all: nuke total difficulty (#30744)

The total difficulty is the sum of all block difficulties from genesis
to a certain block. This value was used in PoW for deciding which chain
is heavier, and thus which chain to select. Since PoS has a different
fork selection algorithm, all blocks since the merge have a difficulty
of 0, and all total difficulties are the same for the past 2 years.

Whilst the TDs are mostly useless nowadays, there was never really a
reason to mess around removing them since they are so tiny. This
reasoning changes when we go down the path of pruned chain history. In
order to reconstruct any TD, we **must** retrieve all the headers from
chain head to genesis and then iterate all the difficulties to compute
the TD.

In a world where we completely prune past chain segments (bodies,
receipts, headers), it is not possible to reconstruct the TD at all. In
a world where we still keep chain headers and prune only the rest,
reconstructing it possible as long as we process (or download) the chain
forward from genesis, but trying to snap sync the head first and
backfill later hits the same issue, the TD becomes impossible to
calculate until genesis is backfilled.

All in all, the TD is a messy out-of-state, out-of-consensus computed
field that is overall useless nowadays, but code relying on it forces
the client into certain modes of operation and prevents other modes or
other optimizations. This PR completely nukes out the TD from the node.
It doesn't compute it, it doesn't operate on it, it's as if it didn't
even exist.

Caveats:

- Whenever we have APIs that return TD (devp2p handshake, tracer, etc.)
we return a TD of 0.
- For era files, we recompute the TD during export time (fairly quick)
to retain the format content.
- It is not possible to "verify" the merge point (i.e. with TD gone, TTD
is useless). Since we're not verifying PoW any more, just blindly trust
it, not verifying but blindly trusting the many year old merge point
seems just the same trust model.
- Our tests still need to be able to generate pre and post merge blocks,
so they need a new way to split the merge without TTD. The PR introduces
a settable ttdBlock field on the consensus object which is used by tests
as the block where originally the TTD happened. This is not needed for
live nodes, we never want to generate old blocks.
- One merge transition consensus test was disabled. With a
non-operational TD, testing how the client reacts to TTD is useless, it
cannot react.

Questions:

- Should we also drop total terminal difficulty from the genesis json?
It's a number we cannot react on any more, so maybe it would be cleaner
to get rid of even more concepts.

---------

Co-authored-by: Gary Rong <[email protected]>

cmd/devp2p/internal/ethtest/chain.go

@@ -143,11 +143,7 @@ func (c *Chain) ForkID() forkid.ID {
 // TD calculates the total difficulty of the chain at the
 // chain head.
 func (c *Chain) TD() *big.Int {
-	sum := new(big.Int)
-	for _, block := range c.blocks[:c.Len()] {
-		sum.Add(sum, block.Difficulty())
-	}
-	return sum
+	return new(big.Int)
 }
 
 // GetBlock returns the block at the specified number.

cmd/devp2p/internal/ethtest/suite.go

@@ -18,7 +18,6 @@ package ethtest
 
 import (
 	"crypto/rand"
-	"math/big"
 	"reflect"
 
 	"github.com/ethereum/go-ethereum/common"
@@ -74,7 +73,6 @@ func (s *Suite) EthTests() []utesting.Test {
 		{Name: "GetBlockBodies", Fn: s.TestGetBlockBodies},
 		// // malicious handshakes + status
 		{Name: "MaliciousHandshake", Fn: s.TestMaliciousHandshake},
-		{Name: "MaliciousStatus", Fn: s.TestMaliciousStatus},
 		// test transactions
 		{Name: "LargeTxRequest", Fn: s.TestLargeTxRequest, Slow: true},
 		{Name: "Transaction", Fn: s.TestTransaction},
@@ -453,42 +451,6 @@ func (s *Suite) TestMaliciousHandshake(t *utesting.T) {
 	}
 }
 
-func (s *Suite) TestMaliciousStatus(t *utesting.T) {
-	t.Log(`This test sends a malicious eth Status message to the node and expects a disconnect.`)
-
-	conn, err := s.dial()
-	if err != nil {
-		t.Fatalf("dial failed: %v", err)
-	}
-	defer conn.Close()
-	if err := conn.handshake(); err != nil {
-		t.Fatalf("handshake failed: %v", err)
-	}
-	// Create status with large total difficulty.
-	status := &eth.StatusPacket{
-		ProtocolVersion: uint32(conn.negotiatedProtoVersion),
-		NetworkID:       s.chain.config.ChainID.Uint64(),
-		TD:              new(big.Int).SetBytes(randBuf(2048)),
-		Head:            s.chain.Head().Hash(),
-		Genesis:         s.chain.GetBlock(0).Hash(),
-		ForkID:          s.chain.ForkID(),
-	}
-	if err := conn.statusExchange(s.chain, status); err != nil {
-		t.Fatalf("status exchange failed: %v", err)
-	}
-	// Wait for disconnect.
-	code, _, err := conn.Read()
-	if err != nil {
-		t.Fatalf("error reading from connection: %v", err)
-	}
-	switch code {
-	case discMsg:
-		break
-	default:
-		t.Fatalf("expected disconnect, got: %d", code)
-	}
-}
-
 func (s *Suite) TestTransaction(t *utesting.T) {
 	t.Log(`This test sends a valid transaction to the node and checks if the
 transaction gets propagated.`)

cmd/utils/cmd.go

@@ -25,6 +25,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"math/big"
 	"os"
 	"os/signal"
 	"path/filepath"
@@ -422,6 +423,10 @@ func ExportHistory(bc *core.BlockChain, dir string, first, last, step uint64) er
 		buf       = bytes.NewBuffer(nil)
 		checksums []string
 	)
+	td := new(big.Int)
+	for i := uint64(0); i < first; i++ {
+		td.Add(td, bc.GetHeaderByNumber(i).Difficulty)
+	}
 	for i := first; i <= last; i += step {
 		err := func() error {
 			filename := filepath.Join(dir, era.Filename(network, int(i/step), common.Hash{}))
@@ -444,11 +449,8 @@ func ExportHistory(bc *core.BlockChain, dir string, first, last, step uint64) er
 				if receipts == nil {
 					return fmt.Errorf("export failed on #%d: receipts not found", n)
 				}
-				td := bc.GetTd(block.Hash(), block.NumberU64())
-				if td == nil {
-					return fmt.Errorf("export failed on #%d: total difficulty not found", n)
-				}
-				if err := w.Add(block, receipts, td); err != nil {
+				td.Add(td, block.Difficulty())
+				if err := w.Add(block, receipts, new(big.Int).Set(td)); err != nil {
 					return err
 				}
 			}

consensus/beacon/consensus.go

@@ -61,7 +61,8 @@ var (
 // is only used for necessary consensus checks. The legacy consensus engine can be any
 // engine implements the consensus interface (except the beacon itself).
 type Beacon struct {
-	ethone consensus.Engine // Original consensus engine used in eth1, e.g. ethash or clique
+	ethone   consensus.Engine // Original consensus engine used in eth1, e.g. ethash or clique
+	ttdblock *uint64          // Merge block-number for testchain generation without TTDs
 }
 
 // New creates a consensus engine with the given embedded eth1 engine.
@@ -72,6 +73,18 @@ func New(ethone consensus.Engine) *Beacon {
 	return &Beacon{ethone: ethone}
 }
 
+// TestingTTDBlock is a replacement mechanism for TTD-based pre-/post-merge
+// splitting. With chain history deletion, TD calculations become impossible.
+// This is fine for progressing the live chain, but to be able to generate test
+// chains, we do need a split point. This method supports setting an explicit
+// block number to use as the splitter *for testing*, instead of having to keep
+// the notion of TDs in the client just for testing.
+//
+// The block with supplied number is regarded as the last pre-merge block.
+func (beacon *Beacon) TestingTTDBlock(number uint64) {
+	beacon.ttdblock = &number
+}
+
 // Author implements consensus.Engine, returning the verified author of the block.
 func (beacon *Beacon) Author(header *types.Header) (common.Address, error) {
 	if !beacon.IsPoSHeader(header) {
@@ -83,78 +96,63 @@ func (beacon *Beacon) Author(header *types.Header) (common.Address, error) {
 // VerifyHeader checks whether a header conforms to the consensus rules of the
 // stock Ethereum consensus engine.
 func (beacon *Beacon) VerifyHeader(chain consensus.ChainHeaderReader, header *types.Header) error {
-	reached, err := IsTTDReached(chain, header.ParentHash, header.Number.Uint64()-1)
-	if err != nil {
-		return err
-	}
-	if !reached {
-		return beacon.ethone.VerifyHeader(chain, header)
-	}
-	// Short circuit if the parent is not known
+	// During the live merge transition, the consensus engine used the terminal
+	// total difficulty to detect when PoW (PoA) switched to PoS. Maintaining the
+	// total difficulty values however require applying all the blocks from the
+	// genesis to build up the TD. This stops being a possibility if the tail of
+	// the chain is pruned already during sync.
+	//
+	// One heuristic that can be used to distinguish pre-merge and post-merge
+	// blocks is whether their *difficulty* is >0 or ==0 respectively. This of
+	// course would mean that we cannot prove anymore for a past chain that it
+	// truly transitioned at the correct TTD, but if we consider that ancient
+	// point in time finalized a long time ago, there should be no attempt from
+	// the consensus client to rewrite very old history.
+	//
+	// One thing that's probably not needed but which we can add to make this
+	// verification even stricter is to enforce that the chain can switch from
+	// >0 to ==0 TD only once by forbidding an ==0 to be followed by a >0.
+
+	// Verify that we're not reverting to pre-merge from post-merge
 	parent := chain.GetHeader(header.ParentHash, header.Number.Uint64()-1)
 	if parent == nil {
 		return consensus.ErrUnknownAncestor
 	}
-	// Sanity checks passed, do a proper verification
-	return beacon.verifyHeader(chain, header, parent)
-}
-
-// errOut constructs an error channel with prefilled errors inside.
-func errOut(n int, err error) chan error {
-	errs := make(chan error, n)
-	for i := 0; i < n; i++ {
-		errs <- err
+	if parent.Difficulty.Sign() == 0 && header.Difficulty.Sign() > 0 {
+		return consensus.ErrInvalidTerminalBlock
+	}
+	// Check >0 TDs with pre-merge, --0 TDs with post-merge rules
+	if header.Difficulty.Sign() > 0 {
+		return beacon.ethone.VerifyHeader(chain, header)
 	}
-	return errs
+	return beacon.verifyHeader(chain, header, parent)
 }
 
 // splitHeaders splits the provided header batch into two parts according to
-// the configured ttd. It requires the parent of header batch along with its
-// td are stored correctly in chain. If ttd is not configured yet, all headers
-// will be treated legacy PoW headers.
+// the difficulty field.
+//
 // Note, this function will not verify the header validity but just split them.
-func (beacon *Beacon) splitHeaders(chain consensus.ChainHeaderReader, headers []*types.Header) ([]*types.Header, []*types.Header, error) {
-	// TTD is not defined yet, all headers should be in legacy format.
-	ttd := chain.Config().TerminalTotalDifficulty
-	ptd := chain.GetTd(headers[0].ParentHash, headers[0].Number.Uint64()-1)
-	if ptd == nil {
-		return nil, nil, consensus.ErrUnknownAncestor
-	}
-	// The entire header batch already crosses the transition.
-	if ptd.Cmp(ttd) >= 0 {
-		return nil, headers, nil
-	}
+func (beacon *Beacon) splitHeaders(headers []*types.Header) ([]*types.Header, []*types.Header) {
 	var (
 		preHeaders  = headers
 		postHeaders []*types.Header
-		td          = new(big.Int).Set(ptd)
-		tdPassed    bool
 	)
 	for i, header := range headers {
-		if tdPassed {
+		if header.Difficulty.Sign() == 0 {
 			preHeaders = headers[:i]
 			postHeaders = headers[i:]
 			break
 		}
-		td = td.Add(td, header.Difficulty)
-		if td.Cmp(ttd) >= 0 {
-			// This is the last PoW header, it still belongs to
-			// the preHeaders, so we cannot split+break yet.
-			tdPassed = true
-		}
 	}
-	return preHeaders, postHeaders, nil
+	return preHeaders, postHeaders
 }
 
 // VerifyHeaders is similar to VerifyHeader, but verifies a batch of headers
 // concurrently. The method returns a quit channel to abort the operations and
 // a results channel to retrieve the async verifications.
 // VerifyHeaders expect the headers to be ordered and continuous.
 func (beacon *Beacon) VerifyHeaders(chain consensus.ChainHeaderReader, headers []*types.Header) (chan<- struct{}, <-chan error) {
-	preHeaders, postHeaders, err := beacon.splitHeaders(chain, headers)
-	if err != nil {
-		return make(chan struct{}), errOut(len(headers), err)
-	}
+	preHeaders, postHeaders := beacon.splitHeaders(headers)
 	if len(postHeaders) == 0 {
 		return beacon.ethone.VerifyHeaders(chain, headers)
 	}
@@ -334,12 +332,15 @@ func (beacon *Beacon) verifyHeaders(chain consensus.ChainHeaderReader, headers [
 // Prepare implements consensus.Engine, initializing the difficulty field of a
 // header to conform to the beacon protocol. The changes are done inline.
 func (beacon *Beacon) Prepare(chain consensus.ChainHeaderReader, header *types.Header) error {
-	// Transition isn't triggered yet, use the legacy rules for preparation.
-	reached, err := IsTTDReached(chain, header.ParentHash, header.Number.Uint64()-1)
-	if err != nil {
-		return err
-	}
-	if !reached {
+	// The beacon engine requires access to total difficulties to be able to
+	// seal pre-merge and post-merge blocks. With the transition to removing
+	// old blocks, TDs become unaccessible, thus making TTD based pre-/post-
+	// merge decisions impossible.
+	//
+	// We do not need to seal non-merge blocks anymore live, but we do need
+	// to be able to generate test chains, thus we're reverting to a testing-
+	// settable field to direct that.
+	if beacon.ttdblock != nil && *beacon.ttdblock >= header.Number.Uint64() {
 		return beacon.ethone.Prepare(chain, header)
 	}
 	header.Difficulty = beaconDifficulty
@@ -449,8 +450,15 @@ func (beacon *Beacon) SealHash(header *types.Header) common.Hash {
 // the difficulty that a new block should have when created at time
 // given the parent block's time and difficulty.
 func (beacon *Beacon) CalcDifficulty(chain consensus.ChainHeaderReader, time uint64, parent *types.Header) *big.Int {
-	// Transition isn't triggered yet, use the legacy rules for calculation
-	if reached, _ := IsTTDReached(chain, parent.Hash(), parent.Number.Uint64()); !reached {
+	// The beacon engine requires access to total difficulties to be able to
+	// seal pre-merge and post-merge blocks. With the transition to removing
+	// old blocks, TDs become unaccessible, thus making TTD based pre-/post-
+	// merge decisions impossible.
+	//
+	// We do not need to seal non-merge blocks anymore live, but we do need
+	// to be able to generate test chains, thus we're reverting to a testing-
+	// settable field to direct that.
+	if beacon.ttdblock != nil && *beacon.ttdblock > parent.Number.Uint64() {
 		return beacon.ethone.CalcDifficulty(chain, time, parent)
 	}
 	return beaconDifficulty
@@ -491,14 +499,3 @@ func (beacon *Beacon) SetThreads(threads int) {
 		th.SetThreads(threads)
 	}
 }
-
-// IsTTDReached checks if the TotalTerminalDifficulty has been surpassed on the `parentHash` block.
-// It depends on the parentHash already being stored in the database.
-// If the parentHash is not stored in the database a UnknownAncestor error is returned.
-func IsTTDReached(chain consensus.ChainHeaderReader, parentHash common.Hash, parentNumber uint64) (bool, error) {
-	td := chain.GetTd(parentHash, parentNumber)
-	if td == nil {
-		return false, consensus.ErrUnknownAncestor
-	}
-	return td.Cmp(chain.Config().TerminalTotalDifficulty) >= 0, nil
-}

consensus/consensus.go

@@ -45,9 +45,6 @@ type ChainHeaderReader interface {
 
 	// GetHeaderByHash retrieves a block header from the database by its hash.
 	GetHeaderByHash(hash common.Hash) *types.Header
-
-	// GetTd retrieves the total difficulty from the database by hash and number.
-	GetTd(hash common.Hash, number uint64) *big.Int
 }
 
 // ChainReader defines a small collection of methods needed to access the local

core/bench_test.go

@@ -285,7 +285,6 @@ func makeChainForBench(db ethdb.Database, genesis *Genesis, full bool, count uin
 
 		rawdb.WriteHeader(db, header)
 		rawdb.WriteCanonicalHash(db, hash, n)
-		rawdb.WriteTd(db, hash, n, big.NewInt(int64(n+1)))
 
 		if n == 0 {
 			rawdb.WriteChainConfig(db, hash, genesis.Config)