplitStore) Start(chain ChainAccessor, us stmgr.UpgradeSchedule) error
 		return xerrors.Errorf("error loading compaction index: %w", err)
 	}
 
-	log.Infow("starting splitstore", "baseEpoch", s.baseEpoch, "warmupEpoch", s.warmupEpoch)
+	log.Infow("starting splitstore", "baseEpoch", s.baseEpoch, "warmupEpoch", s.warmupEpoch.Load())
 
 	if warmup {
 		err = s.warmup(curTs)

blockstore/splitstore/splitstore_check.go

@@ -145,7 +145,7 @@ func (s *SplitStore) doCheck(curTs *types.TipSet) error {
 func (s *SplitStore) Info() map[string]interface{} {
 	info := make(map[string]interface{})
 	info["base epoch"] = s.baseEpoch
-	info["warmup epoch"] = s.warmupEpoch
+	info["warmup epoch"] = s.warmupEpoch.Load()
 	info["compactions"] = s.compactionIndex
 	info["prunes"] = s.pruneIndex
 	info["compacting"] = s.compacting == 1

blockstore/splitstore/splitstore_test.go

@@ -429,7 +429,7 @@ func testSplitStoreReification(t *testing.T, f func(context.Context, blockstore.
 	}
 	defer ss.Close() //nolint
 
-	ss.warmupEpoch = 1
+	ss.warmupEpoch.Store(1)
 	go ss.reifyOrchestrator()
 
 	waitForReification := func() {
@@ -529,7 +529,7 @@ func testSplitStoreReificationLimit(t *testing.T, f func(context.Context, blocks
 	}
 	defer ss.Close() //nolint
 
-	ss.warmupEpoch = 1
+	ss.warmupEpoch.Store(1)
 	go ss.reifyOrchestrator()
 
 	waitForReification := func() {

blockstore/splitstore/splitstore_warmup.go

@@ -136,9 +136,8 @@ func (s *SplitStore) doWarmup(curTs *types.TipSet) error {
 	if err != nil {
 		return xerrors.Errorf("error saving warm up epoch: %w", err)
 	}
-	s.mx.Lock()
-	s.warmupEpoch = epoch
-	s.mx.Unlock()
+
+	s.warmupEpoch.Store(int64(epoch))
 
 	// also save the compactionIndex, as this is used as an indicator of warmup for upgraded nodes
 	err = s.ds.Put(s.ctx, compactionIndexKey, int64ToBytes(s.compactionIndex))

chain/messagepool/messagepool.go

@@ -448,12 +448,8 @@ func New(ctx context.Context, api Provider, ds dtypes.MetadataDS, us stmgr.Upgra
 	return mp, nil
 }
 
-func (mp *MessagePool) TryForEachPendingMessage(f func(cid.Cid) error) error {
-	// avoid deadlocks in splitstore compaction when something else needs to access the blockstore
-	// while holding the mpool lock
-	if !mp.lk.TryLock() {
-		return xerrors.Errorf("mpool TryForEachPendingMessage: could not acquire lock")
-	}
+func (mp *MessagePool) ForEachPendingMessage(f func(cid.Cid) error) error {
+	mp.lk.Lock()
 	defer mp.lk.Unlock()
 
 	for _, mset := range mp.pending {

node/modules/chain.go

@@ -70,7 +70,7 @@ func MessagePool(lc fx.Lifecycle, mctx helpers.MetricsCtx, us stmgr.UpgradeSched
 			return mp.Close()
 		},
 	})
-	protector.AddProtector(mp.TryForEachPendingMessage)
+	protector.AddProtector(mp.ForEachPendingMessage)
 	return mp, nil
 }