add support for validator code prefix in snapshot

cmd/harmony/main.go

@@ -106,7 +106,9 @@ func init() {
 	rootCmd.AddCommand(dumpConfigLegacyCmd)
 	rootCmd.AddCommand(dumpDBCmd)
 	rootCmd.AddCommand(inspectDBCmd)
-
+	rootCmd.AddCommand(snapshotCmd)
+	snapshotCmd.AddCommand(pruneStateCmd)
+	snapshotCmd.AddCommand(verifyStateCmd)
 	if err := registerRootCmdFlags(); err != nil {
 		os.Exit(2)
 	}
@@ -119,6 +121,9 @@ func init() {
 	if err := registerInspectionFlags(); err != nil {
 		os.Exit(2)
 	}
+	if err := registerSnapshotCmdFlags(); err != nil {
+		os.Exit(2)
+	}
 }
 
 func main() {

cmd/harmony/snapshotdb.go

@@ -0,0 +1,168 @@
+package main
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+
+	"github.com/ethereum/go-ethereum/trie"
+	"github.com/spf13/cobra"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/harmony-one/harmony/core/rawdb"
+	"github.com/harmony-one/harmony/core/state/pruner"
+	"github.com/harmony-one/harmony/core/state/snapshot"
+	"github.com/harmony-one/harmony/internal/cli"
+	"github.com/harmony-one/harmony/internal/utils"
+)
+
+var bloomFilterSizeFlag = cli.IntFlag{
+	Name:      "bloomfilter.size",
+	Shorthand: "b",
+	Usage:     "Megabytes of memory allocated to bloom-filter for pruning",
+	DefValue:  2048,
+}
+
+var stateRootFlag = cli.StringFlag{
+	Name:      "stateroot",
+	Shorthand: "r",
+	Usage:     "state root hash",
+	DefValue:  "",
+}
+
+var snapshotCmd = &cobra.Command{
+	Use:   "snapshot",
+	Short: "A set of commands based on the snapshot",
+	Long:  "A set of commands based on the snapshot",
+	Run: func(cmd *cobra.Command, args []string) {
+		fmt.Println("Error: must also specify a subcommand (prune-state, verify, ...)")
+	},
+}
+
+var pruneStateCmd = &cobra.Command{
+	Use:     "prune-state srcdb cachedir",
+	Short:   "prune stale harmony state data based on snapshot",
+	Long:    "will prune historical state data with the help of state snapshot. All trie nodes that do not belong to the specified version state will be deleted from the database",
+	Example: "harmony prune-state /srcDir/harmony_db_0 /prune_cache",
+	Args:    cobra.RangeArgs(1, 2),
+	Run: func(cmd *cobra.Command, args []string) {
+		srcDBDir, cachedir := args[0], args[1]
+		bloomFilterSize := cli.GetIntFlagValue(cmd, bloomFilterSizeFlag)
+		stateRoot := cli.GetStringFlagValue(cmd, stateRootFlag)
+
+		chaindb, err := rawdb.NewLevelDBDatabase(srcDBDir, LEVELDB_CACHE_SIZE, LEVELDB_HANDLES, "", false)
+		if err != nil {
+			fmt.Println("open src db error:", err)
+			os.Exit(-1)
+		}
+		defer chaindb.Close()
+
+		prunerconfig := pruner.Config{
+			Datadir:   ResolvePath(""),
+			Cachedir:  ResolvePath(cachedir),
+			BloomSize: uint64(bloomFilterSize),
+		}
+		pruner, err := pruner.NewPruner(chaindb, prunerconfig)
+		if err != nil {
+			utils.Logger().Error().Err(err).Msg("Failed to open snapshot tree")
+			return
+		}
+
+		var targetRoot common.Hash
+		if len(stateRoot) >= 3 {
+			targetRoot, err = parseRoot(stateRoot)
+			if err != nil {
+				utils.Logger().Error().Err(err).Msg("Failed to resolve state root")
+				return
+			}
+		} else {
+			targetRoot = rawdb.ReadHeadBlockHash(chaindb)
+		}
+
+		if err = pruner.Prune(targetRoot); err != nil {
+			utils.Logger().Error().Err(err).Msg("Failed to prune state")
+			return
+		}
+
+		return
+	},
+}
+
+var verifyStateCmd = &cobra.Command{
+	Use:     "verify-state srcdb",
+	Short:   "Recalculate state hash based on snapshot for verification",
+	Long:    "Recalculate state hash based on snapshot for verification",
+	Example: "harmony verify-state /srcDir/harmony_db_0",
+	Args:    cobra.ExactArgs(1),
+	Run: func(cmd *cobra.Command, args []string) {
+		if len(args) > 1 {
+			fmt.Println("too many arguments")
+			return
+		}
+		srcDBDir := args[0]
+		chaindb, err := rawdb.NewLevelDBDatabase(srcDBDir, LEVELDB_CACHE_SIZE, LEVELDB_HANDLES, "", false)
+		if err != nil {
+			fmt.Println("open src db error:", err)
+			os.Exit(-1)
+		}
+		defer chaindb.Close()
+
+		headRoot := rawdb.ReadHeadBlockHash(chaindb)
+		stateRoot := cli.GetStringFlagValue(cmd, stateRootFlag)
+		var targetRoot common.Hash
+		if len(stateRoot) >= 3 {
+			var err error
+			if targetRoot, err = parseRoot(stateRoot); err != nil {
+				utils.Logger().Error().Err(err).Msg("Failed to resolve state root")
+				return
+			}
+		} else {
+			targetRoot = headRoot
+		}
+		snapconfig := snapshot.Config{
+			CacheSize:  256,
+			Recovery:   false,
+			NoBuild:    true,
+			AsyncBuild: false,
+		}
+		snaptree, err := snapshot.New(snapconfig, chaindb, trie.NewDatabase(chaindb), headRoot)
+		if err != nil {
+			utils.Logger().Error().Err(err).Msg("Failed to open snapshot tree")
+			return
+		}
+		if err := snaptree.Verify(targetRoot); err != nil {
+			utils.Logger().Error().Err(err).Interface("root", targetRoot).Msg("Failed to verify state")
+			return
+		}
+		utils.Logger().Info().Interface("root", targetRoot).Msg("Verified the state")
+		if err := snapshot.CheckDanglingStorage(chaindb); err != nil {
+			utils.Logger().Error().Err(err).Interface("root", targetRoot).Msg("Failed to check dangling storage")
+		}
+		return
+	},
+}
+
+func ResolvePath(filename string) string {
+	if filepath.IsAbs(filename) {
+		return filename
+	}
+	return filepath.Join(filepath.Dir("."), filename)
+}
+
+func parseRoot(input string) (common.Hash, error) {
+	var h common.Hash
+	if err := h.UnmarshalText([]byte(input)); err != nil {
+		return h, err
+	}
+	return h, nil
+}
+
+func registerSnapshotCmdFlags() error {
+	if err := cli.RegisterFlags(pruneStateCmd, []cli.Flag{bloomFilterSizeFlag, stateRootFlag}); err != nil {
+		return err
+	}
+	if err := cli.RegisterFlags(verifyStateCmd, []cli.Flag{stateRootFlag}); err != nil {
+		return err
+	}
+	return nil
+}

core/state/pruner/bloom.go

@@ -0,0 +1,131 @@
+// 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 pruner
+
+import (
+	"encoding/binary"
+	"errors"
+	"os"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/log"
+	"github.com/harmony-one/harmony/core/rawdb"
+	bloomfilter "github.com/holiman/bloomfilter/v2"
+)
+
+// stateBloomHasher is a wrapper around a byte blob to satisfy the interface API
+// requirements of the bloom library used. It's used to convert a trie hash or
+// contract code hash into a 64 bit mini hash.
+type stateBloomHasher []byte
+
+func (f stateBloomHasher) Write(p []byte) (n int, err error) { panic("not implemented") }
+func (f stateBloomHasher) Sum(b []byte) []byte               { panic("not implemented") }
+func (f stateBloomHasher) Reset()                            { panic("not implemented") }
+func (f stateBloomHasher) BlockSize() int                    { panic("not implemented") }
+func (f stateBloomHasher) Size() int                         { return 8 }
+func (f stateBloomHasher) Sum64() uint64                     { return binary.BigEndian.Uint64(f) }
+
+// stateBloom is a bloom filter used during the state conversion(snapshot->state).
+// The keys of all generated entries will be recorded here so that in the pruning
+// stage the entries belong to the specific version can be avoided for deletion.
+//
+// The false-positive is allowed here. The "false-positive" entries means they
+// actually don't belong to the specific version but they are not deleted in the
+// pruning. The downside of the false-positive allowance is we may leave some "dangling"
+// nodes in the disk. But in practice the it's very unlike the dangling node is
+// state root. So in theory this pruned state shouldn't be visited anymore. Another
+// potential issue is for fast sync. If we do another fast sync upon the pruned
+// database, it's problematic which will stop the expansion during the syncing.
+//
+// After the entire state is generated, the bloom filter should be persisted into
+// the disk. It indicates the whole generation procedure is finished.
+type stateBloom struct {
+	bloom *bloomfilter.Filter
+}
+
+// newStateBloomWithSize creates a brand new state bloom for state generation.
+// The bloom filter will be created by the passing bloom filter size. According
+// to the https://hur.st/bloomfilter/?n=600000000&p=&m=2048MB&k=4, the parameters
+// are picked so that the false-positive rate for mainnet is low enough.
+func newStateBloomWithSize(size uint64) (*stateBloom, error) {
+	bloom, err := bloomfilter.New(size*1024*1024*8, 4)
+	if err != nil {
+		return nil, err
+	}
+	log.Info("Initialized state bloom", "size", common.StorageSize(float64(bloom.M()/8)))
+	return &stateBloom{bloom: bloom}, nil
+}
+
+// NewStateBloomFromDisk loads the state bloom from the given file.
+// In this case the assumption is held the bloom filter is complete.
+func NewStateBloomFromDisk(filename string) (*stateBloom, error) {
+	bloom, _, err := bloomfilter.ReadFile(filename)
+	if err != nil {
+		return nil, err
+	}
+	return &stateBloom{bloom: bloom}, nil
+}
+
+// Commit flushes the bloom filter content into the disk and marks the bloom
+// as complete.
+func (bloom *stateBloom) Commit(filename, tempname string) error {
+	// Write the bloom out into a temporary file
+	_, err := bloom.bloom.WriteFile(tempname)
+	if err != nil {
+		return err
+	}
+	// Ensure the file is synced to disk
+	f, err := os.OpenFile(tempname, os.O_RDWR, 0666)
+	if err != nil {
+		return err
+	}
+	if err := f.Sync(); err != nil {
+		f.Close()
+		return err
+	}
+	f.Close()
+
+	// Move the temporary file into it's final location
+	return os.Rename(tempname, filename)
+}
+
+// Put implements the KeyValueWriter interface. But here only the key is needed.
+func (bloom *stateBloom) Put(key []byte, value []byte) error {
+	// If the key length is not 32bytes, ensure it's contract code
+	// entry with new scheme.
+	if len(key) != common.HashLength {
+		isCode, codeKey := rawdb.IsCodeKey(key)
+		if !isCode {
+			return errors.New("invalid entry")
+		}
+		bloom.bloom.Add(stateBloomHasher(codeKey))
+		return nil
+	}
+	bloom.bloom.Add(stateBloomHasher(key))
+	return nil
+}
+
+// Delete removes the key from the key-value data store.
+func (bloom *stateBloom) Delete(key []byte) error { panic("not supported") }
+
+// Contain is the wrapper of the underlying contains function which
+// reports whether the key is contained.
+// - If it says yes, the key may be contained
+// - If it says no, the key is definitely not contained.
+func (bloom *stateBloom) Contain(key []byte) (bool, error) {
+	return bloom.bloom.Contains(stateBloomHasher(key)), nil
+}

core/state/pruner/pruner.go

@@ -0,0 +1,516 @@
+// 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 pruner
+
+import (
+	"bytes"
+	"encoding/binary"
+	"errors"
+	"fmt"
+	"math"
+	"os"
+	"path/filepath"
+	"strings"
+	"time"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/log"
+	"github.com/ethereum/go-ethereum/rlp"
+	"github.com/ethereum/go-ethereum/trie"
+	"github.com/harmony-one/harmony/block"
+	"github.com/harmony-one/harmony/core/rawdb"
+	"github.com/harmony-one/harmony/core/state/snapshot"
+)
+
+const (
+	// stateBloomFilePrefix is the filename prefix of state bloom filter.
+	stateBloomFilePrefix = "statebloom"
+
+	// stateBloomFilePrefix is the filename suffix of state bloom filter.
+	stateBloomFileSuffix = "bf.gz"
+
+	// stateBloomFileTempSuffix is the filename suffix of state bloom filter
+	// while it is being written out to detect write aborts.
+	stateBloomFileTempSuffix = ".tmp"
+
+	// rangeCompactionThreshold is the minimal deleted entry number for
+	// triggering range compaction. It's a quite arbitrary number but just
+	// to avoid triggering range compaction because of small deletion.
+	rangeCompactionThreshold = 100000
+)
+
+// Config includes all the configurations for pruning.
+type Config struct {
+	Datadir   string // The directory of the state database
+	Cachedir  string // The directory of state clean cache
+	BloomSize uint64 // The Megabytes of memory allocated to bloom-filter
+}
+
+// Pruner is an offline tool to prune the stale state with the
+// help of the snapshot. The workflow of pruner is very simple:
+//
+//   - iterate the snapshot, reconstruct the relevant state
+//   - iterate the database, delete all other state entries which
+//     don't belong to the target state and the genesis state
+//
+// It can take several hours(around 2 hours for mainnet) to finish
+// the whole pruning work. It's recommended to run this offline tool
+// periodically in order to release the disk usage and improve the
+// disk read performance to some extent.
+type Pruner struct {
+	config      Config
+	chainHeader *block.Header
+	db          ethdb.Database
+	stateBloom  *stateBloom
+	snaptree    *snapshot.Tree
+}
+
+// NewPruner creates the pruner instance.
+func NewPruner(db ethdb.Database, config Config) (*Pruner, error) {
+	headBlock := rawdb.ReadHeadBlock(db)
+	if headBlock == nil {
+		return nil, errors.New("failed to load head block")
+	}
+	snapconfig := snapshot.Config{
+		CacheSize:  256,
+		Recovery:   false,
+		NoBuild:    false,
+		AsyncBuild: false,
+	}
+	snaptree, err := snapshot.New(snapconfig, db, trie.NewDatabase(db), headBlock.Root())
+	if err != nil {
+		return nil, err // The relevant snapshot(s) might not exist
+	}
+	// Sanitize the bloom filter size if it's too small.
+	if config.BloomSize < 256 {
+		log.Warn("Sanitizing bloomfilter size", "provided(MB)", config.BloomSize, "updated(MB)", 256)
+		config.BloomSize = 256
+	}
+	stateBloom, err := newStateBloomWithSize(config.BloomSize)
+	if err != nil {
+		return nil, err
+	}
+	return &Pruner{
+		config:      config,
+		chainHeader: headBlock.Header(),
+		db:          db,
+		stateBloom:  stateBloom,
+		snaptree:    snaptree,
+	}, nil
+}
+
+func prune(snaptree *snapshot.Tree, root common.Hash, maindb ethdb.Database, stateBloom *stateBloom, bloomPath string, middleStateRoots map[common.Hash]struct{}, start time.Time) error {
+	// Delete all stale trie nodes in the disk. With the help of state bloom
+	// the trie nodes(and codes) belong to the active state will be filtered
+	// out. A very small part of stale tries will also be filtered because of
+	// the false-positive rate of bloom filter. But the assumption is held here
+	// that the false-positive is low enough(~0.05%). The probablity of the
+	// dangling node is the state root is super low. So the dangling nodes in
+	// theory will never ever be visited again.
+	var (
+		count  int
+		size   common.StorageSize
+		pstart = time.Now()
+		logged = time.Now()
+		batch  = maindb.NewBatch()
+		iter   = maindb.NewIterator(nil, nil)
+	)
+	for iter.Next() {
+		key := iter.Key()
+
+		// All state entries don't belong to specific state and genesis are deleted here
+		// - trie node
+		// - legacy contract code
+		// - new-scheme contract code
+		isCode, codeKey := rawdb.IsCodeKey(key)
+		if len(key) == common.HashLength || isCode {
+			checkKey := key
+			if isCode {
+				checkKey = codeKey
+			}
+			if _, exist := middleStateRoots[common.BytesToHash(checkKey)]; exist {
+				log.Debug("Forcibly delete the middle state roots", "hash", common.BytesToHash(checkKey))
+			} else {
+				if ok, err := stateBloom.Contain(checkKey); err != nil {
+					return err
+				} else if ok {
+					continue
+				}
+			}
+			count += 1
+			size += common.StorageSize(len(key) + len(iter.Value()))
+			batch.Delete(key)
+
+			var eta time.Duration // Realistically will never remain uninited
+			if done := binary.BigEndian.Uint64(key[:8]); done > 0 {
+				var (
+					left  = math.MaxUint64 - binary.BigEndian.Uint64(key[:8])
+					speed = done/uint64(time.Since(pstart)/time.Millisecond+1) + 1 // +1s to avoid division by zero
+				)
+				eta = time.Duration(left/speed) * time.Millisecond
+			}
+			if time.Since(logged) > 8*time.Second {
+				log.Info("Pruning state data", "nodes", count, "size", size,
+					"elapsed", common.PrettyDuration(time.Since(pstart)), "eta", common.PrettyDuration(eta))
+				logged = time.Now()
+			}
+			// Recreate the iterator after every batch commit in order
+			// to allow the underlying compactor to delete the entries.
+			if batch.ValueSize() >= ethdb.IdealBatchSize {
+				batch.Write()
+				batch.Reset()
+
+				iter.Release()
+				iter = maindb.NewIterator(nil, key)
+			}
+		}
+	}
+	if batch.ValueSize() > 0 {
+		batch.Write()
+		batch.Reset()
+	}
+	iter.Release()
+	log.Info("Pruned state data", "nodes", count, "size", size, "elapsed", common.PrettyDuration(time.Since(pstart)))
+
+	// Pruning is done, now drop the "useless" layers from the snapshot.
+	// Firstly, flushing the target layer into the disk. After that all
+	// diff layers below the target will all be merged into the disk.
+	if err := snaptree.Cap(root, 0); err != nil {
+		return err
+	}
+	// Secondly, flushing the snapshot journal into the disk. All diff
+	// layers upon are dropped silently. Eventually the entire snapshot
+	// tree is converted into a single disk layer with the pruning target
+	// as the root.
+	if _, err := snaptree.Journal(root); err != nil {
+		return err
+	}
+	// Delete the state bloom, it marks the entire pruning procedure is
+	// finished. If any crashes or manual exit happens before this,
+	// `RecoverPruning` will pick it up in the next restarts to redo all
+	// the things.
+	os.RemoveAll(bloomPath)
+
+	// Start compactions, will remove the deleted data from the disk immediately.
+	// Note for small pruning, the compaction is skipped.
+	if count >= rangeCompactionThreshold {
+		cstart := time.Now()
+		for b := 0x00; b <= 0xf0; b += 0x10 {
+			var (
+				start = []byte{byte(b)}
+				end   = []byte{byte(b + 0x10)}
+			)
+			if b == 0xf0 {
+				end = nil
+			}
+			log.Info("Compacting database", "range", fmt.Sprintf("%#x-%#x", start, end), "elapsed", common.PrettyDuration(time.Since(cstart)))
+			if err := maindb.Compact(start, end); err != nil {
+				log.Error("Database compaction failed", "error", err)
+				return err
+			}
+		}
+		log.Info("Database compaction finished", "elapsed", common.PrettyDuration(time.Since(cstart)))
+	}
+	log.Info("State pruning successful", "pruned", size, "elapsed", common.PrettyDuration(time.Since(start)))
+	return nil
+}
+
+// Prune deletes all historical state nodes except the nodes belong to the
+// specified state version. If user doesn't specify the state version, use
+// the bottom-most snapshot diff layer as the target.
+func (p *Pruner) Prune(root common.Hash) error {
+	// If the state bloom filter is already committed previously,
+	// reuse it for pruning instead of generating a new one. It's
+	// mandatory because a part of state may already be deleted,
+	// the recovery procedure is necessary.
+	_, stateBloomRoot, err := findBloomFilter(p.config.Datadir)
+	if err != nil {
+		return err
+	}
+	if stateBloomRoot != (common.Hash{}) {
+		return RecoverPruning(p.config.Datadir, p.db, p.config.Cachedir)
+	}
+	// If the target state root is not specified, use the HEAD-127 as the
+	// target. The reason for picking it is:
+	// - in most of the normal cases, the related state is available
+	// - the probability of this layer being reorg is very low
+	var layers []snapshot.Snapshot
+	if root == (common.Hash{}) {
+		// Retrieve all snapshot layers from the current HEAD.
+		// In theory there are 128 difflayers + 1 disk layer present,
+		// so 128 diff layers are expected to be returned.
+		layers = p.snaptree.Snapshots(p.chainHeader.Root(), 128, true)
+		if len(layers) != 128 {
+			// Reject if the accumulated diff layers are less than 128. It
+			// means in most of normal cases, there is no associated state
+			// with bottom-most diff layer.
+			return fmt.Errorf("snapshot not old enough yet: need %d more blocks", 128-len(layers))
+		}
+		// Use the bottom-most diff layer as the target
+		root = layers[len(layers)-1].Root()
+	}
+	// Ensure the root is really present. The weak assumption
+	// is the presence of root can indicate the presence of the
+	// entire trie.
+	if !rawdb.HasLegacyTrieNode(p.db, root) {
+		// The special case is for clique based networks(rinkeby, goerli
+		// and some other private networks), it's possible that two
+		// consecutive blocks will have same root. In this case snapshot
+		// difflayer won't be created. So HEAD-127 may not paired with
+		// head-127 layer. Instead the paired layer is higher than the
+		// bottom-most diff layer. Try to find the bottom-most snapshot
+		// layer with state available.
+		//
+		// Note HEAD and HEAD-1 is ignored. Usually there is the associated
+		// state available, but we don't want to use the topmost state
+		// as the pruning target.
+		var found bool
+		for i := len(layers) - 2; i >= 2; i-- {
+			if rawdb.HasLegacyTrieNode(p.db, layers[i].Root()) {
+				root = layers[i].Root()
+				found = true
+				log.Info("Selecting middle-layer as the pruning target", "root", root, "depth", i)
+				break
+			}
+		}
+		if !found {
+			if len(layers) > 0 {
+				return errors.New("no snapshot paired state")
+			}
+			return fmt.Errorf("associated state[%x] is not present", root)
+		}
+	} else {
+		if len(layers) > 0 {
+			log.Info("Selecting bottom-most difflayer as the pruning target", "root", root, "height", p.chainHeader.Number().Uint64()-127)
+		} else {
+			log.Info("Selecting user-specified state as the pruning target", "root", root)
+		}
+	}
+	// Before start the pruning, delete the clean trie cache first.
+	// It's necessary otherwise in the next restart we will hit the
+	// deleted state root in the "clean cache" so that the incomplete
+	// state is picked for usage.
+	deleteCleanTrieCache(p.config.Cachedir)
+
+	// All the state roots of the middle layer should be forcibly pruned,
+	// otherwise the dangling state will be left.
+	middleRoots := make(map[common.Hash]struct{})
+	for _, layer := range layers {
+		if layer.Root() == root {
+			break
+		}
+		middleRoots[layer.Root()] = struct{}{}
+	}
+	// Traverse the target state, re-construct the whole state trie and
+	// commit to the given bloom filter.
+	start := time.Now()
+	if err := snapshot.GenerateTrie(p.snaptree, root, p.db, p.stateBloom); err != nil {
+		return err
+	}
+	// Traverse the genesis, put all genesis state entries into the
+	// bloom filter too.
+	if err := extractGenesis(p.db, p.stateBloom); err != nil {
+		return err
+	}
+	filterName := bloomFilterName(p.config.Datadir, root)
+
+	log.Info("Writing state bloom to disk", "name", filterName)
+	if err := p.stateBloom.Commit(filterName, filterName+stateBloomFileTempSuffix); err != nil {
+		return err
+	}
+	log.Info("State bloom filter committed", "name", filterName)
+	return prune(p.snaptree, root, p.db, p.stateBloom, filterName, middleRoots, start)
+}
+
+// RecoverPruning will resume the pruning procedure during the system restart.
+// This function is used in this case: user tries to prune state data, but the
+// system was interrupted midway because of crash or manual-kill. In this case
+// if the bloom filter for filtering active state is already constructed, the
+// pruning can be resumed. What's more if the bloom filter is constructed, the
+// pruning **has to be resumed**. Otherwise a lot of dangling nodes may be left
+// in the disk.
+func RecoverPruning(datadir string, db ethdb.Database, trieCachePath string) error {
+	stateBloomPath, stateBloomRoot, err := findBloomFilter(datadir)
+	if err != nil {
+		return err
+	}
+	if stateBloomPath == "" {
+		return nil // nothing to recover
+	}
+	headBlock := rawdb.ReadHeadBlock(db)
+	if headBlock == nil {
+		return errors.New("failed to load head block")
+	}
+	// Initialize the snapshot tree in recovery mode to handle this special case:
+	// - Users run the `prune-state` command multiple times
+	// - Neither these `prune-state` running is finished(e.g. interrupted manually)
+	// - The state bloom filter is already generated, a part of state is deleted,
+	//   so that resuming the pruning here is mandatory
+	// - The state HEAD is rewound already because of multiple incomplete `prune-state`
+	// In this case, even the state HEAD is not exactly matched with snapshot, it
+	// still feasible to recover the pruning correctly.
+	snapconfig := snapshot.Config{
+		CacheSize:  256,
+		Recovery:   true,
+		NoBuild:    true,
+		AsyncBuild: false,
+	}
+	snaptree, err := snapshot.New(snapconfig, db, trie.NewDatabase(db), headBlock.Root())
+	if err != nil {
+		return err // The relevant snapshot(s) might not exist
+	}
+	stateBloom, err := NewStateBloomFromDisk(stateBloomPath)
+	if err != nil {
+		return err
+	}
+	log.Info("Loaded state bloom filter", "path", stateBloomPath)
+
+	// Before start the pruning, delete the clean trie cache first.
+	// It's necessary otherwise in the next restart we will hit the
+	// deleted state root in the "clean cache" so that the incomplete
+	// state is picked for usage.
+	deleteCleanTrieCache(trieCachePath)
+
+	// All the state roots of the middle layers should be forcibly pruned,
+	// otherwise the dangling state will be left.
+	var (
+		found       bool
+		layers      = snaptree.Snapshots(headBlock.Root(), 128, true)
+		middleRoots = make(map[common.Hash]struct{})
+	)
+	for _, layer := range layers {
+		if layer.Root() == stateBloomRoot {
+			found = true
+			break
+		}
+		middleRoots[layer.Root()] = struct{}{}
+	}
+	if !found {
+		log.Error("Pruning target state is not existent")
+		return errors.New("non-existent target state")
+	}
+	return prune(snaptree, stateBloomRoot, db, stateBloom, stateBloomPath, middleRoots, time.Now())
+}
+
+// extractGenesis loads the genesis state and commits all the state entries
+// into the given bloomfilter.
+func extractGenesis(db ethdb.Database, stateBloom *stateBloom) error {
+	genesisHash := rawdb.ReadCanonicalHash(db, 0)
+	if genesisHash == (common.Hash{}) {
+		return errors.New("missing genesis hash")
+	}
+	genesis := rawdb.ReadBlock(db, genesisHash, 0)
+	if genesis == nil {
+		return errors.New("missing genesis block")
+	}
+	t, err := trie.NewStateTrie(trie.StateTrieID(genesis.Root()), trie.NewDatabase(db))
+	if err != nil {
+		return err
+	}
+	accIter := t.NodeIterator(nil)
+	for accIter.Next(true) {
+		hash := accIter.Hash()
+
+		// Embedded nodes don't have hash.
+		if hash != (common.Hash{}) {
+			stateBloom.Put(hash.Bytes(), nil)
+		}
+		// If it's a leaf node, yes we are touching an account,
+		// dig into the storage trie further.
+		if accIter.Leaf() {
+			var acc types.StateAccount
+			if err := rlp.DecodeBytes(accIter.LeafBlob(), &acc); err != nil {
+				return err
+			}
+			if acc.Root != types.EmptyRootHash {
+				id := trie.StorageTrieID(genesis.Root(), common.BytesToHash(accIter.LeafKey()), acc.Root)
+				storageTrie, err := trie.NewStateTrie(id, trie.NewDatabase(db))
+				if err != nil {
+					return err
+				}
+				storageIter := storageTrie.NodeIterator(nil)
+				for storageIter.Next(true) {
+					hash := storageIter.Hash()
+					if hash != (common.Hash{}) {
+						stateBloom.Put(hash.Bytes(), nil)
+					}
+				}
+				if storageIter.Error() != nil {
+					return storageIter.Error()
+				}
+			}
+			if !bytes.Equal(acc.CodeHash, types.EmptyCodeHash.Bytes()) {
+				stateBloom.Put(acc.CodeHash, nil)
+			}
+		}
+	}
+	return accIter.Error()
+}
+
+func bloomFilterName(datadir string, hash common.Hash) string {
+	return filepath.Join(datadir, fmt.Sprintf("%s.%s.%s", stateBloomFilePrefix, hash.Hex(), stateBloomFileSuffix))
+}
+
+func isBloomFilter(filename string) (bool, common.Hash) {
+	filename = filepath.Base(filename)
+	if strings.HasPrefix(filename, stateBloomFilePrefix) && strings.HasSuffix(filename, stateBloomFileSuffix) {
+		return true, common.HexToHash(filename[len(stateBloomFilePrefix)+1 : len(filename)-len(stateBloomFileSuffix)-1])
+	}
+	return false, common.Hash{}
+}
+
+func findBloomFilter(datadir string) (string, common.Hash, error) {
+	var (
+		stateBloomPath string
+		stateBloomRoot common.Hash
+	)
+	if err := filepath.Walk(datadir, func(path string, info os.FileInfo, err error) error {
+		if info != nil && !info.IsDir() {
+			ok, root := isBloomFilter(path)
+			if ok {
+				stateBloomPath = path
+				stateBloomRoot = root
+			}
+		}
+		return nil
+	}); err != nil {
+		return "", common.Hash{}, err
+	}
+	return stateBloomPath, stateBloomRoot, nil
+}
+
+const warningLog = `
+
+WARNING!
+
+The clean trie cache is not found. Please delete it by yourself after the 
+pruning. Remember don't start the Geth without deleting the clean trie cache
+otherwise the entire database may be damaged!
+
+Check the command description "harmony prune-state --help" for more details.
+`
+
+func deleteCleanTrieCache(path string) {
+	if !common.FileExist(path) {
+		log.Warn(warningLog)
+		return
+	}
+	os.RemoveAll(path)
+	log.Info("Deleted trie clean cache", "path", path)
+}

core/state/snapshot/account.go

@@ -29,7 +29,7 @@ import (
 // with a byte slice. This format can be used to represent full-consensus format
 // or slim-snapshot format which replaces the empty root and code hash as nil
 // byte slice.
-type Account struct {
+type SlimAccount struct {
 	Nonce    uint64
 	Balance  *big.Int
 	Root     []byte
@@ -37,8 +37,8 @@ type Account struct {
 }
 
 // SlimAccount converts a state.Account content into a slim snapshot account
-func SlimAccount(nonce uint64, balance *big.Int, root common.Hash, codehash []byte) Account {
-	slim := Account{
+func toSlimAccount(nonce uint64, balance *big.Int, root common.Hash, codehash []byte) SlimAccount {
+	slim := SlimAccount{
 		Nonce:   nonce,
 		Balance: balance,
 	}
@@ -54,7 +54,7 @@ func SlimAccount(nonce uint64, balance *big.Int, root common.Hash, codehash []by
 // SlimAccountRLP converts a state.Account content into a slim snapshot
 // version RLP encoded.
 func SlimAccountRLP(nonce uint64, balance *big.Int, root common.Hash, codehash []byte) []byte {
-	data, err := rlp.EncodeToBytes(SlimAccount(nonce, balance, root, codehash))
+	data, err := rlp.EncodeToBytes(toSlimAccount(nonce, balance, root, codehash))
 	if err != nil {
 		panic(err)
 	}
@@ -63,10 +63,10 @@ func SlimAccountRLP(nonce uint64, balance *big.Int, root common.Hash, codehash [
 
 // FullAccount decodes the data on the 'slim RLP' format and return
 // the consensus format account.
-func FullAccount(data []byte) (Account, error) {
-	var account Account
+func FullAccount(data []byte) (SlimAccount, error) {
+	var account SlimAccount
 	if err := rlp.DecodeBytes(data, &account); err != nil {
-		return Account{}, err
+		return SlimAccount{}, err
 	}
 	if len(account.Root) == 0 {
 		account.Root = types.EmptyRootHash[:]

core/state/snapshot/conversion.go

@@ -78,9 +78,15 @@ func GenerateTrie(snaptree *Tree, root common.Hash, src ethdb.Database, dst ethd
 		if codeHash != types.EmptyCodeHash {
 			code := rawdb.ReadCode(src, codeHash)
 			if len(code) == 0 {
-				return common.Hash{}, errors.New("failed to read code")
+				code = rawdb.ReadValidatorCode(src, codeHash)
+				if len(code) == 0 {
+					return common.Hash{}, errors.New("failed to read code")
+				} else {
+					rawdb.WriteValidatorCode(dst, codeHash, code)
+				}
+			} else {
+				rawdb.WriteCode(dst, codeHash, code)
 			}
-			rawdb.WriteCode(dst, codeHash, code)
 		}
 		// Then migrate all storage trie nodes into the tmp db.
 		storageIt, err := snaptree.StorageIterator(root, accountHash, common.Hash{})

core/state/snapshot/difflayer.go

@@ -17,11 +17,11 @@
 package snapshot
 
 import (
+	"bytes"
 	"encoding/binary"
 	"fmt"
 	"math"
 	"math/rand"
-	"sort"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -29,6 +29,7 @@ import (
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/rlp"
 	bloomfilter "github.com/holiman/bloomfilter/v2"
+	"golang.org/x/exp/slices"
 )
 
 var (
@@ -103,7 +104,7 @@ type diffLayer struct {
 	memory uint64     // Approximate guess as to how much memory we use
 
 	root  common.Hash // Root hash to which this snapshot diff belongs to
-	stale uint32      // Signals that the layer became stale (state progressed)
+	stale atomic.Bool // Signals that the layer became stale (state progressed)
 
 	// destructSet is a very special helper marker. If an account is marked as
 	// deleted, then it's recorded in this set. However it's allowed that an account
@@ -267,12 +268,12 @@ func (dl *diffLayer) Parent() snapshot {
 // Stale return whether this layer has become stale (was flattened across) or if
 // it's still live.
 func (dl *diffLayer) Stale() bool {
-	return atomic.LoadUint32(&dl.stale) != 0
+	return dl.stale.Load()
 }
 
 // Account directly retrieves the account associated with a particular hash in
 // the snapshot slim data format.
-func (dl *diffLayer) Account(hash common.Hash) (*Account, error) {
+func (dl *diffLayer) Account(hash common.Hash) (*SlimAccount, error) {
 	data, err := dl.AccountRLP(hash)
 	if err != nil {
 		return nil, err
@@ -280,7 +281,7 @@ func (dl *diffLayer) Account(hash common.Hash) (*Account, error) {
 	if len(data) == 0 { // can be both nil and []byte{}
 		return nil, nil
 	}
-	account := new(Account)
+	account := new(SlimAccount)
 	if err := rlp.DecodeBytes(data, account); err != nil {
 		panic(err)
 	}
@@ -292,9 +293,14 @@ func (dl *diffLayer) Account(hash common.Hash) (*Account, error) {
 //
 // Note the returned account is not a copy, please don't modify it.
 func (dl *diffLayer) AccountRLP(hash common.Hash) ([]byte, error) {
+	// Check staleness before reaching further.
+	dl.lock.RLock()
+	if dl.Stale() {
+		dl.lock.RUnlock()
+		return nil, ErrSnapshotStale
+	}
 	// Check the bloom filter first whether there's even a point in reaching into
 	// all the maps in all the layers below
-	dl.lock.RLock()
 	hit := dl.diffed.Contains(accountBloomHasher(hash))
 	if !hit {
 		hit = dl.diffed.Contains(destructBloomHasher(hash))
@@ -361,6 +367,11 @@ func (dl *diffLayer) Storage(accountHash, storageHash common.Hash) ([]byte, erro
 	// Check the bloom filter first whether there's even a point in reaching into
 	// all the maps in all the layers below
 	dl.lock.RLock()
+	// Check staleness before reaching further.
+	if dl.Stale() {
+		dl.lock.RUnlock()
+		return nil, ErrSnapshotStale
+	}
 	hit := dl.diffed.Contains(storageBloomHasher{accountHash, storageHash})
 	if !hit {
 		hit = dl.diffed.Contains(destructBloomHasher(accountHash))
@@ -449,7 +460,7 @@ func (dl *diffLayer) flatten() snapshot {
 
 	// Before actually writing all our data to the parent, first ensure that the
 	// parent hasn't been 'corrupted' by someone else already flattening into it
-	if atomic.SwapUint32(&parent.stale, 1) != 0 {
+	if parent.stale.Swap(true) {
 		panic("parent diff layer is stale") // we've flattened into the same parent from two children, boo
 	}
 	// Overwrite all the updated accounts blindly, merge the sorted list
@@ -488,6 +499,11 @@ func (dl *diffLayer) flatten() snapshot {
 	}
 }
 
+// Cmp compares two hashes.
+func Cmp(h common.Hash, other common.Hash) int {
+	return bytes.Compare(h[:], other[:])
+}
+
 // AccountList returns a sorted list of all accounts in this diffLayer, including
 // the deleted ones.
 //
@@ -514,7 +530,7 @@ func (dl *diffLayer) AccountList() []common.Hash {
 			dl.accountList = append(dl.accountList, hash)
 		}
 	}
-	sort.Sort(hashes(dl.accountList))
+	slices.SortFunc(dl.accountList, Cmp)
 	dl.memory += uint64(len(dl.accountList) * common.HashLength)
 	return dl.accountList
 }
@@ -552,7 +568,7 @@ func (dl *diffLayer) StorageList(accountHash common.Hash) ([]common.Hash, bool)
 	for k := range storageMap {
 		storageList = append(storageList, k)
 	}
-	sort.Sort(hashes(storageList))
+	slices.SortFunc(storageList, Cmp)
 	dl.storageList[accountHash] = storageList
 	dl.memory += uint64(len(dl.storageList)*common.HashLength + common.HashLength)
 	return storageList, destructed

core/state/snapshot/disklayer.go

@@ -65,7 +65,7 @@ func (dl *diskLayer) Stale() bool {
 
 // Account directly retrieves the account associated with a particular hash in
 // the snapshot slim data format.
-func (dl *diskLayer) Account(hash common.Hash) (*Account, error) {
+func (dl *diskLayer) Account(hash common.Hash) (*SlimAccount, error) {
 	data, err := dl.AccountRLP(hash)
 	if err != nil {
 		return nil, err
@@ -73,7 +73,7 @@ func (dl *diskLayer) Account(hash common.Hash) (*Account, error) {
 	if len(data) == 0 { // can be both nil and []byte{}
 		return nil, nil
 	}
-	account := new(Account)
+	account := new(SlimAccount)
 	if err := rlp.DecodeBytes(data, account); err != nil {
 		panic(err)
 	}

core/state/snapshot/generate.go

@@ -230,7 +230,7 @@ func (dl *diskLayer) proveRange(ctx *generatorContext, trieId *trie.ID, prefix [
 	if origin == nil && !diskMore {
 		stackTr := trie.NewStackTrie(nil)
 		for i, key := range keys {
-			stackTr.TryUpdate(key, vals[i])
+			stackTr.Update(key, vals[i])
 		}
 		if gotRoot := stackTr.Hash(); gotRoot != root {
 			return &proofResult{

core/state/snapshot/generate_test.go

@@ -48,9 +48,9 @@ func TestGeneration(t *testing.T) {
 	var helper = newHelper()
 	stRoot := helper.makeStorageTrie(common.Hash{}, common.Hash{}, []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, false)
 
-	helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 
 	helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
 	helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
@@ -82,16 +82,16 @@ func TestGenerateExistentState(t *testing.T) {
 	var helper = newHelper()
 
 	stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addSnapAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addSnapAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 	helper.addSnapStorage("acc-1", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"})
 
-	helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addSnapAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addSnapAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
 
 	stRoot = helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addSnapAccount("acc-3", &Account{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addSnapAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 	helper.addSnapStorage("acc-3", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"})
 
 	root, snap := helper.CommitAndGenerate()
@@ -157,18 +157,18 @@ func newHelper() *testHelper {
 	}
 }
 
-func (t *testHelper) addTrieAccount(acckey string, acc *Account) {
+func (t *testHelper) addTrieAccount(acckey string, acc *SlimAccount) {
 	val, _ := rlp.EncodeToBytes(acc)
 	t.accTrie.Update([]byte(acckey), val)
 }
 
-func (t *testHelper) addSnapAccount(acckey string, acc *Account) {
+func (t *testHelper) addSnapAccount(acckey string, acc *SlimAccount) {
 	val, _ := rlp.EncodeToBytes(acc)
 	key := hashData([]byte(acckey))
 	rawdb.WriteAccountSnapshot(t.diskdb, key, val)
 }
 
-func (t *testHelper) addAccount(acckey string, acc *Account) {
+func (t *testHelper) addAccount(acckey string, acc *SlimAccount) {
 	t.addTrieAccount(acckey, acc)
 	t.addSnapAccount(acckey, acc)
 }
@@ -233,70 +233,70 @@ func (t *testHelper) CommitAndGenerate() (common.Hash, *diskLayer) {
 func TestGenerateExistentStateWithWrongStorage(t *testing.T) {
 	helper := newHelper()
 
-	// Account one, empty root but non-empty database
-	helper.addAccount("acc-1", &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+	// SlimAccount one, empty root but non-empty database
+	helper.addAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
 	helper.addSnapStorage("acc-1", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"})
 
-	// Account two, non empty root but empty database
+	// SlimAccount two, non empty root but empty database
 	stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-2")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addAccount("acc-2", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addAccount("acc-2", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 
 	// Miss slots
 	{
-		// Account three, non empty root but misses slots in the beginning
+		// SlimAccount three, non empty root but misses slots in the beginning
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-3", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-3", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-3", []string{"key-2", "key-3"}, []string{"val-2", "val-3"})
 
-		// Account four, non empty root but misses slots in the middle
+		// SlimAccount four, non empty root but misses slots in the middle
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-4")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-4", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-4", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-4", []string{"key-1", "key-3"}, []string{"val-1", "val-3"})
 
-		// Account five, non empty root but misses slots in the end
+		// SlimAccount five, non empty root but misses slots in the end
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-5")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-5", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-5", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-5", []string{"key-1", "key-2"}, []string{"val-1", "val-2"})
 	}
 
 	// Wrong storage slots
 	{
-		// Account six, non empty root but wrong slots in the beginning
+		// SlimAccount six, non empty root but wrong slots in the beginning
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-6")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-6", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-6", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-6", []string{"key-1", "key-2", "key-3"}, []string{"badval-1", "val-2", "val-3"})
 
-		// Account seven, non empty root but wrong slots in the middle
+		// SlimAccount seven, non empty root but wrong slots in the middle
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-7")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-7", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-7", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-7", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "badval-2", "val-3"})
 
-		// Account eight, non empty root but wrong slots in the end
+		// SlimAccount eight, non empty root but wrong slots in the end
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-8")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-8", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-8", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-8", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "badval-3"})
 
-		// Account 9, non empty root but rotated slots
+		// SlimAccount 9, non empty root but rotated slots
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-9")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-9", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-9", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-9", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-3", "val-2"})
 	}
 
 	// Extra storage slots
 	{
-		// Account 10, non empty root but extra slots in the beginning
+		// SlimAccount 10, non empty root but extra slots in the beginning
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-10")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-10", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-10", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-10", []string{"key-0", "key-1", "key-2", "key-3"}, []string{"val-0", "val-1", "val-2", "val-3"})
 
-		// Account 11, non empty root but extra slots in the middle
+		// SlimAccount 11, non empty root but extra slots in the middle
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-11")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-11", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-11", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-11", []string{"key-1", "key-2", "key-2-1", "key-3"}, []string{"val-1", "val-2", "val-2-1", "val-3"})
 
-		// Account 12, non empty root but extra slots in the end
+		// SlimAccount 12, non empty root but extra slots in the end
 		helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-12")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-		helper.addAccount("acc-12", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount("acc-12", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		helper.addSnapStorage("acc-12", []string{"key-1", "key-2", "key-3", "key-4"}, []string{"val-1", "val-2", "val-3", "val-4"})
 	}
 
@@ -336,25 +336,25 @@ func TestGenerateExistentStateWithWrongAccounts(t *testing.T) {
 
 	// Missing accounts, only in the trie
 	{
-		helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // Beginning
-		helper.addTrieAccount("acc-4", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // Middle
-		helper.addTrieAccount("acc-6", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // End
+		helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // Beginning
+		helper.addTrieAccount("acc-4", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // Middle
+		helper.addTrieAccount("acc-6", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // End
 	}
 
 	// Wrong accounts
 	{
-		helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-		helper.addSnapAccount("acc-2", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: common.Hex2Bytes("0x1234")})
+		helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addSnapAccount("acc-2", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: common.Hex2Bytes("0x1234")})
 
-		helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-		helper.addSnapAccount("acc-3", &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addSnapAccount("acc-3", &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
 	}
 
 	// Extra accounts, only in the snap
 	{
-		helper.addSnapAccount("acc-0", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})                      // before the beginning
-		helper.addSnapAccount("acc-5", &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: common.Hex2Bytes("0x1234")})  // Middle
-		helper.addSnapAccount("acc-7", &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // after the end
+		helper.addSnapAccount("acc-0", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})                      // before the beginning
+		helper.addSnapAccount("acc-5", &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: common.Hex2Bytes("0x1234")})  // Middle
+		helper.addSnapAccount("acc-7", &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // after the end
 	}
 
 	root, snap := helper.CommitAndGenerate()
@@ -383,9 +383,9 @@ func TestGenerateCorruptAccountTrie(t *testing.T) {
 	// without any storage slots to keep the test smaller.
 	helper := newHelper()
 
-	helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // 0xc7a30f39aff471c95d8a837497ad0e49b65be475cc0953540f80cfcdbdcd9074
-	helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // 0x65145f923027566669a1ae5ccac66f945b55ff6eaeb17d2ea8e048b7d381f2d7
-	helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(3), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // 0x19ead688e907b0fab07176120dceec244a72aff2f0aa51e8b827584e378772f4
+	helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // 0xc7a30f39aff471c95d8a837497ad0e49b65be475cc0953540f80cfcdbdcd9074
+	helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // 0x65145f923027566669a1ae5ccac66f945b55ff6eaeb17d2ea8e048b7d381f2d7
+	helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}) // 0x19ead688e907b0fab07176120dceec244a72aff2f0aa51e8b827584e378772f4
 
 	root := helper.Commit() // Root: 0xa04693ea110a31037fb5ee814308a6f1d76bdab0b11676bdf4541d2de55ba978
 
@@ -418,10 +418,10 @@ func TestGenerateMissingStorageTrie(t *testing.T) {
 	helper := newHelper()
 
 	stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true) // 0xddefcd9376dd029653ef384bd2f0a126bb755fe84fdcc9e7cf421ba454f2bc67
-	helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})                                      // 0x9250573b9c18c664139f3b6a7a8081b7d8f8916a8fcc5d94feec6c29f5fd4e9e
-	helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})                 // 0x65145f923027566669a1ae5ccac66f945b55ff6eaeb17d2ea8e048b7d381f2d7
+	helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})                                  // 0x9250573b9c18c664139f3b6a7a8081b7d8f8916a8fcc5d94feec6c29f5fd4e9e
+	helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})             // 0x65145f923027566669a1ae5ccac66f945b55ff6eaeb17d2ea8e048b7d381f2d7
 	stRoot = helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // 0x50815097425d000edfc8b3a4a13e175fc2bdcfee8bdfbf2d1ff61041d3c235b2
+	helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // 0x50815097425d000edfc8b3a4a13e175fc2bdcfee8bdfbf2d1ff61041d3c235b2
 
 	root := helper.Commit()
 
@@ -452,10 +452,10 @@ func TestGenerateCorruptStorageTrie(t *testing.T) {
 	helper := newHelper()
 
 	stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true) // 0xddefcd9376dd029653ef384bd2f0a126bb755fe84fdcc9e7cf421ba454f2bc67
-	helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})                                      // 0x9250573b9c18c664139f3b6a7a8081b7d8f8916a8fcc5d94feec6c29f5fd4e9e
-	helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})                 // 0x65145f923027566669a1ae5ccac66f945b55ff6eaeb17d2ea8e048b7d381f2d7
+	helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})                                  // 0x9250573b9c18c664139f3b6a7a8081b7d8f8916a8fcc5d94feec6c29f5fd4e9e
+	helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})             // 0x65145f923027566669a1ae5ccac66f945b55ff6eaeb17d2ea8e048b7d381f2d7
 	stRoot = helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // 0x50815097425d000edfc8b3a4a13e175fc2bdcfee8bdfbf2d1ff61041d3c235b2
+	helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}) // 0x50815097425d000edfc8b3a4a13e175fc2bdcfee8bdfbf2d1ff61041d3c235b2
 
 	root := helper.Commit()
 
@@ -481,13 +481,13 @@ func TestGenerateCorruptStorageTrie(t *testing.T) {
 func TestGenerateWithExtraAccounts(t *testing.T) {
 	helper := newHelper()
 	{
-		// Account one in the trie
+		// SlimAccount one in the trie
 		stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")),
 			[]string{"key-1", "key-2", "key-3", "key-4", "key-5"},
 			[]string{"val-1", "val-2", "val-3", "val-4", "val-5"},
 			true,
 		)
-		acc := &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}
+		acc := &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}
 		val, _ := rlp.EncodeToBytes(acc)
 		helper.accTrie.Update([]byte("acc-1"), val) // 0x9250573b9c18c664139f3b6a7a8081b7d8f8916a8fcc5d94feec6c29f5fd4e9e
 
@@ -501,13 +501,13 @@ func TestGenerateWithExtraAccounts(t *testing.T) {
 		rawdb.WriteStorageSnapshot(helper.diskdb, key, hashData([]byte("key-5")), []byte("val-5"))
 	}
 	{
-		// Account two exists only in the snapshot
+		// SlimAccount two exists only in the snapshot
 		stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-2")),
 			[]string{"key-1", "key-2", "key-3", "key-4", "key-5"},
 			[]string{"val-1", "val-2", "val-3", "val-4", "val-5"},
 			true,
 		)
-		acc := &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}
+		acc := &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}
 		val, _ := rlp.EncodeToBytes(acc)
 		key := hashData([]byte("acc-2"))
 		rawdb.WriteAccountSnapshot(helper.diskdb, key, val)
@@ -545,13 +545,13 @@ func TestGenerateWithExtraAccounts(t *testing.T) {
 func TestGenerateWithManyExtraAccounts(t *testing.T) {
 	helper := newHelper()
 	{
-		// Account one in the trie
+		// SlimAccount one in the trie
 		stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")),
 			[]string{"key-1", "key-2", "key-3"},
 			[]string{"val-1", "val-2", "val-3"},
 			true,
 		)
-		acc := &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}
+		acc := &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()}
 		val, _ := rlp.EncodeToBytes(acc)
 		helper.accTrie.Update([]byte("acc-1"), val) // 0x9250573b9c18c664139f3b6a7a8081b7d8f8916a8fcc5d94feec6c29f5fd4e9e
 
@@ -565,7 +565,7 @@ func TestGenerateWithManyExtraAccounts(t *testing.T) {
 	{
 		// 100 accounts exist only in snapshot
 		for i := 0; i < 1000; i++ {
-			acc := &Account{Balance: big.NewInt(int64(i)), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}
+			acc := &SlimAccount{Balance: big.NewInt(int64(i)), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}
 			val, _ := rlp.EncodeToBytes(acc)
 			key := hashData([]byte(fmt.Sprintf("acc-%d", i)))
 			rawdb.WriteAccountSnapshot(helper.diskdb, key, val)
@@ -599,7 +599,7 @@ func TestGenerateWithExtraBeforeAndAfter(t *testing.T) {
 	accountCheckRange = 3
 	helper := newHelper()
 	{
-		acc := &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}
+		acc := &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}
 		val, _ := rlp.EncodeToBytes(acc)
 		helper.accTrie.Update(common.HexToHash("0x03").Bytes(), val)
 		helper.accTrie.Update(common.HexToHash("0x07").Bytes(), val)
@@ -633,7 +633,7 @@ func TestGenerateWithMalformedSnapdata(t *testing.T) {
 	accountCheckRange = 3
 	helper := newHelper()
 	{
-		acc := &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}
+		acc := &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()}
 		val, _ := rlp.EncodeToBytes(acc)
 		helper.accTrie.Update(common.HexToHash("0x03").Bytes(), val)
 
@@ -672,7 +672,7 @@ func TestGenerateFromEmptySnap(t *testing.T) {
 	for i := 0; i < 400; i++ {
 		stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte(fmt.Sprintf("acc-%d", i))), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
 		helper.addTrieAccount(fmt.Sprintf("acc-%d", i),
-			&Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+			&SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 	}
 	root, snap := helper.CommitAndGenerate()
 	t.Logf("Root: %#x\n", root) // Root: 0x6f7af6d2e1a1bf2b84a3beb3f8b64388465fbc1e274ca5d5d3fc787ca78f59e4
@@ -709,7 +709,7 @@ func TestGenerateWithIncompleteStorage(t *testing.T) {
 	for i := 0; i < 8; i++ {
 		accKey := fmt.Sprintf("acc-%d", i)
 		stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte(accKey)), stKeys, stVals, true)
-		helper.addAccount(accKey, &Account{Balance: big.NewInt(int64(i)), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+		helper.addAccount(accKey, &SlimAccount{Balance: big.NewInt(int64(i)), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 		var moddedKeys []string
 		var moddedVals []string
 		for ii := 0; ii < 8; ii++ {
@@ -801,11 +801,11 @@ func TestGenerateCompleteSnapshotWithDanglingStorage(t *testing.T) {
 	var helper = newHelper()
 
 	stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addAccount("acc-2", &Account{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addAccount("acc-2", &SlimAccount{Balance: big.NewInt(1), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
 
 	helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addAccount("acc-3", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addAccount("acc-3", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 
 	helper.addSnapStorage("acc-1", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"})
 	helper.addSnapStorage("acc-3", []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"})
@@ -836,11 +836,11 @@ func TestGenerateBrokenSnapshotWithDanglingStorage(t *testing.T) {
 	var helper = newHelper()
 
 	stRoot := helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-1")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addTrieAccount("acc-1", &Account{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
-	helper.addTrieAccount("acc-2", &Account{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-1", &SlimAccount{Balance: big.NewInt(1), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-2", &SlimAccount{Balance: big.NewInt(2), Root: types.EmptyRootHash.Bytes(), CodeHash: types.EmptyCodeHash.Bytes()})
 
 	helper.makeStorageTrie(common.Hash{}, hashData([]byte("acc-3")), []string{"key-1", "key-2", "key-3"}, []string{"val-1", "val-2", "val-3"}, true)
-	helper.addTrieAccount("acc-3", &Account{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
+	helper.addTrieAccount("acc-3", &SlimAccount{Balance: big.NewInt(3), Root: stRoot, CodeHash: types.EmptyCodeHash.Bytes()})
 
 	populateDangling(helper.diskdb)
 

core/state/snapshot/iterator_fast.go

@@ -22,6 +22,7 @@ import (
 	"sort"
 
 	"github.com/ethereum/go-ethereum/common"
+	"golang.org/x/exp/slices"
 )
 
 // weightedIterator is a iterator with an assigned weight. It is used to prioritise
@@ -32,32 +33,25 @@ type weightedIterator struct {
 	priority int
 }
 
-// weightedIterators is a set of iterators implementing the sort.Interface.
-type weightedIterators []*weightedIterator
-
-// Len implements sort.Interface, returning the number of active iterators.
-func (its weightedIterators) Len() int { return len(its) }
-
-// Less implements sort.Interface, returning which of two iterators in the stack
-// is before the other.
-func (its weightedIterators) Less(i, j int) bool {
+func (it *weightedIterator) Cmp(other *weightedIterator) int {
 	// Order the iterators primarily by the account hashes
-	hashI := its[i].it.Hash()
-	hashJ := its[j].it.Hash()
+	hashI := it.it.Hash()
+	hashJ := other.it.Hash()
 
 	switch bytes.Compare(hashI[:], hashJ[:]) {
 	case -1:
-		return true
+		return -1
 	case 1:
-		return false
+		return 1
 	}
 	// Same account/storage-slot in multiple layers, split by priority
-	return its[i].priority < its[j].priority
-}
-
-// Swap implements sort.Interface, swapping two entries in the iterator stack.
-func (its weightedIterators) Swap(i, j int) {
-	its[i], its[j] = its[j], its[i]
+	if it.priority < other.priority {
+		return -1
+	}
+	if it.priority > other.priority {
+		return 1
+	}
+	return 0
 }
 
 // fastIterator is a more optimized multi-layer iterator which maintains a
@@ -69,7 +63,7 @@ type fastIterator struct {
 	curAccount []byte
 	curSlot    []byte
 
-	iterators weightedIterators
+	iterators []*weightedIterator
 	initiated bool
 	account   bool
 	fail      error
@@ -167,7 +161,7 @@ func (fi *fastIterator) init() {
 		}
 	}
 	// Re-sort the entire list
-	sort.Sort(fi.iterators)
+	slices.SortFunc(fi.iterators, func(a, b *weightedIterator) int { return a.Cmp(b) })
 	fi.initiated = false
 }
 

core/state/snapshot/snapshot.go

@@ -22,7 +22,6 @@ import (
 	"errors"
 	"fmt"
 	"sync"
-	"sync/atomic"
 
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/ethdb"
@@ -103,7 +102,7 @@ type Snapshot interface {
 
 	// Account directly retrieves the account associated with a particular hash in
 	// the snapshot slim data format.
-	Account(hash common.Hash) (*Account, error)
+	Account(hash common.Hash) (*SlimAccount, error)
 
 	// AccountRLP directly retrieves the account RLP associated with a particular
 	// hash in the snapshot slim data format.
@@ -206,8 +205,7 @@ func New(config Config, diskdb ethdb.KeyValueStore, triedb *trie.Database, root
 		utils.Logger().Warn().Err(err).Msg("Snapshot maintenance disabled (syncing)")
 		return snap, nil
 	}
-
-	// Create the building waiter if the background generation is allowed
+	// Create the building waiter iff the background generation is allowed
 	if !config.NoBuild && !config.AsyncBuild {
 		defer snap.waitBuild()
 	}
@@ -273,7 +271,7 @@ func (t *Tree) Disable() {
 		case *diffLayer:
 			// If the layer is a simple diff, simply mark as stale
 			layer.lock.Lock()
-			atomic.StoreUint32(&layer.stale, 1)
+			layer.stale.Store(true)
 			layer.lock.Unlock()
 
 		default:
@@ -565,7 +563,7 @@ func diffToDisk(bottom *diffLayer) *diskLayer {
 			// Ensure we don't delete too much data blindly (contract can be
 			// huge). It's ok to flush, the root will go missing in case of a
 			// crash and we'll detect and regenerate the snapshot.
-			if batch.ValueSize() > ethdb.IdealBatchSize {
+			if batch.ValueSize() > 64*1024*1024 {
 				if err := batch.Write(); err != nil {
 					utils.Logger().Fatal().Err(err).Msg("Failed to write storage deletions")
 				}
@@ -591,7 +589,7 @@ func diffToDisk(bottom *diffLayer) *diskLayer {
 		// Ensure we don't write too much data blindly. It's ok to flush, the
 		// root will go missing in case of a crash and we'll detect and regen
 		// the snapshot.
-		if batch.ValueSize() > ethdb.IdealBatchSize {
+		if batch.ValueSize() > 64*1024*1024 {
 			if err := batch.Write(); err != nil {
 				utils.Logger().Fatal().Err(err).Msg("Failed to write storage deletions")
 			}
@@ -727,7 +725,7 @@ func (t *Tree) Rebuild(root common.Hash) {
 		case *diffLayer:
 			// If the layer is a simple diff, simply mark as stale
 			layer.lock.Lock()
-			atomic.StoreUint32(&layer.stale, 1)
+			layer.stale.Store(true)
 			layer.lock.Unlock()
 
 		default:
@@ -853,3 +851,21 @@ func (t *Tree) DiskRoot() common.Hash {
 
 	return t.diskRoot()
 }
+
+// Size returns the memory usage of the diff layers above the disk layer and the
+// dirty nodes buffered in the disk layer. Currently, the implementation uses a
+// special diff layer (the first) as an aggregator simulating a dirty buffer, so
+// the second return will always be 0. However, this will be made consistent with
+// the pathdb, which will require a second return.
+func (t *Tree) Size() (diffs common.StorageSize, buf common.StorageSize) {
+	t.lock.RLock()
+	defer t.lock.RUnlock()
+
+	var size common.StorageSize
+	for _, layer := range t.layers {
+		if layer, ok := layer.(*diffLayer); ok {
+			size += common.StorageSize(layer.memory)
+		}
+	}
+	return size, 0
+}

core/state/snapshot/snapshot_test.go

@@ -43,11 +43,10 @@ func randomHash() common.Hash {
 
 // randomAccount generates a random account and returns it RLP encoded.
 func randomAccount() []byte {
-	root := randomHash()
-	a := Account{
+	a := &types.StateAccount{
 		Balance:  big.NewInt(rand.Int63()),
 		Nonce:    rand.Uint64(),
-		Root:     root[:],
+		Root:     randomHash(),
 		CodeHash: types.EmptyCodeHash[:],
 	}
 	data, _ := rlp.EncodeToBytes(a)
@@ -118,7 +117,7 @@ func TestDiskLayerExternalInvalidationFullFlatten(t *testing.T) {
 	if err := snaps.Cap(common.HexToHash("0x02"), 0); err != nil {
 		t.Fatalf("failed to merge diff layer onto disk: %v", err)
 	}
-	// Since the base layer was modified, ensure that data retrieval on the external reference fail
+	// Since the base layer was modified, ensure that data retrievals on the external reference fail
 	if acc, err := ref.Account(common.HexToHash("0x01")); err != ErrSnapshotStale {
 		t.Errorf("stale reference returned account: %#x (err: %v)", acc, err)
 	}
@@ -185,6 +184,10 @@ func TestDiskLayerExternalInvalidationPartialFlatten(t *testing.T) {
 // be returned with junk data. This version of the test retains the bottom diff
 // layer to check the usual mode of operation where the accumulator is retained.
 func TestDiffLayerExternalInvalidationPartialFlatten(t *testing.T) {
+	// Un-commenting this triggers the bloom set to be deterministic. The values below
+	// were used to trigger the flaw described in https://github.com/ethereum/go-ethereum/issues/27254.
+	// bloomDestructHasherOffset, bloomAccountHasherOffset, bloomStorageHasherOffset = 14, 24, 5
+
 	// Create an empty base layer and a snapshot tree out of it
 	base := &diskLayer{
 		diskdb: rawdb.NewMemoryDatabase(),
@@ -461,7 +464,7 @@ func TestReadStateDuringFlattening(t *testing.T) {
 	snap := snaps.Snapshot(common.HexToHash("0xa3"))
 
 	// Register the testing hook to access the state after flattening
-	var result = make(chan *Account)
+	var result = make(chan *SlimAccount)
 	snaps.onFlatten = func() {
 		// Spin up a thread to read the account from the pre-created
 		// snapshot handler. It's expected to be blocked.

core/state/snapshot/utils.go

@@ -23,7 +23,6 @@ import (
 
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/ethdb"
-	"github.com/ethereum/go-ethereum/rlp"
 	"github.com/harmony-one/harmony/core/rawdb"
 	"github.com/harmony-one/harmony/internal/utils"
 )
@@ -111,8 +110,8 @@ func CheckJournalAccount(db ethdb.KeyValueStore, hash common.Hash) error {
 	baseRoot := rawdb.ReadSnapshotRoot(db)
 	fmt.Printf("Disklayer: Root: %x\n", baseRoot)
 	if data := rawdb.ReadAccountSnapshot(db, hash); data != nil {
-		account := new(Account)
-		if err := rlp.DecodeBytes(data, account); err != nil {
+		account, err := FullAccount(data)
+		if err != nil {
 			panic(err)
 		}
 		fmt.Printf("\taccount.nonce: %d\n", account.Nonce)
@@ -142,8 +141,8 @@ func CheckJournalAccount(db ethdb.KeyValueStore, hash common.Hash) error {
 		}
 		fmt.Printf("Disklayer+%d: Root: %x, parent %x\n", depth, root, pRoot)
 		if data, ok := accounts[hash]; ok {
-			account := new(Account)
-			if err := rlp.DecodeBytes(data, account); err != nil {
+			account, err := FullAccount(data)
+			if err != nil {
 				panic(err)
 			}
 			fmt.Printf("\taccount.nonce: %d\n", account.Nonce)