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/p2p/server.go

1132 lines
31 KiB

// Copyright 2014 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 p2p implements the Ethereum p2p network protocols.
package p2p
import (
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
"bytes"
"crypto/ecdsa"
"encoding/hex"
"errors"
"fmt"
"net"
"sort"
"sync"
"sync/atomic"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/common/mclock"
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
"github.com/ethereum/go-ethereum/crypto"
"github.com/ethereum/go-ethereum/event"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/p2p/discover"
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
"github.com/ethereum/go-ethereum/p2p/enode"
"github.com/ethereum/go-ethereum/p2p/enr"
"github.com/ethereum/go-ethereum/p2p/nat"
"github.com/ethereum/go-ethereum/p2p/netutil"
)
const (
defaultDialTimeout = 15 * time.Second
// This is the fairness knob for the discovery mixer. When looking for peers, we'll
// wait this long for a single source of candidates before moving on and trying other
// sources.
discmixTimeout = 5 * time.Second
// Connectivity defaults.
defaultMaxPendingPeers = 50
defaultDialRatio = 3
// This time limits inbound connection attempts per source IP.
inboundThrottleTime = 30 * time.Second
// Maximum time allowed for reading a complete message.
// This is effectively the amount of time a connection can be idle.
frameReadTimeout = 30 * time.Second
// Maximum amount of time allowed for writing a complete message.
frameWriteTimeout = 20 * time.Second
)
var errServerStopped = errors.New("server stopped")
// Config holds Server options.
type Config struct {
// This field must be set to a valid secp256k1 private key.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
PrivateKey *ecdsa.PrivateKey `toml:"-"`
// MaxPeers is the maximum number of peers that can be
// connected. It must be greater than zero.
MaxPeers int
// MaxPendingPeers is the maximum number of peers that can be pending in the
// handshake phase, counted separately for inbound and outbound connections.
// Zero defaults to preset values.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
MaxPendingPeers int `toml:",omitempty"`
// DialRatio controls the ratio of inbound to dialed connections.
// Example: a DialRatio of 2 allows 1/2 of connections to be dialed.
// Setting DialRatio to zero defaults it to 3.
DialRatio int `toml:",omitempty"`
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
// NoDiscovery can be used to disable the peer discovery mechanism.
// Disabling is useful for protocol debugging (manual topology).
NoDiscovery bool
// DiscoveryV5 specifies whether the new topic-discovery based V5 discovery
// protocol should be started or not.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
DiscoveryV5 bool `toml:",omitempty"`
// Name sets the node name of this server.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
Name string `toml:"-"`
// BootstrapNodes are used to establish connectivity
// with the rest of the network.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
BootstrapNodes []*enode.Node
// BootstrapNodesV5 are used to establish connectivity
// with the rest of the network using the V5 discovery
// protocol.
BootstrapNodesV5 []*enode.Node `toml:",omitempty"`
// Static nodes are used as pre-configured connections which are always
// maintained and re-connected on disconnects.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
StaticNodes []*enode.Node
// Trusted nodes are used as pre-configured connections which are always
// allowed to connect, even above the peer limit.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
TrustedNodes []*enode.Node
// Connectivity can be restricted to certain IP networks.
// If this option is set to a non-nil value, only hosts which match one of the
// IP networks contained in the list are considered.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
NetRestrict *netutil.Netlist `toml:",omitempty"`
// NodeDatabase is the path to the database containing the previously seen
// live nodes in the network.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
NodeDatabase string `toml:",omitempty"`
// Protocols should contain the protocols supported
// by the server. Matching protocols are launched for
// each peer.
Protocols []Protocol `toml:"-" json:"-"`
// If ListenAddr is set to a non-nil address, the server
// will listen for incoming connections.
//
// If the port is zero, the operating system will pick a port. The
// ListenAddr field will be updated with the actual address when
// the server is started.
ListenAddr string
// If DiscAddr is set to a non-nil value, the server will use ListenAddr
// for TCP and DiscAddr for the UDP discovery protocol.
DiscAddr string
// If set to a non-nil value, the given NAT port mapper
// is used to make the listening port available to the
// Internet.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
NAT nat.Interface `toml:",omitempty"`
// If Dialer is set to a non-nil value, the given Dialer
// is used to dial outbound peer connections.
Dialer NodeDialer `toml:"-"`
// If NoDial is true, the server will not dial any peers.
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
NoDial bool `toml:",omitempty"`
// If EnableMsgEvents is set then the server will emit PeerEvents
// whenever a message is sent to or received from a peer
EnableMsgEvents bool
// Logger is a custom logger to use with the p2p.Server.
Logger log.Logger `toml:",omitempty"`
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
clock mclock.Clock
}
// Server manages all peer connections.
type Server struct {
// Config fields may not be modified while the server is running.
Config
// Hooks for testing. These are useful because we can inhibit
// the whole protocol stack.
newTransport func(net.Conn, *ecdsa.PublicKey) transport
newPeerHook func(*Peer)
listenFunc func(network, addr string) (net.Listener, error)
lock sync.Mutex // protects running
running bool
listener net.Listener
ourHandshake *protoHandshake
loopWG sync.WaitGroup // loop, listenLoop
peerFeed event.Feed
log log.Logger
nodedb *enode.DB
localnode *enode.LocalNode
ntab *discover.UDPv4
DiscV5 *discover.UDPv5
discmix *enode.FairMix
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
dialsched *dialScheduler
// Channels into the run loop.
quit chan struct{}
addtrusted chan *enode.Node
removetrusted chan *enode.Node
peerOp chan peerOpFunc
peerOpDone chan struct{}
delpeer chan peerDrop
checkpointPostHandshake chan *conn
checkpointAddPeer chan *conn
// State of run loop and listenLoop.
inboundHistory expHeap
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
type peerOpFunc func(map[enode.ID]*Peer)
type peerDrop struct {
*Peer
err error
requested bool // true if signaled by the peer
}
type connFlag int32
const (
dynDialedConn connFlag = 1 << iota
staticDialedConn
inboundConn
trustedConn
)
// conn wraps a network connection with information gathered
// during the two handshakes.
type conn struct {
fd net.Conn
transport
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
node *enode.Node
flags connFlag
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
cont chan error // The run loop uses cont to signal errors to SetupConn.
caps []Cap // valid after the protocol handshake
name string // valid after the protocol handshake
}
type transport interface {
// The two handshakes.
doEncHandshake(prv *ecdsa.PrivateKey) (*ecdsa.PublicKey, error)
doProtoHandshake(our *protoHandshake) (*protoHandshake, error)
// The MsgReadWriter can only be used after the encryption
// handshake has completed. The code uses conn.id to track this
// by setting it to a non-nil value after the encryption handshake.
MsgReadWriter
// transports must provide Close because we use MsgPipe in some of
// the tests. Closing the actual network connection doesn't do
// anything in those tests because MsgPipe doesn't use it.
close(err error)
}
func (c *conn) String() string {
s := c.flags.String()
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if (c.node.ID() != enode.ID{}) {
s += " " + c.node.ID().String()
}
s += " " + c.fd.RemoteAddr().String()
return s
}
func (f connFlag) String() string {
s := ""
if f&trustedConn != 0 {
s += "-trusted"
}
if f&dynDialedConn != 0 {
s += "-dyndial"
}
if f&staticDialedConn != 0 {
s += "-staticdial"
}
if f&inboundConn != 0 {
s += "-inbound"
}
if s != "" {
s = s[1:]
}
return s
}
func (c *conn) is(f connFlag) bool {
flags := connFlag(atomic.LoadInt32((*int32)(&c.flags)))
return flags&f != 0
}
func (c *conn) set(f connFlag, val bool) {
for {
oldFlags := connFlag(atomic.LoadInt32((*int32)(&c.flags)))
flags := oldFlags
if val {
flags |= f
} else {
flags &= ^f
}
if atomic.CompareAndSwapInt32((*int32)(&c.flags), int32(oldFlags), int32(flags)) {
return
}
}
}
// LocalNode returns the local node record.
func (srv *Server) LocalNode() *enode.LocalNode {
return srv.localnode
}
// Peers returns all connected peers.
func (srv *Server) Peers() []*Peer {
var ps []*Peer
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
srv.doPeerOp(func(peers map[enode.ID]*Peer) {
for _, p := range peers {
ps = append(ps, p)
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
})
return ps
}
// PeerCount returns the number of connected peers.
func (srv *Server) PeerCount() int {
var count int
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
srv.doPeerOp(func(ps map[enode.ID]*Peer) {
count = len(ps)
})
return count
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// AddPeer adds the given node to the static node set. When there is room in the peer set,
// the server will connect to the node. If the connection fails for any reason, the server
// will attempt to reconnect the peer.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) AddPeer(node *enode.Node) {
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
srv.dialsched.addStatic(node)
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// RemovePeer removes a node from the static node set. It also disconnects from the given
// node if it is currently connected as a peer.
//
// This method blocks until all protocols have exited and the peer is removed. Do not use
// RemovePeer in protocol implementations, call Disconnect on the Peer instead.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) RemovePeer(node *enode.Node) {
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
var (
ch chan *PeerEvent
sub event.Subscription
)
// Disconnect the peer on the main loop.
srv.doPeerOp(func(peers map[enode.ID]*Peer) {
srv.dialsched.removeStatic(node)
if peer := peers[node.ID()]; peer != nil {
ch = make(chan *PeerEvent, 1)
sub = srv.peerFeed.Subscribe(ch)
peer.Disconnect(DiscRequested)
}
})
// Wait for the peer connection to end.
if ch != nil {
defer sub.Unsubscribe()
for ev := range ch {
if ev.Peer == node.ID() && ev.Type == PeerEventTypeDrop {
return
}
}
}
}
// AddTrustedPeer adds the given node to a reserved trusted list which allows the
// node to always connect, even if the slot are full.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) AddTrustedPeer(node *enode.Node) {
select {
case srv.addtrusted <- node:
case <-srv.quit:
}
}
// RemoveTrustedPeer removes the given node from the trusted peer set.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) RemoveTrustedPeer(node *enode.Node) {
select {
case srv.removetrusted <- node:
case <-srv.quit:
}
}
// SubscribeEvents subscribes the given channel to peer events
func (srv *Server) SubscribeEvents(ch chan *PeerEvent) event.Subscription {
return srv.peerFeed.Subscribe(ch)
}
// Self returns the local node's endpoint information.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) Self() *enode.Node {
srv.lock.Lock()
ln := srv.localnode
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
srv.lock.Unlock()
if ln == nil {
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
return enode.NewV4(&srv.PrivateKey.PublicKey, net.ParseIP("0.0.0.0"), 0, 0)
}
return ln.Node()
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
}
// Stop terminates the server and all active peer connections.
// It blocks until all active connections have been closed.
func (srv *Server) Stop() {
srv.lock.Lock()
if !srv.running {
srv.lock.Unlock()
return
}
srv.running = false
if srv.listener != nil {
// this unblocks listener Accept
srv.listener.Close()
}
close(srv.quit)
srv.lock.Unlock()
srv.loopWG.Wait()
}
// sharedUDPConn implements a shared connection. Write sends messages to the underlying connection while read returns
// messages that were found unprocessable and sent to the unhandled channel by the primary listener.
type sharedUDPConn struct {
*net.UDPConn
unhandled chan discover.ReadPacket
}
// ReadFromUDP implements discover.UDPConn
func (s *sharedUDPConn) ReadFromUDP(b []byte) (n int, addr *net.UDPAddr, err error) {
packet, ok := <-s.unhandled
if !ok {
return 0, nil, errors.New("connection was closed")
}
l := len(packet.Data)
if l > len(b) {
l = len(b)
}
copy(b[:l], packet.Data[:l])
return l, packet.Addr, nil
}
// Close implements discover.UDPConn
func (s *sharedUDPConn) Close() error {
return nil
}
// Start starts running the server.
// Servers can not be re-used after stopping.
func (srv *Server) Start() (err error) {
srv.lock.Lock()
defer srv.lock.Unlock()
if srv.running {
return errors.New("server already running")
}
srv.running = true
srv.log = srv.Config.Logger
if srv.log == nil {
srv.log = log.Root()
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
if srv.clock == nil {
srv.clock = mclock.System{}
}
if srv.NoDial && srv.ListenAddr == "" {
srv.log.Warn("P2P server will be useless, neither dialing nor listening")
}
// static fields
if srv.PrivateKey == nil {
return errors.New("Server.PrivateKey must be set to a non-nil key")
}
if srv.newTransport == nil {
srv.newTransport = newRLPX
}
if srv.listenFunc == nil {
srv.listenFunc = net.Listen
}
srv.quit = make(chan struct{})
srv.delpeer = make(chan peerDrop)
srv.checkpointPostHandshake = make(chan *conn)
srv.checkpointAddPeer = make(chan *conn)
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
srv.addtrusted = make(chan *enode.Node)
srv.removetrusted = make(chan *enode.Node)
srv.peerOp = make(chan peerOpFunc)
srv.peerOpDone = make(chan struct{})
if err := srv.setupLocalNode(); err != nil {
return err
}
if srv.ListenAddr != "" {
if err := srv.setupListening(); err != nil {
return err
}
}
if err := srv.setupDiscovery(); err != nil {
return err
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
srv.setupDialScheduler()
srv.loopWG.Add(1)
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
go srv.run()
return nil
}
func (srv *Server) setupLocalNode() error {
// Create the devp2p handshake.
pubkey := crypto.FromECDSAPub(&srv.PrivateKey.PublicKey)
srv.ourHandshake = &protoHandshake{Version: baseProtocolVersion, Name: srv.Name, ID: pubkey[1:]}
for _, p := range srv.Protocols {
srv.ourHandshake.Caps = append(srv.ourHandshake.Caps, p.cap())
}
sort.Sort(capsByNameAndVersion(srv.ourHandshake.Caps))
// Create the local node.
db, err := enode.OpenDB(srv.Config.NodeDatabase)
if err != nil {
return err
}
srv.nodedb = db
srv.localnode = enode.NewLocalNode(db, srv.PrivateKey)
srv.localnode.SetFallbackIP(net.IP{127, 0, 0, 1})
// TODO: check conflicts
for _, p := range srv.Protocols {
for _, e := range p.Attributes {
srv.localnode.Set(e)
}
}
switch srv.NAT.(type) {
case nil:
// No NAT interface, do nothing.
case nat.ExtIP:
// ExtIP doesn't block, set the IP right away.
ip, _ := srv.NAT.ExternalIP()
srv.localnode.SetStaticIP(ip)
default:
// Ask the router about the IP. This takes a while and blocks startup,
// do it in the background.
srv.loopWG.Add(1)
go func() {
defer srv.loopWG.Done()
if ip, err := srv.NAT.ExternalIP(); err == nil {
srv.localnode.SetStaticIP(ip)
}
}()
}
return nil
}
func (srv *Server) setupDiscovery() error {
srv.discmix = enode.NewFairMix(discmixTimeout)
// Add protocol-specific discovery sources.
added := make(map[string]bool)
for _, proto := range srv.Protocols {
if proto.DialCandidates != nil && !added[proto.Name] {
srv.discmix.AddSource(proto.DialCandidates)
added[proto.Name] = true
}
}
// Don't listen on UDP endpoint if DHT is disabled.
if srv.NoDiscovery && !srv.DiscoveryV5 {
return nil
}
listenAddr := srv.ListenAddr
// Use an alternate listening address for UDP if
// a custom discovery address is configured.
if srv.DiscAddr != "" {
listenAddr = srv.DiscAddr
}
addr, err := net.ResolveUDPAddr("udp", listenAddr)
if err != nil {
return err
}
conn, err := net.ListenUDP("udp", addr)
if err != nil {
return err
}
realaddr := conn.LocalAddr().(*net.UDPAddr)
srv.log.Debug("UDP listener up", "addr", realaddr)
if srv.NAT != nil {
if !realaddr.IP.IsLoopback() {
srv.loopWG.Add(1)
go func() {
nat.Map(srv.NAT, srv.quit, "udp", realaddr.Port, realaddr.Port, "ethereum discovery")
srv.loopWG.Done()
}()
}
}
srv.localnode.SetFallbackUDP(realaddr.Port)
// Discovery V4
var unhandled chan discover.ReadPacket
var sconn *sharedUDPConn
cmd/geth: add --config file flag (#13875) * p2p/discover, p2p/discv5: add marshaling methods to Node * p2p/netutil: make Netlist decodable from TOML * common/math: encode nil HexOrDecimal256 as 0x0 * cmd/geth: add --config file flag * cmd/geth: add missing license header * eth: prettify Config again, fix tests * eth: use gasprice.Config instead of duplicating its fields * eth/gasprice: hide nil default from dumpconfig output * cmd/geth: hide genesis block in dumpconfig output * node: make tests compile * console: fix tests * cmd/geth: make TOML keys look exactly like Go struct fields * p2p: use discovery by default This makes the zero Config slightly more useful. It also fixes package node tests because Node detects reuse of the datadir through the NodeDatabase. * cmd/geth: make ethstats URL settable through config file * cmd/faucet: fix configuration * cmd/geth: dedup attach tests * eth: add comment for DefaultConfig * eth: pass downloader.SyncMode in Config This removes the FastSync, LightSync flags in favour of a more general SyncMode flag. * cmd/utils: remove jitvm flags * cmd/utils: make mutually exclusive flag error prettier It now reads: Fatal: flags --dev, --testnet can't be used at the same time * p2p: fix typo * node: add DefaultConfig, use it for geth * mobile: add missing NoDiscovery option * cmd/utils: drop MakeNode This exposed a couple of places that needed to be updated to use node.DefaultConfig. * node: fix typo * eth: make fast sync the default mode * cmd/utils: remove IPCApiFlag (unused) * node: remove default IPC path Set it in the frontends instead. * cmd/geth: add --syncmode * cmd/utils: make --ipcdisable and --ipcpath mutually exclusive * cmd/utils: don't enable WS, HTTP when setting addr * cmd/utils: fix --identity
8 years ago
if !srv.NoDiscovery {
if srv.DiscoveryV5 {
unhandled = make(chan discover.ReadPacket, 100)
sconn = &sharedUDPConn{conn, unhandled}
}
cfg := discover.Config{
PrivateKey: srv.PrivateKey,
NetRestrict: srv.NetRestrict,
Bootnodes: srv.BootstrapNodes,
Unhandled: unhandled,
Log: srv.log,
}
ntab, err := discover.ListenV4(conn, srv.localnode, cfg)
if err != nil {
return err
}
srv.ntab = ntab
srv.discmix.AddSource(ntab.RandomNodes())
}
// Discovery V5
if srv.DiscoveryV5 {
cfg := discover.Config{
PrivateKey: srv.PrivateKey,
NetRestrict: srv.NetRestrict,
Bootnodes: srv.BootstrapNodesV5,
Log: srv.log,
}
var err error
if sconn != nil {
srv.DiscV5, err = discover.ListenV5(sconn, srv.localnode, cfg)
} else {
srv.DiscV5, err = discover.ListenV5(conn, srv.localnode, cfg)
}
if err != nil {
return err
}
}
return nil
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
func (srv *Server) setupDialScheduler() {
config := dialConfig{
self: srv.localnode.ID(),
maxDialPeers: srv.maxDialedConns(),
maxActiveDials: srv.MaxPendingPeers,
log: srv.Logger,
netRestrict: srv.NetRestrict,
dialer: srv.Dialer,
clock: srv.clock,
}
if srv.ntab != nil {
config.resolver = srv.ntab
}
if config.dialer == nil {
config.dialer = tcpDialer{&net.Dialer{Timeout: defaultDialTimeout}}
}
srv.dialsched = newDialScheduler(config, srv.discmix, srv.SetupConn)
for _, n := range srv.StaticNodes {
srv.dialsched.addStatic(n)
}
}
func (srv *Server) maxInboundConns() int {
return srv.MaxPeers - srv.maxDialedConns()
}
func (srv *Server) maxDialedConns() (limit int) {
if srv.NoDial || srv.MaxPeers == 0 {
return 0
}
if srv.DialRatio == 0 {
limit = srv.MaxPeers / defaultDialRatio
} else {
limit = srv.MaxPeers / srv.DialRatio
}
if limit == 0 {
limit = 1
}
return limit
}
func (srv *Server) setupListening() error {
// Launch the listener.
listener, err := srv.listenFunc("tcp", srv.ListenAddr)
if err != nil {
return err
}
srv.listener = listener
srv.ListenAddr = listener.Addr().String()
// Update the local node record and map the TCP listening port if NAT is configured.
if tcp, ok := listener.Addr().(*net.TCPAddr); ok {
srv.localnode.Set(enr.TCP(tcp.Port))
if !tcp.IP.IsLoopback() && srv.NAT != nil {
srv.loopWG.Add(1)
go func() {
nat.Map(srv.NAT, srv.quit, "tcp", tcp.Port, tcp.Port, "ethereum p2p")
srv.loopWG.Done()
}()
}
}
srv.loopWG.Add(1)
go srv.listenLoop()
return nil
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// doPeerOp runs fn on the main loop.
func (srv *Server) doPeerOp(fn peerOpFunc) {
select {
case srv.peerOp <- fn:
<-srv.peerOpDone
case <-srv.quit:
}
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// run is the main loop of the server.
func (srv *Server) run() {
srv.log.Info("Started P2P networking", "self", srv.localnode.Node().URLv4())
defer srv.loopWG.Done()
defer srv.nodedb.Close()
defer srv.discmix.Close()
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
defer srv.dialsched.stop()
var (
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
peers = make(map[enode.ID]*Peer)
inboundCount = 0
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
trusted = make(map[enode.ID]bool, len(srv.TrustedNodes))
)
// Put trusted nodes into a map to speed up checks.
// Trusted peers are loaded on startup or added via AddTrustedPeer RPC.
for _, n := range srv.TrustedNodes {
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
trusted[n.ID()] = true
}
running:
for {
select {
case <-srv.quit:
// The server was stopped. Run the cleanup logic.
break running
case n := <-srv.addtrusted:
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// This channel is used by AddTrustedPeer to add a node
// to the trusted node set.
srv.log.Trace("Adding trusted node", "node", n)
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
trusted[n.ID()] = true
if p, ok := peers[n.ID()]; ok {
p.rw.set(trustedConn, true)
}
case n := <-srv.removetrusted:
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// This channel is used by RemoveTrustedPeer to remove a node
// from the trusted node set.
srv.log.Trace("Removing trusted node", "node", n)
delete(trusted, n.ID())
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if p, ok := peers[n.ID()]; ok {
p.rw.set(trustedConn, false)
}
case op := <-srv.peerOp:
// This channel is used by Peers and PeerCount.
op(peers)
srv.peerOpDone <- struct{}{}
case c := <-srv.checkpointPostHandshake:
// A connection has passed the encryption handshake so
// the remote identity is known (but hasn't been verified yet).
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if trusted[c.node.ID()] {
// Ensure that the trusted flag is set before checking against MaxPeers.
c.flags |= trustedConn
}
// TODO: track in-progress inbound node IDs (pre-Peer) to avoid dialing them.
c.cont <- srv.postHandshakeChecks(peers, inboundCount, c)
case c := <-srv.checkpointAddPeer:
// At this point the connection is past the protocol handshake.
// Its capabilities are known and the remote identity is verified.
err := srv.addPeerChecks(peers, inboundCount, c)
if err == nil {
// The handshakes are done and it passed all checks.
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
p := srv.launchPeer(c)
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
peers[c.node.ID()] = p
srv.log.Debug("Adding p2p peer", "peercount", len(peers), "id", p.ID(), "conn", c.flags, "addr", p.RemoteAddr(), "name", p.Name())
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
srv.dialsched.peerAdded(c)
if p.Inbound() {
inboundCount++
}
}
c.cont <- err
case pd := <-srv.delpeer:
// A peer disconnected.
d := common.PrettyDuration(mclock.Now() - pd.created)
delete(peers, pd.ID())
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
srv.log.Debug("Removing p2p peer", "peercount", len(peers), "id", pd.ID(), "duration", d, "req", pd.requested, "err", pd.err)
srv.dialsched.peerRemoved(pd.rw)
if pd.Inbound() {
inboundCount--
}
}
}
srv.log.Trace("P2P networking is spinning down")
// Terminate discovery. If there is a running lookup it will terminate soon.
if srv.ntab != nil {
srv.ntab.Close()
}
if srv.DiscV5 != nil {
srv.DiscV5.Close()
}
// Disconnect all peers.
for _, p := range peers {
p.Disconnect(DiscQuitting)
}
// Wait for peers to shut down. Pending connections and tasks are
// not handled here and will terminate soon-ish because srv.quit
// is closed.
for len(peers) > 0 {
p := <-srv.delpeer
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
p.log.Trace("<-delpeer (spindown)")
delete(peers, p.ID())
}
}
func (srv *Server) postHandshakeChecks(peers map[enode.ID]*Peer, inboundCount int, c *conn) error {
switch {
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
case !c.is(trustedConn) && len(peers) >= srv.MaxPeers:
return DiscTooManyPeers
case !c.is(trustedConn) && c.is(inboundConn) && inboundCount >= srv.maxInboundConns():
return DiscTooManyPeers
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
case peers[c.node.ID()] != nil:
return DiscAlreadyConnected
case c.node.ID() == srv.localnode.ID():
return DiscSelf
default:
return nil
}
}
func (srv *Server) addPeerChecks(peers map[enode.ID]*Peer, inboundCount int, c *conn) error {
// Drop connections with no matching protocols.
if len(srv.Protocols) > 0 && countMatchingProtocols(srv.Protocols, c.caps) == 0 {
return DiscUselessPeer
}
// Repeat the post-handshake checks because the
// peer set might have changed since those checks were performed.
return srv.postHandshakeChecks(peers, inboundCount, c)
}
// listenLoop runs in its own goroutine and accepts
// inbound connections.
func (srv *Server) listenLoop() {
srv.log.Debug("TCP listener up", "addr", srv.listener.Addr())
// The slots channel limits accepts of new connections.
tokens := defaultMaxPendingPeers
if srv.MaxPendingPeers > 0 {
tokens = srv.MaxPendingPeers
}
slots := make(chan struct{}, tokens)
for i := 0; i < tokens; i++ {
slots <- struct{}{}
}
// Wait for slots to be returned on exit. This ensures all connection goroutines
// are down before listenLoop returns.
defer srv.loopWG.Done()
defer func() {
for i := 0; i < cap(slots); i++ {
<-slots
}
}()
for {
// Wait for a free slot before accepting.
<-slots
var (
fd net.Conn
err error
lastLog time.Time
)
for {
fd, err = srv.listener.Accept()
if netutil.IsTemporaryError(err) {
if time.Since(lastLog) > 1*time.Second {
srv.log.Debug("Temporary read error", "err", err)
lastLog = time.Now()
}
time.Sleep(time.Millisecond * 200)
continue
} else if err != nil {
srv.log.Debug("Read error", "err", err)
slots <- struct{}{}
return
}
break
}
remoteIP := netutil.AddrIP(fd.RemoteAddr())
if err := srv.checkInboundConn(remoteIP); err != nil {
srv.log.Debug("Rejected inbound connection", "addr", fd.RemoteAddr(), "err", err)
fd.Close()
slots <- struct{}{}
continue
}
if remoteIP != nil {
var addr *net.TCPAddr
if tcp, ok := fd.RemoteAddr().(*net.TCPAddr); ok {
addr = tcp
}
fd = newMeteredConn(fd, true, addr)
srv.log.Trace("Accepted connection", "addr", fd.RemoteAddr())
}
go func() {
srv.SetupConn(fd, inboundConn, nil)
slots <- struct{}{}
}()
}
}
func (srv *Server) checkInboundConn(remoteIP net.IP) error {
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
if remoteIP == nil {
return nil
}
// Reject connections that do not match NetRestrict.
if srv.NetRestrict != nil && !srv.NetRestrict.Contains(remoteIP) {
return fmt.Errorf("not in netrestrict list")
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// Reject Internet peers that try too often.
now := srv.clock.Now()
srv.inboundHistory.expire(now, nil)
if !netutil.IsLAN(remoteIP) && srv.inboundHistory.contains(remoteIP.String()) {
return fmt.Errorf("too many attempts")
}
srv.inboundHistory.add(remoteIP.String(), now.Add(inboundThrottleTime))
return nil
}
// SetupConn runs the handshakes and attempts to add the connection
// as a peer. It returns when the connection has been added as a peer
// or the handshakes have failed.
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) SetupConn(fd net.Conn, flags connFlag, dialDest *enode.Node) error {
c := &conn{fd: fd, flags: flags, cont: make(chan error)}
if dialDest == nil {
c.transport = srv.newTransport(fd, nil)
} else {
c.transport = srv.newTransport(fd, dialDest.Pubkey())
}
err := srv.setupConn(c, flags, dialDest)
if err != nil {
c.close(err)
}
return err
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func (srv *Server) setupConn(c *conn, flags connFlag, dialDest *enode.Node) error {
// Prevent leftover pending conns from entering the handshake.
srv.lock.Lock()
running := srv.running
srv.lock.Unlock()
if !running {
return errServerStopped
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
// If dialing, figure out the remote public key.
if dialDest != nil {
dialPubkey := new(ecdsa.PublicKey)
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if err := dialDest.Load((*enode.Secp256k1)(dialPubkey)); err != nil {
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
err = errors.New("dial destination doesn't have a secp256k1 public key")
srv.log.Trace("Setting up connection failed", "addr", c.fd.RemoteAddr(), "conn", c.flags, "err", err)
return err
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
}
}
// Run the RLPx handshake.
remotePubkey, err := c.doEncHandshake(srv.PrivateKey)
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if err != nil {
srv.log.Trace("Failed RLPx handshake", "addr", c.fd.RemoteAddr(), "conn", c.flags, "err", err)
return err
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if dialDest != nil {
c.node = dialDest
} else {
c.node = nodeFromConn(remotePubkey, c.fd)
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
clog := srv.log.New("id", c.node.ID(), "addr", c.fd.RemoteAddr(), "conn", c.flags)
err = srv.checkpoint(c, srv.checkpointPostHandshake)
if err != nil {
clog.Trace("Rejected peer", "err", err)
return err
}
// Run the capability negotiation handshake.
phs, err := c.doProtoHandshake(srv.ourHandshake)
if err != nil {
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
clog.Trace("Failed p2p handshake", "err", err)
return err
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
if id := c.node.ID(); !bytes.Equal(crypto.Keccak256(phs.ID), id[:]) {
clog.Trace("Wrong devp2p handshake identity", "phsid", hex.EncodeToString(phs.ID))
return DiscUnexpectedIdentity
}
c.caps, c.name = phs.Caps, phs.Name
err = srv.checkpoint(c, srv.checkpointAddPeer)
if err != nil {
clog.Trace("Rejected peer", "err", err)
return err
}
return nil
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
func nodeFromConn(pubkey *ecdsa.PublicKey, conn net.Conn) *enode.Node {
var ip net.IP
var port int
if tcp, ok := conn.RemoteAddr().(*net.TCPAddr); ok {
ip = tcp.IP
port = tcp.Port
}
return enode.NewV4(pubkey, ip, port, port)
}
// checkpoint sends the conn to run, which performs the
// post-handshake checks for the stage (posthandshake, addpeer).
func (srv *Server) checkpoint(c *conn, stage chan<- *conn) error {
select {
case stage <- c:
case <-srv.quit:
return errServerStopped
}
return <-c.cont
}
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
func (srv *Server) launchPeer(c *conn) *Peer {
p := newPeer(srv.log, c, srv.Protocols)
if srv.EnableMsgEvents {
// If message events are enabled, pass the peerFeed
// to the peer.
p.events = &srv.peerFeed
}
go srv.runPeer(p)
return p
}
// runPeer runs in its own goroutine for each peer.
func (srv *Server) runPeer(p *Peer) {
if srv.newPeerHook != nil {
srv.newPeerHook(p)
}
srv.peerFeed.Send(&PeerEvent{
Type: PeerEventTypeAdd,
Peer: p.ID(),
RemoteAddress: p.RemoteAddr().String(),
LocalAddress: p.LocalAddr().String(),
})
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// Run the per-peer main loop.
remoteRequested, err := p.run()
p2p: new dial scheduler (#20592) * p2p: new dial scheduler This change replaces the peer-to-peer dial scheduler with a new and improved implementation. The new code is better than the previous implementation in two key aspects: - The time between discovery of a node and dialing that node is significantly lower in the new version. The old dialState kept a buffer of nodes and launched a task to refill it whenever the buffer became empty. This worked well with the discovery interface we used to have, but doesn't really work with the new iterator-based discovery API. - Selection of static dial candidates (created by Server.AddPeer or through static-nodes.json) performs much better for large amounts of static peers. Connections to static nodes are now limited like dynanic dials and can no longer overstep MaxPeers or the dial ratio. * p2p/simulations/adapters: adapt to new NodeDialer interface * p2p: re-add check for self in checkDial * p2p: remove peersetCh * p2p: allow static dials when discovery is disabled * p2p: add test for dialScheduler.removeStatic * p2p: remove blank line * p2p: fix documentation of maxDialPeers * p2p: change "ok" to "added" in static node log * p2p: improve dialTask docs Also increase log level for "Can't resolve node" * p2p: ensure dial resolver is truly nil without discovery * p2p: add "looking for peers" log message * p2p: clean up Server.run comments * p2p: fix maxDialedConns for maxpeers < dialRatio Always allocate at least one dial slot unless dialing is disabled using NoDial or MaxPeers == 0. Most importantly, this fixes MaxPeers == 1 to dedicate the sole slot to dialing instead of listening. * p2p: fix RemovePeer to disconnect the peer again Also make RemovePeer synchronous and add a test. * p2p: remove "Connection set up" log message * p2p: clean up connection logging We previously logged outgoing connection failures up to three times. - in SetupConn() as "Setting up connection failed addr=..." - in setupConn() with an error-specific message and "id=... addr=..." - in dial() as "Dial error task=..." This commit ensures a single log message is emitted per failure and adds "id=... addr=... conn=..." everywhere (id= omitted when the ID isn't known yet). Also avoid printing a log message when a static dial fails but can't be resolved because discv4 is disabled. The light client hit this case all the time, increasing the message count to four lines per failed connection. * p2p: document that RemovePeer blocks
5 years ago
// Announce disconnect on the main loop to update the peer set.
// The main loop waits for existing peers to be sent on srv.delpeer
// before returning, so this send should not select on srv.quit.
srv.delpeer <- peerDrop{p, err, remoteRequested}
// Broadcast peer drop to external subscribers. This needs to be
// after the send to delpeer so subscribers have a consistent view of
// the peer set (i.e. Server.Peers() doesn't include the peer when the
// event is received).
srv.peerFeed.Send(&PeerEvent{
Type: PeerEventTypeDrop,
Peer: p.ID(),
Error: err.Error(),
RemoteAddress: p.RemoteAddr().String(),
LocalAddress: p.LocalAddr().String(),
})
}
// NodeInfo represents a short summary of the information known about the host.
type NodeInfo struct {
ID string `json:"id"` // Unique node identifier (also the encryption key)
Name string `json:"name"` // Name of the node, including client type, version, OS, custom data
Enode string `json:"enode"` // Enode URL for adding this peer from remote peers
ENR string `json:"enr"` // Ethereum Node Record
IP string `json:"ip"` // IP address of the node
Ports struct {
Discovery int `json:"discovery"` // UDP listening port for discovery protocol
Listener int `json:"listener"` // TCP listening port for RLPx
} `json:"ports"`
ListenAddr string `json:"listenAddr"`
Protocols map[string]interface{} `json:"protocols"`
}
// NodeInfo gathers and returns a collection of metadata known about the host.
func (srv *Server) NodeInfo() *NodeInfo {
// Gather and assemble the generic node infos
node := srv.Self()
info := &NodeInfo{
Name: srv.Name,
Enode: node.URLv4(),
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
ID: node.ID().String(),
IP: node.IP().String(),
ListenAddr: srv.ListenAddr,
Protocols: make(map[string]interface{}),
}
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
6 years ago
info.Ports.Discovery = node.UDP()
info.Ports.Listener = node.TCP()
info.ENR = node.String()
// Gather all the running protocol infos (only once per protocol type)
for _, proto := range srv.Protocols {
if _, ok := info.Protocols[proto.Name]; !ok {
nodeInfo := interface{}("unknown")
if query := proto.NodeInfo; query != nil {
nodeInfo = proto.NodeInfo()
}
info.Protocols[proto.Name] = nodeInfo
}
}
return info
}
// PeersInfo returns an array of metadata objects describing connected peers.
func (srv *Server) PeersInfo() []*PeerInfo {
// Gather all the generic and sub-protocol specific infos
infos := make([]*PeerInfo, 0, srv.PeerCount())
for _, peer := range srv.Peers() {
if peer != nil {
infos = append(infos, peer.Info())
}
}
// Sort the result array alphabetically by node identifier
for i := 0; i < len(infos); i++ {
for j := i + 1; j < len(infos); j++ {
if infos[i].ID > infos[j].ID {
infos[i], infos[j] = infos[j], infos[i]
}
}
}
return infos
}