forked from cerc-io/plugeth
d09a6e5421
Removed the managed tx state from the chain manager to the transaction pool where it's much easier to keep track of nonces (and manage them). The transaction pool now also uses the queue and pending txs differently where queued txs are now moved over to the pending queue (i.e. txs ready for processing and propagation).
806 lines
21 KiB
Go
806 lines
21 KiB
Go
package core
|
|
|
|
import (
|
|
"bytes"
|
|
"fmt"
|
|
"io"
|
|
"math/big"
|
|
"os"
|
|
"runtime"
|
|
"sync"
|
|
"time"
|
|
|
|
"github.com/ethereum/go-ethereum/common"
|
|
"github.com/ethereum/go-ethereum/core/state"
|
|
"github.com/ethereum/go-ethereum/core/types"
|
|
"github.com/ethereum/go-ethereum/event"
|
|
"github.com/ethereum/go-ethereum/logger"
|
|
"github.com/ethereum/go-ethereum/logger/glog"
|
|
"github.com/ethereum/go-ethereum/params"
|
|
"github.com/ethereum/go-ethereum/pow"
|
|
"github.com/ethereum/go-ethereum/rlp"
|
|
)
|
|
|
|
var (
|
|
chainlogger = logger.NewLogger("CHAIN")
|
|
jsonlogger = logger.NewJsonLogger()
|
|
|
|
blockHashPre = []byte("block-hash-")
|
|
blockNumPre = []byte("block-num-")
|
|
)
|
|
|
|
const (
|
|
blockCacheLimit = 10000
|
|
maxFutureBlocks = 256
|
|
)
|
|
|
|
func CalcDifficulty(block, parent *types.Header) *big.Int {
|
|
diff := new(big.Int)
|
|
|
|
adjust := new(big.Int).Div(parent.Difficulty, params.DifficultyBoundDivisor)
|
|
if big.NewInt(int64(block.Time)-int64(parent.Time)).Cmp(params.DurationLimit) < 0 {
|
|
diff.Add(parent.Difficulty, adjust)
|
|
} else {
|
|
diff.Sub(parent.Difficulty, adjust)
|
|
}
|
|
|
|
if diff.Cmp(params.MinimumDifficulty) < 0 {
|
|
return params.MinimumDifficulty
|
|
}
|
|
|
|
return diff
|
|
}
|
|
|
|
func CalcTD(block, parent *types.Block) *big.Int {
|
|
if parent == nil {
|
|
return block.Difficulty()
|
|
}
|
|
|
|
td := new(big.Int).Add(parent.Td, block.Header().Difficulty)
|
|
|
|
return td
|
|
}
|
|
|
|
func CalcGasLimit(parent *types.Block) *big.Int {
|
|
decay := new(big.Int).Div(parent.GasLimit(), params.GasLimitBoundDivisor)
|
|
contrib := new(big.Int).Mul(parent.GasUsed(), big.NewInt(3))
|
|
contrib = contrib.Div(contrib, big.NewInt(2))
|
|
contrib = contrib.Div(contrib, params.GasLimitBoundDivisor)
|
|
|
|
gl := new(big.Int).Sub(parent.GasLimit(), decay)
|
|
gl = gl.Add(gl, contrib)
|
|
gl = gl.Add(gl, big.NewInt(1))
|
|
gl = common.BigMax(gl, params.MinGasLimit)
|
|
|
|
if gl.Cmp(params.GenesisGasLimit) < 0 {
|
|
gl2 := new(big.Int).Add(parent.GasLimit(), decay)
|
|
return common.BigMin(params.GenesisGasLimit, gl2)
|
|
}
|
|
return gl
|
|
}
|
|
|
|
type ChainManager struct {
|
|
//eth EthManager
|
|
blockDb common.Database
|
|
stateDb common.Database
|
|
processor types.BlockProcessor
|
|
eventMux *event.TypeMux
|
|
genesisBlock *types.Block
|
|
// Last known total difficulty
|
|
mu sync.RWMutex
|
|
chainmu sync.RWMutex
|
|
tsmu sync.RWMutex
|
|
|
|
td *big.Int
|
|
currentBlock *types.Block
|
|
lastBlockHash common.Hash
|
|
currentGasLimit *big.Int
|
|
|
|
transState *state.StateDB
|
|
txState *state.ManagedState
|
|
|
|
cache *BlockCache
|
|
futureBlocks *BlockCache
|
|
|
|
quit chan struct{}
|
|
wg sync.WaitGroup
|
|
|
|
pow pow.PoW
|
|
}
|
|
|
|
func NewChainManager(blockDb, stateDb common.Database, pow pow.PoW, mux *event.TypeMux) *ChainManager {
|
|
bc := &ChainManager{
|
|
blockDb: blockDb,
|
|
stateDb: stateDb,
|
|
genesisBlock: GenesisBlock(stateDb),
|
|
eventMux: mux,
|
|
quit: make(chan struct{}),
|
|
cache: NewBlockCache(blockCacheLimit),
|
|
pow: pow,
|
|
}
|
|
bc.setLastState()
|
|
|
|
// Check the current state of the block hashes and make sure that we do not have any of the bad blocks in our chain
|
|
for hash, _ := range BadHashes {
|
|
if block := bc.GetBlock(hash); block != nil {
|
|
glog.V(logger.Error).Infof("Found bad hash. Reorganising chain to state %x\n", block.ParentHash().Bytes()[:4])
|
|
block = bc.GetBlock(block.ParentHash())
|
|
if block == nil {
|
|
glog.Fatal("Unable to complete. Parent block not found. Corrupted DB?")
|
|
}
|
|
bc.SetHead(block)
|
|
|
|
glog.V(logger.Error).Infoln("Chain reorg was successfull. Resuming normal operation")
|
|
}
|
|
}
|
|
|
|
bc.transState = bc.State().Copy()
|
|
// Take ownership of this particular state
|
|
bc.txState = state.ManageState(bc.State().Copy())
|
|
|
|
bc.futureBlocks = NewBlockCache(maxFutureBlocks)
|
|
bc.makeCache()
|
|
|
|
go bc.update()
|
|
|
|
return bc
|
|
}
|
|
|
|
func (bc *ChainManager) SetHead(head *types.Block) {
|
|
bc.mu.Lock()
|
|
defer bc.mu.Unlock()
|
|
|
|
for block := bc.currentBlock; block != nil && block.Hash() != head.Hash(); block = bc.GetBlock(block.Header().ParentHash) {
|
|
bc.removeBlock(block)
|
|
}
|
|
|
|
bc.cache = NewBlockCache(blockCacheLimit)
|
|
bc.currentBlock = head
|
|
bc.makeCache()
|
|
|
|
statedb := state.New(head.Root(), bc.stateDb)
|
|
bc.txState = state.ManageState(statedb)
|
|
bc.transState = statedb.Copy()
|
|
bc.setTotalDifficulty(head.Td)
|
|
bc.insert(head)
|
|
bc.setLastState()
|
|
}
|
|
|
|
func (self *ChainManager) Td() *big.Int {
|
|
self.mu.RLock()
|
|
defer self.mu.RUnlock()
|
|
|
|
return self.td
|
|
}
|
|
|
|
func (self *ChainManager) GasLimit() *big.Int {
|
|
// return self.currentGasLimit
|
|
return self.currentBlock.GasLimit()
|
|
}
|
|
|
|
func (self *ChainManager) LastBlockHash() common.Hash {
|
|
self.mu.RLock()
|
|
defer self.mu.RUnlock()
|
|
|
|
return self.lastBlockHash
|
|
}
|
|
|
|
func (self *ChainManager) CurrentBlock() *types.Block {
|
|
self.mu.RLock()
|
|
defer self.mu.RUnlock()
|
|
|
|
return self.currentBlock
|
|
}
|
|
|
|
func (self *ChainManager) Status() (td *big.Int, currentBlock common.Hash, genesisBlock common.Hash) {
|
|
self.mu.RLock()
|
|
defer self.mu.RUnlock()
|
|
|
|
return self.td, self.currentBlock.Hash(), self.genesisBlock.Hash()
|
|
}
|
|
|
|
func (self *ChainManager) SetProcessor(proc types.BlockProcessor) {
|
|
self.processor = proc
|
|
}
|
|
|
|
func (self *ChainManager) State() *state.StateDB {
|
|
return state.New(self.CurrentBlock().Root(), self.stateDb)
|
|
}
|
|
|
|
func (self *ChainManager) TransState() *state.StateDB {
|
|
self.tsmu.RLock()
|
|
defer self.tsmu.RUnlock()
|
|
|
|
return self.transState
|
|
}
|
|
|
|
func (self *ChainManager) setTransState(statedb *state.StateDB) {
|
|
self.transState = statedb
|
|
}
|
|
|
|
func (bc *ChainManager) setLastState() {
|
|
data, _ := bc.blockDb.Get([]byte("LastBlock"))
|
|
if len(data) != 0 {
|
|
block := bc.GetBlock(common.BytesToHash(data))
|
|
if block != nil {
|
|
bc.currentBlock = block
|
|
bc.lastBlockHash = block.Hash()
|
|
} else { // TODO CLEAN THIS UP TMP CODE
|
|
block = bc.GetBlockByNumber(400000)
|
|
if block == nil {
|
|
fmt.Println("Fatal. LastBlock not found. Report this issue")
|
|
os.Exit(1)
|
|
}
|
|
bc.currentBlock = block
|
|
bc.lastBlockHash = block.Hash()
|
|
bc.insert(block)
|
|
}
|
|
} else {
|
|
bc.Reset()
|
|
}
|
|
bc.td = bc.currentBlock.Td
|
|
bc.currentGasLimit = CalcGasLimit(bc.currentBlock)
|
|
|
|
if glog.V(logger.Info) {
|
|
glog.Infof("Last block (#%v) %x TD=%v\n", bc.currentBlock.Number(), bc.currentBlock.Hash(), bc.td)
|
|
}
|
|
}
|
|
|
|
func (bc *ChainManager) makeCache() {
|
|
if bc.cache == nil {
|
|
bc.cache = NewBlockCache(blockCacheLimit)
|
|
}
|
|
// load in last `blockCacheLimit` - 1 blocks. Last block is the current.
|
|
ancestors := bc.GetAncestors(bc.currentBlock, blockCacheLimit-1)
|
|
ancestors = append(ancestors, bc.currentBlock)
|
|
for _, block := range ancestors {
|
|
bc.cache.Push(block)
|
|
}
|
|
}
|
|
|
|
// Block creation & chain handling
|
|
func (bc *ChainManager) NewBlock(coinbase common.Address) *types.Block {
|
|
bc.mu.RLock()
|
|
defer bc.mu.RUnlock()
|
|
|
|
var (
|
|
root common.Hash
|
|
parentHash common.Hash
|
|
)
|
|
|
|
if bc.currentBlock != nil {
|
|
root = bc.currentBlock.Header().Root
|
|
parentHash = bc.lastBlockHash
|
|
}
|
|
|
|
block := types.NewBlock(
|
|
parentHash,
|
|
coinbase,
|
|
root,
|
|
common.BigPow(2, 32),
|
|
0,
|
|
nil)
|
|
block.SetUncles(nil)
|
|
block.SetTransactions(nil)
|
|
block.SetReceipts(nil)
|
|
|
|
parent := bc.currentBlock
|
|
if parent != nil {
|
|
header := block.Header()
|
|
header.Difficulty = CalcDifficulty(block.Header(), parent.Header())
|
|
header.Number = new(big.Int).Add(parent.Header().Number, common.Big1)
|
|
header.GasLimit = CalcGasLimit(parent)
|
|
}
|
|
|
|
return block
|
|
}
|
|
|
|
func (bc *ChainManager) Reset() {
|
|
bc.mu.Lock()
|
|
defer bc.mu.Unlock()
|
|
|
|
for block := bc.currentBlock; block != nil; block = bc.GetBlock(block.Header().ParentHash) {
|
|
bc.removeBlock(block)
|
|
}
|
|
|
|
if bc.cache == nil {
|
|
bc.cache = NewBlockCache(blockCacheLimit)
|
|
}
|
|
|
|
// Prepare the genesis block
|
|
bc.write(bc.genesisBlock)
|
|
bc.insert(bc.genesisBlock)
|
|
bc.currentBlock = bc.genesisBlock
|
|
bc.makeCache()
|
|
|
|
bc.setTotalDifficulty(common.Big("0"))
|
|
}
|
|
|
|
func (bc *ChainManager) removeBlock(block *types.Block) {
|
|
bc.blockDb.Delete(append(blockHashPre, block.Hash().Bytes()...))
|
|
}
|
|
|
|
func (bc *ChainManager) ResetWithGenesisBlock(gb *types.Block) {
|
|
bc.mu.Lock()
|
|
defer bc.mu.Unlock()
|
|
|
|
for block := bc.currentBlock; block != nil; block = bc.GetBlock(block.Header().ParentHash) {
|
|
bc.removeBlock(block)
|
|
}
|
|
|
|
// Prepare the genesis block
|
|
gb.Td = gb.Difficulty()
|
|
bc.genesisBlock = gb
|
|
bc.write(bc.genesisBlock)
|
|
bc.insert(bc.genesisBlock)
|
|
bc.currentBlock = bc.genesisBlock
|
|
bc.makeCache()
|
|
bc.td = gb.Difficulty()
|
|
}
|
|
|
|
// Export writes the active chain to the given writer.
|
|
func (self *ChainManager) Export(w io.Writer) error {
|
|
self.mu.RLock()
|
|
defer self.mu.RUnlock()
|
|
glog.V(logger.Info).Infof("exporting %v blocks...\n", self.currentBlock.Header().Number)
|
|
|
|
last := self.currentBlock.NumberU64()
|
|
|
|
for nr := uint64(1); nr <= last; nr++ {
|
|
block := self.GetBlockByNumber(nr)
|
|
if block == nil {
|
|
return fmt.Errorf("export failed on #%d: not found", nr)
|
|
}
|
|
|
|
if err := block.EncodeRLP(w); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func (bc *ChainManager) insert(block *types.Block) {
|
|
key := append(blockNumPre, block.Number().Bytes()...)
|
|
bc.blockDb.Put(key, block.Hash().Bytes())
|
|
|
|
bc.blockDb.Put([]byte("LastBlock"), block.Hash().Bytes())
|
|
bc.currentBlock = block
|
|
bc.lastBlockHash = block.Hash()
|
|
}
|
|
|
|
func (bc *ChainManager) write(block *types.Block) {
|
|
enc, _ := rlp.EncodeToBytes((*types.StorageBlock)(block))
|
|
key := append(blockHashPre, block.Hash().Bytes()...)
|
|
bc.blockDb.Put(key, enc)
|
|
// Push block to cache
|
|
bc.cache.Push(block)
|
|
}
|
|
|
|
// Accessors
|
|
func (bc *ChainManager) Genesis() *types.Block {
|
|
return bc.genesisBlock
|
|
}
|
|
|
|
// Block fetching methods
|
|
func (bc *ChainManager) HasBlock(hash common.Hash) bool {
|
|
data, _ := bc.blockDb.Get(append(blockHashPre, hash[:]...))
|
|
return len(data) != 0
|
|
}
|
|
|
|
func (self *ChainManager) GetBlockHashesFromHash(hash common.Hash, max uint64) (chain []common.Hash) {
|
|
block := self.GetBlock(hash)
|
|
if block == nil {
|
|
return
|
|
}
|
|
// XXX Could be optimised by using a different database which only holds hashes (i.e., linked list)
|
|
for i := uint64(0); i < max; i++ {
|
|
block = self.GetBlock(block.ParentHash())
|
|
if block == nil {
|
|
break
|
|
}
|
|
|
|
chain = append(chain, block.Hash())
|
|
if block.Number().Cmp(common.Big0) <= 0 {
|
|
break
|
|
}
|
|
}
|
|
|
|
return
|
|
}
|
|
|
|
func (self *ChainManager) GetBlock(hash common.Hash) *types.Block {
|
|
/*
|
|
if block := self.cache.Get(hash); block != nil {
|
|
return block
|
|
}
|
|
*/
|
|
|
|
data, _ := self.blockDb.Get(append(blockHashPre, hash[:]...))
|
|
if len(data) == 0 {
|
|
return nil
|
|
}
|
|
var block types.StorageBlock
|
|
if err := rlp.Decode(bytes.NewReader(data), &block); err != nil {
|
|
glog.V(logger.Error).Infof("invalid block RLP for hash %x: %v", hash, err)
|
|
return nil
|
|
}
|
|
return (*types.Block)(&block)
|
|
}
|
|
|
|
func (self *ChainManager) GetBlockByNumber(num uint64) *types.Block {
|
|
self.mu.RLock()
|
|
defer self.mu.RUnlock()
|
|
|
|
return self.getBlockByNumber(num)
|
|
|
|
}
|
|
|
|
// non blocking version
|
|
func (self *ChainManager) getBlockByNumber(num uint64) *types.Block {
|
|
key, _ := self.blockDb.Get(append(blockNumPre, big.NewInt(int64(num)).Bytes()...))
|
|
if len(key) == 0 {
|
|
return nil
|
|
}
|
|
|
|
return self.GetBlock(common.BytesToHash(key))
|
|
}
|
|
|
|
func (self *ChainManager) GetUnclesInChain(block *types.Block, length int) (uncles []*types.Header) {
|
|
for i := 0; block != nil && i < length; i++ {
|
|
uncles = append(uncles, block.Uncles()...)
|
|
block = self.GetBlock(block.ParentHash())
|
|
}
|
|
|
|
return
|
|
}
|
|
|
|
func (self *ChainManager) GetAncestors(block *types.Block, length int) (blocks []*types.Block) {
|
|
for i := 0; i < length; i++ {
|
|
block = self.GetBlock(block.ParentHash())
|
|
if block == nil {
|
|
break
|
|
}
|
|
|
|
blocks = append(blocks, block)
|
|
}
|
|
|
|
return
|
|
}
|
|
|
|
func (bc *ChainManager) setTotalDifficulty(td *big.Int) {
|
|
//bc.blockDb.Put([]byte("LTD"), td.Bytes())
|
|
bc.td = td
|
|
}
|
|
|
|
func (self *ChainManager) CalcTotalDiff(block *types.Block) (*big.Int, error) {
|
|
parent := self.GetBlock(block.Header().ParentHash)
|
|
if parent == nil {
|
|
return nil, fmt.Errorf("Unable to calculate total diff without known parent %x", block.Header().ParentHash)
|
|
}
|
|
|
|
parentTd := parent.Td
|
|
|
|
uncleDiff := new(big.Int)
|
|
for _, uncle := range block.Uncles() {
|
|
uncleDiff = uncleDiff.Add(uncleDiff, uncle.Difficulty)
|
|
}
|
|
|
|
td := new(big.Int)
|
|
td = td.Add(parentTd, uncleDiff)
|
|
td = td.Add(td, block.Header().Difficulty)
|
|
|
|
return td, nil
|
|
}
|
|
|
|
func (bc *ChainManager) Stop() {
|
|
close(bc.quit)
|
|
|
|
bc.wg.Wait()
|
|
|
|
glog.V(logger.Info).Infoln("Chain manager stopped")
|
|
}
|
|
|
|
type queueEvent struct {
|
|
queue []interface{}
|
|
canonicalCount int
|
|
sideCount int
|
|
splitCount int
|
|
}
|
|
|
|
func (self *ChainManager) procFutureBlocks() {
|
|
var blocks []*types.Block
|
|
self.futureBlocks.Each(func(i int, block *types.Block) {
|
|
blocks = append(blocks, block)
|
|
})
|
|
if len(blocks) > 0 {
|
|
types.BlockBy(types.Number).Sort(blocks)
|
|
self.InsertChain(blocks)
|
|
}
|
|
}
|
|
|
|
// InsertChain will attempt to insert the given chain in to the canonical chain or, otherwise, create a fork. It an error is returned
|
|
// it will return the index number of the failing block as well an error describing what went wrong (for possible errors see core/errors.go).
|
|
func (self *ChainManager) InsertChain(chain types.Blocks) (int, error) {
|
|
self.wg.Add(1)
|
|
defer self.wg.Done()
|
|
|
|
self.chainmu.Lock()
|
|
defer self.chainmu.Unlock()
|
|
|
|
// A queued approach to delivering events. This is generally
|
|
// faster than direct delivery and requires much less mutex
|
|
// acquiring.
|
|
var (
|
|
queue = make([]interface{}, len(chain))
|
|
queueEvent = queueEvent{queue: queue}
|
|
stats struct{ queued, processed, ignored int }
|
|
tstart = time.Now()
|
|
|
|
nonceDone = make(chan nonceResult, len(chain))
|
|
nonceQuit = make(chan struct{})
|
|
nonceChecked = make([]bool, len(chain))
|
|
)
|
|
|
|
// Start the parallel nonce verifier.
|
|
go verifyNonces(self.pow, chain, nonceQuit, nonceDone)
|
|
defer close(nonceQuit)
|
|
|
|
for i, block := range chain {
|
|
bstart := time.Now()
|
|
// Wait for block i's nonce to be verified before processing
|
|
// its state transition.
|
|
for nonceChecked[i] {
|
|
r := <-nonceDone
|
|
nonceChecked[r.i] = true
|
|
if !r.valid {
|
|
block := chain[i]
|
|
return i, ValidationError("Block (#%v / %x) nonce is invalid (= %x)", block.Number(), block.Hash(), block.Nonce)
|
|
}
|
|
}
|
|
|
|
if BadHashes[block.Hash()] {
|
|
err := fmt.Errorf("Found known bad hash in chain %x", block.Hash())
|
|
blockErr(block, err)
|
|
return i, err
|
|
}
|
|
|
|
// Setting block.Td regardless of error (known for example) prevents errors down the line
|
|
// in the protocol handler
|
|
block.Td = new(big.Int).Set(CalcTD(block, self.GetBlock(block.ParentHash())))
|
|
|
|
// Call in to the block processor and check for errors. It's likely that if one block fails
|
|
// all others will fail too (unless a known block is returned).
|
|
logs, err := self.processor.Process(block)
|
|
if err != nil {
|
|
if IsKnownBlockErr(err) {
|
|
stats.ignored++
|
|
continue
|
|
}
|
|
|
|
if err == BlockFutureErr {
|
|
block.SetQueued(true)
|
|
self.futureBlocks.Push(block)
|
|
stats.queued++
|
|
continue
|
|
}
|
|
|
|
if IsParentErr(err) && self.futureBlocks.Has(block.ParentHash()) {
|
|
block.SetQueued(true)
|
|
self.futureBlocks.Push(block)
|
|
stats.queued++
|
|
continue
|
|
}
|
|
|
|
blockErr(block, err)
|
|
|
|
return i, err
|
|
}
|
|
|
|
cblock := self.currentBlock
|
|
// Compare the TD of the last known block in the canonical chain to make sure it's greater.
|
|
// At this point it's possible that a different chain (fork) becomes the new canonical chain.
|
|
if block.Td.Cmp(self.td) > 0 {
|
|
// chain fork
|
|
if block.ParentHash() != cblock.Hash() {
|
|
// during split we merge two different chains and create the new canonical chain
|
|
err := self.merge(cblock, block)
|
|
if err != nil {
|
|
return i, err
|
|
}
|
|
|
|
queue[i] = ChainSplitEvent{block, logs}
|
|
queueEvent.splitCount++
|
|
}
|
|
|
|
self.setTotalDifficulty(block.Td)
|
|
self.insert(block)
|
|
|
|
jsonlogger.LogJson(&logger.EthChainNewHead{
|
|
BlockHash: block.Hash().Hex(),
|
|
BlockNumber: block.Number(),
|
|
ChainHeadHash: cblock.Hash().Hex(),
|
|
BlockPrevHash: block.ParentHash().Hex(),
|
|
})
|
|
|
|
self.setTransState(state.New(block.Root(), self.stateDb))
|
|
self.txState.SetState(state.New(block.Root(), self.stateDb))
|
|
|
|
queue[i] = ChainEvent{block, block.Hash(), logs}
|
|
queueEvent.canonicalCount++
|
|
|
|
if glog.V(logger.Debug) {
|
|
glog.Infof("[%v] inserted block #%d (%d TXs %d UNCs) (%x...). Took %v\n", time.Now().UnixNano(), block.Number(), len(block.Transactions()), len(block.Uncles()), block.Hash().Bytes()[0:4], time.Since(bstart))
|
|
}
|
|
} else {
|
|
if glog.V(logger.Detail) {
|
|
glog.Infof("inserted forked block #%d (TD=%v) (%d TXs %d UNCs) (%x...). Took %v\n", block.Number(), block.Difficulty(), len(block.Transactions()), len(block.Uncles()), block.Hash().Bytes()[0:4], time.Since(bstart))
|
|
}
|
|
|
|
queue[i] = ChainSideEvent{block, logs}
|
|
queueEvent.sideCount++
|
|
}
|
|
// Write block to database. Eventually we'll have to improve on this and throw away blocks that are
|
|
// not in the canonical chain.
|
|
self.write(block)
|
|
// Delete from future blocks
|
|
self.futureBlocks.Delete(block.Hash())
|
|
|
|
stats.processed++
|
|
|
|
}
|
|
|
|
if (stats.queued > 0 || stats.processed > 0 || stats.ignored > 0) && bool(glog.V(logger.Info)) {
|
|
tend := time.Since(tstart)
|
|
start, end := chain[0], chain[len(chain)-1]
|
|
glog.Infof("imported %d block(s) (%d queued %d ignored) in %v. #%v [%x / %x]\n", stats.processed, stats.queued, stats.ignored, tend, end.Number(), start.Hash().Bytes()[:4], end.Hash().Bytes()[:4])
|
|
}
|
|
|
|
go self.eventMux.Post(queueEvent)
|
|
|
|
return 0, nil
|
|
}
|
|
|
|
// diff takes two blocks, an old chain and a new chain and will reconstruct the blocks and inserts them
|
|
// to be part of the new canonical chain.
|
|
func (self *ChainManager) diff(oldBlock, newBlock *types.Block) (types.Blocks, error) {
|
|
var (
|
|
newChain types.Blocks
|
|
commonBlock *types.Block
|
|
oldStart = oldBlock
|
|
newStart = newBlock
|
|
)
|
|
|
|
// first reduce whoever is higher bound
|
|
if oldBlock.NumberU64() > newBlock.NumberU64() {
|
|
// reduce old chain
|
|
for oldBlock = oldBlock; oldBlock != nil && oldBlock.NumberU64() != newBlock.NumberU64(); oldBlock = self.GetBlock(oldBlock.ParentHash()) {
|
|
}
|
|
} else {
|
|
// reduce new chain and append new chain blocks for inserting later on
|
|
for newBlock = newBlock; newBlock != nil && newBlock.NumberU64() != oldBlock.NumberU64(); newBlock = self.GetBlock(newBlock.ParentHash()) {
|
|
newChain = append(newChain, newBlock)
|
|
}
|
|
}
|
|
if oldBlock == nil {
|
|
return nil, fmt.Errorf("Invalid old chain")
|
|
}
|
|
if newBlock == nil {
|
|
return nil, fmt.Errorf("Invalid new chain")
|
|
}
|
|
|
|
numSplit := newBlock.Number()
|
|
for {
|
|
if oldBlock.Hash() == newBlock.Hash() {
|
|
commonBlock = oldBlock
|
|
break
|
|
}
|
|
newChain = append(newChain, newBlock)
|
|
|
|
oldBlock, newBlock = self.GetBlock(oldBlock.ParentHash()), self.GetBlock(newBlock.ParentHash())
|
|
if oldBlock == nil {
|
|
return nil, fmt.Errorf("Invalid old chain")
|
|
}
|
|
if newBlock == nil {
|
|
return nil, fmt.Errorf("Invalid new chain")
|
|
}
|
|
}
|
|
|
|
if glog.V(logger.Info) {
|
|
commonHash := commonBlock.Hash()
|
|
glog.Infof("Fork detected @ %x. Reorganising chain from #%v %x to %x", commonHash[:4], numSplit, oldStart.Hash().Bytes()[:4], newStart.Hash().Bytes()[:4])
|
|
}
|
|
|
|
return newChain, nil
|
|
}
|
|
|
|
// merge merges two different chain to the new canonical chain
|
|
func (self *ChainManager) merge(oldBlock, newBlock *types.Block) error {
|
|
newChain, err := self.diff(oldBlock, newBlock)
|
|
if err != nil {
|
|
return fmt.Errorf("chain reorg failed: %v", err)
|
|
}
|
|
|
|
// insert blocks. Order does not matter. Last block will be written in ImportChain itself which creates the new head properly
|
|
for _, block := range newChain {
|
|
self.insert(block)
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func (self *ChainManager) update() {
|
|
events := self.eventMux.Subscribe(queueEvent{})
|
|
futureTimer := time.Tick(5 * time.Second)
|
|
out:
|
|
for {
|
|
select {
|
|
case ev := <-events.Chan():
|
|
switch ev := ev.(type) {
|
|
case queueEvent:
|
|
for _, event := range ev.queue {
|
|
switch event := event.(type) {
|
|
case ChainEvent:
|
|
// We need some control over the mining operation. Acquiring locks and waiting for the miner to create new block takes too long
|
|
// and in most cases isn't even necessary.
|
|
if self.lastBlockHash == event.Hash {
|
|
self.currentGasLimit = CalcGasLimit(event.Block)
|
|
self.eventMux.Post(ChainHeadEvent{event.Block})
|
|
}
|
|
}
|
|
|
|
self.eventMux.Post(event)
|
|
}
|
|
}
|
|
case <-futureTimer:
|
|
self.procFutureBlocks()
|
|
case <-self.quit:
|
|
break out
|
|
}
|
|
}
|
|
}
|
|
|
|
func blockErr(block *types.Block, err error) {
|
|
h := block.Header()
|
|
glog.V(logger.Error).Infof("Bad block #%v (%x)\n", h.Number, h.Hash().Bytes())
|
|
glog.V(logger.Error).Infoln(err)
|
|
glog.V(logger.Debug).Infoln(verifyNonces)
|
|
}
|
|
|
|
type nonceResult struct {
|
|
i int
|
|
valid bool
|
|
}
|
|
|
|
// block verifies nonces of the given blocks in parallel and returns
|
|
// an error if one of the blocks nonce verifications failed.
|
|
func verifyNonces(pow pow.PoW, blocks []*types.Block, quit <-chan struct{}, done chan<- nonceResult) {
|
|
// Spawn a few workers. They listen for blocks on the in channel
|
|
// and send results on done. The workers will exit in the
|
|
// background when in is closed.
|
|
var (
|
|
in = make(chan int)
|
|
nworkers = runtime.GOMAXPROCS(0)
|
|
)
|
|
defer close(in)
|
|
if len(blocks) < nworkers {
|
|
nworkers = len(blocks)
|
|
}
|
|
for i := 0; i < nworkers; i++ {
|
|
go func() {
|
|
for i := range in {
|
|
done <- nonceResult{i: i, valid: pow.Verify(blocks[i])}
|
|
}
|
|
}()
|
|
}
|
|
// Feed block indices to the workers.
|
|
for i := range blocks {
|
|
select {
|
|
case in <- i:
|
|
continue
|
|
case <-quit:
|
|
return
|
|
}
|
|
}
|
|
}
|