trie: use stacktrie for Derivesha operation (#21407)

core/types: use stacktrie for derivesha

trie: add stacktrie file

trie: fix linter

core/types: use stacktrie for derivesha

rebased: adapt stacktrie to the newer version of DeriveSha
Co-authored-by: Martin Holst Swende <martin@swende.se>

More linter fixes

review feedback: no key offset for nodes converted to hashes

trie: use EncodeRLP for full nodes

core/types: insert txs in order in derivesha

trie: tests for derivesha with stacktrie

trie: make stacktrie use pooled hashers

trie: make stacktrie reuse tmp slice space

trie: minor polishes on stacktrie

trie/stacktrie: less rlp dancing

core/types: explain the contorsions in DeriveSha

ci: fix goimport errors

trie: clear mem on subtrie hashing

squashme: linter fix

stracktrie: use pooling, less allocs (#3)

trie: in-place hex prefix, reduce allocs and add rawNode.EncodeRLP

Reintroduce the `[]node` method, add the missing `EncodeRLP` implementation for `rawNode` and calculate the hex prefix in place.
Co-authored-by: Martin Holst Swende <martin@swende.se>
Co-authored-by: Martin Holst Swende <martin@swende.se>

core/block_validator.go

@@ -62,7 +62,7 @@ func (v *BlockValidator) ValidateBody(block *types.Block) error {
 	if hash := types.CalcUncleHash(block.Uncles()); hash != header.UncleHash {
 		return fmt.Errorf("uncle root hash mismatch: have %x, want %x", hash, header.UncleHash)
 	}
-	if hash := types.DeriveSha(block.Transactions(), new(trie.Trie)); hash != header.TxHash {
+	if hash := types.DeriveSha(block.Transactions(), trie.NewStackTrie(nil)); hash != header.TxHash {
 		return fmt.Errorf("transaction root hash mismatch: have %x, want %x", hash, header.TxHash)
 	}
 	if !v.bc.HasBlockAndState(block.ParentHash(), block.NumberU64()-1) {
@@ -90,7 +90,7 @@ func (v *BlockValidator) ValidateState(block *types.Block, statedb *state.StateD
 		return fmt.Errorf("invalid bloom (remote: %x  local: %x)", header.Bloom, rbloom)
 	}
 	// Tre receipt Trie's root (R = (Tr [[H1, R1], ... [Hn, Rn]]))
-	receiptSha := types.DeriveSha(receipts, new(trie.Trie))
+	receiptSha := types.DeriveSha(receipts, trie.NewStackTrie(nil))
 	if receiptSha != header.ReceiptHash {
 		return fmt.Errorf("invalid receipt root hash (remote: %x local: %x)", header.ReceiptHash, receiptSha)
 	}

core/types/derive_sha.go

@@ -23,7 +23,6 @@ import (
 	"github.com/ethereum/go-ethereum/rlp"
 )
 
-// DerivableList is the interface which can derive the hash.
 type DerivableList interface {
 	Len() int
 	GetRlp(i int) []byte
@@ -39,7 +38,22 @@ type Hasher interface {
 func DeriveSha(list DerivableList, hasher Hasher) common.Hash {
 	hasher.Reset()
 	keybuf := new(bytes.Buffer)
-	for i := 0; i < list.Len(); i++ {
+
+	// StackTrie requires values to be inserted in increasing
+	// hash order, which is not the order that `list` provides
+	// hashes in. This insertion sequence ensures that the
+	// order is correct.
+	for i := 1; i < list.Len() && i <= 0x7f; i++ {
+		keybuf.Reset()
+		rlp.Encode(keybuf, uint(i))
+		hasher.Update(keybuf.Bytes(), list.GetRlp(i))
+	}
+	if list.Len() > 0 {
+		keybuf.Reset()
+		rlp.Encode(keybuf, uint(0))
+		hasher.Update(keybuf.Bytes(), list.GetRlp(0))
+	}
+	for i := 0x80; i < list.Len(); i++ {
 		keybuf.Reset()
 		rlp.Encode(keybuf, uint(i))
 		hasher.Update(keybuf.Bytes(), list.GetRlp(i))

eth/downloader/queue.go

@@ -774,7 +774,7 @@ func (q *queue) DeliverBodies(id string, txLists [][]*types.Transaction, uncleLi
 	q.lock.Lock()
 	defer q.lock.Unlock()
 	validate := func(index int, header *types.Header) error {
-		if types.DeriveSha(types.Transactions(txLists[index]), new(trie.Trie)) != header.TxHash {
+		if types.DeriveSha(types.Transactions(txLists[index]), trie.NewStackTrie(nil)) != header.TxHash {
 			return errInvalidBody
 		}
 		if types.CalcUncleHash(uncleLists[index]) != header.UncleHash {
@@ -799,7 +799,7 @@ func (q *queue) DeliverReceipts(id string, receiptList [][]*types.Receipt) (int,
 	q.lock.Lock()
 	defer q.lock.Unlock()
 	validate := func(index int, header *types.Header) error {
-		if types.DeriveSha(types.Receipts(receiptList[index]), new(trie.Trie)) != header.ReceiptHash {
+		if types.DeriveSha(types.Receipts(receiptList[index]), trie.NewStackTrie(nil)) != header.ReceiptHash {
 			return errInvalidReceipt
 		}
 		return nil

eth/handler.go

@@ -715,7 +715,7 @@ func (pm *ProtocolManager) handleMsg(p *peer) error {
 			log.Warn("Propagated block has invalid uncles", "have", hash, "exp", request.Block.UncleHash())
 			break // TODO(karalabe): return error eventually, but wait a few releases
 		}
-		if hash := types.DeriveSha(request.Block.Transactions(), new(trie.Trie)); hash != request.Block.TxHash() {
+		if hash := types.DeriveSha(request.Block.Transactions(), trie.NewStackTrie(nil)); hash != request.Block.TxHash() {
 			log.Warn("Propagated block has invalid body", "have", hash, "exp", request.Block.TxHash())
 			break // TODO(karalabe): return error eventually, but wait a few releases
 		}

trie/database.go

@@ -99,6 +99,11 @@ type rawNode []byte
 func (n rawNode) cache() (hashNode, bool)   { panic("this should never end up in a live trie") }
 func (n rawNode) fstring(ind string) string { panic("this should never end up in a live trie") }
 
+func (n rawNode) EncodeRLP(w io.Writer) error {
+	_, err := w.Write([]byte(n))
+	return err
+}
+
 // rawFullNode represents only the useful data content of a full node, with the
 // caches and flags stripped out to minimize its data storage. This type honors
 // the same RLP encoding as the original parent.
@@ -199,7 +204,7 @@ func forGatherChildren(n node, onChild func(hash common.Hash)) {
 		}
 	case hashNode:
 		onChild(common.BytesToHash(n))
-	case valueNode, nil:
+	case valueNode, nil, rawNode:
 	default:
 		panic(fmt.Sprintf("unknown node type: %T", n))
 	}

trie/encoding.go

@@ -51,6 +51,35 @@ func hexToCompact(hex []byte) []byte {
 	return buf
 }
 
+// hexToCompactInPlace places the compact key in input buffer, returning the length
+// needed for the representation
+func hexToCompactInPlace(hex []byte) int {
+	var (
+		hexLen    = len(hex) // length of the hex input
+		firstByte = byte(0)
+	)
+	// Check if we have a terminator there
+	if hexLen > 0 && hex[hexLen-1] == 16 {
+		firstByte = 1 << 5
+		hexLen-- // last part was the terminator, ignore that
+	}
+	var (
+		binLen = hexLen/2 + 1
+		ni     = 0 // index in hex
+		bi     = 1 // index in bin (compact)
+	)
+	if hexLen&1 == 1 {
+		firstByte |= 1 << 4 // odd flag
+		firstByte |= hex[0] // first nibble is contained in the first byte
+		ni++
+	}
+	for ; ni < hexLen; bi, ni = bi+1, ni+2 {
+		hex[bi] = hex[ni]<<4 | hex[ni+1]
+	}
+	hex[0] = firstByte
+	return binLen
+}
+
 func compactToHex(compact []byte) []byte {
 	if len(compact) == 0 {
 		return compact

trie/encoding_test.go

@@ -18,6 +18,8 @@ package trie
 
 import (
 	"bytes"
+	"encoding/hex"
+	"math/rand"
 	"testing"
 )
 
@@ -75,6 +77,40 @@ func TestHexKeybytes(t *testing.T) {
 	}
 }
 
+func TestHexToCompactInPlace(t *testing.T) {
+	for i, keyS := range []string{
+		"00",
+		"060a040c0f000a090b040803010801010900080d090a0a0d0903000b10",
+		"10",
+	} {
+		hexBytes, _ := hex.DecodeString(keyS)
+		exp := hexToCompact(hexBytes)
+		sz := hexToCompactInPlace(hexBytes)
+		got := hexBytes[:sz]
+		if !bytes.Equal(exp, got) {
+			t.Fatalf("test %d: encoding err\ninp %v\ngot %x\nexp %x\n", i, keyS, got, exp)
+		}
+	}
+}
+
+func TestHexToCompactInPlaceRandom(t *testing.T) {
+	for i := 0; i < 10000; i++ {
+		l := rand.Intn(128)
+		key := make([]byte, l)
+		rand.Read(key)
+		hexBytes := keybytesToHex(key)
+		hexOrig := []byte(string(hexBytes))
+		exp := hexToCompact(hexBytes)
+		sz := hexToCompactInPlace(hexBytes)
+		got := hexBytes[:sz]
+
+		if !bytes.Equal(exp, got) {
+			t.Fatalf("encoding err \ncpt %x\nhex %x\ngot %x\nexp %x\n",
+				key, hexOrig, got, exp)
+		}
+	}
+}
+
 func BenchmarkHexToCompact(b *testing.B) {
 	testBytes := []byte{0, 15, 1, 12, 11, 8, 16 /*term*/}
 	for i := 0; i < b.N; i++ {

trie/stacktrie_test.go

+package trie
+
+import (
+	"bytes"
+	"fmt"
+	"math/big"
+	mrand "math/rand"
+	"testing"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/common/hexutil"
+	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/crypto"
+	"github.com/ethereum/go-ethereum/ethdb/memorydb"
+)
+
+func TestSizeBug(t *testing.T) {
+	st := NewStackTrie(nil)
+	nt, _ := New(common.Hash{}, NewDatabase(memorydb.New()))
+
+	leaf := common.FromHex("290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
+	value := common.FromHex("94cf40d0d2b44f2b66e07cace1372ca42b73cf21a3")
+
+	nt.TryUpdate(leaf, value)
+	st.TryUpdate(leaf, value)
+
+	if nt.Hash() != st.Hash() {
+		t.Fatalf("error %x != %x", st.Hash(), nt.Hash())
+	}
+}
+
+func TestEmptyBug(t *testing.T) {
+	st := NewStackTrie(nil)
+	nt, _ := New(common.Hash{}, NewDatabase(memorydb.New()))
+
+	//leaf := common.FromHex("290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
+	//value := common.FromHex("94cf40d0d2b44f2b66e07cace1372ca42b73cf21a3")
+	kvs := []struct {
+		K string
+		V string
+	}{
+		{K: "405787fa12a823e0f2b7631cc41b3ba8828b3321ca811111fa75cd3aa3bb5ace", V: "9496f4ec2bf9dab484cac6be589e8417d84781be08"},
+		{K: "40edb63a35fcf86c08022722aa3287cdd36440d671b4918131b2514795fefa9c", V: "01"},
+		{K: "b10e2d527612073b26eecdfd717e6a320cf44b4afac2b0732d9fcbe2b7fa0cf6", V: "947a30f7736e48d6599356464ba4c150d8da0302ff"},
+		{K: "c2575a0e9e593c00f959f8c92f12db2869c3395a3b0502d05e2516446f71f85b", V: "02"},
+	}
+
+	for _, kv := range kvs {
+		nt.TryUpdate(common.FromHex(kv.K), common.FromHex(kv.V))
+		st.TryUpdate(common.FromHex(kv.K), common.FromHex(kv.V))
+	}
+
+	if nt.Hash() != st.Hash() {
+		t.Fatalf("error %x != %x", st.Hash(), nt.Hash())
+	}
+}
+
+func TestValLength56(t *testing.T) {
+	st := NewStackTrie(nil)
+	nt, _ := New(common.Hash{}, NewDatabase(memorydb.New()))
+
+	//leaf := common.FromHex("290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
+	//value := common.FromHex("94cf40d0d2b44f2b66e07cace1372ca42b73cf21a3")
+	kvs := []struct {
+		K string
+		V string
+	}{
+		{K: "405787fa12a823e0f2b7631cc41b3ba8828b3321ca811111fa75cd3aa3bb5ace", V: "1111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111"},
+	}
+
+	for _, kv := range kvs {
+		nt.TryUpdate(common.FromHex(kv.K), common.FromHex(kv.V))
+		st.TryUpdate(common.FromHex(kv.K), common.FromHex(kv.V))
+	}
+
+	if nt.Hash() != st.Hash() {
+		t.Fatalf("error %x != %x", st.Hash(), nt.Hash())
+	}
+}
+
+func genTxs(num uint64) (types.Transactions, error) {
+	key, err := crypto.HexToECDSA("deadbeefdeadbeefdeadbeefdeadbeefdeadbeefdeadbeefdeadbeefdeadbeef")
+	if err != nil {
+		return nil, err
+	}
+	var addr = crypto.PubkeyToAddress(key.PublicKey)
+	newTx := func(i uint64) (*types.Transaction, error) {
+		signer := types.NewEIP155Signer(big.NewInt(18))
+		tx, err := types.SignTx(types.NewTransaction(i, addr, new(big.Int), 0, new(big.Int).SetUint64(10000000), nil), signer, key)
+		return tx, err
+	}
+	var txs types.Transactions
+	for i := uint64(0); i < num; i++ {
+		tx, err := newTx(i)
+		if err != nil {
+			return nil, err
+		}
+		txs = append(txs, tx)
+	}
+	return txs, nil
+}
+
+func TestDeriveSha(t *testing.T) {
+	txs, err := genTxs(0)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for len(txs) < 1000 {
+		exp := types.DeriveSha(txs, newEmpty())
+		got := types.DeriveSha(txs, NewStackTrie(nil))
+		if !bytes.Equal(got[:], exp[:]) {
+			t.Fatalf("%d txs: got %x exp %x", len(txs), got, exp)
+		}
+		newTxs, err := genTxs(uint64(len(txs) + 1))
+		if err != nil {
+			t.Fatal(err)
+		}
+		txs = append(txs, newTxs...)
+	}
+}
+
+func BenchmarkDeriveSha200(b *testing.B) {
+	txs, err := genTxs(200)
+	if err != nil {
+		b.Fatal(err)
+	}
+	var exp common.Hash
+	var got common.Hash
+	b.Run("std_trie", func(b *testing.B) {
+		b.ResetTimer()
+		b.ReportAllocs()
+		for i := 0; i < b.N; i++ {
+			exp = types.DeriveSha(txs, newEmpty())
+		}
+	})
+
+	b.Run("stack_trie", func(b *testing.B) {
+		b.ResetTimer()
+		b.ReportAllocs()
+		for i := 0; i < b.N; i++ {
+			got = types.DeriveSha(txs, NewStackTrie(nil))
+		}
+	})
+	if got != exp {
+		b.Errorf("got %x exp %x", got, exp)
+	}
+}
+
+type dummyDerivableList struct {
+	len  int
+	seed int
+}
+
+func newDummy(seed int) *dummyDerivableList {
+	d := &dummyDerivableList{}
+	src := mrand.NewSource(int64(seed))
+	// don't use lists longer than 4K items
+	d.len = int(src.Int63() & 0x0FFF)
+	d.seed = seed
+	return d
+}
+
+func (d *dummyDerivableList) Len() int {
+	return d.len
+}
+
+func (d *dummyDerivableList) GetRlp(i int) []byte {
+	src := mrand.NewSource(int64(d.seed + i))
+	// max item size 256, at least 1 byte per item
+	size := 1 + src.Int63()&0x00FF
+	data := make([]byte, size)
+	_, err := mrand.New(src).Read(data)
+	if err != nil {
+		panic(err)
+	}
+	return data
+}
+
+func printList(l types.DerivableList) {
+	fmt.Printf("list length: %d\n", l.Len())
+	fmt.Printf("{\n")
+	for i := 0; i < l.Len(); i++ {
+		v := l.GetRlp(i)
+		fmt.Printf("\"0x%x\",\n", v)
+	}
+	fmt.Printf("},\n")
+}
+
+func TestFuzzDeriveSha(t *testing.T) {
+	// increase this for longer runs -- it's set to quite low for travis
+	rndSeed := mrand.Int()
+	for i := 0; i < 10; i++ {
+		seed := rndSeed + i
+		exp := types.DeriveSha(newDummy(i), newEmpty())
+		got := types.DeriveSha(newDummy(i), NewStackTrie(nil))
+		if !bytes.Equal(got[:], exp[:]) {
+			printList(newDummy(seed))
+			t.Fatalf("seed %d: got %x exp %x", seed, got, exp)
+		}
+	}
+}
+
+type flatList struct {
+	rlpvals []string
+}
+
+func newFlatList(rlpvals []string) *flatList {
+	return &flatList{rlpvals}
+}
+func (f *flatList) Len() int {
+	return len(f.rlpvals)
+}
+func (f *flatList) GetRlp(i int) []byte {
+	return hexutil.MustDecode(f.rlpvals[i])
+}
+
+// TestDerivableList contains testcases found via fuzzing
+func TestDerivableList(t *testing.T) {
+	type tcase []string
+	tcs := []tcase{
+		{
+			"0xc041",
+		},
+		{
+			"0xf04cf757812428b0763112efb33b6f4fad7deb445e",
+			"0xf04cf757812428b0763112efb33b6f4fad7deb445e",
+		},
+		{
+			"0xca410605310cdc3bb8d4977ae4f0143df54a724ed873457e2272f39d66e0460e971d9d",
+			"0x6cd850eca0a7ac46bb1748d7b9cb88aa3bd21c57d852c28198ad8fa422c4595032e88a4494b4778b36b944fe47a52b8c5cd312910139dfcb4147ab8e972cc456bcb063f25dd78f54c4d34679e03142c42c662af52947d45bdb6e555751334ace76a5080ab5a0256a1d259855dfc5c0b8023b25befbb13fd3684f9f755cbd3d63544c78ee2001452dd54633a7593ade0b183891a0a4e9c7844e1254005fbe592b1b89149a502c24b6e1dca44c158aebedf01beae9c30cabe16a",
+			"0x14abd5c47c0be87b0454596baad2",
+			"0xca410605310cdc3bb8d4977ae4f0143df54a724ed873457e2272f39d66e0460e971d9d",
+		},
+	}
+	for i, tc := range tcs[1:] {
+		exp := types.DeriveSha(newFlatList(tc), newEmpty())
+		got := types.DeriveSha(newFlatList(tc), NewStackTrie(nil))
+		if !bytes.Equal(got[:], exp[:]) {
+			t.Fatalf("case %d: got %x exp %x", i, got, exp)
+		}
+	}
+}