Official Go implementation of the Ethereum protocol
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
go-ethereum/les/peer.go

1227 lines
39 KiB

// Copyright 2016 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 les
import (
"errors"
"fmt"
"math/big"
"math/rand"
"net"
"sync"
"sync/atomic"
"time"
"github.com/ethereum/go-ethereum/common"
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
"github.com/ethereum/go-ethereum/common/mclock"
"github.com/ethereum/go-ethereum/core"
"github.com/ethereum/go-ethereum/core/forkid"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/les/flowcontrol"
"github.com/ethereum/go-ethereum/les/utils"
vfc "github.com/ethereum/go-ethereum/les/vflux/client"
vfs "github.com/ethereum/go-ethereum/les/vflux/server"
"github.com/ethereum/go-ethereum/light"
"github.com/ethereum/go-ethereum/p2p"
"github.com/ethereum/go-ethereum/params"
"github.com/ethereum/go-ethereum/rlp"
)
var (
errClosed = errors.New("peer set is closed")
errAlreadyRegistered = errors.New("peer is already registered")
errNotRegistered = errors.New("peer is not registered")
)
const (
maxRequestErrors = 20 // number of invalid requests tolerated (makes the protocol less brittle but still avoids spam)
maxResponseErrors = 50 // number of invalid responses tolerated (makes the protocol less brittle but still avoids spam)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
allowedUpdateBytes = 100000 // initial/maximum allowed update size
allowedUpdateRate = time.Millisecond * 10 // time constant for recharging one byte of allowance
freezeTimeBase = time.Millisecond * 700 // fixed component of client freeze time
freezeTimeRandom = time.Millisecond * 600 // random component of client freeze time
freezeCheckPeriod = time.Millisecond * 100 // buffer value recheck period after initial freeze time has elapsed
// If the total encoded size of a sent transaction batch is over txSizeCostLimit
// per transaction then the request cost is calculated as proportional to the
// encoded size instead of the transaction count
txSizeCostLimit = 0x4000
// handshakeTimeout is the timeout LES handshake will be treated as failed.
handshakeTimeout = 5 * time.Second
)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
const (
announceTypeNone = iota
announceTypeSimple
announceTypeSigned
)
type keyValueEntry struct {
Key string
Value rlp.RawValue
}
type keyValueList []keyValueEntry
type keyValueMap map[string]rlp.RawValue
func (l keyValueList) add(key string, val interface{}) keyValueList {
var entry keyValueEntry
entry.Key = key
if val == nil {
val = uint64(0)
}
enc, err := rlp.EncodeToBytes(val)
if err == nil {
entry.Value = enc
}
return append(l, entry)
}
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
func (l keyValueList) decode() (keyValueMap, uint64) {
m := make(keyValueMap)
var size uint64
for _, entry := range l {
m[entry.Key] = entry.Value
size += uint64(len(entry.Key)) + uint64(len(entry.Value)) + 8
}
return m, size
}
func (m keyValueMap) get(key string, val interface{}) error {
enc, ok := m[key]
if !ok {
return errResp(ErrMissingKey, "%s", key)
}
if val == nil {
return nil
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
}
return rlp.DecodeBytes(enc, val)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
}
// peerCommons contains fields needed by both server peer and client peer.
type peerCommons struct {
*p2p.Peer
rw p2p.MsgReadWriter
id string // Peer identity.
version int // Protocol version negotiated.
network uint64 // Network ID being on.
frozen uint32 // Flag whether the peer is frozen.
announceType uint64 // New block announcement type.
serving uint32 // The status indicates the peer is served.
headInfo blockInfo // Last announced block information.
// Background task queue for caching peer tasks and executing in order.
sendQueue *utils.ExecQueue
// Flow control agreement.
fcParams flowcontrol.ServerParams // The config for token bucket.
fcCosts requestCostTable // The Maximum request cost table.
closeCh chan struct{}
lock sync.RWMutex // Lock used to protect all thread-sensitive fields.
}
// isFrozen returns true if the client is frozen or the server has put our
// client in frozen state
func (p *peerCommons) isFrozen() bool {
return atomic.LoadUint32(&p.frozen) != 0
}
// canQueue returns an indicator whether the peer can queue an operation.
func (p *peerCommons) canQueue() bool {
return p.sendQueue.CanQueue() && !p.isFrozen()
}
// queueSend caches a peer operation in the background task queue.
// Please ensure to check `canQueue` before call this function
func (p *peerCommons) queueSend(f func()) bool {
return p.sendQueue.Queue(f)
}
// String implements fmt.Stringer.
func (p *peerCommons) String() string {
return fmt.Sprintf("Peer %s [%s]", p.id, fmt.Sprintf("les/%d", p.version))
}
// PeerInfo represents a short summary of the `eth` sub-protocol metadata known
// about a connected peer.
type PeerInfo struct {
Version int `json:"version"` // Ethereum protocol version negotiated
Difficulty *big.Int `json:"difficulty"` // Total difficulty of the peer's blockchain
Head string `json:"head"` // SHA3 hash of the peer's best owned block
}
// Info gathers and returns a collection of metadata known about a peer.
func (p *peerCommons) Info() *PeerInfo {
return &PeerInfo{
Version: p.version,
Difficulty: p.Td(),
Head: fmt.Sprintf("%x", p.Head()),
}
}
// Head retrieves a copy of the current head (most recent) hash of the peer.
func (p *peerCommons) Head() (hash common.Hash) {
p.lock.RLock()
defer p.lock.RUnlock()
return p.headInfo.Hash
}
// Td retrieves the current total difficulty of a peer.
func (p *peerCommons) Td() *big.Int {
p.lock.RLock()
defer p.lock.RUnlock()
return new(big.Int).Set(p.headInfo.Td)
}
// HeadAndTd retrieves the current head hash and total difficulty of a peer.
func (p *peerCommons) HeadAndTd() (hash common.Hash, td *big.Int) {
p.lock.RLock()
defer p.lock.RUnlock()
return p.headInfo.Hash, new(big.Int).Set(p.headInfo.Td)
}
// sendReceiveHandshake exchanges handshake packet with remote peer and returns any error
// if failed to send or receive packet.
func (p *peerCommons) sendReceiveHandshake(sendList keyValueList) (keyValueList, error) {
var (
errc = make(chan error, 2)
recvList keyValueList
)
// Send out own handshake in a new thread
go func() {
errc <- p2p.Send(p.rw, StatusMsg, sendList)
}()
go func() {
// In the mean time retrieve the remote status message
msg, err := p.rw.ReadMsg()
if err != nil {
errc <- err
return
}
if msg.Code != StatusMsg {
errc <- errResp(ErrNoStatusMsg, "first msg has code %x (!= %x)", msg.Code, StatusMsg)
return
}
if msg.Size > ProtocolMaxMsgSize {
errc <- errResp(ErrMsgTooLarge, "%v > %v", msg.Size, ProtocolMaxMsgSize)
return
}
// Decode the handshake
if err := msg.Decode(&recvList); err != nil {
errc <- errResp(ErrDecode, "msg %v: %v", msg, err)
return
}
errc <- nil
}()
timeout := time.NewTimer(handshakeTimeout)
defer timeout.Stop()
for i := 0; i < 2; i++ {
select {
case err := <-errc:
if err != nil {
return nil, err
}
case <-timeout.C:
return nil, p2p.DiscReadTimeout
}
}
return recvList, nil
}
// handshake executes the les protocol handshake, negotiating version number,
// network IDs, difficulties, head and genesis blocks. Besides the basic handshake
// fields, server and client can exchange and resolve some specified fields through
// two callback functions.
func (p *peerCommons) handshake(td *big.Int, head common.Hash, headNum uint64, genesis common.Hash, forkID forkid.ID, forkFilter forkid.Filter, sendCallback func(*keyValueList), recvCallback func(keyValueMap) error) error {
p.lock.Lock()
defer p.lock.Unlock()
var send keyValueList
// Add some basic handshake fields
send = send.add("protocolVersion", uint64(p.version))
send = send.add("networkId", p.network)
// Note: the head info announced at handshake is only used in case of server peers
// but dummy values are still announced by clients for compatibility with older servers
send = send.add("headTd", td)
send = send.add("headHash", head)
send = send.add("headNum", headNum)
send = send.add("genesisHash", genesis)
// If the protocol version is beyond les4, then pass the forkID
// as well. Check http://eips.ethereum.org/EIPS/eip-2124 for more
// spec detail.
if p.version >= lpv4 {
send = send.add("forkID", forkID)
}
// Add client-specified or server-specified fields
if sendCallback != nil {
sendCallback(&send)
}
// Exchange the handshake packet and resolve the received one.
recvList, err := p.sendReceiveHandshake(send)
if err != nil {
return err
}
recv, size := recvList.decode()
if size > allowedUpdateBytes {
return errResp(ErrRequestRejected, "")
}
var rGenesis common.Hash
var rVersion, rNetwork uint64
if err := recv.get("protocolVersion", &rVersion); err != nil {
return err
}
if err := recv.get("networkId", &rNetwork); err != nil {
return err
}
if err := recv.get("genesisHash", &rGenesis); err != nil {
return err
}
if rGenesis != genesis {
return errResp(ErrGenesisBlockMismatch, "%x (!= %x)", rGenesis[:8], genesis[:8])
}
if rNetwork != p.network {
return errResp(ErrNetworkIdMismatch, "%d (!= %d)", rNetwork, p.network)
}
if int(rVersion) != p.version {
return errResp(ErrProtocolVersionMismatch, "%d (!= %d)", rVersion, p.version)
}
// Check forkID if the protocol version is beyond the les4
if p.version >= lpv4 {
var forkID forkid.ID
if err := recv.get("forkID", &forkID); err != nil {
return err
}
if err := forkFilter(forkID); err != nil {
return errResp(ErrForkIDRejected, "%v", err)
}
}
if recvCallback != nil {
return recvCallback(recv)
}
return nil
}
// close closes the channel and notifies all background routines to exit.
func (p *peerCommons) close() {
close(p.closeCh)
p.sendQueue.Quit()
}
// serverPeer represents each node to which the client is connected.
// The node here refers to the les server.
type serverPeer struct {
peerCommons
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
// Status fields
trusted bool // The flag whether the server is selected as trusted server.
onlyAnnounce bool // The flag whether the server sends announcement only.
chainSince, chainRecent uint64 // The range of chain server peer can serve.
stateSince, stateRecent uint64 // The range of state server peer can serve.
serveTxLookup bool // The server peer can serve tx lookups.
// Advertised checkpoint fields
checkpointNumber uint64 // The block height which the checkpoint is registered.
checkpoint params.TrustedCheckpoint // The advertised checkpoint sent by server.
fcServer *flowcontrol.ServerNode // Client side mirror token bucket.
vtLock sync.Mutex
valueTracker *vfc.ValueTracker
nodeValueTracker *vfc.NodeValueTracker
sentReqs map[uint64]sentReqEntry
// Statistics
errCount utils.LinearExpiredValue // Counter the invalid responses server has replied
updateCount uint64
updateTime mclock.AbsTime
// Test callback hooks
hasBlockHook func(common.Hash, uint64, bool) bool // Used to determine whether the server has the specified block.
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
}
func newServerPeer(version int, network uint64, trusted bool, p *p2p.Peer, rw p2p.MsgReadWriter) *serverPeer {
return &serverPeer{
peerCommons: peerCommons{
Peer: p,
rw: rw,
id: p.ID().String(),
version: version,
network: network,
sendQueue: utils.NewExecQueue(100),
closeCh: make(chan struct{}),
},
trusted: trusted,
errCount: utils.LinearExpiredValue{Rate: mclock.AbsTime(time.Hour)},
}
}
// rejectUpdate returns true if a parameter update has to be rejected because
// the size and/or rate of updates exceed the capacity limitation
func (p *serverPeer) rejectUpdate(size uint64) bool {
now := mclock.Now()
if p.updateCount == 0 {
p.updateTime = now
} else {
dt := now - p.updateTime
p.updateTime = now
r := uint64(dt / mclock.AbsTime(allowedUpdateRate))
if p.updateCount > r {
p.updateCount -= r
} else {
p.updateCount = 0
}
}
p.updateCount += size
return p.updateCount > allowedUpdateBytes
}
// freeze processes Stop messages from the given server and set the status as
// frozen.
func (p *serverPeer) freeze() {
if atomic.CompareAndSwapUint32(&p.frozen, 0, 1) {
p.sendQueue.Clear()
}
}
// unfreeze processes Resume messages from the given server and set the status
// as unfrozen.
func (p *serverPeer) unfreeze() {
atomic.StoreUint32(&p.frozen, 0)
}
// sendRequest send a request to the server based on the given message type
// and content.
func sendRequest(w p2p.MsgWriter, msgcode, reqID uint64, data interface{}) error {
type req struct {
ReqID uint64
Data interface{}
}
return p2p.Send(w, msgcode, req{reqID, data})
}
func (p *serverPeer) sendRequest(msgcode, reqID uint64, data interface{}, amount int) error {
p.sentRequest(reqID, uint32(msgcode), uint32(amount))
return sendRequest(p.rw, msgcode, reqID, data)
}
// requestHeadersByHash fetches a batch of blocks' headers corresponding to the
// specified header query, based on the hash of an origin block.
func (p *serverPeer) requestHeadersByHash(reqID uint64, origin common.Hash, amount int, skip int, reverse bool) error {
p.Log().Debug("Fetching batch of headers", "count", amount, "fromhash", origin, "skip", skip, "reverse", reverse)
return p.sendRequest(GetBlockHeadersMsg, reqID, &GetBlockHeadersData{Origin: hashOrNumber{Hash: origin}, Amount: uint64(amount), Skip: uint64(skip), Reverse: reverse}, amount)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
}
// requestHeadersByNumber fetches a batch of blocks' headers corresponding to the
// specified header query, based on the number of an origin block.
func (p *serverPeer) requestHeadersByNumber(reqID, origin uint64, amount int, skip int, reverse bool) error {
p.Log().Debug("Fetching batch of headers", "count", amount, "fromnum", origin, "skip", skip, "reverse", reverse)
return p.sendRequest(GetBlockHeadersMsg, reqID, &GetBlockHeadersData{Origin: hashOrNumber{Number: origin}, Amount: uint64(amount), Skip: uint64(skip), Reverse: reverse}, amount)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
}
// requestBodies fetches a batch of blocks' bodies corresponding to the hashes
// specified.
func (p *serverPeer) requestBodies(reqID uint64, hashes []common.Hash) error {
p.Log().Debug("Fetching batch of block bodies", "count", len(hashes))
return p.sendRequest(GetBlockBodiesMsg, reqID, hashes, len(hashes))
}
// requestCode fetches a batch of arbitrary data from a node's known state
// data, corresponding to the specified hashes.
func (p *serverPeer) requestCode(reqID uint64, reqs []CodeReq) error {
p.Log().Debug("Fetching batch of codes", "count", len(reqs))
return p.sendRequest(GetCodeMsg, reqID, reqs, len(reqs))
}
// requestReceipts fetches a batch of transaction receipts from a remote node.
func (p *serverPeer) requestReceipts(reqID uint64, hashes []common.Hash) error {
p.Log().Debug("Fetching batch of receipts", "count", len(hashes))
return p.sendRequest(GetReceiptsMsg, reqID, hashes, len(hashes))
}
// requestProofs fetches a batch of merkle proofs from a remote node.
func (p *serverPeer) requestProofs(reqID uint64, reqs []ProofReq) error {
p.Log().Debug("Fetching batch of proofs", "count", len(reqs))
return p.sendRequest(GetProofsV2Msg, reqID, reqs, len(reqs))
}
// requestHelperTrieProofs fetches a batch of HelperTrie merkle proofs from a remote node.
func (p *serverPeer) requestHelperTrieProofs(reqID uint64, reqs []HelperTrieReq) error {
p.Log().Debug("Fetching batch of HelperTrie proofs", "count", len(reqs))
return p.sendRequest(GetHelperTrieProofsMsg, reqID, reqs, len(reqs))
}
// requestTxStatus fetches a batch of transaction status records from a remote node.
func (p *serverPeer) requestTxStatus(reqID uint64, txHashes []common.Hash) error {
p.Log().Debug("Requesting transaction status", "count", len(txHashes))
return p.sendRequest(GetTxStatusMsg, reqID, txHashes, len(txHashes))
}
// sendTxs creates a reply with a batch of transactions to be added to the remote transaction pool.
func (p *serverPeer) sendTxs(reqID uint64, amount int, txs rlp.RawValue) error {
p.Log().Debug("Sending batch of transactions", "amount", amount, "size", len(txs))
sizeFactor := (len(txs) + txSizeCostLimit/2) / txSizeCostLimit
if sizeFactor > amount {
amount = sizeFactor
}
return p.sendRequest(SendTxV2Msg, reqID, txs, amount)
}
// waitBefore implements distPeer interface
func (p *serverPeer) waitBefore(maxCost uint64) (time.Duration, float64) {
return p.fcServer.CanSend(maxCost)
}
// getRequestCost returns an estimated request cost according to the flow control
// rules negotiated between the server and the client.
func (p *serverPeer) getRequestCost(msgcode uint64, amount int) uint64 {
p.lock.RLock()
defer p.lock.RUnlock()
costs := p.fcCosts[msgcode]
if costs == nil {
return 0
}
cost := costs.baseCost + costs.reqCost*uint64(amount)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
if cost > p.fcParams.BufLimit {
cost = p.fcParams.BufLimit
}
return cost
}
// getTxRelayCost returns an estimated relay cost according to the flow control
// rules negotiated between the server and the client.
func (p *serverPeer) getTxRelayCost(amount, size int) uint64 {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
p.lock.RLock()
defer p.lock.RUnlock()
costs := p.fcCosts[SendTxV2Msg]
if costs == nil {
return 0
}
cost := costs.baseCost + costs.reqCost*uint64(amount)
sizeCost := costs.baseCost + costs.reqCost*uint64(size)/txSizeCostLimit
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
if sizeCost > cost {
cost = sizeCost
}
if cost > p.fcParams.BufLimit {
cost = p.fcParams.BufLimit
}
return cost
}
// HasBlock checks if the peer has a given block
func (p *serverPeer) HasBlock(hash common.Hash, number uint64, hasState bool) bool {
p.lock.RLock()
defer p.lock.RUnlock()
if p.hasBlockHook != nil {
return p.hasBlockHook(hash, number, hasState)
}
head := p.headInfo.Number
var since, recent uint64
if hasState {
since = p.stateSince
recent = p.stateRecent
} else {
since = p.chainSince
recent = p.chainRecent
}
return head >= number && number >= since && (recent == 0 || number+recent+4 > head)
}
// updateFlowControl updates the flow control parameters belonging to the server
// node if the announced key/value set contains relevant fields
func (p *serverPeer) updateFlowControl(update keyValueMap) {
p.lock.Lock()
defer p.lock.Unlock()
// If any of the flow control params is nil, refuse to update.
var params flowcontrol.ServerParams
if update.get("flowControl/BL", &params.BufLimit) == nil && update.get("flowControl/MRR", &params.MinRecharge) == nil {
// todo can light client set a minimal acceptable flow control params?
p.fcParams = params
p.fcServer.UpdateParams(params)
}
var MRC RequestCostList
if update.get("flowControl/MRC", &MRC) == nil {
costUpdate := MRC.decode(ProtocolLengths[uint(p.version)])
for code, cost := range costUpdate {
p.fcCosts[code] = cost
}
}
}
// updateHead updates the head information based on the announcement from
// the peer.
func (p *serverPeer) updateHead(hash common.Hash, number uint64, td *big.Int) {
p.lock.Lock()
defer p.lock.Unlock()
p.headInfo = blockInfo{Hash: hash, Number: number, Td: td}
}
// Handshake executes the les protocol handshake, negotiating version number,
// network IDs and genesis blocks.
func (p *serverPeer) Handshake(genesis common.Hash, forkid forkid.ID, forkFilter forkid.Filter) error {
// Note: there is no need to share local head with a server but older servers still
// require these fields so we announce zero values.
return p.handshake(common.Big0, common.Hash{}, 0, genesis, forkid, forkFilter, func(lists *keyValueList) {
// Add some client-specific handshake fields
//
// Enable signed announcement randomly even the server is not trusted.
p.announceType = announceTypeSimple
if p.trusted {
p.announceType = announceTypeSigned
}
*lists = (*lists).add("announceType", p.announceType)
}, func(recv keyValueMap) error {
var (
rHash common.Hash
rNum uint64
rTd *big.Int
)
if err := recv.get("headTd", &rTd); err != nil {
return err
}
if err := recv.get("headHash", &rHash); err != nil {
return err
}
if err := recv.get("headNum", &rNum); err != nil {
return err
}
p.headInfo = blockInfo{Hash: rHash, Number: rNum, Td: rTd}
if recv.get("serveChainSince", &p.chainSince) != nil {
p.onlyAnnounce = true
}
if recv.get("serveRecentChain", &p.chainRecent) != nil {
p.chainRecent = 0
}
if recv.get("serveStateSince", &p.stateSince) != nil {
p.onlyAnnounce = true
}
if recv.get("serveRecentState", &p.stateRecent) != nil {
p.stateRecent = 0
}
if recv.get("txRelay", nil) != nil {
p.onlyAnnounce = true
}
if p.version >= lpv4 {
var recentTx uint
if err := recv.get("recentTxLookup", &recentTx); err != nil {
return err
}
// Note: in the current version we only consider the tx index service useful
// if it is unlimited. This can be made configurable in the future.
p.serveTxLookup = recentTx == txIndexUnlimited
} else {
p.serveTxLookup = true
}
if p.onlyAnnounce && !p.trusted {
return errResp(ErrUselessPeer, "peer cannot serve requests")
}
// Parse flow control handshake packet.
var sParams flowcontrol.ServerParams
if err := recv.get("flowControl/BL", &sParams.BufLimit); err != nil {
return err
}
if err := recv.get("flowControl/MRR", &sParams.MinRecharge); err != nil {
return err
}
var MRC RequestCostList
if err := recv.get("flowControl/MRC", &MRC); err != nil {
return err
}
p.fcParams = sParams
p.fcServer = flowcontrol.NewServerNode(sParams, &mclock.System{})
p.fcCosts = MRC.decode(ProtocolLengths[uint(p.version)])
recv.get("checkpoint/value", &p.checkpoint)
recv.get("checkpoint/registerHeight", &p.checkpointNumber)
if !p.onlyAnnounce {
for msgCode := range reqAvgTimeCost {
if p.fcCosts[msgCode] == nil {
return errResp(ErrUselessPeer, "peer does not support message %d", msgCode)
}
}
}
return nil
})
}
// setValueTracker sets the value tracker references for connected servers. Note that the
// references should be removed upon disconnection by setValueTracker(nil, nil).
func (p *serverPeer) setValueTracker(vt *vfc.ValueTracker, nvt *vfc.NodeValueTracker) {
p.vtLock.Lock()
p.valueTracker = vt
p.nodeValueTracker = nvt
if nvt != nil {
p.sentReqs = make(map[uint64]sentReqEntry)
} else {
p.sentReqs = nil
}
p.vtLock.Unlock()
}
// updateVtParams updates the server's price table in the value tracker.
func (p *serverPeer) updateVtParams() {
p.vtLock.Lock()
defer p.vtLock.Unlock()
if p.nodeValueTracker == nil {
return
}
reqCosts := make([]uint64, len(requestList))
for code, costs := range p.fcCosts {
if m, ok := requestMapping[uint32(code)]; ok {
reqCosts[m.first] = costs.baseCost + costs.reqCost
if m.rest != -1 {
reqCosts[m.rest] = costs.reqCost
}
}
}
p.valueTracker.UpdateCosts(p.nodeValueTracker, reqCosts)
}
// sentReqEntry remembers sent requests and their sending times
type sentReqEntry struct {
reqType, amount uint32
at mclock.AbsTime
}
// sentRequest marks a request sent at the current moment to this server.
func (p *serverPeer) sentRequest(id uint64, reqType, amount uint32) {
p.vtLock.Lock()
if p.sentReqs != nil {
p.sentReqs[id] = sentReqEntry{reqType, amount, mclock.Now()}
}
p.vtLock.Unlock()
}
// answeredRequest marks a request answered at the current moment by this server.
func (p *serverPeer) answeredRequest(id uint64) {
p.vtLock.Lock()
if p.sentReqs == nil {
p.vtLock.Unlock()
return
}
e, ok := p.sentReqs[id]
delete(p.sentReqs, id)
vt := p.valueTracker
nvt := p.nodeValueTracker
p.vtLock.Unlock()
if !ok {
return
}
var (
vtReqs [2]vfc.ServedRequest
reqCount int
)
m := requestMapping[e.reqType]
if m.rest == -1 || e.amount <= 1 {
reqCount = 1
vtReqs[0] = vfc.ServedRequest{ReqType: uint32(m.first), Amount: e.amount}
} else {
reqCount = 2
vtReqs[0] = vfc.ServedRequest{ReqType: uint32(m.first), Amount: 1}
vtReqs[1] = vfc.ServedRequest{ReqType: uint32(m.rest), Amount: e.amount - 1}
}
dt := time.Duration(mclock.Now() - e.at)
vt.Served(nvt, vtReqs[:reqCount], dt)
}
// clientPeer represents each node to which the les server is connected.
// The node here refers to the light client.
type clientPeer struct {
peerCommons
// responseLock ensures that responses are queued in the same order as
// RequestProcessed is called
responseLock sync.Mutex
responseCount uint64 // Counter to generate an unique id for request processing.
balance *vfs.NodeBalance
// invalidLock is used for protecting invalidCount.
invalidLock sync.RWMutex
invalidCount utils.LinearExpiredValue // Counter the invalid request the client peer has made.
server bool
errCh chan error
fcClient *flowcontrol.ClientNode // Server side mirror token bucket.
}
func newClientPeer(version int, network uint64, p *p2p.Peer, rw p2p.MsgReadWriter) *clientPeer {
return &clientPeer{
peerCommons: peerCommons{
Peer: p,
rw: rw,
id: p.ID().String(),
version: version,
network: network,
sendQueue: utils.NewExecQueue(100),
closeCh: make(chan struct{}),
},
invalidCount: utils.LinearExpiredValue{Rate: mclock.AbsTime(time.Hour)},
errCh: make(chan error, 1),
}
}
// freeClientId returns a string identifier for the peer. Multiple peers with
// the same identifier can not be connected in free mode simultaneously.
func (p *clientPeer) freeClientId() string {
if addr, ok := p.RemoteAddr().(*net.TCPAddr); ok {
if addr.IP.IsLoopback() {
// using peer id instead of loopback ip address allows multiple free
// connections from local machine to own server
return p.id
} else {
return addr.IP.String()
}
}
return p.id
}
// sendStop notifies the client about being in frozen state
func (p *clientPeer) sendStop() error {
return p2p.Send(p.rw, StopMsg, struct{}{})
}
// sendResume notifies the client about getting out of frozen state
func (p *clientPeer) sendResume(bv uint64) error {
return p2p.Send(p.rw, ResumeMsg, bv)
}
// freeze temporarily puts the client in a frozen state which means all unprocessed
// and subsequent requests are dropped. Unfreezing happens automatically after a short
// time if the client's buffer value is at least in the slightly positive region.
// The client is also notified about being frozen/unfrozen with a Stop/Resume message.
func (p *clientPeer) freeze() {
if p.version < lpv3 {
// if Stop/Resume is not supported then just drop the peer after setting
// its frozen status permanently
atomic.StoreUint32(&p.frozen, 1)
p.Peer.Disconnect(p2p.DiscUselessPeer)
return
}
if atomic.SwapUint32(&p.frozen, 1) == 0 {
go func() {
p.sendStop()
time.Sleep(freezeTimeBase + time.Duration(rand.Int63n(int64(freezeTimeRandom))))
for {
bufValue, bufLimit := p.fcClient.BufferStatus()
if bufLimit == 0 {
return
}
if bufValue <= bufLimit/8 {
time.Sleep(freezeCheckPeriod)
continue
}
atomic.StoreUint32(&p.frozen, 0)
p.sendResume(bufValue)
return
}
}()
}
}
// reply struct represents a reply with the actual data already RLP encoded and
// only the bv (buffer value) missing. This allows the serving mechanism to
// calculate the bv value which depends on the data size before sending the reply.
type reply struct {
w p2p.MsgWriter
msgcode, reqID uint64
data rlp.RawValue
}
// send sends the reply with the calculated buffer value
func (r *reply) send(bv uint64) error {
type resp struct {
ReqID, BV uint64
Data rlp.RawValue
}
return p2p.Send(r.w, r.msgcode, resp{r.reqID, bv, r.data})
}
// size returns the RLP encoded size of the message data
func (r *reply) size() uint32 {
return uint32(len(r.data))
}
// replyBlockHeaders creates a reply with a batch of block headers
func (p *clientPeer) replyBlockHeaders(reqID uint64, headers []*types.Header) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(headers)
return &reply{p.rw, BlockHeadersMsg, reqID, data}
}
// replyBlockBodiesRLP creates a reply with a batch of block contents from
// an already RLP encoded format.
func (p *clientPeer) replyBlockBodiesRLP(reqID uint64, bodies []rlp.RawValue) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(bodies)
return &reply{p.rw, BlockBodiesMsg, reqID, data}
}
// replyCode creates a reply with a batch of arbitrary internal data, corresponding to the
// hashes requested.
func (p *clientPeer) replyCode(reqID uint64, codes [][]byte) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(codes)
return &reply{p.rw, CodeMsg, reqID, data}
}
// replyReceiptsRLP creates a reply with a batch of transaction receipts, corresponding to the
// ones requested from an already RLP encoded format.
func (p *clientPeer) replyReceiptsRLP(reqID uint64, receipts []rlp.RawValue) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(receipts)
return &reply{p.rw, ReceiptsMsg, reqID, data}
}
// replyProofsV2 creates a reply with a batch of merkle proofs, corresponding to the ones requested.
func (p *clientPeer) replyProofsV2(reqID uint64, proofs light.NodeList) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(proofs)
return &reply{p.rw, ProofsV2Msg, reqID, data}
}
// replyHelperTrieProofs creates a reply with a batch of HelperTrie proofs, corresponding to the ones requested.
func (p *clientPeer) replyHelperTrieProofs(reqID uint64, resp HelperTrieResps) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(resp)
return &reply{p.rw, HelperTrieProofsMsg, reqID, data}
}
// replyTxStatus creates a reply with a batch of transaction status records, corresponding to the ones requested.
func (p *clientPeer) replyTxStatus(reqID uint64, stats []light.TxStatus) *reply {
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
data, _ := rlp.EncodeToBytes(stats)
return &reply{p.rw, TxStatusMsg, reqID, data}
}
// sendAnnounce announces the availability of a number of blocks through
// a hash notification.
func (p *clientPeer) sendAnnounce(request announceData) error {
return p2p.Send(p.rw, AnnounceMsg, request)
}
// allowInactive implements clientPoolPeer
func (p *clientPeer) allowInactive() bool {
return false
}
// updateCapacity updates the request serving capacity assigned to a given client
// and also sends an announcement about the updated flow control parameters
func (p *clientPeer) updateCapacity(cap uint64) {
p.lock.Lock()
defer p.lock.Unlock()
if cap != p.fcParams.MinRecharge {
p.fcParams = flowcontrol.ServerParams{MinRecharge: cap, BufLimit: cap * bufLimitRatio}
p.fcClient.UpdateParams(p.fcParams)
var kvList keyValueList
kvList = kvList.add("flowControl/MRR", cap)
kvList = kvList.add("flowControl/BL", cap*bufLimitRatio)
p.queueSend(func() { p.sendAnnounce(announceData{Update: kvList}) })
}
}
// freezeClient temporarily puts the client in a frozen state which means all
// unprocessed and subsequent requests are dropped. Unfreezing happens automatically
// after a short time if the client's buffer value is at least in the slightly positive
// region. The client is also notified about being frozen/unfrozen with a Stop/Resume
// message.
func (p *clientPeer) freezeClient() {
if p.version < lpv3 {
// if Stop/Resume is not supported then just drop the peer after setting
// its frozen status permanently
atomic.StoreUint32(&p.frozen, 1)
p.Peer.Disconnect(p2p.DiscUselessPeer)
return
}
if atomic.SwapUint32(&p.frozen, 1) == 0 {
go func() {
p.sendStop()
time.Sleep(freezeTimeBase + time.Duration(rand.Int63n(int64(freezeTimeRandom))))
for {
bufValue, bufLimit := p.fcClient.BufferStatus()
if bufLimit == 0 {
return
}
if bufValue <= bufLimit/8 {
time.Sleep(freezeCheckPeriod)
} else {
atomic.StoreUint32(&p.frozen, 0)
p.sendResume(bufValue)
break
}
}
}()
}
}
// Handshake executes the les protocol handshake, negotiating version number,
// network IDs, difficulties, head and genesis blocks.
func (p *clientPeer) Handshake(td *big.Int, head common.Hash, headNum uint64, genesis common.Hash, forkID forkid.ID, forkFilter forkid.Filter, server *LesServer) error {
recentTx := server.handler.blockchain.TxLookupLimit()
if recentTx != txIndexUnlimited {
if recentTx < blockSafetyMargin {
recentTx = txIndexDisabled
} else {
recentTx -= blockSafetyMargin - txIndexRecentOffset
}
}
if server.config.UltraLightOnlyAnnounce {
recentTx = txIndexDisabled
}
if recentTx != txIndexUnlimited && p.version < lpv4 {
return errors.New("Cannot serve old clients without a complete tx index")
}
// Note: clientPeer.headInfo should contain the last head announced to the client by us.
// The values announced in the handshake are dummy values for compatibility reasons and should be ignored.
p.headInfo = blockInfo{Hash: head, Number: headNum, Td: td}
return p.handshake(td, head, headNum, genesis, forkID, forkFilter, func(lists *keyValueList) {
// Add some information which services server can offer.
if !server.config.UltraLightOnlyAnnounce {
*lists = (*lists).add("serveHeaders", nil)
*lists = (*lists).add("serveChainSince", uint64(0))
*lists = (*lists).add("serveStateSince", uint64(0))
// If local ethereum node is running in archive mode, advertise ourselves we have
// all version state data. Otherwise only recent state is available.
stateRecent := uint64(core.TriesInMemory - blockSafetyMargin)
if server.archiveMode {
stateRecent = 0
}
*lists = (*lists).add("serveRecentState", stateRecent)
*lists = (*lists).add("txRelay", nil)
}
if p.version >= lpv4 {
*lists = (*lists).add("recentTxLookup", recentTx)
}
*lists = (*lists).add("flowControl/BL", server.defParams.BufLimit)
*lists = (*lists).add("flowControl/MRR", server.defParams.MinRecharge)
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
var costList RequestCostList
if server.costTracker.testCostList != nil {
costList = server.costTracker.testCostList
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
} else {
costList = server.costTracker.makeCostList(server.costTracker.globalFactor())
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
}
*lists = (*lists).add("flowControl/MRC", costList)
p.fcCosts = costList.decode(ProtocolLengths[uint(p.version)])
les, les/flowcontrol: improved request serving and flow control (#18230) This change - implements concurrent LES request serving even for a single peer. - replaces the request cost estimation method with a cost table based on benchmarks which gives much more consistent results. Until now the allowed number of light peers was just a guess which probably contributed a lot to the fluctuating quality of available service. Everything related to request cost is implemented in a single object, the 'cost tracker'. It uses a fixed cost table with a global 'correction factor'. Benchmark code is included and can be run at any time to adapt costs to low-level implementation changes. - reimplements flowcontrol.ClientManager in a cleaner and more efficient way, with added capabilities: There is now control over bandwidth, which allows using the flow control parameters for client prioritization. Target utilization over 100 percent is now supported to model concurrent request processing. Total serving bandwidth is reduced during block processing to prevent database contention. - implements an RPC API for the LES servers allowing server operators to assign priority bandwidth to certain clients and change prioritized status even while the client is connected. The new API is meant for cases where server operators charge for LES using an off-protocol mechanism. - adds a unit test for the new client manager. - adds an end-to-end test using the network simulator that tests bandwidth control functions through the new API.
6 years ago
p.fcParams = server.defParams
// Add advertised checkpoint and register block height which
// client can verify the checkpoint validity.
if server.oracle != nil && server.oracle.IsRunning() {
cp, height := server.oracle.StableCheckpoint()
if cp != nil {
*lists = (*lists).add("checkpoint/value", cp)
*lists = (*lists).add("checkpoint/registerHeight", height)
}
}
}, func(recv keyValueMap) error {
p.server = recv.get("flowControl/MRR", nil) == nil
if p.server {
p.announceType = announceTypeNone // connected to another server, send no messages
} else {
if recv.get("announceType", &p.announceType) != nil {
// set default announceType on server side
p.announceType = announceTypeSimple
}
p.fcClient = flowcontrol.NewClientNode(server.fcManager, p.fcParams)
}
return nil
})
}
func (p *clientPeer) bumpInvalid() {
p.invalidLock.Lock()
p.invalidCount.Add(1, mclock.Now())
p.invalidLock.Unlock()
}
func (p *clientPeer) getInvalid() uint64 {
p.invalidLock.RLock()
defer p.invalidLock.RUnlock()
return p.invalidCount.Value(mclock.Now())
}
// serverPeerSubscriber is an interface to notify services about added or
// removed server peers
type serverPeerSubscriber interface {
registerPeer(*serverPeer)
unregisterPeer(*serverPeer)
}
// serverPeerSet represents the set of active server peers currently
// participating in the Light Ethereum sub-protocol.
type serverPeerSet struct {
peers map[string]*serverPeer
// subscribers is a batch of subscribers and peerset will notify
// these subscribers when the peerset changes(new server peer is
// added or removed)
subscribers []serverPeerSubscriber
closed bool
lock sync.RWMutex
}
// newServerPeerSet creates a new peer set to track the active server peers.
func newServerPeerSet() *serverPeerSet {
return &serverPeerSet{peers: make(map[string]*serverPeer)}
}
// subscribe adds a service to be notified about added or removed
// peers and also register all active peers into the given service.
func (ps *serverPeerSet) subscribe(sub serverPeerSubscriber) {
ps.lock.Lock()
defer ps.lock.Unlock()
ps.subscribers = append(ps.subscribers, sub)
for _, p := range ps.peers {
sub.registerPeer(p)
}
}
// unSubscribe removes the specified service from the subscriber pool.
func (ps *serverPeerSet) unSubscribe(sub serverPeerSubscriber) {
ps.lock.Lock()
defer ps.lock.Unlock()
for i, s := range ps.subscribers {
if s == sub {
ps.subscribers = append(ps.subscribers[:i], ps.subscribers[i+1:]...)
return
}
}
}
// register adds a new server peer into the set, or returns an error if the
// peer is already known.
func (ps *serverPeerSet) register(peer *serverPeer) error {
ps.lock.Lock()
defer ps.lock.Unlock()
if ps.closed {
return errClosed
}
if _, exist := ps.peers[peer.id]; exist {
return errAlreadyRegistered
}
ps.peers[peer.id] = peer
for _, sub := range ps.subscribers {
sub.registerPeer(peer)
}
return nil
}
// unregister removes a remote peer from the active set, disabling any further
// actions to/from that particular entity. It also initiates disconnection at
// the networking layer.
func (ps *serverPeerSet) unregister(id string) error {
ps.lock.Lock()
defer ps.lock.Unlock()
p, ok := ps.peers[id]
if !ok {
return errNotRegistered
}
delete(ps.peers, id)
for _, sub := range ps.subscribers {
sub.unregisterPeer(p)
}
p.Peer.Disconnect(p2p.DiscRequested)
return nil
}
// ids returns a list of all registered peer IDs
func (ps *serverPeerSet) ids() []string {
ps.lock.RLock()
defer ps.lock.RUnlock()
var ids []string
8 years ago
for id := range ps.peers {
ids = append(ids, id)
}
return ids
}
// peer retrieves the registered peer with the given id.
func (ps *serverPeerSet) peer(id string) *serverPeer {
ps.lock.RLock()
defer ps.lock.RUnlock()
return ps.peers[id]
}
// len returns if the current number of peers in the set.
func (ps *serverPeerSet) len() int {
ps.lock.RLock()
defer ps.lock.RUnlock()
return len(ps.peers)
}
// bestPeer retrieves the known peer with the currently highest total difficulty.
// If the peerset is "client peer set", then nothing meaningful will return. The
// reason is client peer never send back their latest status to server.
func (ps *serverPeerSet) bestPeer() *serverPeer {
ps.lock.RLock()
defer ps.lock.RUnlock()
var (
bestPeer *serverPeer
bestTd *big.Int
)
for _, p := range ps.peers {
if td := p.Td(); bestTd == nil || td.Cmp(bestTd) > 0 {
bestPeer, bestTd = p, td
}
}
return bestPeer
}
// allServerPeers returns all server peers in a list.
func (ps *serverPeerSet) allPeers() []*serverPeer {
ps.lock.RLock()
defer ps.lock.RUnlock()
list := make([]*serverPeer, 0, len(ps.peers))
for _, p := range ps.peers {
list = append(list, p)
}
return list
}
// close disconnects all peers. No new peers can be registered
// after close has returned.
func (ps *serverPeerSet) close() {
ps.lock.Lock()
defer ps.lock.Unlock()
for _, p := range ps.peers {
p.Disconnect(p2p.DiscQuitting)
}
ps.closed = true
}