prysm-pulse/beacon-chain/blockchain/receive_attestation.go
terencechain 7fd2c08be3
Propagate FCU invalid error (#10997)
* Wrap fcu error

* Wrap fcu error

* Wrap error better

* More test

* Add else

* Potuz feedback

* Propagate the correct root for fcu

* Always return true for invalid

Co-authored-by: prylabs-bulldozer[bot] <58059840+prylabs-bulldozer[bot]@users.noreply.github.com>
2022-07-09 20:51:03 +00:00

267 lines
8.9 KiB
Go

package blockchain
import (
"bytes"
"context"
"fmt"
"time"
"github.com/pkg/errors"
"github.com/prysmaticlabs/prysm/async/event"
"github.com/prysmaticlabs/prysm/beacon-chain/core/feed"
"github.com/prysmaticlabs/prysm/beacon-chain/core/helpers"
"github.com/prysmaticlabs/prysm/beacon-chain/state"
"github.com/prysmaticlabs/prysm/config/params"
"github.com/prysmaticlabs/prysm/encoding/bytesutil"
ethpb "github.com/prysmaticlabs/prysm/proto/prysm/v1alpha1"
"github.com/prysmaticlabs/prysm/time/slots"
"github.com/sirupsen/logrus"
"go.opencensus.io/trace"
)
// AttestationStateFetcher allows for retrieving a beacon state corresponding to the block
// root of an attestation's target checkpoint.
type AttestationStateFetcher interface {
AttestationTargetState(ctx context.Context, target *ethpb.Checkpoint) (state.BeaconState, error)
}
// AttestationReceiver interface defines the methods of chain service receive and processing new attestations.
type AttestationReceiver interface {
AttestationStateFetcher
VerifyLmdFfgConsistency(ctx context.Context, att *ethpb.Attestation) error
VerifyFinalizedConsistency(ctx context.Context, root []byte) error
}
// AttestationTargetState returns the pre state of attestation.
func (s *Service) AttestationTargetState(ctx context.Context, target *ethpb.Checkpoint) (state.BeaconState, error) {
ss, err := slots.EpochStart(target.Epoch)
if err != nil {
return nil, err
}
if err := slots.ValidateClock(ss, uint64(s.genesisTime.Unix())); err != nil {
return nil, err
}
return s.getAttPreState(ctx, target)
}
// VerifyLmdFfgConsistency verifies that attestation's LMD and FFG votes are consistency to each other.
func (s *Service) VerifyLmdFfgConsistency(ctx context.Context, a *ethpb.Attestation) error {
targetSlot, err := slots.EpochStart(a.Data.Target.Epoch)
if err != nil {
return err
}
r, err := s.ancestor(ctx, a.Data.BeaconBlockRoot, targetSlot)
if err != nil {
return err
}
if !bytes.Equal(a.Data.Target.Root, r) {
return errors.New("FFG and LMD votes are not consistent")
}
return nil
}
// VerifyFinalizedConsistency verifies input root is consistent with finalized store.
// When the input root is not be consistent with finalized store then we know it is not
// on the finalized check point that leads to current canonical chain and should be rejected accordingly.
func (s *Service) VerifyFinalizedConsistency(ctx context.Context, root []byte) error {
// A canonical root implies the root to has an ancestor that aligns with finalized check point.
// In this case, we could exit early to save on additional computation.
blockRoot := bytesutil.ToBytes32(root)
if s.cfg.ForkChoiceStore.HasNode(blockRoot) && s.cfg.ForkChoiceStore.IsCanonical(blockRoot) {
return nil
}
f := s.FinalizedCheckpt()
ss, err := slots.EpochStart(f.Epoch)
if err != nil {
return err
}
r, err := s.ancestor(ctx, root, ss)
if err != nil {
return err
}
if !bytes.Equal(f.Root, r) {
return errors.New("Root and finalized store are not consistent")
}
return nil
}
// This routine processes fork choice attestations from the pool to account for validator votes and fork choice.
func (s *Service) spawnProcessAttestationsRoutine(stateFeed *event.Feed) {
// Wait for state to be initialized.
stateChannel := make(chan *feed.Event, 1)
stateSub := stateFeed.Subscribe(stateChannel)
go func() {
select {
case <-s.ctx.Done():
stateSub.Unsubscribe()
return
case <-stateChannel:
stateSub.Unsubscribe()
break
}
if s.genesisTime.IsZero() {
log.Warn("ProcessAttestations routine waiting for genesis time")
for s.genesisTime.IsZero() {
if err := s.ctx.Err(); err != nil {
log.WithError(err).Error("Giving up waiting for genesis time")
return
}
time.Sleep(1 * time.Second)
}
log.Warn("Genesis time received, now available to process attestations")
}
st := slots.NewSlotTicker(s.genesisTime, params.BeaconConfig().SecondsPerSlot)
for {
select {
case <-s.ctx.Done():
return
case <-st.C():
if err := s.ForkChoicer().NewSlot(s.ctx, s.CurrentSlot()); err != nil {
log.WithError(err).Error("Could not process new slot")
return
}
if err := s.UpdateHead(s.ctx); err != nil {
log.WithError(err).Error("Could not process attestations and update head")
return
}
}
}
}()
}
// UpdateHead updates the canonical head of the chain based on information from fork-choice attestations and votes.
// It requires no external inputs.
func (s *Service) UpdateHead(ctx context.Context) error {
// Continue when there's no fork choice attestation, there's nothing to process and update head.
// This covers the condition when the node is still initial syncing to the head of the chain.
if s.cfg.AttPool.ForkchoiceAttestationCount() == 0 {
return nil
}
// Only one process can process attestations and update head at a time.
s.processAttestationsLock.Lock()
defer s.processAttestationsLock.Unlock()
s.processAttestations(ctx)
justified := s.ForkChoicer().JustifiedCheckpoint()
balances, err := s.justifiedBalances.get(ctx, justified.Root)
if err != nil {
return err
}
newHeadRoot, err := s.cfg.ForkChoiceStore.Head(ctx, balances)
if err != nil {
log.WithError(err).Warn("Resolving fork due to new attestation")
}
s.headLock.RLock()
if s.headRoot() != newHeadRoot {
log.WithFields(logrus.Fields{
"oldHeadRoot": fmt.Sprintf("%#x", s.headRoot()),
"newHeadRoot": fmt.Sprintf("%#x", newHeadRoot),
}).Debug("Head changed due to attestations")
}
s.headLock.RUnlock()
if err := s.notifyEngineIfChangedHead(ctx, newHeadRoot); err != nil {
return err
}
return nil
}
// This calls notify Forkchoice Update in the event that the head has changed
func (s *Service) notifyEngineIfChangedHead(ctx context.Context, newHeadRoot [32]byte) error {
s.headLock.RLock()
if newHeadRoot == [32]byte{} || s.headRoot() == newHeadRoot {
s.headLock.RUnlock()
return nil
}
s.headLock.RUnlock()
if !s.hasBlockInInitSyncOrDB(ctx, newHeadRoot) {
log.Debug("New head does not exist in DB. Do nothing")
return nil // We don't have the block, don't notify the engine and update head.
}
newHeadBlock, err := s.getBlock(ctx, newHeadRoot)
if err != nil {
log.WithError(err).Error("Could not get new head block")
return nil
}
headState, err := s.cfg.StateGen.StateByRoot(ctx, newHeadRoot)
if err != nil {
log.WithError(err).Error("Could not get state from db")
return nil
}
arg := &notifyForkchoiceUpdateArg{
headState: headState,
headRoot: newHeadRoot,
headBlock: newHeadBlock.Block(),
}
_, err = s.notifyForkchoiceUpdate(s.ctx, arg)
if err != nil {
return err
}
if err := s.saveHead(ctx, newHeadRoot, newHeadBlock, headState); err != nil {
log.WithError(err).Error("could not save head")
}
return nil
}
// This processes fork choice attestations from the pool to account for validator votes and fork choice.
func (s *Service) processAttestations(ctx context.Context) {
atts := s.cfg.AttPool.ForkchoiceAttestations()
for _, a := range atts {
// Based on the spec, don't process the attestation until the subsequent slot.
// This delays consideration in the fork choice until their slot is in the past.
// https://github.com/ethereum/consensus-specs/blob/dev/specs/phase0/fork-choice.md#validate_on_attestation
nextSlot := a.Data.Slot + 1
if err := slots.VerifyTime(uint64(s.genesisTime.Unix()), nextSlot, params.BeaconNetworkConfig().MaximumGossipClockDisparity); err != nil {
continue
}
hasState := s.cfg.BeaconDB.HasStateSummary(ctx, bytesutil.ToBytes32(a.Data.BeaconBlockRoot))
hasBlock := s.hasBlock(ctx, bytesutil.ToBytes32(a.Data.BeaconBlockRoot))
if !(hasState && hasBlock) {
continue
}
if err := s.cfg.AttPool.DeleteForkchoiceAttestation(a); err != nil {
log.WithError(err).Error("Could not delete fork choice attestation in pool")
}
if !helpers.VerifyCheckpointEpoch(a.Data.Target, s.genesisTime) {
continue
}
if err := s.receiveAttestationNoPubsub(ctx, a); err != nil {
log.WithFields(logrus.Fields{
"slot": a.Data.Slot,
"committeeIndex": a.Data.CommitteeIndex,
"beaconBlockRoot": fmt.Sprintf("%#x", bytesutil.Trunc(a.Data.BeaconBlockRoot)),
"targetRoot": fmt.Sprintf("%#x", bytesutil.Trunc(a.Data.Target.Root)),
"aggregationCount": a.AggregationBits.Count(),
}).WithError(err).Warn("Could not process attestation for fork choice")
}
}
}
// receiveAttestationNoPubsub is a function that defines the operations that are performed on
// attestation that is received from regular sync. The operations consist of:
// 1. Validate attestation, update validator's latest vote
// 2. Apply fork choice to the processed attestation
// 3. Save latest head info
func (s *Service) receiveAttestationNoPubsub(ctx context.Context, att *ethpb.Attestation) error {
ctx, span := trace.StartSpan(ctx, "beacon-chain.blockchain.receiveAttestationNoPubsub")
defer span.End()
if err := s.OnAttestation(ctx, att); err != nil {
return errors.Wrap(err, "could not process attestation")
}
return nil
}