plugeth/eth/handler_eth_test.go
rjl493456442 dea1fb3cfc
all: cleanup tests (#25641)
Follow-up to PR #25523 to cleanup all relevant tests.
2022-09-07 20:21:59 +02:00

745 lines
26 KiB
Go

// Copyright 2020 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 (
"fmt"
"math/big"
"math/rand"
"sync/atomic"
"testing"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/consensus"
"github.com/ethereum/go-ethereum/consensus/ethash"
"github.com/ethereum/go-ethereum/core"
"github.com/ethereum/go-ethereum/core/forkid"
"github.com/ethereum/go-ethereum/core/rawdb"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/core/vm"
"github.com/ethereum/go-ethereum/eth/downloader"
"github.com/ethereum/go-ethereum/eth/protocols/eth"
"github.com/ethereum/go-ethereum/event"
"github.com/ethereum/go-ethereum/p2p"
"github.com/ethereum/go-ethereum/p2p/enode"
"github.com/ethereum/go-ethereum/params"
"github.com/ethereum/go-ethereum/rlp"
)
// testEthHandler is a mock event handler to listen for inbound network requests
// on the `eth` protocol and convert them into a more easily testable form.
type testEthHandler struct {
blockBroadcasts event.Feed
txAnnounces event.Feed
txBroadcasts event.Feed
}
func (h *testEthHandler) Chain() *core.BlockChain { panic("no backing chain") }
func (h *testEthHandler) TxPool() eth.TxPool { panic("no backing tx pool") }
func (h *testEthHandler) AcceptTxs() bool { return true }
func (h *testEthHandler) RunPeer(*eth.Peer, eth.Handler) error { panic("not used in tests") }
func (h *testEthHandler) PeerInfo(enode.ID) interface{} { panic("not used in tests") }
func (h *testEthHandler) Handle(peer *eth.Peer, packet eth.Packet) error {
switch packet := packet.(type) {
case *eth.NewBlockPacket:
h.blockBroadcasts.Send(packet.Block)
return nil
case *eth.NewPooledTransactionHashesPacket:
h.txAnnounces.Send(([]common.Hash)(*packet))
return nil
case *eth.TransactionsPacket:
h.txBroadcasts.Send(([]*types.Transaction)(*packet))
return nil
case *eth.PooledTransactionsPacket:
h.txBroadcasts.Send(([]*types.Transaction)(*packet))
return nil
default:
panic(fmt.Sprintf("unexpected eth packet type in tests: %T", packet))
}
}
// Tests that peers are correctly accepted (or rejected) based on the advertised
// fork IDs in the protocol handshake.
func TestForkIDSplit66(t *testing.T) { testForkIDSplit(t, eth.ETH66) }
func testForkIDSplit(t *testing.T, protocol uint) {
t.Parallel()
var (
engine = ethash.NewFaker()
configNoFork = &params.ChainConfig{HomesteadBlock: big.NewInt(1)}
configProFork = &params.ChainConfig{
HomesteadBlock: big.NewInt(1),
EIP150Block: big.NewInt(2),
EIP155Block: big.NewInt(2),
EIP158Block: big.NewInt(2),
ByzantiumBlock: big.NewInt(3),
}
dbNoFork = rawdb.NewMemoryDatabase()
dbProFork = rawdb.NewMemoryDatabase()
gspecNoFork = &core.Genesis{Config: configNoFork}
gspecProFork = &core.Genesis{Config: configProFork}
chainNoFork, _ = core.NewBlockChain(dbNoFork, nil, gspecNoFork, nil, engine, vm.Config{}, nil, nil)
chainProFork, _ = core.NewBlockChain(dbProFork, nil, gspecProFork, nil, engine, vm.Config{}, nil, nil)
_, blocksNoFork, _ = core.GenerateChainWithGenesis(gspecNoFork, engine, 2, nil)
_, blocksProFork, _ = core.GenerateChainWithGenesis(gspecProFork, engine, 2, nil)
ethNoFork, _ = newHandler(&handlerConfig{
Database: dbNoFork,
Chain: chainNoFork,
TxPool: newTestTxPool(),
Merger: consensus.NewMerger(rawdb.NewMemoryDatabase()),
Network: 1,
Sync: downloader.FullSync,
BloomCache: 1,
})
ethProFork, _ = newHandler(&handlerConfig{
Database: dbProFork,
Chain: chainProFork,
TxPool: newTestTxPool(),
Merger: consensus.NewMerger(rawdb.NewMemoryDatabase()),
Network: 1,
Sync: downloader.FullSync,
BloomCache: 1,
})
)
ethNoFork.Start(1000)
ethProFork.Start(1000)
// Clean up everything after ourselves
defer chainNoFork.Stop()
defer chainProFork.Stop()
defer ethNoFork.Stop()
defer ethProFork.Stop()
// Both nodes should allow the other to connect (same genesis, next fork is the same)
p2pNoFork, p2pProFork := p2p.MsgPipe()
defer p2pNoFork.Close()
defer p2pProFork.Close()
peerNoFork := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{1}, "", nil, p2pNoFork), p2pNoFork, nil)
peerProFork := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{2}, "", nil, p2pProFork), p2pProFork, nil)
defer peerNoFork.Close()
defer peerProFork.Close()
errc := make(chan error, 2)
go func(errc chan error) {
errc <- ethNoFork.runEthPeer(peerProFork, func(peer *eth.Peer) error { return nil })
}(errc)
go func(errc chan error) {
errc <- ethProFork.runEthPeer(peerNoFork, func(peer *eth.Peer) error { return nil })
}(errc)
for i := 0; i < 2; i++ {
select {
case err := <-errc:
if err != nil {
t.Fatalf("frontier nofork <-> profork failed: %v", err)
}
case <-time.After(250 * time.Millisecond):
t.Fatalf("frontier nofork <-> profork handler timeout")
}
}
// Progress into Homestead. Fork's match, so we don't care what the future holds
chainNoFork.InsertChain(blocksNoFork[:1])
chainProFork.InsertChain(blocksProFork[:1])
p2pNoFork, p2pProFork = p2p.MsgPipe()
defer p2pNoFork.Close()
defer p2pProFork.Close()
peerNoFork = eth.NewPeer(protocol, p2p.NewPeer(enode.ID{1}, "", nil), p2pNoFork, nil)
peerProFork = eth.NewPeer(protocol, p2p.NewPeer(enode.ID{2}, "", nil), p2pProFork, nil)
defer peerNoFork.Close()
defer peerProFork.Close()
errc = make(chan error, 2)
go func(errc chan error) {
errc <- ethNoFork.runEthPeer(peerProFork, func(peer *eth.Peer) error { return nil })
}(errc)
go func(errc chan error) {
errc <- ethProFork.runEthPeer(peerNoFork, func(peer *eth.Peer) error { return nil })
}(errc)
for i := 0; i < 2; i++ {
select {
case err := <-errc:
if err != nil {
t.Fatalf("homestead nofork <-> profork failed: %v", err)
}
case <-time.After(250 * time.Millisecond):
t.Fatalf("homestead nofork <-> profork handler timeout")
}
}
// Progress into Spurious. Forks mismatch, signalling differing chains, reject
chainNoFork.InsertChain(blocksNoFork[1:2])
chainProFork.InsertChain(blocksProFork[1:2])
p2pNoFork, p2pProFork = p2p.MsgPipe()
defer p2pNoFork.Close()
defer p2pProFork.Close()
peerNoFork = eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{1}, "", nil, p2pNoFork), p2pNoFork, nil)
peerProFork = eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{2}, "", nil, p2pProFork), p2pProFork, nil)
defer peerNoFork.Close()
defer peerProFork.Close()
errc = make(chan error, 2)
go func(errc chan error) {
errc <- ethNoFork.runEthPeer(peerProFork, func(peer *eth.Peer) error { return nil })
}(errc)
go func(errc chan error) {
errc <- ethProFork.runEthPeer(peerNoFork, func(peer *eth.Peer) error { return nil })
}(errc)
var successes int
for i := 0; i < 2; i++ {
select {
case err := <-errc:
if err == nil {
successes++
if successes == 2 { // Only one side disconnects
t.Fatalf("fork ID rejection didn't happen")
}
}
case <-time.After(250 * time.Millisecond):
t.Fatalf("split peers not rejected")
}
}
}
// Tests that received transactions are added to the local pool.
func TestRecvTransactions66(t *testing.T) { testRecvTransactions(t, eth.ETH66) }
func testRecvTransactions(t *testing.T, protocol uint) {
t.Parallel()
// Create a message handler, configure it to accept transactions and watch them
handler := newTestHandler()
defer handler.close()
handler.handler.acceptTxs = 1 // mark synced to accept transactions
txs := make(chan core.NewTxsEvent)
sub := handler.txpool.SubscribeNewTxsEvent(txs)
defer sub.Unsubscribe()
// Create a source peer to send messages through and a sink handler to receive them
p2pSrc, p2pSink := p2p.MsgPipe()
defer p2pSrc.Close()
defer p2pSink.Close()
src := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{1}, "", nil, p2pSrc), p2pSrc, handler.txpool)
sink := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{2}, "", nil, p2pSink), p2pSink, handler.txpool)
defer src.Close()
defer sink.Close()
go handler.handler.runEthPeer(sink, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(handler.handler), peer)
})
// Run the handshake locally to avoid spinning up a source handler
var (
genesis = handler.chain.Genesis()
head = handler.chain.CurrentBlock()
td = handler.chain.GetTd(head.Hash(), head.NumberU64())
)
if err := src.Handshake(1, td, head.Hash(), genesis.Hash(), forkid.NewIDWithChain(handler.chain), forkid.NewFilter(handler.chain)); err != nil {
t.Fatalf("failed to run protocol handshake")
}
// Send the transaction to the sink and verify that it's added to the tx pool
tx := types.NewTransaction(0, common.Address{}, big.NewInt(0), 100000, big.NewInt(0), nil)
tx, _ = types.SignTx(tx, types.HomesteadSigner{}, testKey)
if err := src.SendTransactions([]*types.Transaction{tx}); err != nil {
t.Fatalf("failed to send transaction: %v", err)
}
select {
case event := <-txs:
if len(event.Txs) != 1 {
t.Errorf("wrong number of added transactions: got %d, want 1", len(event.Txs))
} else if event.Txs[0].Hash() != tx.Hash() {
t.Errorf("added wrong tx hash: got %v, want %v", event.Txs[0].Hash(), tx.Hash())
}
case <-time.After(2 * time.Second):
t.Errorf("no NewTxsEvent received within 2 seconds")
}
}
// This test checks that pending transactions are sent.
func TestSendTransactions66(t *testing.T) { testSendTransactions(t, eth.ETH66) }
func testSendTransactions(t *testing.T, protocol uint) {
t.Parallel()
// Create a message handler and fill the pool with big transactions
handler := newTestHandler()
defer handler.close()
insert := make([]*types.Transaction, 100)
for nonce := range insert {
tx := types.NewTransaction(uint64(nonce), common.Address{}, big.NewInt(0), 100000, big.NewInt(0), make([]byte, 10240))
tx, _ = types.SignTx(tx, types.HomesteadSigner{}, testKey)
insert[nonce] = tx
}
go handler.txpool.AddRemotes(insert) // Need goroutine to not block on feed
time.Sleep(250 * time.Millisecond) // Wait until tx events get out of the system (can't use events, tx broadcaster races with peer join)
// Create a source handler to send messages through and a sink peer to receive them
p2pSrc, p2pSink := p2p.MsgPipe()
defer p2pSrc.Close()
defer p2pSink.Close()
src := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{1}, "", nil, p2pSrc), p2pSrc, handler.txpool)
sink := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{2}, "", nil, p2pSink), p2pSink, handler.txpool)
defer src.Close()
defer sink.Close()
go handler.handler.runEthPeer(src, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(handler.handler), peer)
})
// Run the handshake locally to avoid spinning up a source handler
var (
genesis = handler.chain.Genesis()
head = handler.chain.CurrentBlock()
td = handler.chain.GetTd(head.Hash(), head.NumberU64())
)
if err := sink.Handshake(1, td, head.Hash(), genesis.Hash(), forkid.NewIDWithChain(handler.chain), forkid.NewFilter(handler.chain)); err != nil {
t.Fatalf("failed to run protocol handshake")
}
// After the handshake completes, the source handler should stream the sink
// the transactions, subscribe to all inbound network events
backend := new(testEthHandler)
anns := make(chan []common.Hash)
annSub := backend.txAnnounces.Subscribe(anns)
defer annSub.Unsubscribe()
bcasts := make(chan []*types.Transaction)
bcastSub := backend.txBroadcasts.Subscribe(bcasts)
defer bcastSub.Unsubscribe()
go eth.Handle(backend, sink)
// Make sure we get all the transactions on the correct channels
seen := make(map[common.Hash]struct{})
for len(seen) < len(insert) {
switch protocol {
case 66:
select {
case hashes := <-anns:
for _, hash := range hashes {
if _, ok := seen[hash]; ok {
t.Errorf("duplicate transaction announced: %x", hash)
}
seen[hash] = struct{}{}
}
case <-bcasts:
t.Errorf("initial tx broadcast received on post eth/66")
}
default:
panic("unsupported protocol, please extend test")
}
}
for _, tx := range insert {
if _, ok := seen[tx.Hash()]; !ok {
t.Errorf("missing transaction: %x", tx.Hash())
}
}
}
// Tests that transactions get propagated to all attached peers, either via direct
// broadcasts or via announcements/retrievals.
func TestTransactionPropagation66(t *testing.T) { testTransactionPropagation(t, eth.ETH66) }
func testTransactionPropagation(t *testing.T, protocol uint) {
t.Parallel()
// Create a source handler to send transactions from and a number of sinks
// to receive them. We need multiple sinks since a one-to-one peering would
// broadcast all transactions without announcement.
source := newTestHandler()
source.handler.snapSync = 0 // Avoid requiring snap, otherwise some will be dropped below
defer source.close()
sinks := make([]*testHandler, 10)
for i := 0; i < len(sinks); i++ {
sinks[i] = newTestHandler()
defer sinks[i].close()
sinks[i].handler.acceptTxs = 1 // mark synced to accept transactions
}
// Interconnect all the sink handlers with the source handler
for i, sink := range sinks {
sink := sink // Closure for gorotuine below
sourcePipe, sinkPipe := p2p.MsgPipe()
defer sourcePipe.Close()
defer sinkPipe.Close()
sourcePeer := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{byte(i + 1)}, "", nil, sourcePipe), sourcePipe, source.txpool)
sinkPeer := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{0}, "", nil, sinkPipe), sinkPipe, sink.txpool)
defer sourcePeer.Close()
defer sinkPeer.Close()
go source.handler.runEthPeer(sourcePeer, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(source.handler), peer)
})
go sink.handler.runEthPeer(sinkPeer, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(sink.handler), peer)
})
}
// Subscribe to all the transaction pools
txChs := make([]chan core.NewTxsEvent, len(sinks))
for i := 0; i < len(sinks); i++ {
txChs[i] = make(chan core.NewTxsEvent, 1024)
sub := sinks[i].txpool.SubscribeNewTxsEvent(txChs[i])
defer sub.Unsubscribe()
}
// Fill the source pool with transactions and wait for them at the sinks
txs := make([]*types.Transaction, 1024)
for nonce := range txs {
tx := types.NewTransaction(uint64(nonce), common.Address{}, big.NewInt(0), 100000, big.NewInt(0), nil)
tx, _ = types.SignTx(tx, types.HomesteadSigner{}, testKey)
txs[nonce] = tx
}
source.txpool.AddRemotes(txs)
// Iterate through all the sinks and ensure they all got the transactions
for i := range sinks {
for arrived, timeout := 0, false; arrived < len(txs) && !timeout; {
select {
case event := <-txChs[i]:
arrived += len(event.Txs)
case <-time.After(time.Second):
t.Errorf("sink %d: transaction propagation timed out: have %d, want %d", i, arrived, len(txs))
timeout = true
}
}
}
}
// Tests that post eth protocol handshake, clients perform a mutual checkpoint
// challenge to validate each other's chains. Hash mismatches, or missing ones
// during a fast sync should lead to the peer getting dropped.
func TestCheckpointChallenge(t *testing.T) {
tests := []struct {
syncmode downloader.SyncMode
checkpoint bool
timeout bool
empty bool
match bool
drop bool
}{
// If checkpointing is not enabled locally, don't challenge and don't drop
{downloader.FullSync, false, false, false, false, false},
{downloader.SnapSync, false, false, false, false, false},
// If checkpointing is enabled locally and remote response is empty, only drop during fast sync
{downloader.FullSync, true, false, true, false, false},
{downloader.SnapSync, true, false, true, false, true}, // Special case, fast sync, unsynced peer
// If checkpointing is enabled locally and remote response mismatches, always drop
{downloader.FullSync, true, false, false, false, true},
{downloader.SnapSync, true, false, false, false, true},
// If checkpointing is enabled locally and remote response matches, never drop
{downloader.FullSync, true, false, false, true, false},
{downloader.SnapSync, true, false, false, true, false},
// If checkpointing is enabled locally and remote times out, always drop
{downloader.FullSync, true, true, false, true, true},
{downloader.SnapSync, true, true, false, true, true},
}
for _, tt := range tests {
t.Run(fmt.Sprintf("sync %v checkpoint %v timeout %v empty %v match %v", tt.syncmode, tt.checkpoint, tt.timeout, tt.empty, tt.match), func(t *testing.T) {
testCheckpointChallenge(t, tt.syncmode, tt.checkpoint, tt.timeout, tt.empty, tt.match, tt.drop)
})
}
}
func testCheckpointChallenge(t *testing.T, syncmode downloader.SyncMode, checkpoint bool, timeout bool, empty bool, match bool, drop bool) {
// Reduce the checkpoint handshake challenge timeout
defer func(old time.Duration) { syncChallengeTimeout = old }(syncChallengeTimeout)
syncChallengeTimeout = 250 * time.Millisecond
// Create a test handler and inject a CHT into it. The injection is a bit
// ugly, but it beats creating everything manually just to avoid reaching
// into the internals a bit.
handler := newTestHandler()
defer handler.close()
if syncmode == downloader.SnapSync {
atomic.StoreUint32(&handler.handler.snapSync, 1)
} else {
atomic.StoreUint32(&handler.handler.snapSync, 0)
}
var response *types.Header
if checkpoint {
number := (uint64(rand.Intn(500))+1)*params.CHTFrequency - 1
response = &types.Header{Number: big.NewInt(int64(number)), Extra: []byte("valid")}
handler.handler.checkpointNumber = number
handler.handler.checkpointHash = response.Hash()
}
// Create a challenger peer and a challenged one.
p2pLocal, p2pRemote := p2p.MsgPipe()
defer p2pLocal.Close()
defer p2pRemote.Close()
local := eth.NewPeer(eth.ETH66, p2p.NewPeerPipe(enode.ID{1}, "", nil, p2pLocal), p2pLocal, handler.txpool)
remote := eth.NewPeer(eth.ETH66, p2p.NewPeerPipe(enode.ID{2}, "", nil, p2pRemote), p2pRemote, handler.txpool)
defer local.Close()
defer remote.Close()
handlerDone := make(chan struct{})
go func() {
defer close(handlerDone)
handler.handler.runEthPeer(local, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(handler.handler), peer)
})
}()
// Run the handshake locally to avoid spinning up a remote handler.
var (
genesis = handler.chain.Genesis()
head = handler.chain.CurrentBlock()
td = handler.chain.GetTd(head.Hash(), head.NumberU64())
)
if err := remote.Handshake(1, td, head.Hash(), genesis.Hash(), forkid.NewIDWithChain(handler.chain), forkid.NewFilter(handler.chain)); err != nil {
t.Fatalf("failed to run protocol handshake")
}
// Connect a new peer and check that we receive the checkpoint challenge.
if checkpoint {
msg, err := p2pRemote.ReadMsg()
if err != nil {
t.Fatalf("failed to read checkpoint challenge: %v", err)
}
request := new(eth.GetBlockHeadersPacket66)
if err := msg.Decode(request); err != nil {
t.Fatalf("failed to decode checkpoint challenge: %v", err)
}
query := request.GetBlockHeadersPacket
if query.Origin.Number != response.Number.Uint64() || query.Amount != 1 || query.Skip != 0 || query.Reverse {
t.Fatalf("challenge mismatch: have [%d, %d, %d, %v] want [%d, %d, %d, %v]",
query.Origin.Number, query.Amount, query.Skip, query.Reverse,
response.Number.Uint64(), 1, 0, false)
}
// Create a block to reply to the challenge if no timeout is simulated.
if !timeout {
if empty {
if err := remote.ReplyBlockHeadersRLP(request.RequestId, []rlp.RawValue{}); err != nil {
t.Fatalf("failed to answer challenge: %v", err)
}
} else if match {
responseRlp, _ := rlp.EncodeToBytes(response)
if err := remote.ReplyBlockHeadersRLP(request.RequestId, []rlp.RawValue{responseRlp}); err != nil {
t.Fatalf("failed to answer challenge: %v", err)
}
} else {
responseRlp, _ := rlp.EncodeToBytes(&types.Header{Number: response.Number})
if err := remote.ReplyBlockHeadersRLP(request.RequestId, []rlp.RawValue{responseRlp}); err != nil {
t.Fatalf("failed to answer challenge: %v", err)
}
}
}
}
// Wait until the test timeout passes to ensure proper cleanup
time.Sleep(syncChallengeTimeout + 300*time.Millisecond)
// Verify that the remote peer is maintained or dropped.
if drop {
<-handlerDone
if peers := handler.handler.peers.len(); peers != 0 {
t.Fatalf("peer count mismatch: have %d, want %d", peers, 0)
}
} else {
if peers := handler.handler.peers.len(); peers != 1 {
t.Fatalf("peer count mismatch: have %d, want %d", peers, 1)
}
}
}
// Tests that blocks are broadcast to a sqrt number of peers only.
func TestBroadcastBlock1Peer(t *testing.T) { testBroadcastBlock(t, 1, 1) }
func TestBroadcastBlock2Peers(t *testing.T) { testBroadcastBlock(t, 2, 1) }
func TestBroadcastBlock3Peers(t *testing.T) { testBroadcastBlock(t, 3, 1) }
func TestBroadcastBlock4Peers(t *testing.T) { testBroadcastBlock(t, 4, 2) }
func TestBroadcastBlock5Peers(t *testing.T) { testBroadcastBlock(t, 5, 2) }
func TestBroadcastBlock8Peers(t *testing.T) { testBroadcastBlock(t, 9, 3) }
func TestBroadcastBlock12Peers(t *testing.T) { testBroadcastBlock(t, 12, 3) }
func TestBroadcastBlock16Peers(t *testing.T) { testBroadcastBlock(t, 16, 4) }
func TestBroadcastBloc26Peers(t *testing.T) { testBroadcastBlock(t, 26, 5) }
func TestBroadcastBlock100Peers(t *testing.T) { testBroadcastBlock(t, 100, 10) }
func testBroadcastBlock(t *testing.T, peers, bcasts int) {
t.Parallel()
// Create a source handler to broadcast blocks from and a number of sinks
// to receive them.
source := newTestHandlerWithBlocks(1)
defer source.close()
sinks := make([]*testEthHandler, peers)
for i := 0; i < len(sinks); i++ {
sinks[i] = new(testEthHandler)
}
// Interconnect all the sink handlers with the source handler
var (
genesis = source.chain.Genesis()
td = source.chain.GetTd(genesis.Hash(), genesis.NumberU64())
)
for i, sink := range sinks {
sink := sink // Closure for gorotuine below
sourcePipe, sinkPipe := p2p.MsgPipe()
defer sourcePipe.Close()
defer sinkPipe.Close()
sourcePeer := eth.NewPeer(eth.ETH66, p2p.NewPeerPipe(enode.ID{byte(i)}, "", nil, sourcePipe), sourcePipe, nil)
sinkPeer := eth.NewPeer(eth.ETH66, p2p.NewPeerPipe(enode.ID{0}, "", nil, sinkPipe), sinkPipe, nil)
defer sourcePeer.Close()
defer sinkPeer.Close()
go source.handler.runEthPeer(sourcePeer, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(source.handler), peer)
})
if err := sinkPeer.Handshake(1, td, genesis.Hash(), genesis.Hash(), forkid.NewIDWithChain(source.chain), forkid.NewFilter(source.chain)); err != nil {
t.Fatalf("failed to run protocol handshake")
}
go eth.Handle(sink, sinkPeer)
}
// Subscribe to all the transaction pools
blockChs := make([]chan *types.Block, len(sinks))
for i := 0; i < len(sinks); i++ {
blockChs[i] = make(chan *types.Block, 1)
defer close(blockChs[i])
sub := sinks[i].blockBroadcasts.Subscribe(blockChs[i])
defer sub.Unsubscribe()
}
// Initiate a block propagation across the peers
time.Sleep(100 * time.Millisecond)
source.handler.BroadcastBlock(source.chain.CurrentBlock(), true)
// Iterate through all the sinks and ensure the correct number got the block
done := make(chan struct{}, peers)
for _, ch := range blockChs {
ch := ch
go func() {
<-ch
done <- struct{}{}
}()
}
var received int
for {
select {
case <-done:
received++
case <-time.After(100 * time.Millisecond):
if received != bcasts {
t.Errorf("broadcast count mismatch: have %d, want %d", received, bcasts)
}
return
}
}
}
// Tests that a propagated malformed block (uncles or transactions don't match
// with the hashes in the header) gets discarded and not broadcast forward.
func TestBroadcastMalformedBlock66(t *testing.T) { testBroadcastMalformedBlock(t, eth.ETH66) }
func testBroadcastMalformedBlock(t *testing.T, protocol uint) {
t.Parallel()
// Create a source handler to broadcast blocks from and a number of sinks
// to receive them.
source := newTestHandlerWithBlocks(1)
defer source.close()
// Create a source handler to send messages through and a sink peer to receive them
p2pSrc, p2pSink := p2p.MsgPipe()
defer p2pSrc.Close()
defer p2pSink.Close()
src := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{1}, "", nil, p2pSrc), p2pSrc, source.txpool)
sink := eth.NewPeer(protocol, p2p.NewPeerPipe(enode.ID{2}, "", nil, p2pSink), p2pSink, source.txpool)
defer src.Close()
defer sink.Close()
go source.handler.runEthPeer(src, func(peer *eth.Peer) error {
return eth.Handle((*ethHandler)(source.handler), peer)
})
// Run the handshake locally to avoid spinning up a sink handler
var (
genesis = source.chain.Genesis()
td = source.chain.GetTd(genesis.Hash(), genesis.NumberU64())
)
if err := sink.Handshake(1, td, genesis.Hash(), genesis.Hash(), forkid.NewIDWithChain(source.chain), forkid.NewFilter(source.chain)); err != nil {
t.Fatalf("failed to run protocol handshake")
}
// After the handshake completes, the source handler should stream the sink
// the blocks, subscribe to inbound network events
backend := new(testEthHandler)
blocks := make(chan *types.Block, 1)
sub := backend.blockBroadcasts.Subscribe(blocks)
defer sub.Unsubscribe()
go eth.Handle(backend, sink)
// Create various combinations of malformed blocks
head := source.chain.CurrentBlock()
malformedUncles := head.Header()
malformedUncles.UncleHash[0]++
malformedTransactions := head.Header()
malformedTransactions.TxHash[0]++
malformedEverything := head.Header()
malformedEverything.UncleHash[0]++
malformedEverything.TxHash[0]++
// Try to broadcast all malformations and ensure they all get discarded
for _, header := range []*types.Header{malformedUncles, malformedTransactions, malformedEverything} {
block := types.NewBlockWithHeader(header).WithBody(head.Transactions(), head.Uncles())
if err := src.SendNewBlock(block, big.NewInt(131136)); err != nil {
t.Fatalf("failed to broadcast block: %v", err)
}
select {
case <-blocks:
t.Fatalf("malformed block forwarded")
case <-time.After(100 * time.Millisecond):
}
}
}