eth/downloader: implement beacon sync (#23982)
* eth/downloader: implement beacon sync * eth/downloader: fix a crash if the beacon chain is reduced in length * eth/downloader: fix beacon sync start/stop thrashing data race * eth/downloader: use a non-nil pivot even in degenerate sync requests * eth/downloader: don't touch internal state on beacon Head retrieval * eth/downloader: fix spelling mistakes * eth/downloader: fix some typos * eth: integrate legacy/beacon sync switchover and UX * eth: handle UX wise being stuck on post-merge TTD * core, eth: integrate the beacon client with the beacon sync * eth/catalyst: make some warning messages nicer * eth/downloader: remove Ethereum 1&2 notions in favor of merge * core/beacon, eth: clean up engine API returns a bit * eth/downloader: add skeleton extension tests * eth/catalyst: keep non-kiln spec, handle mining on ttd * eth/downloader: add beacon header retrieval tests * eth: fixed spelling, commented failing tests out * eth/downloader: review fixes * eth/downloader: drop peers failing to deliver beacon headers * core/rawdb: track beacon sync data in db inspect * eth: fix review concerns * internal/web3ext: nit Co-authored-by: Marius van der Wijden <m.vanderwijden@live.de>
This commit is contained in:
Péter Szilágyi
GitHub
289
eth/downloader/beaconsync.go
Normal file
289
eth/downloader/beaconsync.go
Normal file
@@ -0,0 +1,289 @@
|
||||
// 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 downloader
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"sync"
|
||||
"sync/atomic"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
)
|
||||
|
||||
// beaconBackfiller is the chain and state backfilling that can be commenced once
|
||||
// the skeleton syncer has successfully reverse downloaded all the headers up to
|
||||
// the genesis block or an existing header in the database. Its operation is fully
|
||||
// directed by the skeleton sync's head/tail events.
|
||||
type beaconBackfiller struct {
|
||||
downloader *Downloader // Downloader to direct via this callback implementation
|
||||
syncMode SyncMode // Sync mode to use for backfilling the skeleton chains
|
||||
success func() // Callback to run on successful sync cycle completion
|
||||
filling bool // Flag whether the downloader is backfilling or not
|
||||
started chan struct{} // Notification channel whether the downloader inited
|
||||
lock sync.Mutex // Mutex protecting the sync lock
|
||||
}
|
||||
|
||||
// newBeaconBackfiller is a helper method to create the backfiller.
|
||||
func newBeaconBackfiller(dl *Downloader, success func()) backfiller {
|
||||
return &beaconBackfiller{
|
||||
downloader: dl,
|
||||
success: success,
|
||||
}
|
||||
}
|
||||
|
||||
// suspend cancels any background downloader threads.
|
||||
func (b *beaconBackfiller) suspend() {
|
||||
// If no filling is running, don't waste cycles
|
||||
b.lock.Lock()
|
||||
filling := b.filling
|
||||
started := b.started
|
||||
b.lock.Unlock()
|
||||
|
||||
if !filling {
|
||||
return
|
||||
}
|
||||
// A previous filling should be running, though it may happen that it hasn't
|
||||
// yet started (being done on a new goroutine). Many concurrent beacon head
|
||||
// announcements can lead to sync start/stop thrashing. In that case we need
|
||||
// to wait for initialization before we can safely cancel it. It is safe to
|
||||
// read this channel multiple times, it gets closed on startup.
|
||||
<-started
|
||||
|
||||
// Now that we're sure the downloader successfully started up, we can cancel
|
||||
// it safely without running the risk of data races.
|
||||
b.downloader.Cancel()
|
||||
}
|
||||
|
||||
// resume starts the downloader threads for backfilling state and chain data.
|
||||
func (b *beaconBackfiller) resume() {
|
||||
b.lock.Lock()
|
||||
if b.filling {
|
||||
// If a previous filling cycle is still running, just ignore this start
|
||||
// request. // TODO(karalabe): We should make this channel driven
|
||||
b.lock.Unlock()
|
||||
return
|
||||
}
|
||||
b.filling = true
|
||||
b.started = make(chan struct{})
|
||||
mode := b.syncMode
|
||||
b.lock.Unlock()
|
||||
|
||||
// Start the backfilling on its own thread since the downloader does not have
|
||||
// its own lifecycle runloop.
|
||||
go func() {
|
||||
// Set the backfiller to non-filling when download completes
|
||||
defer func() {
|
||||
b.lock.Lock()
|
||||
b.filling = false
|
||||
b.lock.Unlock()
|
||||
}()
|
||||
// If the downloader fails, report an error as in beacon chain mode there
|
||||
// should be no errors as long as the chain we're syncing to is valid.
|
||||
if err := b.downloader.synchronise("", common.Hash{}, nil, nil, mode, true, b.started); err != nil {
|
||||
log.Error("Beacon backfilling failed", "err", err)
|
||||
return
|
||||
}
|
||||
// Synchronization succeeded. Since this happens async, notify the outer
|
||||
// context to disable snap syncing and enable transaction propagation.
|
||||
if b.success != nil {
|
||||
b.success()
|
||||
}
|
||||
}()
|
||||
}
|
||||
|
||||
// setMode updates the sync mode from the current one to the requested one. If
|
||||
// there's an active sync in progress, it will be cancelled and restarted.
|
||||
func (b *beaconBackfiller) setMode(mode SyncMode) {
|
||||
// Update the old sync mode and track if it was changed
|
||||
b.lock.Lock()
|
||||
updated := b.syncMode != mode
|
||||
filling := b.filling
|
||||
b.syncMode = mode
|
||||
b.lock.Unlock()
|
||||
|
||||
// If the sync mode was changed mid-sync, restart. This should never ever
|
||||
// really happen, we just handle it to detect programming errors.
|
||||
if !updated || !filling {
|
||||
return
|
||||
}
|
||||
log.Error("Downloader sync mode changed mid-run", "old", mode.String(), "new", mode.String())
|
||||
b.suspend()
|
||||
b.resume()
|
||||
}
|
||||
|
||||
// BeaconSync is the post-merge version of the chain synchronization, where the
|
||||
// chain is not downloaded from genesis onward, rather from trusted head announces
|
||||
// backwards.
|
||||
//
|
||||
// Internally backfilling and state sync is done the same way, but the header
|
||||
// retrieval and scheduling is replaced.
|
||||
func (d *Downloader) BeaconSync(mode SyncMode, head *types.Header) error {
|
||||
return d.beaconSync(mode, head, true)
|
||||
}
|
||||
|
||||
// BeaconExtend is an optimistic version of BeaconSync, where an attempt is made
|
||||
// to extend the current beacon chain with a new header, but in case of a mismatch,
|
||||
// the old sync will not be terminated and reorged, rather the new head is dropped.
|
||||
//
|
||||
// This is useful if a beacon client is feeding us large chunks of payloads to run,
|
||||
// but is not setting the head after each.
|
||||
func (d *Downloader) BeaconExtend(mode SyncMode, head *types.Header) error {
|
||||
return d.beaconSync(mode, head, false)
|
||||
}
|
||||
|
||||
// beaconSync is the post-merge version of the chain synchronization, where the
|
||||
// chain is not downloaded from genesis onward, rather from trusted head announces
|
||||
// backwards.
|
||||
//
|
||||
// Internally backfilling and state sync is done the same way, but the header
|
||||
// retrieval and scheduling is replaced.
|
||||
func (d *Downloader) beaconSync(mode SyncMode, head *types.Header, force bool) error {
|
||||
// When the downloader starts a sync cycle, it needs to be aware of the sync
|
||||
// mode to use (full, snap). To keep the skeleton chain oblivious, inject the
|
||||
// mode into the backfiller directly.
|
||||
//
|
||||
// Super crazy dangerous type cast. Should be fine (TM), we're only using a
|
||||
// different backfiller implementation for skeleton tests.
|
||||
d.skeleton.filler.(*beaconBackfiller).setMode(mode)
|
||||
|
||||
// Signal the skeleton sync to switch to a new head, however it wants
|
||||
if err := d.skeleton.Sync(head, force); err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// findBeaconAncestor tries to locate the common ancestor link of the local chain
|
||||
// and the beacon chain just requested. In the general case when our node was in
|
||||
// sync and on the correct chain, checking the top N links should already get us
|
||||
// a match. In the rare scenario when we ended up on a long reorganisation (i.e.
|
||||
// none of the head links match), we do a binary search to find the ancestor.
|
||||
func (d *Downloader) findBeaconAncestor() uint64 {
|
||||
// Figure out the current local head position
|
||||
var chainHead *types.Header
|
||||
|
||||
switch d.getMode() {
|
||||
case FullSync:
|
||||
chainHead = d.blockchain.CurrentBlock().Header()
|
||||
case SnapSync:
|
||||
chainHead = d.blockchain.CurrentFastBlock().Header()
|
||||
default:
|
||||
chainHead = d.lightchain.CurrentHeader()
|
||||
}
|
||||
number := chainHead.Number.Uint64()
|
||||
|
||||
// If the head is present in the skeleton chain, return that
|
||||
if chainHead.Hash() == d.skeleton.Header(number).Hash() {
|
||||
return number
|
||||
}
|
||||
// Head header not present, binary search to find the ancestor
|
||||
start, end := uint64(0), number
|
||||
|
||||
beaconHead, err := d.skeleton.Head()
|
||||
if err != nil {
|
||||
panic(fmt.Sprintf("failed to read skeleton head: %v", err)) // can't reach this method without a head
|
||||
}
|
||||
if number := beaconHead.Number.Uint64(); end > number {
|
||||
// This shouldn't really happen in a healty network, but if the consensus
|
||||
// clients feeds us a shorter chain as the canonical, we should not attempt
|
||||
// to access non-existent skeleton items.
|
||||
log.Warn("Beacon head lower than local chain", "beacon", number, "local", end)
|
||||
end = number
|
||||
}
|
||||
for start+1 < end {
|
||||
// Split our chain interval in two, and request the hash to cross check
|
||||
check := (start + end) / 2
|
||||
|
||||
h := d.skeleton.Header(check)
|
||||
n := h.Number.Uint64()
|
||||
|
||||
var known bool
|
||||
switch d.getMode() {
|
||||
case FullSync:
|
||||
known = d.blockchain.HasBlock(h.Hash(), n)
|
||||
case SnapSync:
|
||||
known = d.blockchain.HasFastBlock(h.Hash(), n)
|
||||
default:
|
||||
known = d.lightchain.HasHeader(h.Hash(), n)
|
||||
}
|
||||
if !known {
|
||||
end = check
|
||||
continue
|
||||
}
|
||||
start = check
|
||||
}
|
||||
return start
|
||||
}
|
||||
|
||||
// fetchBeaconHeaders feeds skeleton headers to the downloader queue for scheduling
|
||||
// until sync errors or is finished.
|
||||
func (d *Downloader) fetchBeaconHeaders(from uint64) error {
|
||||
head, err := d.skeleton.Head()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
for {
|
||||
// Retrieve a batch of headers and feed it to the header processor
|
||||
var (
|
||||
headers = make([]*types.Header, 0, maxHeadersProcess)
|
||||
hashes = make([]common.Hash, 0, maxHeadersProcess)
|
||||
)
|
||||
for i := 0; i < maxHeadersProcess && from <= head.Number.Uint64(); i++ {
|
||||
headers = append(headers, d.skeleton.Header(from))
|
||||
hashes = append(hashes, headers[i].Hash())
|
||||
from++
|
||||
}
|
||||
if len(headers) > 0 {
|
||||
log.Trace("Scheduling new beacon headers", "count", len(headers), "from", from-uint64(len(headers)))
|
||||
select {
|
||||
case d.headerProcCh <- &headerTask{
|
||||
headers: headers,
|
||||
hashes: hashes,
|
||||
}:
|
||||
case <-d.cancelCh:
|
||||
return errCanceled
|
||||
}
|
||||
}
|
||||
// If we still have headers to import, loop and keep pushing them
|
||||
if from <= head.Number.Uint64() {
|
||||
continue
|
||||
}
|
||||
// If the pivot block is committed, signal header sync termination
|
||||
if atomic.LoadInt32(&d.committed) == 1 {
|
||||
select {
|
||||
case d.headerProcCh <- nil:
|
||||
return nil
|
||||
case <-d.cancelCh:
|
||||
return errCanceled
|
||||
}
|
||||
}
|
||||
// State sync still going, wait a bit for new headers and retry
|
||||
log.Trace("Pivot not yet committed, waiting...")
|
||||
select {
|
||||
case <-time.After(fsHeaderContCheck):
|
||||
case <-d.cancelCh:
|
||||
return errCanceled
|
||||
}
|
||||
head, err = d.skeleton.Head()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user