eth/downloader: increase downloader block body allowance (ethereum#23074)

This change increases the cache size from 64 to 256 Mb for block bodies.
Benchmarks have shown this to be one bottleneck when trying to achieve
higher download speeds.

The commit also includes a minor optimization for header inserts in package
core: previously, the presence of headers in the database was checked for
every header before writing it. With the change, if one header fails the
presence check, all subsequent headers are also assumed to be missing.
This is an improvement because in practice, the headers are almost always
missing during sync.

Author: holiman

core/headerchain.go

@@ -165,6 +165,7 @@ func (hc *HeaderChain) writeHeaders(headers []*types.Header) (result *headerWrit
 	)
 
 	batch := hc.chainDb.NewBatch()
+	parentKnown := true // Set to true to force hc.HasHeader check the first iteration
 	for i, header := range headers {
 		var hash common.Hash
 		// The headers have already been validated at this point, so we already
@@ -178,8 +179,10 @@ func (hc *HeaderChain) writeHeaders(headers []*types.Header) (result *headerWrit
 		number := header.Number.Uint64()
 		newTD.Add(newTD, header.Difficulty)
 
+		// If the parent was not present, store it
 		// If the header is already known, skip it, otherwise store
-		if !hc.HasHeader(hash, number) {
+		alreadyKnown := parentKnown && hc.HasHeader(hash, number)
+		if !alreadyKnown {
 			// Irrelevant of the canonical status, write the TD and header to the database.
 			rawdb.WriteTd(batch, hash, number, newTD)
 			hc.tdCache.Add(hash, new(big.Int).Set(newTD))
@@ -192,6 +195,7 @@ func (hc *HeaderChain) writeHeaders(headers []*types.Header) (result *headerWrit
 				firstInserted = i
 			}
 		}
+		parentKnown = alreadyKnown
 		lastHeader, lastHash, lastNumber = header, hash, number
 	}
 

eth/downloader/queue.go

@@ -40,10 +40,10 @@ const (
 )
 
 var (
-	blockCacheMaxItems     = 8192             // Maximum number of blocks to cache before throttling the download
-	blockCacheInitialItems = 2048             // Initial number of blocks to start fetching, before we know the sizes of the blocks
-	blockCacheMemory       = 64 * 1024 * 1024 // Maximum amount of memory to use for block caching
-	blockCacheSizeWeight   = 0.1              // Multiplier to approximate the average block size based on past ones
+	blockCacheMaxItems     = 8192              // Maximum number of blocks to cache before throttling the download
+	blockCacheInitialItems = 2048              // Initial number of blocks to start fetching, before we know the sizes of the blocks
+	blockCacheMemory       = 256 * 1024 * 1024 // Maximum amount of memory to use for block caching
+	blockCacheSizeWeight   = 0.1               // Multiplier to approximate the average block size based on past ones
 )
 
 var (
@@ -783,8 +783,9 @@ func (q *queue) DeliverHeaders(id string, headers []*types.Header, headerProcCh
 func (q *queue) DeliverBodies(id string, txLists [][]*types.Transaction, uncleLists [][]*types.Header) (int, error) {
 	q.lock.Lock()
 	defer q.lock.Unlock()
+	trieHasher := trie.NewStackTrie(nil)
 	validate := func(index int, header *types.Header) error {
-		if types.DeriveSha(types.Transactions(txLists[index]), trie.NewStackTrie(nil)) != header.TxHash {
+		if types.DeriveSha(types.Transactions(txLists[index]), trieHasher) != header.TxHash {
 			return errInvalidBody
 		}
 		if types.CalcUncleHash(uncleLists[index]) != header.UncleHash {
@@ -808,8 +809,9 @@ func (q *queue) DeliverBodies(id string, txLists [][]*types.Transaction, uncleLi
 func (q *queue) DeliverReceipts(id string, receiptList [][]*types.Receipt) (int, error) {
 	q.lock.Lock()
 	defer q.lock.Unlock()
+	trieHasher := trie.NewStackTrie(nil)
 	validate := func(index int, header *types.Header) error {
-		if types.DeriveSha(types.Receipts(receiptList[index]), trie.NewStackTrie(nil)) != header.ReceiptHash {
+		if types.DeriveSha(types.Receipts(receiptList[index]), trieHasher) != header.ReceiptHash {
 			return errInvalidReceipt
 		}
 		return nil