core, eth, les, trie: add a prefix to contract code (#21080)
This commit is contained in:
129
trie/database.go
129
trie/database.go
@ -27,6 +27,7 @@ import (
|
||||
|
||||
"github.com/VictoriaMetrics/fastcache"
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/rawdb"
|
||||
"github.com/ethereum/go-ethereum/ethdb"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/metrics"
|
||||
@ -57,15 +58,6 @@ var (
|
||||
memcacheCommitSizeMeter = metrics.NewRegisteredMeter("trie/memcache/commit/size", nil)
|
||||
)
|
||||
|
||||
// secureKeyPrefix is the database key prefix used to store trie node preimages.
|
||||
var secureKeyPrefix = []byte("secure-key-")
|
||||
|
||||
// secureKeyPrefixLength is the length of the above prefix
|
||||
const secureKeyPrefixLength = 11
|
||||
|
||||
// secureKeyLength is the length of the above prefix + 32byte hash.
|
||||
const secureKeyLength = secureKeyPrefixLength + 32
|
||||
|
||||
// Database is an intermediate write layer between the trie data structures and
|
||||
// the disk database. The aim is to accumulate trie writes in-memory and only
|
||||
// periodically flush a couple tries to disk, garbage collecting the remainder.
|
||||
@ -78,7 +70,7 @@ type Database struct {
|
||||
diskdb ethdb.KeyValueStore // Persistent storage for matured trie nodes
|
||||
|
||||
cleans *fastcache.Cache // GC friendly memory cache of clean node RLPs
|
||||
dirties map[common.Hash]*cachedNode // Data and references relationships of dirty nodes
|
||||
dirties map[common.Hash]*cachedNode // Data and references relationships of dirty trie nodes
|
||||
oldest common.Hash // Oldest tracked node, flush-list head
|
||||
newest common.Hash // Newest tracked node, flush-list tail
|
||||
|
||||
@ -139,8 +131,8 @@ type rawShortNode struct {
|
||||
func (n rawShortNode) cache() (hashNode, bool) { panic("this should never end up in a live trie") }
|
||||
func (n rawShortNode) fstring(ind string) string { panic("this should never end up in a live trie") }
|
||||
|
||||
// cachedNode is all the information we know about a single cached node in the
|
||||
// memory database write layer.
|
||||
// cachedNode is all the information we know about a single cached trie node
|
||||
// in the memory database write layer.
|
||||
type cachedNode struct {
|
||||
node node // Cached collapsed trie node, or raw rlp data
|
||||
size uint16 // Byte size of the useful cached data
|
||||
@ -161,8 +153,8 @@ var cachedNodeSize = int(reflect.TypeOf(cachedNode{}).Size())
|
||||
// reference map.
|
||||
const cachedNodeChildrenSize = 48
|
||||
|
||||
// rlp returns the raw rlp encoded blob of the cached node, either directly from
|
||||
// the cache, or by regenerating it from the collapsed node.
|
||||
// rlp returns the raw rlp encoded blob of the cached trie node, either directly
|
||||
// from the cache, or by regenerating it from the collapsed node.
|
||||
func (n *cachedNode) rlp() []byte {
|
||||
if node, ok := n.node.(rawNode); ok {
|
||||
return node
|
||||
@ -183,9 +175,9 @@ func (n *cachedNode) obj(hash common.Hash) node {
|
||||
return expandNode(hash[:], n.node)
|
||||
}
|
||||
|
||||
// forChilds invokes the callback for all the tracked children of this node,
|
||||
// both the implicit ones from inside the node as well as the explicit ones
|
||||
//from outside the node.
|
||||
// forChilds invokes the callback for all the tracked children of this node,
|
||||
// both the implicit ones from inside the node as well as the explicit ones
|
||||
// from outside the node.
|
||||
func (n *cachedNode) forChilds(onChild func(hash common.Hash)) {
|
||||
for child := range n.children {
|
||||
onChild(child)
|
||||
@ -305,25 +297,14 @@ func NewDatabaseWithCache(diskdb ethdb.KeyValueStore, cache int, journal string)
|
||||
}
|
||||
|
||||
// DiskDB retrieves the persistent storage backing the trie database.
|
||||
func (db *Database) DiskDB() ethdb.KeyValueReader {
|
||||
func (db *Database) DiskDB() ethdb.KeyValueStore {
|
||||
return db.diskdb
|
||||
}
|
||||
|
||||
// InsertBlob writes a new reference tracked blob to the memory database if it's
|
||||
// yet unknown. This method should only be used for non-trie nodes that require
|
||||
// reference counting, since trie nodes are garbage collected directly through
|
||||
// their embedded children.
|
||||
func (db *Database) InsertBlob(hash common.Hash, blob []byte) {
|
||||
db.lock.Lock()
|
||||
defer db.lock.Unlock()
|
||||
|
||||
db.insert(hash, len(blob), rawNode(blob))
|
||||
}
|
||||
|
||||
// insert inserts a collapsed trie node into the memory database. This method is
|
||||
// a more generic version of InsertBlob, supporting both raw blob insertions as
|
||||
// well ex trie node insertions. The blob size must be specified to allow proper
|
||||
// size tracking.
|
||||
// insert inserts a collapsed trie node into the memory database.
|
||||
// The blob size must be specified to allow proper size tracking.
|
||||
// All nodes inserted by this function will be reference tracked
|
||||
// and in theory should only used for **trie nodes** insertion.
|
||||
func (db *Database) insert(hash common.Hash, size int, node node) {
|
||||
// If the node's already cached, skip
|
||||
if _, ok := db.dirties[hash]; ok {
|
||||
@ -430,39 +411,30 @@ func (db *Database) Node(hash common.Hash) ([]byte, error) {
|
||||
memcacheDirtyMissMeter.Mark(1)
|
||||
|
||||
// Content unavailable in memory, attempt to retrieve from disk
|
||||
enc, err := db.diskdb.Get(hash[:])
|
||||
if err == nil && enc != nil {
|
||||
enc := rawdb.ReadTrieNode(db.diskdb, hash)
|
||||
if len(enc) != 0 {
|
||||
if db.cleans != nil {
|
||||
db.cleans.Set(hash[:], enc)
|
||||
memcacheCleanMissMeter.Mark(1)
|
||||
memcacheCleanWriteMeter.Mark(int64(len(enc)))
|
||||
}
|
||||
return enc, nil
|
||||
}
|
||||
return enc, err
|
||||
return nil, errors.New("not found")
|
||||
}
|
||||
|
||||
// preimage retrieves a cached trie node pre-image from memory. If it cannot be
|
||||
// found cached, the method queries the persistent database for the content.
|
||||
func (db *Database) preimage(hash common.Hash) ([]byte, error) {
|
||||
func (db *Database) preimage(hash common.Hash) []byte {
|
||||
// Retrieve the node from cache if available
|
||||
db.lock.RLock()
|
||||
preimage := db.preimages[hash]
|
||||
db.lock.RUnlock()
|
||||
|
||||
if preimage != nil {
|
||||
return preimage, nil
|
||||
return preimage
|
||||
}
|
||||
// Content unavailable in memory, attempt to retrieve from disk
|
||||
return db.diskdb.Get(secureKey(hash))
|
||||
}
|
||||
|
||||
// secureKey returns the database key for the preimage of key (as a newly
|
||||
// allocated byte-slice)
|
||||
func secureKey(hash common.Hash) []byte {
|
||||
buf := make([]byte, secureKeyLength)
|
||||
copy(buf, secureKeyPrefix)
|
||||
copy(buf[secureKeyPrefixLength:], hash[:])
|
||||
return buf
|
||||
return rawdb.ReadPreimage(db.diskdb, hash)
|
||||
}
|
||||
|
||||
// Nodes retrieves the hashes of all the nodes cached within the memory database.
|
||||
@ -482,6 +454,9 @@ func (db *Database) Nodes() []common.Hash {
|
||||
}
|
||||
|
||||
// Reference adds a new reference from a parent node to a child node.
|
||||
// This function is used to add reference between internal trie node
|
||||
// and external node(e.g. storage trie root), all internal trie nodes
|
||||
// are referenced together by database itself.
|
||||
func (db *Database) Reference(child common.Hash, parent common.Hash) {
|
||||
db.lock.Lock()
|
||||
defer db.lock.Unlock()
|
||||
@ -604,27 +579,16 @@ func (db *Database) Cap(limit common.StorageSize) error {
|
||||
size := db.dirtiesSize + common.StorageSize((len(db.dirties)-1)*cachedNodeSize)
|
||||
size += db.childrenSize - common.StorageSize(len(db.dirties[common.Hash{}].children)*(common.HashLength+2))
|
||||
|
||||
// We reuse an ephemeral buffer for the keys. The batch Put operation
|
||||
// copies it internally, so we can reuse it.
|
||||
var keyBuf [secureKeyLength]byte
|
||||
copy(keyBuf[:], secureKeyPrefix)
|
||||
|
||||
// If the preimage cache got large enough, push to disk. If it's still small
|
||||
// leave for later to deduplicate writes.
|
||||
flushPreimages := db.preimagesSize > 4*1024*1024
|
||||
if flushPreimages {
|
||||
for hash, preimage := range db.preimages {
|
||||
copy(keyBuf[secureKeyPrefixLength:], hash[:])
|
||||
if err := batch.Put(keyBuf[:], preimage); err != nil {
|
||||
log.Error("Failed to commit preimage from trie database", "err", err)
|
||||
rawdb.WritePreimages(batch, db.preimages)
|
||||
if batch.ValueSize() > ethdb.IdealBatchSize {
|
||||
if err := batch.Write(); err != nil {
|
||||
return err
|
||||
}
|
||||
if batch.ValueSize() > ethdb.IdealBatchSize {
|
||||
if err := batch.Write(); err != nil {
|
||||
return err
|
||||
}
|
||||
batch.Reset()
|
||||
}
|
||||
batch.Reset()
|
||||
}
|
||||
}
|
||||
// Keep committing nodes from the flush-list until we're below allowance
|
||||
@ -632,9 +596,8 @@ func (db *Database) Cap(limit common.StorageSize) error {
|
||||
for size > limit && oldest != (common.Hash{}) {
|
||||
// Fetch the oldest referenced node and push into the batch
|
||||
node := db.dirties[oldest]
|
||||
if err := batch.Put(oldest[:], node.rlp()); err != nil {
|
||||
return err
|
||||
}
|
||||
rawdb.WriteTrieNode(batch, oldest, node.rlp())
|
||||
|
||||
// If we exceeded the ideal batch size, commit and reset
|
||||
if batch.ValueSize() >= ethdb.IdealBatchSize {
|
||||
if err := batch.Write(); err != nil {
|
||||
@ -662,8 +625,7 @@ func (db *Database) Cap(limit common.StorageSize) error {
|
||||
defer db.lock.Unlock()
|
||||
|
||||
if flushPreimages {
|
||||
db.preimages = make(map[common.Hash][]byte)
|
||||
db.preimagesSize = 0
|
||||
db.preimages, db.preimagesSize = make(map[common.Hash][]byte), 0
|
||||
}
|
||||
for db.oldest != oldest {
|
||||
node := db.dirties[db.oldest]
|
||||
@ -706,25 +668,13 @@ func (db *Database) Commit(node common.Hash, report bool, callback func(common.H
|
||||
start := time.Now()
|
||||
batch := db.diskdb.NewBatch()
|
||||
|
||||
// We reuse an ephemeral buffer for the keys. The batch Put operation
|
||||
// copies it internally, so we can reuse it.
|
||||
var keyBuf [secureKeyLength]byte
|
||||
copy(keyBuf[:], secureKeyPrefix)
|
||||
|
||||
// Move all of the accumulated preimages into a write batch
|
||||
for hash, preimage := range db.preimages {
|
||||
copy(keyBuf[secureKeyPrefixLength:], hash[:])
|
||||
if err := batch.Put(keyBuf[:], preimage); err != nil {
|
||||
log.Error("Failed to commit preimage from trie database", "err", err)
|
||||
rawdb.WritePreimages(batch, db.preimages)
|
||||
if batch.ValueSize() > ethdb.IdealBatchSize {
|
||||
if err := batch.Write(); err != nil {
|
||||
return err
|
||||
}
|
||||
// If the batch is too large, flush to disk
|
||||
if batch.ValueSize() > ethdb.IdealBatchSize {
|
||||
if err := batch.Write(); err != nil {
|
||||
return err
|
||||
}
|
||||
batch.Reset()
|
||||
}
|
||||
batch.Reset()
|
||||
}
|
||||
// Since we're going to replay trie node writes into the clean cache, flush out
|
||||
// any batched pre-images before continuing.
|
||||
@ -754,8 +704,7 @@ func (db *Database) Commit(node common.Hash, report bool, callback func(common.H
|
||||
batch.Reset()
|
||||
|
||||
// Reset the storage counters and bumpd metrics
|
||||
db.preimages = make(map[common.Hash][]byte)
|
||||
db.preimagesSize = 0
|
||||
db.preimages, db.preimagesSize = make(map[common.Hash][]byte), 0
|
||||
|
||||
memcacheCommitTimeTimer.Update(time.Since(start))
|
||||
memcacheCommitSizeMeter.Mark(int64(storage - db.dirtiesSize))
|
||||
@ -791,13 +740,11 @@ func (db *Database) commit(hash common.Hash, batch ethdb.Batch, uncacher *cleane
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if err := batch.Put(hash[:], node.rlp()); err != nil {
|
||||
return err
|
||||
}
|
||||
// If we've reached an optimal batch size, commit and start over
|
||||
rawdb.WriteTrieNode(batch, hash, node.rlp())
|
||||
if callback != nil {
|
||||
callback(hash)
|
||||
}
|
||||
// If we've reached an optimal batch size, commit and start over
|
||||
if batch.ValueSize() >= ethdb.IdealBatchSize {
|
||||
if err := batch.Write(); err != nil {
|
||||
return err
|
||||
|
@ -130,8 +130,7 @@ func (t *SecureTrie) GetKey(shaKey []byte) []byte {
|
||||
if key, ok := t.getSecKeyCache()[string(shaKey)]; ok {
|
||||
return key
|
||||
}
|
||||
key, _ := t.trie.db.preimage(common.BytesToHash(shaKey))
|
||||
return key
|
||||
return t.trie.db.preimage(common.BytesToHash(shaKey))
|
||||
}
|
||||
|
||||
// Commit writes all nodes and the secure hash pre-images to the trie's database.
|
||||
|
187
trie/sync.go
187
trie/sync.go
@ -22,6 +22,7 @@ import (
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/common/prque"
|
||||
"github.com/ethereum/go-ethereum/core/rawdb"
|
||||
"github.com/ethereum/go-ethereum/ethdb"
|
||||
)
|
||||
|
||||
@ -37,7 +38,7 @@ var ErrAlreadyProcessed = errors.New("already processed")
|
||||
type request struct {
|
||||
hash common.Hash // Hash of the node data content to retrieve
|
||||
data []byte // Data content of the node, cached until all subtrees complete
|
||||
raw bool // Whether this is a raw entry (code) or a trie node
|
||||
code bool // Whether this is a code entry
|
||||
|
||||
parents []*request // Parent state nodes referencing this entry (notify all upon completion)
|
||||
depth int // Depth level within the trie the node is located to prioritise DFS
|
||||
@ -46,8 +47,7 @@ type request struct {
|
||||
callback LeafCallback // Callback to invoke if a leaf node it reached on this branch
|
||||
}
|
||||
|
||||
// SyncResult is a simple list to return missing nodes along with their request
|
||||
// hashes.
|
||||
// SyncResult is a response with requested data along with it's hash.
|
||||
type SyncResult struct {
|
||||
Hash common.Hash // Hash of the originally unknown trie node
|
||||
Data []byte // Data content of the retrieved node
|
||||
@ -56,25 +56,40 @@ type SyncResult struct {
|
||||
// syncMemBatch is an in-memory buffer of successfully downloaded but not yet
|
||||
// persisted data items.
|
||||
type syncMemBatch struct {
|
||||
batch map[common.Hash][]byte // In-memory membatch of recently completed items
|
||||
nodes map[common.Hash][]byte // In-memory membatch of recently completed nodes
|
||||
codes map[common.Hash][]byte // In-memory membatch of recently completed codes
|
||||
}
|
||||
|
||||
// newSyncMemBatch allocates a new memory-buffer for not-yet persisted trie nodes.
|
||||
func newSyncMemBatch() *syncMemBatch {
|
||||
return &syncMemBatch{
|
||||
batch: make(map[common.Hash][]byte),
|
||||
nodes: make(map[common.Hash][]byte),
|
||||
codes: make(map[common.Hash][]byte),
|
||||
}
|
||||
}
|
||||
|
||||
// hasNode reports the trie node with specific hash is already cached.
|
||||
func (batch *syncMemBatch) hasNode(hash common.Hash) bool {
|
||||
_, ok := batch.nodes[hash]
|
||||
return ok
|
||||
}
|
||||
|
||||
// hasCode reports the contract code with specific hash is already cached.
|
||||
func (batch *syncMemBatch) hasCode(hash common.Hash) bool {
|
||||
_, ok := batch.codes[hash]
|
||||
return ok
|
||||
}
|
||||
|
||||
// Sync is the main state trie synchronisation scheduler, which provides yet
|
||||
// unknown trie hashes to retrieve, accepts node data associated with said hashes
|
||||
// and reconstructs the trie step by step until all is done.
|
||||
type Sync struct {
|
||||
database ethdb.KeyValueReader // Persistent database to check for existing entries
|
||||
membatch *syncMemBatch // Memory buffer to avoid frequent database writes
|
||||
requests map[common.Hash]*request // Pending requests pertaining to a key hash
|
||||
nodeReqs map[common.Hash]*request // Pending requests pertaining to a trie node hash
|
||||
codeReqs map[common.Hash]*request // Pending requests pertaining to a code hash
|
||||
queue *prque.Prque // Priority queue with the pending requests
|
||||
bloom *SyncBloom // Bloom filter for fast node existence checks
|
||||
bloom *SyncBloom // Bloom filter for fast state existence checks
|
||||
}
|
||||
|
||||
// NewSync creates a new trie data download scheduler.
|
||||
@ -82,7 +97,8 @@ func NewSync(root common.Hash, database ethdb.KeyValueReader, callback LeafCallb
|
||||
ts := &Sync{
|
||||
database: database,
|
||||
membatch: newSyncMemBatch(),
|
||||
requests: make(map[common.Hash]*request),
|
||||
nodeReqs: make(map[common.Hash]*request),
|
||||
codeReqs: make(map[common.Hash]*request),
|
||||
queue: prque.New(nil),
|
||||
bloom: bloom,
|
||||
}
|
||||
@ -96,13 +112,15 @@ func (s *Sync) AddSubTrie(root common.Hash, depth int, parent common.Hash, callb
|
||||
if root == emptyRoot {
|
||||
return
|
||||
}
|
||||
if _, ok := s.membatch.batch[root]; ok {
|
||||
if s.membatch.hasNode(root) {
|
||||
return
|
||||
}
|
||||
if s.bloom == nil || s.bloom.Contains(root[:]) {
|
||||
// Bloom filter says this might be a duplicate, double check
|
||||
blob, _ := s.database.Get(root[:])
|
||||
if local, err := decodeNode(root[:], blob); local != nil && err == nil {
|
||||
// Bloom filter says this might be a duplicate, double check.
|
||||
// If database says yes, then at least the trie node is present
|
||||
// and we hold the assumption that it's NOT legacy contract code.
|
||||
blob := rawdb.ReadTrieNode(s.database, root)
|
||||
if len(blob) > 0 {
|
||||
return
|
||||
}
|
||||
// False positive, bump fault meter
|
||||
@ -116,7 +134,7 @@ func (s *Sync) AddSubTrie(root common.Hash, depth int, parent common.Hash, callb
|
||||
}
|
||||
// If this sub-trie has a designated parent, link them together
|
||||
if parent != (common.Hash{}) {
|
||||
ancestor := s.requests[parent]
|
||||
ancestor := s.nodeReqs[parent]
|
||||
if ancestor == nil {
|
||||
panic(fmt.Sprintf("sub-trie ancestor not found: %x", parent))
|
||||
}
|
||||
@ -126,21 +144,25 @@ func (s *Sync) AddSubTrie(root common.Hash, depth int, parent common.Hash, callb
|
||||
s.schedule(req)
|
||||
}
|
||||
|
||||
// AddRawEntry schedules the direct retrieval of a state entry that should not be
|
||||
// interpreted as a trie node, but rather accepted and stored into the database
|
||||
// as is. This method's goal is to support misc state metadata retrievals (e.g.
|
||||
// contract code).
|
||||
func (s *Sync) AddRawEntry(hash common.Hash, depth int, parent common.Hash) {
|
||||
// AddCodeEntry schedules the direct retrieval of a contract code that should not
|
||||
// be interpreted as a trie node, but rather accepted and stored into the database
|
||||
// as is.
|
||||
func (s *Sync) AddCodeEntry(hash common.Hash, depth int, parent common.Hash) {
|
||||
// Short circuit if the entry is empty or already known
|
||||
if hash == emptyState {
|
||||
return
|
||||
}
|
||||
if _, ok := s.membatch.batch[hash]; ok {
|
||||
if s.membatch.hasCode(hash) {
|
||||
return
|
||||
}
|
||||
if s.bloom == nil || s.bloom.Contains(hash[:]) {
|
||||
// Bloom filter says this might be a duplicate, double check
|
||||
if ok, _ := s.database.Has(hash[:]); ok {
|
||||
// Bloom filter says this might be a duplicate, double check.
|
||||
// If database says yes, the blob is present for sure.
|
||||
// Note we only check the existence with new code scheme, fast
|
||||
// sync is expected to run with a fresh new node. Even there
|
||||
// exists the code with legacy format, fetch and store with
|
||||
// new scheme anyway.
|
||||
if blob := rawdb.ReadCodeWithPrefix(s.database, hash); len(blob) > 0 {
|
||||
return
|
||||
}
|
||||
// False positive, bump fault meter
|
||||
@ -149,12 +171,12 @@ func (s *Sync) AddRawEntry(hash common.Hash, depth int, parent common.Hash) {
|
||||
// Assemble the new sub-trie sync request
|
||||
req := &request{
|
||||
hash: hash,
|
||||
raw: true,
|
||||
code: true,
|
||||
depth: depth,
|
||||
}
|
||||
// If this sub-trie has a designated parent, link them together
|
||||
if parent != (common.Hash{}) {
|
||||
ancestor := s.requests[parent]
|
||||
ancestor := s.nodeReqs[parent] // the parent of codereq can ONLY be nodereq
|
||||
if ancestor == nil {
|
||||
panic(fmt.Sprintf("raw-entry ancestor not found: %x", parent))
|
||||
}
|
||||
@ -173,61 +195,64 @@ func (s *Sync) Missing(max int) []common.Hash {
|
||||
return requests
|
||||
}
|
||||
|
||||
// Process injects a batch of retrieved trie nodes data, returning if something
|
||||
// was committed to the database and also the index of an entry if its processing
|
||||
// failed.
|
||||
func (s *Sync) Process(results []SyncResult) (bool, int, error) {
|
||||
committed := false
|
||||
|
||||
for i, item := range results {
|
||||
// If the item was not requested, bail out
|
||||
request := s.requests[item.Hash]
|
||||
if request == nil {
|
||||
return committed, i, ErrNotRequested
|
||||
}
|
||||
if request.data != nil {
|
||||
return committed, i, ErrAlreadyProcessed
|
||||
}
|
||||
// If the item is a raw entry request, commit directly
|
||||
if request.raw {
|
||||
request.data = item.Data
|
||||
s.commit(request)
|
||||
committed = true
|
||||
continue
|
||||
}
|
||||
// Process injects the received data for requested item. Note it can
|
||||
// happpen that the single response commits two pending requests(e.g.
|
||||
// there are two requests one for code and one for node but the hash
|
||||
// is same). In this case the second response for the same hash will
|
||||
// be treated as "non-requested" item or "already-processed" item but
|
||||
// there is no downside.
|
||||
func (s *Sync) Process(result SyncResult) error {
|
||||
// If the item was not requested either for code or node, bail out
|
||||
if s.nodeReqs[result.Hash] == nil && s.codeReqs[result.Hash] == nil {
|
||||
return ErrNotRequested
|
||||
}
|
||||
// There is an pending code request for this data, commit directly
|
||||
var filled bool
|
||||
if req := s.codeReqs[result.Hash]; req != nil && req.data == nil {
|
||||
filled = true
|
||||
req.data = result.Data
|
||||
s.commit(req)
|
||||
}
|
||||
// There is an pending node request for this data, fill it.
|
||||
if req := s.nodeReqs[result.Hash]; req != nil && req.data == nil {
|
||||
filled = true
|
||||
// Decode the node data content and update the request
|
||||
node, err := decodeNode(item.Hash[:], item.Data)
|
||||
node, err := decodeNode(result.Hash[:], result.Data)
|
||||
if err != nil {
|
||||
return committed, i, err
|
||||
return err
|
||||
}
|
||||
request.data = item.Data
|
||||
req.data = result.Data
|
||||
|
||||
// Create and schedule a request for all the children nodes
|
||||
requests, err := s.children(request, node)
|
||||
requests, err := s.children(req, node)
|
||||
if err != nil {
|
||||
return committed, i, err
|
||||
return err
|
||||
}
|
||||
if len(requests) == 0 && request.deps == 0 {
|
||||
s.commit(request)
|
||||
committed = true
|
||||
continue
|
||||
}
|
||||
request.deps += len(requests)
|
||||
for _, child := range requests {
|
||||
s.schedule(child)
|
||||
if len(requests) == 0 && req.deps == 0 {
|
||||
s.commit(req)
|
||||
} else {
|
||||
req.deps += len(requests)
|
||||
for _, child := range requests {
|
||||
s.schedule(child)
|
||||
}
|
||||
}
|
||||
}
|
||||
return committed, 0, nil
|
||||
if !filled {
|
||||
return ErrAlreadyProcessed
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// Commit flushes the data stored in the internal membatch out to persistent
|
||||
// storage, returning any occurred error.
|
||||
func (s *Sync) Commit(dbw ethdb.Batch) error {
|
||||
// Dump the membatch into a database dbw
|
||||
for key, value := range s.membatch.batch {
|
||||
if err := dbw.Put(key[:], value); err != nil {
|
||||
return err
|
||||
}
|
||||
for key, value := range s.membatch.nodes {
|
||||
rawdb.WriteTrieNode(dbw, key, value)
|
||||
s.bloom.Add(key[:])
|
||||
}
|
||||
for key, value := range s.membatch.codes {
|
||||
rawdb.WriteCode(dbw, key, value)
|
||||
s.bloom.Add(key[:])
|
||||
}
|
||||
// Drop the membatch data and return
|
||||
@ -237,21 +262,30 @@ func (s *Sync) Commit(dbw ethdb.Batch) error {
|
||||
|
||||
// Pending returns the number of state entries currently pending for download.
|
||||
func (s *Sync) Pending() int {
|
||||
return len(s.requests)
|
||||
return len(s.nodeReqs) + len(s.codeReqs)
|
||||
}
|
||||
|
||||
// schedule inserts a new state retrieval request into the fetch queue. If there
|
||||
// is already a pending request for this node, the new request will be discarded
|
||||
// and only a parent reference added to the old one.
|
||||
func (s *Sync) schedule(req *request) {
|
||||
var reqset = s.nodeReqs
|
||||
if req.code {
|
||||
reqset = s.codeReqs
|
||||
}
|
||||
// If we're already requesting this node, add a new reference and stop
|
||||
if old, ok := s.requests[req.hash]; ok {
|
||||
if old, ok := reqset[req.hash]; ok {
|
||||
old.parents = append(old.parents, req.parents...)
|
||||
return
|
||||
}
|
||||
// Schedule the request for future retrieval
|
||||
reqset[req.hash] = req
|
||||
|
||||
// Schedule the request for future retrieval. This queue is shared
|
||||
// by both node requests and code requests. It can happen that there
|
||||
// is a trie node and code has same hash. In this case two elements
|
||||
// with same hash and same or different depth will be pushed. But it's
|
||||
// ok the worst case is the second response will be treated as duplicated.
|
||||
s.queue.Push(req.hash, int64(req.depth))
|
||||
s.requests[req.hash] = req
|
||||
}
|
||||
|
||||
// children retrieves all the missing children of a state trie entry for future
|
||||
@ -297,12 +331,14 @@ func (s *Sync) children(req *request, object node) ([]*request, error) {
|
||||
if node, ok := (child.node).(hashNode); ok {
|
||||
// Try to resolve the node from the local database
|
||||
hash := common.BytesToHash(node)
|
||||
if _, ok := s.membatch.batch[hash]; ok {
|
||||
if s.membatch.hasNode(hash) {
|
||||
continue
|
||||
}
|
||||
if s.bloom == nil || s.bloom.Contains(node) {
|
||||
// Bloom filter says this might be a duplicate, double check
|
||||
if ok, _ := s.database.Has(node); ok {
|
||||
// Bloom filter says this might be a duplicate, double check.
|
||||
// If database says yes, then at least the trie node is present
|
||||
// and we hold the assumption that it's NOT legacy contract code.
|
||||
if blob := rawdb.ReadTrieNode(s.database, common.BytesToHash(node)); len(blob) > 0 {
|
||||
continue
|
||||
}
|
||||
// False positive, bump fault meter
|
||||
@ -325,10 +361,13 @@ func (s *Sync) children(req *request, object node) ([]*request, error) {
|
||||
// committed themselves.
|
||||
func (s *Sync) commit(req *request) (err error) {
|
||||
// Write the node content to the membatch
|
||||
s.membatch.batch[req.hash] = req.data
|
||||
|
||||
delete(s.requests, req.hash)
|
||||
|
||||
if req.code {
|
||||
s.membatch.codes[req.hash] = req.data
|
||||
delete(s.codeReqs, req.hash)
|
||||
} else {
|
||||
s.membatch.nodes[req.hash] = req.data
|
||||
delete(s.nodeReqs, req.hash)
|
||||
}
|
||||
// Check all parents for completion
|
||||
for _, parent := range req.parents {
|
||||
parent.deps--
|
||||
|
@ -25,6 +25,7 @@ import (
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/rawdb"
|
||||
"github.com/ethereum/go-ethereum/ethdb"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/metrics"
|
||||
@ -41,8 +42,8 @@ var (
|
||||
)
|
||||
|
||||
// syncBloomHasher is a wrapper around a byte blob to satisfy the interface API
|
||||
// requirements of the bloom library used. It's used to convert a trie hash into
|
||||
// a 64 bit mini hash.
|
||||
// requirements of the bloom library used. It's used to convert a trie hash or
|
||||
// contract code hash into a 64 bit mini hash.
|
||||
type syncBloomHasher []byte
|
||||
|
||||
func (f syncBloomHasher) Write(p []byte) (n int, err error) { panic("not implemented") }
|
||||
@ -53,9 +54,9 @@ func (f syncBloomHasher) Size() int { return 8 }
|
||||
func (f syncBloomHasher) Sum64() uint64 { return binary.BigEndian.Uint64(f) }
|
||||
|
||||
// SyncBloom is a bloom filter used during fast sync to quickly decide if a trie
|
||||
// node already exists on disk or not. It self populates from the provided disk
|
||||
// database on creation in a background thread and will only start returning live
|
||||
// results once that's finished.
|
||||
// node or contract code already exists on disk or not. It self populates from the
|
||||
// provided disk database on creation in a background thread and will only start
|
||||
// returning live results once that's finished.
|
||||
type SyncBloom struct {
|
||||
bloom *bloomfilter.Filter
|
||||
inited uint32
|
||||
@ -107,10 +108,16 @@ func (b *SyncBloom) init(database ethdb.Iteratee) {
|
||||
)
|
||||
for it.Next() && atomic.LoadUint32(&b.closed) == 0 {
|
||||
// If the database entry is a trie node, add it to the bloom
|
||||
if key := it.Key(); len(key) == common.HashLength {
|
||||
key := it.Key()
|
||||
if len(key) == common.HashLength {
|
||||
b.bloom.Add(syncBloomHasher(key))
|
||||
bloomLoadMeter.Mark(1)
|
||||
}
|
||||
// If the database entry is a contract code, add it to the bloom
|
||||
if ok, hash := rawdb.IsCodeKey(key); ok {
|
||||
b.bloom.Add(syncBloomHasher(hash))
|
||||
bloomLoadMeter.Mark(1)
|
||||
}
|
||||
// If enough time elapsed since the last iterator swap, restart
|
||||
if time.Since(swap) > 8*time.Second {
|
||||
key := common.CopyBytes(it.Key())
|
||||
|
@ -124,8 +124,10 @@ func testIterativeSync(t *testing.T, count int) {
|
||||
}
|
||||
results[i] = SyncResult{hash, data}
|
||||
}
|
||||
if _, index, err := sched.Process(results); err != nil {
|
||||
t.Fatalf("failed to process result #%d: %v", index, err)
|
||||
for _, result := range results {
|
||||
if err := sched.Process(result); err != nil {
|
||||
t.Fatalf("failed to process result %v", err)
|
||||
}
|
||||
}
|
||||
batch := diskdb.NewBatch()
|
||||
if err := sched.Commit(batch); err != nil {
|
||||
@ -160,8 +162,10 @@ func TestIterativeDelayedSync(t *testing.T) {
|
||||
}
|
||||
results[i] = SyncResult{hash, data}
|
||||
}
|
||||
if _, index, err := sched.Process(results); err != nil {
|
||||
t.Fatalf("failed to process result #%d: %v", index, err)
|
||||
for _, result := range results {
|
||||
if err := sched.Process(result); err != nil {
|
||||
t.Fatalf("failed to process result %v", err)
|
||||
}
|
||||
}
|
||||
batch := diskdb.NewBatch()
|
||||
if err := sched.Commit(batch); err != nil {
|
||||
@ -204,8 +208,10 @@ func testIterativeRandomSync(t *testing.T, count int) {
|
||||
results = append(results, SyncResult{hash, data})
|
||||
}
|
||||
// Feed the retrieved results back and queue new tasks
|
||||
if _, index, err := sched.Process(results); err != nil {
|
||||
t.Fatalf("failed to process result #%d: %v", index, err)
|
||||
for _, result := range results {
|
||||
if err := sched.Process(result); err != nil {
|
||||
t.Fatalf("failed to process result %v", err)
|
||||
}
|
||||
}
|
||||
batch := diskdb.NewBatch()
|
||||
if err := sched.Commit(batch); err != nil {
|
||||
@ -251,8 +257,10 @@ func TestIterativeRandomDelayedSync(t *testing.T) {
|
||||
}
|
||||
}
|
||||
// Feed the retrieved results back and queue new tasks
|
||||
if _, index, err := sched.Process(results); err != nil {
|
||||
t.Fatalf("failed to process result #%d: %v", index, err)
|
||||
for _, result := range results {
|
||||
if err := sched.Process(result); err != nil {
|
||||
t.Fatalf("failed to process result %v", err)
|
||||
}
|
||||
}
|
||||
batch := diskdb.NewBatch()
|
||||
if err := sched.Commit(batch); err != nil {
|
||||
@ -298,8 +306,10 @@ func TestDuplicateAvoidanceSync(t *testing.T) {
|
||||
|
||||
results[i] = SyncResult{hash, data}
|
||||
}
|
||||
if _, index, err := sched.Process(results); err != nil {
|
||||
t.Fatalf("failed to process result #%d: %v", index, err)
|
||||
for _, result := range results {
|
||||
if err := sched.Process(result); err != nil {
|
||||
t.Fatalf("failed to process result %v", err)
|
||||
}
|
||||
}
|
||||
batch := diskdb.NewBatch()
|
||||
if err := sched.Commit(batch); err != nil {
|
||||
@ -336,8 +346,10 @@ func TestIncompleteSync(t *testing.T) {
|
||||
results[i] = SyncResult{hash, data}
|
||||
}
|
||||
// Process each of the trie nodes
|
||||
if _, index, err := sched.Process(results); err != nil {
|
||||
t.Fatalf("failed to process result #%d: %v", index, err)
|
||||
for _, result := range results {
|
||||
if err := sched.Process(result); err != nil {
|
||||
t.Fatalf("failed to process result %v", err)
|
||||
}
|
||||
}
|
||||
batch := diskdb.NewBatch()
|
||||
if err := sched.Commit(batch); err != nil {
|
||||
|
@ -473,3 +473,9 @@ func (t *Trie) hashRoot(db *Database) (node, node, error) {
|
||||
t.unhashed = 0
|
||||
return hashed, cached, nil
|
||||
}
|
||||
|
||||
// Reset drops the referenced root node and cleans all internal state.
|
||||
func (t *Trie) Reset() {
|
||||
t.root = nil
|
||||
t.unhashed = 0
|
||||
}
|
||||
|
Reference in New Issue
Block a user