eth/protocols/snap: generate storage trie from full dirty snap data (#22668)

* eth/protocols/snap: generate storage trie from full dirty snap data

* eth/protocols/snap: get rid of some more dead code

* eth/protocols/snap: less frequent logs, also log during trie generation

* eth/protocols/snap: implement dirty account range stack-hashing

* eth/protocols/snap: don't loop on account trie generation

* eth/protocols/snap: fix account format in trie

* core, eth, ethdb: glue snap packets together, but not chunks

* eth/protocols/snap: print completion log for snap phase

* eth/protocols/snap: extended tests

* eth/protocols/snap: make testcase pass

* eth/protocols/snap: fix account stacktrie commit without defer

* ethdb: fix key counts on reset

* eth/protocols: fix typos

* eth/protocols/snap: make better use of delivered data (#44)

* eth/protocols/snap: make better use of delivered data

* squashme

* eth/protocols/snap: reduce chunking

* squashme

* eth/protocols/snap: reduce chunking further

* eth/protocols/snap: break out hash range calculations

* eth/protocols/snap: use sort.Search instead of looping

* eth/protocols/snap: prevent crash on storage response with no keys

* eth/protocols/snap: nitpicks all around

* eth/protocols/snap: clear heal need on 1-chunk storage completion

* eth/protocols/snap: fix range chunker, add tests
Co-authored-by: NPéter Szilágyi <peterke@gmail.com>

* trie: fix test API error

* eth/protocols/snap: fix some further liter issues

* eth/protocols/snap: fix accidental batch reuse
Co-authored-by: NMartin Holst Swende <martin@swende.se>

core/rawdb/database_test.go

+// 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 rawdb

core/rawdb/table.go

@@ -176,6 +176,11 @@ func (b *tableBatch) Delete(key []byte) error {
 	return b.batch.Delete(append([]byte(b.prefix), key...))
 }
 
+// KeyCount retrieves the number of keys queued up for writing.
+func (b *tableBatch) KeyCount() int {
+	return b.batch.KeyCount()
+}
+
 // ValueSize retrieves the amount of data queued up for writing.
 func (b *tableBatch) ValueSize() int {
 	return b.batch.ValueSize()

eth/protocols/snap/handler.go

@@ -354,7 +354,7 @@ func handleMessage(backend Backend, peer *Peer) error {
 		if err := msg.Decode(res); err != nil {
 			return fmt.Errorf("%w: message %v: %v", errDecode, msg, err)
 		}
-		// Ensure the ranges ae monotonically increasing
+		// Ensure the ranges are monotonically increasing
 		for i, slots := range res.Slots {
 			for j := 1; j < len(slots); j++ {
 				if bytes.Compare(slots[j-1].Hash[:], slots[j].Hash[:]) >= 0 {

eth/protocols/snap/range.go

+// Copyright 2021 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package snap
+
+import (
+	"math/big"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/holiman/uint256"
+)
+
+// hashRange is a utility to handle ranges of hashes, Split up the
+// hash-space into sections, and 'walk' over the sections
+type hashRange struct {
+	current *uint256.Int
+	step    *uint256.Int
+}
+
+// newHashRange creates a new hashRange, initiated at the start position,
+// and with the step set to fill the desired 'num' chunks
+func newHashRange(start common.Hash, num uint64) *hashRange {
+	left := new(big.Int).Sub(hashSpace, start.Big())
+	step := new(big.Int).Div(
+		new(big.Int).Add(left, new(big.Int).SetUint64(num-1)),
+		new(big.Int).SetUint64(num),
+	)
+	step256 := new(uint256.Int)
+	step256.SetFromBig(step)
+
+	return &hashRange{
+		current: uint256.NewInt().SetBytes32(start[:]),
+		step:    step256,
+	}
+}
+
+// Next pushes the hash range to the next interval.
+func (r *hashRange) Next() bool {
+	next := new(uint256.Int)
+	if overflow := next.AddOverflow(r.current, r.step); overflow {
+		return false
+	}
+	r.current = next
+	return true
+}
+
+// Start returns the first hash in the current interval.
+func (r *hashRange) Start() common.Hash {
+	return r.current.Bytes32()
+}
+
+// End returns the last hash in the current interval.
+func (r *hashRange) End() common.Hash {
+	// If the end overflows (non divisible range), return a shorter interval
+	next := new(uint256.Int)
+	if overflow := next.AddOverflow(r.current, r.step); overflow {
+		return common.HexToHash("0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff")
+	}
+	return new(uint256.Int).Sub(next, uint256.NewInt().SetOne()).Bytes32()
+}
+
+// incHash returns the next hash, in lexicographical order (a.k.a plus one)
+func incHash(h common.Hash) common.Hash {
+	a := uint256.NewInt().SetBytes32(h[:])
+	a.Add(a, uint256.NewInt().SetOne())
+	return common.Hash(a.Bytes32())
+}

eth/protocols/snap/range_test.go

+// Copyright 2021 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package snap
+
+import (
+	"testing"
+
+	"github.com/ethereum/go-ethereum/common"
+)
+
+// Tests that given a starting hash and a density, the hash ranger can correctly
+// split up the remaining hash space into a fixed number of chunks.
+func TestHashRanges(t *testing.T) {
+	tests := []struct {
+		head   common.Hash
+		chunks uint64
+		starts []common.Hash
+		ends   []common.Hash
+	}{
+		// Simple test case to split the entire hash range into 4 chunks
+		{
+			head:   common.Hash{},
+			chunks: 4,
+			starts: []common.Hash{
+				{},
+				common.HexToHash("0x4000000000000000000000000000000000000000000000000000000000000000"),
+				common.HexToHash("0x8000000000000000000000000000000000000000000000000000000000000000"),
+				common.HexToHash("0xc000000000000000000000000000000000000000000000000000000000000000"),
+			},
+			ends: []common.Hash{
+				common.HexToHash("0x3fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+				common.HexToHash("0x7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+				common.HexToHash("0xbfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+				common.HexToHash("0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			},
+		},
+		// Split a divisible part of the hash range up into 2 chunks
+		{
+			head:   common.HexToHash("0x2000000000000000000000000000000000000000000000000000000000000000"),
+			chunks: 2,
+			starts: []common.Hash{
+				common.Hash{},
+				common.HexToHash("0x9000000000000000000000000000000000000000000000000000000000000000"),
+			},
+			ends: []common.Hash{
+				common.HexToHash("0x8fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+				common.HexToHash("0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			},
+		},
+		// Split the entire hash range into a non divisible 3 chunks
+		{
+			head:   common.Hash{},
+			chunks: 3,
+			starts: []common.Hash{
+				{},
+				common.HexToHash("0x5555555555555555555555555555555555555555555555555555555555555556"),
+				common.HexToHash("0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaac"),
+			},
+			ends: []common.Hash{
+				common.HexToHash("0x5555555555555555555555555555555555555555555555555555555555555555"),
+				common.HexToHash("0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaab"),
+				common.HexToHash("0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			},
+		},
+		// Split a part of hash range into a non divisible 3 chunks
+		{
+			head:   common.HexToHash("0x2000000000000000000000000000000000000000000000000000000000000000"),
+			chunks: 3,
+			starts: []common.Hash{
+				{},
+				common.HexToHash("0x6aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaab"),
+				common.HexToHash("0xb555555555555555555555555555555555555555555555555555555555555556"),
+			},
+			ends: []common.Hash{
+				common.HexToHash("0x6aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"),
+				common.HexToHash("0xb555555555555555555555555555555555555555555555555555555555555555"),
+				common.HexToHash("0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			},
+		},
+		// Split a part of hash range into a non divisible 3 chunks, but with a
+		// meaningful space size for manual verification.
+		//   - The head being 0xff...f0, we have 14 hashes left in the space
+		//   - Chunking up 14 into 3 pieces is 4.(6), but we need the ceil of 5 to avoid a micro-last-chunk
+		//   - Since the range is not divisible, the last interval will be shrter, capped at 0xff...f
+		//   - The chunk ranges thus needs to be [..0, ..5], [..6, ..b], [..c, ..f]
+		{
+			head:   common.HexToHash("0xfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff0"),
+			chunks: 3,
+			starts: []common.Hash{
+				{},
+				common.HexToHash("0xfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff6"),
+				common.HexToHash("0xfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffc"),
+			},
+			ends: []common.Hash{
+				common.HexToHash("0xfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff5"),
+				common.HexToHash("0xfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffb"),
+				common.HexToHash("0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			},
+		},
+	}
+	for i, tt := range tests {
+		r := newHashRange(tt.head, tt.chunks)
+
+		var (
+			starts = []common.Hash{{}}
+			ends   = []common.Hash{r.End()}
+		)
+		for r.Next() {
+			starts = append(starts, r.Start())
+			ends = append(ends, r.End())
+		}
+		if len(starts) != len(tt.starts) {
+			t.Errorf("test %d: starts count mismatch: have %d, want %d", i, len(starts), len(tt.starts))
+		}
+		for j := 0; j < len(starts) && j < len(tt.starts); j++ {
+			if starts[j] != tt.starts[j] {
+				t.Errorf("test %d, start %d: hash mismatch: have %x, want %x", i, j, starts[j], tt.starts[j])
+			}
+		}
+		if len(ends) != len(tt.ends) {
+			t.Errorf("test %d: ends count mismatch: have %d, want %d", i, len(ends), len(tt.ends))
+		}
+		for j := 0; j < len(ends) && j < len(tt.ends); j++ {
+			if ends[j] != tt.ends[j] {
+				t.Errorf("test %d, end %d: hash mismatch: have %x, want %x", i, j, ends[j], tt.ends[j])
+			}
+		}
+	}
+}

eth/protocols/snap/sync_test.go

@@ -135,6 +135,12 @@ type testPeer struct {
 	trieRequestHandler    trieHandlerFunc
 	codeRequestHandler    codeHandlerFunc
 	term                  func()
+
+	// counters
+	nAccountRequests  int
+	nStorageRequests  int
+	nBytecodeRequests int
+	nTrienodeRequests int
 }
 
 func newTestPeer(id string, t *testing.T, term func()) *testPeer {
@@ -156,19 +162,30 @@ func newTestPeer(id string, t *testing.T, term func()) *testPeer {
 func (t *testPeer) ID() string      { return t.id }
 func (t *testPeer) Log() log.Logger { return t.logger }
 
+func (t *testPeer) Stats() string {
+	return fmt.Sprintf(`Account requests: %d
+Storage requests: %d
+Bytecode requests: %d
+Trienode requests: %d
+`, t.nAccountRequests, t.nStorageRequests, t.nBytecodeRequests, t.nTrienodeRequests)
+}
+
 func (t *testPeer) RequestAccountRange(id uint64, root, origin, limit common.Hash, bytes uint64) error {
 	t.logger.Trace("Fetching range of accounts", "reqid", id, "root", root, "origin", origin, "limit", limit, "bytes", common.StorageSize(bytes))
+	t.nAccountRequests++
 	go t.accountRequestHandler(t, id, root, origin, limit, bytes)
 	return nil
 }
 
 func (t *testPeer) RequestTrieNodes(id uint64, root common.Hash, paths []TrieNodePathSet, bytes uint64) error {
 	t.logger.Trace("Fetching set of trie nodes", "reqid", id, "root", root, "pathsets", len(paths), "bytes", common.StorageSize(bytes))
+	t.nTrienodeRequests++
 	go t.trieRequestHandler(t, id, root, paths, bytes)
 	return nil
 }
 
 func (t *testPeer) RequestStorageRanges(id uint64, root common.Hash, accounts []common.Hash, origin, limit []byte, bytes uint64) error {
+	t.nStorageRequests++
 	if len(accounts) == 1 && origin != nil {
 		t.logger.Trace("Fetching range of large storage slots", "reqid", id, "root", root, "account", accounts[0], "origin", common.BytesToHash(origin), "limit", common.BytesToHash(limit), "bytes", common.StorageSize(bytes))
 	} else {
@@ -179,6 +196,7 @@ func (t *testPeer) RequestStorageRanges(id uint64, root common.Hash, accounts []
 }
 
 func (t *testPeer) RequestByteCodes(id uint64, hashes []common.Hash, bytes uint64) error {
+	t.nBytecodeRequests++
 	t.logger.Trace("Fetching set of byte codes", "reqid", id, "hashes", len(hashes), "bytes", common.StorageSize(bytes))
 	go t.codeRequestHandler(t, id, hashes, bytes)
 	return nil
@@ -1365,7 +1383,7 @@ func makeBoundaryAccountTrie(n int) (*trie.Trie, entrySlice) {
 	step := new(big.Int).Sub(
 		new(big.Int).Div(
 			new(big.Int).Exp(common.Big2, common.Big256, nil),
-			big.NewInt(accountConcurrency),
+			big.NewInt(int64(accountConcurrency)),
 		), common.Big1,
 	)
 	for i := 0; i < accountConcurrency; i++ {
@@ -1529,7 +1547,7 @@ func makeBoundaryStorageTrie(n int, db *trie.Database) (*trie.Trie, entrySlice)
 	step := new(big.Int).Sub(
 		new(big.Int).Div(
 			new(big.Int).Exp(common.Big2, common.Big256, nil),
-			big.NewInt(accountConcurrency),
+			big.NewInt(int64(accountConcurrency)),
 		), common.Big1,
 	)
 	for i := 0; i < accountConcurrency; i++ {
@@ -1605,3 +1623,94 @@ func verifyTrie(db ethdb.KeyValueStore, root common.Hash, t *testing.T) {
 	}
 	t.Logf("accounts: %d, slots: %d", accounts, slots)
 }
+
+// TestSyncAccountPerformance tests how efficient the snap algo is at minimizing
+// state healing
+func TestSyncAccountPerformance(t *testing.T) {
+	// Set the account concurrency to 1. This _should_ result in the
+	// range root to become correct, and there should be no healing needed
+	defer func(old int) { accountConcurrency = old }(accountConcurrency)
+	accountConcurrency = 1
+
+	var (
+		once   sync.Once
+		cancel = make(chan struct{})
+		term   = func() {
+			once.Do(func() {
+				close(cancel)
+			})
+		}
+	)
+	sourceAccountTrie, elems := makeAccountTrieNoStorage(100)
+
+	mkSource := func(name string) *testPeer {
+		source := newTestPeer(name, t, term)
+		source.accountTrie = sourceAccountTrie
+		source.accountValues = elems
+		return source
+	}
+	src := mkSource("source")
+	syncer := setupSyncer(src)
+	if err := syncer.Sync(sourceAccountTrie.Hash(), cancel); err != nil {
+		t.Fatalf("sync failed: %v", err)
+	}
+	verifyTrie(syncer.db, sourceAccountTrie.Hash(), t)
+	// The trie root will always be requested, since it is added when the snap
+	// sync cycle starts. When popping the queue, we do not look it up again.
+	// Doing so would bring this number down to zero in this artificial testcase,
+	// but only add extra IO for no reason in practice.
+	if have, want := src.nTrienodeRequests, 1; have != want {
+		fmt.Printf(src.Stats())
+		t.Errorf("trie node heal requests wrong, want %d, have %d", want, have)
+	}
+}
+
+func TestSlotEstimation(t *testing.T) {
+	for i, tc := range []struct {
+		last  common.Hash
+		count int
+		want  uint64
+	}{
+		{
+			// Half the space
+			common.HexToHash("0x7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			100,
+			100,
+		},
+		{
+			// 1 / 16th
+			common.HexToHash("0x0fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"),
+			100,
+			1500,
+		},
+		{
+			// Bit more than 1 / 16th
+			common.HexToHash("0x1000000000000000000000000000000000000000000000000000000000000000"),
+			100,
+			1499,
+		},
+		{
+			// Almost everything
+			common.HexToHash("0xF000000000000000000000000000000000000000000000000000000000000000"),
+			100,
+			6,
+		},
+		{
+			// Almost nothing -- should lead to error
+			common.HexToHash("0x0000000000000000000000000000000000000000000000000000000000000001"),
+			1,
+			0,
+		},
+		{
+			// Nothing -- should lead to error
+			common.Hash{},
+			100,
+			0,
+		},
+	} {
+		have, _ := estimateRemainingSlots(tc.count, tc.last)
+		if want := tc.want; have != want {
+			t.Errorf("test %d: have %d want %d", i, have, want)
+		}
+	}
+}

ethdb/batch.go

@@ -25,6 +25,9 @@ const IdealBatchSize = 100 * 1024
 type Batch interface {
 	KeyValueWriter
 
+	// KeyCount retrieves the number of keys queued up for writing.
+	KeyCount() int
+
 	// ValueSize retrieves the amount of data queued up for writing.
 	ValueSize() int
 

ethdb/leveldb/leveldb.go

@@ -448,6 +448,7 @@ func (db *Database) meter(refresh time.Duration) {
 type batch struct {
 	db   *leveldb.DB
 	b    *leveldb.Batch
+	keys int
 	size int
 }
 
@@ -461,10 +462,16 @@ func (b *batch) Put(key, value []byte) error {
 // Delete inserts the a key removal into the batch for later committing.
 func (b *batch) Delete(key []byte) error {
 	b.b.Delete(key)
+	b.keys++
 	b.size += len(key)
 	return nil
 }
 
+// KeyCount retrieves the number of keys queued up for writing.
+func (b *batch) KeyCount() int {
+	return b.keys
+}
+
 // ValueSize retrieves the amount of data queued up for writing.
 func (b *batch) ValueSize() int {
 	return b.size
@@ -478,7 +485,7 @@ func (b *batch) Write() error {
 // Reset resets the batch for reuse.
 func (b *batch) Reset() {
 	b.b.Reset()
-	b.size = 0
+	b.keys, b.size = 0, 0
 }
 
 // Replay replays the batch contents.

ethdb/memorydb/memorydb.go

@@ -198,6 +198,7 @@ type keyvalue struct {
 type batch struct {
 	db     *Database
 	writes []keyvalue
+	keys   int
 	size   int
 }
 
@@ -211,10 +212,16 @@ func (b *batch) Put(key, value []byte) error {
 // Delete inserts the a key removal into the batch for later committing.
 func (b *batch) Delete(key []byte) error {
 	b.writes = append(b.writes, keyvalue{common.CopyBytes(key), nil, true})
+	b.keys++
 	b.size += len(key)
 	return nil
 }
 
+// KeyCount retrieves the number of keys queued up for writing.
+func (b *batch) KeyCount() int {
+	return b.keys
+}
+
 // ValueSize retrieves the amount of data queued up for writing.
 func (b *batch) ValueSize() int {
 	return b.size
@@ -238,7 +245,7 @@ func (b *batch) Write() error {
 // Reset resets the batch for reuse.
 func (b *batch) Reset() {
 	b.writes = b.writes[:0]
-	b.size = 0
+	b.keys, b.size = 0, 0
 }
 
 // Replay replays the batch contents.

tests/fuzzers/stacktrie/trie_fuzzer.go

@@ -90,6 +90,7 @@ func (b *spongeBatch) Put(key, value []byte) error {
 	return nil
 }
 func (b *spongeBatch) Delete(key []byte) error             { panic("implement me") }
+func (b *spongeBatch) KeyCount() int                       { panic("not implemented") }
 func (b *spongeBatch) ValueSize() int                      { return 100 }
 func (b *spongeBatch) Write() error                        { return nil }
 func (b *spongeBatch) Reset()                              {}

trie/trie_test.go

@@ -706,6 +706,7 @@ func (b *spongeBatch) Put(key, value []byte) error {
 	return nil
 }
 func (b *spongeBatch) Delete(key []byte) error             { panic("implement me") }
+func (b *spongeBatch) KeyCount() int                       { return 100 }
 func (b *spongeBatch) ValueSize() int                      { return 100 }
 func (b *spongeBatch) Write() error                        { return nil }
 func (b *spongeBatch) Reset()                              {}