core, trie: intermediate mempool between trie and database (#15857)

This commit reduces database I/O by not writing every state trie to disk.

22 files changed, 581 insertions, 190 deletions

diff --git a/core/bench_test.go b/core/bench_test.go
index f976331d1..e23f0d19d 100644
--- a/core/bench_test.go
+++ b/core/bench_test.go
@@ -173,7 +173,7 @@ func benchInsertChain(b *testing.B, disk bool, gen func(int, *BlockGen)) {
 
 	// Time the insertion of the new chain.
 	// State and blocks are stored in the same DB.
-	chainman, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	chainman, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer chainman.Stop()
 	b.ReportAllocs()
 	b.ResetTimer()
@@ -283,7 +283,7 @@ func benchReadChain(b *testing.B, full bool, count uint64) {
 		if err != nil {
 			b.Fatalf("error opening database at %v: %v", dir, err)
 		}
-		chain, err := NewBlockChain(db, params.TestChainConfig, ethash.NewFaker(), vm.Config{})
+		chain, err := NewBlockChain(db, nil, params.TestChainConfig, ethash.NewFaker(), vm.Config{})
 		if err != nil {
 			b.Fatalf("error creating chain: %v", err)
 		}
diff --git a/core/block_validator.go b/core/block_validator.go
index 143728bb8..98958809b 100644
--- a/core/block_validator.go
+++ b/core/block_validator.go
@@ -50,11 +50,14 @@ func NewBlockValidator(config *params.ChainConfig, blockchain *BlockChain, engin
 // validated at this point.
 func (v *BlockValidator) ValidateBody(block *types.Block) error {
 	// Check whether the block's known, and if not, that it's linkable
-	if v.bc.HasBlockAndState(block.Hash()) {
+	if v.bc.HasBlockAndState(block.Hash(), block.NumberU64()) {
 		return ErrKnownBlock
 	}
-	if !v.bc.HasBlockAndState(block.ParentHash()) {
-		return consensus.ErrUnknownAncestor
+	if !v.bc.HasBlockAndState(block.ParentHash(), block.NumberU64()-1) {
+		if !v.bc.HasBlock(block.ParentHash(), block.NumberU64()-1) {
+			return consensus.ErrUnknownAncestor
+		}
+		return consensus.ErrPrunedAncestor
 	}
 	// Header validity is known at this point, check the uncles and transactions
 	header := block.Header()
diff --git a/core/block_validator_test.go b/core/block_validator_test.go
index e668601f3..e334b3c3c 100644
--- a/core/block_validator_test.go
+++ b/core/block_validator_test.go
@@ -42,7 +42,7 @@ func TestHeaderVerification(t *testing.T) {
 		headers[i] = block.Header()
 	}
 	// Run the header checker for blocks one-by-one, checking for both valid and invalid nonces
-	chain, _ := NewBlockChain(testdb, params.TestChainConfig, ethash.NewFaker(), vm.Config{})
+	chain, _ := NewBlockChain(testdb, nil, params.TestChainConfig, ethash.NewFaker(), vm.Config{})
 	defer chain.Stop()
 
 	for i := 0; i < len(blocks); i++ {
@@ -106,11 +106,11 @@ func testHeaderConcurrentVerification(t *testing.T, threads int) {
 		var results <-chan error
 
 		if valid {
-			chain, _ := NewBlockChain(testdb, params.TestChainConfig, ethash.NewFaker(), vm.Config{})
+			chain, _ := NewBlockChain(testdb, nil, params.TestChainConfig, ethash.NewFaker(), vm.Config{})
 			_, results = chain.engine.VerifyHeaders(chain, headers, seals)
 			chain.Stop()
 		} else {
-			chain, _ := NewBlockChain(testdb, params.TestChainConfig, ethash.NewFakeFailer(uint64(len(headers)-1)), vm.Config{})
+			chain, _ := NewBlockChain(testdb, nil, params.TestChainConfig, ethash.NewFakeFailer(uint64(len(headers)-1)), vm.Config{})
 			_, results = chain.engine.VerifyHeaders(chain, headers, seals)
 			chain.Stop()
 		}
@@ -173,7 +173,7 @@ func testHeaderConcurrentAbortion(t *testing.T, threads int) {
 	defer runtime.GOMAXPROCS(old)
 
 	// Start the verifications and immediately abort
-	chain, _ := NewBlockChain(testdb, params.TestChainConfig, ethash.NewFakeDelayer(time.Millisecond), vm.Config{})
+	chain, _ := NewBlockChain(testdb, nil, params.TestChainConfig, ethash.NewFakeDelayer(time.Millisecond), vm.Config{})
 	defer chain.Stop()
 
 	abort, results := chain.engine.VerifyHeaders(chain, headers, seals)
diff --git a/core/blockchain.go b/core/blockchain.go
index d5e139e31..8d141fddb 100644
--- a/core/blockchain.go
+++ b/core/blockchain.go
@@ -42,6 +42,7 @@ import (
 	"github.com/ethereum/go-ethereum/rlp"
 	"github.com/ethereum/go-ethereum/trie"
 	"github.com/hashicorp/golang-lru"
+	"gopkg.in/karalabe/cookiejar.v2/collections/prque"
 )
 
 var (
@@ -56,11 +57,20 @@ const (
 	maxFutureBlocks     = 256
 	maxTimeFutureBlocks = 30
 	badBlockLimit       = 10
+	triesInMemory       = 128
 
 	// BlockChainVersion ensures that an incompatible database forces a resync from scratch.
 	BlockChainVersion = 3
 )
 
+// CacheConfig contains the configuration values for the trie caching/pruning
+// that's resident in a blockchain.
+type CacheConfig struct {
+	Disabled      bool          // Whether to disable trie write caching (archive node)
+	TrieNodeLimit int           // Memory limit (MB) at which to flush the current in-memory trie to disk
+	TrieTimeLimit time.Duration // Time limit after which to flush the current in-memory trie to disk
+}
+
 // BlockChain represents the canonical chain given a database with a genesis
 // block. The Blockchain manages chain imports, reverts, chain reorganisations.
 //
@@ -76,10 +86,14 @@ const (
 // included in the canonical one where as GetBlockByNumber always represents the
 // canonical chain.
 type BlockChain struct {
-	config *params.ChainConfig // chain & network configuration
+	chainConfig *params.ChainConfig // Chain & network configuration
+	cacheConfig *CacheConfig        // Cache configuration for pruning
+
+	db     ethdb.Database // Low level persistent database to store final content in
+	triegc *prque.Prque   // Priority queue mapping block numbers to tries to gc
+	gcproc time.Duration  // Accumulates canonical block processing for trie dumping
 
 	hc            *HeaderChain
-	chainDb       ethdb.Database
 	rmLogsFeed    event.Feed
 	chainFeed     event.Feed
 	chainSideFeed event.Feed
@@ -119,7 +133,13 @@ type BlockChain struct {
 // NewBlockChain returns a fully initialised block chain using information
 // available in the database. It initialises the default Ethereum Validator and
 // Processor.
-func NewBlockChain(chainDb ethdb.Database, config *params.ChainConfig, engine consensus.Engine, vmConfig vm.Config) (*BlockChain, error) {
+func NewBlockChain(db ethdb.Database, cacheConfig *CacheConfig, chainConfig *params.ChainConfig, engine consensus.Engine, vmConfig vm.Config) (*BlockChain, error) {
+	if cacheConfig == nil {
+		cacheConfig = &CacheConfig{
+			TrieNodeLimit: 256 * 1024 * 1024,
+			TrieTimeLimit: 5 * time.Minute,
+		}
+	}
 	bodyCache, _ := lru.New(bodyCacheLimit)
 	bodyRLPCache, _ := lru.New(bodyCacheLimit)
 	blockCache, _ := lru.New(blockCacheLimit)
@@ -127,9 +147,11 @@ func NewBlockChain(chainDb ethdb.Database, config *params.ChainConfig, engine co
 	badBlocks, _ := lru.New(badBlockLimit)
 
 	bc := &BlockChain{
-		config:       config,
-		chainDb:      chainDb,
-		stateCache:   state.NewDatabase(chainDb),
+		chainConfig:  chainConfig,
+		cacheConfig:  cacheConfig,
+		db:           db,
+		triegc:       prque.New(),
+		stateCache:   state.NewDatabase(db),
 		quit:         make(chan struct{}),
 		bodyCache:    bodyCache,
 		bodyRLPCache: bodyRLPCache,
@@ -139,11 +161,11 @@ func NewBlockChain(chainDb ethdb.Database, config *params.ChainConfig, engine co
 		vmConfig:     vmConfig,
 		badBlocks:    badBlocks,
 	}
-	bc.SetValidator(NewBlockValidator(config, bc, engine))
-	bc.SetProcessor(NewStateProcessor(config, bc, engine))
+	bc.SetValidator(NewBlockValidator(chainConfig, bc, engine))
+	bc.SetProcessor(NewStateProcessor(chainConfig, bc, engine))
 
 	var err error
-	bc.hc, err = NewHeaderChain(chainDb, config, engine, bc.getProcInterrupt)
+	bc.hc, err = NewHeaderChain(db, chainConfig, engine, bc.getProcInterrupt)
 	if err != nil {
 		return nil, err
 	}
@@ -180,7 +202,7 @@ func (bc *BlockChain) getProcInterrupt() bool {
 // assumes that the chain manager mutex is held.
 func (bc *BlockChain) loadLastState() error {
 	// Restore the last known head block
-	head := GetHeadBlockHash(bc.chainDb)
+	head := GetHeadBlockHash(bc.db)
 	if head == (common.Hash{}) {
 		// Corrupt or empty database, init from scratch
 		log.Warn("Empty database, resetting chain")
@@ -196,15 +218,17 @@ func (bc *BlockChain) loadLastState() error {
 	// Make sure the state associated with the block is available
 	if _, err := state.New(currentBlock.Root(), bc.stateCache); err != nil {
 		// Dangling block without a state associated, init from scratch
-		log.Warn("Head state missing, resetting chain", "number", currentBlock.Number(), "hash", currentBlock.Hash())
-		return bc.Reset()
+		log.Warn("Head state missing, repairing chain", "number", currentBlock.Number(), "hash", currentBlock.Hash())
+		if err := bc.repair(&currentBlock); err != nil {
+			return err
+		}
 	}
 	// Everything seems to be fine, set as the head block
 	bc.currentBlock = currentBlock
 
 	// Restore the last known head header
 	currentHeader := bc.currentBlock.Header()
-	if head := GetHeadHeaderHash(bc.chainDb); head != (common.Hash{}) {
+	if head := GetHeadHeaderHash(bc.db); head != (common.Hash{}) {
 		if header := bc.GetHeaderByHash(head); header != nil {
 			currentHeader = header
 		}
@@ -213,7 +237,7 @@ func (bc *BlockChain) loadLastState() error {
 
 	// Restore the last known head fast block
 	bc.currentFastBlock = bc.currentBlock
-	if head := GetHeadFastBlockHash(bc.chainDb); head != (common.Hash{}) {
+	if head := GetHeadFastBlockHash(bc.db); head != (common.Hash{}) {
 		if block := bc.GetBlockByHash(head); block != nil {
 			bc.currentFastBlock = block
 		}
@@ -243,7 +267,7 @@ func (bc *BlockChain) SetHead(head uint64) error {
 
 	// Rewind the header chain, deleting all block bodies until then
 	delFn := func(hash common.Hash, num uint64) {
-		DeleteBody(bc.chainDb, hash, num)
+		DeleteBody(bc.db, hash, num)
 	}
 	bc.hc.SetHead(head, delFn)
 	currentHeader := bc.hc.CurrentHeader()
@@ -275,10 +299,10 @@ func (bc *BlockChain) SetHead(head uint64) error {
 	if bc.currentFastBlock == nil {
 		bc.currentFastBlock = bc.genesisBlock
 	}
-	if err := WriteHeadBlockHash(bc.chainDb, bc.currentBlock.Hash()); err != nil {
+	if err := WriteHeadBlockHash(bc.db, bc.currentBlock.Hash()); err != nil {
 		log.Crit("Failed to reset head full block", "err", err)
 	}
-	if err := WriteHeadFastBlockHash(bc.chainDb, bc.currentFastBlock.Hash()); err != nil {
+	if err := WriteHeadFastBlockHash(bc.db, bc.currentFastBlock.Hash()); err != nil {
 		log.Crit("Failed to reset head fast block", "err", err)
 	}
 	return bc.loadLastState()
@@ -292,7 +316,7 @@ func (bc *BlockChain) FastSyncCommitHead(hash common.Hash) error {
 	if block == nil {
 		return fmt.Errorf("non existent block [%x…]", hash[:4])
 	}
-	if _, err := trie.NewSecure(block.Root(), bc.chainDb, 0); err != nil {
+	if _, err := trie.NewSecure(block.Root(), bc.stateCache.TrieDB(), 0); err != nil {
 		return err
 	}
 	// If all checks out, manually set the head block
@@ -387,7 +411,7 @@ func (bc *BlockChain) ResetWithGenesisBlock(genesis *types.Block) error {
 	if err := bc.hc.WriteTd(genesis.Hash(), genesis.NumberU64(), genesis.Difficulty()); err != nil {
 		log.Crit("Failed to write genesis block TD", "err", err)
 	}
-	if err := WriteBlock(bc.chainDb, genesis); err != nil {
+	if err := WriteBlock(bc.db, genesis); err != nil {
 		log.Crit("Failed to write genesis block", "err", err)
 	}
 	bc.genesisBlock = genesis
@@ -400,6 +424,24 @@ func (bc *BlockChain) ResetWithGenesisBlock(genesis *types.Block) error {
 	return nil
 }
 
+// repair tries to repair the current blockchain by rolling back the current block
+// until one with associated state is found. This is needed to fix incomplete db
+// writes caused either by crashes/power outages, or simply non-committed tries.
+//
+// This method only rolls back the current block. The current header and current
+// fast block are left intact.
+func (bc *BlockChain) repair(head **types.Block) error {
+	for {
+		// Abort if we've rewound to a head block that does have associated state
+		if _, err := state.New((*head).Root(), bc.stateCache); err == nil {
+			log.Info("Rewound blockchain to past state", "number", (*head).Number(), "hash", (*head).Hash())
+			return nil
+		}
+		// Otherwise rewind one block and recheck state availability there
+		(*head) = bc.GetBlock((*head).ParentHash(), (*head).NumberU64()-1)
+	}
+}
+
 // Export writes the active chain to the given writer.
 func (bc *BlockChain) Export(w io.Writer) error {
 	return bc.ExportN(w, uint64(0), bc.currentBlock.NumberU64())
@@ -437,13 +479,13 @@ func (bc *BlockChain) ExportN(w io.Writer, first uint64, last uint64) error {
 // Note, this function assumes that the `mu` mutex is held!
 func (bc *BlockChain) insert(block *types.Block) {
 	// If the block is on a side chain or an unknown one, force other heads onto it too
-	updateHeads := GetCanonicalHash(bc.chainDb, block.NumberU64()) != block.Hash()
+	updateHeads := GetCanonicalHash(bc.db, block.NumberU64()) != block.Hash()
 
 	// Add the block to the canonical chain number scheme and mark as the head
-	if err := WriteCanonicalHash(bc.chainDb, block.Hash(), block.NumberU64()); err != nil {
+	if err := WriteCanonicalHash(bc.db, block.Hash(), block.NumberU64()); err != nil {
 		log.Crit("Failed to insert block number", "err", err)
 	}
-	if err := WriteHeadBlockHash(bc.chainDb, block.Hash()); err != nil {
+	if err := WriteHeadBlockHash(bc.db, block.Hash()); err != nil {
 		log.Crit("Failed to insert head block hash", "err", err)
 	}
 	bc.currentBlock = block
@@ -452,7 +494,7 @@ func (bc *BlockChain) insert(block *types.Block) {
 	if updateHeads {
 		bc.hc.SetCurrentHeader(block.Header())
 
-		if err := WriteHeadFastBlockHash(bc.chainDb, block.Hash()); err != nil {
+		if err := WriteHeadFastBlockHash(bc.db, block.Hash()); err != nil {
 			log.Crit("Failed to insert head fast block hash", "err", err)
 		}
 		bc.currentFastBlock = block
@@ -472,7 +514,7 @@ func (bc *BlockChain) GetBody(hash common.Hash) *types.Body {
 		body := cached.(*types.Body)
 		return body
 	}
-	body := GetBody(bc.chainDb, hash, bc.hc.GetBlockNumber(hash))
+	body := GetBody(bc.db, hash, bc.hc.GetBlockNumber(hash))
 	if body == nil {
 		return nil
 	}
@@ -488,7 +530,7 @@ func (bc *BlockChain) GetBodyRLP(hash common.Hash) rlp.RawValue {
 	if cached, ok := bc.bodyRLPCache.Get(hash); ok {
 		return cached.(rlp.RawValue)
 	}
-	body := GetBodyRLP(bc.chainDb, hash, bc.hc.GetBlockNumber(hash))
+	body := GetBodyRLP(bc.db, hash, bc.hc.GetBlockNumber(hash))
 	if len(body) == 0 {
 		return nil
 	}
@@ -502,21 +544,25 @@ func (bc *BlockChain) HasBlock(hash common.Hash, number uint64) bool {
 	if bc.blockCache.Contains(hash) {
 		return true
 	}
-	ok, _ := bc.chainDb.Has(blockBodyKey(hash, number))
+	ok, _ := bc.db.Has(blockBodyKey(hash, number))
 	return ok
 }
 
+// HasState checks if state trie is fully present in the database or not.
+func (bc *BlockChain) HasState(hash common.Hash) bool {
+	_, err := bc.stateCache.OpenTrie(hash)
+	return err == nil
+}
+
 // HasBlockAndState checks if a block and associated state trie is fully present
 // in the database or not, caching it if present.
-func (bc *BlockChain) HasBlockAndState(hash common.Hash) bool {
+func (bc *BlockChain) HasBlockAndState(hash common.Hash, number uint64) bool {
 	// Check first that the block itself is known
-	block := bc.GetBlockByHash(hash)
+	block := bc.GetBlock(hash, number)
 	if block == nil {
 		return false
 	}
-	// Ensure the associated state is also present
-	_, err := bc.stateCache.OpenTrie(block.Root())
-	return err == nil
+	return bc.HasState(block.Root())
 }
 
 // GetBlock retrieves a block from the database by hash and number,
@@ -526,7 +572,7 @@ func (bc *BlockChain) GetBlock(hash common.Hash, number uint64) *types.Block {
 	if block, ok := bc.blockCache.Get(hash); ok {
 		return block.(*types.Block)
 	}
-	block := GetBlock(bc.chainDb, hash, number)
+	block := GetBlock(bc.db, hash, number)
 	if block == nil {
 		return nil
 	}
@@ -543,13 +589,18 @@ func (bc *BlockChain) GetBlockByHash(hash common.Hash) *types.Block {
 // GetBlockByNumber retrieves a block from the database by number, caching it
 // (associated with its hash) if found.
 func (bc *BlockChain) GetBlockByNumber(number uint64) *types.Block {
-	hash := GetCanonicalHash(bc.chainDb, number)
+	hash := GetCanonicalHash(bc.db, number)
 	if hash == (common.Hash{}) {
 		return nil
 	}
 	return bc.GetBlock(hash, number)
 }
 
+// GetReceiptsByHash retrieves the receipts for all transactions in a given block.
+func (bc *BlockChain) GetReceiptsByHash(hash common.Hash) types.Receipts {
+	return GetBlockReceipts(bc.db, hash, GetBlockNumber(bc.db, hash))
+}
+
 // GetBlocksFromHash returns the block corresponding to hash and up to n-1 ancestors.
 // [deprecated by eth/62]
 func (bc *BlockChain) GetBlocksFromHash(hash common.Hash, n int) (blocks []*types.Block) {
@@ -577,6 +628,12 @@ func (bc *BlockChain) GetUnclesInChain(block *types.Block, length int) []*types.
 	return uncles
 }
 
+// TrieNode retrieves a blob of data associated with a trie node (or code hash)
+// either from ephemeral in-memory cache, or from persistent storage.
+func (bc *BlockChain) TrieNode(hash common.Hash) ([]byte, error) {
+	return bc.stateCache.TrieDB().Node(hash)
+}
+
 // Stop stops the blockchain service. If any imports are currently in progress
 // it will abort them using the procInterrupt.
 func (bc *BlockChain) Stop() {
@@ -589,6 +646,33 @@ func (bc *BlockChain) Stop() {
 	atomic.StoreInt32(&bc.procInterrupt, 1)
 
 	bc.wg.Wait()
+
+	// Ensure the state of a recent block is also stored to disk before exiting.
+	// It is fine if this state does not exist (fast start/stop cycle), but it is
+	// advisable to leave an N block gap from the head so 1) a restart loads up
+	// the last N blocks as sync assistance to remote nodes; 2) a restart during
+	// a (small) reorg doesn't require deep reprocesses; 3) chain "repair" from
+	// missing states are constantly tested.
+	//
+	// This may be tuned a bit on mainnet if its too annoying to reprocess the last
+	// N blocks.
+	if !bc.cacheConfig.Disabled {
+		triedb := bc.stateCache.TrieDB()
+		if number := bc.CurrentBlock().NumberU64(); number >= triesInMemory {
+			recent := bc.GetBlockByNumber(bc.CurrentBlock().NumberU64() - triesInMemory + 1)
+
+			log.Info("Writing cached state to disk", "block", recent.Number(), "hash", recent.Hash(), "root", recent.Root())
+			if err := triedb.Commit(recent.Root(), true); err != nil {
+				log.Error("Failed to commit recent state trie", "err", err)
+			}
+		}
+		for !bc.triegc.Empty() {
+			triedb.Dereference(bc.triegc.PopItem().(common.Hash), common.Hash{})
+		}
+		if size := triedb.Size(); size != 0 {
+			log.Error("Dangling trie nodes after full cleanup")
+		}
+	}
 	log.Info("Blockchain manager stopped")
 }
 
@@ -633,11 +717,11 @@ func (bc *BlockChain) Rollback(chain []common.Hash) {
 		}
 		if bc.currentFastBlock.Hash() == hash {
 			bc.currentFastBlock = bc.GetBlock(bc.currentFastBlock.ParentHash(), bc.currentFastBlock.NumberU64()-1)
-			WriteHeadFastBlockHash(bc.chainDb, bc.currentFastBlock.Hash())
+			WriteHeadFastBlockHash(bc.db, bc.currentFastBlock.Hash())
 		}
 		if bc.currentBlock.Hash() == hash {
 			bc.currentBlock = bc.GetBlock(bc.currentBlock.ParentHash(), bc.currentBlock.NumberU64()-1)
-			WriteHeadBlockHash(bc.chainDb, bc.currentBlock.Hash())
+			WriteHeadBlockHash(bc.db, bc.currentBlock.Hash())
 		}
 	}
 }
@@ -696,7 +780,7 @@ func (bc *BlockChain) InsertReceiptChain(blockChain types.Blocks, receiptChain [
 		stats = struct{ processed, ignored int32 }{}
 		start = time.Now()
 		bytes = 0
-		batch = bc.chainDb.NewBatch()
+		batch = bc.db.NewBatch()
 	)
 	for i, block := range blockChain {
 		receipts := receiptChain[i]
@@ -714,7 +798,7 @@ func (bc *BlockChain) InsertReceiptChain(blockChain types.Blocks, receiptChain [
 			continue
 		}
 		// Compute all the non-consensus fields of the receipts
-		SetReceiptsData(bc.config, block, receipts)
+		SetReceiptsData(bc.chainConfig, block, receipts)
 		// Write all the data out into the database
 		if err := WriteBody(batch, block.Hash(), block.NumberU64(), block.Body()); err != nil {
 			return i, fmt.Errorf("failed to write block body: %v", err)
@@ -747,7 +831,7 @@ func (bc *BlockChain) InsertReceiptChain(blockChain types.Blocks, receiptChain [
 	head := blockChain[len(blockChain)-1]
 	if td := bc.GetTd(head.Hash(), head.NumberU64()); td != nil { // Rewind may have occurred, skip in that case
 		if bc.GetTd(bc.currentFastBlock.Hash(), bc.currentFastBlock.NumberU64()).Cmp(td) < 0 {
-			if err := WriteHeadFastBlockHash(bc.chainDb, head.Hash()); err != nil {
+			if err := WriteHeadFastBlockHash(bc.db, head.Hash()); err != nil {
 				log.Crit("Failed to update head fast block hash", "err", err)
 			}
 			bc.currentFastBlock = head
@@ -758,15 +842,33 @@ func (bc *BlockChain) InsertReceiptChain(blockChain types.Blocks, receiptChain [
 	log.Info("Imported new block receipts",
 		"count", stats.processed,
 		"elapsed", common.PrettyDuration(time.Since(start)),
-		"bytes", bytes,
 		"number", head.Number(),
 		"hash", head.Hash(),
+		"size", common.StorageSize(bytes),
 		"ignored", stats.ignored)
 	return 0, nil
 }
 
-// WriteBlock writes the block to the chain.
-func (bc *BlockChain) WriteBlockAndState(block *types.Block, receipts []*types.Receipt, state *state.StateDB) (status WriteStatus, err error) {
+var lastWrite uint64
+
+// WriteBlockWithoutState writes only the block and its metadata to the database,
+// but does not write any state. This is used to construct competing side forks
+// up to the point where they exceed the canonical total difficulty.
+func (bc *BlockChain) WriteBlockWithoutState(block *types.Block, td *big.Int) (err error) {
+	bc.wg.Add(1)
+	defer bc.wg.Done()
+
+	if err := bc.hc.WriteTd(block.Hash(), block.NumberU64(), td); err != nil {
+		return err
+	}
+	if err := WriteBlock(bc.db, block); err != nil {
+		return err
+	}
+	return nil
+}
+
+// WriteBlockWithState writes the block and all associated state to the database.
+func (bc *BlockChain) WriteBlockWithState(block *types.Block, receipts []*types.Receipt, state *state.StateDB) (status WriteStatus, err error) {
 	bc.wg.Add(1)
 	defer bc.wg.Done()
 
@@ -787,17 +889,73 @@ func (bc *BlockChain) WriteBlockAndState(block *types.Block, receipts []*types.R
 		return NonStatTy, err
 	}
 	// Write other block data using a batch.
-	batch := bc.chainDb.NewBatch()
+	batch := bc.db.NewBatch()
 	if err := WriteBlock(batch, block); err != nil {
 		return NonStatTy, err
 	}
-	if _, err := state.CommitTo(batch, bc.config.IsEIP158(block.Number())); err != nil {
+	root, err := state.Commit(bc.chainConfig.IsEIP158(block.Number()))
+	if err != nil {
 		return NonStatTy, err
 	}
+	triedb := bc.stateCache.TrieDB()
+
+	// If we're running an archive node, always flush
+	if bc.cacheConfig.Disabled {
+		if err := triedb.Commit(root, false); err != nil {
+			return NonStatTy, err
+		}
+	} else {
+		// Full but not archive node, do proper garbage collection
+		triedb.Reference(root, common.Hash{}) // metadata reference to keep trie alive
+		bc.triegc.Push(root, -float32(block.NumberU64()))
+
+		if current := block.NumberU64(); current > triesInMemory {
+			// Find the next state trie we need to commit
+			header := bc.GetHeaderByNumber(current - triesInMemory)
+			chosen := header.Number.Uint64()
+
+			// Only write to disk if we exceeded our memory allowance *and* also have at
+			// least a given number of tries gapped.
+			var (
+				size  = triedb.Size()
+				limit = common.StorageSize(bc.cacheConfig.TrieNodeLimit) * 1024 * 1024
+			)
+			if size > limit || bc.gcproc > bc.cacheConfig.TrieTimeLimit {
+				// If we're exceeding limits but haven't reached a large enough memory gap,
+				// warn the user that the system is becoming unstable.
+				if chosen < lastWrite+triesInMemory {
+					switch {
+					case size >= 2*limit:
+						log.Error("Trie memory critical, forcing to disk", "size", size, "limit", limit, "optimum", float64(chosen-lastWrite)/triesInMemory)
+					case bc.gcproc >= 2*bc.cacheConfig.TrieTimeLimit:
+						log.Error("Trie timing critical, forcing to disk", "time", bc.gcproc, "allowance", bc.cacheConfig.TrieTimeLimit, "optimum", float64(chosen-lastWrite)/triesInMemory)
+					case size > limit:
+						log.Warn("Trie memory at dangerous levels", "size", size, "limit", limit, "optimum", float64(chosen-lastWrite)/triesInMemory)
+					case bc.gcproc > bc.cacheConfig.TrieTimeLimit:
+						log.Warn("Trie timing at dangerous levels", "time", bc.gcproc, "limit", bc.cacheConfig.TrieTimeLimit, "optimum", float64(chosen-lastWrite)/triesInMemory)
+					}
+				}
+				// If optimum or critical limits reached, write to disk
+				if chosen >= lastWrite+triesInMemory || size >= 2*limit || bc.gcproc >= 2*bc.cacheConfig.TrieTimeLimit {
+					triedb.Commit(header.Root, true)
+					lastWrite = chosen
+					bc.gcproc = 0
+				}
+			}
+			// Garbage collect anything below our required write retention
+			for !bc.triegc.Empty() {
+				root, number := bc.triegc.Pop()
+				if uint64(-number) > chosen {
+					bc.triegc.Push(root, number)
+					break
+				}
+				triedb.Dereference(root.(common.Hash), common.Hash{})
+			}
+		}
+	}
 	if err := WriteBlockReceipts(batch, block.Hash(), block.NumberU64(), receipts); err != nil {
 		return NonStatTy, err
 	}
-
 	// If the total difficulty is higher than our known, add it to the canonical chain
 	// Second clause in the if statement reduces the vulnerability to selfish mining.
 	// Please refer to http://www.cs.cornell.edu/~ie53/publications/btcProcFC.pdf
@@ -818,7 +976,7 @@ func (bc *BlockChain) WriteBlockAndState(block *types.Block, receipts []*types.R
 			return NonStatTy, err
 		}
 		// Write hash preimages
-		if err := WritePreimages(bc.chainDb, block.NumberU64(), state.Preimages()); err != nil {
+		if err := WritePreimages(bc.db, block.NumberU64(), state.Preimages()); err != nil {
 			return NonStatTy, err
 		}
 		status = CanonStatTy
@@ -910,31 +1068,60 @@ func (bc *BlockChain) insertChain(chain types.Blocks) (int, []interface{}, []*ty
 		if err == nil {
 			err = bc.Validator().ValidateBody(block)
 		}
-		if err != nil {
-			if err == ErrKnownBlock {
-				stats.ignored++
-				continue
+		switch {
+		case err == ErrKnownBlock:
+			stats.ignored++
+			continue
+
+		case err == consensus.ErrFutureBlock:
+			// Allow up to MaxFuture second in the future blocks. If this limit is exceeded
+			// the chain is discarded and processed at a later time if given.
+			max := big.NewInt(time.Now().Unix() + maxTimeFutureBlocks)
+			if block.Time().Cmp(max) > 0 {
+				return i, events, coalescedLogs, fmt.Errorf("future block: %v > %v", block.Time(), max)
 			}
+			bc.futureBlocks.Add(block.Hash(), block)
+			stats.queued++
+			continue
+
+		case err == consensus.ErrUnknownAncestor && bc.futureBlocks.Contains(block.ParentHash()):
+			bc.futureBlocks.Add(block.Hash(), block)
+			stats.queued++
+			continue
 
-			if err == consensus.ErrFutureBlock {
-				// Allow up to MaxFuture second in the future blocks. If this limit
-				// is exceeded the chain is discarded and processed at a later time
-				// if given.
-				max := big.NewInt(time.Now().Unix() + maxTimeFutureBlocks)
-				if block.Time().Cmp(max) > 0 {
-					return i, events, coalescedLogs, fmt.Errorf("future block: %v > %v", block.Time(), max)
+		case err == consensus.ErrPrunedAncestor:
+			// Block competing with the canonical chain, store in the db, but don't process
+			// until the competitor TD goes above the canonical TD
+			localTd := bc.GetTd(bc.currentBlock.Hash(), bc.currentBlock.NumberU64())
+			externTd := new(big.Int).Add(bc.GetTd(block.ParentHash(), block.NumberU64()-1), block.Difficulty())
+			if localTd.Cmp(externTd) > 0 {
+				if err = bc.WriteBlockWithoutState(block, externTd); err != nil {
+					return i, events, coalescedLogs, err
 				}
-				bc.futureBlocks.Add(block.Hash(), block)
-				stats.queued++
 				continue
 			}
+			// Competitor chain beat canonical, gather all blocks from the common ancestor
+			var winner []*types.Block
 
-			if err == consensus.ErrUnknownAncestor && bc.futureBlocks.Contains(block.ParentHash()) {
-				bc.futureBlocks.Add(block.Hash(), block)
-				stats.queued++
-				continue
+			parent := bc.GetBlock(block.ParentHash(), block.NumberU64()-1)
+			for !bc.HasState(parent.Root()) {
+				winner = append(winner, parent)
+				parent = bc.GetBlock(parent.ParentHash(), parent.NumberU64()-1)
+			}
+			for j := 0; j < len(winner)/2; j++ {
+				winner[j], winner[len(winner)-1-j] = winner[len(winner)-1-j], winner[j]
+			}
+			// Import all the pruned blocks to make the state available
+			bc.chainmu.Unlock()
+			_, evs, logs, err := bc.insertChain(winner)
+			bc.chainmu.Lock()
+			events, coalescedLogs = evs, logs
+
+			if err != nil {
+				return i, events, coalescedLogs, err
 			}
 
+		case err != nil:
 			bc.reportBlock(block, nil, err)
 			return i, events, coalescedLogs, err
 		}
@@ -962,8 +1149,10 @@ func (bc *BlockChain) insertChain(chain types.Blocks) (int, []interface{}, []*ty
 			bc.reportBlock(block, receipts, err)
 			return i, events, coalescedLogs, err
 		}
+		proctime := time.Since(bstart)
+
 		// Write the block to the chain and get the status.
-		status, err := bc.WriteBlockAndState(block, receipts, state)
+		status, err := bc.WriteBlockWithState(block, receipts, state)
 		if err != nil {
 			return i, events, coalescedLogs, err
 		}
@@ -977,6 +1166,9 @@ func (bc *BlockChain) insertChain(chain types.Blocks) (int, []interface{}, []*ty
 			events = append(events, ChainEvent{block, block.Hash(), logs})
 			lastCanon = block
 
+			// Only count canonical blocks for GC processing time
+			bc.gcproc += proctime
+
 		case SideStatTy:
 			log.Debug("Inserted forked block", "number", block.Number(), "hash", block.Hash(), "diff", block.Difficulty(), "elapsed",
 				common.PrettyDuration(time.Since(bstart)), "txs", len(block.Transactions()), "gas", block.GasUsed(), "uncles", len(block.Uncles()))
@@ -986,7 +1178,7 @@ func (bc *BlockChain) insertChain(chain types.Blocks) (int, []interface{}, []*ty
 		}
 		stats.processed++
 		stats.usedGas += usedGas
-		stats.report(chain, i)
+		stats.report(chain, i, bc.stateCache.TrieDB().Size())
 	}
 	// Append a single chain head event if we've progressed the chain
 	if lastCanon != nil && bc.CurrentBlock().Hash() == lastCanon.Hash() {
@@ -1009,7 +1201,7 @@ const statsReportLimit = 8 * time.Second
 
 // report prints statistics if some number of blocks have been processed
 // or more than a few seconds have passed since the last message.
-func (st *insertStats) report(chain []*types.Block, index int) {
+func (st *insertStats) report(chain []*types.Block, index int, cache common.StorageSize) {
 	// Fetch the timings for the batch
 	var (
 		now     = mclock.Now()
@@ -1024,7 +1216,7 @@ func (st *insertStats) report(chain []*types.Block, index int) {
 		context := []interface{}{
 			"blocks", st.processed, "txs", txs, "mgas", float64(st.usedGas) / 1000000,
 			"elapsed", common.PrettyDuration(elapsed), "mgasps", float64(st.usedGas) * 1000 / float64(elapsed),
-			"number", end.Number(), "hash", end.Hash(),
+			"number", end.Number(), "hash", end.Hash(), "cache", cache,
 		}
 		if st.queued > 0 {
 			context = append(context, []interface{}{"queued", st.queued}...)
@@ -1060,7 +1252,7 @@ func (bc *BlockChain) reorg(oldBlock, newBlock *types.Block) error {
 		// These logs are later announced as deleted.
 		collectLogs = func(h common.Hash) {
 			// Coalesce logs and set 'Removed'.
-			receipts := GetBlockReceipts(bc.chainDb, h, bc.hc.GetBlockNumber(h))
+			receipts := GetBlockReceipts(bc.db, h, bc.hc.GetBlockNumber(h))
 			for _, receipt := range receipts {
 				for _, log := range receipt.Logs {
 					del := *log
@@ -1129,7 +1321,7 @@ func (bc *BlockChain) reorg(oldBlock, newBlock *types.Block) error {
 		// insert the block in the canonical way, re-writing history
 		bc.insert(newChain[i])
 		// write lookup entries for hash based transaction/receipt searches
-		if err := WriteTxLookupEntries(bc.chainDb, newChain[i]); err != nil {
+		if err := WriteTxLookupEntries(bc.db, newChain[i]); err != nil {
 			return err
 		}
 		addedTxs = append(addedTxs, newChain[i].Transactions()...)
@@ -1139,7 +1331,7 @@ func (bc *BlockChain) reorg(oldBlock, newBlock *types.Block) error {
 	// When transactions get deleted from the database that means the
 	// receipts that were created in the fork must also be deleted
 	for _, tx := range diff {
-		DeleteTxLookupEntry(bc.chainDb, tx.Hash())
+		DeleteTxLookupEntry(bc.db, tx.Hash())
 	}
 	if len(deletedLogs) > 0 {
 		go bc.rmLogsFeed.Send(RemovedLogsEvent{deletedLogs})
@@ -1231,7 +1423,7 @@ Hash: 0x%x
 
 Error: %v
 ##############################
-`, bc.config, block.Number(), block.Hash(), receiptString, err))
+`, bc.chainConfig, block.Number(), block.Hash(), receiptString, err))
 }
 
 // InsertHeaderChain attempts to insert the given header chain in to the local
@@ -1338,7 +1530,7 @@ func (bc *BlockChain) GetHeaderByNumber(number uint64) *types.Header {
 }
 
 // Config retrieves the blockchain's chain configuration.
-func (bc *BlockChain) Config() *params.ChainConfig { return bc.config }
+func (bc *BlockChain) Config() *params.ChainConfig { return bc.chainConfig }
 
 // Engine retrieves the blockchain's consensus engine.
 func (bc *BlockChain) Engine() consensus.Engine { return bc.engine }
diff --git a/core/blockchain_test.go b/core/blockchain_test.go
index cbde3bcd2..635379161 100644
--- a/core/blockchain_test.go
+++ b/core/blockchain_test.go
@@ -46,7 +46,7 @@ func newTestBlockChain(fake bool) *BlockChain {
 	if !fake {
 		engine = ethash.NewTester()
 	}
-	blockchain, err := NewBlockChain(db, gspec.Config, engine, vm.Config{})
+	blockchain, err := NewBlockChain(db, nil, gspec.Config, engine, vm.Config{})
 	if err != nil {
 		panic(err)
 	}
@@ -148,9 +148,9 @@ func testBlockChainImport(chain types.Blocks, blockchain *BlockChain) error {
 			return err
 		}
 		blockchain.mu.Lock()
-		WriteTd(blockchain.chainDb, block.Hash(), block.NumberU64(), new(big.Int).Add(block.Difficulty(), blockchain.GetTdByHash(block.ParentHash())))
-		WriteBlock(blockchain.chainDb, block)
-		statedb.CommitTo(blockchain.chainDb, false)
+		WriteTd(blockchain.db, block.Hash(), block.NumberU64(), new(big.Int).Add(block.Difficulty(), blockchain.GetTdByHash(block.ParentHash())))
+		WriteBlock(blockchain.db, block)
+		statedb.Commit(false)
 		blockchain.mu.Unlock()
 	}
 	return nil
@@ -166,8 +166,8 @@ func testHeaderChainImport(chain []*types.Header, blockchain *BlockChain) error
 		}
 		// Manually insert the header into the database, but don't reorganise (allows subsequent testing)
 		blockchain.mu.Lock()
-		WriteTd(blockchain.chainDb, header.Hash(), header.Number.Uint64(), new(big.Int).Add(header.Difficulty, blockchain.GetTdByHash(header.ParentHash)))
-		WriteHeader(blockchain.chainDb, header)
+		WriteTd(blockchain.db, header.Hash(), header.Number.Uint64(), new(big.Int).Add(header.Difficulty, blockchain.GetTdByHash(header.ParentHash)))
+		WriteHeader(blockchain.db, header)
 		blockchain.mu.Unlock()
 	}
 	return nil
@@ -186,9 +186,9 @@ func TestLastBlock(t *testing.T) {
 	bchain := newTestBlockChain(false)
 	defer bchain.Stop()
 
-	block := makeBlockChain(bchain.CurrentBlock(), 1, ethash.NewFaker(), bchain.chainDb, 0)[0]
+	block := makeBlockChain(bchain.CurrentBlock(), 1, ethash.NewFaker(), bchain.db, 0)[0]
 	bchain.insert(block)
-	if block.Hash() != GetHeadBlockHash(bchain.chainDb) {
+	if block.Hash() != GetHeadBlockHash(bchain.db) {
 		t.Errorf("Write/Get HeadBlockHash failed")
 	}
 }
@@ -496,7 +496,7 @@ func testReorgBadHashes(t *testing.T, full bool) {
 	}
 
 	// Create a new BlockChain and check that it rolled back the state.
-	ncm, err := NewBlockChain(bc.chainDb, bc.config, ethash.NewFaker(), vm.Config{})
+	ncm, err := NewBlockChain(bc.db, nil, bc.chainConfig, ethash.NewFaker(), vm.Config{})
 	if err != nil {
 		t.Fatalf("failed to create new chain manager: %v", err)
 	}
@@ -609,7 +609,7 @@ func TestFastVsFullChains(t *testing.T) {
 	// Import the chain as an archive node for the comparison baseline
 	archiveDb, _ := ethdb.NewMemDatabase()
 	gspec.MustCommit(archiveDb)
-	archive, _ := NewBlockChain(archiveDb, gspec.Config, ethash.NewFaker(), vm.Config{})
+	archive, _ := NewBlockChain(archiveDb, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer archive.Stop()
 
 	if n, err := archive.InsertChain(blocks); err != nil {
@@ -618,7 +618,7 @@ func TestFastVsFullChains(t *testing.T) {
 	// Fast import the chain as a non-archive node to test
 	fastDb, _ := ethdb.NewMemDatabase()
 	gspec.MustCommit(fastDb)
-	fast, _ := NewBlockChain(fastDb, gspec.Config, ethash.NewFaker(), vm.Config{})
+	fast, _ := NewBlockChain(fastDb, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer fast.Stop()
 
 	headers := make([]*types.Header, len(blocks))
@@ -696,7 +696,7 @@ func TestLightVsFastVsFullChainHeads(t *testing.T) {
 	archiveDb, _ := ethdb.NewMemDatabase()
 	gspec.MustCommit(archiveDb)
 
-	archive, _ := NewBlockChain(archiveDb, gspec.Config, ethash.NewFaker(), vm.Config{})
+	archive, _ := NewBlockChain(archiveDb, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	if n, err := archive.InsertChain(blocks); err != nil {
 		t.Fatalf("failed to process block %d: %v", n, err)
 	}
@@ -709,7 +709,7 @@ func TestLightVsFastVsFullChainHeads(t *testing.T) {
 	// Import the chain as a non-archive node and ensure all pointers are updated
 	fastDb, _ := ethdb.NewMemDatabase()
 	gspec.MustCommit(fastDb)
-	fast, _ := NewBlockChain(fastDb, gspec.Config, ethash.NewFaker(), vm.Config{})
+	fast, _ := NewBlockChain(fastDb, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer fast.Stop()
 
 	headers := make([]*types.Header, len(blocks))
@@ -730,7 +730,7 @@ func TestLightVsFastVsFullChainHeads(t *testing.T) {
 	lightDb, _ := ethdb.NewMemDatabase()
 	gspec.MustCommit(lightDb)
 
-	light, _ := NewBlockChain(lightDb, gspec.Config, ethash.NewFaker(), vm.Config{})
+	light, _ := NewBlockChain(lightDb, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	if n, err := light.InsertHeaderChain(headers, 1); err != nil {
 		t.Fatalf("failed to insert header %d: %v", n, err)
 	}
@@ -799,7 +799,7 @@ func TestChainTxReorgs(t *testing.T) {
 		}
 	})
 	// Import the chain. This runs all block validation rules.
-	blockchain, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	if i, err := blockchain.InsertChain(chain); err != nil {
 		t.Fatalf("failed to insert original chain[%d]: %v", i, err)
 	}
@@ -870,7 +870,7 @@ func TestLogReorgs(t *testing.T) {
 		signer  = types.NewEIP155Signer(gspec.Config.ChainId)
 	)
 
-	blockchain, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer blockchain.Stop()
 
 	rmLogsCh := make(chan RemovedLogsEvent)
@@ -917,7 +917,7 @@ func TestReorgSideEvent(t *testing.T) {
 		signer  = types.NewEIP155Signer(gspec.Config.ChainId)
 	)
 
-	blockchain, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer blockchain.Stop()
 
 	chain, _ := GenerateChain(gspec.Config, genesis, ethash.NewFaker(), db, 3, func(i int, gen *BlockGen) {})
@@ -992,7 +992,7 @@ func TestCanonicalBlockRetrieval(t *testing.T) {
 	bc := newTestBlockChain(true)
 	defer bc.Stop()
 
-	chain, _ := GenerateChain(bc.config, bc.genesisBlock, ethash.NewFaker(), bc.chainDb, 10, func(i int, gen *BlockGen) {})
+	chain, _ := GenerateChain(bc.chainConfig, bc.genesisBlock, ethash.NewFaker(), bc.db, 10, func(i int, gen *BlockGen) {})
 
 	var pend sync.WaitGroup
 	pend.Add(len(chain))
@@ -1003,14 +1003,14 @@ func TestCanonicalBlockRetrieval(t *testing.T) {
 
 			// try to retrieve a block by its canonical hash and see if the block data can be retrieved.
 			for {
-				ch := GetCanonicalHash(bc.chainDb, block.NumberU64())
+				ch := GetCanonicalHash(bc.db, block.NumberU64())
 				if ch == (common.Hash{}) {
 					continue // busy wait for canonical hash to be written
 				}
 				if ch != block.Hash() {
 					t.Fatalf("unknown canonical hash, want %s, got %s", block.Hash().Hex(), ch.Hex())
 				}
-				fb := GetBlock(bc.chainDb, ch, block.NumberU64())
+				fb := GetBlock(bc.db, ch, block.NumberU64())
 				if fb == nil {
 					t.Fatalf("unable to retrieve block %d for canonical hash: %s", block.NumberU64(), ch.Hex())
 				}
@@ -1043,7 +1043,7 @@ func TestEIP155Transition(t *testing.T) {
 		genesis = gspec.MustCommit(db)
 	)
 
-	blockchain, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer blockchain.Stop()
 
 	blocks, _ := GenerateChain(gspec.Config, genesis, ethash.NewFaker(), db, 4, func(i int, block *BlockGen) {
@@ -1151,7 +1151,7 @@ func TestEIP161AccountRemoval(t *testing.T) {
 		}
 		genesis = gspec.MustCommit(db)
 	)
-	blockchain, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer blockchain.Stop()
 
 	blocks, _ := GenerateChain(gspec.Config, genesis, ethash.NewFaker(), db, 3, func(i int, block *BlockGen) {
@@ -1226,7 +1226,7 @@ func TestBlockchainHeaderchainReorgConsistency(t *testing.T) {
 	diskdb, _ := ethdb.NewMemDatabase()
 	new(Genesis).MustCommit(diskdb)
 
-	chain, err := NewBlockChain(diskdb, params.TestChainConfig, engine, vm.Config{})
+	chain, err := NewBlockChain(diskdb, nil, params.TestChainConfig, engine, vm.Config{})
 	if err != nil {
 		t.Fatalf("failed to create tester chain: %v", err)
 	}
@@ -1245,3 +1245,102 @@ func TestBlockchainHeaderchainReorgConsistency(t *testing.T) {
 		}
 	}
 }
+
+// Tests that importing small side forks doesn't leave junk in the trie database
+// cache (which would eventually cause memory issues).
+func TestTrieForkGC(t *testing.T) {
+	// Generate a canonical chain to act as the main dataset
+	engine := ethash.NewFaker()
+
+	db, _ := ethdb.NewMemDatabase()
+	genesis := new(Genesis).MustCommit(db)
+	blocks, _ := GenerateChain(params.TestChainConfig, genesis, engine, db, 2*triesInMemory, func(i int, b *BlockGen) { b.SetCoinbase(common.Address{1}) })
+
+	// Generate a bunch of fork blocks, each side forking from the canonical chain
+	forks := make([]*types.Block, len(blocks))
+	for i := 0; i < len(forks); i++ {
+		parent := genesis
+		if i > 0 {
+			parent = blocks[i-1]
+		}
+		fork, _ := GenerateChain(params.TestChainConfig, parent, engine, db, 1, func(i int, b *BlockGen) { b.SetCoinbase(common.Address{2}) })
+		forks[i] = fork[0]
+	}
+	// Import the canonical and fork chain side by side, forcing the trie cache to cache both
+	diskdb, _ := ethdb.NewMemDatabase()
+	new(Genesis).MustCommit(diskdb)
+
+	chain, err := NewBlockChain(diskdb, nil, params.TestChainConfig, engine, vm.Config{})
+	if err != nil {
+		t.Fatalf("failed to create tester chain: %v", err)
+	}
+	for i := 0; i < len(blocks); i++ {
+		if _, err := chain.InsertChain(blocks[i : i+1]); err != nil {
+			t.Fatalf("block %d: failed to insert into chain: %v", i, err)
+		}
+		if _, err := chain.InsertChain(forks[i : i+1]); err != nil {
+			t.Fatalf("fork %d: failed to insert into chain: %v", i, err)
+		}
+	}
+	// Dereference all the recent tries and ensure no past trie is left in
+	for i := 0; i < triesInMemory; i++ {
+		chain.stateCache.TrieDB().Dereference(blocks[len(blocks)-1-i].Root(), common.Hash{})
+		chain.stateCache.TrieDB().Dereference(forks[len(blocks)-1-i].Root(), common.Hash{})
+	}
+	if len(chain.stateCache.TrieDB().Nodes()) > 0 {
+		t.Fatalf("stale tries still alive after garbase collection")
+	}
+}
+
+// Tests that doing large reorgs works even if the state associated with the
+// forking point is not available any more.
+func TestLargeReorgTrieGC(t *testing.T) {
+	// Generate the original common chain segment and the two competing forks
+	engine := ethash.NewFaker()
+
+	db, _ := ethdb.NewMemDatabase()
+	genesis := new(Genesis).MustCommit(db)
+
+	shared, _ := GenerateChain(params.TestChainConfig, genesis, engine, db, 64, func(i int, b *BlockGen) { b.SetCoinbase(common.Address{1}) })
+	original, _ := GenerateChain(params.TestChainConfig, shared[len(shared)-1], engine, db, 2*triesInMemory, func(i int, b *BlockGen) { b.SetCoinbase(common.Address{2}) })
+	competitor, _ := GenerateChain(params.TestChainConfig, shared[len(shared)-1], engine, db, 2*triesInMemory+1, func(i int, b *BlockGen) { b.SetCoinbase(common.Address{3}) })
+
+	// Import the shared chain and the original canonical one
+	diskdb, _ := ethdb.NewMemDatabase()
+	new(Genesis).MustCommit(diskdb)
+
+	chain, err := NewBlockChain(diskdb, nil, params.TestChainConfig, engine, vm.Config{})
+	if err != nil {
+		t.Fatalf("failed to create tester chain: %v", err)
+	}
+	if _, err := chain.InsertChain(shared); err != nil {
+		t.Fatalf("failed to insert shared chain: %v", err)
+	}
+	if _, err := chain.InsertChain(original); err != nil {
+		t.Fatalf("failed to insert shared chain: %v", err)
+	}
+	// Ensure that the state associated with the forking point is pruned away
+	if node, _ := chain.stateCache.TrieDB().Node(shared[len(shared)-1].Root()); node != nil {
+		t.Fatalf("common-but-old ancestor still cache")
+	}
+	// Import the competitor chain without exceeding the canonical's TD and ensure
+	// we have not processed any of the blocks (protection against malicious blocks)
+	if _, err := chain.InsertChain(competitor[:len(competitor)-2]); err != nil {
+		t.Fatalf("failed to insert competitor chain: %v", err)
+	}
+	for i, block := range competitor[:len(competitor)-2] {
+		if node, _ := chain.stateCache.TrieDB().Node(block.Root()); node != nil {
+			t.Fatalf("competitor %d: low TD chain became processed", i)
+		}
+	}
+	// Import the head of the competitor chain, triggering the reorg and ensure we
+	// successfully reprocess all the stashed away blocks.
+	if _, err := chain.InsertChain(competitor[len(competitor)-2:]); err != nil {
+		t.Fatalf("failed to finalize competitor chain: %v", err)
+	}
+	for i, block := range competitor[:len(competitor)-triesInMemory] {
+		if node, _ := chain.stateCache.TrieDB().Node(block.Root()); node != nil {
+			t.Fatalf("competitor %d: competing chain state missing", i)
+		}
+	}
+}
diff --git a/core/chain_indexer.go b/core/chain_indexer.go
index 7fb184aaa..158ed8324 100644
--- a/core/chain_indexer.go
+++ b/core/chain_indexer.go
@@ -203,6 +203,9 @@ func (c *ChainIndexer) eventLoop(currentHeader *types.Header, events chan ChainE
 			if header.ParentHash != prevHash {
 				// Reorg to the common ancestor (might not exist in light sync mode, skip reorg then)
 				// TODO(karalabe, zsfelfoldi): This seems a bit brittle, can we detect this case explicitly?
+
+				// TODO(karalabe): This operation is expensive and might block, causing the event system to
+				// potentially also lock up. We need to do with on a different thread somehow.
 				if h := FindCommonAncestor(c.chainDb, prevHeader, header); h != nil {
 					c.newHead(h.Number.Uint64(), true)
 				}
diff --git a/core/chain_makers.go b/core/chain_makers.go
index 5e264a994..6744428ff 100644
--- a/core/chain_makers.go
+++ b/core/chain_makers.go
@@ -166,7 +166,7 @@ func GenerateChain(config *params.ChainConfig, parent *types.Block, engine conse
 	genblock := func(i int, parent *types.Block, statedb *state.StateDB) (*types.Block, types.Receipts) {
 		// TODO(karalabe): This is needed for clique, which depends on multiple blocks.
 		// It's nonetheless ugly to spin up a blockchain here. Get rid of this somehow.
-		blockchain, _ := NewBlockChain(db, config, engine, vm.Config{})
+		blockchain, _ := NewBlockChain(db, nil, config, engine, vm.Config{})
 		defer blockchain.Stop()
 
 		b := &BlockGen{i: i, parent: parent, chain: blocks, chainReader: blockchain, statedb: statedb, config: config, engine: engine}
@@ -192,10 +192,13 @@ func GenerateChain(config *params.ChainConfig, parent *types.Block, engine conse
 		if b.engine != nil {
 			block, _ := b.engine.Finalize(b.chainReader, b.header, statedb, b.txs, b.uncles, b.receipts)
 			// Write state changes to db
-			_, err := statedb.CommitTo(db, config.IsEIP158(b.header.Number))
+			root, err := statedb.Commit(config.IsEIP158(b.header.Number))
 			if err != nil {
 				panic(fmt.Sprintf("state write error: %v", err))
 			}
+			if err := statedb.Database().TrieDB().Commit(root, false); err != nil {
+				panic(fmt.Sprintf("trie write error: %v", err))
+			}
 			return block, b.receipts
 		}
 		return nil, nil
@@ -246,7 +249,7 @@ func newCanonical(engine consensus.Engine, n int, full bool) (ethdb.Database, *B
 	db, _ := ethdb.NewMemDatabase()
 	genesis := gspec.MustCommit(db)
 
-	blockchain, _ := NewBlockChain(db, params.AllEthashProtocolChanges, engine, vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, params.AllEthashProtocolChanges, engine, vm.Config{})
 	// Create and inject the requested chain
 	if n == 0 {
 		return db, blockchain, nil
diff --git a/core/chain_makers_test.go b/core/chain_makers_test.go
index a3b80da29..93be43ddc 100644
--- a/core/chain_makers_test.go
+++ b/core/chain_makers_test.go
@@ -79,7 +79,7 @@ func ExampleGenerateChain() {
 	})
 
 	// Import the chain. This runs all block validation rules.
-	blockchain, _ := NewBlockChain(db, gspec.Config, ethash.NewFaker(), vm.Config{})
+	blockchain, _ := NewBlockChain(db, nil, gspec.Config, ethash.NewFaker(), vm.Config{})
 	defer blockchain.Stop()
 
 	if i, err := blockchain.InsertChain(chain); err != nil {
diff --git a/core/dao_test.go b/core/dao_test.go
index 43e2982a5..e0a3e3ff3 100644
--- a/core/dao_test.go
+++ b/core/dao_test.go
@@ -45,7 +45,7 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 	proConf.DAOForkBlock = forkBlock
 	proConf.DAOForkSupport = true
 
-	proBc, _ := NewBlockChain(proDb, &proConf, ethash.NewFaker(), vm.Config{})
+	proBc, _ := NewBlockChain(proDb, nil, &proConf, ethash.NewFaker(), vm.Config{})
 	defer proBc.Stop()
 
 	conDb, _ := ethdb.NewMemDatabase()
@@ -55,7 +55,7 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 	conConf.DAOForkBlock = forkBlock
 	conConf.DAOForkSupport = false
 
-	conBc, _ := NewBlockChain(conDb, &conConf, ethash.NewFaker(), vm.Config{})
+	conBc, _ := NewBlockChain(conDb, nil, &conConf, ethash.NewFaker(), vm.Config{})
 	defer conBc.Stop()
 
 	if _, err := proBc.InsertChain(prefix); err != nil {
@@ -69,7 +69,7 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 		// Create a pro-fork block, and try to feed into the no-fork chain
 		db, _ = ethdb.NewMemDatabase()
 		gspec.MustCommit(db)
-		bc, _ := NewBlockChain(db, &conConf, ethash.NewFaker(), vm.Config{})
+		bc, _ := NewBlockChain(db, nil, &conConf, ethash.NewFaker(), vm.Config{})
 		defer bc.Stop()
 
 		blocks := conBc.GetBlocksFromHash(conBc.CurrentBlock().Hash(), int(conBc.CurrentBlock().NumberU64()))
@@ -79,6 +79,9 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 		if _, err := bc.InsertChain(blocks); err != nil {
 			t.Fatalf("failed to import contra-fork chain for expansion: %v", err)
 		}
+		if err := bc.stateCache.TrieDB().Commit(bc.CurrentHeader().Root, true); err != nil {
+			t.Fatalf("failed to commit contra-fork head for expansion: %v", err)
+		}
 		blocks, _ = GenerateChain(&proConf, conBc.CurrentBlock(), ethash.NewFaker(), db, 1, func(i int, gen *BlockGen) {})
 		if _, err := conBc.InsertChain(blocks); err == nil {
 			t.Fatalf("contra-fork chain accepted pro-fork block: %v", blocks[0])
@@ -91,7 +94,7 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 		// Create a no-fork block, and try to feed into the pro-fork chain
 		db, _ = ethdb.NewMemDatabase()
 		gspec.MustCommit(db)
-		bc, _ = NewBlockChain(db, &proConf, ethash.NewFaker(), vm.Config{})
+		bc, _ = NewBlockChain(db, nil, &proConf, ethash.NewFaker(), vm.Config{})
 		defer bc.Stop()
 
 		blocks = proBc.GetBlocksFromHash(proBc.CurrentBlock().Hash(), int(proBc.CurrentBlock().NumberU64()))
@@ -101,6 +104,9 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 		if _, err := bc.InsertChain(blocks); err != nil {
 			t.Fatalf("failed to import pro-fork chain for expansion: %v", err)
 		}
+		if err := bc.stateCache.TrieDB().Commit(bc.CurrentHeader().Root, true); err != nil {
+			t.Fatalf("failed to commit pro-fork head for expansion: %v", err)
+		}
 		blocks, _ = GenerateChain(&conConf, proBc.CurrentBlock(), ethash.NewFaker(), db, 1, func(i int, gen *BlockGen) {})
 		if _, err := proBc.InsertChain(blocks); err == nil {
 			t.Fatalf("pro-fork chain accepted contra-fork block: %v", blocks[0])
@@ -114,7 +120,7 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 	// Verify that contra-forkers accept pro-fork extra-datas after forking finishes
 	db, _ = ethdb.NewMemDatabase()
 	gspec.MustCommit(db)
-	bc, _ := NewBlockChain(db, &conConf, ethash.NewFaker(), vm.Config{})
+	bc, _ := NewBlockChain(db, nil, &conConf, ethash.NewFaker(), vm.Config{})
 	defer bc.Stop()
 
 	blocks := conBc.GetBlocksFromHash(conBc.CurrentBlock().Hash(), int(conBc.CurrentBlock().NumberU64()))
@@ -124,6 +130,9 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 	if _, err := bc.InsertChain(blocks); err != nil {
 		t.Fatalf("failed to import contra-fork chain for expansion: %v", err)
 	}
+	if err := bc.stateCache.TrieDB().Commit(bc.CurrentHeader().Root, true); err != nil {
+		t.Fatalf("failed to commit contra-fork head for expansion: %v", err)
+	}
 	blocks, _ = GenerateChain(&proConf, conBc.CurrentBlock(), ethash.NewFaker(), db, 1, func(i int, gen *BlockGen) {})
 	if _, err := conBc.InsertChain(blocks); err != nil {
 		t.Fatalf("contra-fork chain didn't accept pro-fork block post-fork: %v", err)
@@ -131,7 +140,7 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 	// Verify that pro-forkers accept contra-fork extra-datas after forking finishes
 	db, _ = ethdb.NewMemDatabase()
 	gspec.MustCommit(db)
-	bc, _ = NewBlockChain(db, &proConf, ethash.NewFaker(), vm.Config{})
+	bc, _ = NewBlockChain(db, nil, &proConf, ethash.NewFaker(), vm.Config{})
 	defer bc.Stop()
 
 	blocks = proBc.GetBlocksFromHash(proBc.CurrentBlock().Hash(), int(proBc.CurrentBlock().NumberU64()))
@@ -141,6 +150,9 @@ func TestDAOForkRangeExtradata(t *testing.T) {
 	if _, err := bc.InsertChain(blocks); err != nil {
 		t.Fatalf("failed to import pro-fork chain for expansion: %v", err)
 	}
+	if err := bc.stateCache.TrieDB().Commit(bc.CurrentHeader().Root, true); err != nil {
+		t.Fatalf("failed to commit pro-fork head for expansion: %v", err)
+	}
 	blocks, _ = GenerateChain(&conConf, proBc.CurrentBlock(), ethash.NewFaker(), db, 1, func(i int, gen *BlockGen) {})
 	if _, err := proBc.InsertChain(blocks); err != nil {
 		t.Fatalf("pro-fork chain didn't accept contra-fork block post-fork: %v", err)
diff --git a/core/genesis.go b/core/genesis.go
index e22985b80..b6ead2250 100644
--- a/core/genesis.go
+++ b/core/genesis.go
@@ -169,10 +169,9 @@ func SetupGenesisBlock(db ethdb.Database, genesis *Genesis) (*params.ChainConfig
 
 	// Check whether the genesis block is already written.
 	if genesis != nil {
-		block, _ := genesis.ToBlock()
-		hash := block.Hash()
+		hash := genesis.ToBlock(nil).Hash()
 		if hash != stored {
-			return genesis.Config, block.Hash(), &GenesisMismatchError{stored, hash}
+			return genesis.Config, hash, &GenesisMismatchError{stored, hash}
 		}
 	}
 
@@ -220,9 +219,12 @@ func (g *Genesis) configOrDefault(ghash common.Hash) *params.ChainConfig {
 	}
 }
 
-// ToBlock creates the block and state of a genesis specification.
-func (g *Genesis) ToBlock() (*types.Block, *state.StateDB) {
-	db, _ := ethdb.NewMemDatabase()
+// ToBlock creates the genesis block and writes state of a genesis specification
+// to the given database (or discards it if nil).
+func (g *Genesis) ToBlock(db ethdb.Database) *types.Block {
+	if db == nil {
+		db, _ = ethdb.NewMemDatabase()
+	}
 	statedb, _ := state.New(common.Hash{}, state.NewDatabase(db))
 	for addr, account := range g.Alloc {
 		statedb.AddBalance(addr, account.Balance)
@@ -252,19 +254,19 @@ func (g *Genesis) ToBlock() (*types.Block, *state.StateDB) {
 	if g.Difficulty == nil {
 		head.Difficulty = params.GenesisDifficulty
 	}
-	return types.NewBlock(head, nil, nil, nil), statedb
+	statedb.Commit(false)
+	statedb.Database().TrieDB().Commit(root, true)
+
+	return types.NewBlock(head, nil, nil, nil)
 }
 
 // Commit writes the block and state of a genesis specification to the database.
 // The block is committed as the canonical head block.
 func (g *Genesis) Commit(db ethdb.Database) (*types.Block, error) {
-	block, statedb := g.ToBlock()
+	block := g.ToBlock(db)
 	if block.Number().Sign() != 0 {
 		return nil, fmt.Errorf("can't commit genesis block with number > 0")
 	}
-	if _, err := statedb.CommitTo(db, false); err != nil {
-		return nil, fmt.Errorf("cannot write state: %v", err)
-	}
 	if err := WriteTd(db, block.Hash(), block.NumberU64(), g.Difficulty); err != nil {
 		return nil, err
 	}
diff --git a/core/genesis_test.go b/core/genesis_test.go
index 2fe931b24..cd548d4b1 100644
--- a/core/genesis_test.go
+++ b/core/genesis_test.go
@@ -30,11 +30,11 @@ import (
 )
 
 func TestDefaultGenesisBlock(t *testing.T) {
-	block, _ := DefaultGenesisBlock().ToBlock()
+	block := DefaultGenesisBlock().ToBlock(nil)
 	if block.Hash() != params.MainnetGenesisHash {
 		t.Errorf("wrong mainnet genesis hash, got %v, want %v", block.Hash(), params.MainnetGenesisHash)
 	}
-	block, _ = DefaultTestnetGenesisBlock().ToBlock()
+	block = DefaultTestnetGenesisBlock().ToBlock(nil)
 	if block.Hash() != params.TestnetGenesisHash {
 		t.Errorf("wrong testnet genesis hash, got %v, want %v", block.Hash(), params.TestnetGenesisHash)
 	}
@@ -118,7 +118,7 @@ func TestSetupGenesis(t *testing.T) {
 				// Commit the 'old' genesis block with Homestead transition at #2.
 				// Advance to block #4, past the homestead transition block of customg.
 				genesis := oldcustomg.MustCommit(db)
-				bc, _ := NewBlockChain(db, oldcustomg.Config, ethash.NewFullFaker(), vm.Config{})
+				bc, _ := NewBlockChain(db, nil, oldcustomg.Config, ethash.NewFullFaker(), vm.Config{})
 				defer bc.Stop()
 				bc.SetValidator(bproc{})
 				bc.InsertChain(makeBlockChainWithDiff(genesis, []int{2, 3, 4, 5}, 0))
diff --git a/core/state/database.go b/core/state/database.go
index 946625e76..36926ec69 100644
--- a/core/state/database.go
+++ b/core/state/database.go
@@ -40,16 +40,23 @@ const (
 
 // Database wraps access to tries and contract code.
 type Database interface {
-	// Accessing tries:
 	// OpenTrie opens the main account trie.
-	// OpenStorageTrie opens the storage trie of an account.
 	OpenTrie(root common.Hash) (Trie, error)
+
+	// OpenStorageTrie opens the storage trie of an account.
 	OpenStorageTrie(addrHash, root common.Hash) (Trie, error)
-	// Accessing contract code:
-	ContractCode(addrHash, codeHash common.Hash) ([]byte, error)
-	ContractCodeSize(addrHash, codeHash common.Hash) (int, error)
+
 	// CopyTrie returns an independent copy of the given trie.
 	CopyTrie(Trie) Trie
+
+	// ContractCode retrieves a particular contract's code.
+	ContractCode(addrHash, codeHash common.Hash) ([]byte, error)
+
+	// ContractCodeSize retrieves a particular contracts code's size.
+	ContractCodeSize(addrHash, codeHash common.Hash) (int, error)
+
+	// TrieDB retrieves the low level trie database used for data storage.
+	TrieDB() *trie.Database
 }
 
 // Trie is a Ethereum Merkle Trie.
@@ -57,26 +64,33 @@ type Trie interface {
 	TryGet(key []byte) ([]byte, error)
 	TryUpdate(key, value []byte) error
 	TryDelete(key []byte) error
-	CommitTo(trie.DatabaseWriter) (common.Hash, error)
+	Commit(onleaf trie.LeafCallback) (common.Hash, error)
 	Hash() common.Hash
 	NodeIterator(startKey []byte) trie.NodeIterator
 	GetKey([]byte) []byte // TODO(fjl): remove this when SecureTrie is removed
+	Prove(key []byte, fromLevel uint, proofDb ethdb.Putter) error
 }
 
 // NewDatabase creates a backing store for state. The returned database is safe for
-// concurrent use and retains cached trie nodes in memory.
+// concurrent use and retains cached trie nodes in memory. The pool is an optional
+// intermediate trie-node memory pool between the low level storage layer and the
+// high level trie abstraction.
 func NewDatabase(db ethdb.Database) Database {
 	csc, _ := lru.New(codeSizeCacheSize)
-	return &cachingDB{db: db, codeSizeCache: csc}
+	return &cachingDB{
+		db:            trie.NewDatabase(db),
+		codeSizeCache: csc,
+	}
 }
 
 type cachingDB struct {
-	db            ethdb.Database
+	db            *trie.Database
 	mu            sync.Mutex
 	pastTries     []*trie.SecureTrie
 	codeSizeCache *lru.Cache
 }
 
+// OpenTrie opens the main account trie.
 func (db *cachingDB) OpenTrie(root common.Hash) (Trie, error) {
 	db.mu.Lock()
 	defer db.mu.Unlock()
@@ -105,10 +119,12 @@ func (db *cachingDB) pushTrie(t *trie.SecureTrie) {
 	}
 }
 
+// OpenStorageTrie opens the storage trie of an account.
 func (db *cachingDB) OpenStorageTrie(addrHash, root common.Hash) (Trie, error) {
 	return trie.NewSecure(root, db.db, 0)
 }
 
+// CopyTrie returns an independent copy of the given trie.
 func (db *cachingDB) CopyTrie(t Trie) Trie {
 	switch t := t.(type) {
 	case cachedTrie:
@@ -120,14 +136,16 @@ func (db *cachingDB) CopyTrie(t Trie) Trie {
 	}
 }
 
+// ContractCode retrieves a particular contract's code.
 func (db *cachingDB) ContractCode(addrHash, codeHash common.Hash) ([]byte, error) {
-	code, err := db.db.Get(codeHash[:])
+	code, err := db.db.Node(codeHash)
 	if err == nil {
 		db.codeSizeCache.Add(codeHash, len(code))
 	}
 	return code, err
 }
 
+// ContractCodeSize retrieves a particular contracts code's size.
 func (db *cachingDB) ContractCodeSize(addrHash, codeHash common.Hash) (int, error) {
 	if cached, ok := db.codeSizeCache.Get(codeHash); ok {
 		return cached.(int), nil
@@ -139,16 +157,25 @@ func (db *cachingDB) ContractCodeSize(addrHash, codeHash common.Hash) (int, erro
 	return len(code), err
 }
 
+// TrieDB retrieves any intermediate trie-node caching layer.
+func (db *cachingDB) TrieDB() *trie.Database {
+	return db.db
+}
+
 // cachedTrie inserts its trie into a cachingDB on commit.
 type cachedTrie struct {
 	*trie.SecureTrie
 	db *cachingDB
 }
 
-func (m cachedTrie) CommitTo(dbw trie.DatabaseWriter) (common.Hash, error) {
-	root, err := m.SecureTrie.CommitTo(dbw)
+func (m cachedTrie) Commit(onleaf trie.LeafCallback) (common.Hash, error) {
+	root, err := m.SecureTrie.Commit(onleaf)
 	if err == nil {
 		m.db.pushTrie(m.SecureTrie)
 	}
 	return root, err
 }
+
+func (m cachedTrie) Prove(key []byte, fromLevel uint, proofDb ethdb.Putter) error {
+	return m.SecureTrie.Prove(key, fromLevel, proofDb)
+}
diff --git a/core/state/iterator_test.go b/core/state/iterator_test.go
index ff66ba7a9..9e46c851c 100644
--- a/core/state/iterator_test.go
+++ b/core/state/iterator_test.go
@@ -21,12 +21,13 @@ import (
 	"testing"
 
 	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/ethdb"
 )
 
 // Tests that the node iterator indeed walks over the entire database contents.
 func TestNodeIteratorCoverage(t *testing.T) {
 	// Create some arbitrary test state to iterate
-	db, mem, root, _ := makeTestState()
+	db, root, _ := makeTestState()
 
 	state, err := New(root, db)
 	if err != nil {
@@ -39,14 +40,18 @@ func TestNodeIteratorCoverage(t *testing.T) {
 			hashes[it.Hash] = struct{}{}
 		}
 	}
-
-	// Cross check the hashes and the database itself
+	// Cross check the iterated hashes and the database/nodepool content
 	for hash := range hashes {
-		if _, err := mem.Get(hash.Bytes()); err != nil {
-			t.Errorf("failed to retrieve reported node %x: %v", hash, err)
+		if _, err := db.TrieDB().Node(hash); err != nil {
+			t.Errorf("failed to retrieve reported node %x", hash)
+		}
+	}
+	for _, hash := range db.TrieDB().Nodes() {
+		if _, ok := hashes[hash]; !ok {
+			t.Errorf("state entry not reported %x", hash)
 		}
 	}
-	for _, key := range mem.Keys() {
+	for _, key := range db.TrieDB().DiskDB().(*ethdb.MemDatabase).Keys() {
 		if bytes.HasPrefix(key, []byte("secure-key-")) {
 			continue
 		}
diff --git a/core/state/state_object.go b/core/state/state_object.go
index b2378c69c..b2112bfae 100644
--- a/core/state/state_object.go
+++ b/core/state/state_object.go
@@ -25,7 +25,6 @@ import (
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/crypto"
 	"github.com/ethereum/go-ethereum/rlp"
-	"github.com/ethereum/go-ethereum/trie"
 )
 
 var emptyCodeHash = crypto.Keccak256(nil)
@@ -238,12 +237,12 @@ func (self *stateObject) updateRoot(db Database) {
 
 // CommitTrie the storage trie of the object to dwb.
 // This updates the trie root.
-func (self *stateObject) CommitTrie(db Database, dbw trie.DatabaseWriter) error {
+func (self *stateObject) CommitTrie(db Database) error {
 	self.updateTrie(db)
 	if self.dbErr != nil {
 		return self.dbErr
 	}
-	root, err := self.trie.CommitTo(dbw)
+	root, err := self.trie.Commit(nil)
 	if err == nil {
 		self.data.Root = root
 	}
diff --git a/core/state/state_test.go b/core/state/state_test.go
index bbae3685b..6d42d63d8 100644
--- a/core/state/state_test.go
+++ b/core/state/state_test.go
@@ -48,7 +48,7 @@ func (s *StateSuite) TestDump(c *checker.C) {
 	// write some of them to the trie
 	s.state.updateStateObject(obj1)
 	s.state.updateStateObject(obj2)
-	s.state.CommitTo(s.db, false)
+	s.state.Commit(false)
 
 	// check that dump contains the state objects that are in trie
 	got := string(s.state.Dump())
@@ -97,7 +97,7 @@ func (s *StateSuite) TestNull(c *checker.C) {
 	//value := common.FromHex("0x823140710bf13990e4500136726d8b55")
 	var value common.Hash
 	s.state.SetState(address, common.Hash{}, value)
-	s.state.CommitTo(s.db, false)
+	s.state.Commit(false)
 	value = s.state.GetState(address, common.Hash{})
 	if !common.EmptyHash(value) {
 		c.Errorf("expected empty hash. got %x", value)
@@ -155,7 +155,7 @@ func TestSnapshot2(t *testing.T) {
 	so0.deleted = false
 	state.setStateObject(so0)
 
-	root, _ := state.CommitTo(db, false)
+	root, _ := state.Commit(false)
 	state.Reset(root)
 
 	// and one with deleted == true
diff --git a/core/state/statedb.go b/core/state/statedb.go
index 8e29104d5..776693e24 100644
--- a/core/state/statedb.go
+++ b/core/state/statedb.go
@@ -36,6 +36,14 @@ type revision struct {
 	journalIndex int
 }
 
+var (
+	// emptyState is the known hash of an empty state trie entry.
+	emptyState = crypto.Keccak256Hash(nil)
+
+	// emptyCode is the known hash of the empty EVM bytecode.
+	emptyCode = crypto.Keccak256Hash(nil)
+)
+
 // StateDBs within the ethereum protocol are used to store anything
 // within the merkle trie. StateDBs take care of caching and storing
 // nested states. It's the general query interface to retrieve:
@@ -235,6 +243,11 @@ func (self *StateDB) GetState(a common.Address, b common.Hash) common.Hash {
 	return common.Hash{}
 }
 
+// Database retrieves the low level database supporting the lower level trie ops.
+func (self *StateDB) Database() Database {
+	return self.db
+}
+
 // StorageTrie returns the storage trie of an account.
 // The return value is a copy and is nil for non-existent accounts.
 func (self *StateDB) StorageTrie(a common.Address) Trie {
@@ -568,8 +581,8 @@ func (s *StateDB) clearJournalAndRefund() {
 	s.refund = 0
 }
 
-// CommitTo writes the state to the given database.
-func (s *StateDB) CommitTo(dbw trie.DatabaseWriter, deleteEmptyObjects bool) (root common.Hash, err error) {
+// Commit writes the state to the underlying in-memory trie database.
+func (s *StateDB) Commit(deleteEmptyObjects bool) (root common.Hash, err error) {
 	defer s.clearJournalAndRefund()
 
 	// Commit objects to the trie.
@@ -583,13 +596,11 @@ func (s *StateDB) CommitTo(dbw trie.DatabaseWriter, deleteEmptyObjects bool) (ro
 		case isDirty:
 			// Write any contract code associated with the state object
 			if stateObject.code != nil && stateObject.dirtyCode {
-				if err := dbw.Put(stateObject.CodeHash(), stateObject.code); err != nil {
-					return common.Hash{}, err
-				}
+				s.db.TrieDB().Insert(common.BytesToHash(stateObject.CodeHash()), stateObject.code)
 				stateObject.dirtyCode = false
 			}
 			// Write any storage changes in the state object to its storage trie.
-			if err := stateObject.CommitTrie(s.db, dbw); err != nil {
+			if err := stateObject.CommitTrie(s.db); err != nil {
 				return common.Hash{}, err
 			}
 			// Update the object in the main account trie.
@@ -598,7 +609,20 @@ func (s *StateDB) CommitTo(dbw trie.DatabaseWriter, deleteEmptyObjects bool) (ro
 		delete(s.stateObjectsDirty, addr)
 	}
 	// Write trie changes.
-	root, err = s.trie.CommitTo(dbw)
+	root, err = s.trie.Commit(func(leaf []byte, parent common.Hash) error {
+		var account Account
+		if err := rlp.DecodeBytes(leaf, &account); err != nil {
+			return nil
+		}
+		if account.Root != emptyState {
+			s.db.TrieDB().Reference(account.Root, parent)
+		}
+		code := common.BytesToHash(account.CodeHash)
+		if code != emptyCode {
+			s.db.TrieDB().Reference(code, parent)
+		}
+		return nil
+	})
 	log.Debug("Trie cache stats after commit", "misses", trie.CacheMisses(), "unloads", trie.CacheUnloads())
 	return root, err
 }
diff --git a/core/state/statedb_test.go b/core/state/statedb_test.go
index 5c80e3aa5..d9e3d9b79 100644
--- a/core/state/statedb_test.go
+++ b/core/state/statedb_test.go
@@ -97,10 +97,10 @@ func TestIntermediateLeaks(t *testing.T) {
 	}
 
 	// Commit and cross check the databases.
-	if _, err := transState.CommitTo(transDb, false); err != nil {
+	if _, err := transState.Commit(false); err != nil {
 		t.Fatalf("failed to commit transition state: %v", err)
 	}
-	if _, err := finalState.CommitTo(finalDb, false); err != nil {
+	if _, err := finalState.Commit(false); err != nil {
 		t.Fatalf("failed to commit final state: %v", err)
 	}
 	for _, key := range finalDb.Keys() {
@@ -122,8 +122,8 @@ func TestIntermediateLeaks(t *testing.T) {
 // https://github.com/ethereum/go-ethereum/pull/15549.
 func TestCopy(t *testing.T) {
 	// Create a random state test to copy and modify "independently"
-	mem, _ := ethdb.NewMemDatabase()
-	orig, _ := New(common.Hash{}, NewDatabase(mem))
+	db, _ := ethdb.NewMemDatabase()
+	orig, _ := New(common.Hash{}, NewDatabase(db))
 
 	for i := byte(0); i < 255; i++ {
 		obj := orig.GetOrNewStateObject(common.BytesToAddress([]byte{i}))
@@ -346,11 +346,10 @@ func (test *snapshotTest) run() bool {
 		}
 		action.fn(action, state)
 	}
-
 	// Revert all snapshots in reverse order. Each revert must yield a state
 	// that is equivalent to fresh state with all actions up the snapshot applied.
 	for sindex--; sindex >= 0; sindex-- {
-		checkstate, _ := New(common.Hash{}, NewDatabase(db))
+		checkstate, _ := New(common.Hash{}, state.Database())
 		for _, action := range test.actions[:test.snapshots[sindex]] {
 			action.fn(action, checkstate)
 		}
@@ -409,7 +408,7 @@ func (test *snapshotTest) checkEqual(state, checkstate *StateDB) error {
 
 func (s *StateSuite) TestTouchDelete(c *check.C) {
 	s.state.GetOrNewStateObject(common.Address{})
-	root, _ := s.state.CommitTo(s.db, false)
+	root, _ := s.state.Commit(false)
 	s.state.Reset(root)
 
 	snapshot := s.state.Snapshot()
@@ -417,7 +416,6 @@ func (s *StateSuite) TestTouchDelete(c *check.C) {
 	if len(s.state.stateObjectsDirty) != 1 {
 		c.Fatal("expected one dirty state object")
 	}
-
 	s.state.RevertToSnapshot(snapshot)
 	if len(s.state.stateObjectsDirty) != 0 {
 		c.Fatal("expected no dirty state object")
diff --git a/core/state/sync_test.go b/core/state/sync_test.go
index 06c572ea6..8f14a44e7 100644
--- a/core/state/sync_test.go
+++ b/core/state/sync_test.go
@@ -36,10 +36,10 @@ type testAccount struct {
 }
 
 // makeTestState create a sample test state to test node-wise reconstruction.
-func makeTestState() (Database, *ethdb.MemDatabase, common.Hash, []*testAccount) {
+func makeTestState() (Database, common.Hash, []*testAccount) {
 	// Create an empty state
-	mem, _ := ethdb.NewMemDatabase()
-	db := NewDatabase(mem)
+	diskdb, _ := ethdb.NewMemDatabase()
+	db := NewDatabase(diskdb)
 	state, _ := New(common.Hash{}, db)
 
 	// Fill it with some arbitrary data
@@ -61,10 +61,10 @@ func makeTestState() (Database, *ethdb.MemDatabase, common.Hash, []*testAccount)
 		state.updateStateObject(obj)
 		accounts = append(accounts, acc)
 	}
-	root, _ := state.CommitTo(mem, false)
+	root, _ := state.Commit(false)
 
 	// Return the generated state
-	return db, mem, root, accounts
+	return db, root, accounts
 }
 
 // checkStateAccounts cross references a reconstructed state with an expected
@@ -96,7 +96,7 @@ func checkTrieConsistency(db ethdb.Database, root common.Hash) error {
 	if v, _ := db.Get(root[:]); v == nil {
 		return nil // Consider a non existent state consistent.
 	}
-	trie, err := trie.New(root, db)
+	trie, err := trie.New(root, trie.NewDatabase(db))
 	if err != nil {
 		return err
 	}
@@ -138,7 +138,7 @@ func TestIterativeStateSyncBatched(t *testing.T)    { testIterativeStateSync(t,
 
 func testIterativeStateSync(t *testing.T, batch int) {
 	// Create a random state to copy
-	_, srcMem, srcRoot, srcAccounts := makeTestState()
+	srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb, _ := ethdb.NewMemDatabase()
@@ -148,9 +148,9 @@ func testIterativeStateSync(t *testing.T, batch int) {
 	for len(queue) > 0 {
 		results := make([]trie.SyncResult, len(queue))
 		for i, hash := range queue {
-			data, err := srcMem.Get(hash.Bytes())
+			data, err := srcDb.TrieDB().Node(hash)
 			if err != nil {
-				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+				t.Fatalf("failed to retrieve node data for %x", hash)
 			}
 			results[i] = trie.SyncResult{Hash: hash, Data: data}
 		}
@@ -170,7 +170,7 @@ func testIterativeStateSync(t *testing.T, batch int) {
 // partial results are returned, and the others sent only later.
 func TestIterativeDelayedStateSync(t *testing.T) {
 	// Create a random state to copy
-	_, srcMem, srcRoot, srcAccounts := makeTestState()
+	srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb, _ := ethdb.NewMemDatabase()
@@ -181,9 +181,9 @@ func TestIterativeDelayedStateSync(t *testing.T) {
 		// Sync only half of the scheduled nodes
 		results := make([]trie.SyncResult, len(queue)/2+1)
 		for i, hash := range queue[:len(results)] {
-			data, err := srcMem.Get(hash.Bytes())
+			data, err := srcDb.TrieDB().Node(hash)
 			if err != nil {
-				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+				t.Fatalf("failed to retrieve node data for %x", hash)
 			}
 			results[i] = trie.SyncResult{Hash: hash, Data: data}
 		}
@@ -207,7 +207,7 @@ func TestIterativeRandomStateSyncBatched(t *testing.T)    { testIterativeRandomS
 
 func testIterativeRandomStateSync(t *testing.T, batch int) {
 	// Create a random state to copy
-	_, srcMem, srcRoot, srcAccounts := makeTestState()
+	srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb, _ := ethdb.NewMemDatabase()
@@ -221,9 +221,9 @@ func testIterativeRandomStateSync(t *testing.T, batch int) {
 		// Fetch all the queued nodes in a random order
 		results := make([]trie.SyncResult, 0, len(queue))
 		for hash := range queue {
-			data, err := srcMem.Get(hash.Bytes())
+			data, err := srcDb.TrieDB().Node(hash)
 			if err != nil {
-				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+				t.Fatalf("failed to retrieve node data for %x", hash)
 			}
 			results = append(results, trie.SyncResult{Hash: hash, Data: data})
 		}
@@ -247,7 +247,7 @@ func testIterativeRandomStateSync(t *testing.T, batch int) {
 // partial results are returned (Even those randomly), others sent only later.
 func TestIterativeRandomDelayedStateSync(t *testing.T) {
 	// Create a random state to copy
-	_, srcMem, srcRoot, srcAccounts := makeTestState()
+	srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb, _ := ethdb.NewMemDatabase()
@@ -263,9 +263,9 @@ func TestIterativeRandomDelayedStateSync(t *testing.T) {
 		for hash := range queue {
 			delete(queue, hash)
 
-			data, err := srcMem.Get(hash.Bytes())
+			data, err := srcDb.TrieDB().Node(hash)
 			if err != nil {
-				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+				t.Fatalf("failed to retrieve node data for %x", hash)
 			}
 			results = append(results, trie.SyncResult{Hash: hash, Data: data})
 
@@ -292,9 +292,9 @@ func TestIterativeRandomDelayedStateSync(t *testing.T) {
 // the database.
 func TestIncompleteStateSync(t *testing.T) {
 	// Create a random state to copy
-	_, srcMem, srcRoot, srcAccounts := makeTestState()
+	srcDb, srcRoot, srcAccounts := makeTestState()
 
-	checkTrieConsistency(srcMem, srcRoot)
+	checkTrieConsistency(srcDb.TrieDB().DiskDB().(ethdb.Database), srcRoot)
 
 	// Create a destination state and sync with the scheduler
 	dstDb, _ := ethdb.NewMemDatabase()
@@ -306,9 +306,9 @@ func TestIncompleteStateSync(t *testing.T) {
 		// Fetch a batch of state nodes
 		results := make([]trie.SyncResult, len(queue))
 		for i, hash := range queue {
-			data, err := srcMem.Get(hash.Bytes())
+			data, err := srcDb.TrieDB().Node(hash)
 			if err != nil {
-				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+				t.Fatalf("failed to retrieve node data for %x", hash)
 			}
 			results[i] = trie.SyncResult{Hash: hash, Data: data}
 		}
diff --git a/core/tx_pool_test.go b/core/tx_pool_test.go
index cd11f2ba2..158b9776b 100644
--- a/core/tx_pool_test.go
+++ b/core/tx_pool_test.go
@@ -78,8 +78,8 @@ func pricedTransaction(nonce uint64, gaslimit uint64, gasprice *big.Int, key *ec
 }
 
 func setupTxPool() (*TxPool, *ecdsa.PrivateKey) {
-	db, _ := ethdb.NewMemDatabase()
-	statedb, _ := state.New(common.Hash{}, state.NewDatabase(db))
+	diskdb, _ := ethdb.NewMemDatabase()
+	statedb, _ := state.New(common.Hash{}, state.NewDatabase(diskdb))
 	blockchain := &testBlockChain{statedb, 1000000, new(event.Feed)}
 
 	key, _ := crypto.GenerateKey()
diff --git a/core/types/block.go b/core/types/block.go
index ffe317342..92b868d9d 100644
--- a/core/types/block.go
+++ b/core/types/block.go
@@ -25,6 +25,7 @@ import (
 	"sort"
 	"sync/atomic"
 	"time"
+	"unsafe"
 
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/common/hexutil"
@@ -121,6 +122,12 @@ func (h *Header) HashNoNonce() common.Hash {
 	})
 }
 
+// Size returns the approximate memory used by all internal contents. It is used
+// to approximate and limit the memory consumption of various caches.
+func (h *Header) Size() common.StorageSize {
+	return common.StorageSize(unsafe.Sizeof(*h)) + common.StorageSize(len(h.Extra)+(h.Difficulty.BitLen()+h.Number.BitLen()+h.Time.BitLen())/8)
+}
+
 func rlpHash(x interface{}) (h common.Hash) {
 	hw := sha3.NewKeccak256()
 	rlp.Encode(hw, x)
@@ -322,6 +329,8 @@ func (b *Block) HashNoNonce() common.Hash {
 	return b.header.HashNoNonce()
 }
 
+// Size returns the true RLP encoded storage size of the block, either by encoding
+// and returning it, or returning a previsouly cached value.
 func (b *Block) Size() common.StorageSize {
 	if size := b.size.Load(); size != nil {
 		return size.(common.StorageSize)
diff --git a/core/types/receipt.go b/core/types/receipt.go
index 208d54aaa..f945f6f6a 100644
--- a/core/types/receipt.go
+++ b/core/types/receipt.go
@@ -20,6 +20,7 @@ import (
 	"bytes"
 	"fmt"
 	"io"
+	"unsafe"
 
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/common/hexutil"
@@ -136,6 +137,18 @@ func (r *Receipt) statusEncoding() []byte {
 	return r.PostState
 }
 
+// Size returns the approximate memory used by all internal contents. It is used
+// to approximate and limit the memory consumption of various caches.
+func (r *Receipt) Size() common.StorageSize {
+	size := common.StorageSize(unsafe.Sizeof(*r)) + common.StorageSize(len(r.PostState))
+
+	size += common.StorageSize(len(r.Logs)) * common.StorageSize(unsafe.Sizeof(Log{}))
+	for _, log := range r.Logs {
+		size += common.StorageSize(len(log.Topics)*common.HashLength + len(log.Data))
+	}
+	return size
+}
+
 // String implements the Stringer interface.
 func (r *Receipt) String() string {
 	if len(r.PostState) == 0 {
diff --git a/core/types/transaction.go b/core/types/transaction.go
index a7ed211e4..5660582ba 100644
--- a/core/types/transaction.go
+++ b/core/types/transaction.go
@@ -206,6 +206,8 @@ func (tx *Transaction) Hash() common.Hash {
 	return v
 }
 
+// Size returns the true RLP encoded storage size of the transaction, either by
+// encoding and returning it, or returning a previsouly cached value.
 func (tx *Transaction) Size() common.StorageSize {
 	if size := tx.size.Load(); size != nil {
 		return size.(common.StorageSize)