Merge pull request #1889 from karalabe/fast-sync-rebase

eth/63 fast synchronization algorithm

cmd/geth/main.go

@@ -304,7 +304,7 @@ JavaScript API. See https://github.com/ethereum/go-ethereum/wiki/Javascipt-Conso
 		utils.DataDirFlag,
 		utils.BlockchainVersionFlag,
 		utils.OlympicFlag,
-		utils.EthVersionFlag,
+		utils.FastSyncFlag,
 		utils.CacheFlag,
 		utils.JSpathFlag,
 		utils.ListenPortFlag,
@@ -360,7 +360,6 @@ JavaScript API. See https://github.com/ethereum/go-ethereum/wiki/Javascipt-Conso
 		utils.SetupLogger(ctx)
 		utils.SetupNetwork(ctx)
 		utils.SetupVM(ctx)
-		utils.SetupEth(ctx)
 		if ctx.GlobalBool(utils.PProfEanbledFlag.Name) {
 			utils.StartPProf(ctx)
 		}

cmd/utils/flags.go

@@ -148,10 +148,9 @@ var (
 		Name:  "olympic",
 		Usage: "Use olympic style protocol",
 	}
-	EthVersionFlag = cli.IntFlag{
-		Name:  "eth",
-		Value: 62,
-		Usage: "Highest eth protocol to advertise (temporary, dev option)",
+	FastSyncFlag = cli.BoolFlag{
+		Name:  "fast",
+		Usage: "Enables fast syncing through state downloads",
 	}
 
 	// miner settings
@@ -425,12 +424,13 @@ func MakeEthConfig(clientID, version string, ctx *cli.Context) *eth.Config {
 	if err != nil {
 		glog.V(logger.Error).Infoln("WARNING: No etherbase set and no accounts found as default")
 	}
-
+	// Assemble the entire eth configuration and return
 	cfg := &eth.Config{
 		Name:                    common.MakeName(clientID, version),
 		DataDir:                 MustDataDir(ctx),
 		GenesisNonce:            ctx.GlobalInt(GenesisNonceFlag.Name),
 		GenesisFile:             ctx.GlobalString(GenesisFileFlag.Name),
+		FastSync:                ctx.GlobalBool(FastSyncFlag.Name),
 		BlockChainVersion:       ctx.GlobalInt(BlockchainVersionFlag.Name),
 		DatabaseCache:           ctx.GlobalInt(CacheFlag.Name),
 		SkipBcVersionCheck:      false,
@@ -499,7 +499,6 @@ func MakeEthConfig(clientID, version string, ctx *cli.Context) *eth.Config {
 
 		glog.V(logger.Info).Infoln("dev mode enabled")
 	}
-
 	return cfg
 }
 
@@ -532,18 +531,6 @@ func SetupVM(ctx *cli.Context) {
 	vm.SetJITCacheSize(ctx.GlobalInt(VMJitCacheFlag.Name))
 }
 
-// SetupEth configures the eth packages global settings
-func SetupEth(ctx *cli.Context) {
-	version := ctx.GlobalInt(EthVersionFlag.Name)
-	for len(eth.ProtocolVersions) > 0 && eth.ProtocolVersions[0] > uint(version) {
-		eth.ProtocolVersions = eth.ProtocolVersions[1:]
-		eth.ProtocolLengths = eth.ProtocolLengths[1:]
-	}
-	if len(eth.ProtocolVersions) == 0 {
-		Fatalf("No valid eth protocols remaining")
-	}
-}
-
 // MakeChain creates a chain manager from set command line flags.
 func MakeChain(ctx *cli.Context) (chain *core.BlockChain, chainDb ethdb.Database) {
 	datadir := MustDataDir(ctx)

core/bench_test.go

@@ -163,7 +163,7 @@ func benchInsertChain(b *testing.B, disk bool, gen func(int, *BlockGen)) {
 	// Generate a chain of b.N blocks using the supplied block
 	// generator function.
 	genesis := WriteGenesisBlockForTesting(db, GenesisAccount{benchRootAddr, benchRootFunds})
-	chain := GenerateChain(genesis, db, b.N, gen)
+	chain, _ := GenerateChain(genesis, db, b.N, gen)
 
 	// Time the insertion of the new chain.
 	// State and blocks are stored in the same DB.

core/block_processor.go

@@ -128,7 +128,7 @@ func (self *BlockProcessor) ApplyTransaction(gp *GasPool, statedb *state.StateDB
 	}
 
 	logs := statedb.GetLogs(tx.Hash())
-	receipt.SetLogs(logs)
+	receipt.Logs = logs
 	receipt.Bloom = types.CreateBloom(types.Receipts{receipt})
 
 	glog.V(logger.Debug).Infoln(receipt)
@@ -212,14 +212,16 @@ func (sm *BlockProcessor) Process(block *types.Block) (logs vm.Logs, receipts ty
 	defer sm.mutex.Unlock()
 
 	if sm.bc.HasBlock(block.Hash()) {
-		return nil, nil, &KnownBlockError{block.Number(), block.Hash()}
+		if _, err := state.New(block.Root(), sm.chainDb); err == nil {
+			return nil, nil, &KnownBlockError{block.Number(), block.Hash()}
+		}
 	}
-
-	if !sm.bc.HasBlock(block.ParentHash()) {
-		return nil, nil, ParentError(block.ParentHash())
+	if parent := sm.bc.GetBlock(block.ParentHash()); parent != nil {
+		if _, err := state.New(parent.Root(), sm.chainDb); err == nil {
+			return sm.processWithParent(block, parent)
+		}
 	}
-	parent := sm.bc.GetBlock(block.ParentHash())
-	return sm.processWithParent(block, parent)
+	return nil, nil, ParentError(block.ParentHash())
 }
 
 func (sm *BlockProcessor) processWithParent(block, parent *types.Block) (logs vm.Logs, receipts types.Receipts, err error) {
@@ -381,18 +383,40 @@ func (sm *BlockProcessor) GetLogs(block *types.Block) (logs vm.Logs, err error)
 	receipts := GetBlockReceipts(sm.chainDb, block.Hash())
 	// coalesce logs
 	for _, receipt := range receipts {
-		logs = append(logs, receipt.Logs()...)
+		logs = append(logs, receipt.Logs...)
 	}
 	return logs, nil
 }
 
+// ValidateHeader verifies the validity of a header, relying on the database and
+// POW behind the block processor.
+func (sm *BlockProcessor) ValidateHeader(header *types.Header, checkPow, uncle bool) error {
+	// Short circuit if the header's already known or its parent missing
+	if sm.bc.HasHeader(header.Hash()) {
+		return nil
+	}
+	if parent := sm.bc.GetHeader(header.ParentHash); parent == nil {
+		return ParentError(header.ParentHash)
+	} else {
+		return ValidateHeader(sm.Pow, header, parent, checkPow, uncle)
+	}
+}
+
+// ValidateHeaderWithParent verifies the validity of a header, relying on the database and
+// POW behind the block processor.
+func (sm *BlockProcessor) ValidateHeaderWithParent(header, parent *types.Header, checkPow, uncle bool) error {
+	if sm.bc.HasHeader(header.Hash()) {
+		return nil
+	}
+	return ValidateHeader(sm.Pow, header, parent, checkPow, uncle)
+}
+
 // See YP section 4.3.4. "Block Header Validity"
 // Validates a header. Returns an error if the header is invalid.
 func ValidateHeader(pow pow.PoW, header *types.Header, parent *types.Header, checkPow, uncle bool) error {
 	if big.NewInt(int64(len(header.Extra))).Cmp(params.MaximumExtraDataSize) == 1 {
 		return fmt.Errorf("Header extra data too long (%d)", len(header.Extra))
 	}
-
 	if uncle {
 		if header.Time.Cmp(common.MaxBig) == 1 {
 			return BlockTSTooBigErr
@@ -429,7 +453,7 @@ func ValidateHeader(pow pow.PoW, header *types.Header, parent *types.Header, che
 	if checkPow {
 		// Verify the nonce of the header. Return an error if it's not valid
 		if !pow.Verify(types.NewBlockWithHeader(header)) {
-			return ValidationError("Header's nonce is invalid (= %x)", header.Nonce)
+			return &BlockNonceErr{Hash: header.Hash(), Number: header.Number, Nonce: header.Nonce.Uint64()}
 		}
 	}
 	return nil

core/block_processor_test.go

@@ -70,16 +70,16 @@ func TestPutReceipt(t *testing.T) {
 	hash[0] = 2
 
 	receipt := new(types.Receipt)
-	receipt.SetLogs(vm.Logs{&vm.Log{
-		Address:   addr,
-		Topics:    []common.Hash{hash},
-		Data:      []byte("hi"),
-		Number:    42,
-		TxHash:    hash,
-		TxIndex:   0,
-		BlockHash: hash,
-		Index:     0,
-	}})
+	receipt.Logs = vm.Logs{&vm.Log{
+		Address:     addr,
+		Topics:      []common.Hash{hash},
+		Data:        []byte("hi"),
+		BlockNumber: 42,
+		TxHash:      hash,
+		TxIndex:     0,
+		BlockHash:   hash,
+		Index:       0,
+	}}
 
 	PutReceipts(db, types.Receipts{receipt})
 	receipt = GetReceipt(db, common.Hash{})

core/blockchain.go

@@ -18,10 +18,14 @@
 package core
 
 import (
+	crand "crypto/rand"
 	"errors"
 	"fmt"
 	"io"
+	"math"
 	"math/big"
+	mrand "math/rand"
+	"runtime"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -29,6 +33,7 @@ import (
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/core/state"
 	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/crypto"
 	"github.com/ethereum/go-ethereum/ethdb"
 	"github.com/ethereum/go-ethereum/event"
 	"github.com/ethereum/go-ethereum/logger"
@@ -36,6 +41,7 @@ import (
 	"github.com/ethereum/go-ethereum/metrics"
 	"github.com/ethereum/go-ethereum/pow"
 	"github.com/ethereum/go-ethereum/rlp"
+	"github.com/ethereum/go-ethereum/trie"
 	"github.com/hashicorp/golang-lru"
 )
 
@@ -67,9 +73,10 @@ type BlockChain struct {
 	chainmu sync.RWMutex
 	tsmu    sync.RWMutex
 
-	td              *big.Int
-	currentBlock    *types.Block
-	currentGasLimit *big.Int
+	checkpoint       int           // checkpoint counts towards the new checkpoint
+	currentHeader    *types.Header // Current head of the header chain (may be above the block chain!)
+	currentBlock     *types.Block  // Current head of the block chain
+	currentFastBlock *types.Block  // Current head of the fast-sync chain (may be above the block chain!)
 
 	headerCache  *lru.Cache // Cache for the most recent block headers
 	bodyCache    *lru.Cache // Cache for the most recent block bodies
@@ -84,7 +91,8 @@ type BlockChain struct {
 	procInterrupt int32 // interrupt signaler for block processing
 	wg            sync.WaitGroup
 
-	pow pow.PoW
+	pow  pow.PoW
+	rand *mrand.Rand
 }
 
 func NewBlockChain(chainDb ethdb.Database, pow pow.PoW, mux *event.TypeMux) (*BlockChain, error) {
@@ -107,6 +115,12 @@ func NewBlockChain(chainDb ethdb.Database, pow pow.PoW, mux *event.TypeMux) (*Bl
 		futureBlocks: futureBlocks,
 		pow:          pow,
 	}
+	// Seed a fast but crypto originating random generator
+	seed, err := crand.Int(crand.Reader, big.NewInt(math.MaxInt64))
+	if err != nil {
+		return nil, err
+	}
+	bc.rand = mrand.New(mrand.NewSource(seed.Int64()))
 
 	bc.genesisBlock = bc.GetBlockByNumber(0)
 	if bc.genesisBlock == nil {
@@ -120,20 +134,15 @@ func NewBlockChain(chainDb ethdb.Database, pow pow.PoW, mux *event.TypeMux) (*Bl
 		}
 		glog.V(logger.Info).Infoln("WARNING: Wrote default ethereum genesis block")
 	}
-	if err := bc.setLastState(); err != nil {
+	if err := bc.loadLastState(); err != nil {
 		return nil, err
 	}
 	// Check the current state of the block hashes and make sure that we do not have any of the bad blocks in our chain
 	for hash, _ := range BadHashes {
-		if block := bc.GetBlock(hash); block != nil {
-			glog.V(logger.Error).Infof("Found bad hash. Reorganising chain to state %x\n", block.ParentHash().Bytes()[:4])
-			block = bc.GetBlock(block.ParentHash())
-			if block == nil {
-				glog.Fatal("Unable to complete. Parent block not found. Corrupted DB?")
-			}
-			bc.SetHead(block)
-
-			glog.V(logger.Error).Infoln("Chain reorg was successfull. Resuming normal operation")
+		if header := bc.GetHeader(hash); header != nil {
+			glog.V(logger.Error).Infof("Found bad hash, rewinding chain to block #%d [%x…]", header.Number, header.ParentHash[:4])
+			bc.SetHead(header.Number.Uint64() - 1)
+			glog.V(logger.Error).Infoln("Chain rewind was successful, resuming normal operation")
 		}
 	}
 	// Take ownership of this particular state
@@ -141,30 +150,146 @@ func NewBlockChain(chainDb ethdb.Database, pow pow.PoW, mux *event.TypeMux) (*Bl
 	return bc, nil
 }
 
-func (bc *BlockChain) SetHead(head *types.Block) {
+// loadLastState loads the last known chain state from the database. This method
+// assumes that the chain manager mutex is held.
+func (self *BlockChain) loadLastState() error {
+	// Restore the last known head block
+	head := GetHeadBlockHash(self.chainDb)
+	if head == (common.Hash{}) {
+		// Corrupt or empty database, init from scratch
+		self.Reset()
+	} else {
+		if block := self.GetBlock(head); block != nil {
+			// Block found, set as the current head
+			self.currentBlock = block
+		} else {
+			// Corrupt or empty database, init from scratch
+			self.Reset()
+		}
+	}
+	// Restore the last known head header
+	self.currentHeader = self.currentBlock.Header()
+	if head := GetHeadHeaderHash(self.chainDb); head != (common.Hash{}) {
+		if header := self.GetHeader(head); header != nil {
+			self.currentHeader = header
+		}
+	}
+	// Restore the last known head fast block
+	self.currentFastBlock = self.currentBlock
+	if head := GetHeadFastBlockHash(self.chainDb); head != (common.Hash{}) {
+		if block := self.GetBlock(head); block != nil {
+			self.currentFastBlock = block
+		}
+	}
+	// Issue a status log and return
+	headerTd := self.GetTd(self.currentHeader.Hash())
+	blockTd := self.GetTd(self.currentBlock.Hash())
+	fastTd := self.GetTd(self.currentFastBlock.Hash())
+
+	glog.V(logger.Info).Infof("Last header: #%d [%x…] TD=%v", self.currentHeader.Number, self.currentHeader.Hash().Bytes()[:4], headerTd)
+	glog.V(logger.Info).Infof("Last block: #%d [%x…] TD=%v", self.currentBlock.Number(), self.currentBlock.Hash().Bytes()[:4], blockTd)
+	glog.V(logger.Info).Infof("Fast block: #%d [%x…] TD=%v", self.currentFastBlock.Number(), self.currentFastBlock.Hash().Bytes()[:4], fastTd)
+
+	return nil
+}
+
+// SetHead rewinds the local chain to a new head. In the case of headers, everything
+// above the new head will be deleted and the new one set. In the case of blocks
+// though, the head may be further rewound if block bodies are missing (non-archive
+// nodes after a fast sync).
+func (bc *BlockChain) SetHead(head uint64) {
 	bc.mu.Lock()
 	defer bc.mu.Unlock()
 
-	for block := bc.currentBlock; block != nil && block.Hash() != head.Hash(); block = bc.GetBlock(block.ParentHash()) {
-		DeleteBlock(bc.chainDb, block.Hash())
+	// Figure out the highest known canonical headers and/or blocks
+	height := uint64(0)
+	if bc.currentHeader != nil {
+		if hh := bc.currentHeader.Number.Uint64(); hh > height {
+			height = hh
+		}
+	}
+	if bc.currentBlock != nil {
+		if bh := bc.currentBlock.NumberU64(); bh > height {
+			height = bh
+		}
+	}
+	if bc.currentFastBlock != nil {
+		if fbh := bc.currentFastBlock.NumberU64(); fbh > height {
+			height = fbh
+		}
+	}
+	// Gather all the hashes that need deletion
+	drop := make(map[common.Hash]struct{})
+
+	for bc.currentHeader != nil && bc.currentHeader.Number.Uint64() > head {
+		drop[bc.currentHeader.Hash()] = struct{}{}
+		bc.currentHeader = bc.GetHeader(bc.currentHeader.ParentHash)
+	}
+	for bc.currentBlock != nil && bc.currentBlock.NumberU64() > head {
+		drop[bc.currentBlock.Hash()] = struct{}{}
+		bc.currentBlock = bc.GetBlock(bc.currentBlock.ParentHash())
+	}
+	for bc.currentFastBlock != nil && bc.currentFastBlock.NumberU64() > head {
+		drop[bc.currentFastBlock.Hash()] = struct{}{}
+		bc.currentFastBlock = bc.GetBlock(bc.currentFastBlock.ParentHash())
+	}
+	// Roll back the canonical chain numbering
+	for i := height; i > head; i-- {
+		DeleteCanonicalHash(bc.chainDb, i)
 	}
+	// Delete everything found by the above rewind
+	for hash, _ := range drop {
+		DeleteHeader(bc.chainDb, hash)
+		DeleteBody(bc.chainDb, hash)
+		DeleteTd(bc.chainDb, hash)
+	}
+	// Clear out any stale content from the caches
 	bc.headerCache.Purge()
 	bc.bodyCache.Purge()
 	bc.bodyRLPCache.Purge()
 	bc.blockCache.Purge()
 	bc.futureBlocks.Purge()
 
-	bc.currentBlock = head
-	bc.setTotalDifficulty(bc.GetTd(head.Hash()))
-	bc.insert(head)
-	bc.setLastState()
+	// Update all computed fields to the new head
+	if bc.currentBlock == nil {
+		bc.currentBlock = bc.genesisBlock
+	}
+	if bc.currentHeader == nil {
+		bc.currentHeader = bc.genesisBlock.Header()
+	}
+	if bc.currentFastBlock == nil {
+		bc.currentFastBlock = bc.genesisBlock
+	}
+	if err := WriteHeadBlockHash(bc.chainDb, bc.currentBlock.Hash()); err != nil {
+		glog.Fatalf("failed to reset head block hash: %v", err)
+	}
+	if err := WriteHeadHeaderHash(bc.chainDb, bc.currentHeader.Hash()); err != nil {
+		glog.Fatalf("failed to reset head header hash: %v", err)
+	}
+	if err := WriteHeadFastBlockHash(bc.chainDb, bc.currentFastBlock.Hash()); err != nil {
+		glog.Fatalf("failed to reset head fast block hash: %v", err)
+	}
+	bc.loadLastState()
 }
 
-func (self *BlockChain) Td() *big.Int {
-	self.mu.RLock()
-	defer self.mu.RUnlock()
+// FastSyncCommitHead sets the current head block to the one defined by the hash
+// irrelevant what the chain contents were prior.
+func (self *BlockChain) FastSyncCommitHead(hash common.Hash) error {
+	// Make sure that both the block as well at its state trie exists
+	block := self.GetBlock(hash)
+	if block == nil {
+		return fmt.Errorf("non existent block [%x…]", hash[:4])
+	}
+	if _, err := trie.NewSecure(block.Root(), self.chainDb); err != nil {
+		return err
+	}
+	// If all checks out, manually set the head block
+	self.mu.Lock()
+	self.currentBlock = block
+	self.mu.Unlock()
 
-	return new(big.Int).Set(self.td)
+	glog.V(logger.Info).Infof("committed block #%d [%x…] as new head", block.Number(), hash[:4])
+	return nil
 }
 
 func (self *BlockChain) GasLimit() *big.Int {
@@ -181,6 +306,17 @@ func (self *BlockChain) LastBlockHash() common.Hash {
 	return self.currentBlock.Hash()
 }
 
+// CurrentHeader retrieves the current head header of the canonical chain. The
+// header is retrieved from the blockchain's internal cache.
+func (self *BlockChain) CurrentHeader() *types.Header {
+	self.mu.RLock()
+	defer self.mu.RUnlock()
+
+	return self.currentHeader
+}
+
+// CurrentBlock retrieves the current head block of the canonical chain. The
+// block is retrieved from the blockchain's internal cache.
 func (self *BlockChain) CurrentBlock() *types.Block {
 	self.mu.RLock()
 	defer self.mu.RUnlock()
@@ -188,11 +324,20 @@ func (self *BlockChain) CurrentBlock() *types.Block {
 	return self.currentBlock
 }
 
+// CurrentFastBlock retrieves the current fast-sync head block of the canonical
+// chain. The block is retrieved from the blockchain's internal cache.
+func (self *BlockChain) CurrentFastBlock() *types.Block {
+	self.mu.RLock()
+	defer self.mu.RUnlock()
+
+	return self.currentFastBlock
+}
+
 func (self *BlockChain) Status() (td *big.Int, currentBlock common.Hash, genesisBlock common.Hash) {
 	self.mu.RLock()
 	defer self.mu.RUnlock()
 
-	return new(big.Int).Set(self.td), self.currentBlock.Hash(), self.genesisBlock.Hash()
+	return self.GetTd(self.currentBlock.Hash()), self.currentBlock.Hash(), self.genesisBlock.Hash()
 }
 
 func (self *BlockChain) SetProcessor(proc types.BlockProcessor) {
@@ -203,26 +348,6 @@ func (self *BlockChain) State() (*state.StateDB, error) {
 	return state.New(self.CurrentBlock().Root(), self.chainDb)
 }
 
-func (bc *BlockChain) setLastState() error {
-	head := GetHeadBlockHash(bc.chainDb)
-	if head != (common.Hash{}) {
-		block := bc.GetBlock(head)
-		if block != nil {
-			bc.currentBlock = block
-		}
-	} else {
-		bc.Reset()
-	}
-	bc.td = bc.GetTd(bc.currentBlock.Hash())
-	bc.currentGasLimit = CalcGasLimit(bc.currentBlock)
-
-	if glog.V(logger.Info) {
-		glog.Infof("Last block (#%v) %x TD=%v\n", bc.currentBlock.Number(), bc.currentBlock.Hash(), bc.td)
-	}
-
-	return nil
-}
-
 // Reset purges the entire blockchain, restoring it to its genesis state.
 func (bc *BlockChain) Reset() {
 	bc.ResetWithGenesisBlock(bc.genesisBlock)
@@ -231,20 +356,13 @@ func (bc *BlockChain) Reset() {
 // ResetWithGenesisBlock purges the entire blockchain, restoring it to the
 // specified genesis state.
 func (bc *BlockChain) ResetWithGenesisBlock(genesis *types.Block) {
+	// Dump the entire block chain and purge the caches
+	bc.SetHead(0)
+
 	bc.mu.Lock()
 	defer bc.mu.Unlock()
 
-	// Dump the entire block chain and purge the caches
-	for block := bc.currentBlock; block != nil; block = bc.GetBlock(block.ParentHash()) {
-		DeleteBlock(bc.chainDb, block.Hash())
-	}
-	bc.headerCache.Purge()
-	bc.bodyCache.Purge()
-	bc.bodyRLPCache.Purge()
-	bc.blockCache.Purge()
-	bc.futureBlocks.Purge()
-
-	// Prepare the genesis block and reinitialize the chain
+	// Prepare the genesis block and reinitialise the chain
 	if err := WriteTd(bc.chainDb, genesis.Hash(), genesis.Difficulty()); err != nil {
 		glog.Fatalf("failed to write genesis block TD: %v", err)
 	}
@@ -254,7 +372,8 @@ func (bc *BlockChain) ResetWithGenesisBlock(genesis *types.Block) {
 	bc.genesisBlock = genesis
 	bc.insert(bc.genesisBlock)
 	bc.currentBlock = bc.genesisBlock
-	bc.setTotalDifficulty(genesis.Difficulty())
+	bc.currentHeader = bc.genesisBlock.Header()
+	bc.currentFastBlock = bc.genesisBlock
 }
 
 // Export writes the active chain to the given writer.
@@ -290,17 +409,30 @@ func (self *BlockChain) ExportN(w io.Writer, first uint64, last uint64) error {
 	return nil
 }
 
-// insert injects a block into the current chain block chain. Note, this function
-// assumes that the `mu` mutex is held!
+// insert injects a new head block into the current block chain. This method
+// assumes that the block is indeed a true head. It will also reset the head
+// header and the head fast sync block to this very same block to prevent them
+// from pointing to a possibly old canonical chain (i.e. side chain by now).
+//
+// Note, this function assumes that the `mu` mutex is held!
 func (bc *BlockChain) insert(block *types.Block) {
 	// 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 {
 		glog.Fatalf("failed to insert block number: %v", err)
 	}
 	if err := WriteHeadBlockHash(bc.chainDb, block.Hash()); err != nil {
-		glog.Fatalf("failed to insert block number: %v", err)
+		glog.Fatalf("failed to insert head block hash: %v", err)
+	}
+	if err := WriteHeadHeaderHash(bc.chainDb, block.Hash()); err != nil {
+		glog.Fatalf("failed to insert head header hash: %v", err)
 	}
+	if err := WriteHeadFastBlockHash(bc.chainDb, block.Hash()); err != nil {
+		glog.Fatalf("failed to insert head fast block hash: %v", err)
+	}
+	// Update the internal state with the head block
 	bc.currentBlock = block
+	bc.currentHeader = block.Header()
+	bc.currentFastBlock = block
 }
 
 // Accessors
@@ -456,19 +588,15 @@ func (self *BlockChain) GetBlocksFromHash(hash common.Hash, n int) (blocks []*ty
 	return
 }
 
-func (self *BlockChain) GetUnclesInChain(block *types.Block, length int) (uncles []*types.Header) {
+// GetUnclesInChain retrieves all the uncles from a given block backwards until
+// a specific distance is reached.
+func (self *BlockChain) GetUnclesInChain(block *types.Block, length int) []*types.Header {
+	uncles := []*types.Header{}
 	for i := 0; block != nil && i < length; i++ {
 		uncles = append(uncles, block.Uncles()...)
 		block = self.GetBlock(block.ParentHash())
 	}
-
-	return
-}
-
-// setTotalDifficulty updates the TD of the chain manager. Note, this function
-// assumes that the `mu` mutex is held!
-func (bc *BlockChain) setTotalDifficulty(td *big.Int) {
-	bc.td = new(big.Int).Set(td)
+	return uncles
 }
 
 func (bc *BlockChain) Stop() {
@@ -504,6 +632,337 @@ const (
 	SideStatTy
 )
 
+// writeHeader writes a header into the local chain, given that its parent is
+// already known. If the total difficulty of the newly inserted header becomes
+// greater than the current known TD, the canonical chain is re-routed.
+//
+// Note: This method is not concurrent-safe with inserting blocks simultaneously
+// into the chain, as side effects caused by reorganisations cannot be emulated
+// without the real blocks. Hence, writing headers directly should only be done
+// in two scenarios: pure-header mode of operation (light clients), or properly
+// separated header/block phases (non-archive clients).
+func (self *BlockChain) writeHeader(header *types.Header) error {
+	self.wg.Add(1)
+	defer self.wg.Done()
+
+	// Calculate the total difficulty of the header
+	ptd := self.GetTd(header.ParentHash)
+	if ptd == nil {
+		return ParentError(header.ParentHash)
+	}
+	td := new(big.Int).Add(header.Difficulty, ptd)
+
+	// Make sure no inconsistent state is leaked during insertion
+	self.mu.Lock()
+	defer self.mu.Unlock()
+
+	// If the total difficulty is higher than our known, add it to the canonical chain
+	if td.Cmp(self.GetTd(self.currentHeader.Hash())) > 0 {
+		// Delete any canonical number assignments above the new head
+		for i := header.Number.Uint64() + 1; GetCanonicalHash(self.chainDb, i) != (common.Hash{}); i++ {
+			DeleteCanonicalHash(self.chainDb, i)
+		}
+		// Overwrite any stale canonical number assignments
+		head := self.GetHeader(header.ParentHash)
+		for GetCanonicalHash(self.chainDb, head.Number.Uint64()) != head.Hash() {
+			WriteCanonicalHash(self.chainDb, head.Hash(), head.Number.Uint64())
+			head = self.GetHeader(head.ParentHash)
+		}
+		// Extend the canonical chain with the new header
+		if err := WriteCanonicalHash(self.chainDb, header.Hash(), header.Number.Uint64()); err != nil {
+			glog.Fatalf("failed to insert header number: %v", err)
+		}
+		if err := WriteHeadHeaderHash(self.chainDb, header.Hash()); err != nil {
+			glog.Fatalf("failed to insert head header hash: %v", err)
+		}
+		self.currentHeader = types.CopyHeader(header)
+	}
+	// Irrelevant of the canonical status, write the header itself to the database
+	if err := WriteTd(self.chainDb, header.Hash(), td); err != nil {
+		glog.Fatalf("failed to write header total difficulty: %v", err)
+	}
+	if err := WriteHeader(self.chainDb, header); err != nil {
+		glog.Fatalf("filed to write header contents: %v", err)
+	}
+	return nil
+}
+
+// InsertHeaderChain attempts to insert the given header chain in to the local
+// chain, possibly creating a reorg. If an error is returned, it will return the
+// index number of the failing header as well an error describing what went wrong.
+//
+// The verify parameter can be used to fine tune whether nonce verification
+// should be done or not. The reason behind the optional check is because some
+// of the header retrieval mechanisms already need to verfy nonces, as well as
+// because nonces can be verified sparsely, not needing to check each.
+func (self *BlockChain) InsertHeaderChain(chain []*types.Header, checkFreq int) (int, error) {
+	self.wg.Add(1)
+	defer self.wg.Done()
+
+	// Make sure only one thread manipulates the chain at once
+	self.chainmu.Lock()
+	defer self.chainmu.Unlock()
+
+	// Collect some import statistics to report on
+	stats := struct{ processed, ignored int }{}
+	start := time.Now()
+
+	// Generate the list of headers that should be POW verified
+	verify := make([]bool, len(chain))
+	for i := 0; i < len(verify)/checkFreq; i++ {
+		index := i*checkFreq + self.rand.Intn(checkFreq)
+		if index >= len(verify) {
+			index = len(verify) - 1
+		}
+		verify[index] = true
+	}
+	verify[len(verify)-1] = true // Last should always be verified to avoid junk
+
+	// Create the header verification task queue and worker functions
+	tasks := make(chan int, len(chain))
+	for i := 0; i < len(chain); i++ {
+		tasks <- i
+	}
+	close(tasks)
+
+	errs, failed := make([]error, len(tasks)), int32(0)
+	process := func(worker int) {
+		for index := range tasks {
+			header, hash := chain[index], chain[index].Hash()
+
+			// Short circuit insertion if shutting down or processing failed
+			if atomic.LoadInt32(&self.procInterrupt) == 1 {
+				return
+			}
+			if atomic.LoadInt32(&failed) > 0 {
+				return
+			}
+			// Short circuit if the header is bad or already known
+			if BadHashes[hash] {
+				errs[index] = BadHashError(hash)
+				atomic.AddInt32(&failed, 1)
+				return
+			}
+			if self.HasHeader(hash) {
+				continue
+			}
+			// Verify that the header honors the chain parameters
+			checkPow := verify[index]
+
+			var err error
+			if index == 0 {
+				err = self.processor.ValidateHeader(header, checkPow, false)
+			} else {
+				err = self.processor.ValidateHeaderWithParent(header, chain[index-1], checkPow, false)
+			}
+			if err != nil {
+				errs[index] = err
+				atomic.AddInt32(&failed, 1)
+				return
+			}
+		}
+	}
+	// Start as many worker threads as goroutines allowed
+	pending := new(sync.WaitGroup)
+	for i := 0; i < runtime.GOMAXPROCS(0); i++ {
+		pending.Add(1)
+		go func(id int) {
+			defer pending.Done()
+			process(id)
+		}(i)
+	}
+	pending.Wait()
+
+	// If anything failed, report
+	if failed > 0 {
+		for i, err := range errs {
+			if err != nil {
+				return i, err
+			}
+		}
+	}
+	// All headers passed verification, import them into the database
+	for i, header := range chain {
+		// Short circuit insertion if shutting down
+		if atomic.LoadInt32(&self.procInterrupt) == 1 {
+			glog.V(logger.Debug).Infoln("premature abort during header chain processing")
+			break
+		}
+		hash := header.Hash()
+
+		// If the header's already known, skip it, otherwise store
+		if self.HasHeader(hash) {
+			stats.ignored++
+			continue
+		}
+		if err := self.writeHeader(header); err != nil {
+			return i, err
+		}
+		stats.processed++
+	}
+	// Report some public statistics so the user has a clue what's going on
+	first, last := chain[0], chain[len(chain)-1]
+	glog.V(logger.Info).Infof("imported %d header(s) (%d ignored) in %v. #%v [%x… / %x…]", stats.processed, stats.ignored,
+		time.Since(start), last.Number, first.Hash().Bytes()[:4], last.Hash().Bytes()[:4])
+
+	return 0, nil
+}
+
+// Rollback is designed to remove a chain of links from the database that aren't
+// certain enough to be valid.
+func (self *BlockChain) Rollback(chain []common.Hash) {
+	self.mu.Lock()
+	defer self.mu.Unlock()
+
+	for i := len(chain) - 1; i >= 0; i-- {
+		hash := chain[i]
+
+		if self.currentHeader.Hash() == hash {
+			self.currentHeader = self.GetHeader(self.currentHeader.ParentHash)
+			WriteHeadHeaderHash(self.chainDb, self.currentHeader.Hash())
+		}
+		if self.currentFastBlock.Hash() == hash {
+			self.currentFastBlock = self.GetBlock(self.currentFastBlock.ParentHash())
+			WriteHeadFastBlockHash(self.chainDb, self.currentFastBlock.Hash())
+		}
+		if self.currentBlock.Hash() == hash {
+			self.currentBlock = self.GetBlock(self.currentBlock.ParentHash())
+			WriteHeadBlockHash(self.chainDb, self.currentBlock.Hash())
+		}
+	}
+}
+
+// InsertReceiptChain attempts to complete an already existing header chain with
+// transaction and receipt data.
+func (self *BlockChain) InsertReceiptChain(blockChain types.Blocks, receiptChain []types.Receipts) (int, error) {
+	self.wg.Add(1)
+	defer self.wg.Done()
+
+	// Collect some import statistics to report on
+	stats := struct{ processed, ignored int32 }{}
+	start := time.Now()
+
+	// Create the block importing task queue and worker functions
+	tasks := make(chan int, len(blockChain))
+	for i := 0; i < len(blockChain) && i < len(receiptChain); i++ {
+		tasks <- i
+	}
+	close(tasks)
+
+	errs, failed := make([]error, len(tasks)), int32(0)
+	process := func(worker int) {
+		for index := range tasks {
+			block, receipts := blockChain[index], receiptChain[index]
+
+			// Short circuit insertion if shutting down or processing failed
+			if atomic.LoadInt32(&self.procInterrupt) == 1 {
+				return
+			}
+			if atomic.LoadInt32(&failed) > 0 {
+				return
+			}
+			// Short circuit if the owner header is unknown
+			if !self.HasHeader(block.Hash()) {
+				errs[index] = fmt.Errorf("containing header #%d [%x…] unknown", block.Number(), block.Hash().Bytes()[:4])
+				atomic.AddInt32(&failed, 1)
+				return
+			}
+			// Skip if the entire data is already known
+			if self.HasBlock(block.Hash()) {
+				atomic.AddInt32(&stats.ignored, 1)
+				continue
+			}
+			// Compute all the non-consensus fields of the receipts
+			transactions, logIndex := block.Transactions(), uint(0)
+			for j := 0; j < len(receipts); j++ {
+				// The transaction hash can be retrieved from the transaction itself
+				receipts[j].TxHash = transactions[j].Hash()
+
+				// The contract address can be derived from the transaction itself
+				if MessageCreatesContract(transactions[j]) {
+					from, _ := transactions[j].From()
+					receipts[j].ContractAddress = crypto.CreateAddress(from, transactions[j].Nonce())
+				}
+				// The used gas can be calculated based on previous receipts
+				if j == 0 {
+					receipts[j].GasUsed = new(big.Int).Set(receipts[j].CumulativeGasUsed)
+				} else {
+					receipts[j].GasUsed = new(big.Int).Sub(receipts[j].CumulativeGasUsed, receipts[j-1].CumulativeGasUsed)
+				}
+				// The derived log fields can simply be set from the block and transaction
+				for k := 0; k < len(receipts[j].Logs); k++ {
+					receipts[j].Logs[k].BlockNumber = block.NumberU64()
+					receipts[j].Logs[k].BlockHash = block.Hash()
+					receipts[j].Logs[k].TxHash = receipts[j].TxHash
+					receipts[j].Logs[k].TxIndex = uint(j)
+					receipts[j].Logs[k].Index = logIndex
+					logIndex++
+				}
+			}
+			// Write all the data out into the database
+			if err := WriteBody(self.chainDb, block.Hash(), &types.Body{block.Transactions(), block.Uncles()}); err != nil {
+				errs[index] = fmt.Errorf("failed to write block body: %v", err)
+				atomic.AddInt32(&failed, 1)
+				glog.Fatal(errs[index])
+				return
+			}
+			if err := PutBlockReceipts(self.chainDb, block.Hash(), receipts); err != nil {
+				errs[index] = fmt.Errorf("failed to write block receipts: %v", err)
+				atomic.AddInt32(&failed, 1)
+				glog.Fatal(errs[index])
+				return
+			}
+			if err := WriteMipmapBloom(self.chainDb, block.NumberU64(), receipts); err != nil {
+				errs[index] = fmt.Errorf("failed to write log blooms: %v", err)
+				atomic.AddInt32(&failed, 1)
+				glog.Fatal(errs[index])
+				return
+			}
+			atomic.AddInt32(&stats.processed, 1)
+		}
+	}
+	// Start as many worker threads as goroutines allowed
+	pending := new(sync.WaitGroup)
+	for i := 0; i < runtime.GOMAXPROCS(0); i++ {
+		pending.Add(1)
+		go func(id int) {
+			defer pending.Done()
+			process(id)
+		}(i)
+	}
+	pending.Wait()
+
+	// If anything failed, report
+	if failed > 0 {
+		for i, err := range errs {
+			if err != nil {
+				return i, err
+			}
+		}
+	}
+	if atomic.LoadInt32(&self.procInterrupt) == 1 {
+		glog.V(logger.Debug).Infoln("premature abort during receipt chain processing")
+		return 0, nil
+	}
+	// Update the head fast sync block if better
+	self.mu.Lock()
+	head := blockChain[len(errs)-1]
+	if self.GetTd(self.currentFastBlock.Hash()).Cmp(self.GetTd(head.Hash())) < 0 {
+		if err := WriteHeadFastBlockHash(self.chainDb, head.Hash()); err != nil {
+			glog.Fatalf("failed to update head fast block hash: %v", err)
+		}
+		self.currentFastBlock = head
+	}
+	self.mu.Unlock()
+
+	// Report some public statistics so the user has a clue what's going on
+	first, last := blockChain[0], blockChain[len(blockChain)-1]
+	glog.V(logger.Info).Infof("imported %d receipt(s) (%d ignored) in %v. #%d [%x… / %x…]", stats.processed, stats.ignored,
+		time.Since(start), last.Number(), first.Hash().Bytes()[:4], last.Hash().Bytes()[:4])
+
+	return 0, nil
+}
+
 // WriteBlock writes the block to the chain.
 func (self *BlockChain) WriteBlock(block *types.Block) (status writeStatus, err error) {
 	self.wg.Add(1)
@@ -516,38 +975,31 @@ func (self *BlockChain) WriteBlock(block *types.Block) (status writeStatus, err
 	}
 	td := new(big.Int).Add(block.Difficulty(), ptd)
 
-	self.mu.RLock()
-	cblock := self.currentBlock
-	self.mu.RUnlock()
-
-	// Compare the TD of the last known block in the canonical chain to make sure it's greater.
-	// At this point it's possible that a different chain (fork) becomes the new canonical chain.
-	if td.Cmp(self.Td()) > 0 {
-		// chain fork
-		if block.ParentHash() != cblock.Hash() {
-			// during split we merge two different chains and create the new canonical chain
-			err := self.reorg(cblock, block)
-			if err != nil {
+	// Make sure no inconsistent state is leaked during insertion
+	self.mu.Lock()
+	defer self.mu.Unlock()
+
+	// If the total difficulty is higher than our known, add it to the canonical chain
+	if td.Cmp(self.GetTd(self.currentBlock.Hash())) > 0 {
+		// Reorganize the chain if the parent is not the head block
+		if block.ParentHash() != self.currentBlock.Hash() {
+			if err := self.reorg(self.currentBlock, block); err != nil {
 				return NonStatTy, err
 			}
 		}
-		status = CanonStatTy
-
-		self.mu.Lock()
-		self.setTotalDifficulty(td)
+		// Insert the block as the new head of the chain
 		self.insert(block)
-		self.mu.Unlock()
+		status = CanonStatTy
 	} else {
 		status = SideStatTy
 	}
-
+	// Irrelevant of the canonical status, write the block itself to the database
 	if err := WriteTd(self.chainDb, block.Hash(), td); err != nil {
 		glog.Fatalf("failed to write block total difficulty: %v", err)
 	}
 	if err := WriteBlock(self.chainDb, block); err != nil {
 		glog.Fatalf("filed to write block contents: %v", err)
 	}
-	// Delete from future blocks
 	self.futureBlocks.Remove(block.Hash())
 
 	return
@@ -580,7 +1032,7 @@ func (self *BlockChain) InsertChain(chain types.Blocks) (int, error) {
 	txcount := 0
 	for i, block := range chain {
 		if atomic.LoadInt32(&self.procInterrupt) == 1 {
-			glog.V(logger.Debug).Infoln("Premature abort during chain processing")
+			glog.V(logger.Debug).Infoln("Premature abort during block chain processing")
 			break
 		}
 
@@ -636,7 +1088,7 @@ func (self *BlockChain) InsertChain(chain types.Blocks) (int, error) {
 
 			return i, err
 		}
-		if err := PutBlockReceipts(self.chainDb, block, receipts); err != nil {
+		if err := PutBlockReceipts(self.chainDb, block.Hash(), receipts); err != nil {
 			glog.V(logger.Warn).Infoln("error writing block receipts:", err)
 		}
 
@@ -691,9 +1143,6 @@ func (self *BlockChain) InsertChain(chain types.Blocks) (int, error) {
 // to be part of the new canonical chain and accumulates potential missing transactions and post an
 // event about them
 func (self *BlockChain) reorg(oldBlock, newBlock *types.Block) error {
-	self.mu.Lock()
-	defer self.mu.Unlock()
-
 	var (
 		newChain    types.Blocks
 		commonBlock *types.Block
@@ -788,8 +1237,7 @@ func (self *BlockChain) postChainEvents(events []interface{}) {
 		if event, ok := event.(ChainEvent); ok {
 			// We need some control over the mining operation. Acquiring locks and waiting for the miner to create new block takes too long
 			// and in most cases isn't even necessary.
-			if self.currentBlock.Hash() == event.Hash {
-				self.currentGasLimit = CalcGasLimit(event.Block)
+			if self.LastBlockHash() == event.Hash {
 				self.eventMux.Post(ChainHeadEvent{event.Block})
 			}
 		}

core/blockchain_test.go

@@ -64,44 +64,58 @@ func theBlockChain(db ethdb.Database, t *testing.T) *BlockChain {
 }
 
 // Test fork of length N starting from block i
-func testFork(t *testing.T, bman *BlockProcessor, i, N int, f func(td1, td2 *big.Int)) {
-	// switch databases to process the new chain
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	// copy old chain up to i into new db with deterministic canonical
-	bman2, err := newCanonical(i, db)
+func testFork(t *testing.T, processor *BlockProcessor, i, n int, full bool, comparator func(td1, td2 *big.Int)) {
+	// Copy old chain up to #i into a new db
+	db, processor2, err := newCanonical(i, full)
 	if err != nil {
 		t.Fatal("could not make new canonical in testFork", err)
 	}
-	// assert the bmans have the same block at i
-	bi1 := bman.bc.GetBlockByNumber(uint64(i)).Hash()
-	bi2 := bman2.bc.GetBlockByNumber(uint64(i)).Hash()
-	if bi1 != bi2 {
-		fmt.Printf("%+v\n%+v\n\n", bi1, bi2)
-		t.Fatal("chains do not have the same hash at height", i)
+	// Assert the chains have the same header/block at #i
+	var hash1, hash2 common.Hash
+	if full {
+		hash1 = processor.bc.GetBlockByNumber(uint64(i)).Hash()
+		hash2 = processor2.bc.GetBlockByNumber(uint64(i)).Hash()
+	} else {
+		hash1 = processor.bc.GetHeaderByNumber(uint64(i)).Hash()
+		hash2 = processor2.bc.GetHeaderByNumber(uint64(i)).Hash()
 	}
-	bman2.bc.SetProcessor(bman2)
-
-	// extend the fork
-	parent := bman2.bc.CurrentBlock()
-	chainB := makeChain(parent, N, db, forkSeed)
-	_, err = bman2.bc.InsertChain(chainB)
-	if err != nil {
-		t.Fatal("Insert chain error for fork:", err)
+	if hash1 != hash2 {
+		t.Errorf("chain content mismatch at %d: have hash %v, want hash %v", i, hash2, hash1)
 	}
-
-	tdpre := bman.bc.Td()
-	// Test the fork's blocks on the original chain
-	td, err := testChain(chainB, bman)
-	if err != nil {
-		t.Fatal("expected chainB not to give errors:", err)
+	// Extend the newly created chain
+	var (
+		blockChainB  []*types.Block
+		headerChainB []*types.Header
+	)
+	if full {
+		blockChainB = makeBlockChain(processor2.bc.CurrentBlock(), n, db, forkSeed)
+		if _, err := processor2.bc.InsertChain(blockChainB); err != nil {
+			t.Fatalf("failed to insert forking chain: %v", err)
+		}
+	} else {
+		headerChainB = makeHeaderChain(processor2.bc.CurrentHeader(), n, db, forkSeed)
+		if _, err := processor2.bc.InsertHeaderChain(headerChainB, 1); err != nil {
+			t.Fatalf("failed to insert forking chain: %v", err)
+		}
 	}
-	// Compare difficulties
-	f(tdpre, td)
+	// Sanity check that the forked chain can be imported into the original
+	var tdPre, tdPost *big.Int
 
-	// Loop over parents making sure reconstruction is done properly
+	if full {
+		tdPre = processor.bc.GetTd(processor.bc.CurrentBlock().Hash())
+		if err := testBlockChainImport(blockChainB, processor); err != nil {
+			t.Fatalf("failed to import forked block chain: %v", err)
+		}
+		tdPost = processor.bc.GetTd(blockChainB[len(blockChainB)-1].Hash())
+	} else {
+		tdPre = processor.bc.GetTd(processor.bc.CurrentHeader().Hash())
+		if err := testHeaderChainImport(headerChainB, processor); err != nil {
+			t.Fatalf("failed to import forked header chain: %v", err)
+		}
+		tdPost = processor.bc.GetTd(headerChainB[len(headerChainB)-1].Hash())
+	}
+	// Compare the total difficulties of the chains
+	comparator(tdPre, tdPost)
 }
 
 func printChain(bc *BlockChain) {
@@ -111,22 +125,41 @@ func printChain(bc *BlockChain) {
 	}
 }
 
-// process blocks against a chain
-func testChain(chainB types.Blocks, bman *BlockProcessor) (*big.Int, error) {
-	for _, block := range chainB {
-		_, _, err := bman.bc.processor.Process(block)
-		if err != nil {
+// testBlockChainImport tries to process a chain of blocks, writing them into
+// the database if successful.
+func testBlockChainImport(chain []*types.Block, processor *BlockProcessor) error {
+	for _, block := range chain {
+		// Try and process the block
+		if _, _, err := processor.Process(block); err != nil {
 			if IsKnownBlockErr(err) {
 				continue
 			}
-			return nil, err
+			return err
 		}
-		bman.bc.mu.Lock()
-		WriteTd(bman.bc.chainDb, block.Hash(), new(big.Int).Add(block.Difficulty(), bman.bc.GetTd(block.ParentHash())))
-		WriteBlock(bman.bc.chainDb, block)
-		bman.bc.mu.Unlock()
+		// Manually insert the block into the database, but don't reorganize (allows subsequent testing)
+		processor.bc.mu.Lock()
+		WriteTd(processor.chainDb, block.Hash(), new(big.Int).Add(block.Difficulty(), processor.bc.GetTd(block.ParentHash())))
+		WriteBlock(processor.chainDb, block)
+		processor.bc.mu.Unlock()
 	}
-	return bman.bc.GetTd(chainB[len(chainB)-1].Hash()), nil
+	return nil
+}
+
+// testHeaderChainImport tries to process a chain of header, writing them into
+// the database if successful.
+func testHeaderChainImport(chain []*types.Header, processor *BlockProcessor) error {
+	for _, header := range chain {
+		// Try and validate the header
+		if err := processor.ValidateHeader(header, false, false); err != nil {
+			return err
+		}
+		// Manually insert the header into the database, but don't reorganize (allows subsequent testing)
+		processor.bc.mu.Lock()
+		WriteTd(processor.chainDb, header.Hash(), new(big.Int).Add(header.Difficulty, processor.bc.GetTd(header.ParentHash)))
+		WriteHeader(processor.chainDb, header)
+		processor.bc.mu.Unlock()
+	}
+	return nil
 }
 
 func loadChain(fn string, t *testing.T) (types.Blocks, error) {
@@ -154,139 +187,147 @@ func insertChain(done chan bool, blockchain *BlockChain, chain types.Blocks, t *
 }
 
 func TestLastBlock(t *testing.T) {
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
+	db, _ := ethdb.NewMemDatabase()
+
 	bchain := theBlockChain(db, t)
-	block := makeChain(bchain.CurrentBlock(), 1, db, 0)[0]
+	block := makeBlockChain(bchain.CurrentBlock(), 1, db, 0)[0]
 	bchain.insert(block)
 	if block.Hash() != GetHeadBlockHash(db) {
 		t.Errorf("Write/Get HeadBlockHash failed")
 	}
 }
 
-func TestExtendCanonical(t *testing.T) {
-	CanonicalLength := 5
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	// make first chain starting from genesis
-	bman, err := newCanonical(CanonicalLength, db)
+// Tests that given a starting canonical chain of a given size, it can be extended
+// with various length chains.
+func TestExtendCanonicalHeaders(t *testing.T) { testExtendCanonical(t, false) }
+func TestExtendCanonicalBlocks(t *testing.T)  { testExtendCanonical(t, true) }
+
+func testExtendCanonical(t *testing.T, full bool) {
+	length := 5
+
+	// Make first chain starting from genesis
+	_, processor, err := newCanonical(length, full)
 	if err != nil {
-		t.Fatal("Could not make new canonical chain:", err)
+		t.Fatalf("failed to make new canonical chain: %v", err)
 	}
-	f := func(td1, td2 *big.Int) {
+	// Define the difficulty comparator
+	better := func(td1, td2 *big.Int) {
 		if td2.Cmp(td1) <= 0 {
-			t.Error("expected chainB to have higher difficulty. Got", td2, "expected more than", td1)
+			t.Errorf("total difficulty mismatch: have %v, expected more than %v", td2, td1)
 		}
 	}
-	// Start fork from current height (CanonicalLength)
-	testFork(t, bman, CanonicalLength, 1, f)
-	testFork(t, bman, CanonicalLength, 2, f)
-	testFork(t, bman, CanonicalLength, 5, f)
-	testFork(t, bman, CanonicalLength, 10, f)
+	// Start fork from current height
+	testFork(t, processor, length, 1, full, better)
+	testFork(t, processor, length, 2, full, better)
+	testFork(t, processor, length, 5, full, better)
+	testFork(t, processor, length, 10, full, better)
 }
 
-func TestShorterFork(t *testing.T) {
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	// make first chain starting from genesis
-	bman, err := newCanonical(10, db)
+// Tests that given a starting canonical chain of a given size, creating shorter
+// forks do not take canonical ownership.
+func TestShorterForkHeaders(t *testing.T) { testShorterFork(t, false) }
+func TestShorterForkBlocks(t *testing.T)  { testShorterFork(t, true) }
+
+func testShorterFork(t *testing.T, full bool) {
+	length := 10
+
+	// Make first chain starting from genesis
+	_, processor, err := newCanonical(length, full)
 	if err != nil {
-		t.Fatal("Could not make new canonical chain:", err)
+		t.Fatalf("failed to make new canonical chain: %v", err)
 	}
-	f := func(td1, td2 *big.Int) {
+	// Define the difficulty comparator
+	worse := func(td1, td2 *big.Int) {
 		if td2.Cmp(td1) >= 0 {
-			t.Error("expected chainB to have lower difficulty. Got", td2, "expected less than", td1)
+			t.Errorf("total difficulty mismatch: have %v, expected less than %v", td2, td1)
 		}
 	}
-	// Sum of numbers must be less than 10
-	// for this to be a shorter fork
-	testFork(t, bman, 0, 3, f)
-	testFork(t, bman, 0, 7, f)
-	testFork(t, bman, 1, 1, f)
-	testFork(t, bman, 1, 7, f)
-	testFork(t, bman, 5, 3, f)
-	testFork(t, bman, 5, 4, f)
+	// Sum of numbers must be less than `length` for this to be a shorter fork
+	testFork(t, processor, 0, 3, full, worse)
+	testFork(t, processor, 0, 7, full, worse)
+	testFork(t, processor, 1, 1, full, worse)
+	testFork(t, processor, 1, 7, full, worse)
+	testFork(t, processor, 5, 3, full, worse)
+	testFork(t, processor, 5, 4, full, worse)
 }
 
-func TestLongerFork(t *testing.T) {
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	// make first chain starting from genesis
-	bman, err := newCanonical(10, db)
+// Tests that given a starting canonical chain of a given size, creating longer
+// forks do take canonical ownership.
+func TestLongerForkHeaders(t *testing.T) { testLongerFork(t, false) }
+func TestLongerForkBlocks(t *testing.T)  { testLongerFork(t, true) }
+
+func testLongerFork(t *testing.T, full bool) {
+	length := 10
+
+	// Make first chain starting from genesis
+	_, processor, err := newCanonical(length, full)
 	if err != nil {
-		t.Fatal("Could not make new canonical chain:", err)
+		t.Fatalf("failed to make new canonical chain: %v", err)
 	}
-	f := func(td1, td2 *big.Int) {
+	// Define the difficulty comparator
+	better := func(td1, td2 *big.Int) {
 		if td2.Cmp(td1) <= 0 {
-			t.Error("expected chainB to have higher difficulty. Got", td2, "expected more than", td1)
+			t.Errorf("total difficulty mismatch: have %v, expected more than %v", td2, td1)
 		}
 	}
-	// Sum of numbers must be greater than 10
-	// for this to be a longer fork
-	testFork(t, bman, 0, 11, f)
-	testFork(t, bman, 0, 15, f)
-	testFork(t, bman, 1, 10, f)
-	testFork(t, bman, 1, 12, f)
-	testFork(t, bman, 5, 6, f)
-	testFork(t, bman, 5, 8, f)
+	// Sum of numbers must be greater than `length` for this to be a longer fork
+	testFork(t, processor, 0, 11, full, better)
+	testFork(t, processor, 0, 15, full, better)
+	testFork(t, processor, 1, 10, full, better)
+	testFork(t, processor, 1, 12, full, better)
+	testFork(t, processor, 5, 6, full, better)
+	testFork(t, processor, 5, 8, full, better)
 }
 
-func TestEqualFork(t *testing.T) {
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	bman, err := newCanonical(10, db)
+// Tests that given a starting canonical chain of a given size, creating equal
+// forks do take canonical ownership.
+func TestEqualForkHeaders(t *testing.T) { testEqualFork(t, false) }
+func TestEqualForkBlocks(t *testing.T)  { testEqualFork(t, true) }
+
+func testEqualFork(t *testing.T, full bool) {
+	length := 10
+
+	// Make first chain starting from genesis
+	_, processor, err := newCanonical(length, full)
 	if err != nil {
-		t.Fatal("Could not make new canonical chain:", err)
+		t.Fatalf("failed to make new canonical chain: %v", err)
 	}
-	f := func(td1, td2 *big.Int) {
+	// Define the difficulty comparator
+	equal := func(td1, td2 *big.Int) {
 		if td2.Cmp(td1) != 0 {
-			t.Error("expected chainB to have equal difficulty. Got", td2, "expected ", td1)
+			t.Errorf("total difficulty mismatch: have %v, want %v", td2, td1)
 		}
 	}
-	// Sum of numbers must be equal to 10
-	// for this to be an equal fork
-	testFork(t, bman, 0, 10, f)
-	testFork(t, bman, 1, 9, f)
-	testFork(t, bman, 2, 8, f)
-	testFork(t, bman, 5, 5, f)
-	testFork(t, bman, 6, 4, f)
-	testFork(t, bman, 9, 1, f)
+	// Sum of numbers must be equal to `length` for this to be an equal fork
+	testFork(t, processor, 0, 10, full, equal)
+	testFork(t, processor, 1, 9, full, equal)
+	testFork(t, processor, 2, 8, full, equal)
+	testFork(t, processor, 5, 5, full, equal)
+	testFork(t, processor, 6, 4, full, equal)
+	testFork(t, processor, 9, 1, full, equal)
 }
 
-func TestBrokenChain(t *testing.T) {
-	db, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	bman, err := newCanonical(10, db)
-	if err != nil {
-		t.Fatal("Could not make new canonical chain:", err)
-	}
-	db2, err := ethdb.NewMemDatabase()
-	if err != nil {
-		t.Fatal("Failed to create db:", err)
-	}
-	bman2, err := newCanonical(10, db2)
+// Tests that chains missing links do not get accepted by the processor.
+func TestBrokenHeaderChain(t *testing.T) { testBrokenChain(t, false) }
+func TestBrokenBlockChain(t *testing.T)  { testBrokenChain(t, true) }
+
+func testBrokenChain(t *testing.T, full bool) {
+	// Make chain starting from genesis
+	db, processor, err := newCanonical(10, full)
 	if err != nil {
-		t.Fatal("Could not make new canonical chain:", err)
+		t.Fatalf("failed to make new canonical chain: %v", err)
 	}
-	bman2.bc.SetProcessor(bman2)
-	parent := bman2.bc.CurrentBlock()
-	chainB := makeChain(parent, 5, db2, forkSeed)
-	chainB = chainB[1:]
-	_, err = testChain(chainB, bman)
-	if err == nil {
-		t.Error("expected broken chain to return error")
+	// Create a forked chain, and try to insert with a missing link
+	if full {
+		chain := makeBlockChain(processor.bc.CurrentBlock(), 5, db, forkSeed)[1:]
+		if err := testBlockChainImport(chain, processor); err == nil {
+			t.Errorf("broken block chain not reported")
+		}
+	} else {
+		chain := makeHeaderChain(processor.bc.CurrentHeader(), 5, db, forkSeed)[1:]
+		if err := testHeaderChainImport(chain, processor); err == nil {
+			t.Errorf("broken header chain not reported")
+		}
 	}
 }
 
@@ -374,15 +415,29 @@ func TestChainMultipleInsertions(t *testing.T) {
 
 type bproc struct{}
 
-func (bproc) Process(*types.Block) (vm.Logs, types.Receipts, error) { return nil, nil, nil }
+func (bproc) Process(*types.Block) (vm.Logs, types.Receipts, error)                   { return nil, nil, nil }
+func (bproc) ValidateHeader(*types.Header, bool, bool) error                          { return nil }
+func (bproc) ValidateHeaderWithParent(*types.Header, *types.Header, bool, bool) error { return nil }
 
-func makeChainWithDiff(genesis *types.Block, d []int, seed byte) []*types.Block {
+func makeHeaderChainWithDiff(genesis *types.Block, d []int, seed byte) []*types.Header {
+	blocks := makeBlockChainWithDiff(genesis, d, seed)
+	headers := make([]*types.Header, len(blocks))
+	for i, block := range blocks {
+		headers[i] = block.Header()
+	}
+	return headers
+}
+
+func makeBlockChainWithDiff(genesis *types.Block, d []int, seed byte) []*types.Block {
 	var chain []*types.Block
 	for i, difficulty := range d {
 		header := &types.Header{
-			Coinbase:   common.Address{seed},
-			Number:     big.NewInt(int64(i + 1)),
-			Difficulty: big.NewInt(int64(difficulty)),
+			Coinbase:    common.Address{seed},
+			Number:      big.NewInt(int64(i + 1)),
+			Difficulty:  big.NewInt(int64(difficulty)),
+			UncleHash:   types.EmptyUncleHash,
+			TxHash:      types.EmptyRootHash,
+			ReceiptHash: types.EmptyRootHash,
 		}
 		if i == 0 {
 			header.ParentHash = genesis.Hash()
@@ -397,7 +452,7 @@ func makeChainWithDiff(genesis *types.Block, d []int, seed byte) []*types.Block
 
 func chm(genesis *types.Block, db ethdb.Database) *BlockChain {
 	var eventMux event.TypeMux
-	bc := &BlockChain{chainDb: db, genesisBlock: genesis, eventMux: &eventMux, pow: FakePow{}}
+	bc := &BlockChain{chainDb: db, genesisBlock: genesis, eventMux: &eventMux, pow: FakePow{}, rand: rand.New(rand.NewSource(0))}
 	bc.headerCache, _ = lru.New(100)
 	bc.bodyCache, _ = lru.New(100)
 	bc.bodyRLPCache, _ = lru.New(100)
@@ -410,147 +465,381 @@ func chm(genesis *types.Block, db ethdb.Database) *BlockChain {
 	return bc
 }
 
-func TestReorgLongest(t *testing.T) {
-	db, _ := ethdb.NewMemDatabase()
+// Tests that reorganizing a long difficult chain after a short easy one
+// overwrites the canonical numbers and links in the database.
+func TestReorgLongHeaders(t *testing.T) { testReorgLong(t, false) }
+func TestReorgLongBlocks(t *testing.T)  { testReorgLong(t, true) }
 
-	genesis, err := WriteTestNetGenesisBlock(db, 0)
-	if err != nil {
-		t.Error(err)
-		t.FailNow()
-	}
-	bc := chm(genesis, db)
+func testReorgLong(t *testing.T, full bool) {
+	testReorg(t, []int{1, 2, 4}, []int{1, 2, 3, 4}, 10, full)
+}
 
-	chain1 := makeChainWithDiff(genesis, []int{1, 2, 4}, 10)
-	chain2 := makeChainWithDiff(genesis, []int{1, 2, 3, 4}, 11)
+// Tests that reorganizing a short difficult chain after a long easy one
+// overwrites the canonical numbers and links in the database.
+func TestReorgShortHeaders(t *testing.T) { testReorgShort(t, false) }
+func TestReorgShortBlocks(t *testing.T)  { testReorgShort(t, true) }
 
-	bc.InsertChain(chain1)
-	bc.InsertChain(chain2)
+func testReorgShort(t *testing.T, full bool) {
+	testReorg(t, []int{1, 2, 3, 4}, []int{1, 10}, 11, full)
+}
 
-	prev := bc.CurrentBlock()
-	for block := bc.GetBlockByNumber(bc.CurrentBlock().NumberU64() - 1); block.NumberU64() != 0; prev, block = block, bc.GetBlockByNumber(block.NumberU64()-1) {
-		if prev.ParentHash() != block.Hash() {
-			t.Errorf("parent hash mismatch %x - %x", prev.ParentHash(), block.Hash())
+func testReorg(t *testing.T, first, second []int, td int64, full bool) {
+	// Create a pristine block chain
+	db, _ := ethdb.NewMemDatabase()
+	genesis, _ := WriteTestNetGenesisBlock(db, 0)
+	bc := chm(genesis, db)
+
+	// Insert an easy and a difficult chain afterwards
+	if full {
+		bc.InsertChain(makeBlockChainWithDiff(genesis, first, 11))
+		bc.InsertChain(makeBlockChainWithDiff(genesis, second, 22))
+	} else {
+		bc.InsertHeaderChain(makeHeaderChainWithDiff(genesis, first, 11), 1)
+		bc.InsertHeaderChain(makeHeaderChainWithDiff(genesis, second, 22), 1)
+	}
+	// Check that the chain is valid number and link wise
+	if full {
+		prev := bc.CurrentBlock()
+		for block := bc.GetBlockByNumber(bc.CurrentBlock().NumberU64() - 1); block.NumberU64() != 0; prev, block = block, bc.GetBlockByNumber(block.NumberU64()-1) {
+			if prev.ParentHash() != block.Hash() {
+				t.Errorf("parent block hash mismatch: have %x, want %x", prev.ParentHash(), block.Hash())
+			}
+		}
+	} else {
+		prev := bc.CurrentHeader()
+		for header := bc.GetHeaderByNumber(bc.CurrentHeader().Number.Uint64() - 1); header.Number.Uint64() != 0; prev, header = header, bc.GetHeaderByNumber(header.Number.Uint64()-1) {
+			if prev.ParentHash != header.Hash() {
+				t.Errorf("parent header hash mismatch: have %x, want %x", prev.ParentHash, header.Hash())
+			}
+		}
+	}
+	// Make sure the chain total difficulty is the correct one
+	want := new(big.Int).Add(genesis.Difficulty(), big.NewInt(td))
+	if full {
+		if have := bc.GetTd(bc.CurrentBlock().Hash()); have.Cmp(want) != 0 {
+			t.Errorf("total difficulty mismatch: have %v, want %v", have, want)
+		}
+	} else {
+		if have := bc.GetTd(bc.CurrentHeader().Hash()); have.Cmp(want) != 0 {
+			t.Errorf("total difficulty mismatch: have %v, want %v", have, want)
 		}
 	}
 }
 
-func TestBadHashes(t *testing.T) {
+// Tests that the insertion functions detect banned hashes.
+func TestBadHeaderHashes(t *testing.T) { testBadHashes(t, false) }
+func TestBadBlockHashes(t *testing.T)  { testBadHashes(t, true) }
+
+func testBadHashes(t *testing.T, full bool) {
+	// Create a pristine block chain
 	db, _ := ethdb.NewMemDatabase()
-	genesis, err := WriteTestNetGenesisBlock(db, 0)
-	if err != nil {
-		t.Error(err)
-		t.FailNow()
-	}
+	genesis, _ := WriteTestNetGenesisBlock(db, 0)
 	bc := chm(genesis, db)
 
-	chain := makeChainWithDiff(genesis, []int{1, 2, 4}, 10)
-	BadHashes[chain[2].Header().Hash()] = true
-
-	_, err = bc.InsertChain(chain)
+	// Create a chain, ban a hash and try to import
+	var err error
+	if full {
+		blocks := makeBlockChainWithDiff(genesis, []int{1, 2, 4}, 10)
+		BadHashes[blocks[2].Header().Hash()] = true
+		_, err = bc.InsertChain(blocks)
+	} else {
+		headers := makeHeaderChainWithDiff(genesis, []int{1, 2, 4}, 10)
+		BadHashes[headers[2].Hash()] = true
+		_, err = bc.InsertHeaderChain(headers, 1)
+	}
 	if !IsBadHashError(err) {
 		t.Errorf("error mismatch: want: BadHashError, have: %v", err)
 	}
 }
 
-func TestReorgBadHashes(t *testing.T) {
+// Tests that bad hashes are detected on boot, and the chan rolled back to a
+// good state prior to the bad hash.
+func TestReorgBadHeaderHashes(t *testing.T) { testReorgBadHashes(t, false) }
+func TestReorgBadBlockHashes(t *testing.T)  { testReorgBadHashes(t, true) }
+
+func testReorgBadHashes(t *testing.T, full bool) {
+	// Create a pristine block chain
 	db, _ := ethdb.NewMemDatabase()
-	genesis, err := WriteTestNetGenesisBlock(db, 0)
-	if err != nil {
-		t.Error(err)
-		t.FailNow()
-	}
+	genesis, _ := WriteTestNetGenesisBlock(db, 0)
 	bc := chm(genesis, db)
 
-	chain := makeChainWithDiff(genesis, []int{1, 2, 3, 4}, 11)
-	bc.InsertChain(chain)
-
-	if chain[3].Header().Hash() != bc.LastBlockHash() {
-		t.Errorf("last block hash mismatch: want: %x, have: %x", chain[3].Header().Hash(), bc.LastBlockHash())
-	}
-
-	// NewChainManager should check BadHashes when loading it db
-	BadHashes[chain[3].Header().Hash()] = true
+	// Create a chain, import and ban aferwards
+	headers := makeHeaderChainWithDiff(genesis, []int{1, 2, 3, 4}, 10)
+	blocks := makeBlockChainWithDiff(genesis, []int{1, 2, 3, 4}, 10)
 
-	var eventMux event.TypeMux
-	ncm, err := NewBlockChain(db, FakePow{}, &eventMux)
-	if err != nil {
-		t.Errorf("NewChainManager err: %s", err)
-	}
-
-	// check it set head to (valid) parent of bad hash block
-	if chain[2].Header().Hash() != ncm.LastBlockHash() {
-		t.Errorf("last block hash mismatch: want: %x, have: %x", chain[2].Header().Hash(), ncm.LastBlockHash())
-	}
-
-	if chain[2].Header().GasLimit.Cmp(ncm.GasLimit()) != 0 {
-		t.Errorf("current block gasLimit mismatch: want: %x, have: %x", chain[2].Header().GasLimit, ncm.GasLimit())
+	if full {
+		if _, err := bc.InsertChain(blocks); err != nil {
+			t.Fatalf("failed to import blocks: %v", err)
+		}
+		if bc.CurrentBlock().Hash() != blocks[3].Hash() {
+			t.Errorf("last block hash mismatch: have: %x, want %x", bc.CurrentBlock().Hash(), blocks[3].Header().Hash())
+		}
+		BadHashes[blocks[3].Header().Hash()] = true
+		defer func() { delete(BadHashes, blocks[3].Header().Hash()) }()
+	} else {
+		if _, err := bc.InsertHeaderChain(headers, 1); err != nil {
+			t.Fatalf("failed to import headers: %v", err)
+		}
+		if bc.CurrentHeader().Hash() != headers[3].Hash() {
+			t.Errorf("last header hash mismatch: have: %x, want %x", bc.CurrentHeader().Hash(), headers[3].Hash())
+		}
+		BadHashes[headers[3].Hash()] = true
+		defer func() { delete(BadHashes, headers[3].Hash()) }()
 	}
-}
-
-func TestReorgShortest(t *testing.T) {
-	db, _ := ethdb.NewMemDatabase()
-	genesis, err := WriteTestNetGenesisBlock(db, 0)
+	// Create a new chain manager and check it rolled back the state
+	ncm, err := NewBlockChain(db, FakePow{}, new(event.TypeMux))
 	if err != nil {
-		t.Error(err)
-		t.FailNow()
+		t.Fatalf("failed to create new chain manager: %v", err)
 	}
-	bc := chm(genesis, db)
-
-	chain1 := makeChainWithDiff(genesis, []int{1, 2, 3, 4}, 10)
-	chain2 := makeChainWithDiff(genesis, []int{1, 10}, 11)
-
-	bc.InsertChain(chain1)
-	bc.InsertChain(chain2)
-
-	prev := bc.CurrentBlock()
-	for block := bc.GetBlockByNumber(bc.CurrentBlock().NumberU64() - 1); block.NumberU64() != 0; prev, block = block, bc.GetBlockByNumber(block.NumberU64()-1) {
-		if prev.ParentHash() != block.Hash() {
-			t.Errorf("parent hash mismatch %x - %x", prev.ParentHash(), block.Hash())
+	if full {
+		if ncm.CurrentBlock().Hash() != blocks[2].Header().Hash() {
+			t.Errorf("last block hash mismatch: have: %x, want %x", ncm.CurrentBlock().Hash(), blocks[2].Header().Hash())
+		}
+		if blocks[2].Header().GasLimit.Cmp(ncm.GasLimit()) != 0 {
+			t.Errorf("last  block gasLimit mismatch: have: %x, want %x", ncm.GasLimit(), blocks[2].Header().GasLimit)
+		}
+	} else {
+		if ncm.CurrentHeader().Hash() != headers[2].Hash() {
+			t.Errorf("last header hash mismatch: have: %x, want %x", ncm.CurrentHeader().Hash(), headers[2].Hash())
 		}
 	}
 }
 
-func TestInsertNonceError(t *testing.T) {
+// Tests chain insertions in the face of one entity containing an invalid nonce.
+func TestHeadersInsertNonceError(t *testing.T) { testInsertNonceError(t, false) }
+func TestBlocksInsertNonceError(t *testing.T)  { testInsertNonceError(t, true) }
+
+func testInsertNonceError(t *testing.T, full bool) {
 	for i := 1; i < 25 && !t.Failed(); i++ {
-		db, _ := ethdb.NewMemDatabase()
-		genesis, err := WriteTestNetGenesisBlock(db, 0)
+		// Create a pristine chain and database
+		db, processor, err := newCanonical(0, full)
 		if err != nil {
-			t.Error(err)
-			t.FailNow()
+			t.Fatalf("failed to create pristine chain: %v", err)
 		}
-		bc := chm(genesis, db)
-		bc.processor = NewBlockProcessor(db, bc.pow, bc, bc.eventMux)
-		blocks := makeChain(bc.currentBlock, i, db, 0)
+		bc := processor.bc
+
+		// Create and insert a chain with a failing nonce
+		var (
+			failAt   int
+			failRes  int
+			failNum  uint64
+			failHash common.Hash
+		)
+		if full {
+			blocks := makeBlockChain(processor.bc.CurrentBlock(), i, db, 0)
+
+			failAt = rand.Int() % len(blocks)
+			failNum = blocks[failAt].NumberU64()
+			failHash = blocks[failAt].Hash()
+
+			processor.bc.pow = failPow{failNum}
+			processor.Pow = failPow{failNum}
+
+			failRes, err = processor.bc.InsertChain(blocks)
+		} else {
+			headers := makeHeaderChain(processor.bc.CurrentHeader(), i, db, 0)
 
-		fail := rand.Int() % len(blocks)
-		failblock := blocks[fail]
-		bc.pow = failPow{failblock.NumberU64()}
-		n, err := bc.InsertChain(blocks)
+			failAt = rand.Int() % len(headers)
+			failNum = headers[failAt].Number.Uint64()
+			failHash = headers[failAt].Hash()
 
+			processor.bc.pow = failPow{failNum}
+			processor.Pow = failPow{failNum}
+
+			failRes, err = processor.bc.InsertHeaderChain(headers, 1)
+		}
 		// Check that the returned error indicates the nonce failure.
-		if n != fail {
-			t.Errorf("(i=%d) wrong failed block index: got %d, want %d", i, n, fail)
+		if failRes != failAt {
+			t.Errorf("test %d: failure index mismatch: have %d, want %d", i, failRes, failAt)
 		}
 		if !IsBlockNonceErr(err) {
-			t.Fatalf("(i=%d) got %q, want a nonce error", i, err)
+			t.Fatalf("test %d: error mismatch: have %v, want nonce error", i, err)
 		}
 		nerr := err.(*BlockNonceErr)
-		if nerr.Number.Cmp(failblock.Number()) != 0 {
-			t.Errorf("(i=%d) wrong block number in error, got %v, want %v", i, nerr.Number, failblock.Number())
+		if nerr.Number.Uint64() != failNum {
+			t.Errorf("test %d: number mismatch: have %v, want %v", i, nerr.Number, failNum)
 		}
-		if nerr.Hash != failblock.Hash() {
-			t.Errorf("(i=%d) wrong block hash in error, got %v, want %v", i, nerr.Hash, failblock.Hash())
+		if nerr.Hash != failHash {
+			t.Errorf("test %d: hash mismatch: have %x, want %x", i, nerr.Hash[:4], failHash[:4])
 		}
-
 		// Check that all no blocks after the failing block have been inserted.
-		for _, block := range blocks[fail:] {
-			if bc.HasBlock(block.Hash()) {
-				t.Errorf("(i=%d) invalid block %d present in chain", i, block.NumberU64())
+		for j := 0; j < i-failAt; j++ {
+			if full {
+				if block := bc.GetBlockByNumber(failNum + uint64(j)); block != nil {
+					t.Errorf("test %d: invalid block in chain: %v", i, block)
+				}
+			} else {
+				if header := bc.GetHeaderByNumber(failNum + uint64(j)); header != nil {
+					t.Errorf("test %d: invalid header in chain: %v", i, header)
+				}
 			}
 		}
 	}
 }
 
+// Tests that fast importing a block chain produces the same chain data as the
+// classical full block processing.
+func TestFastVsFullChains(t *testing.T) {
+	// Configure and generate a sample block chain
+	var (
+		gendb, _ = ethdb.NewMemDatabase()
+		key, _   = crypto.HexToECDSA("b71c71a67e1177ad4e901695e1b4b9ee17ae16c6668d313eac2f96dbcda3f291")
+		address  = crypto.PubkeyToAddress(key.PublicKey)
+		funds    = big.NewInt(1000000000)
+		genesis  = GenesisBlockForTesting(gendb, address, funds)
+	)
+	blocks, receipts := GenerateChain(genesis, gendb, 1024, func(i int, block *BlockGen) {
+		block.SetCoinbase(common.Address{0x00})
+
+		// If the block number is multiple of 3, send a few bonus transactions to the miner
+		if i%3 == 2 {
+			for j := 0; j < i%4+1; j++ {
+				tx, err := types.NewTransaction(block.TxNonce(address), common.Address{0x00}, big.NewInt(1000), params.TxGas, nil, nil).SignECDSA(key)
+				if err != nil {
+					panic(err)
+				}
+				block.AddTx(tx)
+			}
+		}
+		// If the block number is a multiple of 5, add a few bonus uncles to the block
+		if i%5 == 5 {
+			block.AddUncle(&types.Header{ParentHash: block.PrevBlock(i - 1).Hash(), Number: big.NewInt(int64(i - 1))})
+		}
+	})
+	// Import the chain as an archive node for the comparison baseline
+	archiveDb, _ := ethdb.NewMemDatabase()
+	WriteGenesisBlockForTesting(archiveDb, GenesisAccount{address, funds})
+
+	archive, _ := NewBlockChain(archiveDb, FakePow{}, new(event.TypeMux))
+	archive.SetProcessor(NewBlockProcessor(archiveDb, FakePow{}, archive, new(event.TypeMux)))
+
+	if n, err := archive.InsertChain(blocks); err != nil {
+		t.Fatalf("failed to process block %d: %v", n, err)
+	}
+	// Fast import the chain as a non-archive node to test
+	fastDb, _ := ethdb.NewMemDatabase()
+	WriteGenesisBlockForTesting(fastDb, GenesisAccount{address, funds})
+	fast, _ := NewBlockChain(fastDb, FakePow{}, new(event.TypeMux))
+	fast.SetProcessor(NewBlockProcessor(fastDb, FakePow{}, fast, new(event.TypeMux)))
+
+	headers := make([]*types.Header, len(blocks))
+	for i, block := range blocks {
+		headers[i] = block.Header()
+	}
+	if n, err := fast.InsertHeaderChain(headers, 1); err != nil {
+		t.Fatalf("failed to insert header %d: %v", n, err)
+	}
+	if n, err := fast.InsertReceiptChain(blocks, receipts); err != nil {
+		t.Fatalf("failed to insert receipt %d: %v", n, err)
+	}
+	// Iterate over all chain data components, and cross reference
+	for i := 0; i < len(blocks); i++ {
+		num, hash := blocks[i].NumberU64(), blocks[i].Hash()
+
+		if ftd, atd := fast.GetTd(hash), archive.GetTd(hash); ftd.Cmp(atd) != 0 {
+			t.Errorf("block #%d [%x]: td mismatch: have %v, want %v", num, hash, ftd, atd)
+		}
+		if fheader, aheader := fast.GetHeader(hash), archive.GetHeader(hash); fheader.Hash() != aheader.Hash() {
+			t.Errorf("block #%d [%x]: header mismatch: have %v, want %v", num, hash, fheader, aheader)
+		}
+		if fblock, ablock := fast.GetBlock(hash), archive.GetBlock(hash); fblock.Hash() != ablock.Hash() {
+			t.Errorf("block #%d [%x]: block mismatch: have %v, want %v", num, hash, fblock, ablock)
+		} else if types.DeriveSha(fblock.Transactions()) != types.DeriveSha(ablock.Transactions()) {
+			t.Errorf("block #%d [%x]: transactions mismatch: have %v, want %v", num, hash, fblock.Transactions(), ablock.Transactions())
+		} else if types.CalcUncleHash(fblock.Uncles()) != types.CalcUncleHash(ablock.Uncles()) {
+			t.Errorf("block #%d [%x]: uncles mismatch: have %v, want %v", num, hash, fblock.Uncles(), ablock.Uncles())
+		}
+		if freceipts, areceipts := GetBlockReceipts(fastDb, hash), GetBlockReceipts(archiveDb, hash); types.DeriveSha(freceipts) != types.DeriveSha(areceipts) {
+			t.Errorf("block #%d [%x]: receipts mismatch: have %v, want %v", num, hash, freceipts, areceipts)
+		}
+	}
+	// Check that the canonical chains are the same between the databases
+	for i := 0; i < len(blocks)+1; i++ {
+		if fhash, ahash := GetCanonicalHash(fastDb, uint64(i)), GetCanonicalHash(archiveDb, uint64(i)); fhash != ahash {
+			t.Errorf("block #%d: canonical hash mismatch: have %v, want %v", i, fhash, ahash)
+		}
+	}
+}
+
+// Tests that various import methods move the chain head pointers to the correct
+// positions.
+func TestLightVsFastVsFullChainHeads(t *testing.T) {
+	// Configure and generate a sample block chain
+	var (
+		gendb, _ = ethdb.NewMemDatabase()
+		key, _   = crypto.HexToECDSA("b71c71a67e1177ad4e901695e1b4b9ee17ae16c6668d313eac2f96dbcda3f291")
+		address  = crypto.PubkeyToAddress(key.PublicKey)
+		funds    = big.NewInt(1000000000)
+		genesis  = GenesisBlockForTesting(gendb, address, funds)
+	)
+	height := uint64(1024)
+	blocks, receipts := GenerateChain(genesis, gendb, int(height), nil)
+
+	// Configure a subchain to roll back
+	remove := []common.Hash{}
+	for _, block := range blocks[height/2:] {
+		remove = append(remove, block.Hash())
+	}
+	// Create a small assertion method to check the three heads
+	assert := func(t *testing.T, kind string, chain *BlockChain, header uint64, fast uint64, block uint64) {
+		if num := chain.CurrentBlock().NumberU64(); num != block {
+			t.Errorf("%s head block mismatch: have #%v, want #%v", kind, num, block)
+		}
+		if num := chain.CurrentFastBlock().NumberU64(); num != fast {
+			t.Errorf("%s head fast-block mismatch: have #%v, want #%v", kind, num, fast)
+		}
+		if num := chain.CurrentHeader().Number.Uint64(); num != header {
+			t.Errorf("%s head header mismatch: have #%v, want #%v", kind, num, header)
+		}
+	}
+	// Import the chain as an archive node and ensure all pointers are updated
+	archiveDb, _ := ethdb.NewMemDatabase()
+	WriteGenesisBlockForTesting(archiveDb, GenesisAccount{address, funds})
+
+	archive, _ := NewBlockChain(archiveDb, FakePow{}, new(event.TypeMux))
+	archive.SetProcessor(NewBlockProcessor(archiveDb, FakePow{}, archive, new(event.TypeMux)))
+
+	if n, err := archive.InsertChain(blocks); err != nil {
+		t.Fatalf("failed to process block %d: %v", n, err)
+	}
+	assert(t, "archive", archive, height, height, height)
+	archive.Rollback(remove)
+	assert(t, "archive", archive, height/2, height/2, height/2)
+
+	// Import the chain as a non-archive node and ensure all pointers are updated
+	fastDb, _ := ethdb.NewMemDatabase()
+	WriteGenesisBlockForTesting(fastDb, GenesisAccount{address, funds})
+	fast, _ := NewBlockChain(fastDb, FakePow{}, new(event.TypeMux))
+	fast.SetProcessor(NewBlockProcessor(fastDb, FakePow{}, fast, new(event.TypeMux)))
+
+	headers := make([]*types.Header, len(blocks))
+	for i, block := range blocks {
+		headers[i] = block.Header()
+	}
+	if n, err := fast.InsertHeaderChain(headers, 1); err != nil {
+		t.Fatalf("failed to insert header %d: %v", n, err)
+	}
+	if n, err := fast.InsertReceiptChain(blocks, receipts); err != nil {
+		t.Fatalf("failed to insert receipt %d: %v", n, err)
+	}
+	assert(t, "fast", fast, height, height, 0)
+	fast.Rollback(remove)
+	assert(t, "fast", fast, height/2, height/2, 0)
+
+	// Import the chain as a light node and ensure all pointers are updated
+	lightDb, _ := ethdb.NewMemDatabase()
+	WriteGenesisBlockForTesting(lightDb, GenesisAccount{address, funds})
+	light, _ := NewBlockChain(lightDb, FakePow{}, new(event.TypeMux))
+	light.SetProcessor(NewBlockProcessor(lightDb, FakePow{}, light, new(event.TypeMux)))
+
+	if n, err := light.InsertHeaderChain(headers, 1); err != nil {
+		t.Fatalf("failed to insert header %d: %v", n, err)
+	}
+	assert(t, "light", light, height, 0, 0)
+	light.Rollback(remove)
+	assert(t, "light", light, height/2, 0, 0)
+}
+
 // Tests that chain reorganizations handle transaction removals and reinsertions.
 func TestChainTxReorgs(t *testing.T) {
 	params.MinGasLimit = big.NewInt(125000)      // Minimum the gas limit may ever be.
@@ -587,7 +876,7 @@ func TestChainTxReorgs(t *testing.T) {
 	//  - futureAdd: transaction added after the reorg has already finished
 	var pastAdd, freshAdd, futureAdd *types.Transaction
 
-	chain := GenerateChain(genesis, db, 3, func(i int, gen *BlockGen) {
+	chain, _ := GenerateChain(genesis, db, 3, func(i int, gen *BlockGen) {
 		switch i {
 		case 0:
 			pastDrop, _ = types.NewTransaction(gen.TxNonce(addr2), addr2, big.NewInt(1000), params.TxGas, nil, nil).SignECDSA(key2)
@@ -613,7 +902,7 @@ func TestChainTxReorgs(t *testing.T) {
 	}
 
 	// overwrite the old chain
-	chain = GenerateChain(genesis, db, 5, func(i int, gen *BlockGen) {
+	chain, _ = GenerateChain(genesis, db, 5, func(i int, gen *BlockGen) {
 		switch i {
 		case 0:
 			pastAdd, _ = types.NewTransaction(gen.TxNonce(addr3), addr3, big.NewInt(1000), params.TxGas, nil, nil).SignECDSA(key3)

core/chain_makers.go

@@ -98,7 +98,7 @@ func (b *BlockGen) AddTx(tx *types.Transaction) {
 	b.header.GasUsed.Add(b.header.GasUsed, gas)
 	receipt := types.NewReceipt(root.Bytes(), b.header.GasUsed)
 	logs := b.statedb.GetLogs(tx.Hash())
-	receipt.SetLogs(logs)
+	receipt.Logs = logs
 	receipt.Bloom = types.CreateBloom(types.Receipts{receipt})
 	b.txs = append(b.txs, tx)
 	b.receipts = append(b.receipts, receipt)
@@ -163,13 +163,13 @@ func (b *BlockGen) OffsetTime(seconds int64) {
 // Blocks created by GenerateChain do not contain valid proof of work
 // values. Inserting them into BlockChain requires use of FakePow or
 // a similar non-validating proof of work implementation.
-func GenerateChain(parent *types.Block, db ethdb.Database, n int, gen func(int, *BlockGen)) []*types.Block {
+func GenerateChain(parent *types.Block, db ethdb.Database, n int, gen func(int, *BlockGen)) ([]*types.Block, []types.Receipts) {
 	statedb, err := state.New(parent.Root(), db)
 	if err != nil {
 		panic(err)
 	}
-	blocks := make(types.Blocks, n)
-	genblock := func(i int, h *types.Header) *types.Block {
+	blocks, receipts := make(types.Blocks, n), make([]types.Receipts, n)
+	genblock := func(i int, h *types.Header) (*types.Block, types.Receipts) {
 		b := &BlockGen{parent: parent, i: i, chain: blocks, header: h, statedb: statedb}
 		if gen != nil {
 			gen(i, b)
@@ -180,15 +180,16 @@ func GenerateChain(parent *types.Block, db ethdb.Database, n int, gen func(int,
 			panic(fmt.Sprintf("state write error: %v", err))
 		}
 		h.Root = root
-		return types.NewBlock(h, b.txs, b.uncles, b.receipts)
+		return types.NewBlock(h, b.txs, b.uncles, b.receipts), b.receipts
 	}
 	for i := 0; i < n; i++ {
 		header := makeHeader(parent, statedb)
-		block := genblock(i, header)
+		block, receipt := genblock(i, header)
 		blocks[i] = block
+		receipts[i] = receipt
 		parent = block
 	}
-	return blocks
+	return blocks, receipts
 }
 
 func makeHeader(parent *types.Block, state *state.StateDB) *types.Header {
@@ -210,26 +211,51 @@ func makeHeader(parent *types.Block, state *state.StateDB) *types.Header {
 	}
 }
 
-// newCanonical creates a new deterministic canonical chain by running
-// InsertChain on the result of makeChain.
-func newCanonical(n int, db ethdb.Database) (*BlockProcessor, error) {
+// newCanonical creates a chain database, and injects a deterministic canonical
+// chain. Depending on the full flag, if creates either a full block chain or a
+// header only chain.
+func newCanonical(n int, full bool) (ethdb.Database, *BlockProcessor, error) {
+	// Create te new chain database
+	db, _ := ethdb.NewMemDatabase()
 	evmux := &event.TypeMux{}
 
-	WriteTestNetGenesisBlock(db, 0)
-	chainman, _ := NewBlockChain(db, FakePow{}, evmux)
-	bman := NewBlockProcessor(db, FakePow{}, chainman, evmux)
-	bman.bc.SetProcessor(bman)
-	parent := bman.bc.CurrentBlock()
+	// Initialize a fresh chain with only a genesis block
+	genesis, _ := WriteTestNetGenesisBlock(db, 0)
+
+	blockchain, _ := NewBlockChain(db, FakePow{}, evmux)
+	processor := NewBlockProcessor(db, FakePow{}, blockchain, evmux)
+	processor.bc.SetProcessor(processor)
+
+	// Create and inject the requested chain
 	if n == 0 {
-		return bman, nil
+		return db, processor, nil
+	}
+	if full {
+		// Full block-chain requested
+		blocks := makeBlockChain(genesis, n, db, canonicalSeed)
+		_, err := blockchain.InsertChain(blocks)
+		return db, processor, err
 	}
-	lchain := makeChain(parent, n, db, canonicalSeed)
-	_, err := bman.bc.InsertChain(lchain)
-	return bman, err
+	// Header-only chain requested
+	headers := makeHeaderChain(genesis.Header(), n, db, canonicalSeed)
+	_, err := blockchain.InsertHeaderChain(headers, 1)
+	return db, processor, err
 }
 
-func makeChain(parent *types.Block, n int, db ethdb.Database, seed int) []*types.Block {
-	return GenerateChain(parent, db, n, func(i int, b *BlockGen) {
+// makeHeaderChain creates a deterministic chain of headers rooted at parent.
+func makeHeaderChain(parent *types.Header, n int, db ethdb.Database, seed int) []*types.Header {
+	blocks := makeBlockChain(types.NewBlockWithHeader(parent), n, db, seed)
+	headers := make([]*types.Header, len(blocks))
+	for i, block := range blocks {
+		headers[i] = block.Header()
+	}
+	return headers
+}
+
+// makeBlockChain creates a deterministic chain of blocks rooted at parent.
+func makeBlockChain(parent *types.Block, n int, db ethdb.Database, seed int) []*types.Block {
+	blocks, _ := GenerateChain(parent, db, n, func(i int, b *BlockGen) {
 		b.SetCoinbase(common.Address{0: byte(seed), 19: byte(i)})
 	})
+	return blocks
 }

core/chain_makers_test.go

@@ -47,7 +47,7 @@ func ExampleGenerateChain() {
 	// This call generates a chain of 5 blocks. The function runs for
 	// each block and adds different features to gen based on the
 	// block index.
-	chain := GenerateChain(genesis, db, 5, func(i int, gen *BlockGen) {
+	chain, _ := GenerateChain(genesis, db, 5, func(i int, gen *BlockGen) {
 		switch i {
 		case 0:
 			// In block 1, addr1 sends addr2 some ether.

core/chain_pow_test.go

@@ -60,7 +60,7 @@ func TestPowVerification(t *testing.T) {
 	var (
 		testdb, _ = ethdb.NewMemDatabase()
 		genesis   = GenesisBlockForTesting(testdb, common.Address{}, new(big.Int))
-		blocks    = GenerateChain(genesis, testdb, 8, nil)
+		blocks, _ = GenerateChain(genesis, testdb, 8, nil)
 	)
 	headers := make([]*types.Header, len(blocks))
 	for i, block := range blocks {
@@ -115,7 +115,7 @@ func testPowConcurrentVerification(t *testing.T, threads int) {
 	var (
 		testdb, _ = ethdb.NewMemDatabase()
 		genesis   = GenesisBlockForTesting(testdb, common.Address{}, new(big.Int))
-		blocks    = GenerateChain(genesis, testdb, 8, nil)
+		blocks, _ = GenerateChain(genesis, testdb, 8, nil)
 	)
 	headers := make([]*types.Header, len(blocks))
 	for i, block := range blocks {
@@ -186,7 +186,7 @@ func testPowConcurrentAbortion(t *testing.T, threads int) {
 	var (
 		testdb, _ = ethdb.NewMemDatabase()
 		genesis   = GenesisBlockForTesting(testdb, common.Address{}, new(big.Int))
-		blocks    = GenerateChain(genesis, testdb, 1024, nil)
+		blocks, _ = GenerateChain(genesis, testdb, 1024, nil)
 	)
 	headers := make([]*types.Header, len(blocks))
 	for i, block := range blocks {

core/chain_util.go

@@ -34,6 +34,7 @@ import (
 var (
 	headHeaderKey = []byte("LastHeader")
 	headBlockKey  = []byte("LastBlock")
+	headFastKey   = []byte("LastFast")
 
 	blockPrefix    = []byte("block-")
 	blockNumPrefix = []byte("block-num-")
@@ -129,7 +130,7 @@ func GetCanonicalHash(db ethdb.Database, number uint64) common.Hash {
 // header. The difference between this and GetHeadBlockHash is that whereas the
 // last block hash is only updated upon a full block import, the last header
 // hash is updated already at header import, allowing head tracking for the
-// fast synchronization mechanism.
+// light synchronization mechanism.
 func GetHeadHeaderHash(db ethdb.Database) common.Hash {
 	data, _ := db.Get(headHeaderKey)
 	if len(data) == 0 {
@@ -147,6 +148,18 @@ func GetHeadBlockHash(db ethdb.Database) common.Hash {
 	return common.BytesToHash(data)
 }
 
+// GetHeadFastBlockHash retrieves the hash of the current canonical head block during
+// fast synchronization. The difference between this and GetHeadBlockHash is that
+// whereas the last block hash is only updated upon a full block import, the last
+// fast hash is updated when importing pre-processed blocks.
+func GetHeadFastBlockHash(db ethdb.Database) common.Hash {
+	data, _ := db.Get(headFastKey)
+	if len(data) == 0 {
+		return common.Hash{}
+	}
+	return common.BytesToHash(data)
+}
+
 // GetHeaderRLP retrieves a block header in its raw RLP database encoding, or nil
 // if the header's not found.
 func GetHeaderRLP(db ethdb.Database, hash common.Hash) rlp.RawValue {
@@ -249,6 +262,15 @@ func WriteHeadBlockHash(db ethdb.Database, hash common.Hash) error {
 	return nil
 }
 
+// WriteHeadFastBlockHash stores the fast head block's hash.
+func WriteHeadFastBlockHash(db ethdb.Database, hash common.Hash) error {
+	if err := db.Put(headFastKey, hash.Bytes()); err != nil {
+		glog.Fatalf("failed to store last fast block's hash into database: %v", err)
+		return err
+	}
+	return nil
+}
+
 // WriteHeader serializes a block header into the database.
 func WriteHeader(db ethdb.Database, header *types.Header) error {
 	data, err := rlp.EncodeToBytes(header)
@@ -372,7 +394,7 @@ func WriteMipmapBloom(db ethdb.Database, number uint64, receipts types.Receipts)
 		bloomDat, _ := db.Get(key)
 		bloom := types.BytesToBloom(bloomDat)
 		for _, receipt := range receipts {
-			for _, log := range receipt.Logs() {
+			for _, log := range receipt.Logs {
 				bloom.Add(log.Address.Big())
 			}
 		}

core/chain_util_test.go

@@ -163,7 +163,12 @@ func TestBlockStorage(t *testing.T) {
 	db, _ := ethdb.NewMemDatabase()
 
 	// Create a test block to move around the database and make sure it's really new
-	block := types.NewBlockWithHeader(&types.Header{Extra: []byte("test block")})
+	block := types.NewBlockWithHeader(&types.Header{
+		Extra:       []byte("test block"),
+		UncleHash:   types.EmptyUncleHash,
+		TxHash:      types.EmptyRootHash,
+		ReceiptHash: types.EmptyRootHash,
+	})
 	if entry := GetBlock(db, block.Hash()); entry != nil {
 		t.Fatalf("Non existent block returned: %v", entry)
 	}
@@ -208,8 +213,12 @@ func TestBlockStorage(t *testing.T) {
 // Tests that partial block contents don't get reassembled into full blocks.
 func TestPartialBlockStorage(t *testing.T) {
 	db, _ := ethdb.NewMemDatabase()
-	block := types.NewBlockWithHeader(&types.Header{Extra: []byte("test block")})
-
+	block := types.NewBlockWithHeader(&types.Header{
+		Extra:       []byte("test block"),
+		UncleHash:   types.EmptyUncleHash,
+		TxHash:      types.EmptyRootHash,
+		ReceiptHash: types.EmptyRootHash,
+	})
 	// Store a header and check that it's not recognized as a block
 	if err := WriteHeader(db, block.Header()); err != nil {
 		t.Fatalf("Failed to write header into database: %v", err)
@@ -298,6 +307,7 @@ func TestHeadStorage(t *testing.T) {
 
 	blockHead := types.NewBlockWithHeader(&types.Header{Extra: []byte("test block header")})
 	blockFull := types.NewBlockWithHeader(&types.Header{Extra: []byte("test block full")})
+	blockFast := types.NewBlockWithHeader(&types.Header{Extra: []byte("test block fast")})
 
 	// Check that no head entries are in a pristine database
 	if entry := GetHeadHeaderHash(db); entry != (common.Hash{}) {
@@ -306,6 +316,9 @@ func TestHeadStorage(t *testing.T) {
 	if entry := GetHeadBlockHash(db); entry != (common.Hash{}) {
 		t.Fatalf("Non head block entry returned: %v", entry)
 	}
+	if entry := GetHeadFastBlockHash(db); entry != (common.Hash{}) {
+		t.Fatalf("Non fast head block entry returned: %v", entry)
+	}
 	// Assign separate entries for the head header and block
 	if err := WriteHeadHeaderHash(db, blockHead.Hash()); err != nil {
 		t.Fatalf("Failed to write head header hash: %v", err)
@@ -313,6 +326,9 @@ func TestHeadStorage(t *testing.T) {
 	if err := WriteHeadBlockHash(db, blockFull.Hash()); err != nil {
 		t.Fatalf("Failed to write head block hash: %v", err)
 	}
+	if err := WriteHeadFastBlockHash(db, blockFast.Hash()); err != nil {
+		t.Fatalf("Failed to write fast head block hash: %v", err)
+	}
 	// Check that both heads are present, and different (i.e. two heads maintained)
 	if entry := GetHeadHeaderHash(db); entry != blockHead.Hash() {
 		t.Fatalf("Head header hash mismatch: have %v, want %v", entry, blockHead.Hash())
@@ -320,21 +336,24 @@ func TestHeadStorage(t *testing.T) {
 	if entry := GetHeadBlockHash(db); entry != blockFull.Hash() {
 		t.Fatalf("Head block hash mismatch: have %v, want %v", entry, blockFull.Hash())
 	}
+	if entry := GetHeadFastBlockHash(db); entry != blockFast.Hash() {
+		t.Fatalf("Fast head block hash mismatch: have %v, want %v", entry, blockFast.Hash())
+	}
 }
 
 func TestMipmapBloom(t *testing.T) {
 	db, _ := ethdb.NewMemDatabase()
 
 	receipt1 := new(types.Receipt)
-	receipt1.SetLogs(vm.Logs{
+	receipt1.Logs = vm.Logs{
 		&vm.Log{Address: common.BytesToAddress([]byte("test"))},
 		&vm.Log{Address: common.BytesToAddress([]byte("address"))},
-	})
+	}
 	receipt2 := new(types.Receipt)
-	receipt2.SetLogs(vm.Logs{
+	receipt2.Logs = vm.Logs{
 		&vm.Log{Address: common.BytesToAddress([]byte("test"))},
 		&vm.Log{Address: common.BytesToAddress([]byte("address1"))},
-	})
+	}
 
 	WriteMipmapBloom(db, 1, types.Receipts{receipt1})
 	WriteMipmapBloom(db, 2, types.Receipts{receipt2})
@@ -349,15 +368,15 @@ func TestMipmapBloom(t *testing.T) {
 	// reset
 	db, _ = ethdb.NewMemDatabase()
 	receipt := new(types.Receipt)
-	receipt.SetLogs(vm.Logs{
+	receipt.Logs = vm.Logs{
 		&vm.Log{Address: common.BytesToAddress([]byte("test"))},
-	})
+	}
 	WriteMipmapBloom(db, 999, types.Receipts{receipt1})
 
 	receipt = new(types.Receipt)
-	receipt.SetLogs(vm.Logs{
+	receipt.Logs = vm.Logs{
 		&vm.Log{Address: common.BytesToAddress([]byte("test 1"))},
-	})
+	}
 	WriteMipmapBloom(db, 1000, types.Receipts{receipt})
 
 	bloom := GetMipmapBloom(db, 1000, 1000)
@@ -384,22 +403,22 @@ func TestMipmapChain(t *testing.T) {
 	defer db.Close()
 
 	genesis := WriteGenesisBlockForTesting(db, GenesisAccount{addr, big.NewInt(1000000)})
-	chain := GenerateChain(genesis, db, 1010, func(i int, gen *BlockGen) {
+	chain, receipts := GenerateChain(genesis, db, 1010, func(i int, gen *BlockGen) {
 		var receipts types.Receipts
 		switch i {
 		case 1:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{
+			receipt.Logs = vm.Logs{
 				&vm.Log{
 					Address: addr,
 					Topics:  []common.Hash{hash1},
 				},
-			})
+			}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		case 1000:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{&vm.Log{Address: addr2}})
+			receipt.Logs = vm.Logs{&vm.Log{Address: addr2}}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 
@@ -412,7 +431,7 @@ func TestMipmapChain(t *testing.T) {
 		}
 		WriteMipmapBloom(db, uint64(i+1), receipts)
 	})
-	for _, block := range chain {
+	for i, block := range chain {
 		WriteBlock(db, block)
 		if err := WriteCanonicalHash(db, block.Hash(), block.NumberU64()); err != nil {
 			t.Fatalf("failed to insert block number: %v", err)
@@ -420,7 +439,7 @@ func TestMipmapChain(t *testing.T) {
 		if err := WriteHeadBlockHash(db, block.Hash()); err != nil {
 			t.Fatalf("failed to insert block number: %v", err)
 		}
-		if err := PutBlockReceipts(db, block, block.Receipts()); err != nil {
+		if err := PutBlockReceipts(db, block.Hash(), receipts[i]); err != nil {
 			t.Fatal("error writing block receipts:", err)
 		}
 	}

core/error.go

@@ -111,7 +111,7 @@ type BlockNonceErr struct {
 }
 
 func (err *BlockNonceErr) Error() string {
-	return fmt.Sprintf("block %d (%v) nonce is invalid (got %d)", err.Number, err.Hash, err.Nonce)
+	return fmt.Sprintf("nonce for #%d [%x…] is invalid (got %d)", err.Number, err.Hash, err.Nonce)
 }
 
 // IsBlockNonceErr returns true for invalid block nonce errors.

core/genesis.go

@@ -103,7 +103,7 @@ func WriteGenesisBlock(chainDb ethdb.Database, reader io.Reader) (*types.Block,
 	if err := WriteBlock(chainDb, block); err != nil {
 		return nil, err
 	}
-	if err := PutBlockReceipts(chainDb, block, nil); err != nil {
+	if err := PutBlockReceipts(chainDb, block.Hash(), nil); err != nil {
 		return nil, err
 	}
 	if err := WriteCanonicalHash(chainDb, block.Hash(), block.NumberU64()); err != nil {

core/state/sync.go

@@ -0,0 +1,70 @@
+// Copyright 2015 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 state
+
+import (
+	"bytes"
+	"math/big"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/rlp"
+	"github.com/ethereum/go-ethereum/trie"
+)
+
+// StateSync is the main state synchronisation scheduler, which provides yet the
+// unknown state hashes to retrieve, accepts node data associated with said hashes
+// and reconstructs the state database step by step until all is done.
+type StateSync trie.TrieSync
+
+// NewStateSync create a new state trie download scheduler.
+func NewStateSync(root common.Hash, database ethdb.Database) *StateSync {
+	var syncer *trie.TrieSync
+
+	callback := func(leaf []byte, parent common.Hash) error {
+		var obj struct {
+			Nonce    uint64
+			Balance  *big.Int
+			Root     common.Hash
+			CodeHash []byte
+		}
+		if err := rlp.Decode(bytes.NewReader(leaf), &obj); err != nil {
+			return err
+		}
+		syncer.AddSubTrie(obj.Root, 64, parent, nil)
+		syncer.AddRawEntry(common.BytesToHash(obj.CodeHash), 64, parent)
+
+		return nil
+	}
+	syncer = trie.NewTrieSync(root, database, callback)
+	return (*StateSync)(syncer)
+}
+
+// Missing retrieves the known missing nodes from the state trie for retrieval.
+func (s *StateSync) Missing(max int) []common.Hash {
+	return (*trie.TrieSync)(s).Missing(max)
+}
+
+// Process injects a batch of retrieved trie nodes data.
+func (s *StateSync) Process(list []trie.SyncResult) (int, error) {
+	return (*trie.TrieSync)(s).Process(list)
+}
+
+// Pending returns the number of state entries currently pending for download.
+func (s *StateSync) Pending() int {
+	return (*trie.TrieSync)(s).Pending()
+}

core/state/sync_test.go

@@ -0,0 +1,238 @@
+// Copyright 2015 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 state
+
+import (
+	"bytes"
+	"math/big"
+	"testing"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/trie"
+)
+
+// testAccount is the data associated with an account used by the state tests.
+type testAccount struct {
+	address common.Address
+	balance *big.Int
+	nonce   uint64
+	code    []byte
+}
+
+// makeTestState create a sample test state to test node-wise reconstruction.
+func makeTestState() (ethdb.Database, common.Hash, []*testAccount) {
+	// Create an empty state
+	db, _ := ethdb.NewMemDatabase()
+	state, _ := New(common.Hash{}, db)
+
+	// Fill it with some arbitrary data
+	accounts := []*testAccount{}
+	for i := byte(0); i < 255; i++ {
+		obj := state.GetOrNewStateObject(common.BytesToAddress([]byte{i}))
+		acc := &testAccount{address: common.BytesToAddress([]byte{i})}
+
+		obj.AddBalance(big.NewInt(int64(11 * i)))
+		acc.balance = big.NewInt(int64(11 * i))
+
+		obj.SetNonce(uint64(42 * i))
+		acc.nonce = uint64(42 * i)
+
+		if i%3 == 0 {
+			obj.SetCode([]byte{i, i, i, i, i})
+			acc.code = []byte{i, i, i, i, i}
+		}
+		state.UpdateStateObject(obj)
+		accounts = append(accounts, acc)
+	}
+	root, _ := state.Commit()
+
+	// Return the generated state
+	return db, root, accounts
+}
+
+// checkStateAccounts cross references a reconstructed state with an expected
+// account array.
+func checkStateAccounts(t *testing.T, db ethdb.Database, root common.Hash, accounts []*testAccount) {
+	state, _ := New(root, db)
+	for i, acc := range accounts {
+
+		if balance := state.GetBalance(acc.address); balance.Cmp(acc.balance) != 0 {
+			t.Errorf("account %d: balance mismatch: have %v, want %v", i, balance, acc.balance)
+		}
+		if nonce := state.GetNonce(acc.address); nonce != acc.nonce {
+			t.Errorf("account %d: nonce mismatch: have %v, want %v", i, nonce, acc.nonce)
+		}
+		if code := state.GetCode(acc.address); bytes.Compare(code, acc.code) != 0 {
+			t.Errorf("account %d: code mismatch: have %x, want %x", i, code, acc.code)
+		}
+	}
+}
+
+// Tests that an empty state is not scheduled for syncing.
+func TestEmptyStateSync(t *testing.T) {
+	empty := common.HexToHash("56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421")
+	db, _ := ethdb.NewMemDatabase()
+	if req := NewStateSync(empty, db).Missing(1); len(req) != 0 {
+		t.Errorf("content requested for empty state: %v", req)
+	}
+}
+
+// Tests that given a root hash, a state can sync iteratively on a single thread,
+// requesting retrieval tasks and returning all of them in one go.
+func TestIterativeStateSyncIndividual(t *testing.T) { testIterativeStateSync(t, 1) }
+func TestIterativeStateSyncBatched(t *testing.T)    { testIterativeStateSync(t, 100) }
+
+func testIterativeStateSync(t *testing.T, batch int) {
+	// Create a random state to copy
+	srcDb, srcRoot, srcAccounts := makeTestState()
+
+	// Create a destination state and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewStateSync(srcRoot, dstDb)
+
+	queue := append([]common.Hash{}, sched.Missing(batch)...)
+	for len(queue) > 0 {
+		results := make([]trie.SyncResult, len(queue))
+		for i, hash := range queue {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results[i] = trie.SyncResult{hash, data}
+		}
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = append(queue[:0], sched.Missing(batch)...)
+	}
+	// Cross check that the two states are in sync
+	checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
+}
+
+// Tests that the trie scheduler can correctly reconstruct the state even if only
+// 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()
+
+	// Create a destination state and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewStateSync(srcRoot, dstDb)
+
+	queue := append([]common.Hash{}, sched.Missing(0)...)
+	for len(queue) > 0 {
+		// Sync only half of the scheduled nodes
+		results := make([]trie.SyncResult, len(queue)/2+1)
+		for i, hash := range queue[:len(results)] {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results[i] = trie.SyncResult{hash, data}
+		}
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = append(queue[len(results):], sched.Missing(0)...)
+	}
+	// Cross check that the two states are in sync
+	checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
+}
+
+// Tests that given a root hash, a trie can sync iteratively on a single thread,
+// requesting retrieval tasks and returning all of them in one go, however in a
+// random order.
+func TestIterativeRandomStateSyncIndividual(t *testing.T) { testIterativeRandomStateSync(t, 1) }
+func TestIterativeRandomStateSyncBatched(t *testing.T)    { testIterativeRandomStateSync(t, 100) }
+
+func testIterativeRandomStateSync(t *testing.T, batch int) {
+	// Create a random state to copy
+	srcDb, srcRoot, srcAccounts := makeTestState()
+
+	// Create a destination state and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewStateSync(srcRoot, dstDb)
+
+	queue := make(map[common.Hash]struct{})
+	for _, hash := range sched.Missing(batch) {
+		queue[hash] = struct{}{}
+	}
+	for len(queue) > 0 {
+		// Fetch all the queued nodes in a random order
+		results := make([]trie.SyncResult, 0, len(queue))
+		for hash, _ := range queue {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results = append(results, trie.SyncResult{hash, data})
+		}
+		// Feed the retrieved results back and queue new tasks
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = make(map[common.Hash]struct{})
+		for _, hash := range sched.Missing(batch) {
+			queue[hash] = struct{}{}
+		}
+	}
+	// Cross check that the two states are in sync
+	checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
+}
+
+// Tests that the trie scheduler can correctly reconstruct the state even if only
+// 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()
+
+	// Create a destination state and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewStateSync(srcRoot, dstDb)
+
+	queue := make(map[common.Hash]struct{})
+	for _, hash := range sched.Missing(0) {
+		queue[hash] = struct{}{}
+	}
+	for len(queue) > 0 {
+		// Sync only half of the scheduled nodes, even those in random order
+		results := make([]trie.SyncResult, 0, len(queue)/2+1)
+		for hash, _ := range queue {
+			delete(queue, hash)
+
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results = append(results, trie.SyncResult{hash, data})
+
+			if len(results) >= cap(results) {
+				break
+			}
+		}
+		// Feed the retrieved results back and queue new tasks
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		for _, hash := range sched.Missing(0) {
+			queue[hash] = struct{}{}
+		}
+	}
+	// Cross check that the two states are in sync
+	checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
+}

core/transaction_util.go

@@ -140,11 +140,14 @@ func GetBlockReceipts(db ethdb.Database, hash common.Hash) types.Receipts {
 	if len(data) == 0 {
 		return nil
 	}
-
-	var receipts types.Receipts
-	err := rlp.DecodeBytes(data, &receipts)
-	if err != nil {
-		glog.V(logger.Core).Infoln("GetReceiptse err", err)
+	rs := []*types.ReceiptForStorage{}
+	if err := rlp.DecodeBytes(data, &rs); err != nil {
+		glog.V(logger.Error).Infof("invalid receipt array RLP for hash %x: %v", hash, err)
+		return nil
+	}
+	receipts := make(types.Receipts, len(rs))
+	for i, receipt := range rs {
+		receipts[i] = (*types.Receipt)(receipt)
 	}
 	return receipts
 }
@@ -152,7 +155,7 @@ func GetBlockReceipts(db ethdb.Database, hash common.Hash) types.Receipts {
 // PutBlockReceipts stores the block's transactions associated receipts
 // and stores them by block hash in a single slice. This is required for
 // forks and chain reorgs
-func PutBlockReceipts(db ethdb.Database, block *types.Block, receipts types.Receipts) error {
+func PutBlockReceipts(db ethdb.Database, hash common.Hash, receipts types.Receipts) error {
 	rs := make([]*types.ReceiptForStorage, len(receipts))
 	for i, receipt := range receipts {
 		rs[i] = (*types.ReceiptForStorage)(receipt)
@@ -161,12 +164,9 @@ func PutBlockReceipts(db ethdb.Database, block *types.Block, receipts types.Rece
 	if err != nil {
 		return err
 	}
-
-	hash := block.Hash()
 	err = db.Put(append(blockReceiptsPre, hash[:]...), bytes)
 	if err != nil {
 		return err
 	}
-
 	return nil
 }

core/types/block.go

@@ -128,7 +128,6 @@ type Block struct {
 	header       *Header
 	uncles       []*Header
 	transactions Transactions
-	receipts     Receipts
 
 	// caches
 	hash atomic.Value
@@ -172,8 +171,8 @@ type storageblock struct {
 }
 
 var (
-	emptyRootHash  = DeriveSha(Transactions{})
-	emptyUncleHash = CalcUncleHash(nil)
+	EmptyRootHash  = DeriveSha(Transactions{})
+	EmptyUncleHash = CalcUncleHash(nil)
 )
 
 // NewBlock creates a new block. The input data is copied,
@@ -184,11 +183,11 @@ var (
 // are ignored and set to values derived from the given txs, uncles
 // and receipts.
 func NewBlock(header *Header, txs []*Transaction, uncles []*Header, receipts []*Receipt) *Block {
-	b := &Block{header: copyHeader(header), td: new(big.Int)}
+	b := &Block{header: CopyHeader(header), td: new(big.Int)}
 
 	// TODO: panic if len(txs) != len(receipts)
 	if len(txs) == 0 {
-		b.header.TxHash = emptyRootHash
+		b.header.TxHash = EmptyRootHash
 	} else {
 		b.header.TxHash = DeriveSha(Transactions(txs))
 		b.transactions = make(Transactions, len(txs))
@@ -196,21 +195,19 @@ func NewBlock(header *Header, txs []*Transaction, uncles []*Header, receipts []*
 	}
 
 	if len(receipts) == 0 {
-		b.header.ReceiptHash = emptyRootHash
+		b.header.ReceiptHash = EmptyRootHash
 	} else {
 		b.header.ReceiptHash = DeriveSha(Receipts(receipts))
 		b.header.Bloom = CreateBloom(receipts)
-		b.receipts = make([]*Receipt, len(receipts))
-		copy(b.receipts, receipts)
 	}
 
 	if len(uncles) == 0 {
-		b.header.UncleHash = emptyUncleHash
+		b.header.UncleHash = EmptyUncleHash
 	} else {
 		b.header.UncleHash = CalcUncleHash(uncles)
 		b.uncles = make([]*Header, len(uncles))
 		for i := range uncles {
-			b.uncles[i] = copyHeader(uncles[i])
+			b.uncles[i] = CopyHeader(uncles[i])
 		}
 	}
 
@@ -221,10 +218,12 @@ func NewBlock(header *Header, txs []*Transaction, uncles []*Header, receipts []*
 // header data is copied, changes to header and to the field values
 // will not affect the block.
 func NewBlockWithHeader(header *Header) *Block {
-	return &Block{header: copyHeader(header)}
+	return &Block{header: CopyHeader(header)}
 }
 
-func copyHeader(h *Header) *Header {
+// CopyHeader creates a deep copy of a block header to prevent side effects from
+// modifying a header variable.
+func CopyHeader(h *Header) *Header {
 	cpy := *h
 	if cpy.Time = new(big.Int); h.Time != nil {
 		cpy.Time.Set(h.Time)
@@ -297,7 +296,6 @@ func (b *StorageBlock) DecodeRLP(s *rlp.Stream) error {
 // TODO: copies
 func (b *Block) Uncles() []*Header          { return b.uncles }
 func (b *Block) Transactions() Transactions { return b.transactions }
-func (b *Block) Receipts() Receipts         { return b.receipts }
 
 func (b *Block) Transaction(hash common.Hash) *Transaction {
 	for _, transaction := range b.transactions {
@@ -326,7 +324,7 @@ func (b *Block) ReceiptHash() common.Hash { return b.header.ReceiptHash }
 func (b *Block) UncleHash() common.Hash   { return b.header.UncleHash }
 func (b *Block) Extra() []byte            { return common.CopyBytes(b.header.Extra) }
 
-func (b *Block) Header() *Header { return copyHeader(b.header) }
+func (b *Block) Header() *Header { return CopyHeader(b.header) }
 
 func (b *Block) HashNoNonce() common.Hash {
 	return b.header.HashNoNonce()
@@ -362,7 +360,6 @@ func (b *Block) WithMiningResult(nonce uint64, mixDigest common.Hash) *Block {
 	return &Block{
 		header:       &cpy,
 		transactions: b.transactions,
-		receipts:     b.receipts,
 		uncles:       b.uncles,
 	}
 }
@@ -370,13 +367,13 @@ func (b *Block) WithMiningResult(nonce uint64, mixDigest common.Hash) *Block {
 // WithBody returns a new block with the given transaction and uncle contents.
 func (b *Block) WithBody(transactions []*Transaction, uncles []*Header) *Block {
 	block := &Block{
-		header:       copyHeader(b.header),
+		header:       CopyHeader(b.header),
 		transactions: make([]*Transaction, len(transactions)),
 		uncles:       make([]*Header, len(uncles)),
 	}
 	copy(block.transactions, transactions)
 	for i := range uncles {
-		block.uncles[i] = copyHeader(uncles[i])
+		block.uncles[i] = CopyHeader(uncles[i])
 	}
 	return block
 }

core/types/bloom9.go

@@ -72,7 +72,7 @@ func (b Bloom) TestBytes(test []byte) bool {
 func CreateBloom(receipts Receipts) Bloom {
 	bin := new(big.Int)
 	for _, receipt := range receipts {
-		bin.Or(bin, LogsBloom(receipt.logs))
+		bin.Or(bin, LogsBloom(receipt.Logs))
 	}
 
 	return BytesToBloom(bin.Bytes())

core/types/common.go

@@ -20,4 +20,6 @@ import "github.com/ethereum/go-ethereum/core/vm"
 
 type BlockProcessor interface {
 	Process(*Block) (vm.Logs, Receipts, error)
+	ValidateHeader(*Header, bool, bool) error
+	ValidateHeaderWithParent(*Header, *Header, bool, bool) error
 }

core/types/receipt.go

@@ -17,7 +17,6 @@
 package types
 
 import (
-	"bytes"
 	"fmt"
 	"io"
 	"math/big"
@@ -27,89 +26,116 @@ import (
 	"github.com/ethereum/go-ethereum/rlp"
 )
 
+// Receipt represents the results of a transaction.
 type Receipt struct {
+	// Consensus fields
 	PostState         []byte
 	CumulativeGasUsed *big.Int
 	Bloom             Bloom
-	TxHash            common.Hash
-	ContractAddress   common.Address
-	logs              vm.Logs
-	GasUsed           *big.Int
-}
-
-func NewReceipt(root []byte, cumalativeGasUsed *big.Int) *Receipt {
-	return &Receipt{PostState: common.CopyBytes(root), CumulativeGasUsed: new(big.Int).Set(cumalativeGasUsed)}
-}
+	Logs              vm.Logs
 
-func (self *Receipt) SetLogs(logs vm.Logs) {
-	self.logs = logs
+	// Implementation fields
+	TxHash          common.Hash
+	ContractAddress common.Address
+	GasUsed         *big.Int
 }
 
-func (self *Receipt) Logs() vm.Logs {
-	return self.logs
+// NewReceipt creates a barebone transaction receipt, copying the init fields.
+func NewReceipt(root []byte, cumulativeGasUsed *big.Int) *Receipt {
+	return &Receipt{PostState: common.CopyBytes(root), CumulativeGasUsed: new(big.Int).Set(cumulativeGasUsed)}
 }
 
-func (self *Receipt) EncodeRLP(w io.Writer) error {
-	return rlp.Encode(w, []interface{}{self.PostState, self.CumulativeGasUsed, self.Bloom, self.logs})
+// EncodeRLP implements rlp.Encoder, and flattens the consensus fields of a receipt
+// into an RLP stream.
+func (r *Receipt) EncodeRLP(w io.Writer) error {
+	return rlp.Encode(w, []interface{}{r.PostState, r.CumulativeGasUsed, r.Bloom, r.Logs})
 }
 
-func (self *Receipt) DecodeRLP(s *rlp.Stream) error {
-	var r struct {
+// DecodeRLP implements rlp.Decoder, and loads the consensus fields of a receipt
+// from an RLP stream.
+func (r *Receipt) DecodeRLP(s *rlp.Stream) error {
+	var receipt struct {
 		PostState         []byte
 		CumulativeGasUsed *big.Int
 		Bloom             Bloom
-		TxHash            common.Hash
-		ContractAddress   common.Address
 		Logs              vm.Logs
-		GasUsed           *big.Int
 	}
-	if err := s.Decode(&r); err != nil {
+	if err := s.Decode(&receipt); err != nil {
 		return err
 	}
-	self.PostState, self.CumulativeGasUsed, self.Bloom, self.TxHash, self.ContractAddress, self.logs, self.GasUsed = r.PostState, r.CumulativeGasUsed, r.Bloom, r.TxHash, r.ContractAddress, r.Logs, r.GasUsed
-
+	r.PostState, r.CumulativeGasUsed, r.Bloom, r.Logs = receipt.PostState, receipt.CumulativeGasUsed, receipt.Bloom, receipt.Logs
 	return nil
 }
 
-type ReceiptForStorage Receipt
-
-func (self *ReceiptForStorage) EncodeRLP(w io.Writer) error {
-	storageLogs := make([]*vm.LogForStorage, len(self.logs))
-	for i, log := range self.logs {
-		storageLogs[i] = (*vm.LogForStorage)(log)
-	}
-	return rlp.Encode(w, []interface{}{self.PostState, self.CumulativeGasUsed, self.Bloom, self.TxHash, self.ContractAddress, storageLogs, self.GasUsed})
-}
-
-func (self *Receipt) RlpEncode() []byte {
-	bytes, err := rlp.EncodeToBytes(self)
+// RlpEncode implements common.RlpEncode required for SHA3 derivation.
+func (r *Receipt) RlpEncode() []byte {
+	bytes, err := rlp.EncodeToBytes(r)
 	if err != nil {
-		fmt.Println("TMP -- RECEIPT ENCODE ERROR", err)
+		panic(err)
 	}
 	return bytes
 }
 
-func (self *Receipt) Cmp(other *Receipt) bool {
-	if bytes.Compare(self.PostState, other.PostState) != 0 {
-		return false
-	}
+// String implements the Stringer interface.
+func (r *Receipt) String() string {
+	return fmt.Sprintf("receipt{med=%x cgas=%v bloom=%x logs=%v}", r.PostState, r.CumulativeGasUsed, r.Bloom, r.Logs)
+}
+
+// ReceiptForStorage is a wrapper around a Receipt that flattens and parses the
+// entire content of a receipt, as opposed to only the consensus fields originally.
+type ReceiptForStorage Receipt
 
-	return true
+// EncodeRLP implements rlp.Encoder, and flattens all content fields of a receipt
+// into an RLP stream.
+func (r *ReceiptForStorage) EncodeRLP(w io.Writer) error {
+	logs := make([]*vm.LogForStorage, len(r.Logs))
+	for i, log := range r.Logs {
+		logs[i] = (*vm.LogForStorage)(log)
+	}
+	return rlp.Encode(w, []interface{}{r.PostState, r.CumulativeGasUsed, r.Bloom, r.TxHash, r.ContractAddress, logs, r.GasUsed})
 }
 
-func (self *Receipt) String() string {
-	return fmt.Sprintf("receipt{med=%x cgas=%v bloom=%x logs=%v}", self.PostState, self.CumulativeGasUsed, self.Bloom, self.logs)
+// DecodeRLP implements rlp.Decoder, and loads both consensus and implementation
+// fields of a receipt from an RLP stream.
+func (r *ReceiptForStorage) DecodeRLP(s *rlp.Stream) error {
+	var receipt struct {
+		PostState         []byte
+		CumulativeGasUsed *big.Int
+		Bloom             Bloom
+		TxHash            common.Hash
+		ContractAddress   common.Address
+		Logs              []*vm.LogForStorage
+		GasUsed           *big.Int
+	}
+	if err := s.Decode(&receipt); err != nil {
+		return err
+	}
+	// Assign the consensus fields
+	r.PostState, r.CumulativeGasUsed, r.Bloom = receipt.PostState, receipt.CumulativeGasUsed, receipt.Bloom
+	r.Logs = make(vm.Logs, len(receipt.Logs))
+	for i, log := range receipt.Logs {
+		r.Logs[i] = (*vm.Log)(log)
+	}
+	// Assign the implementation fields
+	r.TxHash, r.ContractAddress, r.GasUsed = receipt.TxHash, receipt.ContractAddress, receipt.GasUsed
+
+	return nil
 }
 
+// Receipts is a wrapper around a Receipt array to implement types.DerivableList.
 type Receipts []*Receipt
 
-func (self Receipts) RlpEncode() []byte {
-	bytes, err := rlp.EncodeToBytes(self)
+// RlpEncode implements common.RlpEncode required for SHA3 derivation.
+func (r Receipts) RlpEncode() []byte {
+	bytes, err := rlp.EncodeToBytes(r)
 	if err != nil {
-		fmt.Println("TMP -- RECEIPTS ENCODE ERROR", err)
+		panic(err)
 	}
 	return bytes
 }
 
-func (self Receipts) Len() int            { return len(self) }
-func (self Receipts) GetRlp(i int) []byte { return common.Rlp(self[i]) }
+// Len returns the number of receipts in this list.
+func (r Receipts) Len() int { return len(r) }
+
+// GetRlp returns the RLP encoding of one receipt from the list.
+func (r Receipts) GetRlp(i int) []byte { return common.Rlp(r[i]) }

core/vm/log.go

@@ -25,42 +25,47 @@ import (
 )
 
 type Log struct {
+	// Consensus fields
 	Address common.Address
 	Topics  []common.Hash
 	Data    []byte
-	Number  uint64
 
-	TxHash    common.Hash
-	TxIndex   uint
-	BlockHash common.Hash
-	Index     uint
+	// Derived fields (don't reorder!)
+	BlockNumber uint64
+	TxHash      common.Hash
+	TxIndex     uint
+	BlockHash   common.Hash
+	Index       uint
 }
 
 func NewLog(address common.Address, topics []common.Hash, data []byte, number uint64) *Log {
-	return &Log{Address: address, Topics: topics, Data: data, Number: number}
+	return &Log{Address: address, Topics: topics, Data: data, BlockNumber: number}
 }
 
-func (self *Log) EncodeRLP(w io.Writer) error {
-	return rlp.Encode(w, []interface{}{self.Address, self.Topics, self.Data})
+func (l *Log) EncodeRLP(w io.Writer) error {
+	return rlp.Encode(w, []interface{}{l.Address, l.Topics, l.Data})
 }
 
-func (self *Log) String() string {
-	return fmt.Sprintf(`log: %x %x %x %x %d %x %d`, self.Address, self.Topics, self.Data, self.TxHash, self.TxIndex, self.BlockHash, self.Index)
+func (l *Log) DecodeRLP(s *rlp.Stream) error {
+	var log struct {
+		Address common.Address
+		Topics  []common.Hash
+		Data    []byte
+	}
+	if err := s.Decode(&log); err != nil {
+		return err
+	}
+	l.Address, l.Topics, l.Data = log.Address, log.Topics, log.Data
+	return nil
+}
+
+func (l *Log) String() string {
+	return fmt.Sprintf(`log: %x %x %x %x %d %x %d`, l.Address, l.Topics, l.Data, l.TxHash, l.TxIndex, l.BlockHash, l.Index)
 }
 
 type Logs []*Log
 
+// LogForStorage is a wrapper around a Log that flattens and parses the entire
+// content of a log, as opposed to only the consensus fields originally (by hiding
+// the rlp interface methods).
 type LogForStorage Log
-
-func (self *LogForStorage) EncodeRLP(w io.Writer) error {
-	return rlp.Encode(w, []interface{}{
-		self.Address,
-		self.Topics,
-		self.Data,
-		self.Number,
-		self.TxHash,
-		self.TxIndex,
-		self.BlockHash,
-		self.Index,
-	})
-}

eth/backend.go

@@ -88,6 +88,7 @@ type Config struct {
 	GenesisNonce int
 	GenesisFile  string
 	GenesisBlock *types.Block // used by block tests
+	FastSync     bool
 	Olympic      bool
 
 	BlockChainVersion  int
@@ -390,7 +391,6 @@ func New(config *Config) (*Ethereum, error) {
 		if err == core.ErrNoGenesis {
 			return nil, fmt.Errorf(`Genesis block not found. Please supply a genesis block with the "--genesis /path/to/file" argument`)
 		}
-
 		return nil, err
 	}
 	newPool := core.NewTxPool(eth.EventMux(), eth.blockchain.State, eth.blockchain.GasLimit)
@@ -398,8 +398,9 @@ func New(config *Config) (*Ethereum, error) {
 
 	eth.blockProcessor = core.NewBlockProcessor(chainDb, eth.pow, eth.blockchain, eth.EventMux())
 	eth.blockchain.SetProcessor(eth.blockProcessor)
-	eth.protocolManager = NewProtocolManager(config.NetworkId, eth.eventMux, eth.txPool, eth.pow, eth.blockchain, chainDb)
-
+	if eth.protocolManager, err = NewProtocolManager(config.FastSync, config.NetworkId, eth.eventMux, eth.txPool, eth.pow, eth.blockchain, chainDb); err != nil {
+		return nil, err
+	}
 	eth.miner = miner.New(eth, eth.EventMux(), eth.pow)
 	eth.miner.SetGasPrice(config.GasPrice)
 	eth.miner.SetExtra(config.ExtraData)
@@ -462,7 +463,7 @@ func (s *Ethereum) NodeInfo() *NodeInfo {
 		DiscPort:   int(node.UDP),
 		TCPPort:    int(node.TCP),
 		ListenAddr: s.net.ListenAddr,
-		Td:         s.BlockChain().Td().String(),
+		Td:         s.BlockChain().GetTd(s.BlockChain().CurrentBlock().Hash()).String(),
 	}
 }
 

eth/backend_test.go

@@ -16,17 +16,17 @@ func TestMipmapUpgrade(t *testing.T) {
 	addr := common.BytesToAddress([]byte("jeff"))
 	genesis := core.WriteGenesisBlockForTesting(db)
 
-	chain := core.GenerateChain(genesis, db, 10, func(i int, gen *core.BlockGen) {
+	chain, receipts := core.GenerateChain(genesis, db, 10, func(i int, gen *core.BlockGen) {
 		var receipts types.Receipts
 		switch i {
 		case 1:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{&vm.Log{Address: addr}})
+			receipt.Logs = vm.Logs{&vm.Log{Address: addr}}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		case 2:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{&vm.Log{Address: addr}})
+			receipt.Logs = vm.Logs{&vm.Log{Address: addr}}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		}
@@ -37,7 +37,7 @@ func TestMipmapUpgrade(t *testing.T) {
 			t.Fatal(err)
 		}
 	})
-	for _, block := range chain {
+	for i, block := range chain {
 		core.WriteBlock(db, block)
 		if err := core.WriteCanonicalHash(db, block.Hash(), block.NumberU64()); err != nil {
 			t.Fatalf("failed to insert block number: %v", err)
@@ -45,7 +45,7 @@ func TestMipmapUpgrade(t *testing.T) {
 		if err := core.WriteHeadBlockHash(db, block.Hash()); err != nil {
 			t.Fatalf("failed to insert block number: %v", err)
 		}
-		if err := core.PutBlockReceipts(db, block, block.Receipts()); err != nil {
+		if err := core.PutBlockReceipts(db, block.Hash(), receipts[i]); err != nil {
 			t.Fatal("error writing block receipts:", err)
 		}
 	}

eth/downloader/metrics.go

@@ -42,4 +42,14 @@ var (
 	bodyReqTimer     = metrics.NewTimer("eth/downloader/bodies/req")
 	bodyDropMeter    = metrics.NewMeter("eth/downloader/bodies/drop")
 	bodyTimeoutMeter = metrics.NewMeter("eth/downloader/bodies/timeout")
+
+	receiptInMeter      = metrics.NewMeter("eth/downloader/receipts/in")
+	receiptReqTimer     = metrics.NewTimer("eth/downloader/receipts/req")
+	receiptDropMeter    = metrics.NewMeter("eth/downloader/receipts/drop")
+	receiptTimeoutMeter = metrics.NewMeter("eth/downloader/receipts/timeout")
+
+	stateInMeter      = metrics.NewMeter("eth/downloader/states/in")
+	stateReqTimer     = metrics.NewTimer("eth/downloader/states/req")
+	stateDropMeter    = metrics.NewMeter("eth/downloader/states/drop")
+	stateTimeoutMeter = metrics.NewMeter("eth/downloader/states/timeout")
 )

eth/downloader/modes.go

@@ -0,0 +1,26 @@
+// Copyright 2015 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 downloader
+
+// SyncMode represents the synchronisation mode of the downloader.
+type SyncMode int
+
+const (
+	FullSync  SyncMode = iota // Synchronise the entire blockchain history from full blocks
+	FastSync                  // Quickly download the headers, full sync only at the chain head
+	LightSync                 // Download only the headers and terminate afterwards
+)

eth/downloader/peer.go

@@ -36,10 +36,12 @@ type relativeHashFetcherFn func(common.Hash) error
 type absoluteHashFetcherFn func(uint64, int) error
 type blockFetcherFn func([]common.Hash) error
 
-// Block header and body fethers belonging to eth/62 and above
+// Block header and body fetchers belonging to eth/62 and above
 type relativeHeaderFetcherFn func(common.Hash, int, int, bool) error
 type absoluteHeaderFetcherFn func(uint64, int, int, bool) error
 type blockBodyFetcherFn func([]common.Hash) error
+type receiptFetcherFn func([]common.Hash) error
+type stateFetcherFn func([]common.Hash) error
 
 var (
 	errAlreadyFetching   = errors.New("already fetching blocks from peer")
@@ -52,11 +54,18 @@ type peer struct {
 	id   string      // Unique identifier of the peer
 	head common.Hash // Hash of the peers latest known block
 
-	idle int32 // Current activity state of the peer (idle = 0, active = 1)
-	rep  int32 // Simple peer reputation
+	blockIdle   int32 // Current block activity state of the peer (idle = 0, active = 1)
+	receiptIdle int32 // Current receipt activity state of the peer (idle = 0, active = 1)
+	stateIdle   int32 // Current node data activity state of the peer (idle = 0, active = 1)
+	rep         int32 // Simple peer reputation
 
-	capacity int32     // Number of blocks allowed to fetch per request
-	started  time.Time // Time instance when the last fetch was started
+	blockCapacity   int32 // Number of blocks (bodies) allowed to fetch per request
+	receiptCapacity int32 // Number of receipts allowed to fetch per request
+	stateCapacity   int32 // Number of node data pieces allowed to fetch per request
+
+	blockStarted   time.Time // Time instance when the last block (body)fetch was started
+	receiptStarted time.Time // Time instance when the last receipt fetch was started
+	stateStarted   time.Time // Time instance when the last node data fetch was started
 
 	ignored *set.Set // Set of hashes not to request (didn't have previously)
 
@@ -68,6 +77,9 @@ type peer struct {
 	getAbsHeaders  absoluteHeaderFetcherFn // [eth/62] Method to retrieve a batch of headers from an absolute position
 	getBlockBodies blockBodyFetcherFn      // [eth/62] Method to retrieve a batch of block bodies
 
+	getReceipts receiptFetcherFn // [eth/63] Method to retrieve a batch of block transaction receipts
+	getNodeData stateFetcherFn   // [eth/63] Method to retrieve a batch of state trie data
+
 	version int // Eth protocol version number to switch strategies
 }
 
@@ -75,12 +87,15 @@ type peer struct {
 // mechanisms.
 func newPeer(id string, version int, head common.Hash,
 	getRelHashes relativeHashFetcherFn, getAbsHashes absoluteHashFetcherFn, getBlocks blockFetcherFn, // eth/61 callbacks, remove when upgrading
-	getRelHeaders relativeHeaderFetcherFn, getAbsHeaders absoluteHeaderFetcherFn, getBlockBodies blockBodyFetcherFn) *peer {
+	getRelHeaders relativeHeaderFetcherFn, getAbsHeaders absoluteHeaderFetcherFn, getBlockBodies blockBodyFetcherFn,
+	getReceipts receiptFetcherFn, getNodeData stateFetcherFn) *peer {
 	return &peer{
-		id:       id,
-		head:     head,
-		capacity: 1,
-		ignored:  set.New(),
+		id:              id,
+		head:            head,
+		blockCapacity:   1,
+		receiptCapacity: 1,
+		stateCapacity:   1,
+		ignored:         set.New(),
 
 		getRelHashes: getRelHashes,
 		getAbsHashes: getAbsHashes,
@@ -90,24 +105,34 @@ func newPeer(id string, version int, head common.Hash,
 		getAbsHeaders:  getAbsHeaders,
 		getBlockBodies: getBlockBodies,
 
+		getReceipts: getReceipts,
+		getNodeData: getNodeData,
+
 		version: version,
 	}
 }
 
 // Reset clears the internal state of a peer entity.
 func (p *peer) Reset() {
-	atomic.StoreInt32(&p.idle, 0)
-	atomic.StoreInt32(&p.capacity, 1)
+	atomic.StoreInt32(&p.blockIdle, 0)
+	atomic.StoreInt32(&p.receiptIdle, 0)
+	atomic.StoreInt32(&p.blockCapacity, 1)
+	atomic.StoreInt32(&p.receiptCapacity, 1)
+	atomic.StoreInt32(&p.stateCapacity, 1)
 	p.ignored.Clear()
 }
 
 // Fetch61 sends a block retrieval request to the remote peer.
 func (p *peer) Fetch61(request *fetchRequest) error {
+	// Sanity check the protocol version
+	if p.version != 61 {
+		panic(fmt.Sprintf("block fetch [eth/61] requested on eth/%d", p.version))
+	}
 	// Short circuit if the peer is already fetching
-	if !atomic.CompareAndSwapInt32(&p.idle, 0, 1) {
+	if !atomic.CompareAndSwapInt32(&p.blockIdle, 0, 1) {
 		return errAlreadyFetching
 	}
-	p.started = time.Now()
+	p.blockStarted = time.Now()
 
 	// Convert the hash set to a retrievable slice
 	hashes := make([]common.Hash, 0, len(request.Hashes))
@@ -119,13 +144,17 @@ func (p *peer) Fetch61(request *fetchRequest) error {
 	return nil
 }
 
-// Fetch sends a block body retrieval request to the remote peer.
-func (p *peer) Fetch(request *fetchRequest) error {
+// FetchBodies sends a block body retrieval request to the remote peer.
+func (p *peer) FetchBodies(request *fetchRequest) error {
+	// Sanity check the protocol version
+	if p.version < 62 {
+		panic(fmt.Sprintf("body fetch [eth/62+] requested on eth/%d", p.version))
+	}
 	// Short circuit if the peer is already fetching
-	if !atomic.CompareAndSwapInt32(&p.idle, 0, 1) {
+	if !atomic.CompareAndSwapInt32(&p.blockIdle, 0, 1) {
 		return errAlreadyFetching
 	}
-	p.started = time.Now()
+	p.blockStarted = time.Now()
 
 	// Convert the header set to a retrievable slice
 	hashes := make([]common.Hash, 0, len(request.Headers))
@@ -137,55 +166,97 @@ func (p *peer) Fetch(request *fetchRequest) error {
 	return nil
 }
 
-// SetIdle61 sets the peer to idle, allowing it to execute new retrieval requests.
-// Its block retrieval allowance will also be updated either up- or downwards,
-// depending on whether the previous fetch completed in time or not.
-func (p *peer) SetIdle61() {
-	// Update the peer's download allowance based on previous performance
-	scale := 2.0
-	if time.Since(p.started) > blockSoftTTL {
-		scale = 0.5
-		if time.Since(p.started) > blockHardTTL {
-			scale = 1 / float64(MaxBlockFetch) // reduces capacity to 1
-		}
+// FetchReceipts sends a receipt retrieval request to the remote peer.
+func (p *peer) FetchReceipts(request *fetchRequest) error {
+	// Sanity check the protocol version
+	if p.version < 63 {
+		panic(fmt.Sprintf("body fetch [eth/63+] requested on eth/%d", p.version))
 	}
-	for {
-		// Calculate the new download bandwidth allowance
-		prev := atomic.LoadInt32(&p.capacity)
-		next := int32(math.Max(1, math.Min(float64(MaxBlockFetch), float64(prev)*scale)))
+	// Short circuit if the peer is already fetching
+	if !atomic.CompareAndSwapInt32(&p.receiptIdle, 0, 1) {
+		return errAlreadyFetching
+	}
+	p.receiptStarted = time.Now()
 
-		// Try to update the old value
-		if atomic.CompareAndSwapInt32(&p.capacity, prev, next) {
-			// If we're having problems at 1 capacity, try to find better peers
-			if next == 1 {
-				p.Demote()
-			}
-			break
-		}
+	// Convert the header set to a retrievable slice
+	hashes := make([]common.Hash, 0, len(request.Headers))
+	for _, header := range request.Headers {
+		hashes = append(hashes, header.Hash())
 	}
-	// Set the peer to idle to allow further block requests
-	atomic.StoreInt32(&p.idle, 0)
+	go p.getReceipts(hashes)
+
+	return nil
 }
 
-// SetIdle sets the peer to idle, allowing it to execute new retrieval requests.
+// FetchNodeData sends a node state data retrieval request to the remote peer.
+func (p *peer) FetchNodeData(request *fetchRequest) error {
+	// Sanity check the protocol version
+	if p.version < 63 {
+		panic(fmt.Sprintf("node data fetch [eth/63+] requested on eth/%d", p.version))
+	}
+	// Short circuit if the peer is already fetching
+	if !atomic.CompareAndSwapInt32(&p.stateIdle, 0, 1) {
+		return errAlreadyFetching
+	}
+	p.stateStarted = time.Now()
+
+	// Convert the hash set to a retrievable slice
+	hashes := make([]common.Hash, 0, len(request.Hashes))
+	for hash, _ := range request.Hashes {
+		hashes = append(hashes, hash)
+	}
+	go p.getNodeData(hashes)
+
+	return nil
+}
+
+// SetBlocksIdle sets the peer to idle, allowing it to execute new retrieval requests.
+// Its block retrieval allowance will also be updated either up- or downwards,
+// depending on whether the previous fetch completed in time.
+func (p *peer) SetBlocksIdle() {
+	p.setIdle(p.blockStarted, blockSoftTTL, blockHardTTL, MaxBlockFetch, &p.blockCapacity, &p.blockIdle)
+}
+
+// SetBodiesIdle sets the peer to idle, allowing it to execute new retrieval requests.
 // Its block body retrieval allowance will also be updated either up- or downwards,
-// depending on whether the previous fetch completed in time or not.
-func (p *peer) SetIdle() {
+// depending on whether the previous fetch completed in time.
+func (p *peer) SetBodiesIdle() {
+	p.setIdle(p.blockStarted, bodySoftTTL, bodyHardTTL, MaxBodyFetch, &p.blockCapacity, &p.blockIdle)
+}
+
+// SetReceiptsIdle sets the peer to idle, allowing it to execute new retrieval requests.
+// Its receipt retrieval allowance will also be updated either up- or downwards,
+// depending on whether the previous fetch completed in time.
+func (p *peer) SetReceiptsIdle() {
+	p.setIdle(p.receiptStarted, receiptSoftTTL, receiptHardTTL, MaxReceiptFetch, &p.receiptCapacity, &p.receiptIdle)
+}
+
+// SetNodeDataIdle sets the peer to idle, allowing it to execute new retrieval
+// requests. Its node data retrieval allowance will also be updated either up- or
+// downwards, depending on whether the previous fetch completed in time.
+func (p *peer) SetNodeDataIdle() {
+	p.setIdle(p.stateStarted, stateSoftTTL, stateSoftTTL, MaxStateFetch, &p.stateCapacity, &p.stateIdle)
+}
+
+// setIdle sets the peer to idle, allowing it to execute new retrieval requests.
+// Its data retrieval allowance will also be updated either up- or downwards,
+// depending on whether the previous fetch completed in time.
+func (p *peer) setIdle(started time.Time, softTTL, hardTTL time.Duration, maxFetch int, capacity, idle *int32) {
 	// Update the peer's download allowance based on previous performance
 	scale := 2.0
-	if time.Since(p.started) > bodySoftTTL {
+	if time.Since(started) > softTTL {
 		scale = 0.5
-		if time.Since(p.started) > bodyHardTTL {
-			scale = 1 / float64(MaxBodyFetch) // reduces capacity to 1
+		if time.Since(started) > hardTTL {
+			scale = 1 / float64(maxFetch) // reduces capacity to 1
 		}
 	}
 	for {
 		// Calculate the new download bandwidth allowance
-		prev := atomic.LoadInt32(&p.capacity)
-		next := int32(math.Max(1, math.Min(float64(MaxBodyFetch), float64(prev)*scale)))
+		prev := atomic.LoadInt32(capacity)
+		next := int32(math.Max(1, math.Min(float64(maxFetch), float64(prev)*scale)))
 
 		// Try to update the old value
-		if atomic.CompareAndSwapInt32(&p.capacity, prev, next) {
+		if atomic.CompareAndSwapInt32(capacity, prev, next) {
 			// If we're having problems at 1 capacity, try to find better peers
 			if next == 1 {
 				p.Demote()
@@ -193,14 +264,26 @@ func (p *peer) SetIdle() {
 			break
 		}
 	}
-	// Set the peer to idle to allow further block requests
-	atomic.StoreInt32(&p.idle, 0)
+	// Set the peer to idle to allow further fetch requests
+	atomic.StoreInt32(idle, 0)
+}
+
+// BlockCapacity retrieves the peers block download allowance based on its
+// previously discovered bandwidth capacity.
+func (p *peer) BlockCapacity() int {
+	return int(atomic.LoadInt32(&p.blockCapacity))
+}
+
+// ReceiptCapacity retrieves the peers block download allowance based on its
+// previously discovered bandwidth capacity.
+func (p *peer) ReceiptCapacity() int {
+	return int(atomic.LoadInt32(&p.receiptCapacity))
 }
 
-// Capacity retrieves the peers block download allowance based on its previously
-// discovered bandwidth capacity.
-func (p *peer) Capacity() int {
-	return int(atomic.LoadInt32(&p.capacity))
+// NodeDataCapacity retrieves the peers block download allowance based on its
+// previously discovered bandwidth capacity.
+func (p *peer) NodeDataCapacity() int {
+	return int(atomic.LoadInt32(&p.stateCapacity))
 }
 
 // Promote increases the peer's reputation.
@@ -226,7 +309,8 @@ func (p *peer) Demote() {
 func (p *peer) String() string {
 	return fmt.Sprintf("Peer %s [%s]", p.id,
 		fmt.Sprintf("reputation %3d, ", atomic.LoadInt32(&p.rep))+
-			fmt.Sprintf("capacity %3d, ", atomic.LoadInt32(&p.capacity))+
+			fmt.Sprintf("block cap %3d, ", atomic.LoadInt32(&p.blockCapacity))+
+			fmt.Sprintf("receipt cap %3d, ", atomic.LoadInt32(&p.receiptCapacity))+
 			fmt.Sprintf("ignored %4d", p.ignored.Size()),
 	)
 }
@@ -310,26 +394,63 @@ func (ps *peerSet) AllPeers() []*peer {
 	return list
 }
 
-// IdlePeers retrieves a flat list of all the currently idle peers within the
+// BlockIdlePeers retrieves a flat list of all the currently idle peers within the
 // active peer set, ordered by their reputation.
-func (ps *peerSet) IdlePeers(version int) []*peer {
+func (ps *peerSet) BlockIdlePeers() ([]*peer, int) {
+	idle := func(p *peer) bool {
+		return atomic.LoadInt32(&p.blockIdle) == 0
+	}
+	return ps.idlePeers(61, 61, idle)
+}
+
+// BodyIdlePeers retrieves a flat list of all the currently body-idle peers within
+// the active peer set, ordered by their reputation.
+func (ps *peerSet) BodyIdlePeers() ([]*peer, int) {
+	idle := func(p *peer) bool {
+		return atomic.LoadInt32(&p.blockIdle) == 0
+	}
+	return ps.idlePeers(62, 64, idle)
+}
+
+// ReceiptIdlePeers retrieves a flat list of all the currently receipt-idle peers
+// within the active peer set, ordered by their reputation.
+func (ps *peerSet) ReceiptIdlePeers() ([]*peer, int) {
+	idle := func(p *peer) bool {
+		return atomic.LoadInt32(&p.receiptIdle) == 0
+	}
+	return ps.idlePeers(63, 64, idle)
+}
+
+// NodeDataIdlePeers retrieves a flat list of all the currently node-data-idle
+// peers within the active peer set, ordered by their reputation.
+func (ps *peerSet) NodeDataIdlePeers() ([]*peer, int) {
+	idle := func(p *peer) bool {
+		return atomic.LoadInt32(&p.stateIdle) == 0
+	}
+	return ps.idlePeers(63, 64, idle)
+}
+
+// idlePeers retrieves a flat list of all currently idle peers satisfying the
+// protocol version constraints, using the provided function to check idleness.
+func (ps *peerSet) idlePeers(minProtocol, maxProtocol int, idleCheck func(*peer) bool) ([]*peer, int) {
 	ps.lock.RLock()
 	defer ps.lock.RUnlock()
 
-	list := make([]*peer, 0, len(ps.peers))
+	idle, total := make([]*peer, 0, len(ps.peers)), 0
 	for _, p := range ps.peers {
-		if (version == eth61 && p.version == eth61) || (version >= eth62 && p.version >= eth62) {
-			if atomic.LoadInt32(&p.idle) == 0 {
-				list = append(list, p)
+		if p.version >= minProtocol && p.version <= maxProtocol {
+			if idleCheck(p) {
+				idle = append(idle, p)
 			}
+			total++
 		}
 	}
-	for i := 0; i < len(list); i++ {
-		for j := i + 1; j < len(list); j++ {
-			if atomic.LoadInt32(&list[i].rep) < atomic.LoadInt32(&list[j].rep) {
-				list[i], list[j] = list[j], list[i]
+	for i := 0; i < len(idle); i++ {
+		for j := i + 1; j < len(idle); j++ {
+			if atomic.LoadInt32(&idle[i].rep) < atomic.LoadInt32(&idle[j].rep) {
+				idle[i], idle[j] = idle[j], idle[i]
 			}
 		}
 	}
-	return list
+	return idle, total
 }

eth/downloader/types.go

@@ -0,0 +1,140 @@
+// Copyright 2015 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 downloader
+
+import (
+	"fmt"
+	"math/big"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core/types"
+)
+
+// headerCheckFn is a callback type for verifying a header's presence in the local chain.
+type headerCheckFn func(common.Hash) bool
+
+// blockCheckFn is a callback type for verifying a block's presence in the local chain.
+type blockCheckFn func(common.Hash) bool
+
+// headerRetrievalFn is a callback type for retrieving a header from the local chain.
+type headerRetrievalFn func(common.Hash) *types.Header
+
+// blockRetrievalFn is a callback type for retrieving a block from the local chain.
+type blockRetrievalFn func(common.Hash) *types.Block
+
+// headHeaderRetrievalFn is a callback type for retrieving the head header from the local chain.
+type headHeaderRetrievalFn func() *types.Header
+
+// headBlockRetrievalFn is a callback type for retrieving the head block from the local chain.
+type headBlockRetrievalFn func() *types.Block
+
+// headFastBlockRetrievalFn is a callback type for retrieving the head fast block from the local chain.
+type headFastBlockRetrievalFn func() *types.Block
+
+// headBlockCommitterFn is a callback for directly committing the head block to a certain entity.
+type headBlockCommitterFn func(common.Hash) error
+
+// tdRetrievalFn is a callback type for retrieving the total difficulty of a local block.
+type tdRetrievalFn func(common.Hash) *big.Int
+
+// headerChainInsertFn is a callback type to insert a batch of headers into the local chain.
+type headerChainInsertFn func([]*types.Header, int) (int, error)
+
+// blockChainInsertFn is a callback type to insert a batch of blocks into the local chain.
+type blockChainInsertFn func(types.Blocks) (int, error)
+
+// receiptChainInsertFn is a callback type to insert a batch of receipts into the local chain.
+type receiptChainInsertFn func(types.Blocks, []types.Receipts) (int, error)
+
+// chainRollbackFn is a callback type to remove a few recently added elements from the local chain.
+type chainRollbackFn func([]common.Hash)
+
+// peerDropFn is a callback type for dropping a peer detected as malicious.
+type peerDropFn func(id string)
+
+// dataPack is a data message returned by a peer for some query.
+type dataPack interface {
+	PeerId() string
+	Items() int
+	Stats() string
+}
+
+// hashPack is a batch of block hashes returned by a peer (eth/61).
+type hashPack struct {
+	peerId string
+	hashes []common.Hash
+}
+
+func (p *hashPack) PeerId() string { return p.peerId }
+func (p *hashPack) Items() int     { return len(p.hashes) }
+func (p *hashPack) Stats() string  { return fmt.Sprintf("%d", len(p.hashes)) }
+
+// blockPack is a batch of blocks returned by a peer (eth/61).
+type blockPack struct {
+	peerId string
+	blocks []*types.Block
+}
+
+func (p *blockPack) PeerId() string { return p.peerId }
+func (p *blockPack) Items() int     { return len(p.blocks) }
+func (p *blockPack) Stats() string  { return fmt.Sprintf("%d", len(p.blocks)) }
+
+// headerPack is a batch of block headers returned by a peer.
+type headerPack struct {
+	peerId  string
+	headers []*types.Header
+}
+
+func (p *headerPack) PeerId() string { return p.peerId }
+func (p *headerPack) Items() int     { return len(p.headers) }
+func (p *headerPack) Stats() string  { return fmt.Sprintf("%d", len(p.headers)) }
+
+// bodyPack is a batch of block bodies returned by a peer.
+type bodyPack struct {
+	peerId       string
+	transactions [][]*types.Transaction
+	uncles       [][]*types.Header
+}
+
+func (p *bodyPack) PeerId() string { return p.peerId }
+func (p *bodyPack) Items() int {
+	if len(p.transactions) <= len(p.uncles) {
+		return len(p.transactions)
+	}
+	return len(p.uncles)
+}
+func (p *bodyPack) Stats() string { return fmt.Sprintf("%d:%d", len(p.transactions), len(p.uncles)) }
+
+// receiptPack is a batch of receipts returned by a peer.
+type receiptPack struct {
+	peerId   string
+	receipts [][]*types.Receipt
+}
+
+func (p *receiptPack) PeerId() string { return p.peerId }
+func (p *receiptPack) Items() int     { return len(p.receipts) }
+func (p *receiptPack) Stats() string  { return fmt.Sprintf("%d", len(p.receipts)) }
+
+// statePack is a batch of states returned by a peer.
+type statePack struct {
+	peerId string
+	states [][]byte
+}
+
+func (p *statePack) PeerId() string { return p.peerId }
+func (p *statePack) Items() int     { return len(p.states) }
+func (p *statePack) Stats() string  { return fmt.Sprintf("%d", len(p.states)) }

eth/fetcher/fetcher.go

@@ -142,9 +142,11 @@ type Fetcher struct {
 	dropPeer       peerDropFn         // Drops a peer for misbehaving
 
 	// Testing hooks
-	fetchingHook   func([]common.Hash) // Method to call upon starting a block (eth/61) or header (eth/62) fetch
-	completingHook func([]common.Hash) // Method to call upon starting a block body fetch (eth/62)
-	importedHook   func(*types.Block)  // Method to call upon successful block import (both eth/61 and eth/62)
+	announceChangeHook func(common.Hash, bool) // Method to call upon adding or deleting a hash from the announce list
+	queueChangeHook    func(common.Hash, bool) // Method to call upon adding or deleting a block from the import queue
+	fetchingHook       func([]common.Hash)     // Method to call upon starting a block (eth/61) or header (eth/62) fetch
+	completingHook     func([]common.Hash)     // Method to call upon starting a block body fetch (eth/62)
+	importedHook       func(*types.Block)      // Method to call upon successful block import (both eth/61 and eth/62)
 }
 
 // New creates a block fetcher to retrieve blocks based on hash announcements.
@@ -324,11 +326,16 @@ func (f *Fetcher) loop() {
 		height := f.chainHeight()
 		for !f.queue.Empty() {
 			op := f.queue.PopItem().(*inject)
-
+			if f.queueChangeHook != nil {
+				f.queueChangeHook(op.block.Hash(), false)
+			}
 			// If too high up the chain or phase, continue later
 			number := op.block.NumberU64()
 			if number > height+1 {
 				f.queue.Push(op, -float32(op.block.NumberU64()))
+				if f.queueChangeHook != nil {
+					f.queueChangeHook(op.block.Hash(), true)
+				}
 				break
 			}
 			// Otherwise if fresh and still unknown, try and import
@@ -372,6 +379,9 @@ func (f *Fetcher) loop() {
 			}
 			f.announces[notification.origin] = count
 			f.announced[notification.hash] = append(f.announced[notification.hash], notification)
+			if f.announceChangeHook != nil && len(f.announced[notification.hash]) == 1 {
+				f.announceChangeHook(notification.hash, true)
+			}
 			if len(f.announced) == 1 {
 				f.rescheduleFetch(fetchTimer)
 			}
@@ -714,7 +724,9 @@ func (f *Fetcher) enqueue(peer string, block *types.Block) {
 		f.queues[peer] = count
 		f.queued[hash] = op
 		f.queue.Push(op, -float32(block.NumberU64()))
-
+		if f.queueChangeHook != nil {
+			f.queueChangeHook(op.block.Hash(), true)
+		}
 		if glog.V(logger.Debug) {
 			glog.Infof("Peer %s: queued block #%d [%x…], total %v", peer, block.NumberU64(), hash.Bytes()[:4], f.queue.Size())
 		}
@@ -781,7 +793,9 @@ func (f *Fetcher) forgetHash(hash common.Hash) {
 		}
 	}
 	delete(f.announced, hash)
-
+	if f.announceChangeHook != nil {
+		f.announceChangeHook(hash, false)
+	}
 	// Remove any pending fetches and decrement the DOS counters
 	if announce := f.fetching[hash]; announce != nil {
 		f.announces[announce.origin]--

eth/fetcher/fetcher_test.go

@@ -45,7 +45,7 @@ var (
 // contains a transaction and every 5th an uncle to allow testing correct block
 // reassembly.
 func makeChain(n int, seed byte, parent *types.Block) ([]common.Hash, map[common.Hash]*types.Block) {
-	blocks := core.GenerateChain(parent, testdb, n, func(i int, block *core.BlockGen) {
+	blocks, _ := core.GenerateChain(parent, testdb, n, func(i int, block *core.BlockGen) {
 		block.SetCoinbase(common.Address{seed})
 
 		// If the block number is multiple of 3, send a bonus transaction to the miner
@@ -145,6 +145,9 @@ func (f *fetcherTester) insertChain(blocks types.Blocks) (int, error) {
 // dropPeer is an emulator for the peer removal, simply accumulating the various
 // peers dropped by the fetcher.
 func (f *fetcherTester) dropPeer(peer string) {
+	f.lock.Lock()
+	defer f.lock.Unlock()
+
 	f.drops[peer] = true
 }
 
@@ -608,8 +611,11 @@ func TestDistantPropagationDiscarding(t *testing.T) {
 
 	// Create a tester and simulate a head block being the middle of the above chain
 	tester := newTester()
+
+	tester.lock.Lock()
 	tester.hashes = []common.Hash{head}
 	tester.blocks = map[common.Hash]*types.Block{head: blocks[head]}
+	tester.lock.Unlock()
 
 	// Ensure that a block with a lower number than the threshold is discarded
 	tester.fetcher.Enqueue("lower", blocks[hashes[low]])
@@ -641,8 +647,11 @@ func testDistantAnnouncementDiscarding(t *testing.T, protocol int) {
 
 	// Create a tester and simulate a head block being the middle of the above chain
 	tester := newTester()
+
+	tester.lock.Lock()
 	tester.hashes = []common.Hash{head}
 	tester.blocks = map[common.Hash]*types.Block{head: blocks[head]}
+	tester.lock.Unlock()
 
 	headerFetcher := tester.makeHeaderFetcher(blocks, -gatherSlack)
 	bodyFetcher := tester.makeBodyFetcher(blocks, 0)
@@ -687,14 +696,22 @@ func testInvalidNumberAnnouncement(t *testing.T, protocol int) {
 	tester.fetcher.Notify("bad", hashes[0], 2, time.Now().Add(-arriveTimeout), nil, headerFetcher, bodyFetcher)
 	verifyImportEvent(t, imported, false)
 
-	if !tester.drops["bad"] {
+	tester.lock.RLock()
+	dropped := tester.drops["bad"]
+	tester.lock.RUnlock()
+
+	if !dropped {
 		t.Fatalf("peer with invalid numbered announcement not dropped")
 	}
 	// Make sure a good announcement passes without a drop
 	tester.fetcher.Notify("good", hashes[0], 1, time.Now().Add(-arriveTimeout), nil, headerFetcher, bodyFetcher)
 	verifyImportEvent(t, imported, true)
 
-	if tester.drops["good"] {
+	tester.lock.RLock()
+	dropped = tester.drops["good"]
+	tester.lock.RUnlock()
+
+	if dropped {
 		t.Fatalf("peer with valid numbered announcement dropped")
 	}
 	verifyImportDone(t, imported)
@@ -752,9 +769,15 @@ func testHashMemoryExhaustionAttack(t *testing.T, protocol int) {
 	// Create a tester with instrumented import hooks
 	tester := newTester()
 
-	imported := make(chan *types.Block)
+	imported, announces := make(chan *types.Block), int32(0)
 	tester.fetcher.importedHook = func(block *types.Block) { imported <- block }
-
+	tester.fetcher.announceChangeHook = func(hash common.Hash, added bool) {
+		if added {
+			atomic.AddInt32(&announces, 1)
+		} else {
+			atomic.AddInt32(&announces, -1)
+		}
+	}
 	// Create a valid chain and an infinite junk chain
 	targetBlocks := hashLimit + 2*maxQueueDist
 	hashes, blocks := makeChain(targetBlocks, 0, genesis)
@@ -782,8 +805,8 @@ func testHashMemoryExhaustionAttack(t *testing.T, protocol int) {
 			tester.fetcher.Notify("attacker", attack[i], 1 /* don't distance drop */, time.Now(), nil, attackerHeaderFetcher, attackerBodyFetcher)
 		}
 	}
-	if len(tester.fetcher.announced) != hashLimit+maxQueueDist {
-		t.Fatalf("queued announce count mismatch: have %d, want %d", len(tester.fetcher.announced), hashLimit+maxQueueDist)
+	if count := atomic.LoadInt32(&announces); count != hashLimit+maxQueueDist {
+		t.Fatalf("queued announce count mismatch: have %d, want %d", count, hashLimit+maxQueueDist)
 	}
 	// Wait for fetches to complete
 	verifyImportCount(t, imported, maxQueueDist)
@@ -807,9 +830,15 @@ func TestBlockMemoryExhaustionAttack(t *testing.T) {
 	// Create a tester with instrumented import hooks
 	tester := newTester()
 
-	imported := make(chan *types.Block)
+	imported, enqueued := make(chan *types.Block), int32(0)
 	tester.fetcher.importedHook = func(block *types.Block) { imported <- block }
-
+	tester.fetcher.queueChangeHook = func(hash common.Hash, added bool) {
+		if added {
+			atomic.AddInt32(&enqueued, 1)
+		} else {
+			atomic.AddInt32(&enqueued, -1)
+		}
+	}
 	// Create a valid chain and a batch of dangling (but in range) blocks
 	targetBlocks := hashLimit + 2*maxQueueDist
 	hashes, blocks := makeChain(targetBlocks, 0, genesis)
@@ -825,7 +854,7 @@ func TestBlockMemoryExhaustionAttack(t *testing.T) {
 		tester.fetcher.Enqueue("attacker", block)
 	}
 	time.Sleep(200 * time.Millisecond)
-	if queued := tester.fetcher.queue.Size(); queued != blockLimit {
+	if queued := atomic.LoadInt32(&enqueued); queued != blockLimit {
 		t.Fatalf("queued block count mismatch: have %d, want %d", queued, blockLimit)
 	}
 	// Queue up a batch of valid blocks, and check that a new peer is allowed to do so
@@ -833,7 +862,7 @@ func TestBlockMemoryExhaustionAttack(t *testing.T) {
 		tester.fetcher.Enqueue("valid", blocks[hashes[len(hashes)-3-i]])
 	}
 	time.Sleep(100 * time.Millisecond)
-	if queued := tester.fetcher.queue.Size(); queued != blockLimit+maxQueueDist-1 {
+	if queued := atomic.LoadInt32(&enqueued); queued != blockLimit+maxQueueDist-1 {
 		t.Fatalf("queued block count mismatch: have %d, want %d", queued, blockLimit+maxQueueDist-1)
 	}
 	// Insert the missing piece (and sanity check the import)

eth/filters/filter.go

@@ -138,7 +138,7 @@ func (self *Filter) getLogs(start, end uint64) (logs vm.Logs) {
 				unfiltered vm.Logs
 			)
 			for _, receipt := range receipts {
-				unfiltered = append(unfiltered, receipt.Logs()...)
+				unfiltered = append(unfiltered, receipt.Logs...)
 			}
 			logs = append(logs, self.FilterLogs(unfiltered)...)
 		}

eth/filters/filter_test.go

@@ -16,9 +16,9 @@ import (
 
 func makeReceipt(addr common.Address) *types.Receipt {
 	receipt := types.NewReceipt(nil, new(big.Int))
-	receipt.SetLogs(vm.Logs{
+	receipt.Logs = vm.Logs{
 		&vm.Log{Address: addr},
-	})
+	}
 	receipt.Bloom = types.CreateBloom(types.Receipts{receipt})
 	return receipt
 }
@@ -41,7 +41,7 @@ func BenchmarkMipmaps(b *testing.B) {
 	defer db.Close()
 
 	genesis := core.WriteGenesisBlockForTesting(db, core.GenesisAccount{addr1, big.NewInt(1000000)})
-	chain := core.GenerateChain(genesis, db, 100010, func(i int, gen *core.BlockGen) {
+	chain, receipts := core.GenerateChain(genesis, db, 100010, func(i int, gen *core.BlockGen) {
 		var receipts types.Receipts
 		switch i {
 		case 2403:
@@ -70,7 +70,7 @@ func BenchmarkMipmaps(b *testing.B) {
 		}
 		core.WriteMipmapBloom(db, uint64(i+1), receipts)
 	})
-	for _, block := range chain {
+	for i, block := range chain {
 		core.WriteBlock(db, block)
 		if err := core.WriteCanonicalHash(db, block.Hash(), block.NumberU64()); err != nil {
 			b.Fatalf("failed to insert block number: %v", err)
@@ -78,11 +78,10 @@ func BenchmarkMipmaps(b *testing.B) {
 		if err := core.WriteHeadBlockHash(db, block.Hash()); err != nil {
 			b.Fatalf("failed to insert block number: %v", err)
 		}
-		if err := core.PutBlockReceipts(db, block, block.Receipts()); err != nil {
+		if err := core.PutBlockReceipts(db, block.Hash(), receipts[i]); err != nil {
 			b.Fatal("error writing block receipts:", err)
 		}
 	}
-
 	b.ResetTimer()
 
 	filter := New(db)
@@ -118,47 +117,47 @@ func TestFilters(t *testing.T) {
 	defer db.Close()
 
 	genesis := core.WriteGenesisBlockForTesting(db, core.GenesisAccount{addr, big.NewInt(1000000)})
-	chain := core.GenerateChain(genesis, db, 1000, func(i int, gen *core.BlockGen) {
+	chain, receipts := core.GenerateChain(genesis, db, 1000, func(i int, gen *core.BlockGen) {
 		var receipts types.Receipts
 		switch i {
 		case 1:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{
+			receipt.Logs = vm.Logs{
 				&vm.Log{
 					Address: addr,
 					Topics:  []common.Hash{hash1},
 				},
-			})
+			}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		case 2:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{
+			receipt.Logs = vm.Logs{
 				&vm.Log{
 					Address: addr,
 					Topics:  []common.Hash{hash2},
 				},
-			})
+			}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		case 998:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{
+			receipt.Logs = vm.Logs{
 				&vm.Log{
 					Address: addr,
 					Topics:  []common.Hash{hash3},
 				},
-			})
+			}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		case 999:
 			receipt := types.NewReceipt(nil, new(big.Int))
-			receipt.SetLogs(vm.Logs{
+			receipt.Logs = vm.Logs{
 				&vm.Log{
 					Address: addr,
 					Topics:  []common.Hash{hash4},
 				},
-			})
+			}
 			gen.AddUncheckedReceipt(receipt)
 			receipts = types.Receipts{receipt}
 		}
@@ -173,7 +172,7 @@ func TestFilters(t *testing.T) {
 		// by one
 		core.WriteMipmapBloom(db, uint64(i+1), receipts)
 	})
-	for _, block := range chain {
+	for i, block := range chain {
 		core.WriteBlock(db, block)
 		if err := core.WriteCanonicalHash(db, block.Hash(), block.NumberU64()); err != nil {
 			t.Fatalf("failed to insert block number: %v", err)
@@ -181,7 +180,7 @@ func TestFilters(t *testing.T) {
 		if err := core.WriteHeadBlockHash(db, block.Hash()); err != nil {
 			t.Fatalf("failed to insert block number: %v", err)
 		}
-		if err := core.PutBlockReceipts(db, block, block.Receipts()); err != nil {
+		if err := core.PutBlockReceipts(db, block.Hash(), receipts[i]); err != nil {
 			t.Fatal("error writing block receipts:", err)
 		}
 	}

eth/handler.go

@@ -17,6 +17,7 @@
 package eth
 
 import (
+	"errors"
 	"fmt"
 	"math"
 	"math/big"
@@ -42,6 +43,10 @@ const (
 	estHeaderRlpSize  = 500             // Approximate size of an RLP encoded block header
 )
 
+// errIncompatibleConfig is returned if the requested protocols and configs are
+// not compatible (low protocol version restrictions and high requirements).
+var errIncompatibleConfig = errors.New("incompatible configuration")
+
 func errResp(code errCode, format string, v ...interface{}) error {
 	return fmt.Errorf("%v - %v", code, fmt.Sprintf(format, v...))
 }
@@ -49,17 +54,8 @@ func errResp(code errCode, format string, v ...interface{}) error {
 type hashFetcherFn func(common.Hash) error
 type blockFetcherFn func([]common.Hash) error
 
-// extProt is an interface which is passed around so we can expose GetHashes and GetBlock without exposing it to the rest of the protocol
-// extProt is passed around to peers which require to GetHashes and GetBlocks
-type extProt struct {
-	getHashes hashFetcherFn
-	getBlocks blockFetcherFn
-}
-
-func (ep extProt) GetHashes(hash common.Hash) error    { return ep.getHashes(hash) }
-func (ep extProt) GetBlock(hashes []common.Hash) error { return ep.getBlocks(hashes) }
-
 type ProtocolManager struct {
+	fastSync   bool
 	txpool     txPool
 	blockchain *core.BlockChain
 	chaindb    ethdb.Database
@@ -87,9 +83,15 @@ type ProtocolManager struct {
 
 // NewProtocolManager returns a new ethereum sub protocol manager. The Ethereum sub protocol manages peers capable
 // with the ethereum network.
-func NewProtocolManager(networkId int, mux *event.TypeMux, txpool txPool, pow pow.PoW, blockchain *core.BlockChain, chaindb ethdb.Database) *ProtocolManager {
+func NewProtocolManager(fastSync bool, networkId int, mux *event.TypeMux, txpool txPool, pow pow.PoW, blockchain *core.BlockChain, chaindb ethdb.Database) (*ProtocolManager, error) {
+	// Figure out whether to allow fast sync or not
+	if fastSync && blockchain.CurrentBlock().NumberU64() > 0 {
+		glog.V(logger.Info).Infof("blockchain not empty, fast sync disabled")
+		fastSync = false
+	}
 	// Create the protocol manager with the base fields
 	manager := &ProtocolManager{
+		fastSync:   fastSync,
 		eventMux:   mux,
 		txpool:     txpool,
 		blockchain: blockchain,
@@ -100,11 +102,15 @@ func NewProtocolManager(networkId int, mux *event.TypeMux, txpool txPool, pow po
 		quitSync:   make(chan struct{}),
 	}
 	// Initiate a sub-protocol for every implemented version we can handle
-	manager.SubProtocols = make([]p2p.Protocol, len(ProtocolVersions))
-	for i := 0; i < len(manager.SubProtocols); i++ {
-		version := ProtocolVersions[i]
-
-		manager.SubProtocols[i] = p2p.Protocol{
+	manager.SubProtocols = make([]p2p.Protocol, 0, len(ProtocolVersions))
+	for i, version := range ProtocolVersions {
+		// Skip protocol version if incompatible with the mode of operation
+		if fastSync && version < eth63 {
+			continue
+		}
+		// Compatible; initialise the sub-protocol
+		version := version // Closure for the run
+		manager.SubProtocols = append(manager.SubProtocols, p2p.Protocol{
 			Name:    "eth",
 			Version: version,
 			Length:  ProtocolLengths[i],
@@ -113,20 +119,25 @@ func NewProtocolManager(networkId int, mux *event.TypeMux, txpool txPool, pow po
 				manager.newPeerCh <- peer
 				return manager.handle(peer)
 			},
-		}
+		})
+	}
+	if len(manager.SubProtocols) == 0 {
+		return nil, errIncompatibleConfig
 	}
 	// Construct the different synchronisation mechanisms
-	manager.downloader = downloader.New(manager.eventMux, manager.blockchain.HasBlock, manager.blockchain.GetBlock, manager.blockchain.CurrentBlock, manager.blockchain.GetTd, manager.blockchain.InsertChain, manager.removePeer)
+	manager.downloader = downloader.New(chaindb, manager.eventMux, blockchain.HasHeader, blockchain.HasBlock, blockchain.GetHeader, blockchain.GetBlock,
+		blockchain.CurrentHeader, blockchain.CurrentBlock, blockchain.CurrentFastBlock, blockchain.FastSyncCommitHead, blockchain.GetTd,
+		blockchain.InsertHeaderChain, blockchain.InsertChain, blockchain.InsertReceiptChain, blockchain.Rollback, manager.removePeer)
 
 	validator := func(block *types.Block, parent *types.Block) error {
 		return core.ValidateHeader(pow, block.Header(), parent.Header(), true, false)
 	}
 	heighter := func() uint64 {
-		return manager.blockchain.CurrentBlock().NumberU64()
+		return blockchain.CurrentBlock().NumberU64()
 	}
-	manager.fetcher = fetcher.New(manager.blockchain.GetBlock, validator, manager.BroadcastBlock, heighter, manager.blockchain.InsertChain, manager.removePeer)
+	manager.fetcher = fetcher.New(blockchain.GetBlock, validator, manager.BroadcastBlock, heighter, blockchain.InsertChain, manager.removePeer)
 
-	return manager
+	return manager, nil
 }
 
 func (pm *ProtocolManager) removePeer(id string) {
@@ -205,8 +216,8 @@ func (pm *ProtocolManager) handle(p *peer) error {
 
 	// Register the peer in the downloader. If the downloader considers it banned, we disconnect
 	if err := pm.downloader.RegisterPeer(p.id, p.version, p.Head(),
-		p.RequestHashes, p.RequestHashesFromNumber, p.RequestBlocks,
-		p.RequestHeadersByHash, p.RequestHeadersByNumber, p.RequestBodies); err != nil {
+		p.RequestHashes, p.RequestHashesFromNumber, p.RequestBlocks, p.RequestHeadersByHash,
+		p.RequestHeadersByNumber, p.RequestBodies, p.RequestReceipts, p.RequestNodeData); err != nil {
 		return err
 	}
 	// Propagate existing transactions. new transactions appearing
@@ -292,7 +303,7 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 			break
 		}
 		// Deliver them all to the downloader for queuing
-		err := pm.downloader.DeliverHashes61(p.id, hashes)
+		err := pm.downloader.DeliverHashes(p.id, hashes)
 		if err != nil {
 			glog.V(logger.Debug).Infoln(err)
 		}
@@ -338,7 +349,7 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 		}
 		// Filter out any explicitly requested blocks, deliver the rest to the downloader
 		if blocks := pm.fetcher.FilterBlocks(blocks); len(blocks) > 0 {
-			pm.downloader.DeliverBlocks61(p.id, blocks)
+			pm.downloader.DeliverBlocks(p.id, blocks)
 		}
 
 	// Block header query, collect the requested headers and reply
@@ -424,28 +435,6 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 			}
 		}
 
-	case p.version >= eth62 && msg.Code == BlockBodiesMsg:
-		// A batch of block bodies arrived to one of our previous requests
-		var request blockBodiesData
-		if err := msg.Decode(&request); err != nil {
-			return errResp(ErrDecode, "msg %v: %v", msg, err)
-		}
-		// Deliver them all to the downloader for queuing
-		trasactions := make([][]*types.Transaction, len(request))
-		uncles := make([][]*types.Header, len(request))
-
-		for i, body := range request {
-			trasactions[i] = body.Transactions
-			uncles[i] = body.Uncles
-		}
-		// Filter out any explicitly requested bodies, deliver the rest to the downloader
-		if trasactions, uncles := pm.fetcher.FilterBodies(trasactions, uncles, time.Now()); len(trasactions) > 0 || len(uncles) > 0 {
-			err := pm.downloader.DeliverBodies(p.id, trasactions, uncles)
-			if err != nil {
-				glog.V(logger.Debug).Infoln(err)
-			}
-		}
-
 	case p.version >= eth62 && msg.Code == GetBlockBodiesMsg:
 		// Decode the retrieval message
 		msgStream := rlp.NewStream(msg.Payload, uint64(msg.Size))
@@ -473,6 +462,28 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 		}
 		return p.SendBlockBodiesRLP(bodies)
 
+	case p.version >= eth62 && msg.Code == BlockBodiesMsg:
+		// A batch of block bodies arrived to one of our previous requests
+		var request blockBodiesData
+		if err := msg.Decode(&request); err != nil {
+			return errResp(ErrDecode, "msg %v: %v", msg, err)
+		}
+		// Deliver them all to the downloader for queuing
+		trasactions := make([][]*types.Transaction, len(request))
+		uncles := make([][]*types.Header, len(request))
+
+		for i, body := range request {
+			trasactions[i] = body.Transactions
+			uncles[i] = body.Uncles
+		}
+		// Filter out any explicitly requested bodies, deliver the rest to the downloader
+		if trasactions, uncles := pm.fetcher.FilterBodies(trasactions, uncles, time.Now()); len(trasactions) > 0 || len(uncles) > 0 {
+			err := pm.downloader.DeliverBodies(p.id, trasactions, uncles)
+			if err != nil {
+				glog.V(logger.Debug).Infoln(err)
+			}
+		}
+
 	case p.version >= eth63 && msg.Code == GetNodeDataMsg:
 		// Decode the retrieval message
 		msgStream := rlp.NewStream(msg.Payload, uint64(msg.Size))
@@ -500,6 +511,17 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 		}
 		return p.SendNodeData(data)
 
+	case p.version >= eth63 && msg.Code == NodeDataMsg:
+		// A batch of node state data arrived to one of our previous requests
+		var data [][]byte
+		if err := msg.Decode(&data); err != nil {
+			return errResp(ErrDecode, "msg %v: %v", msg, err)
+		}
+		// Deliver all to the downloader
+		if err := pm.downloader.DeliverNodeData(p.id, data); err != nil {
+			glog.V(logger.Debug).Infof("failed to deliver node state data: %v", err)
+		}
+
 	case p.version >= eth63 && msg.Code == GetReceiptsMsg:
 		// Decode the retrieval message
 		msgStream := rlp.NewStream(msg.Payload, uint64(msg.Size))
@@ -510,22 +532,42 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 		var (
 			hash     common.Hash
 			bytes    int
-			receipts []*types.Receipt
+			receipts []rlp.RawValue
 		)
-		for bytes < softResponseLimit && len(receipts) < downloader.MaxReceiptsFetch {
-			// Retrieve the hash of the next transaction receipt
+		for bytes < softResponseLimit && len(receipts) < downloader.MaxReceiptFetch {
+			// Retrieve the hash of the next block
 			if err := msgStream.Decode(&hash); err == rlp.EOL {
 				break
 			} else if err != nil {
 				return errResp(ErrDecode, "msg %v: %v", msg, err)
 			}
-			// Retrieve the requested receipt, stopping if enough was found
-			if receipt := core.GetReceipt(pm.chaindb, hash); receipt != nil {
-				receipts = append(receipts, receipt)
-				bytes += len(receipt.RlpEncode())
+			// Retrieve the requested block's receipts, skipping if unknown to us
+			results := core.GetBlockReceipts(pm.chaindb, hash)
+			if results == nil {
+				if header := pm.blockchain.GetHeader(hash); header == nil || header.ReceiptHash != types.EmptyRootHash {
+					continue
+				}
+			}
+			// If known, encode and queue for response packet
+			if encoded, err := rlp.EncodeToBytes(results); err != nil {
+				glog.V(logger.Error).Infof("failed to encode receipt: %v", err)
+			} else {
+				receipts = append(receipts, encoded)
+				bytes += len(encoded)
 			}
 		}
-		return p.SendReceipts(receipts)
+		return p.SendReceiptsRLP(receipts)
+
+	case p.version >= eth63 && msg.Code == ReceiptsMsg:
+		// A batch of receipts arrived to one of our previous requests
+		var receipts [][]*types.Receipt
+		if err := msg.Decode(&receipts); err != nil {
+			return errResp(ErrDecode, "msg %v: %v", msg, err)
+		}
+		// Deliver all to the downloader
+		if err := pm.downloader.DeliverReceipts(p.id, receipts); err != nil {
+			glog.V(logger.Debug).Infof("failed to deliver receipts: %v", err)
+		}
 
 	case msg.Code == NewBlockHashesMsg:
 		// Retrieve and deseralize the remote new block hashes notification
@@ -585,15 +627,6 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 		}
 		request.Block.ReceivedAt = msg.ReceivedAt
 
-		// Mark the block's arrival for whatever reason
-		_, chainHead, _ := pm.blockchain.Status()
-		jsonlogger.LogJson(&logger.EthChainReceivedNewBlock{
-			BlockHash:     request.Block.Hash().Hex(),
-			BlockNumber:   request.Block.Number(),
-			ChainHeadHash: chainHead.Hex(),
-			BlockPrevHash: request.Block.ParentHash().Hex(),
-			RemoteId:      p.ID().String(),
-		})
 		// Mark the peer as owning the block and schedule it for import
 		p.MarkBlock(request.Block.Hash())
 		p.SetHead(request.Block.Hash())
@@ -603,7 +636,8 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 		// Update the peers total difficulty if needed, schedule a download if gapped
 		if request.TD.Cmp(p.Td()) > 0 {
 			p.SetTd(request.TD)
-			if request.TD.Cmp(new(big.Int).Add(pm.blockchain.Td(), request.Block.Difficulty())) > 0 {
+			td := pm.blockchain.GetTd(pm.blockchain.CurrentBlock().Hash())
+			if request.TD.Cmp(new(big.Int).Add(td, request.Block.Difficulty())) > 0 {
 				go pm.synchronise(p)
 			}
 		}
@@ -620,12 +654,6 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 				return errResp(ErrDecode, "transaction %d is nil", i)
 			}
 			p.MarkTransaction(tx.Hash())
-
-			// Log it's arrival for later analysis
-			jsonlogger.LogJson(&logger.EthTxReceived{
-				TxHash:   tx.Hash().Hex(),
-				RemoteId: p.ID().String(),
-			})
 		}
 		pm.txpool.AddTransactions(txs)
 

eth/handler_test.go

@@ -17,12 +17,41 @@ import (
 	"github.com/ethereum/go-ethereum/params"
 )
 
+// Tests that protocol versions and modes of operations are matched up properly.
+func TestProtocolCompatibility(t *testing.T) {
+	// Define the compatibility chart
+	tests := []struct {
+		version    uint
+		fastSync   bool
+		compatible bool
+	}{
+		{61, false, true}, {62, false, true}, {63, false, true},
+		{61, true, false}, {62, true, false}, {63, true, true},
+	}
+	// Make sure anything we screw up is restored
+	backup := ProtocolVersions
+	defer func() { ProtocolVersions = backup }()
+
+	// Try all available compatibility configs and check for errors
+	for i, tt := range tests {
+		ProtocolVersions = []uint{tt.version}
+
+		pm, err := newTestProtocolManager(tt.fastSync, 0, nil, nil)
+		if pm != nil {
+			defer pm.Stop()
+		}
+		if (err == nil && !tt.compatible) || (err != nil && tt.compatible) {
+			t.Errorf("test %d: compatibility mismatch: have error %v, want compatibility %v", i, err, tt.compatible)
+		}
+	}
+}
+
 // Tests that hashes can be retrieved from a remote chain by hashes in reverse
 // order.
 func TestGetBlockHashes61(t *testing.T) { testGetBlockHashes(t, 61) }
 
 func testGetBlockHashes(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(downloader.MaxHashFetch+15, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, downloader.MaxHashFetch+15, nil, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
@@ -65,7 +94,7 @@ func testGetBlockHashes(t *testing.T, protocol int) {
 func TestGetBlockHashesFromNumber61(t *testing.T) { testGetBlockHashesFromNumber(t, 61) }
 
 func testGetBlockHashesFromNumber(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(downloader.MaxHashFetch+15, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, downloader.MaxHashFetch+15, nil, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
@@ -105,7 +134,7 @@ func testGetBlockHashesFromNumber(t *testing.T, protocol int) {
 func TestGetBlocks61(t *testing.T) { testGetBlocks(t, 61) }
 
 func testGetBlocks(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(downloader.MaxHashFetch+15, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, downloader.MaxHashFetch+15, nil, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
@@ -174,10 +203,9 @@ func testGetBlocks(t *testing.T, protocol int) {
 // Tests that block headers can be retrieved from a remote chain based on user queries.
 func TestGetBlockHeaders62(t *testing.T) { testGetBlockHeaders(t, 62) }
 func TestGetBlockHeaders63(t *testing.T) { testGetBlockHeaders(t, 63) }
-func TestGetBlockHeaders64(t *testing.T) { testGetBlockHeaders(t, 64) }
 
 func testGetBlockHeaders(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(downloader.MaxHashFetch+15, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, downloader.MaxHashFetch+15, nil, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
@@ -300,10 +328,9 @@ func testGetBlockHeaders(t *testing.T, protocol int) {
 // Tests that block contents can be retrieved from a remote chain based on their hashes.
 func TestGetBlockBodies62(t *testing.T) { testGetBlockBodies(t, 62) }
 func TestGetBlockBodies63(t *testing.T) { testGetBlockBodies(t, 63) }
-func TestGetBlockBodies64(t *testing.T) { testGetBlockBodies(t, 64) }
 
 func testGetBlockBodies(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(downloader.MaxBlockFetch+15, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, downloader.MaxBlockFetch+15, nil, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
@@ -372,7 +399,6 @@ func testGetBlockBodies(t *testing.T, protocol int) {
 
 // Tests that the node state database can be retrieved based on hashes.
 func TestGetNodeData63(t *testing.T) { testGetNodeData(t, 63) }
-func TestGetNodeData64(t *testing.T) { testGetNodeData(t, 64) }
 
 func testGetNodeData(t *testing.T, protocol int) {
 	// Define three accounts to simulate transactions with
@@ -410,14 +436,16 @@ func testGetNodeData(t *testing.T, protocol int) {
 		}
 	}
 	// Assemble the test environment
-	pm := newTestProtocolManager(4, generator, nil)
+	pm := newTestProtocolManagerMust(t, false, 4, generator, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
 	// Fetch for now the entire chain db
 	hashes := []common.Hash{}
 	for _, key := range pm.chaindb.(*ethdb.MemDatabase).Keys() {
-		hashes = append(hashes, common.BytesToHash(key))
+		if len(key) == len(common.Hash{}) {
+			hashes = append(hashes, common.BytesToHash(key))
+		}
 	}
 	p2p.Send(peer.app, 0x0d, hashes)
 	msg, err := peer.app.ReadMsg()
@@ -462,7 +490,6 @@ func testGetNodeData(t *testing.T, protocol int) {
 
 // Tests that the transaction receipts can be retrieved based on hashes.
 func TestGetReceipt63(t *testing.T) { testGetReceipt(t, 63) }
-func TestGetReceipt64(t *testing.T) { testGetReceipt(t, 64) }
 
 func testGetReceipt(t *testing.T, protocol int) {
 	// Define three accounts to simulate transactions with
@@ -500,20 +527,17 @@ func testGetReceipt(t *testing.T, protocol int) {
 		}
 	}
 	// Assemble the test environment
-	pm := newTestProtocolManager(4, generator, nil)
+	pm := newTestProtocolManagerMust(t, false, 4, generator, nil)
 	peer, _ := newTestPeer("peer", protocol, pm, true)
 	defer peer.close()
 
 	// Collect the hashes to request, and the response to expect
-	hashes := []common.Hash{}
+	hashes, receipts := []common.Hash{}, []types.Receipts{}
 	for i := uint64(0); i <= pm.blockchain.CurrentBlock().NumberU64(); i++ {
-		for _, tx := range pm.blockchain.GetBlockByNumber(i).Transactions() {
-			hashes = append(hashes, tx.Hash())
-		}
-	}
-	receipts := make([]*types.Receipt, len(hashes))
-	for i, hash := range hashes {
-		receipts[i] = core.GetReceipt(pm.chaindb, hash)
+		block := pm.blockchain.GetBlockByNumber(i)
+
+		hashes = append(hashes, block.Hash())
+		receipts = append(receipts, core.GetBlockReceipts(pm.chaindb, block.Hash()))
 	}
 	// Send the hash request and verify the response
 	p2p.Send(peer.app, 0x0f, hashes)

eth/helper_test.go

@@ -28,7 +28,7 @@ var (
 // newTestProtocolManager creates a new protocol manager for testing purposes,
 // with the given number of blocks already known, and potential notification
 // channels for different events.
-func newTestProtocolManager(blocks int, generator func(int, *core.BlockGen), newtx chan<- []*types.Transaction) *ProtocolManager {
+func newTestProtocolManager(fastSync bool, blocks int, generator func(int, *core.BlockGen), newtx chan<- []*types.Transaction) (*ProtocolManager, error) {
 	var (
 		evmux         = new(event.TypeMux)
 		pow           = new(core.FakePow)
@@ -38,12 +38,27 @@ func newTestProtocolManager(blocks int, generator func(int, *core.BlockGen), new
 		blockproc     = core.NewBlockProcessor(db, pow, blockchain, evmux)
 	)
 	blockchain.SetProcessor(blockproc)
-	chain := core.GenerateChain(genesis, db, blocks, generator)
+	chain, _ := core.GenerateChain(genesis, db, blocks, generator)
 	if _, err := blockchain.InsertChain(chain); err != nil {
 		panic(err)
 	}
-	pm := NewProtocolManager(NetworkId, evmux, &testTxPool{added: newtx}, pow, blockchain, db)
+	pm, err := NewProtocolManager(fastSync, NetworkId, evmux, &testTxPool{added: newtx}, pow, blockchain, db)
+	if err != nil {
+		return nil, err
+	}
 	pm.Start()
+	return pm, nil
+}
+
+// newTestProtocolManagerMust creates a new protocol manager for testing purposes,
+// with the given number of blocks already known, and potential notification
+// channels for different events. In case of an error, the constructor force-
+// fails the test.
+func newTestProtocolManagerMust(t *testing.T, fastSync bool, blocks int, generator func(int, *core.BlockGen), newtx chan<- []*types.Transaction) *ProtocolManager {
+	pm, err := newTestProtocolManager(fastSync, blocks, generator, newtx)
+	if err != nil {
+		t.Fatalf("Failed to create protocol manager: %v", err)
+	}
 	return pm
 }
 

eth/metrics.go

@@ -101,7 +101,7 @@ func (rw *meteredMsgReadWriter) ReadMsg() (p2p.Msg, error) {
 		packets, traffic = reqBlockInPacketsMeter, reqBlockInTrafficMeter
 
 	case rw.version >= eth62 && msg.Code == BlockHeadersMsg:
-		packets, traffic = reqBlockInPacketsMeter, reqBlockInTrafficMeter
+		packets, traffic = reqHeaderInPacketsMeter, reqHeaderInTrafficMeter
 	case rw.version >= eth62 && msg.Code == BlockBodiesMsg:
 		packets, traffic = reqBodyInPacketsMeter, reqBodyInTrafficMeter
 

eth/peer.go

@@ -191,15 +191,15 @@ func (p *peer) SendBlockBodiesRLP(bodies []rlp.RawValue) error {
 	return p2p.Send(p.rw, BlockBodiesMsg, bodies)
 }
 
-// SendNodeData sends a batch of arbitrary internal data, corresponding to the
+// SendNodeDataRLP sends a batch of arbitrary internal data, corresponding to the
 // hashes requested.
 func (p *peer) SendNodeData(data [][]byte) error {
 	return p2p.Send(p.rw, NodeDataMsg, data)
 }
 
-// SendReceipts sends a batch of transaction receipts, corresponding to the ones
-// requested.
-func (p *peer) SendReceipts(receipts []*types.Receipt) error {
+// SendReceiptsRLP sends a batch of transaction receipts, corresponding to the
+// ones requested from an already RLP encoded format.
+func (p *peer) SendReceiptsRLP(receipts []rlp.RawValue) error {
 	return p2p.Send(p.rw, ReceiptsMsg, receipts)
 }
 

eth/protocol.go

@@ -31,14 +31,13 @@ const (
 	eth61 = 61
 	eth62 = 62
 	eth63 = 63
-	eth64 = 64
 )
 
 // Supported versions of the eth protocol (first is primary).
-var ProtocolVersions = []uint{eth64, eth63, eth62, eth61}
+var ProtocolVersions = []uint{eth63, eth62, eth61}
 
 // Number of implemented message corresponding to different protocol versions.
-var ProtocolLengths = []uint64{15, 12, 8, 9}
+var ProtocolLengths = []uint64{17, 8, 9}
 
 const (
 	NetworkId          = 1
@@ -73,11 +72,6 @@ const (
 	NodeDataMsg    = 0x0e
 	GetReceiptsMsg = 0x0f
 	ReceiptsMsg    = 0x10
-
-	// Protocol messages belonging to eth/64
-	GetAcctProofMsg     = 0x11
-	GetStorageDataProof = 0x12
-	Proof               = 0x13
 )
 
 type errCode int

eth/protocol_test.go

@@ -41,10 +41,9 @@ var testAccount = crypto.NewKey(rand.Reader)
 func TestStatusMsgErrors61(t *testing.T) { testStatusMsgErrors(t, 61) }
 func TestStatusMsgErrors62(t *testing.T) { testStatusMsgErrors(t, 62) }
 func TestStatusMsgErrors63(t *testing.T) { testStatusMsgErrors(t, 63) }
-func TestStatusMsgErrors64(t *testing.T) { testStatusMsgErrors(t, 64) }
 
 func testStatusMsgErrors(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(0, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, 0, nil, nil)
 	td, currentBlock, genesis := pm.blockchain.Status()
 	defer pm.Stop()
 
@@ -95,11 +94,10 @@ func testStatusMsgErrors(t *testing.T, protocol int) {
 func TestRecvTransactions61(t *testing.T) { testRecvTransactions(t, 61) }
 func TestRecvTransactions62(t *testing.T) { testRecvTransactions(t, 62) }
 func TestRecvTransactions63(t *testing.T) { testRecvTransactions(t, 63) }
-func TestRecvTransactions64(t *testing.T) { testRecvTransactions(t, 64) }
 
 func testRecvTransactions(t *testing.T, protocol int) {
 	txAdded := make(chan []*types.Transaction)
-	pm := newTestProtocolManager(0, nil, txAdded)
+	pm := newTestProtocolManagerMust(t, false, 0, nil, txAdded)
 	p, _ := newTestPeer("peer", protocol, pm, true)
 	defer pm.Stop()
 	defer p.close()
@@ -124,10 +122,9 @@ func testRecvTransactions(t *testing.T, protocol int) {
 func TestSendTransactions61(t *testing.T) { testSendTransactions(t, 61) }
 func TestSendTransactions62(t *testing.T) { testSendTransactions(t, 62) }
 func TestSendTransactions63(t *testing.T) { testSendTransactions(t, 63) }
-func TestSendTransactions64(t *testing.T) { testSendTransactions(t, 64) }
 
 func testSendTransactions(t *testing.T, protocol int) {
-	pm := newTestProtocolManager(0, nil, nil)
+	pm := newTestProtocolManagerMust(t, false, 0, nil, nil)
 	defer pm.Stop()
 
 	// Fill the pool with big transactions.

eth/sync.go

@@ -22,6 +22,7 @@ import (
 
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/eth/downloader"
 	"github.com/ethereum/go-ethereum/logger"
 	"github.com/ethereum/go-ethereum/logger/glog"
 	"github.com/ethereum/go-ethereum/p2p/discover"
@@ -160,9 +161,25 @@ func (pm *ProtocolManager) synchronise(peer *peer) {
 		return
 	}
 	// Make sure the peer's TD is higher than our own. If not drop.
-	if peer.Td().Cmp(pm.blockchain.Td()) <= 0 {
+	td := pm.blockchain.GetTd(pm.blockchain.CurrentBlock().Hash())
+	if peer.Td().Cmp(td) <= 0 {
 		return
 	}
 	// Otherwise try to sync with the downloader
-	pm.downloader.Synchronise(peer.id, peer.Head(), peer.Td())
+	mode := downloader.FullSync
+	if pm.fastSync {
+		mode = downloader.FastSync
+	}
+	pm.downloader.Synchronise(peer.id, peer.Head(), peer.Td(), mode)
+
+	// If fast sync was enabled, and we synced up, disable it
+	if pm.fastSync {
+		for pm.downloader.Synchronising() {
+			time.Sleep(100 * time.Millisecond)
+		}
+		if pm.blockchain.CurrentBlock().NumberU64() > 0 {
+			glog.V(logger.Info).Infof("fast sync complete, auto disabling")
+			pm.fastSync = false
+		}
+	}
 }

eth/sync_test.go

@@ -0,0 +1,53 @@
+// Copyright 2015 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 eth
+
+import (
+	"testing"
+	"time"
+
+	"github.com/ethereum/go-ethereum/p2p"
+	"github.com/ethereum/go-ethereum/p2p/discover"
+)
+
+// Tests that fast sync gets disabled as soon as a real block is successfully
+// imported into the blockchain.
+func TestFastSyncDisabling(t *testing.T) {
+	// Create a pristine protocol manager, check that fast sync is left enabled
+	pmEmpty := newTestProtocolManagerMust(t, true, 0, nil, nil)
+	if !pmEmpty.fastSync {
+		t.Fatalf("fast sync disabled on pristine blockchain")
+	}
+	// Create a full protocol manager, check that fast sync gets disabled
+	pmFull := newTestProtocolManagerMust(t, true, 1024, nil, nil)
+	if pmFull.fastSync {
+		t.Fatalf("fast sync not disabled on non-empty blockchain")
+	}
+	// Sync up the two peers
+	io1, io2 := p2p.MsgPipe()
+
+	go pmFull.handle(pmFull.newPeer(63, NetworkId, p2p.NewPeer(discover.NodeID{}, "empty", nil), io2))
+	go pmEmpty.handle(pmEmpty.newPeer(63, NetworkId, p2p.NewPeer(discover.NodeID{}, "full", nil), io1))
+
+	time.Sleep(250 * time.Millisecond)
+	pmEmpty.synchronise(pmEmpty.peers.BestPeer())
+
+	// Check that fast sync was disabled
+	if pmEmpty.fastSync {
+		t.Fatalf("fast sync not disabled after successful synchronisation")
+	}
+}

ethdb/memory_database.go

@@ -17,7 +17,9 @@
 package ethdb
 
 import (
+	"errors"
 	"fmt"
+	"sync"
 
 	"github.com/ethereum/go-ethereum/common"
 )
@@ -26,29 +28,45 @@ import (
  * This is a test memory database. Do not use for any production it does not get persisted
  */
 type MemDatabase struct {
-	db map[string][]byte
+	db   map[string][]byte
+	lock sync.RWMutex
 }
 
 func NewMemDatabase() (*MemDatabase, error) {
-	db := &MemDatabase{db: make(map[string][]byte)}
-
-	return db, nil
+	return &MemDatabase{
+		db: make(map[string][]byte),
+	}, nil
 }
 
 func (db *MemDatabase) Put(key []byte, value []byte) error {
+	db.lock.Lock()
+	defer db.lock.Unlock()
+
 	db.db[string(key)] = common.CopyBytes(value)
 	return nil
 }
 
 func (db *MemDatabase) Set(key []byte, value []byte) {
+	db.lock.Lock()
+	defer db.lock.Unlock()
+
 	db.Put(key, value)
 }
 
 func (db *MemDatabase) Get(key []byte) ([]byte, error) {
-	return db.db[string(key)], nil
+	db.lock.RLock()
+	defer db.lock.RUnlock()
+
+	if entry, ok := db.db[string(key)]; ok {
+		return entry, nil
+	}
+	return nil, errors.New("not found")
 }
 
 func (db *MemDatabase) Keys() [][]byte {
+	db.lock.RLock()
+	defer db.lock.RUnlock()
+
 	keys := [][]byte{}
 	for key, _ := range db.db {
 		keys = append(keys, []byte(key))
@@ -65,12 +83,17 @@ func (db *MemDatabase) GetKeys() []*common.Key {
 */
 
 func (db *MemDatabase) Delete(key []byte) error {
-	delete(db.db, string(key))
+	db.lock.Lock()
+	defer db.lock.Unlock()
 
+	delete(db.db, string(key))
 	return nil
 }
 
 func (db *MemDatabase) Print() {
+	db.lock.RLock()
+	defer db.lock.RUnlock()
+
 	for key, val := range db.db {
 		fmt.Printf("%x(%d): ", key, len(key))
 		node := common.NewValueFromBytes(val)
@@ -83,11 +106,9 @@ func (db *MemDatabase) Close() {
 
 func (db *MemDatabase) LastKnownTD() []byte {
 	data, _ := db.Get([]byte("LastKnownTotalDifficulty"))
-
 	if len(data) == 0 || data == nil {
 		data = []byte{0x0}
 	}
-
 	return data
 }
 
@@ -100,16 +121,26 @@ type kv struct{ k, v []byte }
 type memBatch struct {
 	db     *MemDatabase
 	writes []kv
+	lock   sync.RWMutex
 }
 
-func (w *memBatch) Put(key, value []byte) error {
-	w.writes = append(w.writes, kv{key, common.CopyBytes(value)})
+func (b *memBatch) Put(key, value []byte) error {
+	b.lock.Lock()
+	defer b.lock.Unlock()
+
+	b.writes = append(b.writes, kv{key, common.CopyBytes(value)})
 	return nil
 }
 
-func (w *memBatch) Write() error {
-	for _, kv := range w.writes {
-		w.db.db[string(kv.k)] = kv.v
+func (b *memBatch) Write() error {
+	b.lock.RLock()
+	defer b.lock.RUnlock()
+
+	b.db.lock.Lock()
+	defer b.db.lock.Unlock()
+
+	for _, kv := range b.writes {
+		b.db.db[string(kv.k)] = kv.v
 	}
 	return nil
 }

miner/worker.go

@@ -312,7 +312,7 @@ func (self *worker) wait() {
 						self.mux.Post(core.ChainHeadEvent{block})
 						self.mux.Post(logs)
 					}
-					if err := core.PutBlockReceipts(self.chainDb, block, receipts); err != nil {
+					if err := core.PutBlockReceipts(self.chainDb, block.Hash(), receipts); err != nil {
 						glog.V(logger.Warn).Infoln("error writing block receipts:", err)
 					}
 				}(block, work.state.Logs(), work.receipts)

rpc/api/debug.go

@@ -146,13 +146,7 @@ func (self *debugApi) SetHead(req *shared.Request) (interface{}, error) {
 	if err := self.codec.Decode(req.Params, &args); err != nil {
 		return nil, shared.NewDecodeParamError(err.Error())
 	}
-
-	block := self.xeth.EthBlockByNumber(args.BlockNumber)
-	if block == nil {
-		return nil, fmt.Errorf("block #%d not found", args.BlockNumber)
-	}
-
-	self.ethereum.BlockChain().SetHead(block)
+	self.ethereum.BlockChain().SetHead(uint64(args.BlockNumber))
 
 	return nil, nil
 }

rpc/api/eth.go

@@ -168,9 +168,7 @@ func (self *ethApi) IsMining(req *shared.Request) (interface{}, error) {
 }
 
 func (self *ethApi) IsSyncing(req *shared.Request) (interface{}, error) {
-	current := self.ethereum.BlockChain().CurrentBlock().NumberU64()
-	origin, height := self.ethereum.Downloader().Boundaries()
-
+	origin, current, height := self.ethereum.Downloader().Progress()
 	if current < height {
 		return map[string]interface{}{
 			"startingBlock": newHexNum(big.NewInt(int64(origin)).Bytes()),

rpc/api/eth_args.go

@@ -838,7 +838,7 @@ func NewLogRes(log *vm.Log) LogRes {
 	}
 	l.Address = newHexData(log.Address)
 	l.Data = newHexData(log.Data)
-	l.BlockNumber = newHexNum(log.Number)
+	l.BlockNumber = newHexNum(log.BlockNumber)
 	l.LogIndex = newHexNum(log.Index)
 	l.TransactionHash = newHexData(log.TxHash)
 	l.TransactionIndex = newHexNum(log.TxIndex)

rpc/api/parsing.go

@@ -453,8 +453,8 @@ func NewReceiptRes(rec *types.Receipt) *ReceiptRes {
 		v.ContractAddress = newHexData(rec.ContractAddress)
 	}
 
-	logs := make([]interface{}, len(rec.Logs()))
-	for i, log := range rec.Logs() {
+	logs := make([]interface{}, len(rec.Logs))
+	for i, log := range rec.Logs {
 		logs[i] = NewLogRes(log)
 	}
 	v.Logs = &logs

trie/sync.go

@@ -0,0 +1,285 @@
+// Copyright 2015 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 (
+	"fmt"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/ethdb"
+	"gopkg.in/karalabe/cookiejar.v2/collections/prque"
+)
+
+// request represents a scheduled or already in-flight state retrieval request.
+type request struct {
+	hash   common.Hash // Hash of the node data content to retrieve
+	data   []byte      // Data content of the node, cached until all subtrees complete
+	object *node       // Target node to populate with retrieved data (hashnode originally)
+
+	parents []*request // Parent state nodes referencing this entry (notify all upon completion)
+	depth   int        // Depth level within the trie the node is located to prioritise DFS
+	deps    int        // Number of dependencies before allowed to commit this node
+
+	callback TrieSyncLeafCallback // Callback to invoke if a leaf node it reached on this branch
+}
+
+// SyncResult is a simple list to return missing nodes along with their request
+// hashes.
+type SyncResult struct {
+	Hash common.Hash // Hash of the originally unknown trie node
+	Data []byte      // Data content of the retrieved node
+}
+
+// TrieSyncLeafCallback is a callback type invoked when a trie sync reaches a
+// leaf node. It's used by state syncing to check if the leaf node requires some
+// further data syncing.
+type TrieSyncLeafCallback func(leaf []byte, parent common.Hash) error
+
+// TrieSync is the main state trie synchronisation scheduler, which provides yet
+// unknown trie hashes to retrieve, accepts node data associated with said hashes
+// and reconstructs the trie step by step until all is done.
+type TrieSync struct {
+	database ethdb.Database           // State database for storing all the assembled node data
+	requests map[common.Hash]*request // Pending requests pertaining to a key hash
+	queue    *prque.Prque             // Priority queue with the pending requests
+}
+
+// NewTrieSync creates a new trie data download scheduler.
+func NewTrieSync(root common.Hash, database ethdb.Database, callback TrieSyncLeafCallback) *TrieSync {
+	ts := &TrieSync{
+		database: database,
+		requests: make(map[common.Hash]*request),
+		queue:    prque.New(),
+	}
+	ts.AddSubTrie(root, 0, common.Hash{}, callback)
+	return ts
+}
+
+// AddSubTrie registers a new trie to the sync code, rooted at the designated parent.
+func (s *TrieSync) AddSubTrie(root common.Hash, depth int, parent common.Hash, callback TrieSyncLeafCallback) {
+	// Short circuit if the trie is empty or already known
+	if root == emptyRoot {
+		return
+	}
+	blob, _ := s.database.Get(root.Bytes())
+	if local, err := decodeNode(blob); local != nil && err == nil {
+		return
+	}
+	// Assemble the new sub-trie sync request
+	node := node(hashNode(root.Bytes()))
+	req := &request{
+		object:   &node,
+		hash:     root,
+		depth:    depth,
+		callback: callback,
+	}
+	// If this sub-trie has a designated parent, link them together
+	if parent != (common.Hash{}) {
+		ancestor := s.requests[parent]
+		if ancestor == nil {
+			panic(fmt.Sprintf("sub-trie ancestor not found: %x", parent))
+		}
+		ancestor.deps++
+		req.parents = append(req.parents, ancestor)
+	}
+	s.schedule(req)
+}
+
+// AddRawEntry schedules the direct retrieval of a state entry that should not be
+// interpreted as a trie node, but rather accepted and stored into the database
+// as is. This method's goal is to support misc state metadata retrievals (e.g.
+// contract code).
+func (s *TrieSync) AddRawEntry(hash common.Hash, depth int, parent common.Hash) {
+	// Short circuit if the entry is empty or already known
+	if hash == emptyState {
+		return
+	}
+	if blob, _ := s.database.Get(hash.Bytes()); blob != nil {
+		return
+	}
+	// Assemble the new sub-trie sync request
+	req := &request{
+		hash:  hash,
+		depth: depth,
+	}
+	// If this sub-trie has a designated parent, link them together
+	if parent != (common.Hash{}) {
+		ancestor := s.requests[parent]
+		if ancestor == nil {
+			panic(fmt.Sprintf("raw-entry ancestor not found: %x", parent))
+		}
+		ancestor.deps++
+		req.parents = append(req.parents, ancestor)
+	}
+	s.schedule(req)
+}
+
+// Missing retrieves the known missing nodes from the trie for retrieval.
+func (s *TrieSync) Missing(max int) []common.Hash {
+	requests := []common.Hash{}
+	for !s.queue.Empty() && (max == 0 || len(requests) < max) {
+		requests = append(requests, s.queue.PopItem().(common.Hash))
+	}
+	return requests
+}
+
+// Process injects a batch of retrieved trie nodes data.
+func (s *TrieSync) Process(results []SyncResult) (int, error) {
+	for i, item := range results {
+		// If the item was not requested, bail out
+		request := s.requests[item.Hash]
+		if request == nil {
+			return i, fmt.Errorf("not requested: %x", item.Hash)
+		}
+		// If the item is a raw entry request, commit directly
+		if request.object == nil {
+			request.data = item.Data
+			s.commit(request, nil)
+			continue
+		}
+		// Decode the node data content and update the request
+		node, err := decodeNode(item.Data)
+		if err != nil {
+			return i, err
+		}
+		*request.object = node
+		request.data = item.Data
+
+		// Create and schedule a request for all the children nodes
+		requests, err := s.children(request)
+		if err != nil {
+			return i, err
+		}
+		if len(requests) == 0 && request.deps == 0 {
+			s.commit(request, nil)
+			continue
+		}
+		request.deps += len(requests)
+		for _, child := range requests {
+			s.schedule(child)
+		}
+	}
+	return 0, nil
+}
+
+// Pending returns the number of state entries currently pending for download.
+func (s *TrieSync) Pending() int {
+	return len(s.requests)
+}
+
+// schedule inserts a new state retrieval request into the fetch queue. If there
+// is already a pending request for this node, the new request will be discarded
+// and only a parent reference added to the old one.
+func (s *TrieSync) schedule(req *request) {
+	// If we're already requesting this node, add a new reference and stop
+	if old, ok := s.requests[req.hash]; ok {
+		old.parents = append(old.parents, req.parents...)
+		return
+	}
+	// Schedule the request for future retrieval
+	s.queue.Push(req.hash, float32(req.depth))
+	s.requests[req.hash] = req
+}
+
+// children retrieves all the missing children of a state trie entry for future
+// retrieval scheduling.
+func (s *TrieSync) children(req *request) ([]*request, error) {
+	// Gather all the children of the node, irrelevant whether known or not
+	type child struct {
+		node  *node
+		depth int
+	}
+	children := []child{}
+
+	switch node := (*req.object).(type) {
+	case shortNode:
+		children = []child{{
+			node:  &node.Val,
+			depth: req.depth + len(node.Key),
+		}}
+	case fullNode:
+		for i := 0; i < 17; i++ {
+			if node[i] != nil {
+				children = append(children, child{
+					node:  &node[i],
+					depth: req.depth + 1,
+				})
+			}
+		}
+	default:
+		panic(fmt.Sprintf("unknown node: %+v", node))
+	}
+	// Iterate over the children, and request all unknown ones
+	requests := make([]*request, 0, len(children))
+	for _, child := range children {
+		// Notify any external watcher of a new key/value node
+		if req.callback != nil {
+			if node, ok := (*child.node).(valueNode); ok {
+				if err := req.callback(node, req.hash); err != nil {
+					return nil, err
+				}
+			}
+		}
+		// If the child references another node, resolve or schedule
+		if node, ok := (*child.node).(hashNode); ok {
+			// Try to resolve the node from the local database
+			blob, _ := s.database.Get(node)
+			if local, err := decodeNode(blob); local != nil && err == nil {
+				*child.node = local
+				continue
+			}
+			// Locally unknown node, schedule for retrieval
+			requests = append(requests, &request{
+				object:   child.node,
+				hash:     common.BytesToHash(node),
+				parents:  []*request{req},
+				depth:    child.depth,
+				callback: req.callback,
+			})
+		}
+	}
+	return requests, nil
+}
+
+// commit finalizes a retrieval request and stores it into the database. If any
+// of the referencing parent requests complete due to this commit, they are also
+// committed themselves.
+func (s *TrieSync) commit(req *request, batch ethdb.Batch) (err error) {
+	// Create a new batch if none was specified
+	if batch == nil {
+		batch = s.database.NewBatch()
+		defer func() {
+			err = batch.Write()
+		}()
+	}
+	// Write the node content to disk
+	if err := batch.Put(req.hash[:], req.data); err != nil {
+		return err
+	}
+	delete(s.requests, req.hash)
+
+	// Check all parents for completion
+	for _, parent := range req.parents {
+		parent.deps--
+		if parent.deps == 0 {
+			if err := s.commit(parent, batch); err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}

trie/sync_test.go

@@ -0,0 +1,257 @@
+// Copyright 2015 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 (
+	"bytes"
+	"testing"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/ethdb"
+)
+
+// makeTestTrie create a sample test trie to test node-wise reconstruction.
+func makeTestTrie() (ethdb.Database, *Trie, map[string][]byte) {
+	// Create an empty trie
+	db, _ := ethdb.NewMemDatabase()
+	trie, _ := New(common.Hash{}, db)
+
+	// Fill it with some arbitrary data
+	content := make(map[string][]byte)
+	for i := byte(0); i < 255; i++ {
+		key, val := common.LeftPadBytes([]byte{1, i}, 32), []byte{i}
+		content[string(key)] = val
+		trie.Update(key, val)
+
+		key, val = common.LeftPadBytes([]byte{2, i}, 32), []byte{i}
+		content[string(key)] = val
+		trie.Update(key, val)
+	}
+	trie.Commit()
+
+	// Return the generated trie
+	return db, trie, content
+}
+
+// checkTrieContents cross references a reconstructed trie with an expected data
+// content map.
+func checkTrieContents(t *testing.T, db Database, root []byte, content map[string][]byte) {
+	trie, err := New(common.BytesToHash(root), db)
+	if err != nil {
+		t.Fatalf("failed to create trie at %x: %v", root, err)
+	}
+	for key, val := range content {
+		if have := trie.Get([]byte(key)); bytes.Compare(have, val) != 0 {
+			t.Errorf("entry %x: content mismatch: have %x, want %x", key, have, val)
+		}
+	}
+}
+
+// Tests that an empty trie is not scheduled for syncing.
+func TestEmptyTrieSync(t *testing.T) {
+	emptyA, _ := New(common.Hash{}, nil)
+	emptyB, _ := New(emptyRoot, nil)
+
+	for i, trie := range []*Trie{emptyA, emptyB} {
+		db, _ := ethdb.NewMemDatabase()
+		if req := NewTrieSync(common.BytesToHash(trie.Root()), db, nil).Missing(1); len(req) != 0 {
+			t.Errorf("test %d: content requested for empty trie: %v", i, req)
+		}
+	}
+}
+
+// Tests that given a root hash, a trie can sync iteratively on a single thread,
+// requesting retrieval tasks and returning all of them in one go.
+func TestIterativeTrieSyncIndividual(t *testing.T) { testIterativeTrieSync(t, 1) }
+func TestIterativeTrieSyncBatched(t *testing.T)    { testIterativeTrieSync(t, 100) }
+
+func testIterativeTrieSync(t *testing.T, batch int) {
+	// Create a random trie to copy
+	srcDb, srcTrie, srcData := makeTestTrie()
+
+	// Create a destination trie and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewTrieSync(common.BytesToHash(srcTrie.Root()), dstDb, nil)
+
+	queue := append([]common.Hash{}, sched.Missing(batch)...)
+	for len(queue) > 0 {
+		results := make([]SyncResult, len(queue))
+		for i, hash := range queue {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results[i] = SyncResult{hash, data}
+		}
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = append(queue[:0], sched.Missing(batch)...)
+	}
+	// Cross check that the two tries re in sync
+	checkTrieContents(t, dstDb, srcTrie.Root(), srcData)
+}
+
+// Tests that the trie scheduler can correctly reconstruct the state even if only
+// partial results are returned, and the others sent only later.
+func TestIterativeDelayedTrieSync(t *testing.T) {
+	// Create a random trie to copy
+	srcDb, srcTrie, srcData := makeTestTrie()
+
+	// Create a destination trie and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewTrieSync(common.BytesToHash(srcTrie.Root()), dstDb, nil)
+
+	queue := append([]common.Hash{}, sched.Missing(10000)...)
+	for len(queue) > 0 {
+		// Sync only half of the scheduled nodes
+		results := make([]SyncResult, len(queue)/2+1)
+		for i, hash := range queue[:len(results)] {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results[i] = SyncResult{hash, data}
+		}
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = append(queue[len(results):], sched.Missing(10000)...)
+	}
+	// Cross check that the two tries re in sync
+	checkTrieContents(t, dstDb, srcTrie.Root(), srcData)
+}
+
+// Tests that given a root hash, a trie can sync iteratively on a single thread,
+// requesting retrieval tasks and returning all of them in one go, however in a
+// random order.
+func TestIterativeRandomTrieSyncIndividual(t *testing.T) { testIterativeRandomTrieSync(t, 1) }
+func TestIterativeRandomTrieSyncBatched(t *testing.T)    { testIterativeRandomTrieSync(t, 100) }
+
+func testIterativeRandomTrieSync(t *testing.T, batch int) {
+	// Create a random trie to copy
+	srcDb, srcTrie, srcData := makeTestTrie()
+
+	// Create a destination trie and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewTrieSync(common.BytesToHash(srcTrie.Root()), dstDb, nil)
+
+	queue := make(map[common.Hash]struct{})
+	for _, hash := range sched.Missing(batch) {
+		queue[hash] = struct{}{}
+	}
+	for len(queue) > 0 {
+		// Fetch all the queued nodes in a random order
+		results := make([]SyncResult, 0, len(queue))
+		for hash, _ := range queue {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results = append(results, SyncResult{hash, data})
+		}
+		// Feed the retrieved results back and queue new tasks
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = make(map[common.Hash]struct{})
+		for _, hash := range sched.Missing(batch) {
+			queue[hash] = struct{}{}
+		}
+	}
+	// Cross check that the two tries re in sync
+	checkTrieContents(t, dstDb, srcTrie.Root(), srcData)
+}
+
+// Tests that the trie scheduler can correctly reconstruct the state even if only
+// partial results are returned (Even those randomly), others sent only later.
+func TestIterativeRandomDelayedTrieSync(t *testing.T) {
+	// Create a random trie to copy
+	srcDb, srcTrie, srcData := makeTestTrie()
+
+	// Create a destination trie and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewTrieSync(common.BytesToHash(srcTrie.Root()), dstDb, nil)
+
+	queue := make(map[common.Hash]struct{})
+	for _, hash := range sched.Missing(10000) {
+		queue[hash] = struct{}{}
+	}
+	for len(queue) > 0 {
+		// Sync only half of the scheduled nodes, even those in random order
+		results := make([]SyncResult, 0, len(queue)/2+1)
+		for hash, _ := range queue {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			results = append(results, SyncResult{hash, data})
+
+			if len(results) >= cap(results) {
+				break
+			}
+		}
+		// Feed the retrieved results back and queue new tasks
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		for _, result := range results {
+			delete(queue, result.Hash)
+		}
+		for _, hash := range sched.Missing(10000) {
+			queue[hash] = struct{}{}
+		}
+	}
+	// Cross check that the two tries re in sync
+	checkTrieContents(t, dstDb, srcTrie.Root(), srcData)
+}
+
+// Tests that a trie sync will not request nodes multiple times, even if they
+// have such references.
+func TestDuplicateAvoidanceTrieSync(t *testing.T) {
+	// Create a random trie to copy
+	srcDb, srcTrie, srcData := makeTestTrie()
+
+	// Create a destination trie and sync with the scheduler
+	dstDb, _ := ethdb.NewMemDatabase()
+	sched := NewTrieSync(common.BytesToHash(srcTrie.Root()), dstDb, nil)
+
+	queue := append([]common.Hash{}, sched.Missing(0)...)
+	requested := make(map[common.Hash]struct{})
+
+	for len(queue) > 0 {
+		results := make([]SyncResult, len(queue))
+		for i, hash := range queue {
+			data, err := srcDb.Get(hash.Bytes())
+			if err != nil {
+				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
+			}
+			if _, ok := requested[hash]; ok {
+				t.Errorf("hash %x already requested once", hash)
+			}
+			requested[hash] = struct{}{}
+
+			results[i] = SyncResult{hash, data}
+		}
+		if index, err := sched.Process(results); err != nil {
+			t.Fatalf("failed to process result #%d: %v", index, err)
+		}
+		queue = append(queue[:0], sched.Missing(0)...)
+	}
+	// Cross check that the two tries re in sync
+	checkTrieContents(t, dstDb, srcTrie.Root(), srcData)
+}

trie/trie.go

@@ -24,6 +24,7 @@ import (
 	"hash"
 
 	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/crypto"
 	"github.com/ethereum/go-ethereum/crypto/sha3"
 	"github.com/ethereum/go-ethereum/logger"
 	"github.com/ethereum/go-ethereum/logger/glog"
@@ -35,8 +36,12 @@ const defaultCacheCapacity = 800
 var (
 	// The global cache stores decoded trie nodes by hash as they get loaded.
 	globalCache = newARC(defaultCacheCapacity)
+
 	// This is the known root hash of an empty trie.
 	emptyRoot = common.HexToHash("56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421")
+
+	// This is the known hash of an empty state trie entry.
+	emptyState = crypto.Sha3Hash(nil)
 )
 
 var ErrMissingRoot = errors.New("missing root node")