core,light: implement fast lookup of ancestor hashes

core/blockchain.go

@@ -1679,6 +1679,7 @@ func (bc *BlockChain) insertChain(chain types.Blocks, verifySeals bool) (int, []
 
 		switch status {
 		case CanonStatTy:
+			bc.hc.hashHistory.Set(block.Header())
 			log.Debug("Inserted new block", "number", block.Number(), "hash", block.Hash(),
 				"uncles", len(block.Uncles()), "txs", len(block.Transactions()), "gas", block.GasUsed(),
 				"elapsed", common.PrettyDuration(time.Since(start)),
@@ -1967,6 +1968,8 @@ func (bc *BlockChain) reorg(oldBlock, newBlock *types.Block) error {
 		// Write lookup entries for hash based transaction/receipt searches
 		rawdb.WriteTxLookupEntries(bc.db, newChain[i])
 		addedTxs = append(addedTxs, newChain[i].Transactions()...)
+		// Add to hash history
+		bc.hc.hashHistory.Set(newChain[i].Header())
 	}
 	// When transactions get deleted from the database, the receipts that were
 	// created in the fork must also be deleted
@@ -2224,3 +2227,8 @@ func (bc *BlockChain) SubscribeLogsEvent(ch chan<- []*types.Log) event.Subscript
 func (bc *BlockChain) SubscribeBlockProcessingEvent(ch chan<- bool) event.Subscription {
 	return bc.scope.Track(bc.blockProcFeed.Subscribe(ch))
 }
+
+//GetAncestorHash return the hash of ancestor at the given number
+func (bc *BlockChain) GetAncestorHash(ref *types.Header, target uint64) common.Hash {
+	return bc.hc.GetAncestorHash(ref, target)
+}

core/evm.go

@@ -31,8 +31,8 @@ type ChainContext interface {
 	// Engine retrieves the chain's consensus engine.
 	Engine() consensus.Engine
 
-	// GetHeader returns the hash corresponding to their hash.
-	GetHeader(common.Hash, uint64) *types.Header
+	// GetAncestorHash return the hash of ancestor at the given number
+	GetAncestorHash(child *types.Header, number uint64) common.Hash
 }
 
 // NewEVMContext creates a new context for use in the EVM.
@@ -60,27 +60,8 @@ func NewEVMContext(msg Message, header *types.Header, chain ChainContext, author
 
 // GetHashFn returns a GetHashFunc which retrieves header hashes by number
 func GetHashFn(ref *types.Header, chain ChainContext) func(n uint64) common.Hash {
-	var cache map[uint64]common.Hash
-
 	return func(n uint64) common.Hash {
-		// If there's no hash cache yet, make one
-		if cache == nil {
-			cache = map[uint64]common.Hash{
-				ref.Number.Uint64() - 1: ref.ParentHash,
-			}
-		}
-		// Try to fulfill the request from the cache
-		if hash, ok := cache[n]; ok {
-			return hash
-		}
-		// Not cached, iterate the blocks and cache the hashes
-		for header := chain.GetHeader(ref.ParentHash, ref.Number.Uint64()-1); header != nil; header = chain.GetHeader(header.ParentHash, header.Number.Uint64()-1) {
-			cache[header.Number.Uint64()-1] = header.ParentHash
-			if n == header.Number.Uint64()-1 {
-				return header.ParentHash
-			}
-		}
-		return common.Hash{}
+		return chain.GetAncestorHash(ref, n)
 	}
 }
 

core/hashbuffer.go

@@ -0,0 +1,173 @@
+// Copyright 2019 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package core
+
+import (
+	"sync"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/log"
+	"github.com/ethereum/go-ethereum/metrics"
+)
+
+// For EVM execution, we need around 256 items. We add a few more to allow reorgs.
+// For mainnet, a couple more would suffice, but a few more added for
+// testnets/private nets
+// For LES, we use a larger buffer. 500K * 32 bytes = 16M
+const hashBufferElems = 500000
+
+var (
+	hashHitCounter  = metrics.NewRegisteredGauge("chain/headerhash/hit", nil)
+	hashMissCounter = metrics.NewRegisteredGauge("chain/headerhash/miss", nil)
+	hashHeadGauge   = metrics.NewRegisteredGauge("chain/headerhash/head", nil)
+	hashTailGauge   = metrics.NewRegisteredGauge("chain/headerhash/tail", nil)
+)
+
+// hashBuffer implements a storage for chains of hashes, intended to be used for quick lookup of block hashes.
+// Internally, it uses an array of hashes in a circular buffer.
+// It enforces that all hashes added have a contiguous parent-child relation, and supports rollbacks
+// It is thread-safe.
+type hashBuffer struct {
+	// The data holds the hashes. The hashes are sequential, but also a
+	// circular buffer.
+	// The `head` points to the position of the latest hash.
+	// The parent, if present, is located 32 bytes back.
+	// [.., .., ..., head-2 , head-1, head, oldest, ... ]
+	data [hashBufferElems]common.Hash
+
+	head uint64 // index of hash for the head block
+
+	headNumber uint64 // The block number for head (the most recent block)
+	tailNumber uint64 // The block number for tail (the oldest block)
+
+	mu sync.RWMutex
+}
+
+// newHashBuffer creates a new storage with a header in it.
+// Since we take a header here, a hash storage can never be empty.
+// This makes things easier later on (in Set)
+func newHashBuffer(header *types.Header) *hashBuffer {
+	return &hashBuffer{
+		headNumber: header.Number.Uint64(),
+		tailNumber: header.Number.Uint64(),
+		data:       [hashBufferElems]common.Hash{header.Hash()},
+	}
+}
+
+// Get locates the hash for the requested number
+func (hs *hashBuffer) Get(number uint64) (common.Hash, bool) {
+	hs.mu.RLock()
+	defer hs.mu.RUnlock()
+	if !hs.has(number) {
+		hashMissCounter.Inc(1)
+		return common.Hash{}, false
+	}
+	hashHitCounter.Inc(1)
+	distance := hs.headNumber - number
+	index := (hs.head + hashBufferElems - distance) % hashBufferElems
+	return hs.data[index], true
+}
+
+// has returns if the storage has a hash for the given number
+func (hs *hashBuffer) has(number uint64) bool {
+	return number <= hs.headNumber && number >= hs.tailNumber
+}
+
+// Contains checks if the hash at the given number matches the expected
+func (hs *hashBuffer) Contains(number uint64, expected common.Hash) bool {
+	hs.mu.RLock()
+	defer hs.mu.RUnlock()
+	if hs.contains(number, expected) {
+		hashHitCounter.Inc(1)
+		return true
+	} else {
+		hashMissCounter.Inc(1)
+		return false
+	}
+}
+
+// contains is the non-concurrency safe internal version of Contains
+func (hs *hashBuffer) contains(number uint64, expected common.Hash) bool {
+	if !hs.has(number) {
+		return false
+	}
+	distance := hs.headNumber - number
+	index := (hs.head + hashBufferElems - distance) % hashBufferElems
+	return hs.data[index] == expected
+}
+
+// Newest returns the most recent (number, hash) stored
+func (hs *hashBuffer) Newest() (uint64, common.Hash) {
+	hs.mu.RLock()
+	defer hs.mu.RUnlock()
+	return hs.headNumber, hs.data[hs.head]
+}
+
+// Oldest returns the oldest (number, hash) found
+func (hs *hashBuffer) Oldest() (uint64, common.Hash) {
+	hs.mu.RLock()
+	defer hs.mu.RUnlock()
+	distance := hs.headNumber - hs.tailNumber
+	index := (hs.head + hashBufferElems - distance) % hashBufferElems
+	return hs.tailNumber, hs.data[index]
+}
+
+// Set inserts a new header (hash) to the storage.
+// If
+// a) Header already exists, this is a no-op
+// b) Number is occupied by other header, the new header replaces it, and also
+// truncates any descendants
+//
+// If the new header does not have any ancestors, it replaces the entire storage.
+func (hs *hashBuffer) Set(header *types.Header) {
+	var (
+		number = header.Number.Uint64()
+		index  uint64
+		hash   = header.Hash()
+	)
+	hs.mu.Lock()
+	defer hs.mu.Unlock()
+	if hs.contains(number-1, header.ParentHash) {
+		if hs.headNumber >= number {
+			distance := hs.headNumber - number
+			index = (hs.head + hashBufferElems - distance) % hashBufferElems
+			if hs.data[index] == hash {
+				return
+			}
+			// Continue by replacing this number and wipe descendants
+		} else {
+			// head is parent of this new header - regular append
+			index = (hs.head + 1) % hashBufferElems
+		}
+	} else {
+		// This should not normally happen, and indicates a programming error
+		log.Error("Hash storage wiping ancestors", "oldhead", hs.headNumber, "newhead", number, "oldtail", hs.tailNumber)
+		// Wipe ancestors
+		hs.tailNumber = number
+	}
+	hs.head = index
+	hs.headNumber = number
+	hs.data[hs.head] = hash
+
+	if number-hs.tailNumber == hashBufferElems {
+		// It's full, need to move the tail
+		hs.tailNumber++
+	}
+	hashTailGauge.Update(int64(hs.tailNumber))
+	hashHeadGauge.Update(int64(hs.headNumber))
+}

core/hashbuffer_test.go

@@ -0,0 +1,204 @@
+// Copyright 2019 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package core
+
+import (
+	"bytes"
+	"math/big"
+	"math/rand"
+	"testing"
+	"time"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core/types"
+)
+
+func init() {
+	rand.Seed(time.Now().Unix())
+}
+func dummyHeader(n int, prev common.Hash) *types.Header {
+	return &types.Header{
+		Number:     new(big.Int).SetUint64(uint64(n)),
+		ParentHash: prev,
+	}
+}
+
+func uniqueHeader(n int, prev common.Hash) *types.Header {
+	return &types.Header{
+		Number:     new(big.Int).SetUint64(uint64(n)),
+		ParentHash: prev,
+		GasUsed:    rand.Uint64(),
+	}
+}
+
+// TestConsecutiveHashes does inserts and rollback, but uses contiguous chains
+func TestConsecutiveHashes(t *testing.T) {
+	t.Parallel()
+	// This should we swapped out very quickly
+	hs := newHashBuffer(uniqueHeader(0, common.Hash{0xaa}))
+
+	parent := common.Hash{}
+	expected := make(map[int]common.Hash)
+
+	assertEmpty := func(num int) {
+		h, found := hs.Get(uint64(num))
+		if found {
+			t.Fatalf("expected %d not to be present", num)
+		}
+		if h != (common.Hash{}) {
+			t.Fatalf("expected empty hash, got %x", h)
+		}
+
+	}
+
+	// test 10 entries
+	for n := 1; n < 10; n++ {
+		h := dummyHeader(n, parent)
+		hs.Set(h)
+		if _, lh := hs.Newest(); lh != h.Hash() {
+			t.Fatalf("num %d, wrong last hash, got %x exp %x", n, lh, h.Hash())
+		}
+		expected[n] = h.Hash()
+		parent = h.Hash()
+	}
+
+	n, _ := hs.Oldest()
+	if n != 1 {
+		t.Fatalf("wrong oldest, expected %d got %d", 1, n)
+	}
+
+	for n := 1; n < 10; n++ {
+		got, _ := hs.Get(uint64(n))
+		exp := expected[n]
+		if got != exp {
+			t.Errorf("num %d, got %x expected %x", n, got, exp)
+		}
+	}
+	assertEmpty(11)
+	assertEmpty(0)
+
+	// Write another 300, overflowing the storage
+	for n := 10; n < hashBufferElems+10; n++ {
+		h := dummyHeader(n, parent)
+		hs.Set(h)
+		if _, lh := hs.Newest(); lh != h.Hash() {
+			t.Fatalf("num %d, wrong last hash, got %x exp %x", n, lh, h.Hash())
+		}
+		expected[n] = h.Hash()
+		parent = h.Hash()
+	}
+	x, _ := hs.Oldest()
+	if x != 10 {
+		t.Fatalf("wrong oldest, expected %d got %d", 10, x)
+	}
+
+	// The last 256 should be available
+	for n := hashBufferElems + 10 - 1; n > 10; n-- {
+		got, found := hs.Get(uint64(n))
+		exp := expected[n]
+		if !found {
+			t.Fatalf("expected %d to be found", n)
+		}
+		if got != exp {
+			t.Fatalf("num %d, got %x expected %x", n, got, exp)
+		}
+	}
+	// The older ones should be flushed
+	for ; n > 0; n-- {
+		got, found := hs.Get(n)
+		if found {
+			t.Fatalf("expected %d to be flushed", n)
+		}
+		if got != (common.Hash{}) {
+			t.Fatalf("expected empty hash, got %x", got)
+		}
+	}
+}
+
+func TestHashStorageNonContiguous(t *testing.T) {
+
+	hs := newHashBuffer(uniqueHeader(0, common.Hash{0xff}))
+	parent := common.Hash{}
+	expected := make(map[int]common.Hash)
+
+	for n := 1; n < 10; n++ {
+		hs.Set(uniqueHeader(n, parent))
+		hdr := uniqueHeader(n, parent)
+		hs.Set(hdr)
+		parent = hdr.Hash()
+		expected[n] = hdr.Hash()
+	}
+	n, _ := hs.Oldest()
+	if n != 1 {
+		t.Fatalf("wrong oldest, expected %d got %d", 1, n)
+	}
+	for n := 1; n < 10; n++ {
+		got, _ := hs.Get(uint64(n))
+		exp := expected[n]
+		if got != exp {
+			t.Errorf("num %d, got %x expected %x", n, got, exp)
+		}
+	}
+	// 9 headers there [ 1,2,3,4a,5,6,7,8,9]
+	// Setting a new in the middle should change it to
+	// [ 1, 2, 3, 4b]
+	{
+		parent = expected[4]
+		hdr := uniqueHeader(5, parent)
+		hs.Set(hdr)
+		if hs.headNumber != 5 {
+			t.Fatalf("expected head num 5, got %d", hs.headNumber)
+		}
+		got, found := hs.Get(5)
+		if !found {
+			t.Fatalf("expected hash to exist")
+		}
+		if !bytes.Equal(got[:], hdr.Hash().Bytes()) {
+			t.Fatalf("expected %x, got %x", hdr.Hash(), got)
+		}
+	}
+
+	// Set a totally new header at 3, should clean out everything else
+	{
+		hdr := uniqueHeader(4, common.Hash{0x1})
+		hs.Set(hdr)
+		if hs.headNumber != 4 {
+			t.Fatalf("expected head num 4, got %d", hs.headNumber)
+		}
+		if hs.tailNumber != 4 {
+			t.Fatalf("expected head num 4, got %d", hs.tailNumber)
+		}
+		if _, exist := hs.Get(3); exist {
+			t.Fatalf("should be gone: %d", 3)
+		}
+		if _, exist := hs.Get(5); exist {
+			t.Fatalf("should be gone: %d", 5)
+		}
+		if _, exist := hs.Get(0); exist {
+			t.Fatalf("should be gone: %d", 0)
+		}
+		// 4 should be there
+		got, found := hs.Get(4)
+		if !found {
+			t.Fatalf("expected hash to exist")
+		}
+		if !bytes.Equal(got[:], hdr.Hash().Bytes()) {
+			t.Fatalf("expected %x, got %x", hdr.Hash(), got)
+		}
+	}
+
+}

core/headerchain.go

@@ -60,6 +60,8 @@ type HeaderChain struct {
 	tdCache     *lru.Cache // Cache for the most recent block total difficulties
 	numberCache *lru.Cache // Cache for the most recent block numbers
 
+	hashHistory *hashBuffer // Cache for recent hashes
+
 	procInterrupt func() bool
 
 	rand   *mrand.Rand
@@ -96,6 +98,7 @@ func NewHeaderChain(chainDb ethdb.Database, config *params.ChainConfig, engine c
 	if hc.genesisHeader == nil {
 		return nil, ErrNoGenesis
 	}
+	hc.hashHistory = newHashBuffer(hc.genesisHeader)
 
 	hc.currentHeader.Store(hc.genesisHeader)
 	if head := rawdb.ReadHeadBlockHash(chainDb); head != (common.Hash{}) {
@@ -287,6 +290,7 @@ func (hc *HeaderChain) InsertHeaderChain(chain []*types.Header, writeHeader WhCa
 		if err := writeHeader(header); err != nil {
 			return i, err
 		}
+		hc.hashHistory.Set(header)
 		stats.processed++
 	}
 	// Report some public statistics so the user has a clue what's going on
@@ -339,35 +343,19 @@ func (hc *HeaderChain) GetAncestor(hash common.Hash, number, ancestor uint64, ma
 	if ancestor > number {
 		return common.Hash{}, 0
 	}
-	if ancestor == 1 {
-		// in this case it is cheaper to just read the header
-		if header := hc.GetHeader(hash, number); header != nil {
-			return header.ParentHash, number - 1
-		} else {
-			return common.Hash{}, 0
-		}
+	ref := hc.GetHeader(hash, number)
+	if ref == nil {
+		return common.Hash{}, 0
 	}
-	for ancestor != 0 {
-		if rawdb.ReadCanonicalHash(hc.chainDb, number) == hash {
-			ancestorHash := rawdb.ReadCanonicalHash(hc.chainDb, number-ancestor)
-			if rawdb.ReadCanonicalHash(hc.chainDb, number) == hash {
-				number -= ancestor
-				return ancestorHash, number
-			}
-		}
-		if *maxNonCanonical == 0 {
-			return common.Hash{}, 0
-		}
-		*maxNonCanonical--
-		ancestor--
-		header := hc.GetHeader(hash, number)
-		if header == nil {
-			return common.Hash{}, 0
-		}
-		hash = header.ParentHash
-		number--
+	if number == 0 {
+		return ref.Hash(), number
 	}
-	return hash, number
+	target := number - ancestor
+	ancestorHash := hc.GetAncestorHash(ref, target)
+	if ancestorHash == (common.Hash{}) {
+		return common.Hash{}, 0
+	}
+	return ancestorHash, target
 }
 
 // GetTd retrieves a block's total difficulty in the canonical chain from the
@@ -464,6 +452,7 @@ func (hc *HeaderChain) SetCurrentHeader(head *types.Header) {
 
 	hc.currentHeader.Store(head)
 	hc.currentHeaderHash = head.Hash()
+	hc.hashHistory.Set(head)
 	headHeaderGauge.Update(head.Number.Int64())
 }
 
@@ -543,3 +532,49 @@ func (hc *HeaderChain) Engine() consensus.Engine { return hc.engine }
 func (hc *HeaderChain) GetBlock(hash common.Hash, number uint64) *types.Block {
 	return nil
 }
+
+// GetAncestorHash return the hash of ancestor at the given number
+func (hc *HeaderChain) GetAncestorHash(ref *types.Header, target uint64) common.Hash {
+	number := ref.Number.Uint64() - 1
+	hash := ref.ParentHash
+	if target == number {
+		return hash
+	}
+	if target > number {
+		// Should never happen
+		log.Error("Ancestor number must be <= descendant", "target", target, "descendant", number)
+		return common.Hash{}
+	}
+	var (
+		maxNonCanonLookups = uint64(100)
+	)
+	if hashHistoryTail, _ := hc.hashHistory.Oldest(); hashHistoryTail <= target {
+		// Iterate the chain until we hit the target or we hit a ancestor
+		// within the storage
+		for ; !hc.hashHistory.Contains(number, hash); maxNonCanonLookups-- {
+			if maxNonCanonLookups == 0 {
+				return common.Hash{}
+			}
+			header := hc.GetHeader(hash, number)
+			if header == nil {
+				return common.Hash{}
+			}
+			number, hash = header.Number.Uint64()-1, header.ParentHash
+			if number == target {
+				return hash
+			}
+		}
+		// The hash storage has the right ancestor chain
+		if h, ok := hc.hashHistory.Get(target); ok {
+			return h
+		}
+	}
+	// At this point, we have failed to find the ancestor in our hash history lookup.
+	// Either it's some very long sidefork, or the 'ref' is a very old header, too
+	// old to be in the history. In that case, it should exist in the canon chain.
+	if rawdb.ReadCanonicalHash(hc.chainDb, number) == hash {
+		return rawdb.ReadCanonicalHash(hc.chainDb, target)
+	}
+	// They are requesting a very old header which is not in the canon chain,
+	return common.Hash{}
+}

eth/handler_test.go

@@ -179,18 +179,18 @@ func testGetBlockHeaders(t *testing.T, protocol int) {
 			headers = append(headers, pm.blockchain.GetBlockByHash(hash).Header())
 		}
 		// Send the hash request and verify the response
-		p2p.Send(peer.app, 0x03, tt.query)
-		if err := p2p.ExpectMsg(peer.app, 0x04, headers); err != nil {
-			t.Errorf("test %d: headers mismatch: %v", i, err)
+		p2p.Send(peer.app, GetBlockHeadersMsg, tt.query)
+		if err := p2p.ExpectMsg(peer.app, BlockHeadersMsg, headers); err != nil {
+			t.Fatalf("test %d: headers mismatch: %v", i, err)
 		}
 		// If the test used number origins, repeat with hashes as the too
 		if tt.query.Origin.Hash == (common.Hash{}) {
 			if origin := pm.blockchain.GetBlockByNumber(tt.query.Origin.Number); origin != nil {
 				tt.query.Origin.Hash, tt.query.Origin.Number = origin.Hash(), 0
 
-				p2p.Send(peer.app, 0x03, tt.query)
-				if err := p2p.ExpectMsg(peer.app, 0x04, headers); err != nil {
-					t.Errorf("test %d: headers mismatch: %v", i, err)
+				p2p.Send(peer.app, GetBlockHeadersMsg, tt.query)
+				if err := p2p.ExpectMsg(peer.app, BlockHeadersMsg, headers); err != nil {
+					t.Fatalf("test %d: headers mismatch: %v", i, err)
 				}
 			}
 		}

light/lightchain.go

@@ -542,3 +542,12 @@ func (lc *LightChain) DisableCheckFreq() {
 func (lc *LightChain) EnableCheckFreq() {
 	atomic.StoreInt32(&lc.disableCheckFreq, 0)
 }
+
+// GetAncestorHash return the hash of ancestor at the given number
+// This implementation is slow, since it does not use a hashstorage
+// lookup like blockchain. However, the light clients does not require
+// fast lookups for EVM execution. This should probably be improved if this
+// method becomes more used.
+func (lc *LightChain) GetAncestorHash(ref *types.Header, target uint64) common.Hash {
+	return lc.hc.GetAncestorHash(ref, target)
+}