core, eth, trie: abstract node scheme #25532 (#1123)

This PR introduces a node scheme abstraction. The interface is only implemented by `hashScheme` at the moment, but will be extended by `pathScheme` very soon.

Apart from that, a few changes are also included which is worth mentioning:

-  port the changes in the stacktrie, tracking the path prefix of nodes during commit
-  use ethdb.Database for constructing trie.Database. This is not necessary right now, but it is required for path-based used to open reverse diff freezer

Co-authored-by: rjl493456442 <garyrong0905@gmail.com>

cmd/XDC/chaincmd.go

@@ -39,6 +39,7 @@ import (
 	xdc_genesis "github.com/XinFinOrg/XDPoSChain/genesis"
 	"github.com/XinFinOrg/XDPoSChain/log"
 	"github.com/XinFinOrg/XDPoSChain/node"
+	"github.com/XinFinOrg/XDPoSChain/trie"
 	"github.com/urfave/cli/v2"
 )
 
@@ -48,7 +49,10 @@ var (
 		Name:      "init",
 		Usage:     "Bootstrap and initialize a new genesis block",
 		ArgsUsage: "<genesisPath>",
-		Flags:     slices.Concat(utils.NetworkFlags, utils.DatabaseFlags),
+		Flags: slices.Concat(
+			[]cli.Flag{utils.CachePreimagesFlag},
+			utils.NetworkFlags,
+			utils.DatabaseFlags),
 		Description: `
 The init command initializes a new genesis block and definition for the network.
 This is a destructive action and changes the network in which you will be
@@ -424,7 +428,10 @@ func dump(ctx *cli.Context) error {
 	if err != nil {
 		return err
 	}
-	state, err := state.New(root, state.NewDatabase(db))
+	config := &trie.Config{
+		Preimages: true, // always enable preimage lookup
+	}
+	state, err := state.New(root, state.NewDatabaseWithConfig(db, config))
 	if err != nil {
 		return err
 	}

core/blockchain.go

@@ -168,9 +168,11 @@ type BlockChain struct {
 	cacheConfig *CacheConfig        // Cache configuration for pruning
 
 	db         ethdb.Database                   // Low level persistent database to store final content in
-	XDCxDb ethdb.XDCxDatabase
+	XDCxDb     ethdb.XDCxDatabase               // XDCx database
 	triegc     *prque.Prque[int64, common.Hash] // Priority queue mapping block numbers to tries to gc
 	gcproc     time.Duration                    // Accumulates canonical block processing for trie dumping
+	triedb     *trie.Database                   // The database handler for maintaining trie nodes.
+	stateCache state.Database                   // State database to reuse between imports (contains state cache)
 
 	hc            *HeaderChain
 	rmLogsFeed    event.Feed
@@ -188,8 +190,6 @@ type BlockChain struct {
 	currentBlock     atomic.Value // Current head of the block chain
 	currentFastBlock atomic.Value // Current head of the fast-sync chain (may be above the block chain!)
 
-	stateCache state.Database // State database to reuse between imports (contains state cache)
-
 	bodyCache        *lru.Cache[common.Hash, *types.Body]         // Cache for the most recent block bodies
 	bodyRLPCache     *lru.Cache[common.Hash, rlp.RawValue]        // Cache for the most recent block bodies in RLP encoded format
 	receiptsCache    *lru.Cache[common.Hash, types.Receipts]      // Cache for the most recent block receipts
@@ -239,10 +239,16 @@ func NewBlockChain(db ethdb.Database, cacheConfig *CacheConfig, chainConfig *par
 		}
 	}
 
+	// Open trie database with provided config
+	triedb := trie.NewDatabaseWithConfig(db, &trie.Config{
+		Cache:     cacheConfig.TrieCleanLimit,
+		Preimages: cacheConfig.Preimages,
+	})
 	bc := &BlockChain{
 		chainConfig: chainConfig,
 		cacheConfig: cacheConfig,
 		db:          db,
+		triedb:      triedb,
 		triegc:      prque.New[int64, common.Hash](nil),
 		stateCache: state.NewDatabaseWithConfig(db, &trie.Config{
 			Cache:     cacheConfig.TrieCleanLimit,
@@ -268,6 +274,7 @@ func NewBlockChain(db ethdb.Database, cacheConfig *CacheConfig, chainConfig *par
 		rejectedLendingItem: lru.NewCache[common.Hash, interface{}](tradingstate.OrderCacheLimit),
 		finalizedTrade:      lru.NewCache[common.Hash, interface{}](tradingstate.OrderCacheLimit),
 	}
+	bc.stateCache = state.NewDatabaseWithNodeDB(bc.db, bc.triedb)
 	bc.validator = NewBlockValidator(chainConfig, bc, engine)
 	bc.prefetcher = newStatePrefetcher(chainConfig, bc, engine)
 	bc.processor = NewStateProcessor(chainConfig, bc, engine)
@@ -373,8 +380,7 @@ func (bc *BlockChain) loadLastState() error {
 	}
 	// Make sure the state associated with the block is available
 	repair := false
-	_, err := state.New(currentBlock.Root(), bc.stateCache)
+	if !bc.HasState(currentBlock.Root()) {
-	if err != nil {
 		repair = true
 	} else {
 		engine, ok := bc.Engine().(*XDPoS.XDPoS)
@@ -487,7 +493,7 @@ func (bc *BlockChain) setHeadBeyondRoot(head uint64) error {
 			if newHeadBlock == nil {
 				newHeadBlock = bc.genesisBlock
 			} else {
-				if _, err := state.New(newHeadBlock.Root(), bc.stateCache); err != nil {
+				if !bc.HasState(newHeadBlock.Root()) {
 					// Rewound state missing, rolled back to before pivot, reset to genesis
 					newHeadBlock = bc.genesisBlock
 				}
@@ -712,7 +718,7 @@ func (bc *BlockChain) repair(head **types.Block) error {
 	for {
 		// Abort if we've rewound to a head block that does have associated state
 		if (common.RollbackNumber == 0) || ((*head).Number().Uint64() < common.RollbackNumber) {
-			if _, err := state.New((*head).Root(), bc.stateCache); err == nil {
+			if bc.HasState((*head).Root()) {
 				log.Info("Rewound blockchain to past state", "number", (*head).Number(), "hash", (*head).Hash())
 				engine, ok := bc.Engine().(*XDPoS.XDPoS)
 				if ok {
@@ -1082,10 +1088,10 @@ func (bc *BlockChain) saveData() {
 	if !bc.cacheConfig.TrieDirtyDisabled {
 		var tradingTriedb *trie.Database
 		var lendingTriedb *trie.Database
-		engine, _ := bc.Engine().(*XDPoS.XDPoS)
-		triedb := bc.stateCache.TrieDB()
 		var tradingService utils.TradingService
 		var lendingService utils.LendingService
+		triedb := bc.triedb
+		engine, _ := bc.Engine().(*XDPoS.XDPoS)
 		if bc.Config().IsTIPXDCX(bc.CurrentBlock().Number()) && bc.chainConfig.XDPoS != nil && bc.CurrentBlock().NumberU64() > bc.chainConfig.XDPoS.Epoch && engine != nil {
 			tradingService = engine.GetXDCXService()
 			if tradingService != nil && tradingService.GetStateCache() != nil {
@@ -1439,7 +1445,6 @@ func (bc *BlockChain) writeBlockWithState(block *types.Block, receipts []*types.
 	if err != nil {
 		return NonStatTy, err
 	}
-	triedb := bc.stateCache.TrieDB()
 
 	tradingRoot := common.Hash{}
 	if tradingState != nil {
@@ -1474,7 +1479,7 @@ func (bc *BlockChain) writeBlockWithState(block *types.Block, receipts []*types.
 
 	// If we're running an archive node, always flush
 	if bc.cacheConfig.TrieDirtyDisabled {
-		if err := triedb.Commit(root, false); err != nil {
+		if err := bc.triedb.Commit(root, false); err != nil {
 			return NonStatTy, err
 		}
 		if tradingTrieDb != nil {
@@ -1489,7 +1494,7 @@ func (bc *BlockChain) writeBlockWithState(block *types.Block, receipts []*types.
 		}
 	} else {
 		// Full but not archive node, do proper garbage collection
-		triedb.Reference(root, common.Hash{}) // metadata reference to keep trie alive
+		bc.triedb.Reference(root, common.Hash{}) // metadata reference to keep trie alive
 		bc.triegc.Push(root, -int64(block.NumberU64()))
 		if tradingTrieDb != nil {
 			tradingTrieDb.Reference(tradingRoot, common.Hash{})
@@ -1516,11 +1521,11 @@ func (bc *BlockChain) writeBlockWithState(block *types.Block, receipts []*types.
 			//	size = size + lendingTrieDb.Size()
 			//}
 			var (
-				nodes, imgs = triedb.Size()
+				nodes, imgs = bc.triedb.Size()
 				limit       = common.StorageSize(bc.cacheConfig.TrieDirtyLimit) * 1024 * 1024
 			)
 			if nodes > limit || imgs > 4*1024*1024 {
-				triedb.Cap(limit - ethdb.IdealBatchSize)
+				bc.triedb.Cap(limit - ethdb.IdealBatchSize)
 			}
 			if bc.gcproc > bc.cacheConfig.TrieTimeLimit || chosen > lastWrite+triesInMemory {
 				// If the header is missing (canonical chain behind), we're reorging a low
@@ -1535,7 +1540,7 @@ func (bc *BlockChain) writeBlockWithState(block *types.Block, receipts []*types.
 						log.Info("State in memory for too long, committing", "time", bc.gcproc, "allowance", bc.cacheConfig.TrieTimeLimit, "optimum", float64(chosen-lastWrite)/triesInMemory)
 					}
 					// Flush an entire trie and restart the counters
-					triedb.Commit(header.Root, true)
+					bc.triedb.Commit(header.Root, true)
 					lastWrite = chosen
 					bc.gcproc = 0
 					if tradingTrieDb != nil && lendingTrieDb != nil {
@@ -1555,7 +1560,7 @@ func (bc *BlockChain) writeBlockWithState(block *types.Block, receipts []*types.
 					bc.triegc.Push(root, number)
 					break
 				}
-				triedb.Dereference(root)
+				bc.triedb.Dereference(root)
 			}
 			if tradingService != nil {
 				for !tradingService.GetTriegc().Empty() {
@@ -1831,7 +1836,7 @@ func (bc *BlockChain) insertChain(chain types.Blocks, verifySeals bool) (int, []
 			bc.UpdateBlocksHashCache(block)
 		}
 
-		dirty, _ := bc.stateCache.TrieDB().Size()
+		dirty, _ := bc.triedb.Size()
 		stats.report(chain, it.index, dirty)
 		if bc.chainConfig.XDPoS != nil {
 			engine, _ := bc.Engine().(*XDPoS.XDPoS)
@@ -2287,7 +2292,7 @@ func (bc *BlockChain) insertBlock(block *types.Block) ([]interface{}, []*types.L
 	}
 	stats.processed++
 	stats.usedGas += result.usedGas
-	dirty, _ := bc.stateCache.TrieDB().Size()
+	dirty, _ := bc.triedb.Size()
 	stats.report(types.Blocks{block}, 0, dirty)
 	if bc.chainConfig.XDPoS != nil {
 		// epoch block

core/blockchain_reader.go

@@ -0,0 +1,26 @@
+// Copyright 2021 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package core
+
+import (
+	"github.com/XinFinOrg/XDPoSChain/trie"
+)
+
+// TrieDB retrieves the low level trie database used for data storage.
+func (bc *BlockChain) TrieDB() *trie.Database {
+	return bc.triedb
+}

core/state/database.go

@@ -130,23 +130,33 @@ func NewDatabase(db ethdb.Database) Database {
 // large memory cache.
 func NewDatabaseWithConfig(db ethdb.Database, config *trie.Config) Database {
 	return &cachingDB{
-		db:            trie.NewDatabaseWithConfig(db, config),
 		disk:          db,
-		codeCache:     lru.NewSizeConstrainedCache[common.Hash, []byte](codeCacheSize),
 		codeSizeCache: lru.NewCache[common.Hash, int](codeSizeCacheSize),
+		codeCache:     lru.NewSizeConstrainedCache[common.Hash, []byte](codeCacheSize),
+		triedb:        trie.NewDatabaseWithConfig(db, config),
+	}
+}
+
+// NewDatabaseWithNodeDB creates a state database with an already initialized node database.
+func NewDatabaseWithNodeDB(db ethdb.Database, triedb *trie.Database) Database {
+	return &cachingDB{
+		disk:          db,
+		codeSizeCache: lru.NewCache[common.Hash, int](codeSizeCacheSize),
+		codeCache:     lru.NewSizeConstrainedCache[common.Hash, []byte](codeCacheSize),
+		triedb:        triedb,
 	}
 }
 
 type cachingDB struct {
-	db            *trie.Database
 	disk          ethdb.KeyValueStore
-	codeCache     *lru.SizeConstrainedCache[common.Hash, []byte]
 	codeSizeCache *lru.Cache[common.Hash, int]
+	codeCache     *lru.SizeConstrainedCache[common.Hash, []byte]
+	triedb        *trie.Database
 }
 
 // OpenTrie opens the main account trie at a specific root hash.
 func (db *cachingDB) OpenTrie(root common.Hash) (Trie, error) {
-	tr, err := trie.NewStateTrie(trie.StateTrieID(root), db.db)
+	tr, err := trie.NewStateTrie(trie.StateTrieID(root), db.triedb)
 	if err != nil {
 		return nil, err
 	}
@@ -155,7 +165,7 @@ func (db *cachingDB) OpenTrie(root common.Hash) (Trie, error) {
 
 // OpenStorageTrie opens the storage trie of an account.
 func (db *cachingDB) OpenStorageTrie(stateRoot common.Hash, addrHash, root common.Hash) (Trie, error) {
-	tr, err := trie.NewStateTrie(trie.StorageTrieID(stateRoot, addrHash, root), db.db)
+	tr, err := trie.NewStateTrie(trie.StorageTrieID(stateRoot, addrHash, root), db.triedb)
 	if err != nil {
 		return nil, err
 	}
@@ -218,5 +228,5 @@ func (db *cachingDB) DiskDB() ethdb.KeyValueStore {
 
 // TrieDB retrieves any intermediate trie-node caching layer.
 func (db *cachingDB) TrieDB() *trie.Database {
-	return db.db
+	return db.triedb
 }

core/state/iterator_test.go

@@ -21,15 +21,15 @@ import (
 	"testing"
 
 	"github.com/XinFinOrg/XDPoSChain/common"
-	"github.com/XinFinOrg/XDPoSChain/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, root, _ := makeTestState()
+	db, sdb, root, _ := makeTestState()
+	sdb.TrieDB().Commit(root, false)
 
-	state, err := New(root, db)
+	state, err := New(root, sdb)
 	if err != nil {
 		t.Fatalf("failed to create state trie at %x: %v", root, err)
 	}
@@ -42,19 +42,19 @@ func TestNodeIteratorCoverage(t *testing.T) {
 	}
 	// Cross check the iterated hashes and the database/nodepool content
 	for hash := range hashes {
-		if _, err = db.TrieDB().Node(hash); err != nil {
+		if _, err = sdb.TrieDB().Node(hash); err != nil {
-			_, err = db.ContractCode(common.Hash{}, hash)
+			_, err = sdb.ContractCode(common.Hash{}, hash)
 		}
 		if err != nil {
 			t.Errorf("failed to retrieve reported node %x", hash)
 		}
 	}
-	for _, hash := range db.TrieDB().Nodes() {
+	for _, hash := range sdb.TrieDB().Nodes() {
 		if _, ok := hashes[hash]; !ok {
 			t.Errorf("state entry not reported %x", hash)
 		}
 	}
-	it := db.DiskDB().(ethdb.Database).NewIterator(nil, nil)
+	it := db.NewIterator(nil, nil)
 	for it.Next() {
 		key := it.Key()
 		if bytes.HasPrefix(key, []byte("secure-key-")) {

core/state/sync.go

@@ -27,7 +27,7 @@ import (
 )
 
 // NewStateSync create a new state trie download scheduler.
-func NewStateSync(root common.Hash, database ethdb.KeyValueReader, onLeaf func(keys [][]byte, leaf []byte) error) *trie.Sync {
+func NewStateSync(root common.Hash, database ethdb.KeyValueReader, onLeaf func(keys [][]byte, leaf []byte) error, scheme trie.NodeScheme) *trie.Sync {
 	// Register the storage slot callback if the external callback is specified.
 	var onSlot func(keys [][]byte, path []byte, leaf []byte, parent common.Hash, parentPath []byte) error
 	if onLeaf != nil {
@@ -52,6 +52,6 @@ func NewStateSync(root common.Hash, database ethdb.KeyValueReader, onLeaf func(k
 		syncer.AddCodeEntry(common.BytesToHash(obj.CodeHash), path, parent, parentPath)
 		return nil
 	}
-	syncer = trie.NewSync(root, database, onAccount)
+	syncer = trie.NewSync(root, database, onAccount, scheme)
 	return syncer
 }

core/state/sync_test.go

@@ -40,10 +40,11 @@ type testAccount struct {
 }
 
 // makeTestState create a sample test state to test node-wise reconstruction.
-func makeTestState() (Database, common.Hash, []*testAccount) {
+func makeTestState() (ethdb.Database, Database, common.Hash, []*testAccount) {
 	// Create an empty state
-	db := NewDatabase(rawdb.NewMemoryDatabase())
+	db := rawdb.NewMemoryDatabase()
-	state, _ := New(types.EmptyRootHash, db)
+	sdb := NewDatabase(db)
+	state, _ := New(types.EmptyRootHash, sdb)
 
 	// Fill it with some arbitrary data
 	var accounts []*testAccount
@@ -64,7 +65,7 @@ func makeTestState() (Database, common.Hash, []*testAccount) {
 		if i%5 == 0 {
 			for j := byte(0); j < 5; j++ {
 				hash := crypto.Keccak256Hash([]byte{i, i, i, i, i, j, j})
-				obj.SetState(db, hash, hash)
+				obj.SetState(sdb, hash, hash)
 			}
 		}
 		state.updateStateObject(obj)
@@ -73,7 +74,7 @@ func makeTestState() (Database, common.Hash, []*testAccount) {
 	root, _ := state.Commit(false)
 
 	// Return the generated state
-	return db, root, accounts
+	return db, sdb, root, accounts
 }
 
 // checkStateAccounts cross references a reconstructed state with an expected
@@ -101,7 +102,7 @@ func checkStateAccounts(t *testing.T, db ethdb.Database, root common.Hash, accou
 }
 
 // checkTrieConsistency checks that all nodes in a (sub-)trie are indeed present.
-func checkTrieConsistency(db ethdb.KeyValueStore, root common.Hash) error {
+func checkTrieConsistency(db ethdb.Database, root common.Hash) error {
 	if v, _ := db.Get(root[:]); v == nil {
 		return nil // Consider a non existent state consistent.
 	}
@@ -133,7 +134,9 @@ func checkStateConsistency(db ethdb.Database, root common.Hash) error {
 
 // Tests that an empty state is not scheduled for syncing.
 func TestEmptyStateSync(t *testing.T) {
-	sync := NewStateSync(types.EmptyRootHash, rawdb.NewMemoryDatabase(), nil)
+	db := trie.NewDatabase(rawdb.NewMemoryDatabase())
+	empty := common.HexToHash("56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421")
+	sync := NewStateSync(empty, rawdb.NewMemoryDatabase(), nil, db.Scheme())
 	if paths, nodes, codes := sync.Missing(1); len(paths) != 0 || len(nodes) != 0 || len(codes) != 0 {
 		t.Errorf("content requested for empty state: %v, %v, %v", nodes, paths, codes)
 	}
@@ -170,7 +173,7 @@ type stateElement struct {
 
 func testIterativeStateSync(t *testing.T, count int, commit bool, bypath bool) {
 	// Create a random state to copy
-	srcDb, srcRoot, srcAccounts := makeTestState()
+	_, srcDb, srcRoot, srcAccounts := makeTestState()
 	if commit {
 		srcDb.TrieDB().Commit(srcRoot, false)
 	}
@@ -178,7 +181,7 @@ func testIterativeStateSync(t *testing.T, count int, commit bool, bypath bool) {
 
 	// Create a destination state and sync with the scheduler
 	dstDb := rawdb.NewMemoryDatabase()
-	sched := NewStateSync(srcRoot, dstDb, nil)
+	sched := NewStateSync(srcRoot, dstDb, nil, srcDb.TrieDB().Scheme())
 
 	var (
 		nodeElements []stateElement
@@ -281,11 +284,11 @@ func testIterativeStateSync(t *testing.T, count int, commit bool, bypath bool) {
 // partial results are returned, and the others sent only later.
 func TestIterativeDelayedStateSync(t *testing.T) {
 	// Create a random state to copy
-	srcDb, srcRoot, srcAccounts := makeTestState()
+	_, srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb := rawdb.NewMemoryDatabase()
-	sched := NewStateSync(srcRoot, dstDb, nil)
+	sched := NewStateSync(srcRoot, dstDb, nil, srcDb.TrieDB().Scheme())
 
 	var (
 		nodeElements []stateElement
@@ -374,11 +377,11 @@ func TestIterativeRandomStateSyncBatched(t *testing.T)    { testIterativeRandomS
 
 func testIterativeRandomStateSync(t *testing.T, count int) {
 	// Create a random state to copy
-	srcDb, srcRoot, srcAccounts := makeTestState()
+	_, srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb := rawdb.NewMemoryDatabase()
-	sched := NewStateSync(srcRoot, dstDb, nil)
+	sched := NewStateSync(srcRoot, dstDb, nil, srcDb.TrieDB().Scheme())
 
 	nodeQueue := make(map[string]stateElement)
 	codeQueue := make(map[common.Hash]struct{})
@@ -454,11 +457,11 @@ func testIterativeRandomStateSync(t *testing.T, count int) {
 // partial results are returned (Even those randomly), others sent only later.
 func TestIterativeRandomDelayedStateSync(t *testing.T) {
 	// Create a random state to copy
-	srcDb, srcRoot, srcAccounts := makeTestState()
+	_, srcDb, srcRoot, srcAccounts := makeTestState()
 
 	// Create a destination state and sync with the scheduler
 	dstDb := rawdb.NewMemoryDatabase()
-	sched := NewStateSync(srcRoot, dstDb, nil)
+	sched := NewStateSync(srcRoot, dstDb, nil, srcDb.TrieDB().Scheme())
 
 	nodeQueue := make(map[string]stateElement)
 	codeQueue := make(map[common.Hash]struct{})
@@ -544,7 +547,7 @@ func TestIterativeRandomDelayedStateSync(t *testing.T) {
 // the database.
 func TestIncompleteStateSync(t *testing.T) {
 	// Create a random state to copy
-	srcDb, srcRoot, srcAccounts := makeTestState()
+	db, srcDb, srcRoot, srcAccounts := makeTestState()
 
 	/// isCodeLookup to save some hashing
 	var isCode = make(map[common.Hash]struct{})
@@ -554,15 +557,16 @@ func TestIncompleteStateSync(t *testing.T) {
 		}
 	}
 	isCode[types.EmptyCodeHash] = struct{}{}
-	checkTrieConsistency(srcDb.DiskDB(), srcRoot)
+	checkTrieConsistency(db, srcRoot)
 
 	// Create a destination state and sync with the scheduler
 	dstDb := rawdb.NewMemoryDatabase()
-	sched := NewStateSync(srcRoot, dstDb, nil)
+	sched := NewStateSync(srcRoot, dstDb, nil, srcDb.TrieDB().Scheme())
 
 	var (
 		addedCodes  []common.Hash
-		addedNodes []common.Hash
+		addedPaths  []string
+		addedHashes []common.Hash
 	)
 	nodeQueue := make(map[string]stateElement)
 	codeQueue := make(map[common.Hash]struct{})
@@ -599,15 +603,16 @@ func TestIncompleteStateSync(t *testing.T) {
 		var nodehashes []common.Hash
 		if len(nodeQueue) > 0 {
 			results := make([]trie.NodeSyncResult, 0, len(nodeQueue))
-			for key, element := range nodeQueue {
+			for path, element := range nodeQueue {
 				data, err := srcDb.TrieDB().Node(element.hash)
 				if err != nil {
 					t.Fatalf("failed to retrieve node data for %x", element.hash)
 				}
-				results = append(results, trie.NodeSyncResult{Path: key, Data: data})
+				results = append(results, trie.NodeSyncResult{Path: path, Data: data})
 
 				if element.hash != srcRoot {
-					addedNodes = append(addedNodes, element.hash)
+					addedPaths = append(addedPaths, element.path)
+					addedHashes = append(addedHashes, element.hash)
 				}
 				nodehashes = append(nodehashes, element.hash)
 			}
@@ -655,12 +660,18 @@ func TestIncompleteStateSync(t *testing.T) {
 		}
 		rawdb.WriteCode(dstDb, node, val)
 	}
-	for _, node := range addedNodes {
+	scheme := srcDb.TrieDB().Scheme()
-		val := rawdb.ReadTrieNode(dstDb, node)
+	for i, path := range addedPaths {
-		rawdb.DeleteTrieNode(dstDb, node)
+		owner, inner := trie.ResolvePath([]byte(path))
+		hash := addedHashes[i]
+		val := scheme.ReadTrieNode(dstDb, owner, inner, hash)
+		if val == nil {
+			t.Error("missing trie node")
+		}
+		scheme.DeleteTrieNode(dstDb, owner, inner, hash)
 		if err := checkStateConsistency(dstDb, srcRoot); err == nil {
-			t.Errorf("trie inconsistency not caught, missing: %v", node.Hex())
+			t.Errorf("trie inconsistency not caught, missing: %v", path)
 		}
-		rawdb.WriteTrieNode(dstDb, node, val)
+		scheme.WriteTrieNode(dstDb, owner, inner, hash, val)
 	}
 }

eth/downloader/downloader.go

@@ -34,6 +34,7 @@ import (
 	"github.com/XinFinOrg/XDPoSChain/log"
 	"github.com/XinFinOrg/XDPoSChain/metrics"
 	"github.com/XinFinOrg/XDPoSChain/params"
+	"github.com/XinFinOrg/XDPoSChain/trie"
 )
 
 // proposeBlockHandlerFn is a callback type to handle a block by the consensus
@@ -205,6 +206,10 @@ type BlockChain interface {
 
 	// InsertReceiptChain inserts a batch of receipts into the local chain.
 	InsertReceiptChain(types.Blocks, []types.Receipts) (int, error)
+
+	// TrieDB retrieves the low level trie database used for interacting
+	// with trie nodes.
+	TrieDB() *trie.Database
 }
 
 // New creates a new downloader to fetch hashes and blocks from remote peers.

eth/downloader/downloader_test.go

@@ -44,8 +44,10 @@ func init() {
 }
 
 // downloadTester is a test simulator for mocking out local block chain.
+// TODO(daniel): remove field triedb
 type downloadTester struct {
 	downloader *Downloader
+	triedb     *trie.Database
 
 	genesis *types.Block   // Genesis blocks used by the tester and peers
 	stateDb ethdb.Database // Database used by the tester for syncing from peers
@@ -74,11 +76,16 @@ func newTester() *downloadTester {
 		ownChainTd:  map[common.Hash]*big.Int{testGenesis.Hash(): testGenesis.Difficulty()},
 	}
 	tester.stateDb = rawdb.NewMemoryDatabase()
+	tester.triedb = trie.NewDatabase(tester.stateDb)
 	tester.stateDb.Put(testGenesis.Root().Bytes(), []byte{0x00})
 	tester.downloader = New(tester.stateDb, new(event.TypeMux), tester, nil, tester.dropPeer, tester.handleProposedBlock)
 	return tester
 }
 
+func (db *downloadTester) TrieDB() *trie.Database {
+	return db.triedb
+}
+
 // terminate aborts any operations on the embedded downloader and releases all
 // held resources.
 func (dl *downloadTester) terminate() {

eth/downloader/statesync.go

@@ -291,18 +291,19 @@ type codeTask struct {
 // newStateSync creates a new state trie download scheduler. This method does not
 // yet start the sync. The user needs to call run to initiate.
 // only use fast sync but XDC only run full sync
+// TODO(daniel): remove field sched
 func newStateSync(d *Downloader, root common.Hash) *stateSync {
 	return &stateSync{
 		d:         d,
-		sched:     state.NewStateSync(root, d.stateDB, nil),
+		root:      root,
+		cancel:    make(chan struct{}),
+		done:      make(chan struct{}),
+		started:   make(chan struct{}),
+		sched:     state.NewStateSync(root, d.stateDB, nil, d.blockchain.TrieDB().Scheme()),
 		keccak:    sha3.NewLegacyKeccak256(),
 		trieTasks: make(map[string]*trieTask),
 		codeTasks: make(map[common.Hash]*codeTask),
 		deliver:   make(chan *stateReq),
-		cancel:    make(chan struct{}),
-		done:      make(chan struct{}),
-		started:   make(chan struct{}),
-		root:      root,
 	}
 }
 

tests/block_test_util.go

@@ -103,10 +103,8 @@ func (t *BlockTest) Run() error {
 
 	// import pre accounts & construct test genesis block & state root
 	db := rawdb.NewMemoryDatabase()
-	gblock, err := t.genesis(config).Commit(db)
+	gspec := t.genesis(config)
-	if err != nil {
+	gblock := gspec.MustCommit(db)
-		return err
-	}
 	if gblock.Hash() != t.json.Genesis.Hash {
 		return fmt.Errorf("genesis block hash doesn't match test: computed=%x, test=%x", gblock.Hash().Bytes(), t.json.Genesis.Hash)
 	}

trie/database.go

@@ -67,7 +67,7 @@ var (
 // behind this split design is to provide read access to RPC handlers and sync
 // servers even while the trie is executing expensive garbage collection.
 type Database struct {
-	diskdb ethdb.KeyValueStore // Persistent storage for matured trie nodes
+	diskdb ethdb.Database // Persistent storage for matured trie nodes
 
 	cleans  *fastcache.Cache            // GC friendly memory Cache of clean Node RLPs
 	dirties map[common.Hash]*cachedNode // Data and references relationships of dirty trie nodes
@@ -271,14 +271,14 @@ type Config struct {
 // NewDatabase creates a new trie database to store ephemeral trie content before
 // its written out to disk or garbage collected. No read Cache is created, so all
 // data retrievals will hit the underlying disk database.
-func NewDatabase(diskdb ethdb.KeyValueStore) *Database {
+func NewDatabase(diskdb ethdb.Database) *Database {
 	return NewDatabaseWithConfig(diskdb, nil)
 }
 
 // NewDatabaseWithConfig creates a new trie database to store ephemeral trie content
 // before its written out to disk or garbage collected. It also acts as a read cache
 // for nodes loaded from disk.
-func NewDatabaseWithConfig(diskdb ethdb.KeyValueStore, config *Config) *Database {
+func NewDatabaseWithConfig(diskdb ethdb.Database, config *Config) *Database {
 	var cleans *fastcache.Cache
 	if config != nil && config.Cache > 0 {
 		cleans = fastcache.New(config.Cache * 1024 * 1024)
@@ -891,3 +891,8 @@ func (db *Database) CommitPreimages() error {
 	}
 	return db.preimages.commit(true)
 }
+
+// Scheme returns the node scheme used in the database.
+func (db *Database) Scheme() NodeScheme {
+	return &hashScheme{}
+}

trie/database_test.go

@@ -20,13 +20,13 @@ import (
 	"testing"
 
 	"github.com/XinFinOrg/XDPoSChain/common"
-	"github.com/XinFinOrg/XDPoSChain/ethdb/memorydb"
+	"github.com/XinFinOrg/XDPoSChain/core/rawdb"
 )
 
 // Tests that the trie database returns a missing trie Node error if attempting
 // to retrieve the meta root.
 func TestDatabaseMetarootFetch(t *testing.T) {
-	db := NewDatabase(memorydb.New())
+	db := NewDatabase(rawdb.NewMemoryDatabase())
 	if _, err := db.Node(common.Hash{}); err == nil {
 		t.Fatalf("metaroot retrieval succeeded")
 	}

trie/iterator_test.go

@@ -327,7 +327,7 @@ func TestIteratorContinueAfterErrorDisk(t *testing.T)    { testIteratorContinueA
 func TestIteratorContinueAfterErrorMemonly(t *testing.T) { testIteratorContinueAfterError(t, true) }
 
 func testIteratorContinueAfterError(t *testing.T, memonly bool) {
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
 
 	tr := NewEmpty(triedb)
@@ -418,8 +418,8 @@ func TestIteratorContinueAfterSeekErrorMemonly(t *testing.T) {
 }
 
 func testIteratorContinueAfterSeekError(t *testing.T, memonly bool) {
-	// Commit test trie to Db, then remove the Node containing "bars".
+	// Commit test trie to db, then remove the node containing "bars".
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
 
 	ctr := NewEmpty(triedb)
@@ -528,7 +528,7 @@ func (l *loggingDb) Close() error {
 func makeLargeTestTrie() (*Database, *StateTrie, *loggingDb) {
 	// Create an empty trie
 	logDb := &loggingDb{0, memorydb.New()}
-	triedb := NewDatabase(logDb)
+	triedb := NewDatabase(rawdb.NewDatabase(logDb))
 	trie, _ := NewStateTrie(TrieID(common.Hash{}), triedb)
 
 	// Fill it with some arbitrary data
@@ -563,7 +563,7 @@ func TestNodeIteratorLargeTrie(t *testing.T) {
 
 func TestIteratorNodeBlob(t *testing.T) {
 	var (
-		db     = memorydb.New()
+		db     = rawdb.NewMemoryDatabase()
 		triedb = NewDatabase(db)
 		trie   = NewEmpty(triedb)
 	)

trie/schema.go

@@ -0,0 +1,96 @@
+// Copyright 2021 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package trie
+
+import (
+	"github.com/XinFinOrg/XDPoSChain/common"
+	"github.com/XinFinOrg/XDPoSChain/core/rawdb"
+	"github.com/XinFinOrg/XDPoSChain/ethdb"
+)
+
+const (
+	HashScheme = "hashScheme" // Identifier of hash based node scheme
+
+	// Path-based scheme will be introduced in the following PRs.
+	// PathScheme = "pathScheme" // Identifier of path based node scheme
+)
+
+// NodeScheme describes the scheme for interacting nodes in disk.
+type NodeScheme interface {
+	// Name returns the identifier of node scheme.
+	Name() string
+
+	// HasTrieNode checks the trie node presence with the provided node info and
+	// the associated node hash.
+	HasTrieNode(db ethdb.KeyValueReader, owner common.Hash, path []byte, hash common.Hash) bool
+
+	// ReadTrieNode retrieves the trie node from database with the provided node
+	// info and the associated node hash.
+	ReadTrieNode(db ethdb.KeyValueReader, owner common.Hash, path []byte, hash common.Hash) []byte
+
+	// WriteTrieNode writes the trie node into database with the provided node
+	// info and associated node hash.
+	WriteTrieNode(db ethdb.KeyValueWriter, owner common.Hash, path []byte, hash common.Hash, node []byte)
+
+	// DeleteTrieNode deletes the trie node from database with the provided node
+	// info and associated node hash.
+	DeleteTrieNode(db ethdb.KeyValueWriter, owner common.Hash, path []byte, hash common.Hash)
+
+	// IsTrieNode returns an indicator if the given database key is the key of
+	// trie node according to the scheme.
+	IsTrieNode(key []byte) (bool, []byte)
+}
+
+type hashScheme struct{}
+
+// Name returns the identifier of hash based scheme.
+func (scheme *hashScheme) Name() string {
+	return HashScheme
+}
+
+// HasTrieNode checks the trie node presence with the provided node info and
+// the associated node hash.
+func (scheme *hashScheme) HasTrieNode(db ethdb.KeyValueReader, owner common.Hash, path []byte, hash common.Hash) bool {
+	return rawdb.HasTrieNode(db, hash)
+}
+
+// ReadTrieNode retrieves the trie node from database with the provided node info
+// and associated node hash.
+func (scheme *hashScheme) ReadTrieNode(db ethdb.KeyValueReader, owner common.Hash, path []byte, hash common.Hash) []byte {
+	return rawdb.ReadTrieNode(db, hash)
+}
+
+// WriteTrieNode writes the trie node into database with the provided node info
+// and associated node hash.
+func (scheme *hashScheme) WriteTrieNode(db ethdb.KeyValueWriter, owner common.Hash, path []byte, hash common.Hash, node []byte) {
+	rawdb.WriteTrieNode(db, hash, node)
+}
+
+// DeleteTrieNode deletes the trie node from database with the provided node info
+// and associated node hash.
+func (scheme *hashScheme) DeleteTrieNode(db ethdb.KeyValueWriter, owner common.Hash, path []byte, hash common.Hash) {
+	rawdb.DeleteTrieNode(db, hash)
+}
+
+// IsTrieNode returns an indicator if the given database key is the key of trie
+// node according to the scheme.
+func (scheme *hashScheme) IsTrieNode(key []byte) (bool, []byte) {
+	if len(key) == common.HashLength {
+		return true, key
+	}
+	return false, nil
+}

trie/secure_trie_test.go

@@ -24,19 +24,19 @@ import (
 	"testing"
 
 	"github.com/XinFinOrg/XDPoSChain/common"
+	"github.com/XinFinOrg/XDPoSChain/core/rawdb"
 	"github.com/XinFinOrg/XDPoSChain/crypto"
-	"github.com/XinFinOrg/XDPoSChain/ethdb/memorydb"
 )
 
 func newEmptySecure() *StateTrie {
-	trie, _ := NewStateTrie(TrieID(common.Hash{}), NewDatabase(memorydb.New()))
+	trie, _ := NewStateTrie(TrieID(common.Hash{}), NewDatabase(rawdb.NewMemoryDatabase()))
 	return trie
 }
 
 // makeTestStateTrie creates a large enough secure trie for testing.
 func makeTestStateTrie() (*Database, *StateTrie, map[string][]byte) {
 	// Create an empty trie
-	triedb := NewDatabase(memorydb.New())
+	triedb := NewDatabase(rawdb.NewMemoryDatabase())
 	trie, _ := NewStateTrie(TrieID(common.Hash{}), triedb)
 
 	// Fill it with some arbitrary data

trie/stacktrie.go

@@ -26,7 +26,6 @@ import (
 
 	"github.com/XinFinOrg/XDPoSChain/common"
 	"github.com/XinFinOrg/XDPoSChain/core/types"
-	"github.com/XinFinOrg/XDPoSChain/ethdb"
 	"github.com/XinFinOrg/XDPoSChain/log"
 )
 
@@ -38,10 +37,14 @@ var stPool = sync.Pool{
 	},
 }
 
-func stackTrieFromPool(db ethdb.KeyValueWriter, owner common.Hash) *StackTrie {
+// NodeWriteFunc is used to provide all information of a dirty node for committing
+// so that callers can flush nodes into database with desired scheme.
+type NodeWriteFunc = func(owner common.Hash, path []byte, hash common.Hash, blob []byte)
+
+func stackTrieFromPool(writeFn NodeWriteFunc, owner common.Hash) *StackTrie {
 	st := stPool.Get().(*StackTrie)
-	st.db = db
 	st.owner = owner
+	st.writeFn = writeFn
 	return st
 }
 
@@ -59,36 +62,36 @@ type StackTrie struct {
 	val      []byte         // value contained by this node if it's a leaf
 	key      []byte         // key chunk covered by this (leaf|ext) node
 	children [16]*StackTrie // list of children (for branch and exts)
-	db       ethdb.KeyValueWriter // Pointer to the commit db, can be nil
+	writeFn  NodeWriteFunc  // function for committing nodes, can be nil
 }
 
 // NewStackTrie allocates and initializes an empty trie.
-func NewStackTrie(db ethdb.KeyValueWriter) *StackTrie {
+func NewStackTrie(writeFn NodeWriteFunc) *StackTrie {
 	return &StackTrie{
 		nodeType: emptyNode,
-		db:       db,
+		writeFn:  writeFn,
 	}
 }
 
 // NewStackTrieWithOwner allocates and initializes an empty trie, but with
 // the additional owner field.
-func NewStackTrieWithOwner(db ethdb.KeyValueWriter, owner common.Hash) *StackTrie {
+func NewStackTrieWithOwner(writeFn NodeWriteFunc, owner common.Hash) *StackTrie {
 	return &StackTrie{
 		owner:    owner,
 		nodeType: emptyNode,
-		db:       db,
+		writeFn:  writeFn,
 	}
 }
 
 // NewFromBinary initialises a serialized stacktrie with the given db.
-func NewFromBinary(data []byte, db ethdb.KeyValueWriter) (*StackTrie, error) {
+func NewFromBinary(data []byte, writeFn NodeWriteFunc) (*StackTrie, error) {
 	var st StackTrie
 	if err := st.UnmarshalBinary(data); err != nil {
 		return nil, err
 	}
 	// If a database is used, we need to recursively add it to every child
-	if db != nil {
+	if writeFn != nil {
-		st.setDb(db)
+		st.setWriter(writeFn)
 	}
 	return &st, nil
 }
@@ -161,25 +164,25 @@ func (st *StackTrie) unmarshalBinary(r io.Reader) error {
 	return nil
 }
 
-func (st *StackTrie) setDb(db ethdb.KeyValueWriter) {
+func (st *StackTrie) setWriter(writeFn NodeWriteFunc) {
-	st.db = db
+	st.writeFn = writeFn
 	for _, child := range st.children {
 		if child != nil {
-			child.setDb(db)
+			child.setWriter(writeFn)
 		}
 	}
 }
 
-func newLeaf(owner common.Hash, key, val []byte, db ethdb.KeyValueWriter) *StackTrie {
+func newLeaf(owner common.Hash, key, val []byte, writeFn NodeWriteFunc) *StackTrie {
-	st := stackTrieFromPool(db, owner)
+	st := stackTrieFromPool(writeFn, owner)
 	st.nodeType = leafNode
 	st.key = append(st.key, key...)
 	st.val = val
 	return st
 }
 
-func newExt(owner common.Hash, key []byte, child *StackTrie, db ethdb.KeyValueWriter) *StackTrie {
+func newExt(owner common.Hash, key []byte, child *StackTrie, writeFn NodeWriteFunc) *StackTrie {
-	st := stackTrieFromPool(db, owner)
+	st := stackTrieFromPool(writeFn, owner)
 	st.nodeType = extNode
 	st.key = append(st.key, key...)
 	st.children[0] = child
@@ -201,7 +204,7 @@ func (st *StackTrie) TryUpdate(key, value []byte) error {
 	if len(value) == 0 {
 		panic("deletion not supported")
 	}
-	st.insert(k[:len(k)-1], value)
+	st.insert(k[:len(k)-1], value, nil)
 	return nil
 }
 
@@ -213,7 +216,7 @@ func (st *StackTrie) Update(key, value []byte) {
 
 func (st *StackTrie) Reset() {
 	st.owner = common.Hash{}
-	st.db = nil
+	st.writeFn = nil
 	st.key = st.key[:0]
 	st.val = nil
 	for i := range st.children {
@@ -236,7 +239,7 @@ func (st *StackTrie) getDiffIndex(key []byte) int {
 
 // Helper function to that inserts a (key, value) pair into
 // the trie.
-func (st *StackTrie) insert(key, value []byte) {
+func (st *StackTrie) insert(key, value []byte, prefix []byte) {
 	switch st.nodeType {
 	case branchNode: /* Branch */
 		idx := int(key[0])
@@ -245,7 +248,7 @@ func (st *StackTrie) insert(key, value []byte) {
 		for i := idx - 1; i >= 0; i-- {
 			if st.children[i] != nil {
 				if st.children[i].nodeType != hashedNode {
-					st.children[i].hash()
+					st.children[i].hash(append(prefix, byte(i)))
 				}
 				break
 			}
@@ -253,9 +256,9 @@ func (st *StackTrie) insert(key, value []byte) {
 
 		// Add new child
 		if st.children[idx] == nil {
-			st.children[idx] = newLeaf(st.owner, key[1:], value, st.db)
+			st.children[idx] = newLeaf(st.owner, key[1:], value, st.writeFn)
 		} else {
-			st.children[idx].insert(key[1:], value)
+			st.children[idx].insert(key[1:], value, append(prefix, key[0]))
 		}
 
 	case extNode: /* Ext */
@@ -270,7 +273,7 @@ func (st *StackTrie) insert(key, value []byte) {
 		if diffidx == len(st.key) {
 			// Ext key and key segment are identical, recurse into
 			// the child node.
-			st.children[0].insert(key[diffidx:], value)
+			st.children[0].insert(key[diffidx:], value, append(prefix, key[:diffidx]...))
 			return
 		}
 		// Save the original part. Depending if the break is
@@ -279,14 +282,19 @@ func (st *StackTrie) insert(key, value []byte) {
 		// node directly.
 		var n *StackTrie
 		if diffidx < len(st.key)-1 {
-			n = newExt(st.owner, st.key[diffidx+1:], st.children[0], st.db)
+			// Break on the non-last byte, insert an intermediate
+			// extension. The path prefix of the newly-inserted
+			// extension should also contain the different byte.
+			n = newExt(st.owner, st.key[diffidx+1:], st.children[0], st.writeFn)
+			n.hash(append(prefix, st.key[:diffidx+1]...))
 		} else {
 			// Break on the last byte, no need to insert
-			// an extension node: reuse the current node
+			// an extension node: reuse the current node.
+			// The path prefix of the original part should
+			// still be same.
 			n = st.children[0]
+			n.hash(append(prefix, st.key...))
 		}
-		// Convert to hash
-		n.hash()
 		var p *StackTrie
 		if diffidx == 0 {
 			// the break is on the first byte, so
@@ -299,12 +307,12 @@ func (st *StackTrie) insert(key, value []byte) {
 			// the common prefix is at least one byte
 			// long, insert a new intermediate branch
 			// node.
-			st.children[0] = stackTrieFromPool(st.db, st.owner)
+			st.children[0] = stackTrieFromPool(st.writeFn, st.owner)
 			st.children[0].nodeType = branchNode
 			p = st.children[0]
 		}
 		// Create a leaf for the inserted part
-		o := newLeaf(st.owner, key[diffidx+1:], value, st.db)
+		o := newLeaf(st.owner, key[diffidx+1:], value, st.writeFn)
 
 		// Insert both child leaves where they belong:
 		origIdx := st.key[diffidx]
@@ -340,7 +348,7 @@ func (st *StackTrie) insert(key, value []byte) {
 			// Convert current node into an ext,
 			// and insert a child branch node.
 			st.nodeType = extNode
-			st.children[0] = NewStackTrieWithOwner(st.db, st.owner)
+			st.children[0] = NewStackTrieWithOwner(st.writeFn, st.owner)
 			st.children[0].nodeType = branchNode
 			p = st.children[0]
 		}
@@ -349,11 +357,11 @@ func (st *StackTrie) insert(key, value []byte) {
 		// value and another containing the new value. The child leaf
 		// is hashed directly in order to free up some memory.
 		origIdx := st.key[diffidx]
-		p.children[origIdx] = newLeaf(st.owner, st.key[diffidx+1:], st.val, st.db)
+		p.children[origIdx] = newLeaf(st.owner, st.key[diffidx+1:], st.val, st.writeFn)
-		p.children[origIdx].hash()
+		p.children[origIdx].hash(append(prefix, st.key[:diffidx+1]...))
 
 		newIdx := key[diffidx]
-		p.children[newIdx] = newLeaf(st.owner, key[diffidx+1:], value, st.db)
+		p.children[newIdx] = newLeaf(st.owner, key[diffidx+1:], value, st.writeFn)
 
 		// Finally, cut off the key part that has been passed
 		// over to the children.
@@ -384,14 +392,14 @@ func (st *StackTrie) insert(key, value []byte) {
 //   - And the 'st.type' will be 'hashedNode' AGAIN
 //
 // This method also sets 'st.type' to hashedNode, and clears 'st.key'.
-func (st *StackTrie) hash() {
+func (st *StackTrie) hash(path []byte) {
 	h := newHasher(false)
 	defer returnHasherToPool(h)
 
-	st.hashRec(h)
+	st.hashRec(h, path)
 }
 
-func (st *StackTrie) hashRec(hasher *hasher) {
+func (st *StackTrie) hashRec(hasher *hasher, path []byte) {
 	// The switch below sets this to the RLP-encoding of this node.
 	var encodedNode []byte
 
@@ -412,8 +420,7 @@ func (st *StackTrie) hashRec(hasher *hasher) {
 				nodes[i] = nilValueNode
 				continue
 			}
-
+			child.hashRec(hasher, append(path, byte(i)))
-			child.hashRec(hasher)
 			if len(child.val) < 32 {
 				nodes[i] = rawNode(child.val)
 			} else {
@@ -429,10 +436,9 @@ func (st *StackTrie) hashRec(hasher *hasher) {
 		encodedNode = hasher.encodedBytes()
 
 	case extNode:
-		st.children[0].hashRec(hasher)
+		st.children[0].hashRec(hasher, append(path, st.key...))
 
-		sz := hexToCompactInPlace(st.key)
+		n := rawShortNode{Key: hexToCompact(st.key)}
-		n := rawShortNode{Key: st.key[:sz]}
 		if len(st.children[0].val) < 32 {
 			n.Val = rawNode(st.children[0].val)
 		} else {
@@ -448,8 +454,7 @@ func (st *StackTrie) hashRec(hasher *hasher) {
 
 	case leafNode:
 		st.key = append(st.key, byte(16))
-		sz := hexToCompactInPlace(st.key)
+		n := rawShortNode{Key: hexToCompact(st.key), Val: valueNode(st.val)}
-		n := rawShortNode{Key: st.key[:sz], Val: valueNode(st.val)}
 
 		n.encode(hasher.encbuf)
 		encodedNode = hasher.encodedBytes()
@@ -468,10 +473,8 @@ func (st *StackTrie) hashRec(hasher *hasher) {
 	// Write the hash to the 'val'. We allocate a new val here to not mutate
 	// input values
 	st.val = hasher.hashData(encodedNode)
-	if st.db != nil {
+	if st.writeFn != nil {
-		// TODO! Is it safe to Put the slice here?
+		st.writeFn(st.owner, path, common.BytesToHash(st.val), encodedNode)
-		// Do all db implementations copy the value provided?
-		st.db.Put(st.val, encodedNode)
 	}
 }
 
@@ -480,12 +483,11 @@ func (st *StackTrie) Hash() (h common.Hash) {
 	hasher := newHasher(false)
 	defer returnHasherToPool(hasher)
 
-	st.hashRec(hasher)
+	st.hashRec(hasher, nil)
 	if len(st.val) == 32 {
 		copy(h[:], st.val)
 		return h
 	}
-
 	// If the node's RLP isn't 32 bytes long, the node will not
 	// be hashed, and instead contain the  rlp-encoding of the
 	// node. For the top level node, we need to force the hashing.
@@ -503,25 +505,24 @@ func (st *StackTrie) Hash() (h common.Hash) {
 // The associated database is expected, otherwise the whole commit
 // functionality should be disabled.
 func (st *StackTrie) Commit() (h common.Hash, err error) {
-	if st.db == nil {
+	if st.writeFn == nil {
 		return common.Hash{}, ErrCommitDisabled
 	}
-
 	hasher := newHasher(false)
 	defer returnHasherToPool(hasher)
 
-	st.hashRec(hasher)
+	st.hashRec(hasher, nil)
 	if len(st.val) == 32 {
 		copy(h[:], st.val)
 		return h, nil
 	}
-
 	// If the node's RLP isn't 32 bytes long, the node will not
 	// be hashed (and committed), and instead contain the rlp-encoding of the
 	// node. For the top level node, we need to force the hashing+commit.
 	hasher.sha.Reset()
 	hasher.sha.Write(st.val)
 	hasher.sha.Read(h[:])
-	st.db.Put(h[:], st.val)
+
+	st.writeFn(st.owner, nil, h, st.val)
 	return h, nil
 }

trie/stacktrie_test.go

@@ -22,8 +22,8 @@ import (
 	"testing"
 
 	"github.com/XinFinOrg/XDPoSChain/common"
+	"github.com/XinFinOrg/XDPoSChain/core/rawdb"
 	"github.com/XinFinOrg/XDPoSChain/crypto"
-	"github.com/XinFinOrg/XDPoSChain/ethdb/memorydb"
 )
 
 func TestStackTrieInsertAndHash(t *testing.T) {
@@ -188,7 +188,7 @@ func TestStackTrieInsertAndHash(t *testing.T) {
 
 func TestSizeBug(t *testing.T) {
 	st := NewStackTrie(nil)
-	nt := NewEmpty(NewDatabase(memorydb.New()))
+	nt := NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
 
 	leaf := common.FromHex("290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
 	value := common.FromHex("94cf40d0d2b44f2b66e07cace1372ca42b73cf21a3")
@@ -203,7 +203,7 @@ func TestSizeBug(t *testing.T) {
 
 func TestEmptyBug(t *testing.T) {
 	st := NewStackTrie(nil)
-	nt := NewEmpty(NewDatabase(memorydb.New()))
+	nt := NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
 
 	//leaf := common.FromHex("290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
 	//value := common.FromHex("94cf40d0d2b44f2b66e07cace1372ca42b73cf21a3")
@@ -229,7 +229,7 @@ func TestEmptyBug(t *testing.T) {
 
 func TestValLength56(t *testing.T) {
 	st := NewStackTrie(nil)
-	nt := NewEmpty(NewDatabase(memorydb.New()))
+	nt := NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
 
 	//leaf := common.FromHex("290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
 	//value := common.FromHex("94cf40d0d2b44f2b66e07cace1372ca42b73cf21a3")
@@ -254,7 +254,7 @@ func TestValLength56(t *testing.T) {
 // which causes a lot of node-within-node. This case was found via fuzzing.
 func TestUpdateSmallNodes(t *testing.T) {
 	st := NewStackTrie(nil)
-	nt := NewEmpty(NewDatabase(memorydb.New()))
+	nt := NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
 
 	kvs := []struct {
 		K string
@@ -283,7 +283,7 @@ func TestUpdateSmallNodes(t *testing.T) {
 func TestUpdateVariableKeys(t *testing.T) {
 	t.SkipNow()
 	st := NewStackTrie(nil)
-	nt := NewEmpty(NewDatabase(memorydb.New()))
+	nt := NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
 
 	kvs := []struct {
 		K string
@@ -353,7 +353,7 @@ func TestStacktrieNotModifyValues(t *testing.T) {
 func TestStacktrieSerialization(t *testing.T) {
 	var (
 		st       = NewStackTrie(nil)
-		nt       = NewEmpty(NewDatabase(memorydb.New()))
+		nt       = NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
 		keyB     = big.NewInt(1)
 		keyDelta = big.NewInt(1)
 		vals     [][]byte

trie/sync.go

@@ -65,7 +65,7 @@ type SyncPath [][]byte
 // version that can be sent over the network.
 func NewSyncPath(path []byte) SyncPath {
 	// If the hash is from the account trie, append a single item, if it
-	// is from the a storage trie, append a tuple. Note, the length 64 is
+	// is from a storage trie, append a tuple. Note, the length 64 is
 	// clashing between account leaf and storage root. It's fine though
 	// because having a trie node at 64 depth means a hash collision was
 	// found and we're long dead.
@@ -75,6 +75,22 @@ func NewSyncPath(path []byte) SyncPath {
 	return SyncPath{hexToKeybytes(path[:64]), hexToCompact(path[64:])}
 }
 
+// LeafCallback is a callback type invoked when a trie operation reaches a leaf
+// node.
+//
+// The keys is a path tuple identifying a particular trie node either in a single
+// trie (account) or a layered trie (account -> storage). Each key in the tuple
+// is in the raw format(32 bytes).
+//
+// The path is a composite hexary path identifying the trie node. All the key
+// bytes are converted to the hexary nibbles and composited with the parent path
+// if the trie node is in a layered trie.
+//
+// It's used by state sync and commit to allow handling external references
+// between account and storage tries. And also it's used in the state healing
+// for extracting the raw states(leaf nodes) with corresponding paths.
+type LeafCallback func(keys [][]byte, path []byte, leaf []byte, parent common.Hash, parentPath []byte) error
+
 // nodeRequest represents a scheduled or already in-flight trie node retrieval request.
 type nodeRequest struct {
 	hash common.Hash // Hash of the trie node to retrieve
@@ -139,6 +155,7 @@ func (batch *syncMemBatch) hasCode(hash common.Hash) bool {
 // unknown trie hashes to retrieve, accepts Node data associated with said hashes
 // and reconstructs the trie step by step until all is done.
 type Sync struct {
+	scheme   NodeScheme                   // Node scheme descriptor used in database.
 	database ethdb.KeyValueReader         // Persistent database to check for existing entries
 	membatch *syncMemBatch                // Memory buffer to avoid frequent database writes
 	nodeReqs map[string]*nodeRequest      // Pending requests pertaining to a trie node path
@@ -148,8 +165,9 @@ type Sync struct {
 }
 
 // NewSync creates a new trie data download scheduler.
-func NewSync(root common.Hash, database ethdb.KeyValueReader, callback LeafCallback) *Sync {
+func NewSync(root common.Hash, database ethdb.KeyValueReader, callback LeafCallback, scheme NodeScheme) *Sync {
 	ts := &Sync{
+		scheme:   scheme,
 		database: database,
 		membatch: newSyncMemBatch(),
 		nodeReqs: make(map[string]*nodeRequest),
@@ -172,7 +190,8 @@ func (s *Sync) AddSubTrie(root common.Hash, path []byte, parent common.Hash, par
 	if s.membatch.hasNode(path) {
 		return
 	}
-	if rawdb.HasTrieNode(s.database, root) {
+	owner, inner := ResolvePath(path)
+	if s.scheme.HasTrieNode(s.database, owner, inner, root) {
 		return
 	}
 	// Assemble the new sub-trie sync request
@@ -205,7 +224,7 @@ func (s *Sync) AddCodeEntry(hash common.Hash, path []byte, parent common.Hash, p
 		return
 	}
 	// If database says duplicate, the blob is present for sure.
-	// Note we only check the existence with new code scheme, fast
+	// Note we only check the existence with new code scheme, snap
 	// sync is expected to run with a fresh new node. Even there
 	// exists the code with legacy format, fetch and store with
 	// new scheme anyway.
@@ -329,7 +348,8 @@ func (s *Sync) ProcessNode(result NodeSyncResult) error {
 func (s *Sync) Commit(dbw ethdb.Batch) error {
 	// Dump the membatch into a database dbw
 	for path, value := range s.membatch.nodes {
-		rawdb.WriteTrieNode(dbw, s.membatch.hashes[path], value)
+		owner, inner := ResolvePath([]byte(path))
+		s.scheme.WriteTrieNode(dbw, owner, inner, s.membatch.hashes[path], value)
 	}
 	for hash, value := range s.membatch.codes {
 		rawdb.WriteCode(dbw, hash, value)
@@ -445,8 +465,11 @@ func (s *Sync) children(req *nodeRequest, object node) ([]*nodeRequest, error) {
 
 				// If database says duplicate, then at least the trie node is present
 				// and we hold the assumption that it's NOT legacy contract code.
-				chash := common.BytesToHash(node)
+				var (
-				if rawdb.HasTrieNode(s.database, chash) {
+					chash        = common.BytesToHash(node)
+					owner, inner = ResolvePath(child.path)
+				)
+				if s.scheme.HasTrieNode(s.database, owner, inner, chash) {
 					return
 				}
 				// Locally unknown node, schedule for retrieval
@@ -516,3 +539,14 @@ func (s *Sync) commitCodeRequest(req *codeRequest) error {
 	}
 	return nil
 }
+
+// ResolvePath resolves the provided composite node path by separating the
+// path in account trie if it's existent.
+func ResolvePath(path []byte) (common.Hash, []byte) {
+	var owner common.Hash
+	if len(path) >= 2*common.HashLength {
+		owner = common.BytesToHash(hexToKeybytes(path[:2*common.HashLength]))
+		path = path[2*common.HashLength:]
+	}
+	return owner, path
+}

trie/sync_test.go

@@ -22,6 +22,7 @@ import (
 	"testing"
 
 	"github.com/XinFinOrg/XDPoSChain/common"
+	"github.com/XinFinOrg/XDPoSChain/core/rawdb"
 	"github.com/XinFinOrg/XDPoSChain/core/types"
 	"github.com/XinFinOrg/XDPoSChain/crypto"
 	"github.com/XinFinOrg/XDPoSChain/ethdb/memorydb"
@@ -30,7 +31,7 @@ import (
 // makeTestTrie create a sample test trie to test node-wise reconstruction.
 func makeTestTrie() (*Database, *StateTrie, map[string][]byte) {
 	// Create an empty trie
-	triedb := NewDatabase(memorydb.New())
+	triedb := NewDatabase(rawdb.NewMemoryDatabase())
 	trie, _ := NewStateTrie(TrieID(common.Hash{}), triedb)
 
 	// Fill it with some arbitrary data
@@ -104,13 +105,13 @@ type trieElement struct {
 
 // Tests that an empty trie is not scheduled for syncing.
 func TestEmptySync(t *testing.T) {
-	dbA := NewDatabase(memorydb.New())
+	dbA := NewDatabase(rawdb.NewMemoryDatabase())
-	dbB := NewDatabase(memorydb.New())
+	dbB := NewDatabase(rawdb.NewMemoryDatabase())
 	emptyA, _ := New(TrieID(common.Hash{}), dbA)
 	emptyB, _ := New(TrieID(types.EmptyRootHash), dbB)
 
 	for i, trie := range []*Trie{emptyA, emptyB} {
-		sync := NewSync(trie.Hash(), memorydb.New(), nil)
+		sync := NewSync(trie.Hash(), memorydb.New(), nil, []*Database{dbA, dbB}[i].Scheme())
 		if paths, nodes, codes := sync.Missing(1); len(paths) != 0 || len(nodes) != 0 || len(codes) != 0 {
 			t.Errorf("test %d: content requested for empty trie: %v, %v, %v", i, paths, nodes, codes)
 		}
@@ -129,9 +130,9 @@ func testIterativeSync(t *testing.T, count int, bypath bool) {
 	srcDb, srcTrie, srcData := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
@@ -195,9 +196,9 @@ func TestIterativeDelayedSync(t *testing.T) {
 	srcDb, srcTrie, srcData := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
@@ -256,9 +257,9 @@ func testIterativeRandomSync(t *testing.T, count int) {
 	srcDb, srcTrie, srcData := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
@@ -314,9 +315,9 @@ func TestIterativeRandomDelayedSync(t *testing.T) {
 	srcDb, srcTrie, srcData := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
@@ -377,9 +378,9 @@ func TestDuplicateAvoidanceSync(t *testing.T) {
 	srcDb, srcTrie, srcData := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
@@ -440,9 +441,9 @@ func TestIncompleteSync(t *testing.T) {
 	srcDb, srcTrie, _ := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
@@ -520,9 +521,9 @@ func TestSyncOrdering(t *testing.T) {
 	srcDb, srcTrie, srcData := makeTestTrie()
 
 	// Create a destination trie and sync with the scheduler, tracking the requests
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
-	sched := NewSync(srcTrie.Hash(), diskdb, nil)
+	sched := NewSync(srcTrie.Hash(), diskdb, nil, srcDb.Scheme())
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.

trie/trie.go

@@ -27,22 +27,6 @@ import (
 	"github.com/XinFinOrg/XDPoSChain/log"
 )
 
-// LeafCallback is a callback type invoked when a trie operation reaches a leaf
-// node.
-//
-// The keys is a path tuple identifying a particular trie node either in a single
-// trie (account) or a layered trie (account -> storage). Each key in the tuple
-// is in the raw format(32 bytes).
-//
-// The path is a composite hexary path identifying the trie node. All the key
-// bytes are converted to the hexary nibbles and composited with the parent path
-// if the trie node is in a layered trie.
-//
-// It's used by state sync and commit to allow handling external references
-// between account and storage tries. And also it's used in the state healing
-// for extracting the raw states(leaf nodes) with corresponding paths.
-type LeafCallback func(keys [][]byte, path []byte, leaf []byte, parent common.Hash, parentPath []byte) error
-
 // Trie is a Merkle Patricia Trie. Use New to create a trie that sits on
 // top of a database. Whenever trie performs a commit operation, the generated
 // nodes will be gathered and returned in a set. Once the trie is committed,

trie/trie_test.go

@@ -35,7 +35,6 @@ import (
 	"github.com/XinFinOrg/XDPoSChain/core/types"
 	"github.com/XinFinOrg/XDPoSChain/crypto"
 	"github.com/XinFinOrg/XDPoSChain/ethdb"
-	"github.com/XinFinOrg/XDPoSChain/ethdb/memorydb"
 	"github.com/XinFinOrg/XDPoSChain/rlp"
 	"github.com/davecgh/go-spew/spew"
 	"golang.org/x/crypto/sha3"
@@ -67,7 +66,7 @@ func TestNull(t *testing.T) {
 
 func TestMissingRoot(t *testing.T) {
 	root := common.HexToHash("0beec7b5ea3f0fdbc95d0dd47f3c5bc275da8a33")
-	trie, err := New(TrieID(root), NewDatabase(memorydb.New()))
+	trie, err := New(TrieID(root), NewDatabase(rawdb.NewMemoryDatabase()))
 	if trie != nil {
 		t.Error("New returned non-nil trie for invalid root")
 	}
@@ -80,7 +79,7 @@ func TestMissingNodeDisk(t *testing.T)    { testMissingNode(t, false) }
 func TestMissingNodeMemonly(t *testing.T) { testMissingNode(t, true) }
 
 func testMissingNode(t *testing.T, memonly bool) {
-	diskdb := memorydb.New()
+	diskdb := rawdb.NewMemoryDatabase()
 	triedb := NewDatabase(diskdb)
 
 	trie := NewEmpty(triedb)
@@ -435,7 +434,7 @@ func runRandTestBool(rt randTest) bool {
 
 func runRandTest(rt randTest) error {
 	var (
-		triedb   = NewDatabase(memorydb.New())
+		triedb   = NewDatabase(rawdb.NewMemoryDatabase())
 		tr       = NewEmpty(triedb)
 		values   = make(map[string]string) // tracks content of the trie
 		origTrie = NewEmpty(triedb)
@@ -815,11 +814,13 @@ func TestCommitSequenceStackTrie(t *testing.T) {
 		prng := rand.New(rand.NewSource(int64(count)))
 		// This spongeDb is used to check the sequence of disk-db-writes
 		s := &spongeDb{sponge: sha3.NewLegacyKeccak256(), id: "a"}
-		db := NewDatabase(s)
+		db := NewDatabase(rawdb.NewDatabase(s))
 		trie := NewEmpty(db)
 		// Another sponge is used for the stacktrie commits
 		stackTrieSponge := &spongeDb{sponge: sha3.NewLegacyKeccak256(), id: "b"}
-		stTrie := NewStackTrie(stackTrieSponge)
+		stTrie := NewStackTrie(func(owner common.Hash, path []byte, hash common.Hash, blob []byte) {
+			db.Scheme().WriteTrieNode(stackTrieSponge, owner, path, hash, blob)
+		})
 		// Fill the trie with elements
 		for i := 0; i < count; i++ {
 			// For the stack trie, we need to do inserts in proper order
@@ -872,11 +873,13 @@ func TestCommitSequenceStackTrie(t *testing.T) {
 // not fit into 32 bytes, rlp-encoded. However, it's still the correct thing to do.
 func TestCommitSequenceSmallRoot(t *testing.T) {
 	s := &spongeDb{sponge: sha3.NewLegacyKeccak256(), id: "a"}
-	db := NewDatabase(s)
+	db := NewDatabase(rawdb.NewDatabase(s))
 	trie := NewEmpty(db)
 	// Another sponge is used for the stacktrie commits
 	stackTrieSponge := &spongeDb{sponge: sha3.NewLegacyKeccak256(), id: "b"}
-	stTrie := NewStackTrie(stackTrieSponge)
+	stTrie := NewStackTrie(func(owner common.Hash, path []byte, hash common.Hash, blob []byte) {
+		db.Scheme().WriteTrieNode(stackTrieSponge, owner, path, hash, blob)
+	})
 	// Add a single small-element to the trie(s)
 	key := make([]byte, 5)
 	key[0] = 1