path: root/core/state/snapshot/difflayer.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 snapshot

import (
	"encoding/binary"
	"fmt"
	"math"
	"math/rand"
	"sort"
	"sync"
	"sync/atomic"
	"time"

	"github.com/ethereum/go-ethereum/common"
	"github.com/ethereum/go-ethereum/rlp"
	"github.com/steakknife/bloomfilter"
)

var (
	// aggregatorMemoryLimit is the maximum size of the bottom-most diff layer
	// that aggregates the writes from above until it's flushed into the disk
	// layer.
	//
	// Note, bumping this up might drastically increase the size of the bloom
	// filters that's stored in every diff layer. Don't do that without fully
	// understanding all the implications.
	aggregatorMemoryLimit = uint64(4 * 1024 * 1024)

	// aggregatorItemLimit is an approximate number of items that will end up
	// in the agregator layer before it's flushed out to disk. A plain account
	// weighs around 14B (+hash), a storage slot 32B (+hash), a deleted slot
	// 0B (+hash). Slots are mostly set/unset in lockstep, so thet average at
	// 16B (+hash). All in all, the average entry seems to be 15+32=47B. Use a
	// smaller number to be on the safe side.
	aggregatorItemLimit = aggregatorMemoryLimit / 42

	// bloomTargetError is the target false positive rate when the aggregator
	// layer is at its fullest. The actual value will probably move around up
	// and down from this number, it's mostly a ballpark figure.
	//
	// Note, dropping this down might drastically increase the size of the bloom
	// filters that's stored in every diff layer. Don't do that without fully
	// understanding all the implications.
	bloomTargetError = 0.02

	// bloomSize is the ideal bloom filter size given the maximum number of items
	// it's expected to hold and the target false positive error rate.
	bloomSize = math.Ceil(float64(aggregatorItemLimit) * math.Log(bloomTargetError) / math.Log(1/math.Pow(2, math.Log(2))))

	// bloomFuncs is the ideal number of bits a single entry should set in the
	// bloom filter to keep its size to a minimum (given it's size and maximum
	// entry count).
	bloomFuncs = math.Round((bloomSize / float64(aggregatorItemLimit)) * math.Log(2))

	// the bloom offsets are runtime constants which determines which part of the
	// the account/storage hash the hasher functions looks at, to determine the
	// bloom key for an account/slot. This is randomized at init(), so that the
	// global population of nodes do not all display the exact same behaviour with
	// regards to bloom content
	bloomDestructHasherOffset = 0
	bloomAccountHasherOffset  = 0
	bloomStorageHasherOffset  = 0
)

func init() {
	// Init the bloom offsets in the range [0:24] (requires 8 bytes)
	bloomDestructHasherOffset = rand.Intn(25)
	bloomAccountHasherOffset = rand.Intn(25)
	bloomStorageHasherOffset = rand.Intn(25)

	// The destruct and account blooms must be different, as the storage slots
	// will check for destruction too for every bloom miss. It should not collide
	// with modified accounts.
	for bloomAccountHasherOffset == bloomDestructHasherOffset {
		bloomAccountHasherOffset = rand.Intn(25)
	}
}

// diffLayer represents a collection of modifications made to a state snapshot
// after running a block on top. It contains one sorted list for the account trie
// and one-one list for each storage tries.
//
// The goal of a diff layer is to act as a journal, tracking recent modifications
// made to the state, that have not yet graduated into a semi-immutable state.
type diffLayer struct {
	origin *diskLayer // Base disk layer to directly use on bloom misses
	parent snapshot   // Parent snapshot modified by this one, never nil
	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)

	// 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
	// is included here but still available in other sets(e.g. storageData). The
	// reason is the diff layer includes all the changes in a *block*. It can
	// happen that in the tx_1, account A is self-destructed while in the tx_2
	// it's recreated. But we still need this marker to indicate the "old" A is
	// deleted, all data in other set belongs to the "new" A.
	destructSet map[common.Hash]struct{}               // Keyed markers for deleted (and potentially) recreated accounts
	accountList []common.Hash                          // List of account for iteration. If it exists, it's sorted, otherwise it's nil
	accountData map[common.Hash][]byte                 // Keyed accounts for direct retrival (nil means deleted)
	storageList map[common.Hash][]common.Hash          // List of storage slots for iterated retrievals, one per account. Any existing lists are sorted if non-nil
	storageData map[common.Hash]map[common.Hash][]byte // Keyed storage slots for direct retrival. one per account (nil means deleted)

	diffed *bloomfilter.Filter // Bloom filter tracking all the diffed items up to the disk layer

	lock sync.RWMutex
}

// destructBloomHasher is a wrapper around a common.Hash to satisfy the interface
// API requirements of the bloom library used. It's used to convert a destruct
// event into a 64 bit mini hash.
type destructBloomHasher common.Hash

func (h destructBloomHasher) Write(p []byte) (n int, err error) { panic("not implemented") }
func (h destructBloomHasher) Sum(b []byte) []byte               { panic("not implemented") }
func (h destructBloomHasher) Reset()                            { panic("not implemented") }
func (h destructBloomHasher) BlockSize() int                    { panic("not implemented") }
func (h destructBloomHasher) Size() int                         { return 8 }
func (h destructBloomHasher) Sum64() uint64 {
	return binary.BigEndian.Uint64(h[bloomDestructHasherOffset : bloomDestructHasherOffset+8])
}

// accountBloomHasher is a wrapper around a common.Hash to satisfy the interface
// API requirements of the bloom library used. It's used to convert an account
// hash into a 64 bit mini hash.
type accountBloomHasher common.Hash

func (h accountBloomHasher) Write(p []byte) (n int, err error) { panic("not implemented") }
func (h accountBloomHasher) Sum(b []byte) []byte               { panic("not implemented") }
func (h accountBloomHasher) Reset()                            { panic("not implemented") }
func (h accountBloomHasher) BlockSize() int                    { panic("not implemented") }
func (h accountBloomHasher) Size() int                         { return 8 }
func (h accountBloomHasher) Sum64() uint64 {
	return binary.BigEndian.Uint64(h[bloomAccountHasherOffset : bloomAccountHasherOffset+8])
}

// storageBloomHasher is a wrapper around a [2]common.Hash to satisfy the interface
// API requirements of the bloom library used. It's used to convert an account
// hash into a 64 bit mini hash.
type storageBloomHasher [2]common.Hash

func (h storageBloomHasher) Write(p []byte) (n int, err error) { panic("not implemented") }
func (h storageBloomHasher) Sum(b []byte) []byte               { panic("not implemented") }
func (h storageBloomHasher) Reset()                            { panic("not implemented") }
func (h storageBloomHasher) BlockSize() int                    { panic("not implemented") }
func (h storageBloomHasher) Size() int                         { return 8 }
func (h storageBloomHasher) Sum64() uint64 {
	return binary.BigEndian.Uint64(h[0][bloomStorageHasherOffset:bloomStorageHasherOffset+8]) ^
		binary.BigEndian.Uint64(h[1][bloomStorageHasherOffset:bloomStorageHasherOffset+8])
}

// newDiffLayer creates a new diff on top of an existing snapshot, whether that's a low
// level persistent database or a hierarchical diff already.
func newDiffLayer(parent snapshot, root common.Hash, destructs map[common.Hash]struct{}, accounts map[common.Hash][]byte, storage map[common.Hash]map[common.Hash][]byte) *diffLayer {
	// Create the new layer with some pre-allocated data segments
	dl := &diffLayer{
		parent:      parent,
		root:        root,
		destructSet: destructs,
		accountData: accounts,
		storageData: storage,
		storageList: make(map[common.Hash][]common.Hash),
	}
	switch parent := parent.(type) {
	case *diskLayer:
		dl.rebloom(parent)
	case *diffLayer:
		dl.rebloom(parent.origin)
	default:
		panic("unknown parent type")
	}
	// Sanity check that accounts or storage slots are never nil
	for accountHash, blob := range accounts {
		if blob == nil {
			panic(fmt.Sprintf("account %#x nil", accountHash))
		}
	}
	for accountHash, slots := range storage {
		if slots == nil {
			panic(fmt.Sprintf("storage %#x nil", accountHash))
		}
	}
	// Determine memory size and track the dirty writes
	for _, data := range accounts {
		dl.memory += uint64(common.HashLength + len(data))
		snapshotDirtyAccountWriteMeter.Mark(int64(len(data)))
	}
	// Determine memory size and track the dirty writes
	for _, slots := range storage {
		for _, data := range slots {
			dl.memory += uint64(common.HashLength + len(data))