go-ethereum/internal/era/builder.go
lightclient 85938dda09
internal/era: update block index format to be based on record offset (#28959)
As mentioned in #26621, the block index format for era1 is not in line with the regular era block index. This change modifies the index so all relative offsets are based against the beginning of the block index record.
2024-02-09 07:42:50 +01:00

225 lines
7.3 KiB
Go

// Copyright 2023 The go-ethereum Authors
// This file is part of go-ethereum.
//
// go-ethereum is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// go-ethereum 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 General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with go-ethereum. If not, see <http://www.gnu.org/licenses/>.
package era
import (
"bytes"
"encoding/binary"
"fmt"
"io"
"math/big"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/internal/era/e2store"
"github.com/ethereum/go-ethereum/rlp"
"github.com/golang/snappy"
)
// Builder is used to create Era1 archives of block data.
//
// Era1 files are themselves e2store files. For more information on this format,
// see https://github.com/status-im/nimbus-eth2/blob/stable/docs/e2store.md.
//
// The overall structure of an Era1 file follows closely the structure of an Era file
// which contains consensus Layer data (and as a byproduct, EL data after the merge).
//
// The structure can be summarized through this definition:
//
// era1 := Version | block-tuple* | other-entries* | Accumulator | BlockIndex
// block-tuple := CompressedHeader | CompressedBody | CompressedReceipts | TotalDifficulty
//
// Each basic element is its own entry:
//
// Version = { type: [0x65, 0x32], data: nil }
// CompressedHeader = { type: [0x03, 0x00], data: snappyFramed(rlp(header)) }
// CompressedBody = { type: [0x04, 0x00], data: snappyFramed(rlp(body)) }
// CompressedReceipts = { type: [0x05, 0x00], data: snappyFramed(rlp(receipts)) }
// TotalDifficulty = { type: [0x06, 0x00], data: uint256(header.total_difficulty) }
// AccumulatorRoot = { type: [0x07, 0x00], data: accumulator-root }
// BlockIndex = { type: [0x32, 0x66], data: block-index }
//
// Accumulator is computed by constructing an SSZ list of header-records of length at most
// 8192 and then calculating the hash_tree_root of that list.
//
// header-record := { block-hash: Bytes32, total-difficulty: Uint256 }
// accumulator := hash_tree_root([]header-record, 8192)
//
// BlockIndex stores relative offsets to each compressed block entry. The
// format is:
//
// block-index := starting-number | index | index | index ... | count
//
// starting-number is the first block number in the archive. Every index is a
// defined relative to beginning of the record. The total number of block
// entries in the file is recorded with count.
//
// Due to the accumulator size limit of 8192, the maximum number of blocks in
// an Era1 batch is also 8192.
type Builder struct {
w *e2store.Writer
startNum *uint64
startTd *big.Int
indexes []uint64
hashes []common.Hash
tds []*big.Int
written int
buf *bytes.Buffer
snappy *snappy.Writer
}
// NewBuilder returns a new Builder instance.
func NewBuilder(w io.Writer) *Builder {
buf := bytes.NewBuffer(nil)
return &Builder{
w: e2store.NewWriter(w),
buf: buf,
snappy: snappy.NewBufferedWriter(buf),
}
}
// Add writes a compressed block entry and compressed receipts entry to the
// underlying e2store file.
func (b *Builder) Add(block *types.Block, receipts types.Receipts, td *big.Int) error {
eh, err := rlp.EncodeToBytes(block.Header())
if err != nil {
return err
}
eb, err := rlp.EncodeToBytes(block.Body())
if err != nil {
return err
}
er, err := rlp.EncodeToBytes(receipts)
if err != nil {
return err
}
return b.AddRLP(eh, eb, er, block.NumberU64(), block.Hash(), td, block.Difficulty())
}
// AddRLP writes a compressed block entry and compressed receipts entry to the
// underlying e2store file.
func (b *Builder) AddRLP(header, body, receipts []byte, number uint64, hash common.Hash, td, difficulty *big.Int) error {
// Write Era1 version entry before first block.
if b.startNum == nil {
n, err := b.w.Write(TypeVersion, nil)
if err != nil {
return err
}
startNum := number
b.startNum = &startNum
b.startTd = new(big.Int).Sub(td, difficulty)
b.written += n
}
if len(b.indexes) >= MaxEra1Size {
return fmt.Errorf("exceeds maximum batch size of %d", MaxEra1Size)
}
b.indexes = append(b.indexes, uint64(b.written))
b.hashes = append(b.hashes, hash)
b.tds = append(b.tds, td)
// Write block data.
if err := b.snappyWrite(TypeCompressedHeader, header); err != nil {
return err
}
if err := b.snappyWrite(TypeCompressedBody, body); err != nil {
return err
}
if err := b.snappyWrite(TypeCompressedReceipts, receipts); err != nil {
return err
}
// Also write total difficulty, but don't snappy encode.
btd := bigToBytes32(td)
n, err := b.w.Write(TypeTotalDifficulty, btd[:])
b.written += n
if err != nil {
return err
}
return nil
}
// Finalize computes the accumulator and block index values, then writes the
// corresponding e2store entries.
func (b *Builder) Finalize() (common.Hash, error) {
if b.startNum == nil {
return common.Hash{}, fmt.Errorf("finalize called on empty builder")
}
// Compute accumulator root and write entry.
root, err := ComputeAccumulator(b.hashes, b.tds)
if err != nil {
return common.Hash{}, fmt.Errorf("error calculating accumulator root: %w", err)
}
n, err := b.w.Write(TypeAccumulator, root[:])
b.written += n
if err != nil {
return common.Hash{}, fmt.Errorf("error writing accumulator: %w", err)
}
// Get beginning of index entry to calculate block relative offset.
base := int64(b.written)
// Construct block index. Detailed format described in Builder
// documentation, but it is essentially encoded as:
// "start | index | index | ... | count"
var (
count = len(b.indexes)
index = make([]byte, 16+count*8)
)
binary.LittleEndian.PutUint64(index, *b.startNum)
// Each offset is relative from the position it is encoded in the
// index. This means that even if the same block was to be included in
// the index twice (this would be invalid anyways), the relative offset
// would be different. The idea with this is that after reading a
// relative offset, the corresponding block can be quickly read by
// performing a seek relative to the current position.
for i, offset := range b.indexes {
relative := int64(offset) - base
binary.LittleEndian.PutUint64(index[8+i*8:], uint64(relative))
}
binary.LittleEndian.PutUint64(index[8+count*8:], uint64(count))
// Finally, write the block index entry.
if _, err := b.w.Write(TypeBlockIndex, index); err != nil {
return common.Hash{}, fmt.Errorf("unable to write block index: %w", err)
}
return root, nil
}
// snappyWrite is a small helper to take care snappy encoding and writing an e2store entry.
func (b *Builder) snappyWrite(typ uint16, in []byte) error {
var (
buf = b.buf
s = b.snappy
)
buf.Reset()
s.Reset(buf)
if _, err := b.snappy.Write(in); err != nil {
return fmt.Errorf("error snappy encoding: %w", err)
}
if err := s.Flush(); err != nil {
return fmt.Errorf("error flushing snappy encoding: %w", err)
}
n, err := b.w.Write(typ, b.buf.Bytes())
b.written += n
if err != nil {
return fmt.Errorf("error writing e2store entry: %w", err)
}
return nil
}