core: fix race conditions in txpool

core/tx_list.go

@@ -21,6 +21,8 @@ import (
 	"math"
 	"math/big"
 	"sort"
+	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/ethereum/go-ethereum/common"
@@ -478,9 +480,10 @@ func (h *priceHeap) Pop() interface{} {
 // better candidates for inclusion while in other cases (at the top of the baseFee peak)
 // the floating heap is better. When baseFee is decreasing they behave similarly.
 type txPricedList struct {
-	all              *txLookup // Pointer to the map of all transactions
-	urgent, floating priceHeap // Heaps of prices of all the stored **remote** transactions
-	stales           int       // Number of stale price points to (re-heap trigger)
+	all              *txLookup  // Pointer to the map of all transactions
+	urgent, floating priceHeap  // Heaps of prices of all the stored **remote** transactions
+	stales           int64      // Number of stale price points to (re-heap trigger)
+	reheapMu         sync.Mutex // Mutex asserts that only one routine is reheaping the list
 }
 
 const (
@@ -510,8 +513,8 @@ func (l *txPricedList) Put(tx *types.Transaction, local bool) {
 // the heap if a large enough ratio of transactions go stale.
 func (l *txPricedList) Removed(count int) {
 	// Bump the stale counter, but exit if still too low (< 25%)
-	l.stales += count
-	if l.stales <= (len(l.urgent.list)+len(l.floating.list))/4 {
+	stales := atomic.AddInt64(&l.stales, int64(count))
+	if int(stales) <= (len(l.urgent.list)+len(l.floating.list))/4 {
 		return
 	}
 	// Seems we've reached a critical number of stale transactions, reheap
@@ -535,7 +538,7 @@ func (l *txPricedList) underpricedFor(h *priceHeap, tx *types.Transaction) bool
 	for len(h.list) > 0 {
 		head := h.list[0]
 		if l.all.GetRemote(head.Hash()) == nil { // Removed or migrated
-			l.stales--
+			atomic.AddInt64(&l.stales, -1)
 			heap.Pop(h)
 			continue
 		}
@@ -561,7 +564,7 @@ func (l *txPricedList) Discard(slots int, force bool) (types.Transactions, bool)
 			// Discard stale transactions if found during cleanup
 			tx := heap.Pop(&l.urgent).(*types.Transaction)
 			if l.all.GetRemote(tx.Hash()) == nil { // Removed or migrated
-				l.stales--
+				atomic.AddInt64(&l.stales, -1)
 				continue
 			}
 			// Non stale transaction found, move to floating heap
@@ -574,7 +577,7 @@ func (l *txPricedList) Discard(slots int, force bool) (types.Transactions, bool)
 			// Discard stale transactions if found during cleanup
 			tx := heap.Pop(&l.floating).(*types.Transaction)
 			if l.all.GetRemote(tx.Hash()) == nil { // Removed or migrated
-				l.stales--
+				atomic.AddInt64(&l.stales, -1)
 				continue
 			}
 			// Non stale transaction found, discard it
@@ -594,8 +597,10 @@ func (l *txPricedList) Discard(slots int, force bool) (types.Transactions, bool)
 
 // Reheap forcibly rebuilds the heap based on the current remote transaction set.
 func (l *txPricedList) Reheap() {
+	l.reheapMu.Lock()
+	defer l.reheapMu.Unlock()
 	start := time.Now()
-	l.stales = 0
+	atomic.StoreInt64(&l.stales, 0)
 	l.urgent.list = make([]*types.Transaction, 0, l.all.RemoteCount())
 	l.all.Range(func(hash common.Hash, tx *types.Transaction, local bool) bool {
 		l.urgent.list = append(l.urgent.list, tx)

core/tx_pool.go

@@ -22,6 +22,7 @@ import (
 	"math/big"
 	"sort"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/ethereum/go-ethereum/common"
@@ -264,6 +265,7 @@ type TxPool struct {
 	reorgDoneCh     chan chan struct{}
 	reorgShutdownCh chan struct{}  // requests shutdown of scheduleReorgLoop
 	wg              sync.WaitGroup // tracks loop, scheduleReorgLoop
+	initDoneCh      chan struct{}  // is closed once the pool is initialized (for tests)
 
 	changesSinceReorg int // A counter for how many drops we've performed in-between reorg.
 }
@@ -294,6 +296,7 @@ func NewTxPool(config TxPoolConfig, chainconfig *params.ChainConfig, chain block
 		queueTxEventCh:  make(chan *types.Transaction),
 		reorgDoneCh:     make(chan chan struct{}),
 		reorgShutdownCh: make(chan struct{}),
+		initDoneCh:      make(chan struct{}),
 		gasPrice:        new(big.Int).SetUint64(config.PriceLimit),
 	}
 	pool.locals = newAccountSet(pool.signer)
@@ -347,6 +350,8 @@ func (pool *TxPool) loop() {
 	defer evict.Stop()
 	defer journal.Stop()
 
+	// Notify tests that the init phase is done
+	close(pool.initDoneCh)
 	for {
 		select {
 		// Handle ChainHeadEvent
@@ -365,8 +370,8 @@ func (pool *TxPool) loop() {
 		case <-report.C:
 			pool.mu.RLock()
 			pending, queued := pool.stats()
-			stales := pool.priced.stales
 			pool.mu.RUnlock()
+			stales := int(atomic.LoadInt64(&pool.priced.stales))
 
 			if pending != prevPending || queued != prevQueued || stales != prevStales {
 				log.Debug("Transaction pool status report", "executable", pending, "queued", queued, "stales", stales)

core/tx_pool_test.go

@@ -24,6 +24,7 @@ import (
 	"math/big"
 	"math/rand"
 	"os"
+	"sync/atomic"
 	"testing"
 	"time"
 
@@ -64,7 +65,7 @@ type testBlockChain struct {
 
 func (bc *testBlockChain) CurrentBlock() *types.Block {
 	return types.NewBlock(&types.Header{
-		GasLimit: bc.gasLimit,
+		GasLimit: atomic.LoadUint64(&bc.gasLimit),
 	}, nil, nil, nil, trie.NewStackTrie(nil))
 }
 
@@ -123,6 +124,8 @@ func setupTxPoolWithConfig(config *params.ChainConfig) (*TxPool, *ecdsa.PrivateK
 	key, _ := crypto.GenerateKey()
 	pool := NewTxPool(testTxPoolConfig, config, blockchain)
 
+	// wait for the pool to initialize
+	<-pool.initDoneCh
 	return pool, key
 }
 
@@ -625,7 +628,7 @@ func TestTransactionDropping(t *testing.T) {
 		t.Errorf("total transaction mismatch: have %d, want %d", pool.all.Count(), 4)
 	}
 	// Reduce the block gas limit, check that invalidated transactions are dropped
-	pool.chain.(*testBlockChain).gasLimit = 100
+	atomic.StoreUint64(&pool.chain.(*testBlockChain).gasLimit, 100)
 	<-pool.requestReset(nil, nil)
 
 	if _, ok := pool.pending[account].txs.items[tx0.Nonce()]; !ok {