bsc/p2p/discover/database.go
Felix Lange c73b654fd1 p2p/discover: move bond logic from table to transport (#17048)
* p2p/discover: move bond logic from table to transport

This commit moves node endpoint verification (bonding) from the table to
the UDP transport implementation. Previously, adding a node to the table
entailed pinging the node if needed. With this change, the ping-back
logic is embedded in the packet handler at a lower level.

It is easy to verify that the basic protocol is unchanged: we still
require a valid pong reply from the node before findnode is accepted.

The node database tracked the time of last ping sent to the node and
time of last valid pong received from the node. Node endpoints are
considered verified when a valid pong is received and the time of last
pong was called 'bond time'. The time of last ping sent was unused. In
this commit, the last ping database entry is repurposed to mean last
ping _received_. This entry is now used to track whether the node needs
to be pinged back.

The other big change is how nodes are added to the table. We used to add
nodes in Table.bond, which ran when a remote node pinged us or when we
encountered the node in a neighbors reply. The transport now adds to the
table directly after the endpoint is verified through ping. To ensure
that the Table can't be filled just by pinging the node repeatedly, we
retain the isInitDone check. During init, only nodes from neighbors
replies are added.

* p2p/discover: reduce findnode failure counter on success

* p2p/discover: remove unused parameter of loadSeedNodes

* p2p/discover: improve ping-back check and comments

* p2p/discover: add neighbors reply nodes always, not just during init
2018-07-03 16:24:12 +03:00

371 lines
11 KiB
Go

// Copyright 2015 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/>.
// Contains the node database, storing previously seen nodes and any collected
// metadata about them for QoS purposes.
package discover
import (
"bytes"
"crypto/rand"
"encoding/binary"
"os"
"sync"
"time"
"github.com/ethereum/go-ethereum/crypto"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/rlp"
"github.com/syndtr/goleveldb/leveldb"
"github.com/syndtr/goleveldb/leveldb/errors"
"github.com/syndtr/goleveldb/leveldb/iterator"
"github.com/syndtr/goleveldb/leveldb/opt"
"github.com/syndtr/goleveldb/leveldb/storage"
"github.com/syndtr/goleveldb/leveldb/util"
)
var (
nodeDBNilNodeID = NodeID{} // Special node ID to use as a nil element.
nodeDBNodeExpiration = 24 * time.Hour // Time after which an unseen node should be dropped.
nodeDBCleanupCycle = time.Hour // Time period for running the expiration task.
nodeDBVersion = 5
)
// nodeDB stores all nodes we know about.
type nodeDB struct {
lvl *leveldb.DB // Interface to the database itself
self NodeID // Own node id to prevent adding it into the database
runner sync.Once // Ensures we can start at most one expirer
quit chan struct{} // Channel to signal the expiring thread to stop
}
// Schema layout for the node database
var (
nodeDBVersionKey = []byte("version") // Version of the database to flush if changes
nodeDBItemPrefix = []byte("n:") // Identifier to prefix node entries with
nodeDBDiscoverRoot = ":discover"
nodeDBDiscoverPing = nodeDBDiscoverRoot + ":lastping"
nodeDBDiscoverPong = nodeDBDiscoverRoot + ":lastpong"
nodeDBDiscoverFindFails = nodeDBDiscoverRoot + ":findfail"
)
// newNodeDB creates a new node database for storing and retrieving infos about
// known peers in the network. If no path is given, an in-memory, temporary
// database is constructed.
func newNodeDB(path string, version int, self NodeID) (*nodeDB, error) {
if path == "" {
return newMemoryNodeDB(self)
}
return newPersistentNodeDB(path, version, self)
}
// newMemoryNodeDB creates a new in-memory node database without a persistent
// backend.
func newMemoryNodeDB(self NodeID) (*nodeDB, error) {
db, err := leveldb.Open(storage.NewMemStorage(), nil)
if err != nil {
return nil, err
}
return &nodeDB{
lvl: db,
self: self,
quit: make(chan struct{}),
}, nil
}
// newPersistentNodeDB creates/opens a leveldb backed persistent node database,
// also flushing its contents in case of a version mismatch.
func newPersistentNodeDB(path string, version int, self NodeID) (*nodeDB, error) {
opts := &opt.Options{OpenFilesCacheCapacity: 5}
db, err := leveldb.OpenFile(path, opts)
if _, iscorrupted := err.(*errors.ErrCorrupted); iscorrupted {
db, err = leveldb.RecoverFile(path, nil)
}
if err != nil {
return nil, err
}
// The nodes contained in the cache correspond to a certain protocol version.
// Flush all nodes if the version doesn't match.
currentVer := make([]byte, binary.MaxVarintLen64)
currentVer = currentVer[:binary.PutVarint(currentVer, int64(version))]
blob, err := db.Get(nodeDBVersionKey, nil)
switch err {
case leveldb.ErrNotFound:
// Version not found (i.e. empty cache), insert it
if err := db.Put(nodeDBVersionKey, currentVer, nil); err != nil {
db.Close()
return nil, err
}
case nil:
// Version present, flush if different
if !bytes.Equal(blob, currentVer) {
db.Close()
if err = os.RemoveAll(path); err != nil {
return nil, err
}
return newPersistentNodeDB(path, version, self)
}
}
return &nodeDB{
lvl: db,
self: self,
quit: make(chan struct{}),
}, nil
}
// makeKey generates the leveldb key-blob from a node id and its particular
// field of interest.
func makeKey(id NodeID, field string) []byte {
if bytes.Equal(id[:], nodeDBNilNodeID[:]) {
return []byte(field)
}
return append(nodeDBItemPrefix, append(id[:], field...)...)
}
// splitKey tries to split a database key into a node id and a field part.
func splitKey(key []byte) (id NodeID, field string) {
// If the key is not of a node, return it plainly
if !bytes.HasPrefix(key, nodeDBItemPrefix) {
return NodeID{}, string(key)
}
// Otherwise split the id and field
item := key[len(nodeDBItemPrefix):]
copy(id[:], item[:len(id)])
field = string(item[len(id):])
return id, field
}
// fetchInt64 retrieves an integer instance associated with a particular
// database key.
func (db *nodeDB) fetchInt64(key []byte) int64 {
blob, err := db.lvl.Get(key, nil)
if err != nil {
return 0
}
val, read := binary.Varint(blob)
if read <= 0 {
return 0
}
return val
}
// storeInt64 update a specific database entry to the current time instance as a
// unix timestamp.
func (db *nodeDB) storeInt64(key []byte, n int64) error {
blob := make([]byte, binary.MaxVarintLen64)
blob = blob[:binary.PutVarint(blob, n)]
return db.lvl.Put(key, blob, nil)
}
// node retrieves a node with a given id from the database.
func (db *nodeDB) node(id NodeID) *Node {
blob, err := db.lvl.Get(makeKey(id, nodeDBDiscoverRoot), nil)
if err != nil {
return nil
}
node := new(Node)
if err := rlp.DecodeBytes(blob, node); err != nil {
log.Error("Failed to decode node RLP", "err", err)
return nil
}
node.sha = crypto.Keccak256Hash(node.ID[:])
return node
}
// updateNode inserts - potentially overwriting - a node into the peer database.
func (db *nodeDB) updateNode(node *Node) error {
blob, err := rlp.EncodeToBytes(node)
if err != nil {
return err
}
return db.lvl.Put(makeKey(node.ID, nodeDBDiscoverRoot), blob, nil)
}
// deleteNode deletes all information/keys associated with a node.
func (db *nodeDB) deleteNode(id NodeID) error {
deleter := db.lvl.NewIterator(util.BytesPrefix(makeKey(id, "")), nil)
for deleter.Next() {
if err := db.lvl.Delete(deleter.Key(), nil); err != nil {
return err
}
}
return nil
}
// ensureExpirer is a small helper method ensuring that the data expiration
// mechanism is running. If the expiration goroutine is already running, this
// method simply returns.
//
// The goal is to start the data evacuation only after the network successfully
// bootstrapped itself (to prevent dumping potentially useful seed nodes). Since
// it would require significant overhead to exactly trace the first successful
// convergence, it's simpler to "ensure" the correct state when an appropriate
// condition occurs (i.e. a successful bonding), and discard further events.
func (db *nodeDB) ensureExpirer() {
db.runner.Do(func() { go db.expirer() })
}
// expirer should be started in a go routine, and is responsible for looping ad
// infinitum and dropping stale data from the database.
func (db *nodeDB) expirer() {
tick := time.NewTicker(nodeDBCleanupCycle)
defer tick.Stop()
for {
select {
case <-tick.C:
if err := db.expireNodes(); err != nil {
log.Error("Failed to expire nodedb items", "err", err)
}
case <-db.quit:
return
}
}
}
// expireNodes iterates over the database and deletes all nodes that have not
// been seen (i.e. received a pong from) for some allotted time.
func (db *nodeDB) expireNodes() error {
threshold := time.Now().Add(-nodeDBNodeExpiration)
// Find discovered nodes that are older than the allowance
it := db.lvl.NewIterator(nil, nil)
defer it.Release()
for it.Next() {
// Skip the item if not a discovery node
id, field := splitKey(it.Key())
if field != nodeDBDiscoverRoot {
continue
}
// Skip the node if not expired yet (and not self)
if !bytes.Equal(id[:], db.self[:]) {
if seen := db.lastPongReceived(id); seen.After(threshold) {
continue
}
}
// Otherwise delete all associated information
db.deleteNode(id)
}
return nil
}
// lastPingReceived retrieves the time of the last ping packet sent by the remote node.
func (db *nodeDB) lastPingReceived(id NodeID) time.Time {
return time.Unix(db.fetchInt64(makeKey(id, nodeDBDiscoverPing)), 0)
}
// updateLastPing updates the last time remote node pinged us.
func (db *nodeDB) updateLastPingReceived(id NodeID, instance time.Time) error {
return db.storeInt64(makeKey(id, nodeDBDiscoverPing), instance.Unix())
}
// lastPongReceived retrieves the time of the last successful pong from remote node.
func (db *nodeDB) lastPongReceived(id NodeID) time.Time {
return time.Unix(db.fetchInt64(makeKey(id, nodeDBDiscoverPong)), 0)
}
// hasBond reports whether the given node is considered bonded.
func (db *nodeDB) hasBond(id NodeID) bool {
return time.Since(db.lastPongReceived(id)) < nodeDBNodeExpiration
}
// updateLastPongReceived updates the last pong time of a node.
func (db *nodeDB) updateLastPongReceived(id NodeID, instance time.Time) error {
return db.storeInt64(makeKey(id, nodeDBDiscoverPong), instance.Unix())
}
// findFails retrieves the number of findnode failures since bonding.
func (db *nodeDB) findFails(id NodeID) int {
return int(db.fetchInt64(makeKey(id, nodeDBDiscoverFindFails)))
}
// updateFindFails updates the number of findnode failures since bonding.
func (db *nodeDB) updateFindFails(id NodeID, fails int) error {
return db.storeInt64(makeKey(id, nodeDBDiscoverFindFails), int64(fails))
}
// querySeeds retrieves random nodes to be used as potential seed nodes
// for bootstrapping.
func (db *nodeDB) querySeeds(n int, maxAge time.Duration) []*Node {
var (
now = time.Now()
nodes = make([]*Node, 0, n)
it = db.lvl.NewIterator(nil, nil)
id NodeID
)
defer it.Release()
seek:
for seeks := 0; len(nodes) < n && seeks < n*5; seeks++ {
// Seek to a random entry. The first byte is incremented by a
// random amount each time in order to increase the likelihood
// of hitting all existing nodes in very small databases.
ctr := id[0]
rand.Read(id[:])
id[0] = ctr + id[0]%16
it.Seek(makeKey(id, nodeDBDiscoverRoot))
n := nextNode(it)
if n == nil {
id[0] = 0
continue seek // iterator exhausted
}
if n.ID == db.self {
continue seek
}
if now.Sub(db.lastPongReceived(n.ID)) > maxAge {
continue seek
}
for i := range nodes {
if nodes[i].ID == n.ID {
continue seek // duplicate
}
}
nodes = append(nodes, n)
}
return nodes
}
// reads the next node record from the iterator, skipping over other
// database entries.
func nextNode(it iterator.Iterator) *Node {
for end := false; !end; end = !it.Next() {
id, field := splitKey(it.Key())
if field != nodeDBDiscoverRoot {
continue
}
var n Node
if err := rlp.DecodeBytes(it.Value(), &n); err != nil {
log.Warn("Failed to decode node RLP", "id", id, "err", err)
continue
}
return &n
}
return nil
}
// close flushes and closes the database files.
func (db *nodeDB) close() {
close(db.quit)
db.lvl.Close()
}