bsc/eth/sync.go
zjubfd 2ce00adb55
[R4R] performance improvement in many aspects (#257)
* focus on performance improvement in many aspects.

1. Do BlockBody verification concurrently;
2. Do calculation of intermediate root concurrently;
3. Preload accounts before processing blocks;
4. Make the snapshot layers configurable.
5. Reuse some object to reduce GC.

add

* rlp: improve decoder stream implementation (#22858)

This commit makes various cleanup changes to rlp.Stream.

* rlp: shrink Stream struct

This removes a lot of unused padding space in Stream by reordering the
fields. The size of Stream changes from 120 bytes to 88 bytes. Stream
instances are internally cached and reused using sync.Pool, so this does
not improve performance.

* rlp: simplify list stack

The list stack kept track of the size of the current list context as
well as the current offset into it. The size had to be stored in the
stack in order to subtract it from the remaining bytes of any enclosing
list in ListEnd. It seems that this can be implemented in a simpler
way: just subtract the size from the enclosing list context in List instead.

* rlp: use atomic.Value for type cache (#22902)

All encoding/decoding operations read the type cache to find the
writer/decoder function responsible for a type. When analyzing CPU
profiles of geth during sync, I found that the use of sync.RWMutex in
cache lookups appears in the profiles. It seems we are running into
CPU cache contention problems when package rlp is heavily used
on all CPU cores during sync.

This change makes it use atomic.Value + a writer lock instead of
sync.RWMutex. In the common case where the typeinfo entry is present in
the cache, we simply fetch the map and lookup the type.

* rlp: optimize byte array handling (#22924)

This change improves the performance of encoding/decoding [N]byte.

    name                     old time/op    new time/op    delta
    DecodeByteArrayStruct-8     336ns ± 0%     246ns ± 0%  -26.98%  (p=0.000 n=9+10)
    EncodeByteArrayStruct-8     225ns ± 1%     148ns ± 1%  -34.12%  (p=0.000 n=10+10)

    name                     old alloc/op   new alloc/op   delta
    DecodeByteArrayStruct-8      120B ± 0%       48B ± 0%  -60.00%  (p=0.000 n=10+10)
    EncodeByteArrayStruct-8     0.00B          0.00B          ~     (all equal)

* rlp: optimize big.Int decoding for size <= 32 bytes (#22927)

This change grows the static integer buffer in Stream to 32 bytes,
making it possible to decode 256bit integers without allocating a
temporary buffer.

In the recent commit 088da24, Stream struct size decreased from 120
bytes down to 88 bytes. This commit grows the struct to 112 bytes again,
but the size change will not degrade performance because Stream
instances are internally cached in sync.Pool.

    name             old time/op    new time/op    delta
    DecodeBigInts-8    12.2µs ± 0%     8.6µs ± 4%  -29.58%  (p=0.000 n=9+10)

    name             old speed      new speed      delta
    DecodeBigInts-8   230MB/s ± 0%   326MB/s ± 4%  +42.04%  (p=0.000 n=9+10)

* eth/protocols/eth, les: avoid Raw() when decoding HashOrNumber (#22841)

Getting the raw value is not necessary to decode this type, and
decoding it directly from the stream is faster.

* fix testcase

* debug no lazy

* fix can not repair

* address comments

Co-authored-by: Felix Lange <fjl@twurst.com>
2021-07-29 17:16:53 +08:00

358 lines
12 KiB
Go

// Copyright 2015 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 eth
import (
"math/big"
"math/rand"
"sync/atomic"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/common/gopool"
"github.com/ethereum/go-ethereum/core/rawdb"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/eth/downloader"
"github.com/ethereum/go-ethereum/eth/protocols/eth"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/p2p/enode"
)
const (
forceSyncCycle = 10 * time.Second // Time interval to force syncs, even if few peers are available
defaultMinSyncPeers = 5 // Amount of peers desired to start syncing
// This is the target size for the packs of transactions sent by txsyncLoop64.
// A pack can get larger than this if a single transactions exceeds this size.
txsyncPackSize = 100 * 1024
)
type txsync struct {
p *eth.Peer
txs []*types.Transaction
}
// syncTransactions starts sending all currently pending transactions to the given peer.
func (h *handler) syncTransactions(p *eth.Peer) {
// Assemble the set of transaction to broadcast or announce to the remote
// peer. Fun fact, this is quite an expensive operation as it needs to sort
// the transactions if the sorting is not cached yet. However, with a random
// order, insertions could overflow the non-executable queues and get dropped.
//
// TODO(karalabe): Figure out if we could get away with random order somehow
var txs types.Transactions
pending, _ := h.txpool.Pending()
for _, batch := range pending {
txs = append(txs, batch...)
}
if len(txs) == 0 {
return
}
// The eth/65 protocol introduces proper transaction announcements, so instead
// of dripping transactions across multiple peers, just send the entire list as
// an announcement and let the remote side decide what they need (likely nothing).
if p.Version() >= eth.ETH65 {
hashes := make([]common.Hash, len(txs))
for i, tx := range txs {
hashes[i] = tx.Hash()
}
p.AsyncSendPooledTransactionHashes(hashes)
return
}
// Out of luck, peer is running legacy protocols, drop the txs over
select {
case h.txsyncCh <- &txsync{p: p, txs: txs}:
case <-h.quitSync:
}
}
// txsyncLoop64 takes care of the initial transaction sync for each new
// connection. When a new peer appears, we relay all currently pending
// transactions. In order to minimise egress bandwidth usage, we send
// the transactions in small packs to one peer at a time.
func (h *handler) txsyncLoop64() {
defer h.wg.Done()
var (
pending = make(map[enode.ID]*txsync)
sending = false // whether a send is active
pack = new(txsync) // the pack that is being sent
done = make(chan error, 1) // result of the send
)
// send starts a sending a pack of transactions from the sync.
send := func(s *txsync) {
if s.p.Version() >= eth.ETH65 {
panic("initial transaction syncer running on eth/65+")
}
// Fill pack with transactions up to the target size.
size := common.StorageSize(0)
pack.p = s.p
pack.txs = pack.txs[:0]
for i := 0; i < len(s.txs) && size < txsyncPackSize; i++ {
pack.txs = append(pack.txs, s.txs[i])
size += s.txs[i].Size()
}
// Remove the transactions that will be sent.
s.txs = s.txs[:copy(s.txs, s.txs[len(pack.txs):])]
if len(s.txs) == 0 {
delete(pending, s.p.Peer.ID())
}
// Send the pack in the background.
s.p.Log().Trace("Sending batch of transactions", "count", len(pack.txs), "bytes", size)
sending = true
gopool.Submit(func() { done <- pack.p.SendTransactions(pack.txs) })
}
// pick chooses the next pending sync.
pick := func() *txsync {
if len(pending) == 0 {
return nil
}
n := rand.Intn(len(pending)) + 1
for _, s := range pending {
if n--; n == 0 {
return s
}
}
return nil
}
for {
select {
case s := <-h.txsyncCh:
pending[s.p.Peer.ID()] = s
if !sending {
send(s)
}
case err := <-done:
sending = false
// Stop tracking peers that cause send failures.
if err != nil {
pack.p.Log().Debug("Transaction send failed", "err", err)
delete(pending, pack.p.Peer.ID())
}
// Schedule the next send.
if s := pick(); s != nil {
send(s)
}
case <-h.quitSync:
return
}
}
}
// chainSyncer coordinates blockchain sync components.
type chainSyncer struct {
handler *handler
force *time.Timer
forced bool // true when force timer fired
peerEventCh chan struct{}
doneCh chan error // non-nil when sync is running
}
// chainSyncOp is a scheduled sync operation.
type chainSyncOp struct {
mode downloader.SyncMode
peer *eth.Peer
td *big.Int
head common.Hash
}
// newChainSyncer creates a chainSyncer.
func newChainSyncer(handler *handler) *chainSyncer {
return &chainSyncer{
handler: handler,
peerEventCh: make(chan struct{}),
}
}
// handlePeerEvent notifies the syncer about a change in the peer set.
// This is called for new peers and every time a peer announces a new
// chain head.
func (cs *chainSyncer) handlePeerEvent(peer *eth.Peer) bool {
select {
case cs.peerEventCh <- struct{}{}:
return true
case <-cs.handler.quitSync:
return false
}
}
// loop runs in its own goroutine and launches the sync when necessary.
func (cs *chainSyncer) loop() {
defer cs.handler.wg.Done()
cs.handler.blockFetcher.Start()
cs.handler.txFetcher.Start()
defer cs.handler.blockFetcher.Stop()
defer cs.handler.txFetcher.Stop()
defer cs.handler.downloader.Terminate()
// The force timer lowers the peer count threshold down to one when it fires.
// This ensures we'll always start sync even if there aren't enough peers.
cs.force = time.NewTimer(forceSyncCycle)
defer cs.force.Stop()
for {
if op := cs.nextSyncOp(); op != nil {
cs.startSync(op)
}
select {
case <-cs.peerEventCh:
// Peer information changed, recheck.
case <-cs.doneCh:
cs.doneCh = nil
cs.force.Reset(forceSyncCycle)
cs.forced = false
case <-cs.force.C:
cs.forced = true
case <-cs.handler.quitSync:
// Disable all insertion on the blockchain. This needs to happen before
// terminating the downloader because the downloader waits for blockchain
// inserts, and these can take a long time to finish.
cs.handler.chain.StopInsert()
cs.handler.downloader.Terminate()
if cs.doneCh != nil {
<-cs.doneCh
}
return
}
}
}
// nextSyncOp determines whether sync is required at this time.
func (cs *chainSyncer) nextSyncOp() *chainSyncOp {
if cs.doneCh != nil {
return nil // Sync already running.
}
// Ensure we're at minimum peer count.
minPeers := defaultMinSyncPeers
if cs.forced {
minPeers = 1
} else if minPeers > cs.handler.maxPeers {
minPeers = cs.handler.maxPeers
}
if cs.handler.peers.len() < minPeers {
return nil
}
// We have enough peers, check TD
peer := cs.handler.peers.peerWithHighestTD()
if peer == nil {
return nil
}
mode, ourTD := cs.modeAndLocalHead()
if mode == downloader.FastSync && atomic.LoadUint32(&cs.handler.snapSync) == 1 {
// Fast sync via the snap protocol
mode = downloader.SnapSync
}
op := peerToSyncOp(mode, peer)
if op.td.Cmp(ourTD) <= 0 {
return nil // We're in sync.
}
return op
}
func peerToSyncOp(mode downloader.SyncMode, p *eth.Peer) *chainSyncOp {
peerHead, peerTD := p.Head()
return &chainSyncOp{mode: mode, peer: p, td: peerTD, head: peerHead}
}
func (cs *chainSyncer) modeAndLocalHead() (downloader.SyncMode, *big.Int) {
// If we're in fast sync mode, return that directly
if atomic.LoadUint32(&cs.handler.fastSync) == 1 {
block := cs.handler.chain.CurrentFastBlock()
td := cs.handler.chain.GetTdByHash(block.Hash())
return downloader.FastSync, td
}
// We are probably in full sync, but we might have rewound to before the
// fast sync pivot, check if we should reenable
if pivot := rawdb.ReadLastPivotNumber(cs.handler.database); pivot != nil {
if head := cs.handler.chain.CurrentBlock(); head.NumberU64() < *pivot {
block := cs.handler.chain.CurrentFastBlock()
td := cs.handler.chain.GetTdByHash(block.Hash())
return downloader.FastSync, td
}
}
// Nope, we're really full syncing
head := cs.handler.chain.CurrentBlock()
td := cs.handler.chain.GetTd(head.Hash(), head.NumberU64())
return downloader.FullSync, td
}
// startSync launches doSync in a new goroutine.
func (cs *chainSyncer) startSync(op *chainSyncOp) {
cs.doneCh = make(chan error, 1)
go func() { cs.doneCh <- cs.handler.doSync(op) }()
}
// doSync synchronizes the local blockchain with a remote peer.
func (h *handler) doSync(op *chainSyncOp) error {
if op.mode == downloader.FastSync || op.mode == downloader.SnapSync {
// Before launch the fast sync, we have to ensure user uses the same
// txlookup limit.
// The main concern here is: during the fast sync Geth won't index the
// block(generate tx indices) before the HEAD-limit. But if user changes
// the limit in the next fast sync(e.g. user kill Geth manually and
// restart) then it will be hard for Geth to figure out the oldest block
// has been indexed. So here for the user-experience wise, it's non-optimal
// that user can't change limit during the fast sync. If changed, Geth
// will just blindly use the original one.
limit := h.chain.TxLookupLimit()
if stored := rawdb.ReadFastTxLookupLimit(h.database); stored == nil {
rawdb.WriteFastTxLookupLimit(h.database, limit)
} else if *stored != limit {
h.chain.SetTxLookupLimit(*stored)
log.Warn("Update txLookup limit", "provided", limit, "updated", *stored)
}
}
// Run the sync cycle, and disable fast sync if we're past the pivot block
err := h.downloader.Synchronise(op.peer.ID(), op.head, op.td, op.mode)
if err != nil {
return err
}
if atomic.LoadUint32(&h.fastSync) == 1 {
log.Info("Fast sync complete, auto disabling")
atomic.StoreUint32(&h.fastSync, 0)
}
if atomic.LoadUint32(&h.snapSync) == 1 {
log.Info("Snap sync complete, auto disabling")
atomic.StoreUint32(&h.snapSync, 0)
}
// If we've successfully finished a sync cycle and passed any required checkpoint,
// enable accepting transactions from the network.
head := h.chain.CurrentBlock()
if head.NumberU64() >= h.checkpointNumber {
// Checkpoint passed, sanity check the timestamp to have a fallback mechanism
// for non-checkpointed (number = 0) private networks.
if head.Time() >= uint64(time.Now().AddDate(0, -1, 0).Unix()) {
atomic.StoreUint32(&h.acceptTxs, 1)
}
}
if head.NumberU64() > 0 {
// We've completed a sync cycle, notify all peers of new state. This path is
// essential in star-topology networks where a gateway node needs to notify
// all its out-of-date peers of the availability of a new block. This failure
// scenario will most often crop up in private and hackathon networks with
// degenerate connectivity, but it should be healthy for the mainnet too to
// more reliably update peers or the local TD state.
h.BroadcastBlock(head, false)
}
return nil
}