dash/src/net.cpp

2881 lines
91 KiB
C++
Raw Normal View History

// Copyright (c) 2009-2010 Satoshi Nakamoto
// Copyright (c) 2009-2015 The Bitcoin Core developers
2016-12-20 14:26:45 +01:00
// Copyright (c) 2014-2017 The Dash Core developers
2014-12-13 05:09:33 +01:00
// Distributed under the MIT software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
#if defined(HAVE_CONFIG_H)
2015-04-03 00:51:08 +02:00
#include "config/dash-config.h"
#endif
2011-05-14 23:20:30 +02:00
#include "net.h"
#include "addrman.h"
#include "chainparams.h"
#include "clientversion.h"
#include "consensus/consensus.h"
#include "crypto/common.h"
#include "hash.h"
2014-11-18 22:03:02 +01:00
#include "primitives/transaction.h"
#include "netbase.h"
#include "scheduler.h"
2012-04-15 22:10:54 +02:00
#include "ui_interface.h"
#include "utilstrencodings.h"
#include "instantx.h"
#include "masternode-sync.h"
#include "masternodeman.h"
#include "privatesend.h"
#ifdef WIN32
#include <string.h>
#else
#include <fcntl.h>
#endif
#ifdef USE_UPNP
#include <miniupnpc/miniupnpc.h>
#include <miniupnpc/miniwget.h>
#include <miniupnpc/upnpcommands.h>
#include <miniupnpc/upnperrors.h>
#endif
#include <math.h>
// Dump addresses to peers.dat and banlist.dat every 15 minutes (900s)
#define DUMP_ADDRESSES_INTERVAL 900
// We add a random period time (0 to 1 seconds) to feeler connections to prevent synchronization.
#define FEELER_SLEEP_WINDOW 1
#if !defined(HAVE_MSG_NOSIGNAL) && !defined(MSG_NOSIGNAL)
#define MSG_NOSIGNAL 0
#endif
// Fix for ancient MinGW versions, that don't have defined these in ws2tcpip.h.
// Todo: Can be removed when our pull-tester is upgraded to a modern MinGW version.
#ifdef WIN32
#ifndef PROTECTION_LEVEL_UNRESTRICTED
#define PROTECTION_LEVEL_UNRESTRICTED 10
#endif
#ifndef IPV6_PROTECTION_LEVEL
#define IPV6_PROTECTION_LEVEL 23
#endif
#endif
const static std::string NET_MESSAGE_COMMAND_OTHER = "*other*";
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
constexpr const CConnman::CFullyConnectedOnly CConnman::FullyConnectedOnly;
constexpr const CConnman::CAllNodes CConnman::AllNodes;
//
// Global state variables
//
2012-05-24 19:02:21 +02:00
bool fDiscover = true;
bool fListen = true;
bool fRelayTxes = true;
CCriticalSection cs_mapLocalHost;
std::map<CNetAddr, LocalServiceInfo> mapLocalHost;
static bool vfLimited[NET_MAX] = {};
static CNode* pnodeLocalHost = NULL;
std::string strSubVersion;
limitedmap<uint256, int64_t> mapAlreadyAskedFor(MAX_INV_SZ);
// Signals for message handling
static CNodeSignals g_signals;
CNodeSignals& GetNodeSignals() { return g_signals; }
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::AddOneShot(const std::string& strDest)
{
LOCK(cs_vOneShots);
vOneShots.push_back(strDest);
}
unsigned short GetListenPort()
{
return (unsigned short)(GetArg("-port", Params().GetDefaultPort()));
}
2012-02-12 13:45:24 +01:00
// find 'best' local address for a particular peer
bool GetLocal(CService& addr, const CNetAddr *paddrPeer)
2012-02-12 13:45:24 +01:00
{
if (!fListen)
2012-02-12 13:45:24 +01:00
return false;
int nBestScore = -1;
2012-02-12 13:45:24 +01:00
int nBestReachability = -1;
{
LOCK(cs_mapLocalHost);
for (std::map<CNetAddr, LocalServiceInfo>::iterator it = mapLocalHost.begin(); it != mapLocalHost.end(); it++)
2012-02-12 13:45:24 +01:00
{
int nScore = (*it).second.nScore;
2012-02-12 13:45:24 +01:00
int nReachability = (*it).first.GetReachabilityFrom(paddrPeer);
if (nReachability > nBestReachability || (nReachability == nBestReachability && nScore > nBestScore))
2012-02-12 13:45:24 +01:00
{
addr = CService((*it).first, (*it).second.nPort);
2012-02-12 13:45:24 +01:00
nBestReachability = nReachability;
nBestScore = nScore;
2012-02-12 13:45:24 +01:00
}
}
}
return nBestScore >= 0;
2012-02-12 13:45:24 +01:00
}
//! Convert the pnSeeds6 array into usable address objects.
static std::vector<CAddress> convertSeed6(const std::vector<SeedSpec6> &vSeedsIn)
{
// It'll only connect to one or two seed nodes because once it connects,
// it'll get a pile of addresses with newer timestamps.
// Seed nodes are given a random 'last seen time' of between one and two
// weeks ago.
const int64_t nOneWeek = 7*24*60*60;
std::vector<CAddress> vSeedsOut;
vSeedsOut.reserve(vSeedsIn.size());
for (std::vector<SeedSpec6>::const_iterator i(vSeedsIn.begin()); i != vSeedsIn.end(); ++i)
{
struct in6_addr ip;
memcpy(&ip, i->addr, sizeof(ip));
CAddress addr(CService(ip, i->port), NODE_NETWORK);
addr.nTime = GetTime() - GetRand(nOneWeek) - nOneWeek;
vSeedsOut.push_back(addr);
}
return vSeedsOut;
}
2012-02-12 13:45:24 +01:00
// get best local address for a particular peer as a CAddress
// Otherwise, return the unroutable 0.0.0.0 but filled in with
// the normal parameters, since the IP may be changed to a useful
// one by discovery.
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CAddress GetLocalAddress(const CNetAddr *paddrPeer, ServiceFlags nLocalServices)
2012-02-12 13:45:24 +01:00
{
CAddress ret(CService(CNetAddr(),GetListenPort()), NODE_NONE);
CService addr;
2012-02-12 13:45:24 +01:00
if (GetLocal(addr, paddrPeer))
{
ret = CAddress(addr, nLocalServices);
2012-02-12 13:45:24 +01:00
}
ret.nTime = GetAdjustedTime();
2012-02-12 13:45:24 +01:00
return ret;
}
int GetnScore(const CService& addr)
2012-02-12 13:45:24 +01:00
{
LOCK(cs_mapLocalHost);
if (mapLocalHost.count(addr) == LOCAL_NONE)
return 0;
return mapLocalHost[addr].nScore;
}
// Is our peer's addrLocal potentially useful as an external IP source?
bool IsPeerAddrLocalGood(CNode *pnode)
{
return fDiscover && pnode->addr.IsRoutable() && pnode->addrLocal.IsRoutable() &&
!IsLimited(pnode->addrLocal.GetNetwork());
}
// pushes our own address to a peer
void AdvertiseLocal(CNode *pnode)
{
if (fListen && pnode->fSuccessfullyConnected)
2012-02-12 13:45:24 +01:00
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CAddress addrLocal = GetLocalAddress(&pnode->addr, pnode->GetLocalServices());
// If discovery is enabled, sometimes give our peer the address it
// tells us that it sees us as in case it has a better idea of our
// address than we do.
if (IsPeerAddrLocalGood(pnode) && (!addrLocal.IsRoutable() ||
GetRand((GetnScore(addrLocal) > LOCAL_MANUAL) ? 8:2) == 0))
2012-02-12 13:45:24 +01:00
{
addrLocal.SetIP(pnode->addrLocal);
}
if (addrLocal.IsRoutable())
{
LogPrintf("AdvertiseLocal: advertising address %s\n", addrLocal.ToString());
pnode->PushAddress(addrLocal);
2012-02-12 13:45:24 +01:00
}
}
}
// learn a new local address
bool AddLocal(const CService& addr, int nScore)
2012-02-12 13:45:24 +01:00
{
if (!addr.IsRoutable())
return false;
2012-05-24 19:02:21 +02:00
if (!fDiscover && nScore < LOCAL_MANUAL)
2012-05-13 14:11:53 +02:00
return false;
if (IsLimited(addr))
2012-05-13 15:11:51 +02:00
return false;
LogPrintf("AddLocal(%s,%i)\n", addr.ToString(), nScore);
2012-02-12 13:45:24 +01:00
{
LOCK(cs_mapLocalHost);
bool fAlready = mapLocalHost.count(addr) > 0;
LocalServiceInfo &info = mapLocalHost[addr];
if (!fAlready || nScore >= info.nScore) {
2012-08-29 02:33:25 +02:00
info.nScore = nScore + (fAlready ? 1 : 0);
info.nPort = addr.GetPort();
}
2012-02-12 13:45:24 +01:00
}
return true;
}
bool AddLocal(const CNetAddr &addr, int nScore)
{
return AddLocal(CService(addr, GetListenPort()), nScore);
}
bool RemoveLocal(const CService& addr)
{
LOCK(cs_mapLocalHost);
LogPrintf("RemoveLocal(%s)\n", addr.ToString());
mapLocalHost.erase(addr);
return true;
}
/** Make a particular network entirely off-limits (no automatic connects to it) */
void SetLimited(enum Network net, bool fLimited)
{
2012-05-14 17:15:58 +02:00
if (net == NET_UNROUTABLE)
return;
LOCK(cs_mapLocalHost);
vfLimited[net] = fLimited;
}
2012-05-14 17:15:58 +02:00
bool IsLimited(enum Network net)
{
LOCK(cs_mapLocalHost);
2012-05-14 17:15:58 +02:00
return vfLimited[net];
}
bool IsLimited(const CNetAddr &addr)
{
return IsLimited(addr.GetNetwork());
}
/** vote for a local address */
bool SeenLocal(const CService& addr)
2012-02-12 13:45:24 +01:00
{
{
LOCK(cs_mapLocalHost);
if (mapLocalHost.count(addr) == 0)
return false;
mapLocalHost[addr].nScore++;
2012-02-12 13:45:24 +01:00
}
return true;
}
/** check whether a given address is potentially local */
bool IsLocal(const CService& addr)
2012-02-12 13:45:24 +01:00
{
LOCK(cs_mapLocalHost);
return mapLocalHost.count(addr) > 0;
}
/** check whether a given network is one we can probably connect to */
bool IsReachable(enum Network net)
{
LOCK(cs_mapLocalHost);
return !vfLimited[net];
}
/** check whether a given address is in a network we can probably connect to */
bool IsReachable(const CNetAddr& addr)
{
enum Network net = addr.GetNetwork();
return IsReachable(net);
}
std::vector<unsigned char> CNode::vchSecretKey;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* CConnman::FindNode(const CNetAddr& ip)
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
if ((CNetAddr)pnode->addr == ip)
return (pnode);
return NULL;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* CConnman::FindNode(const CSubNet& subNet)
2015-05-25 20:03:51 +02:00
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
if (subNet.Match((CNetAddr)pnode->addr))
return (pnode);
return NULL;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* CConnman::FindNode(const std::string& addrName)
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
if (pnode->addrName == addrName)
return (pnode);
return NULL;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* CConnman::FindNode(const CService& addr)
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
if((CService)pnode->addr == addr)
return (pnode);
return NULL;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::CheckIncomingNonce(uint64_t nonce)
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes) {
if (!pnode->fSuccessfullyConnected && !pnode->fInbound && pnode->GetLocalNonce() == nonce)
return false;
}
return true;
}
CNode* CConnman::ConnectNode(CAddress addrConnect, const char *pszDest, bool fCountFailure, bool fConnectToMasternode)
{
Backport Bitcoin Qt/Gui changes up to 0.14.x part 3 (#1617) * Merge #8996: Network activity toggle 19f46f1 Qt: New network_disabled icon (Luke Dashjr) 54cf997 RPC/Net: Use boolean consistently for networkactive, and remove from getinfo (Luke Dashjr) b2b33d9 Overhaul network activity toggle (Jonas Schnelli) 32efa79 Qt: Add GUI feedback and control of network activity state. (Jon Lund Steffensen) e38993b RPC: Add "togglenetwork" method to toggle network activity temporarily (Jon Lund Steffensen) 7c9a98a Allow network activity to be temporarily suspended. (Jon Lund Steffensen) * Revert on-click behavior of network status icon to showing peers list Stay with the way Dash handled clicking on the status icon * Add theme support for network disabled icon * Merge #8874: Multiple Selection for peer and ban tables 1077577 Fix auto-deselection of peers (Andrew Chow) addfdeb Multiple Selection for peer and ban tables (Andrew Chow) * Merge #9190: qt: Plug many memory leaks ed998ea qt: Avoid OpenSSL certstore-related memory leak (Wladimir J. van der Laan) 5204598 qt: Avoid shutdownwindow-related memory leak (Wladimir J. van der Laan) e4f126a qt: Avoid splash-screen related memory leak (Wladimir J. van der Laan) 693384e qt: Prevent thread/memory leak on exiting RPCConsole (Wladimir J. van der Laan) 47db075 qt: Plug many memory leaks (Wladimir J. van der Laan) * Merge #9218: qt: Show progress overlay when clicking spinner icon 042f9fa qt: Show progress overlay when clicking spinner icon (Wladimir J. van der Laan) 827d9a3 qt: Replace NetworkToggleStatusBarControl with generic ClickableLabel (Wladimir J. van der Laan) * Merge #9266: Bugfix: Qt/RPCConsole: Put column enum in the right places df17fe0 Bugfix: Qt/RPCConsole: Put column enum in the right places (Luke Dashjr) * Merge #9255: qt: layoutAboutToChange signal is called layoutAboutToBeChanged f36349e qt: Remove on_toggleNetworkActiveButton_clicked from RPCConsole (Wladimir J. van der Laan) 297cc20 qt: layoutAboutToChange signal is called layoutAboutToBeChanged (Wladimir J. van der Laan) * Use UniValue until bitcoin PR #8788 is backported Network active toggle was already based on "[RPC] Give RPC commands more information about the RPC request" We need to use the old UniValue style until that one is backported * Merge #8906: [qt] sync-overlay: Don't show progress twice fafeec3 [qt] sync-overlay: Don't show progress twice (MarcoFalke) * Merge #8985: Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() 3154d6e [Qt] use NotifyHeaderTip's height and date for the progress update (Jonas Schnelli) 0a261b6 Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() (Jonas Schnelli) * Merge #9280: [Qt] Show ModalOverlay by pressing the progress bar, allow hiding 89a3723 [Qt] Show ModalOverlay by pressing the progress bar, disabled show() in sync mode (Jonas Schnelli) * Merge #9461: [Qt] Improve progress display during headers-sync and peer-finding 40ec7c7 [Qt] Improve progress display during headers-sync and peer-finding (Jonas Schnelli) * Merge #9588: qt: Use nPowTargetSpacing constant fa4d478 qt: Use nPowTargetSpacing constant (MarcoFalke) * Hide modal overlay forever when syncing has catched up Don't allow to open it again by clicking on the progress bar and spinner icon. Currently the overlay does not show meaningful information about masternode sync and it gives the impression of being stuck after the block chain sync is done. * Don't include chainparams.h in sendcoinsdialog.cpp This was just a remainder of a backported PR which meant to change some calculation in this file which does not apply to Dash. * Also check for fNetworkActive in ConnectNode * Merge #9528: [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) 988d300 [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) (practicalswift) * Merge #11237: qt: Fixing division by zero in time remaining c8d38abd6 Refactor tipUpdate as per style guide (MeshCollider) 3b69a08c5 Fix division by zero in time remaining (MeshCollider) Pull request description: Fixes https://github.com/bitcoin/bitcoin/issues/10291, https://github.com/bitcoin/bitcoin/issues/11265 progressDelta may be 0 (or even negative according to 11265), this checks for that and prints unknown if it is, because we cannot calculate an estimate for the time remaining (would be infinite or negative). Tree-SHA512: bc5708e5ed6e4670d008219558c5fbb25709bd99a32c98ec39bb74f94a0b7fa058f3d03389ccdd39e6723e6b5b48e34b13ceee7c051c2db631e51d8ec3e1d68c
2017-09-11 15:38:14 +02:00
// TODO: This is different from what we have in Bitcoin which only calls ConnectNode from OpenNetworkConnection
// If we ever switch to using OpenNetworkConnection for MNs as well, this can be removed
if (!fNetworkActive) {
return NULL;
}
if (pszDest == NULL) {
// we clean masternode connections in CMasternodeMan::ProcessMasternodeConnections()
// so should be safe to skip this and connect to local Hot MN on CActiveMasternode::ManageState()
if (IsLocal(addrConnect) && !fConnectToMasternode)
return NULL;
LOCK(cs_vNodes);
// Look for an existing connection
CNode* pnode = FindNode((CService)addrConnect);
if (pnode)
{
// we have existing connection to this node but it was not a connection to masternode,
// change flag and add reference so that we can correctly clear it later
if(fConnectToMasternode && !pnode->fMasternode) {
pnode->AddRef();
pnode->fMasternode = true;
}
return pnode;
}
}
/// debug print
LogPrint("net", "trying connection %s lastseen=%.1fhrs\n",
pszDest ? pszDest : addrConnect.ToString(),
pszDest ? 0.0 : (double)(GetAdjustedTime() - addrConnect.nTime)/3600.0);
// Connect
SOCKET hSocket;
bool proxyConnectionFailed = false;
if (pszDest ? ConnectSocketByName(addrConnect, hSocket, pszDest, Params().GetDefaultPort(), nConnectTimeout, &proxyConnectionFailed) :
ConnectSocket(addrConnect, hSocket, nConnectTimeout, &proxyConnectionFailed))
{
if (!IsSelectableSocket(hSocket)) {
LogPrintf("Cannot create connection: non-selectable socket created (fd >= FD_SETSIZE ?)\n");
CloseSocket(hSocket);
return NULL;
}
if (pszDest && addrConnect.IsValid()) {
// It is possible that we already have a connection to the IP/port pszDest resolved to.
// In that case, drop the connection that was just created, and return the existing CNode instead.
// Also store the name we used to connect in that CNode, so that future FindNode() calls to that
// name catch this early.
LOCK(cs_vNodes);
CNode* pnode = FindNode((CService)addrConnect);
if (pnode)
{
// we have existing connection to this node but it was not a connection to masternode,
// change flag and add reference so that we can correctly clear it later
if(fConnectToMasternode && !pnode->fMasternode) {
pnode->AddRef();
pnode->fMasternode = true;
}
if (pnode->addrName.empty()) {
pnode->addrName = std::string(pszDest);
}
CloseSocket(hSocket);
return pnode;
}
}
addrman.Attempt(addrConnect, fCountFailure);
// Add node
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* pnode = new CNode(GetNewNodeId(), nLocalServices, GetBestHeight(), hSocket, addrConnect, pszDest ? pszDest : "", false, true);
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
pnode->nServicesExpected = ServiceFlags(addrConnect.nServices & nRelevantServices);
pnode->nTimeConnected = GetSystemTimeInSeconds();
if(fConnectToMasternode) {
pnode->AddRef();
pnode->fMasternode = true;
}
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
GetNodeSignals().InitializeNode(pnode, *this);
LOCK(cs_vNodes);
vNodes.push_back(pnode);
return pnode;
} else if (!proxyConnectionFailed) {
// If connecting to the node failed, and failure is not caused by a problem connecting to
// the proxy, mark this as an attempt.
addrman.Attempt(addrConnect, fCountFailure);
}
return NULL;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::DumpBanlist()
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
SweepBanned(); // clean unused entries (if bantime has expired)
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (!BannedSetIsDirty())
return;
int64_t nStart = GetTimeMillis();
CBanDB bandb;
banmap_t banmap;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
SetBannedSetDirty(false);
GetBanned(banmap);
if (!bandb.Write(banmap))
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
SetBannedSetDirty(true);
LogPrint("net", "Flushed %d banned node ips/subnets to banlist.dat %dms\n",
banmap.size(), GetTimeMillis() - nStart);
}
void CNode::CloseSocketDisconnect()
{
fDisconnect = true;
if (hSocket != INVALID_SOCKET)
{
LogPrint("net", "disconnecting peer=%d\n", id);
CloseSocket(hSocket);
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ClearBanned()
{
{
LOCK(cs_setBanned);
setBanned.clear();
setBannedIsDirty = true;
}
DumpBanlist(); //store banlist to disk
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if(clientInterface)
clientInterface->BannedListChanged();
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::IsBanned(CNetAddr ip)
{
bool fResult = false;
{
LOCK(cs_setBanned);
for (banmap_t::iterator it = setBanned.begin(); it != setBanned.end(); it++)
{
2015-05-25 20:03:51 +02:00
CSubNet subNet = (*it).first;
CBanEntry banEntry = (*it).second;
2015-05-25 20:03:51 +02:00
if(subNet.Match(ip) && GetTime() < banEntry.nBanUntil)
fResult = true;
}
}
return fResult;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::IsBanned(CSubNet subnet)
2015-05-25 20:03:51 +02:00
{
bool fResult = false;
{
2013-11-18 01:25:17 +01:00
LOCK(cs_setBanned);
banmap_t::iterator i = setBanned.find(subnet);
if (i != setBanned.end())
{
CBanEntry banEntry = (*i).second;
if (GetTime() < banEntry.nBanUntil)
fResult = true;
}
}
return fResult;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::Ban(const CNetAddr& addr, const BanReason &banReason, int64_t bantimeoffset, bool sinceUnixEpoch) {
CSubNet subNet(addr);
Ban(subNet, banReason, bantimeoffset, sinceUnixEpoch);
2015-05-25 20:03:51 +02:00
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::Ban(const CSubNet& subNet, const BanReason &banReason, int64_t bantimeoffset, bool sinceUnixEpoch) {
CBanEntry banEntry(GetTime());
banEntry.banReason = banReason;
if (bantimeoffset <= 0)
{
bantimeoffset = GetArg("-bantime", DEFAULT_MISBEHAVING_BANTIME);
sinceUnixEpoch = false;
}
banEntry.nBanUntil = (sinceUnixEpoch ? 0 : GetTime() )+bantimeoffset;
{
LOCK(cs_setBanned);
if (setBanned[subNet].nBanUntil < banEntry.nBanUntil) {
setBanned[subNet] = banEntry;
setBannedIsDirty = true;
}
else
return;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if(clientInterface)
clientInterface->BannedListChanged();
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes) {
if (subNet.Match((CNetAddr)pnode->addr))
pnode->fDisconnect = true;
}
}
if(banReason == BanReasonManuallyAdded)
DumpBanlist(); //store banlist to disk immediately if user requested ban
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::Unban(const CNetAddr &addr) {
CSubNet subNet(addr);
2015-05-25 20:03:51 +02:00
return Unban(subNet);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::Unban(const CSubNet &subNet) {
2015-06-19 15:27:37 +02:00
{
LOCK(cs_setBanned);
if (!setBanned.erase(subNet))
return false;
2015-06-19 15:27:37 +02:00
setBannedIsDirty = true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if(clientInterface)
clientInterface->BannedListChanged();
DumpBanlist(); //store banlist to disk immediately
return true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::GetBanned(banmap_t &banMap)
{
LOCK(cs_setBanned);
banMap = setBanned; //create a thread safe copy
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::SetBanned(const banmap_t &banMap)
2015-06-19 15:27:37 +02:00
{
LOCK(cs_setBanned);
setBanned = banMap;
setBannedIsDirty = true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::SweepBanned()
2015-06-19 15:27:37 +02:00
{
int64_t now = GetTime();
LOCK(cs_setBanned);
banmap_t::iterator it = setBanned.begin();
2015-06-19 15:27:37 +02:00
while(it != setBanned.end())
{
CSubNet subNet = (*it).first;
CBanEntry banEntry = (*it).second;
if(now > banEntry.nBanUntil)
2015-06-19 15:27:37 +02:00
{
setBanned.erase(it++);
setBannedIsDirty = true;
LogPrint("net", "%s: Removed banned node ip/subnet from banlist.dat: %s\n", __func__, subNet.ToString());
2015-06-19 15:27:37 +02:00
}
else
++it;
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::BannedSetIsDirty()
2015-06-19 15:27:37 +02:00
{
LOCK(cs_setBanned);
return setBannedIsDirty;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::SetBannedSetDirty(bool dirty)
2015-06-19 15:27:37 +02:00
{
LOCK(cs_setBanned); //reuse setBanned lock for the isDirty flag
setBannedIsDirty = dirty;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::IsWhitelistedRange(const CNetAddr &addr) {
LOCK(cs_vWhitelistedRange);
BOOST_FOREACH(const CSubNet& subnet, vWhitelistedRange) {
if (subnet.Match(addr))
return true;
}
return false;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::AddWhitelistedRange(const CSubNet &subnet) {
LOCK(cs_vWhitelistedRange);
vWhitelistedRange.push_back(subnet);
}
#undef X
#define X(name) stats.name = name
void CNode::copyStats(CNodeStats &stats)
{
2013-11-18 01:25:17 +01:00
stats.nodeid = this->GetId();
X(nServices);
Backport Bitcoin Qt/Gui changes up to 0.14.x part 2 (#1615) * Merge #7506: Use CCoinControl selection in CWallet::FundTransaction d6cc6a1 Use CCoinControl selection in CWallet::FundTransaction (João Barbosa) * Merge #7732: [Qt] Debug window: replace "Build date" with "Datadir" fc737d1 [Qt] remove unused formatBuildDate method (Jonas Schnelli) 4856f1d [Qt] Debug window: replace "Build date" with "Datadir" (Jonas Schnelli) * Merge #7707: [RPC][QT] UI support for abandoned transactions 8efed3b [Qt] Support for abandoned/abandoning transactions (Jonas Schnelli) * Merge #7688: List solvability in listunspent output and improve help c3932b3 List solvability in listunspent output and improve help (Pieter Wuille) * Merge #8006: Qt: Add option to disable the system tray icon 8b0e497 Qt: Add option to hide the system tray icon (Tyler Hardin) * Merge #8073: qt: askpassphrasedialog: Clear pass fields on accept 02ce2a3 qt: askpassphrasedialog: Clear pass fields on accept (Pavel Vasin) * Merge #8231: [Qt] fix a bug where the SplashScreen will not be hidden during startup b3e1348 [Qt] fix a bug where the SplashScreen will not be hidden during startup (Jonas Schnelli) * Merge #8257: Do not ask a UI question from bitcoind 1acf1db Do not ask a UI question from bitcoind (Pieter Wuille) * Merge #8463: [qt] Remove Priority from coincontrol dialog fa8dd78 [qt] Remove Priority from coincontrol dialog (MarcoFalke) * Merge #8678: [Qt][CoinControl] fix UI bug that could result in paying unexpected fee 0480293 [Qt][CoinControl] fix UI bug that could result in paying unexpected fee (Jonas Schnelli) * Merge #8672: Qt: Show transaction size in transaction details window c015634 qt: Adding transaction size to transaction details window (Hampus Sjöberg) \-- merge fix for s/size/total size/ fdf82fb Adding method GetTotalSize() to CTransaction (Hampus Sjöberg) * Merge #8371: [Qt] Add out-of-sync modal info layer 08827df [Qt] modalinfolayer: removed unused comments, renamed signal, code style overhaul (Jonas Schnelli) d8b062e [Qt] only update "amount of blocks left" when the header chain is in-sync (Jonas Schnelli) e3245b4 [Qt] add out-of-sync modal info layer (Jonas Schnelli) e47052f [Qt] ClientModel add method to get the height of the header chain (Jonas Schnelli) a001f18 [Qt] Always pass the numBlocksChanged signal for headers tip changed (Jonas Schnelli) bd44a04 [Qt] make Out-Of-Sync warning icon clickable (Jonas Schnelli) 0904c3c [Refactor] refactor function that forms human readable text out of a timeoffset (Jonas Schnelli) * Merge #8805: Trivial: Grammar and capitalization c9ce17b Trivial: Grammar and capitalization (Derek Miller) * Merge #8885: gui: fix ban from qt console cb78c60 gui: fix ban from qt console (Cory Fields) * Merge #8821: [qt] sync-overlay: Don't block during reindex fa85e86 [qt] sync-overlay: Don't show estimated number of headers left (MarcoFalke) faa4de2 [qt] sync-overlay: Don't block during reindex (MarcoFalke) * Support themes for new transaction_abandoned icon * Fix constructor call to COutput * Merge #7842: RPC: do not print minping time in getpeerinfo when no ping received yet 62a6486 RPC: do not print ping info in getpeerinfo when no ping received yet, fix help (Pavel Janík) * Merge #8918: Qt: Add "Copy URI" to payment request context menu 21f5a63 Qt: Add "Copy URI" to payment request context menu (Luke Dashjr) * Merge #8925: qt: Display minimum ping in debug window. 1724a40 Display minimum ping in debug window. (R E Broadley) * Merge #8972: [Qt] make warnings label selectable (jonasschnelli) ef0c9ee [Qt] make warnings label selectable (Jonas Schnelli) * Make background of warning icon transparent in modaloverlay * Merge #9088: Reduce ambiguity of warning message 77cbbd9 Make warning message about wallet balance possibly being incorrect less ambiguous. (R E Broadley) * Replace Bitcoin with Dash in modal overlay * Remove clicked signals from labelWalletStatus and labelTransactionsStatus As both are really just labels, clicking on those is not possible. This is different in Bitcoin, where these labels are actually buttons. * Pull out modaloverlay show/hide into it's own if/else block and switch to time based check Also don't use masternodeSync.IsBlockchainSynced() for now as it won't report the blockchain being synced before the first block (or other MN data?) arrives. This would otherwise give the impression that sync is being stuck.
2017-09-09 09:04:02 +02:00
X(addr);
2015-11-21 00:51:44 +01:00
X(fRelayTxes);
X(nLastSend);
X(nLastRecv);
X(nTimeConnected);
2014-12-15 11:06:15 +01:00
X(nTimeOffset);
X(addrName);
X(nVersion);
2013-11-26 12:52:21 +01:00
X(cleanSubVer);
X(fInbound);
X(nStartingHeight);
X(nSendBytes);
X(mapSendBytesPerMsgCmd);
X(nRecvBytes);
X(mapRecvBytesPerMsgCmd);
X(fWhitelisted);
2013-11-15 12:24:34 +01:00
// It is common for nodes with good ping times to suddenly become lagged,
// due to a new block arriving or other large transfer.
// Merely reporting pingtime might fool the caller into thinking the node was still responsive,
// since pingtime does not update until the ping is complete, which might take a while.
// So, if a ping is taking an unusually long time in flight,
// the caller can immediately detect that this is happening.
int64_t nPingUsecWait = 0;
if ((0 != nPingNonceSent) && (0 != nPingUsecStart)) {
nPingUsecWait = GetTimeMicros() - nPingUsecStart;
}
2013-11-15 12:24:34 +01:00
2015-03-18 00:06:58 +01:00
// Raw ping time is in microseconds, but show it to user as whole seconds (Dash users should be well used to small numbers with many decimal places by now :)
stats.dPingTime = (((double)nPingUsecTime) / 1e6);
Backport Bitcoin Qt/Gui changes up to 0.14.x part 2 (#1615) * Merge #7506: Use CCoinControl selection in CWallet::FundTransaction d6cc6a1 Use CCoinControl selection in CWallet::FundTransaction (João Barbosa) * Merge #7732: [Qt] Debug window: replace "Build date" with "Datadir" fc737d1 [Qt] remove unused formatBuildDate method (Jonas Schnelli) 4856f1d [Qt] Debug window: replace "Build date" with "Datadir" (Jonas Schnelli) * Merge #7707: [RPC][QT] UI support for abandoned transactions 8efed3b [Qt] Support for abandoned/abandoning transactions (Jonas Schnelli) * Merge #7688: List solvability in listunspent output and improve help c3932b3 List solvability in listunspent output and improve help (Pieter Wuille) * Merge #8006: Qt: Add option to disable the system tray icon 8b0e497 Qt: Add option to hide the system tray icon (Tyler Hardin) * Merge #8073: qt: askpassphrasedialog: Clear pass fields on accept 02ce2a3 qt: askpassphrasedialog: Clear pass fields on accept (Pavel Vasin) * Merge #8231: [Qt] fix a bug where the SplashScreen will not be hidden during startup b3e1348 [Qt] fix a bug where the SplashScreen will not be hidden during startup (Jonas Schnelli) * Merge #8257: Do not ask a UI question from bitcoind 1acf1db Do not ask a UI question from bitcoind (Pieter Wuille) * Merge #8463: [qt] Remove Priority from coincontrol dialog fa8dd78 [qt] Remove Priority from coincontrol dialog (MarcoFalke) * Merge #8678: [Qt][CoinControl] fix UI bug that could result in paying unexpected fee 0480293 [Qt][CoinControl] fix UI bug that could result in paying unexpected fee (Jonas Schnelli) * Merge #8672: Qt: Show transaction size in transaction details window c015634 qt: Adding transaction size to transaction details window (Hampus Sjöberg) \-- merge fix for s/size/total size/ fdf82fb Adding method GetTotalSize() to CTransaction (Hampus Sjöberg) * Merge #8371: [Qt] Add out-of-sync modal info layer 08827df [Qt] modalinfolayer: removed unused comments, renamed signal, code style overhaul (Jonas Schnelli) d8b062e [Qt] only update "amount of blocks left" when the header chain is in-sync (Jonas Schnelli) e3245b4 [Qt] add out-of-sync modal info layer (Jonas Schnelli) e47052f [Qt] ClientModel add method to get the height of the header chain (Jonas Schnelli) a001f18 [Qt] Always pass the numBlocksChanged signal for headers tip changed (Jonas Schnelli) bd44a04 [Qt] make Out-Of-Sync warning icon clickable (Jonas Schnelli) 0904c3c [Refactor] refactor function that forms human readable text out of a timeoffset (Jonas Schnelli) * Merge #8805: Trivial: Grammar and capitalization c9ce17b Trivial: Grammar and capitalization (Derek Miller) * Merge #8885: gui: fix ban from qt console cb78c60 gui: fix ban from qt console (Cory Fields) * Merge #8821: [qt] sync-overlay: Don't block during reindex fa85e86 [qt] sync-overlay: Don't show estimated number of headers left (MarcoFalke) faa4de2 [qt] sync-overlay: Don't block during reindex (MarcoFalke) * Support themes for new transaction_abandoned icon * Fix constructor call to COutput * Merge #7842: RPC: do not print minping time in getpeerinfo when no ping received yet 62a6486 RPC: do not print ping info in getpeerinfo when no ping received yet, fix help (Pavel Janík) * Merge #8918: Qt: Add "Copy URI" to payment request context menu 21f5a63 Qt: Add "Copy URI" to payment request context menu (Luke Dashjr) * Merge #8925: qt: Display minimum ping in debug window. 1724a40 Display minimum ping in debug window. (R E Broadley) * Merge #8972: [Qt] make warnings label selectable (jonasschnelli) ef0c9ee [Qt] make warnings label selectable (Jonas Schnelli) * Make background of warning icon transparent in modaloverlay * Merge #9088: Reduce ambiguity of warning message 77cbbd9 Make warning message about wallet balance possibly being incorrect less ambiguous. (R E Broadley) * Replace Bitcoin with Dash in modal overlay * Remove clicked signals from labelWalletStatus and labelTransactionsStatus As both are really just labels, clicking on those is not possible. This is different in Bitcoin, where these labels are actually buttons. * Pull out modaloverlay show/hide into it's own if/else block and switch to time based check Also don't use masternodeSync.IsBlockchainSynced() for now as it won't report the blockchain being synced before the first block (or other MN data?) arrives. This would otherwise give the impression that sync is being stuck.
2017-09-09 09:04:02 +02:00
stats.dMinPing = (((double)nMinPingUsecTime) / 1e6);
stats.dPingWait = (((double)nPingUsecWait) / 1e6);
2013-11-15 12:24:34 +01:00
// Leave string empty if addrLocal invalid (not filled in yet)
stats.addrLocal = addrLocal.IsValid() ? addrLocal.ToString() : "";
}
#undef X
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CNode::ReceiveMsgBytes(const char *pch, unsigned int nBytes, bool& complete)
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
complete = false;
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
int64_t nTimeMicros = GetTimeMicros();
nLastRecv = nTimeMicros / 1000000;
nRecvBytes += nBytes;
while (nBytes > 0) {
// get current incomplete message, or create a new one
if (vRecvMsg.empty() ||
vRecvMsg.back().complete())
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
vRecvMsg.push_back(CNetMessage(Params().MessageStart(), SER_NETWORK, INIT_PROTO_VERSION));
CNetMessage& msg = vRecvMsg.back();
// absorb network data
int handled;
if (!msg.in_data)
handled = msg.readHeader(pch, nBytes);
else
handled = msg.readData(pch, nBytes);
if (handled < 0)
return false;
if (msg.in_data && msg.hdr.nMessageSize > MAX_PROTOCOL_MESSAGE_LENGTH) {
LogPrint("net", "Oversized message from peer=%i, disconnecting\n", GetId());
return false;
}
pch += handled;
nBytes -= handled;
if (msg.complete()) {
//store received bytes per message command
//to prevent a memory DOS, only allow valid commands
mapMsgCmdSize::iterator i = mapRecvBytesPerMsgCmd.find(msg.hdr.pchCommand);
if (i == mapRecvBytesPerMsgCmd.end())
i = mapRecvBytesPerMsgCmd.find(NET_MESSAGE_COMMAND_OTHER);
assert(i != mapRecvBytesPerMsgCmd.end());
i->second += msg.hdr.nMessageSize + CMessageHeader::HEADER_SIZE;
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
msg.nTime = nTimeMicros;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
complete = true;
}
}
return true;
}
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
void CNode::SetSendVersion(int nVersionIn)
{
// Send version may only be changed in the version message, and
// only one version message is allowed per session. We can therefore
// treat this value as const and even atomic as long as it's only used
// once a version message has been successfully processed. Any attempt to
// set this twice is an error.
if (nSendVersion != 0) {
error("Send version already set for node: %i. Refusing to change from %i to %i", id, nSendVersion, nVersionIn);
} else {
nSendVersion = nVersionIn;
}
}
int CNode::GetSendVersion() const
{
// The send version should always be explicitly set to
// INIT_PROTO_VERSION rather than using this value until SetSendVersion
// has been called.
if (nSendVersion == 0) {
error("Requesting unset send version for node: %i. Using %i", id, INIT_PROTO_VERSION);
return INIT_PROTO_VERSION;
}
return nSendVersion;
}
int CNetMessage::readHeader(const char *pch, unsigned int nBytes)
{
// copy data to temporary parsing buffer
unsigned int nRemaining = 24 - nHdrPos;
unsigned int nCopy = std::min(nRemaining, nBytes);
memcpy(&hdrbuf[nHdrPos], pch, nCopy);
nHdrPos += nCopy;
// if header incomplete, exit
if (nHdrPos < 24)
return nCopy;
// deserialize to CMessageHeader
try {
hdrbuf >> hdr;
}
catch (const std::exception&) {
return -1;
}
// reject messages larger than MAX_SIZE
if (hdr.nMessageSize > MAX_SIZE)
return -1;
// switch state to reading message data
in_data = true;
return nCopy;
}
int CNetMessage::readData(const char *pch, unsigned int nBytes)
{
unsigned int nRemaining = hdr.nMessageSize - nDataPos;
unsigned int nCopy = std::min(nRemaining, nBytes);
2014-06-21 17:00:38 +02:00
if (vRecv.size() < nDataPos + nCopy) {
// Allocate up to 256 KiB ahead, but never more than the total message size.
vRecv.resize(std::min(hdr.nMessageSize, nDataPos + nCopy + 256 * 1024));
}
memcpy(&vRecv[nDataPos], pch, nCopy);
nDataPos += nCopy;
return nCopy;
}
// requires LOCK(cs_vSend)
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
size_t CConnman::SocketSendData(CNode *pnode)
{
std::deque<CSerializeData>::iterator it = pnode->vSendMsg.begin();
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
size_t nSentSize = 0;
while (it != pnode->vSendMsg.end()) {
const CSerializeData &data = *it;
assert(data.size() > pnode->nSendOffset);
int nBytes = send(pnode->hSocket, &data[pnode->nSendOffset], data.size() - pnode->nSendOffset, MSG_NOSIGNAL | MSG_DONTWAIT);
if (nBytes > 0) {
pnode->nLastSend = GetSystemTimeInSeconds();
pnode->nSendBytes += nBytes;
pnode->nSendOffset += nBytes;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
nSentSize += nBytes;
if (pnode->nSendOffset == data.size()) {
pnode->nSendOffset = 0;
pnode->nSendSize -= data.size();
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
pnode->fPauseSend = pnode->nSendSize > nSendBufferMaxSize;
it++;
} else {
// could not send full message; stop sending more
break;
}
} else {
if (nBytes < 0) {
// error
int nErr = WSAGetLastError();
if (nErr != WSAEWOULDBLOCK && nErr != WSAEMSGSIZE && nErr != WSAEINTR && nErr != WSAEINPROGRESS)
{
LogPrintf("socket send error %s\n", NetworkErrorString(nErr));
pnode->fDisconnect = true;
}
}
// couldn't send anything at all
break;
}
}
if (it == pnode->vSendMsg.end()) {
assert(pnode->nSendOffset == 0);
assert(pnode->nSendSize == 0);
}
pnode->vSendMsg.erase(pnode->vSendMsg.begin(), it);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
return nSentSize;
}
struct NodeEvictionCandidate
{
NodeEvictionCandidate(CNode* pnode)
: id(pnode->id),
nTimeConnected(pnode->nTimeConnected),
nMinPingUsecTime(pnode->nMinPingUsecTime),
nLastBlockTime(pnode->nLastBlockTime),
nLastTXTime(pnode->nLastTXTime),
fNetworkNode(pnode->fNetworkNode),
fRelayTxes(pnode->fRelayTxes),
fBloomFilter(pnode->pfilter != NULL),
vchNetGroup(pnode->addr.GetGroup()),
vchKeyedNetGroup(pnode->vchKeyedNetGroup)
{}
int id;
int64_t nTimeConnected;
int64_t nMinPingUsecTime;
int64_t nLastBlockTime;
int64_t nLastTXTime;
bool fNetworkNode;
bool fRelayTxes;
bool fBloomFilter;
std::vector<unsigned char> vchNetGroup;
std::vector<unsigned char> vchKeyedNetGroup;
2015-08-21 02:29:04 +02:00
};
static bool ReverseCompareNodeMinPingTime(const NodeEvictionCandidate& a, const NodeEvictionCandidate& b)
2015-08-13 11:58:58 +02:00
{
return a.nMinPingUsecTime > b.nMinPingUsecTime;
2015-08-13 11:58:58 +02:00
}
static bool ReverseCompareNodeTimeConnected(const NodeEvictionCandidate& a, const NodeEvictionCandidate& b)
2015-08-13 11:58:58 +02:00
{
return a.nTimeConnected > b.nTimeConnected;
2015-08-13 11:58:58 +02:00
}
static bool CompareKeyedNetGroup(const NodeEvictionCandidate& a, const NodeEvictionCandidate& b)
2015-08-13 11:58:58 +02:00
{
return a.vchKeyedNetGroup < b.vchKeyedNetGroup;
}
2015-08-13 11:58:58 +02:00
static bool CompareNodeBlockTime(const NodeEvictionCandidate &a, const NodeEvictionCandidate &b)
{
// There is a fall-through here because it is common for a node to have many peers which have not yet relayed a block.
if (a.nLastBlockTime != b.nLastBlockTime) return a.nLastBlockTime < b.nLastBlockTime;
if (a.fNetworkNode != b.fNetworkNode) return b.fNetworkNode;
return a.nTimeConnected > b.nTimeConnected;
}
static bool CompareNodeTXTime(const NodeEvictionCandidate &a, const NodeEvictionCandidate &b)
{
// There is a fall-through here because it is common for a node to have more than a few peers that have not yet relayed txn.
if (a.nLastTXTime != b.nLastTXTime) return a.nLastTXTime < b.nLastTXTime;
if (a.fRelayTxes != b.fRelayTxes) return b.fRelayTxes;
if (a.fBloomFilter != b.fBloomFilter) return a.fBloomFilter;
return a.nTimeConnected > b.nTimeConnected;
}
/** Try to find a connection to evict when the node is full.
* Extreme care must be taken to avoid opening the node to attacker
* triggered network partitioning.
* The strategy used here is to protect a small number of peers
* for each of several distinct characteristics which are difficult
* to forge. In order to partition a node the attacker must be
* simultaneously better at all of them than honest peers.
*/
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::AttemptToEvictConnection()
{
std::vector<NodeEvictionCandidate> vEvictionCandidates;
2015-08-13 11:58:58 +02:00
{
LOCK(cs_vNodes);
for(size_t i = 0; i < vNodes.size(); ++i) {
CNode* pnode = vNodes[i];
if (pnode->fWhitelisted)
2015-08-13 11:58:58 +02:00
continue;
if (!pnode->fInbound)
2015-08-13 11:58:58 +02:00
continue;
if (pnode->fDisconnect)
2015-08-13 11:58:58 +02:00
continue;
vEvictionCandidates.push_back(NodeEvictionCandidate(pnode));
2015-08-13 11:58:58 +02:00
}
}
if (vEvictionCandidates.empty()) return false;
2015-08-13 11:58:58 +02:00
// Protect connections with certain characteristics
// Deterministically select 4 peers to protect by netgroup.
// An attacker cannot predict which netgroups will be protected.
std::sort(vEvictionCandidates.begin(), vEvictionCandidates.end(), CompareKeyedNetGroup);
2015-08-13 11:58:58 +02:00
vEvictionCandidates.erase(vEvictionCandidates.end() - std::min(4, static_cast<int>(vEvictionCandidates.size())), vEvictionCandidates.end());
if (vEvictionCandidates.empty()) return false;
// Protect the 8 nodes with the lowest minimum ping time.
// An attacker cannot manipulate this metric without physically moving nodes closer to the target.
2015-08-13 11:58:58 +02:00
std::sort(vEvictionCandidates.begin(), vEvictionCandidates.end(), ReverseCompareNodeMinPingTime);
vEvictionCandidates.erase(vEvictionCandidates.end() - std::min(8, static_cast<int>(vEvictionCandidates.size())), vEvictionCandidates.end());
if (vEvictionCandidates.empty()) return false;
// Protect 4 nodes that most recently sent us transactions.
// An attacker cannot manipulate this metric without performing useful work.
std::sort(vEvictionCandidates.begin(), vEvictionCandidates.end(), CompareNodeTXTime);
vEvictionCandidates.erase(vEvictionCandidates.end() - std::min(4, static_cast<int>(vEvictionCandidates.size())), vEvictionCandidates.end());
if (vEvictionCandidates.empty()) return false;
// Protect 4 nodes that most recently sent us blocks.
// An attacker cannot manipulate this metric without performing useful work.
std::sort(vEvictionCandidates.begin(), vEvictionCandidates.end(), CompareNodeBlockTime);
vEvictionCandidates.erase(vEvictionCandidates.end() - std::min(4, static_cast<int>(vEvictionCandidates.size())), vEvictionCandidates.end());
if (vEvictionCandidates.empty()) return false;
2015-08-26 01:31:13 +02:00
// Protect the half of the remaining nodes which have been connected the longest.
// This replicates the non-eviction implicit behavior, and precludes attacks that start later.
2015-08-13 11:58:58 +02:00
std::sort(vEvictionCandidates.begin(), vEvictionCandidates.end(), ReverseCompareNodeTimeConnected);
vEvictionCandidates.erase(vEvictionCandidates.end() - static_cast<int>(vEvictionCandidates.size() / 2), vEvictionCandidates.end());
2015-08-13 11:58:58 +02:00
if (vEvictionCandidates.empty()) return false;
2015-08-13 11:58:58 +02:00
// Identify the network group with the most connections and youngest member.
// (vEvictionCandidates is already sorted by reverse connect time)
std::vector<unsigned char> naMostConnections;
2015-08-13 11:58:58 +02:00
unsigned int nMostConnections = 0;
int64_t nMostConnectionsTime = 0;
std::map<std::vector<unsigned char>, std::vector<NodeEvictionCandidate> > mapAddrCounts;
for(size_t i = 0; i < vEvictionCandidates.size(); ++i) {
const NodeEvictionCandidate& candidate = vEvictionCandidates[i];
mapAddrCounts[candidate.vchNetGroup].push_back(candidate);
int64_t grouptime = mapAddrCounts[candidate.vchNetGroup][0].nTimeConnected;
size_t groupsize = mapAddrCounts[candidate.vchNetGroup].size();
2015-08-13 11:58:58 +02:00
if (groupsize > nMostConnections || (groupsize == nMostConnections && grouptime > nMostConnectionsTime)) {
nMostConnections = groupsize;
nMostConnectionsTime = grouptime;
naMostConnections = candidate.vchNetGroup;
2015-08-13 11:58:58 +02:00
}
}
// Reduce to the network group with the most connections
std::vector<NodeEvictionCandidate> vEvictionNodes = mapAddrCounts[naMostConnections];
// Do not disconnect peers if there is only 1 connection from their network group
if(vEvictionNodes.empty()) {
return false;
}
2015-08-13 11:58:58 +02:00
// Disconnect from the network group with the most connections
int nEvictionId = vEvictionNodes[0].id;
{
LOCK(cs_vNodes);
for(size_t i = 0; i < vNodes.size(); ++i) {
CNode* pnode = vNodes[i];
if(pnode->id == nEvictionId) {
pnode->fDisconnect = true;
return true;
}
}
}
2015-08-13 11:58:58 +02:00
return false;
2015-08-13 11:58:58 +02:00
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::AcceptConnection(const ListenSocket& hListenSocket) {
2015-08-13 11:00:10 +02:00
struct sockaddr_storage sockaddr;
socklen_t len = sizeof(sockaddr);
SOCKET hSocket = accept(hListenSocket.socket, (struct sockaddr*)&sockaddr, &len);
CAddress addr;
int nInbound = 0;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
int nMaxInbound = nMaxConnections - (nMaxOutbound + nMaxFeeler);
2015-08-13 11:00:10 +02:00
if (hSocket != INVALID_SOCKET)
if (!addr.SetSockAddr((const struct sockaddr*)&sockaddr))
LogPrintf("Warning: Unknown socket family\n");
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool whitelisted = hListenSocket.whitelisted || IsWhitelistedRange(addr);
2015-08-13 11:00:10 +02:00
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
if (pnode->fInbound)
nInbound++;
}
if (hSocket == INVALID_SOCKET)
{
int nErr = WSAGetLastError();
if (nErr != WSAEWOULDBLOCK)
LogPrintf("socket error accept failed: %s\n", NetworkErrorString(nErr));
return;
2015-08-13 11:00:10 +02:00
}
Backport Bitcoin Qt/Gui changes up to 0.14.x part 3 (#1617) * Merge #8996: Network activity toggle 19f46f1 Qt: New network_disabled icon (Luke Dashjr) 54cf997 RPC/Net: Use boolean consistently for networkactive, and remove from getinfo (Luke Dashjr) b2b33d9 Overhaul network activity toggle (Jonas Schnelli) 32efa79 Qt: Add GUI feedback and control of network activity state. (Jon Lund Steffensen) e38993b RPC: Add "togglenetwork" method to toggle network activity temporarily (Jon Lund Steffensen) 7c9a98a Allow network activity to be temporarily suspended. (Jon Lund Steffensen) * Revert on-click behavior of network status icon to showing peers list Stay with the way Dash handled clicking on the status icon * Add theme support for network disabled icon * Merge #8874: Multiple Selection for peer and ban tables 1077577 Fix auto-deselection of peers (Andrew Chow) addfdeb Multiple Selection for peer and ban tables (Andrew Chow) * Merge #9190: qt: Plug many memory leaks ed998ea qt: Avoid OpenSSL certstore-related memory leak (Wladimir J. van der Laan) 5204598 qt: Avoid shutdownwindow-related memory leak (Wladimir J. van der Laan) e4f126a qt: Avoid splash-screen related memory leak (Wladimir J. van der Laan) 693384e qt: Prevent thread/memory leak on exiting RPCConsole (Wladimir J. van der Laan) 47db075 qt: Plug many memory leaks (Wladimir J. van der Laan) * Merge #9218: qt: Show progress overlay when clicking spinner icon 042f9fa qt: Show progress overlay when clicking spinner icon (Wladimir J. van der Laan) 827d9a3 qt: Replace NetworkToggleStatusBarControl with generic ClickableLabel (Wladimir J. van der Laan) * Merge #9266: Bugfix: Qt/RPCConsole: Put column enum in the right places df17fe0 Bugfix: Qt/RPCConsole: Put column enum in the right places (Luke Dashjr) * Merge #9255: qt: layoutAboutToChange signal is called layoutAboutToBeChanged f36349e qt: Remove on_toggleNetworkActiveButton_clicked from RPCConsole (Wladimir J. van der Laan) 297cc20 qt: layoutAboutToChange signal is called layoutAboutToBeChanged (Wladimir J. van der Laan) * Use UniValue until bitcoin PR #8788 is backported Network active toggle was already based on "[RPC] Give RPC commands more information about the RPC request" We need to use the old UniValue style until that one is backported * Merge #8906: [qt] sync-overlay: Don't show progress twice fafeec3 [qt] sync-overlay: Don't show progress twice (MarcoFalke) * Merge #8985: Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() 3154d6e [Qt] use NotifyHeaderTip's height and date for the progress update (Jonas Schnelli) 0a261b6 Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() (Jonas Schnelli) * Merge #9280: [Qt] Show ModalOverlay by pressing the progress bar, allow hiding 89a3723 [Qt] Show ModalOverlay by pressing the progress bar, disabled show() in sync mode (Jonas Schnelli) * Merge #9461: [Qt] Improve progress display during headers-sync and peer-finding 40ec7c7 [Qt] Improve progress display during headers-sync and peer-finding (Jonas Schnelli) * Merge #9588: qt: Use nPowTargetSpacing constant fa4d478 qt: Use nPowTargetSpacing constant (MarcoFalke) * Hide modal overlay forever when syncing has catched up Don't allow to open it again by clicking on the progress bar and spinner icon. Currently the overlay does not show meaningful information about masternode sync and it gives the impression of being stuck after the block chain sync is done. * Don't include chainparams.h in sendcoinsdialog.cpp This was just a remainder of a backported PR which meant to change some calculation in this file which does not apply to Dash. * Also check for fNetworkActive in ConnectNode * Merge #9528: [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) 988d300 [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) (practicalswift) * Merge #11237: qt: Fixing division by zero in time remaining c8d38abd6 Refactor tipUpdate as per style guide (MeshCollider) 3b69a08c5 Fix division by zero in time remaining (MeshCollider) Pull request description: Fixes https://github.com/bitcoin/bitcoin/issues/10291, https://github.com/bitcoin/bitcoin/issues/11265 progressDelta may be 0 (or even negative according to 11265), this checks for that and prints unknown if it is, because we cannot calculate an estimate for the time remaining (would be infinite or negative). Tree-SHA512: bc5708e5ed6e4670d008219558c5fbb25709bd99a32c98ec39bb74f94a0b7fa058f3d03389ccdd39e6723e6b5b48e34b13ceee7c051c2db631e51d8ec3e1d68c
2017-09-11 15:38:14 +02:00
if (!fNetworkActive) {
LogPrintf("connection from %s dropped: not accepting new connections\n", addr.ToString());
CloseSocket(hSocket);
return;
}
if (!IsSelectableSocket(hSocket))
2015-08-13 11:00:10 +02:00
{
LogPrintf("connection from %s dropped: non-selectable socket\n", addr.ToString());
CloseSocket(hSocket);
return;
2015-08-13 11:00:10 +02:00
}
// According to the internet TCP_NODELAY is not carried into accepted sockets
// on all platforms. Set it again here just to be sure.
int set = 1;
#ifdef WIN32
setsockopt(hSocket, IPPROTO_TCP, TCP_NODELAY, (const char*)&set, sizeof(int));
#else
setsockopt(hSocket, IPPROTO_TCP, TCP_NODELAY, (void*)&set, sizeof(int));
#endif
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (IsBanned(addr) && !whitelisted)
2015-08-13 11:00:10 +02:00
{
LogPrintf("connection from %s dropped (banned)\n", addr.ToString());
2015-08-13 11:00:10 +02:00
CloseSocket(hSocket);
return;
2015-08-13 11:00:10 +02:00
}
if (nInbound >= nMaxInbound)
2015-08-13 11:00:10 +02:00
{
if (!AttemptToEvictConnection()) {
2015-08-13 11:58:58 +02:00
// No connection to evict, disconnect the new connection
LogPrint("net", "failed to find an eviction candidate - connection dropped (full)\n");
CloseSocket(hSocket);
return;
}
2015-08-13 11:00:10 +02:00
}
// don't accept incoming connections until fully synced
if(fMasterNode && !masternodeSync.IsSynced()) {
LogPrintf("AcceptConnection -- masternode is not synced yet, skipping inbound connection attempt\n");
CloseSocket(hSocket);
return;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* pnode = new CNode(GetNewNodeId(), nLocalServices, GetBestHeight(), hSocket, addr, "", true);
pnode->fWhitelisted = whitelisted;
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
GetNodeSignals().InitializeNode(pnode, *this);
2015-08-13 11:00:10 +02:00
LogPrint("net", "connection from %s accepted\n", addr.ToString());
{
LOCK(cs_vNodes);
vNodes.push_back(pnode);
2015-08-13 11:00:10 +02:00
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ThreadSocketHandler()
{
unsigned int nPrevNodeCount = 0;
while (!interruptNet)
{
//
// Disconnect nodes
//
{
LOCK(cs_vNodes);
// Disconnect unused nodes
std::vector<CNode*> vNodesCopy = vNodes;
BOOST_FOREACH(CNode* pnode, vNodesCopy)
{
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
if (pnode->fDisconnect)
{
LogPrintf("ThreadSocketHandler -- removing node: peer=%d addr=%s nRefCount=%d fNetworkNode=%d fInbound=%d fMasternode=%d\n",
pnode->id, pnode->addr.ToString(), pnode->GetRefCount(), pnode->fNetworkNode, pnode->fInbound, pnode->fMasternode);
// remove from vNodes
vNodes.erase(remove(vNodes.begin(), vNodes.end(), pnode), vNodes.end());
// release outbound grant (if any)
pnode->grantOutbound.Release();
pnode->grantMasternodeOutbound.Release();
// close socket and cleanup
pnode->CloseSocketDisconnect();
// hold in disconnected pool until all refs are released
if (pnode->fNetworkNode || pnode->fInbound)
pnode->Release();
if (pnode->fMasternode)
pnode->Release();
vNodesDisconnected.push_back(pnode);
}
}
}
{
// Delete disconnected nodes
std::list<CNode*> vNodesDisconnectedCopy = vNodesDisconnected;
BOOST_FOREACH(CNode* pnode, vNodesDisconnectedCopy)
{
// wait until threads are done using it
if (pnode->GetRefCount() <= 0)
{
bool fDelete = false;
{
TRY_LOCK(pnode->cs_vSend, lockSend);
if (lockSend)
{
TRY_LOCK(pnode->cs_inventory, lockInv);
if (lockInv)
fDelete = true;
}
}
if (fDelete)
{
vNodesDisconnected.remove(pnode);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
DeleteNode(pnode);
}
}
}
}
size_t vNodesSize;
{
LOCK(cs_vNodes);
vNodesSize = vNodes.size();
}
if(vNodesSize != nPrevNodeCount) {
nPrevNodeCount = vNodesSize;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if(clientInterface)
clientInterface->NotifyNumConnectionsChanged(nPrevNodeCount);
}
//
// Find which sockets have data to receive
//
struct timeval timeout;
timeout.tv_sec = 0;
timeout.tv_usec = 50000; // frequency to poll pnode->vSend
fd_set fdsetRecv;
fd_set fdsetSend;
fd_set fdsetError;
FD_ZERO(&fdsetRecv);
FD_ZERO(&fdsetSend);
FD_ZERO(&fdsetError);
SOCKET hSocketMax = 0;
bool have_fds = false;
2010-12-22 14:08:00 +01:00
BOOST_FOREACH(const ListenSocket& hListenSocket, vhListenSocket) {
FD_SET(hListenSocket.socket, &fdsetRecv);
hSocketMax = std::max(hSocketMax, hListenSocket.socket);
have_fds = true;
}
{
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
{
if (pnode->hSocket == INVALID_SOCKET)
continue;
FD_SET(pnode->hSocket, &fdsetError);
hSocketMax = std::max(hSocketMax, pnode->hSocket);
have_fds = true;
// Implement the following logic:
// * If there is data to send, select() for sending data. As this only
// happens when optimistic write failed, we choose to first drain the
// write buffer in this case before receiving more. This avoids
// needlessly queueing received data, if the remote peer is not themselves
// receiving data. This means properly utilizing TCP flow control signalling.
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
// * Otherwise, if there is space left in the receive buffer, select() for
// receiving data.
// * Hand off all complete messages to the processor, to be handled without
// blocking here.
{
TRY_LOCK(pnode->cs_vSend, lockSend);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (lockSend) {
if (!pnode->vSendMsg.empty()) {
FD_SET(pnode->hSocket, &fdsetSend);
continue;
}
}
}
{
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
if (!pnode->fPauseRecv)
FD_SET(pnode->hSocket, &fdsetRecv);
}
}
}
int nSelect = select(have_fds ? hSocketMax + 1 : 0,
&fdsetRecv, &fdsetSend, &fdsetError, &timeout);
if (interruptNet)
return;
if (nSelect == SOCKET_ERROR)
{
if (have_fds)
{
int nErr = WSAGetLastError();
LogPrintf("socket select error %s\n", NetworkErrorString(nErr));
for (unsigned int i = 0; i <= hSocketMax; i++)
FD_SET(i, &fdsetRecv);
}
FD_ZERO(&fdsetSend);
FD_ZERO(&fdsetError);
if (!interruptNet.sleep_for(std::chrono::milliseconds(timeout.tv_usec/1000)))
return;
}
//
// Accept new connections
//
BOOST_FOREACH(const ListenSocket& hListenSocket, vhListenSocket)
{
if (hListenSocket.socket != INVALID_SOCKET && FD_ISSET(hListenSocket.socket, &fdsetRecv))
{
2015-08-13 11:00:10 +02:00
AcceptConnection(hListenSocket);
}
}
//
// Service each socket
//
std::vector<CNode*> vNodesCopy = CopyNodeVector();
BOOST_FOREACH(CNode* pnode, vNodesCopy)
{
if (interruptNet)
return;
//
// Receive
//
if (pnode->hSocket == INVALID_SOCKET)
continue;
if (FD_ISSET(pnode->hSocket, &fdsetRecv) || FD_ISSET(pnode->hSocket, &fdsetError))
{
{
{
// typical socket buffer is 8K-64K
char pchBuf[0x10000];
int nBytes = recv(pnode->hSocket, pchBuf, sizeof(pchBuf), MSG_DONTWAIT);
if (nBytes > 0)
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool notify = false;
if (!pnode->ReceiveMsgBytes(pchBuf, nBytes, notify))
pnode->CloseSocketDisconnect();
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
RecordBytesRecv(nBytes);
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
if (notify) {
size_t nSizeAdded = 0;
auto it(pnode->vRecvMsg.begin());
for (; it != pnode->vRecvMsg.end(); ++it) {
if (!it->complete())
break;
nSizeAdded += it->vRecv.size() + CMessageHeader::HEADER_SIZE;
}
{
LOCK(pnode->cs_vProcessMsg);
pnode->vProcessMsg.splice(pnode->vProcessMsg.end(), pnode->vRecvMsg, pnode->vRecvMsg.begin(), it);
pnode->nProcessQueueSize += nSizeAdded;
pnode->fPauseRecv = pnode->nProcessQueueSize > nReceiveFloodSize;
}
WakeMessageHandler();
}
}
else if (nBytes == 0)
{
// socket closed gracefully
if (!pnode->fDisconnect)
LogPrint("net", "socket closed\n");
pnode->CloseSocketDisconnect();
}
else if (nBytes < 0)
{
// error
int nErr = WSAGetLastError();
if (nErr != WSAEWOULDBLOCK && nErr != WSAEMSGSIZE && nErr != WSAEINTR && nErr != WSAEINPROGRESS)
{
if (!pnode->fDisconnect)
LogPrintf("socket recv error %s\n", NetworkErrorString(nErr));
pnode->CloseSocketDisconnect();
}
}
}
}
}
//
// Send
//
if (pnode->hSocket == INVALID_SOCKET)
continue;
if (FD_ISSET(pnode->hSocket, &fdsetSend))
{
TRY_LOCK(pnode->cs_vSend, lockSend);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (lockSend) {
size_t nBytes = SocketSendData(pnode);
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
if (nBytes) {
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
RecordBytesSent(nBytes);
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
}
}
//
// Inactivity checking
//
int64_t nTime = GetSystemTimeInSeconds();
if (nTime - pnode->nTimeConnected > 60)
{
if (pnode->nLastRecv == 0 || pnode->nLastSend == 0)
{
LogPrint("net", "socket no message in first 60 seconds, %d %d from %d\n", pnode->nLastRecv != 0, pnode->nLastSend != 0, pnode->id);
pnode->fDisconnect = true;
}
else if (nTime - pnode->nLastSend > TIMEOUT_INTERVAL)
{
LogPrintf("socket sending timeout: %is\n", nTime - pnode->nLastSend);
pnode->fDisconnect = true;
}
else if (nTime - pnode->nLastRecv > (pnode->nVersion > BIP0031_VERSION ? TIMEOUT_INTERVAL : 90*60))
{
LogPrintf("socket receive timeout: %is\n", nTime - pnode->nLastRecv);
pnode->fDisconnect = true;
}
else if (pnode->nPingNonceSent && pnode->nPingUsecStart + TIMEOUT_INTERVAL * 1000000 < GetTimeMicros())
{
LogPrintf("ping timeout: %fs\n", 0.000001 * (GetTimeMicros() - pnode->nPingUsecStart));
pnode->fDisconnect = true;
}
}
}
ReleaseNodeVector(vNodesCopy);
}
}
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
void CConnman::WakeMessageHandler()
{
{
std::lock_guard<std::mutex> lock(mutexMsgProc);
fMsgProcWake = true;
}
condMsgProc.notify_one();
}
#ifdef USE_UPNP
void ThreadMapPort()
{
std::string port = strprintf("%u", GetListenPort());
const char * multicastif = 0;
const char * minissdpdpath = 0;
struct UPNPDev * devlist = 0;
char lanaddr[64];
#ifndef UPNPDISCOVER_SUCCESS
/* miniupnpc 1.5 */
devlist = upnpDiscover(2000, multicastif, minissdpdpath, 0);
2015-12-09 10:06:41 +01:00
#elif MINIUPNPC_API_VERSION < 14
/* miniupnpc 1.6 */
int error = 0;
devlist = upnpDiscover(2000, multicastif, minissdpdpath, 0, 0, &error);
2015-12-09 10:06:41 +01:00
#else
/* miniupnpc 1.9.20150730 */
int error = 0;
devlist = upnpDiscover(2000, multicastif, minissdpdpath, 0, 0, 2, &error);
#endif
struct UPNPUrls urls;
struct IGDdatas data;
int r;
r = UPNP_GetValidIGD(devlist, &urls, &data, lanaddr, sizeof(lanaddr));
if (r == 1)
{
2012-05-24 19:02:21 +02:00
if (fDiscover) {
char externalIPAddress[40];
r = UPNP_GetExternalIPAddress(urls.controlURL, data.first.servicetype, externalIPAddress);
if(r != UPNPCOMMAND_SUCCESS)
LogPrintf("UPnP: GetExternalIPAddress() returned %d\n", r);
else
{
if(externalIPAddress[0])
{
CNetAddr resolved;
if(LookupHost(externalIPAddress, resolved, false)) {
LogPrintf("UPnP: ExternalIPAddress = %s\n", resolved.ToString().c_str());
AddLocal(resolved, LOCAL_UPNP);
}
}
else
LogPrintf("UPnP: GetExternalIPAddress failed.\n");
}
}
std::string strDesc = "Dash Core " + FormatFullVersion();
try {
while (true) {
#ifndef UPNPDISCOVER_SUCCESS
/* miniupnpc 1.5 */
r = UPNP_AddPortMapping(urls.controlURL, data.first.servicetype,
port.c_str(), port.c_str(), lanaddr, strDesc.c_str(), "TCP", 0);
#else
/* miniupnpc 1.6 */
r = UPNP_AddPortMapping(urls.controlURL, data.first.servicetype,
port.c_str(), port.c_str(), lanaddr, strDesc.c_str(), "TCP", 0, "0");
#endif
if(r!=UPNPCOMMAND_SUCCESS)
LogPrintf("AddPortMapping(%s, %s, %s) failed with code %d (%s)\n",
port, port, lanaddr, r, strupnperror(r));
else
2017-09-07 17:59:00 +02:00
LogPrintf("UPnP Port Mapping successful.\n");
MilliSleep(20*60*1000); // Refresh every 20 minutes
}
}
catch (const boost::thread_interrupted&)
{
r = UPNP_DeletePortMapping(urls.controlURL, data.first.servicetype, port.c_str(), "TCP", 0);
LogPrintf("UPNP_DeletePortMapping() returned: %d\n", r);
freeUPNPDevlist(devlist); devlist = 0;
FreeUPNPUrls(&urls);
throw;
}
} else {
LogPrintf("No valid UPnP IGDs found\n");
freeUPNPDevlist(devlist); devlist = 0;
if (r != 0)
FreeUPNPUrls(&urls);
}
}
void MapPort(bool fUseUPnP)
{
static boost::thread* upnp_thread = NULL;
if (fUseUPnP)
{
if (upnp_thread) {
upnp_thread->interrupt();
upnp_thread->join();
delete upnp_thread;
}
upnp_thread = new boost::thread(boost::bind(&TraceThread<void (*)()>, "upnp", &ThreadMapPort));
}
else if (upnp_thread) {
upnp_thread->interrupt();
upnp_thread->join();
delete upnp_thread;
upnp_thread = NULL;
}
}
#else
void MapPort(bool)
{
// Intentionally left blank.
}
#endif
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ThreadDNSAddressSeed()
2011-11-21 18:25:00 +01:00
{
// goal: only query DNS seeds if address need is acute
if ((addrman.size() > 0) &&
(!GetBoolArg("-forcednsseed", DEFAULT_FORCEDNSSEED))) {
if (!interruptNet.sleep_for(std::chrono::seconds(11)))
return;
LOCK(cs_vNodes);
if (vNodes.size() >= 2) {
LogPrintf("P2P peers available. Skipped DNS seeding.\n");
return;
}
}
const std::vector<CDNSSeedData> &vSeeds = Params().DNSSeeds();
2011-03-09 04:40:50 +01:00
int found = 0;
LogPrintf("Loading addresses from DNS seeds (could take a while)\n");
2013-01-30 05:13:17 +01:00
BOOST_FOREACH(const CDNSSeedData &seed, vSeeds) {
2013-01-30 05:13:17 +01:00
if (HaveNameProxy()) {
AddOneShot(seed.host);
2013-01-30 05:13:17 +01:00
} else {
std::vector<CNetAddr> vIPs;
std::vector<CAddress> vAdd;
if (LookupHost(seed.host.c_str(), vIPs, 0, true))
2013-01-30 05:13:17 +01:00
{
2015-05-31 15:36:44 +02:00
BOOST_FOREACH(const CNetAddr& ip, vIPs)
{
2013-01-30 05:13:17 +01:00
int nOneDay = 24*3600;
CAddress addr = CAddress(CService(ip, Params().GetDefaultPort()), NODE_NETWORK);
2013-01-30 05:13:17 +01:00
addr.nTime = GetTime() - 3*nOneDay - GetRand(4*nOneDay); // use a random age between 3 and 7 days old
vAdd.push_back(addr);
found++;
}
2011-03-09 04:40:50 +01:00
}
// TODO: The seed name resolve may fail, yielding an IP of [::], which results in
// addrman assigning the same source to results from different seeds.
// This should switch to a hard-coded stable dummy IP for each seed name, so that the
// resolve is not required at all.
if (!vIPs.empty()) {
CService seedSource;
Lookup(seed.name.c_str(), seedSource, 0, true);
addrman.Add(vAdd, seedSource);
}
2011-03-09 04:40:50 +01:00
}
}
LogPrintf("%d addresses found from DNS seeds\n", found);
2011-03-09 04:40:50 +01:00
}
2011-11-21 18:25:00 +01:00
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::DumpAddresses()
{
int64_t nStart = GetTimeMillis();
CAddrDB adb;
adb.Write(addrman);
LogPrint("net", "Flushed %d addresses to peers.dat %dms\n",
addrman.size(), GetTimeMillis() - nStart);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::DumpData()
2015-06-19 15:27:37 +02:00
{
DumpAddresses();
DumpBanlist();
2015-06-19 15:27:37 +02:00
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ProcessOneShot()
{
std::string strDest;
{
LOCK(cs_vOneShots);
if (vOneShots.empty())
return;
strDest = vOneShots.front();
vOneShots.pop_front();
}
CAddress addr;
CSemaphoreGrant grant(*semOutbound, true);
if (grant) {
if (!OpenNetworkConnection(addr, false, &grant, strDest.c_str(), true))
AddOneShot(strDest);
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ThreadOpenConnections()
{
// Connect to specific addresses
2012-08-21 17:32:04 +02:00
if (mapArgs.count("-connect") && mapMultiArgs["-connect"].size() > 0)
{
for (int64_t nLoop = 0;; nLoop++)
{
ProcessOneShot();
2015-05-31 15:36:44 +02:00
BOOST_FOREACH(const std::string& strAddr, mapMultiArgs["-connect"])
{
CAddress addr(CService(), NODE_NONE);
OpenNetworkConnection(addr, false, NULL, strAddr.c_str());
for (int i = 0; i < 10 && i < nLoop; i++)
{
if (!interruptNet.sleep_for(std::chrono::milliseconds(500)))
return;
}
}
if (!interruptNet.sleep_for(std::chrono::milliseconds(500)))
return;
}
}
// Initiate network connections
int64_t nStart = GetTime();
// Minimum time before next feeler connection (in microseconds).
int64_t nNextFeeler = PoissonNextSend(nStart*1000*1000, FEELER_INTERVAL);
while (!interruptNet)
{
ProcessOneShot();
if (!interruptNet.sleep_for(std::chrono::milliseconds(500)))
return;
CSemaphoreGrant grant(*semOutbound);
if (interruptNet)
return;
// Add seed nodes if DNS seeds are all down (an infrastructure attack?).
if (addrman.size() == 0 && (GetTime() - nStart > 60)) {
static bool done = false;
if (!done) {
LogPrintf("Adding fixed seed nodes as DNS doesn't seem to be available.\n");
CNetAddr local;
LookupHost("127.0.0.1", local, false);
addrman.Add(convertSeed6(Params().FixedSeeds()), local);
done = true;
}
}
//
// Choose an address to connect to based on most recently seen
//
CAddress addrConnect;
// Only connect out to one peer per network group (/16 for IPv4).
// Do this here so we don't have to critsect vNodes inside mapAddresses critsect.
Implement named devnets (#1791) * Initial devnet * Move genesis block adding into its own method * Introduce -allowprivatenet to lift limitation on RFC1918 addresses Normally, RFC1918 (192.168.x.x/10.x.x.x/...) addresses are not allowed to be relayed. Also, masternodes won't start when the address is considered invalid. This is needed to test local devnet or regtest based networks. * Lift the requirement of minimum MN age for regtest/devnet * Implement named devnets This allows the creation of multiple independent devnets. Each one is identified by a name which is hardened into a "devnet genesis" block, which is automatically positioned at height 1. Validation rules will ensure that a node from devnet=test1 never be able to accept blocks from devnet=test2. This is done by checking the expected devnet genesis block. The genesis block of the devnet is the same as the one from regtest. This starts the devnet with a very low difficulty, allowing us to fill up needed balances for masternodes very fast. Also, the devnet name is put into the sub-version of the VERSION message. If a node connects to the wrong network, it will immediately be disconnected. * Allow to select multiple addresses from the same group in devnet/regtest The selection code normally only allows to select addresses from the same group (e.g. 192.168.x.x) once. This results in connecting to only a single node in devnet/regtest. * Show the devnet name in the title bar and on the loading screen * Add AllowMultipleAddressesFromGroup to chainparams and use it in net.cpp * Remove unused/unneeded scripts from devnet geneses creation 1. OP_RETURN not needed in input script of devnet genesis 2. genesisOutputScript was unused * Fix copy/paste error in -allowprivatenet description * Improve -devnet parameter error handling - Only allow one of -devnet, -regtest or -testnet - Only allow -devnet=name to be specified once * Use different datadir for each devnet * Fix `devnet-devnet` issue * Fix devnet splashscreen (should use testnet img) * Avoid passing devNetName around (most of the time) * Remove nMaxTipAge from CDevNetParams Not present anymore after rebase on develop
2017-12-20 12:45:01 +01:00
// This is only done for mainnet and testnet
int nOutbound = 0;
std::set<std::vector<unsigned char> > setConnected;
Implement named devnets (#1791) * Initial devnet * Move genesis block adding into its own method * Introduce -allowprivatenet to lift limitation on RFC1918 addresses Normally, RFC1918 (192.168.x.x/10.x.x.x/...) addresses are not allowed to be relayed. Also, masternodes won't start when the address is considered invalid. This is needed to test local devnet or regtest based networks. * Lift the requirement of minimum MN age for regtest/devnet * Implement named devnets This allows the creation of multiple independent devnets. Each one is identified by a name which is hardened into a "devnet genesis" block, which is automatically positioned at height 1. Validation rules will ensure that a node from devnet=test1 never be able to accept blocks from devnet=test2. This is done by checking the expected devnet genesis block. The genesis block of the devnet is the same as the one from regtest. This starts the devnet with a very low difficulty, allowing us to fill up needed balances for masternodes very fast. Also, the devnet name is put into the sub-version of the VERSION message. If a node connects to the wrong network, it will immediately be disconnected. * Allow to select multiple addresses from the same group in devnet/regtest The selection code normally only allows to select addresses from the same group (e.g. 192.168.x.x) once. This results in connecting to only a single node in devnet/regtest. * Show the devnet name in the title bar and on the loading screen * Add AllowMultipleAddressesFromGroup to chainparams and use it in net.cpp * Remove unused/unneeded scripts from devnet geneses creation 1. OP_RETURN not needed in input script of devnet genesis 2. genesisOutputScript was unused * Fix copy/paste error in -allowprivatenet description * Improve -devnet parameter error handling - Only allow one of -devnet, -regtest or -testnet - Only allow -devnet=name to be specified once * Use different datadir for each devnet * Fix `devnet-devnet` issue * Fix devnet splashscreen (should use testnet img) * Avoid passing devNetName around (most of the time) * Remove nMaxTipAge from CDevNetParams Not present anymore after rebase on develop
2017-12-20 12:45:01 +01:00
if (!Params().AllowMultipleAddressesFromGroup()) {
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes) {
if (!pnode->fInbound && !pnode->fMasternode) {
setConnected.insert(pnode->addr.GetGroup());
nOutbound++;
}
}
}
// Feeler Connections
//
// Design goals:
// * Increase the number of connectable addresses in the tried table.
//
// Method:
// * Choose a random address from new and attempt to connect to it if we can connect
// successfully it is added to tried.
// * Start attempting feeler connections only after node finishes making outbound
// connections.
// * Only make a feeler connection once every few minutes.
//
bool fFeeler = false;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (nOutbound >= nMaxOutbound) {
int64_t nTime = GetTimeMicros(); // The current time right now (in microseconds).
if (nTime > nNextFeeler) {
nNextFeeler = PoissonNextSend(nTime, FEELER_INTERVAL);
fFeeler = true;
} else {
continue;
}
}
int64_t nANow = GetAdjustedTime();
int nTries = 0;
while (!interruptNet)
{
CAddrInfo addr = addrman.Select(fFeeler);
// if we selected an invalid address, restart
if (!addr.IsValid() || setConnected.count(addr.GetGroup()) || IsLocal(addr))
break;
2012-08-21 17:32:04 +02:00
// If we didn't find an appropriate destination after trying 100 addresses fetched from addrman,
// stop this loop, and let the outer loop run again (which sleeps, adds seed nodes, recalculates
// already-connected network ranges, ...) before trying new addrman addresses.
nTries++;
2012-08-21 17:32:04 +02:00
if (nTries > 100)
break;
if (IsLimited(addr))
continue;
// only connect to full nodes
if ((addr.nServices & REQUIRED_SERVICES) != REQUIRED_SERVICES)
continue;
// only consider very recently tried nodes after 30 failed attempts
if (nANow - addr.nLastTry < 600 && nTries < 30)
continue;
// do not allow non-default ports, unless after 50 invalid addresses selected already
if (addr.GetPort() != Params().GetDefaultPort() && nTries < 50)
continue;
addrConnect = addr;
break;
}
if (addrConnect.IsValid()) {
if (fFeeler) {
// Add small amount of random noise before connection to avoid synchronization.
int randsleep = GetRandInt(FEELER_SLEEP_WINDOW * 1000);
if (!interruptNet.sleep_for(std::chrono::milliseconds(randsleep)))
return;
LogPrint("net", "Making feeler connection to %s\n", addrConnect.ToString());
}
OpenNetworkConnection(addrConnect, (int)setConnected.size() >= std::min(nMaxConnections - 1, 2), &grant, NULL, false, fFeeler);
}
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
std::vector<AddedNodeInfo> CConnman::GetAddedNodeInfo()
{
std::vector<AddedNodeInfo> ret;
std::list<std::string> lAddresses(0);
{
LOCK(cs_vAddedNodes);
ret.reserve(vAddedNodes.size());
BOOST_FOREACH(const std::string& strAddNode, vAddedNodes)
lAddresses.push_back(strAddNode);
}
// Build a map of all already connected addresses (by IP:port and by name) to inbound/outbound and resolved CService
std::map<CService, bool> mapConnected;
std::map<std::string, std::pair<bool, CService>> mapConnectedByName;
{
LOCK(cs_vNodes);
for (const CNode* pnode : vNodes) {
if (pnode->addr.IsValid()) {
mapConnected[pnode->addr] = pnode->fInbound;
}
if (!pnode->addrName.empty()) {
mapConnectedByName[pnode->addrName] = std::make_pair(pnode->fInbound, static_cast<const CService&>(pnode->addr));
}
}
}
BOOST_FOREACH(const std::string& strAddNode, lAddresses) {
CService service(LookupNumeric(strAddNode.c_str(), Params().GetDefaultPort()));
if (service.IsValid()) {
// strAddNode is an IP:port
auto it = mapConnected.find(service);
if (it != mapConnected.end()) {
ret.push_back(AddedNodeInfo{strAddNode, service, true, it->second});
} else {
ret.push_back(AddedNodeInfo{strAddNode, CService(), false, false});
}
} else {
// strAddNode is a name
auto it = mapConnectedByName.find(strAddNode);
if (it != mapConnectedByName.end()) {
ret.push_back(AddedNodeInfo{strAddNode, it->second.second, true, it->second.first});
} else {
ret.push_back(AddedNodeInfo{strAddNode, CService(), false, false});
}
}
}
return ret;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ThreadOpenAddedConnections()
{
{
LOCK(cs_vAddedNodes);
vAddedNodes = mapMultiArgs["-addnode"];
}
for (unsigned int i = 0; true; i++)
{
std::vector<AddedNodeInfo> vInfo = GetAddedNodeInfo();
for (const AddedNodeInfo& info : vInfo) {
if (!info.fConnected) {
CSemaphoreGrant grant(*semOutbound);
// If strAddedNode is an IP/port, decode it immediately, so
// OpenNetworkConnection can detect existing connections to that IP/port.
CService service(LookupNumeric(info.strAddedNode.c_str(), Params().GetDefaultPort()));
OpenNetworkConnection(CAddress(service, NODE_NONE), false, &grant, info.strAddedNode.c_str(), false);
if (!interruptNet.sleep_for(std::chrono::milliseconds(500)))
return;
}
}
if (!interruptNet.sleep_for(std::chrono::minutes(2)))
return;
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ThreadMnbRequestConnections()
{
// Connecting to specific addresses, no masternode connections available
if (mapArgs.count("-connect") && mapMultiArgs["-connect"].size() > 0)
return;
while (!interruptNet)
{
if (!interruptNet.sleep_for(std::chrono::milliseconds(1000)))
return;
CSemaphoreGrant grant(*semMasternodeOutbound);
if (interruptNet)
return;
std::pair<CService, std::set<uint256> > p = mnodeman.PopScheduledMnbRequestConnection();
if(p.first == CService() || p.second.empty()) continue;
ConnectNode(CAddress(p.first, NODE_NETWORK), NULL, false, true);
LOCK(cs_vNodes);
CNode *pnode = FindNode(p.first);
if(!pnode || pnode->fDisconnect) continue;
grant.MoveTo(pnode->grantMasternodeOutbound);
// compile request vector
std::vector<CInv> vToFetch;
std::set<uint256>::iterator it = p.second.begin();
while(it != p.second.end()) {
if(*it != uint256()) {
vToFetch.push_back(CInv(MSG_MASTERNODE_ANNOUNCE, *it));
LogPrint("masternode", "ThreadMnbRequestConnections -- asking for mnb %s from addr=%s\n", it->ToString(), p.first.ToString());
}
++it;
}
// ask for data
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
PushMessage(pnode, NetMsgType::GETDATA, vToFetch);
}
}
2012-07-26 02:48:39 +02:00
// if successful, this moves the passed grant to the constructed node
bool CConnman::OpenNetworkConnection(const CAddress& addrConnect, bool fCountFailure, CSemaphoreGrant *grantOutbound, const char *pszDest, bool fOneShot, bool fFeeler)
{
//
// Initiate outbound network connection
//
if (interruptNet) {
return false;
}
Backport Bitcoin Qt/Gui changes up to 0.14.x part 3 (#1617) * Merge #8996: Network activity toggle 19f46f1 Qt: New network_disabled icon (Luke Dashjr) 54cf997 RPC/Net: Use boolean consistently for networkactive, and remove from getinfo (Luke Dashjr) b2b33d9 Overhaul network activity toggle (Jonas Schnelli) 32efa79 Qt: Add GUI feedback and control of network activity state. (Jon Lund Steffensen) e38993b RPC: Add "togglenetwork" method to toggle network activity temporarily (Jon Lund Steffensen) 7c9a98a Allow network activity to be temporarily suspended. (Jon Lund Steffensen) * Revert on-click behavior of network status icon to showing peers list Stay with the way Dash handled clicking on the status icon * Add theme support for network disabled icon * Merge #8874: Multiple Selection for peer and ban tables 1077577 Fix auto-deselection of peers (Andrew Chow) addfdeb Multiple Selection for peer and ban tables (Andrew Chow) * Merge #9190: qt: Plug many memory leaks ed998ea qt: Avoid OpenSSL certstore-related memory leak (Wladimir J. van der Laan) 5204598 qt: Avoid shutdownwindow-related memory leak (Wladimir J. van der Laan) e4f126a qt: Avoid splash-screen related memory leak (Wladimir J. van der Laan) 693384e qt: Prevent thread/memory leak on exiting RPCConsole (Wladimir J. van der Laan) 47db075 qt: Plug many memory leaks (Wladimir J. van der Laan) * Merge #9218: qt: Show progress overlay when clicking spinner icon 042f9fa qt: Show progress overlay when clicking spinner icon (Wladimir J. van der Laan) 827d9a3 qt: Replace NetworkToggleStatusBarControl with generic ClickableLabel (Wladimir J. van der Laan) * Merge #9266: Bugfix: Qt/RPCConsole: Put column enum in the right places df17fe0 Bugfix: Qt/RPCConsole: Put column enum in the right places (Luke Dashjr) * Merge #9255: qt: layoutAboutToChange signal is called layoutAboutToBeChanged f36349e qt: Remove on_toggleNetworkActiveButton_clicked from RPCConsole (Wladimir J. van der Laan) 297cc20 qt: layoutAboutToChange signal is called layoutAboutToBeChanged (Wladimir J. van der Laan) * Use UniValue until bitcoin PR #8788 is backported Network active toggle was already based on "[RPC] Give RPC commands more information about the RPC request" We need to use the old UniValue style until that one is backported * Merge #8906: [qt] sync-overlay: Don't show progress twice fafeec3 [qt] sync-overlay: Don't show progress twice (MarcoFalke) * Merge #8985: Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() 3154d6e [Qt] use NotifyHeaderTip's height and date for the progress update (Jonas Schnelli) 0a261b6 Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() (Jonas Schnelli) * Merge #9280: [Qt] Show ModalOverlay by pressing the progress bar, allow hiding 89a3723 [Qt] Show ModalOverlay by pressing the progress bar, disabled show() in sync mode (Jonas Schnelli) * Merge #9461: [Qt] Improve progress display during headers-sync and peer-finding 40ec7c7 [Qt] Improve progress display during headers-sync and peer-finding (Jonas Schnelli) * Merge #9588: qt: Use nPowTargetSpacing constant fa4d478 qt: Use nPowTargetSpacing constant (MarcoFalke) * Hide modal overlay forever when syncing has catched up Don't allow to open it again by clicking on the progress bar and spinner icon. Currently the overlay does not show meaningful information about masternode sync and it gives the impression of being stuck after the block chain sync is done. * Don't include chainparams.h in sendcoinsdialog.cpp This was just a remainder of a backported PR which meant to change some calculation in this file which does not apply to Dash. * Also check for fNetworkActive in ConnectNode * Merge #9528: [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) 988d300 [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) (practicalswift) * Merge #11237: qt: Fixing division by zero in time remaining c8d38abd6 Refactor tipUpdate as per style guide (MeshCollider) 3b69a08c5 Fix division by zero in time remaining (MeshCollider) Pull request description: Fixes https://github.com/bitcoin/bitcoin/issues/10291, https://github.com/bitcoin/bitcoin/issues/11265 progressDelta may be 0 (or even negative according to 11265), this checks for that and prints unknown if it is, because we cannot calculate an estimate for the time remaining (would be infinite or negative). Tree-SHA512: bc5708e5ed6e4670d008219558c5fbb25709bd99a32c98ec39bb74f94a0b7fa058f3d03389ccdd39e6723e6b5b48e34b13ceee7c051c2db631e51d8ec3e1d68c
2017-09-11 15:38:14 +02:00
if (!fNetworkActive) {
return false;
}
if (!pszDest) {
2012-02-12 13:45:24 +01:00
if (IsLocal(addrConnect) ||
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
FindNode((CNetAddr)addrConnect) || IsBanned(addrConnect) ||
FindNode(addrConnect.ToStringIPPort()))
return false;
} else if (FindNode(std::string(pszDest)))
return false;
CNode* pnode = ConnectNode(addrConnect, pszDest, fCountFailure);
if (!pnode)
return false;
if (grantOutbound)
grantOutbound->MoveTo(pnode->grantOutbound);
if (fOneShot)
pnode->fOneShot = true;
if (fFeeler)
pnode->fFeeler = true;
return true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ThreadMessageHandler()
{
while (!flagInterruptMsgProc)
{
std::vector<CNode*> vNodesCopy = CopyNodeVector();
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
bool fMoreWork = false;
2013-11-15 12:24:34 +01:00
BOOST_FOREACH(CNode* pnode, vNodesCopy)
{
if (pnode->fDisconnect)
continue;
// Receive messages
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
bool fMoreNodeWork = GetNodeSignals().ProcessMessages(pnode, *this, flagInterruptMsgProc);
fMoreWork |= (fMoreNodeWork && !pnode->fPauseSend);
if (flagInterruptMsgProc)
return;
// Send messages
{
TRY_LOCK(pnode->cs_vSend, lockSend);
if (lockSend)
GetNodeSignals().SendMessages(pnode, *this, flagInterruptMsgProc);
}
if (flagInterruptMsgProc)
return;
}
ReleaseNodeVector(vNodesCopy);
2013-11-15 12:24:34 +01:00
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
std::unique_lock<std::mutex> lock(mutexMsgProc);
if (!fMoreWork) {
condMsgProc.wait_until(lock, std::chrono::steady_clock::now() + std::chrono::milliseconds(100), [this] { return fMsgProcWake; });
}
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
fMsgProcWake = false;
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::BindListenPort(const CService &addrBind, std::string& strError, bool fWhitelisted)
{
strError = "";
int nOne = 1;
// Create socket for listening for incoming connections
struct sockaddr_storage sockaddr;
socklen_t len = sizeof(sockaddr);
if (!addrBind.GetSockAddr((struct sockaddr*)&sockaddr, &len))
{
strError = strprintf("Error: Bind address family for %s not supported", addrBind.ToString());
LogPrintf("%s\n", strError);
return false;
}
SOCKET hListenSocket = socket(((struct sockaddr*)&sockaddr)->sa_family, SOCK_STREAM, IPPROTO_TCP);
if (hListenSocket == INVALID_SOCKET)
{
strError = strprintf("Error: Couldn't open socket for incoming connections (socket returned error %s)", NetworkErrorString(WSAGetLastError()));
LogPrintf("%s\n", strError);
return false;
}
if (!IsSelectableSocket(hListenSocket))
{
strError = "Error: Couldn't create a listenable socket for incoming connections";
LogPrintf("%s\n", strError);
return false;
}
#ifndef WIN32
#ifdef SO_NOSIGPIPE
// Different way of disabling SIGPIPE on BSD
setsockopt(hListenSocket, SOL_SOCKET, SO_NOSIGPIPE, (void*)&nOne, sizeof(int));
#endif
// Allow binding if the port is still in TIME_WAIT state after
// the program was closed and restarted.
setsockopt(hListenSocket, SOL_SOCKET, SO_REUSEADDR, (void*)&nOne, sizeof(int));
// Disable Nagle's algorithm
setsockopt(hListenSocket, IPPROTO_TCP, TCP_NODELAY, (void*)&nOne, sizeof(int));
#else
setsockopt(hListenSocket, SOL_SOCKET, SO_REUSEADDR, (const char*)&nOne, sizeof(int));
setsockopt(hListenSocket, IPPROTO_TCP, TCP_NODELAY, (const char*)&nOne, sizeof(int));
#endif
2012-07-26 02:48:39 +02:00
// Set to non-blocking, incoming connections will also inherit this
if (!SetSocketNonBlocking(hListenSocket, true)) {
strError = strprintf("BindListenPort: Setting listening socket to non-blocking failed, error %s\n", NetworkErrorString(WSAGetLastError()));
LogPrintf("%s\n", strError);
return false;
}
// some systems don't have IPV6_V6ONLY but are always v6only; others do have the option
// and enable it by default or not. Try to enable it, if possible.
if (addrBind.IsIPv6()) {
#ifdef IPV6_V6ONLY
#ifdef WIN32
setsockopt(hListenSocket, IPPROTO_IPV6, IPV6_V6ONLY, (const char*)&nOne, sizeof(int));
#else
setsockopt(hListenSocket, IPPROTO_IPV6, IPV6_V6ONLY, (void*)&nOne, sizeof(int));
#endif
#endif
#ifdef WIN32
int nProtLevel = PROTECTION_LEVEL_UNRESTRICTED;
setsockopt(hListenSocket, IPPROTO_IPV6, IPV6_PROTECTION_LEVEL, (const char*)&nProtLevel, sizeof(int));
#endif
}
if (::bind(hListenSocket, (struct sockaddr*)&sockaddr, len) == SOCKET_ERROR)
{
int nErr = WSAGetLastError();
if (nErr == WSAEADDRINUSE)
Merge #7192: Unify product name to as few places as possible 027fdb8 When/if the copyright line does not mention Bitcoin Core developers, add a second line to copyrights in -version, About dialog, and splash screen (Luke Dashjr) cc2095e Rewrite FormatParagraph to handle newlines within input strings correctly (Luke Dashjr) cddffaf Bugfix: Include COPYRIGHT_HOLDERS_SUBSTITUTION in Makefile substitutions so it gets passed to extract-strings correctly (Luke Dashjr) 29598e4 Move PACKAGE_URL to configure.ac (Luke Dashjr) 78ec83d splashscreen: Resize text to fit exactly (Luke Dashjr) 3cae140 Bugfix: Actually use _COPYRIGHT_HOLDERS_SUBSTITUTION everywhere (Luke Dashjr) 4d5a3df Bugfix: gitian-descriptors: Add missing python-setuptools requirement for OS X (biplist module) (Luke Dashjr) e4ab5e5 Bugfix: Correct copyright year in Mac DMG background image (Luke Dashjr) 917b1d0 Set copyright holders displayed in notices separately from the package name (Luke Dashjr) c39a6ff Travis & gitian-osx: Use depends for ds_store and mac_alias modules (Luke Dashjr) 902ccde depends: Add mac_alias to depends (Luke Dashjr) 82a2d98 depends: Add ds_store to depends (Cory Fields) de619a3 depends: Pass PYTHONPATH along to configure (Cory Fields) e611b6e macdeploy: Use rsvg-convert rather than cairosvg (Luke Dashjr) 63bcdc5 More complicated package name substitution for Mac deployment (Luke Dashjr) 1a6c67c Parameterise 2009 in translatable copyright strings (Luke Dashjr) d5f4683 Unify package name to as few places as possible without major changes (Luke Dashjr)
2016-02-04 13:41:58 +01:00
strError = strprintf(_("Unable to bind to %s on this computer. %s is probably already running."), addrBind.ToString(), _(PACKAGE_NAME));
else
strError = strprintf(_("Unable to bind to %s on this computer (bind returned error %s)"), addrBind.ToString(), NetworkErrorString(nErr));
LogPrintf("%s\n", strError);
CloseSocket(hListenSocket);
return false;
}
LogPrintf("Bound to %s\n", addrBind.ToString());
// Listen for incoming connections
if (listen(hListenSocket, SOMAXCONN) == SOCKET_ERROR)
{
strError = strprintf(_("Error: Listening for incoming connections failed (listen returned error %s)"), NetworkErrorString(WSAGetLastError()));
LogPrintf("%s\n", strError);
CloseSocket(hListenSocket);
return false;
}
vhListenSocket.push_back(ListenSocket(hListenSocket, fWhitelisted));
if (addrBind.IsRoutable() && fDiscover && !fWhitelisted)
AddLocal(addrBind, LOCAL_BIND);
return true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void Discover(boost::thread_group& threadGroup)
{
2012-05-24 19:02:21 +02:00
if (!fDiscover)
return;
#ifdef WIN32
2012-07-26 02:48:39 +02:00
// Get local host IP
char pszHostName[256] = "";
if (gethostname(pszHostName, sizeof(pszHostName)) != SOCKET_ERROR)
{
std::vector<CNetAddr> vaddr;
if (LookupHost(pszHostName, vaddr, 0, true))
{
BOOST_FOREACH (const CNetAddr &addr, vaddr)
{
if (AddLocal(addr, LOCAL_IF))
LogPrintf("%s: %s - %s\n", __func__, pszHostName, addr.ToString());
}
}
}
#else
// Get local host ip
struct ifaddrs* myaddrs;
if (getifaddrs(&myaddrs) == 0)
{
for (struct ifaddrs* ifa = myaddrs; ifa != NULL; ifa = ifa->ifa_next)
{
if (ifa->ifa_addr == NULL) continue;
if ((ifa->ifa_flags & IFF_UP) == 0) continue;
if (strcmp(ifa->ifa_name, "lo") == 0) continue;
if (strcmp(ifa->ifa_name, "lo0") == 0) continue;
if (ifa->ifa_addr->sa_family == AF_INET)
{
struct sockaddr_in* s4 = (struct sockaddr_in*)(ifa->ifa_addr);
2012-02-12 13:45:24 +01:00
CNetAddr addr(s4->sin_addr);
if (AddLocal(addr, LOCAL_IF))
LogPrintf("%s: IPv4 %s: %s\n", __func__, ifa->ifa_name, addr.ToString());
}
else if (ifa->ifa_addr->sa_family == AF_INET6)
{
struct sockaddr_in6* s6 = (struct sockaddr_in6*)(ifa->ifa_addr);
2012-02-12 13:45:24 +01:00
CNetAddr addr(s6->sin6_addr);
if (AddLocal(addr, LOCAL_IF))
LogPrintf("%s: IPv6 %s: %s\n", __func__, ifa->ifa_name, addr.ToString());
}
}
freeifaddrs(myaddrs);
}
#endif
}
Backport Bitcoin Qt/Gui changes up to 0.14.x part 3 (#1617) * Merge #8996: Network activity toggle 19f46f1 Qt: New network_disabled icon (Luke Dashjr) 54cf997 RPC/Net: Use boolean consistently for networkactive, and remove from getinfo (Luke Dashjr) b2b33d9 Overhaul network activity toggle (Jonas Schnelli) 32efa79 Qt: Add GUI feedback and control of network activity state. (Jon Lund Steffensen) e38993b RPC: Add "togglenetwork" method to toggle network activity temporarily (Jon Lund Steffensen) 7c9a98a Allow network activity to be temporarily suspended. (Jon Lund Steffensen) * Revert on-click behavior of network status icon to showing peers list Stay with the way Dash handled clicking on the status icon * Add theme support for network disabled icon * Merge #8874: Multiple Selection for peer and ban tables 1077577 Fix auto-deselection of peers (Andrew Chow) addfdeb Multiple Selection for peer and ban tables (Andrew Chow) * Merge #9190: qt: Plug many memory leaks ed998ea qt: Avoid OpenSSL certstore-related memory leak (Wladimir J. van der Laan) 5204598 qt: Avoid shutdownwindow-related memory leak (Wladimir J. van der Laan) e4f126a qt: Avoid splash-screen related memory leak (Wladimir J. van der Laan) 693384e qt: Prevent thread/memory leak on exiting RPCConsole (Wladimir J. van der Laan) 47db075 qt: Plug many memory leaks (Wladimir J. van der Laan) * Merge #9218: qt: Show progress overlay when clicking spinner icon 042f9fa qt: Show progress overlay when clicking spinner icon (Wladimir J. van der Laan) 827d9a3 qt: Replace NetworkToggleStatusBarControl with generic ClickableLabel (Wladimir J. van der Laan) * Merge #9266: Bugfix: Qt/RPCConsole: Put column enum in the right places df17fe0 Bugfix: Qt/RPCConsole: Put column enum in the right places (Luke Dashjr) * Merge #9255: qt: layoutAboutToChange signal is called layoutAboutToBeChanged f36349e qt: Remove on_toggleNetworkActiveButton_clicked from RPCConsole (Wladimir J. van der Laan) 297cc20 qt: layoutAboutToChange signal is called layoutAboutToBeChanged (Wladimir J. van der Laan) * Use UniValue until bitcoin PR #8788 is backported Network active toggle was already based on "[RPC] Give RPC commands more information about the RPC request" We need to use the old UniValue style until that one is backported * Merge #8906: [qt] sync-overlay: Don't show progress twice fafeec3 [qt] sync-overlay: Don't show progress twice (MarcoFalke) * Merge #8985: Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() 3154d6e [Qt] use NotifyHeaderTip's height and date for the progress update (Jonas Schnelli) 0a261b6 Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() (Jonas Schnelli) * Merge #9280: [Qt] Show ModalOverlay by pressing the progress bar, allow hiding 89a3723 [Qt] Show ModalOverlay by pressing the progress bar, disabled show() in sync mode (Jonas Schnelli) * Merge #9461: [Qt] Improve progress display during headers-sync and peer-finding 40ec7c7 [Qt] Improve progress display during headers-sync and peer-finding (Jonas Schnelli) * Merge #9588: qt: Use nPowTargetSpacing constant fa4d478 qt: Use nPowTargetSpacing constant (MarcoFalke) * Hide modal overlay forever when syncing has catched up Don't allow to open it again by clicking on the progress bar and spinner icon. Currently the overlay does not show meaningful information about masternode sync and it gives the impression of being stuck after the block chain sync is done. * Don't include chainparams.h in sendcoinsdialog.cpp This was just a remainder of a backported PR which meant to change some calculation in this file which does not apply to Dash. * Also check for fNetworkActive in ConnectNode * Merge #9528: [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) 988d300 [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) (practicalswift) * Merge #11237: qt: Fixing division by zero in time remaining c8d38abd6 Refactor tipUpdate as per style guide (MeshCollider) 3b69a08c5 Fix division by zero in time remaining (MeshCollider) Pull request description: Fixes https://github.com/bitcoin/bitcoin/issues/10291, https://github.com/bitcoin/bitcoin/issues/11265 progressDelta may be 0 (or even negative according to 11265), this checks for that and prints unknown if it is, because we cannot calculate an estimate for the time remaining (would be infinite or negative). Tree-SHA512: bc5708e5ed6e4670d008219558c5fbb25709bd99a32c98ec39bb74f94a0b7fa058f3d03389ccdd39e6723e6b5b48e34b13ceee7c051c2db631e51d8ec3e1d68c
2017-09-11 15:38:14 +02:00
void CConnman::SetNetworkActive(bool active)
{
if (fDebug) {
LogPrint("net", "SetNetworkActive: %s\n", active);
}
if (!active) {
fNetworkActive = false;
LOCK(cs_vNodes);
// Close sockets to all nodes
BOOST_FOREACH(CNode* pnode, vNodes) {
pnode->CloseSocketDisconnect();
}
} else {
fNetworkActive = true;
}
uiInterface.NotifyNetworkActiveChanged(fNetworkActive);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CConnman::CConnman()
{
Backport Bitcoin Qt/Gui changes up to 0.14.x part 3 (#1617) * Merge #8996: Network activity toggle 19f46f1 Qt: New network_disabled icon (Luke Dashjr) 54cf997 RPC/Net: Use boolean consistently for networkactive, and remove from getinfo (Luke Dashjr) b2b33d9 Overhaul network activity toggle (Jonas Schnelli) 32efa79 Qt: Add GUI feedback and control of network activity state. (Jon Lund Steffensen) e38993b RPC: Add "togglenetwork" method to toggle network activity temporarily (Jon Lund Steffensen) 7c9a98a Allow network activity to be temporarily suspended. (Jon Lund Steffensen) * Revert on-click behavior of network status icon to showing peers list Stay with the way Dash handled clicking on the status icon * Add theme support for network disabled icon * Merge #8874: Multiple Selection for peer and ban tables 1077577 Fix auto-deselection of peers (Andrew Chow) addfdeb Multiple Selection for peer and ban tables (Andrew Chow) * Merge #9190: qt: Plug many memory leaks ed998ea qt: Avoid OpenSSL certstore-related memory leak (Wladimir J. van der Laan) 5204598 qt: Avoid shutdownwindow-related memory leak (Wladimir J. van der Laan) e4f126a qt: Avoid splash-screen related memory leak (Wladimir J. van der Laan) 693384e qt: Prevent thread/memory leak on exiting RPCConsole (Wladimir J. van der Laan) 47db075 qt: Plug many memory leaks (Wladimir J. van der Laan) * Merge #9218: qt: Show progress overlay when clicking spinner icon 042f9fa qt: Show progress overlay when clicking spinner icon (Wladimir J. van der Laan) 827d9a3 qt: Replace NetworkToggleStatusBarControl with generic ClickableLabel (Wladimir J. van der Laan) * Merge #9266: Bugfix: Qt/RPCConsole: Put column enum in the right places df17fe0 Bugfix: Qt/RPCConsole: Put column enum in the right places (Luke Dashjr) * Merge #9255: qt: layoutAboutToChange signal is called layoutAboutToBeChanged f36349e qt: Remove on_toggleNetworkActiveButton_clicked from RPCConsole (Wladimir J. van der Laan) 297cc20 qt: layoutAboutToChange signal is called layoutAboutToBeChanged (Wladimir J. van der Laan) * Use UniValue until bitcoin PR #8788 is backported Network active toggle was already based on "[RPC] Give RPC commands more information about the RPC request" We need to use the old UniValue style until that one is backported * Merge #8906: [qt] sync-overlay: Don't show progress twice fafeec3 [qt] sync-overlay: Don't show progress twice (MarcoFalke) * Merge #8985: Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() 3154d6e [Qt] use NotifyHeaderTip's height and date for the progress update (Jonas Schnelli) 0a261b6 Use pindexBestHeader instead of setBlockIndexCandidates for NotifyHeaderTip() (Jonas Schnelli) * Merge #9280: [Qt] Show ModalOverlay by pressing the progress bar, allow hiding 89a3723 [Qt] Show ModalOverlay by pressing the progress bar, disabled show() in sync mode (Jonas Schnelli) * Merge #9461: [Qt] Improve progress display during headers-sync and peer-finding 40ec7c7 [Qt] Improve progress display during headers-sync and peer-finding (Jonas Schnelli) * Merge #9588: qt: Use nPowTargetSpacing constant fa4d478 qt: Use nPowTargetSpacing constant (MarcoFalke) * Hide modal overlay forever when syncing has catched up Don't allow to open it again by clicking on the progress bar and spinner icon. Currently the overlay does not show meaningful information about masternode sync and it gives the impression of being stuck after the block chain sync is done. * Don't include chainparams.h in sendcoinsdialog.cpp This was just a remainder of a backported PR which meant to change some calculation in this file which does not apply to Dash. * Also check for fNetworkActive in ConnectNode * Merge #9528: [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) 988d300 [qt] Rename formateNiceTimeOffset(qint64) to formatNiceTimeOffset(qint64) (practicalswift) * Merge #11237: qt: Fixing division by zero in time remaining c8d38abd6 Refactor tipUpdate as per style guide (MeshCollider) 3b69a08c5 Fix division by zero in time remaining (MeshCollider) Pull request description: Fixes https://github.com/bitcoin/bitcoin/issues/10291, https://github.com/bitcoin/bitcoin/issues/11265 progressDelta may be 0 (or even negative according to 11265), this checks for that and prints unknown if it is, because we cannot calculate an estimate for the time remaining (would be infinite or negative). Tree-SHA512: bc5708e5ed6e4670d008219558c5fbb25709bd99a32c98ec39bb74f94a0b7fa058f3d03389ccdd39e6723e6b5b48e34b13ceee7c051c2db631e51d8ec3e1d68c
2017-09-11 15:38:14 +02:00
fNetworkActive = true;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
setBannedIsDirty = false;
fAddressesInitialized = false;
nLastNodeId = 0;
nSendBufferMaxSize = 0;
nReceiveFloodSize = 0;
semOutbound = NULL;
semMasternodeOutbound = NULL;
nMaxConnections = 0;
nMaxOutbound = 0;
nBestHeight = 0;
clientInterface = NULL;
flagInterruptMsgProc = false;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
}
NodeId CConnman::GetNewNodeId()
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
return nLastNodeId.fetch_add(1, std::memory_order_relaxed);
}
bool CConnman::Start(CScheduler& scheduler, std::string& strNodeError, Options connOptions)
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
{
nTotalBytesRecv = 0;
nTotalBytesSent = 0;
nMaxOutboundLimit = 0;
nMaxOutboundTotalBytesSentInCycle = 0;
nMaxOutboundTimeframe = 60*60*24; //1 day
nMaxOutboundCycleStartTime = 0;
nRelevantServices = connOptions.nRelevantServices;
nLocalServices = connOptions.nLocalServices;
nMaxConnections = connOptions.nMaxConnections;
nMaxOutbound = std::min((connOptions.nMaxOutbound), nMaxConnections);
nMaxFeeler = connOptions.nMaxFeeler;
nSendBufferMaxSize = connOptions.nSendBufferMaxSize;
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
nReceiveFloodSize = connOptions.nReceiveFloodSize;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
SetBestHeight(connOptions.nBestHeight);
clientInterface = connOptions.uiInterface;
if (clientInterface)
clientInterface->InitMessage(_("Loading addresses..."));
// Load addresses from peers.dat
int64_t nStart = GetTimeMillis();
{
CAddrDB adb;
if (adb.Read(addrman))
LogPrintf("Loaded %i addresses from peers.dat %dms\n", addrman.size(), GetTimeMillis() - nStart);
else {
addrman.Clear(); // Addrman can be in an inconsistent state after failure, reset it
LogPrintf("Invalid or missing peers.dat; recreating\n");
DumpAddresses();
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (clientInterface)
clientInterface->InitMessage(_("Loading banlist..."));
// Load addresses from banlist.dat
nStart = GetTimeMillis();
2015-06-19 15:27:37 +02:00
CBanDB bandb;
banmap_t banmap;
if (bandb.Read(banmap)) {
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
SetBanned(banmap); // thread save setter
SetBannedSetDirty(false); // no need to write down, just read data
SweepBanned(); // sweep out unused entries
LogPrint("net", "Loaded %d banned node ips/subnets from banlist.dat %dms\n",
banmap.size(), GetTimeMillis() - nStart);
} else {
2015-06-19 15:27:37 +02:00
LogPrintf("Invalid or missing banlist.dat; recreating\n");
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
SetBannedSetDirty(true); // force write
DumpBanlist();
}
2015-06-19 15:27:37 +02:00
fAddressesInitialized = true;
if (semOutbound == NULL) {
// initialize semaphore
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
semOutbound = new CSemaphore(std::min((nMaxOutbound + nMaxFeeler), nMaxConnections));
}
if (semMasternodeOutbound == NULL) {
// initialize semaphore
semMasternodeOutbound = new CSemaphore(MAX_OUTBOUND_MASTERNODE_CONNECTIONS);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (pnodeLocalHost == NULL) {
CNetAddr local;
LookupHost("127.0.0.1", local, false);
pnodeLocalHost = new CNode(GetNewNodeId(), nLocalServices, GetBestHeight(), INVALID_SOCKET, CAddress(CService(local, 0), nLocalServices));
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
GetNodeSignals().InitializeNode(pnodeLocalHost, *this);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
}
//
// Start threads
//
InterruptSocks5(false);
interruptNet.reset();
flagInterruptMsgProc = false;
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
{
std::unique_lock<std::mutex> lock(mutexMsgProc);
fMsgProcWake = false;
}
// Send and receive from sockets, accept connections
threadSocketHandler = std::thread(&TraceThread<std::function<void()> >, "net", std::function<void()>(std::bind(&CConnman::ThreadSocketHandler, this)));
if (!GetBoolArg("-dnsseed", true))
LogPrintf("DNS seeding disabled\n");
2011-11-21 18:25:00 +01:00
else
threadDNSAddressSeed = std::thread(&TraceThread<std::function<void()> >, "dnsseed", std::function<void()>(std::bind(&CConnman::ThreadDNSAddressSeed, this)));
// Initiate outbound connections from -addnode
threadOpenAddedConnections = std::thread(&TraceThread<std::function<void()> >, "addcon", std::function<void()>(std::bind(&CConnman::ThreadOpenAddedConnections, this)));
// Initiate outbound connections
threadOpenConnections = std::thread(&TraceThread<std::function<void()> >, "opencon", std::function<void()>(std::bind(&CConnman::ThreadOpenConnections, this)));
// Initiate masternode connections
threadMnbRequestConnections = std::thread(&TraceThread<std::function<void()> >, "mnbcon", std::function<void()>(std::bind(&CConnman::ThreadMnbRequestConnections, this)));
// Process messages
threadMessageHandler = std::thread(&TraceThread<std::function<void()> >, "msghand", std::function<void()>(std::bind(&CConnman::ThreadMessageHandler, this)));
// Dump network addresses
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
scheduler.scheduleEvery(boost::bind(&CConnman::DumpData, this), DUMP_ADDRESSES_INTERVAL);
return true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
class CNetCleanup
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
public:
CNetCleanup() {}
~CNetCleanup()
{
#ifdef WIN32
// Shutdown Windows Sockets
WSACleanup();
#endif
}
}
instance_of_cnetcleanup;
void CExplicitNetCleanup::callCleanup()
{
// Explicit call to destructor of CNetCleanup because it's not implicitly called
// when the wallet is restarted from within the wallet itself.
CNetCleanup *tmp = new CNetCleanup();
delete tmp; // Stroustrup's gonna kill me for that
}
void CConnman::Interrupt()
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
{
{
std::lock_guard<std::mutex> lock(mutexMsgProc);
flagInterruptMsgProc = true;
}
condMsgProc.notify_all();
interruptNet();
InterruptSocks5(true);
if (semOutbound)
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
for (int i=0; i<(nMaxOutbound + nMaxFeeler); i++)
semOutbound->post();
}
void CConnman::Stop()
{
if (threadMessageHandler.joinable())
threadMessageHandler.join();
if (threadMnbRequestConnections.joinable())
threadMnbRequestConnections.join();
if (threadOpenConnections.joinable())
threadOpenConnections.join();
if (threadOpenAddedConnections.joinable())
threadOpenAddedConnections.join();
if (threadDNSAddressSeed.joinable())
threadDNSAddressSeed.join();
if (threadSocketHandler.joinable())
threadSocketHandler.join();
if (semMasternodeOutbound)
for (int i=0; i<MAX_OUTBOUND_MASTERNODE_CONNECTIONS; i++)
semMasternodeOutbound->post();
if (fAddressesInitialized)
{
2015-06-19 15:27:37 +02:00
DumpData();
fAddressesInitialized = false;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
// Close sockets
BOOST_FOREACH(CNode* pnode, vNodes)
if (pnode->hSocket != INVALID_SOCKET)
CloseSocket(pnode->hSocket);
BOOST_FOREACH(ListenSocket& hListenSocket, vhListenSocket)
if (hListenSocket.socket != INVALID_SOCKET)
if (!CloseSocket(hListenSocket.socket))
LogPrintf("CloseSocket(hListenSocket) failed with error %s\n", NetworkErrorString(WSAGetLastError()));
// clean up some globals (to help leak detection)
BOOST_FOREACH(CNode *pnode, vNodes) {
DeleteNode(pnode);
}
BOOST_FOREACH(CNode *pnode, vNodesDisconnected) {
DeleteNode(pnode);
}
vNodes.clear();
vNodesDisconnected.clear();
vhListenSocket.clear();
delete semOutbound;
semOutbound = NULL;
delete semMasternodeOutbound;
semMasternodeOutbound = NULL;
if(pnodeLocalHost)
DeleteNode(pnodeLocalHost);
pnodeLocalHost = NULL;
}
void CConnman::DeleteNode(CNode* pnode)
{
assert(pnode);
bool fUpdateConnectionTime = false;
GetNodeSignals().FinalizeNode(pnode->GetId(), fUpdateConnectionTime);
if(fUpdateConnectionTime)
addrman.Connected(pnode->addr);
delete pnode;
}
CConnman::~CConnman()
{
Interrupt();
Stop();
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
}
size_t CConnman::GetAddressCount() const
{
return addrman.size();
}
void CConnman::SetServices(const CService &addr, ServiceFlags nServices)
{
addrman.SetServices(addr, nServices);
}
void CConnman::MarkAddressGood(const CAddress& addr)
{
addrman.Good(addr);
}
void CConnman::AddNewAddress(const CAddress& addr, const CAddress& addrFrom, int64_t nTimePenalty)
{
addrman.Add(addr, addrFrom, nTimePenalty);
}
void CConnman::AddNewAddresses(const std::vector<CAddress>& vAddr, const CAddress& addrFrom, int64_t nTimePenalty)
{
addrman.Add(vAddr, addrFrom, nTimePenalty);
}
std::vector<CAddress> CConnman::GetAddresses()
{
return addrman.GetAddr();
}
bool CConnman::AddNode(const std::string& strNode)
{
LOCK(cs_vAddedNodes);
for(std::vector<std::string>::const_iterator it = vAddedNodes.begin(); it != vAddedNodes.end(); ++it) {
if (strNode == *it)
return false;
}
vAddedNodes.push_back(strNode);
return true;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::RemoveAddedNode(const std::string& strNode)
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
LOCK(cs_vAddedNodes);
for(std::vector<std::string>::iterator it = vAddedNodes.begin(); it != vAddedNodes.end(); ++it) {
if (strNode == *it) {
vAddedNodes.erase(it);
return true;
}
}
return false;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
size_t CConnman::GetNodeCount(NumConnections flags)
{
LOCK(cs_vNodes);
if (flags == CConnman::CONNECTIONS_ALL) // Shortcut if we want total
return vNodes.size();
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
int nNum = 0;
for(std::vector<CNode*>::const_iterator it = vNodes.begin(); it != vNodes.end(); ++it)
if (flags & ((*it)->fInbound ? CONNECTIONS_IN : CONNECTIONS_OUT))
nNum++;
return nNum;
}
void CConnman::GetNodeStats(std::vector<CNodeStats>& vstats)
{
vstats.clear();
LOCK(cs_vNodes);
vstats.reserve(vNodes.size());
for(std::vector<CNode*>::iterator it = vNodes.begin(); it != vNodes.end(); ++it) {
CNode* pnode = *it;
CNodeStats stats;
pnode->copyStats(stats);
vstats.push_back(stats);
}
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::DisconnectNode(const std::string& strNode)
{
LOCK(cs_vNodes);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
if (CNode* pnode = FindNode(strNode)) {
pnode->fDisconnect = true;
return true;
}
return false;
}
bool CConnman::DisconnectNode(NodeId id)
{
LOCK(cs_vNodes);
for(CNode* pnode : vNodes) {
if (id == pnode->id) {
pnode->fDisconnect = true;
return true;
}
}
return false;
2015-05-25 22:59:38 +02:00
}
void CConnman::RelayTransaction(const CTransaction& tx)
{
uint256 hash = tx.GetHash();
int nInv = static_cast<bool>(CPrivateSend::GetDSTX(hash)) ? MSG_DSTX :
(instantsend.HasTxLockRequest(hash) ? MSG_TXLOCK_REQUEST : MSG_TX);
CInv inv(nInv, hash);
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
{
pnode->PushInventory(inv);
}
}
2013-08-22 18:09:32 +02:00
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::RelayInv(CInv &inv, const int minProtoVersion) {
2015-07-08 02:37:23 +02:00
LOCK(cs_vNodes);
BOOST_FOREACH(CNode* pnode, vNodes)
if(pnode->nVersion >= minProtoVersion)
pnode->PushInventory(inv);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::RecordBytesRecv(uint64_t bytes)
2013-08-22 18:09:32 +02:00
{
LOCK(cs_totalBytesRecv);
nTotalBytesRecv += bytes;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::RecordBytesSent(uint64_t bytes)
2013-08-22 18:09:32 +02:00
{
LOCK(cs_totalBytesSent);
nTotalBytesSent += bytes;
uint64_t now = GetTime();
if (nMaxOutboundCycleStartTime + nMaxOutboundTimeframe < now)
{
// timeframe expired, reset cycle
nMaxOutboundCycleStartTime = now;
nMaxOutboundTotalBytesSentInCycle = 0;
}
// TODO, exclude whitebind peers
nMaxOutboundTotalBytesSentInCycle += bytes;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::SetMaxOutboundTarget(uint64_t limit)
{
LOCK(cs_totalBytesSent);
uint64_t recommendedMinimum = (nMaxOutboundTimeframe / 600) * MaxBlockSize(true);
nMaxOutboundLimit = limit;
if (limit > 0 && limit < recommendedMinimum)
LogPrintf("Max outbound target is very small (%s bytes) and will be overshot. Recommended minimum is %s bytes.\n", nMaxOutboundLimit, recommendedMinimum);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
uint64_t CConnman::GetMaxOutboundTarget()
{
LOCK(cs_totalBytesSent);
return nMaxOutboundLimit;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
uint64_t CConnman::GetMaxOutboundTimeframe()
{
LOCK(cs_totalBytesSent);
return nMaxOutboundTimeframe;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
uint64_t CConnman::GetMaxOutboundTimeLeftInCycle()
{
LOCK(cs_totalBytesSent);
if (nMaxOutboundLimit == 0)
return 0;
if (nMaxOutboundCycleStartTime == 0)
return nMaxOutboundTimeframe;
uint64_t cycleEndTime = nMaxOutboundCycleStartTime + nMaxOutboundTimeframe;
uint64_t now = GetTime();
return (cycleEndTime < now) ? 0 : cycleEndTime - GetTime();
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::SetMaxOutboundTimeframe(uint64_t timeframe)
{
LOCK(cs_totalBytesSent);
if (nMaxOutboundTimeframe != timeframe)
{
// reset measure-cycle in case of changing
// the timeframe
nMaxOutboundCycleStartTime = GetTime();
}
nMaxOutboundTimeframe = timeframe;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
bool CConnman::OutboundTargetReached(bool historicalBlockServingLimit)
{
LOCK(cs_totalBytesSent);
if (nMaxOutboundLimit == 0)
return false;
if (historicalBlockServingLimit)
{
2017-09-07 17:59:00 +02:00
// keep a large enough buffer to at least relay each block once
uint64_t timeLeftInCycle = GetMaxOutboundTimeLeftInCycle();
uint64_t buffer = timeLeftInCycle / 600 * MaxBlockSize(fDIP0001ActiveAtTip);
if (buffer >= nMaxOutboundLimit || nMaxOutboundTotalBytesSentInCycle >= nMaxOutboundLimit - buffer)
return true;
}
else if (nMaxOutboundTotalBytesSentInCycle >= nMaxOutboundLimit)
return true;
return false;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
uint64_t CConnman::GetOutboundTargetBytesLeft()
{
LOCK(cs_totalBytesSent);
if (nMaxOutboundLimit == 0)
return 0;
return (nMaxOutboundTotalBytesSentInCycle >= nMaxOutboundLimit) ? 0 : nMaxOutboundLimit - nMaxOutboundTotalBytesSentInCycle;
2013-08-22 18:09:32 +02:00
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
uint64_t CConnman::GetTotalBytesRecv()
2013-08-22 18:09:32 +02:00
{
LOCK(cs_totalBytesRecv);
return nTotalBytesRecv;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
uint64_t CConnman::GetTotalBytesSent()
2013-08-22 18:09:32 +02:00
{
LOCK(cs_totalBytesSent);
return nTotalBytesSent;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
ServiceFlags CConnman::GetLocalServices() const
{
return nLocalServices;
}
void CConnman::SetBestHeight(int height)
{
nBestHeight.store(height, std::memory_order_release);
}
int CConnman::GetBestHeight() const
{
return nBestHeight.load(std::memory_order_acquire);
}
unsigned int CConnman::GetReceiveFloodSize() const { return nReceiveFloodSize; }
unsigned int CConnman::GetSendBufferSize() const{ return nSendBufferMaxSize; }
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode::CNode(NodeId idIn, ServiceFlags nLocalServicesIn, int nMyStartingHeightIn, SOCKET hSocketIn, const CAddress& addrIn, const std::string& addrNameIn, bool fInboundIn, bool fNetworkNodeIn) :
addrKnown(5000, 0.001),
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
filterInventoryKnown(50000, 0.000001),
nSendVersion(0)
{
nServices = NODE_NONE;
nServicesExpected = NODE_NONE;
hSocket = hSocketIn;
nRecvVersion = INIT_PROTO_VERSION;
nLastSend = 0;
nLastRecv = 0;
nSendBytes = 0;
nRecvBytes = 0;
nTimeConnected = GetSystemTimeInSeconds();
2014-12-15 11:06:15 +01:00
nTimeOffset = 0;
addr = addrIn;
addrName = addrNameIn == "" ? addr.ToStringIPPort() : addrNameIn;
nVersion = 0;
nNumWarningsSkipped = 0;
nLastWarningTime = 0;
strSubVer = "";
fWhitelisted = false;
fOneShot = false;
fClient = false; // set by version message
fFeeler = false;
fInbound = fInboundIn;
fNetworkNode = fNetworkNodeIn;
fSuccessfullyConnected = false;
fDisconnect = false;
nRefCount = 0;
nSendSize = 0;
nSendOffset = 0;
hashContinue = uint256();
nStartingHeight = -1;
filterInventoryKnown.reset();
fSendMempool = false;
fGetAddr = false;
nNextLocalAddrSend = 0;
nNextAddrSend = 0;
nNextInvSend = 0;
fRelayTxes = false;
fSentAddr = false;
pfilter = new CBloomFilter();
timeLastMempoolReq = 0;
nLastBlockTime = 0;
nLastTXTime = 0;
nPingNonceSent = 0;
nPingUsecStart = 0;
nPingUsecTime = 0;
fPingQueued = false;
fMasternode = false;
nMinPingUsecTime = std::numeric_limits<int64_t>::max();
minFeeFilter = 0;
lastSentFeeFilter = 0;
nextSendTimeFeeFilter = 0;
vchKeyedNetGroup = CalculateKeyedNetGroup(addr);
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
id = idIn;
nLocalServices = nLocalServicesIn;
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
fPauseRecv = false;
fPauseSend = false;
nProcessQueueSize = 0;
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
GetRandBytes((unsigned char*)&nLocalHostNonce, sizeof(nLocalHostNonce));
nMyStartingHeight = nMyStartingHeightIn;
BOOST_FOREACH(const std::string &msg, getAllNetMessageTypes())
mapRecvBytesPerMsgCmd[msg] = 0;
mapRecvBytesPerMsgCmd[NET_MESSAGE_COMMAND_OTHER] = 0;
if(fNetworkNode || fInbound)
AddRef();
if (fLogIPs)
LogPrint("net", "Added connection to %s peer=%d\n", addrName, id);
else
LogPrint("net", "Added connection peer=%d\n", id);
}
CNode::~CNode()
{
CloseSocket(hSocket);
if (pfilter)
delete pfilter;
}
void CNode::AskFor(const CInv& inv)
{
if (mapAskFor.size() > MAPASKFOR_MAX_SZ || setAskFor.size() > SETASKFOR_MAX_SZ) {
int64_t nNow = GetTime();
if(nNow - nLastWarningTime > WARNING_INTERVAL) {
LogPrintf("CNode::AskFor -- WARNING: inventory message dropped: mapAskFor.size = %d, setAskFor.size = %d, MAPASKFOR_MAX_SZ = %d, SETASKFOR_MAX_SZ = %d, nSkipped = %d, peer=%d\n",
mapAskFor.size(), setAskFor.size(), MAPASKFOR_MAX_SZ, SETASKFOR_MAX_SZ, nNumWarningsSkipped, id);
nLastWarningTime = nNow;
nNumWarningsSkipped = 0;
}
else {
++nNumWarningsSkipped;
}
return;
}
// a peer may not have multiple non-responded queue positions for a single inv item
if (!setAskFor.insert(inv.hash).second)
return;
// We're using mapAskFor as a priority queue,
// the key is the earliest time the request can be sent
int64_t nRequestTime;
limitedmap<uint256, int64_t>::const_iterator it = mapAlreadyAskedFor.find(inv.hash);
if (it != mapAlreadyAskedFor.end())
nRequestTime = it->second;
else
nRequestTime = 0;
LogPrint("net", "askfor %s %d (%s) peer=%d\n", inv.ToString(), nRequestTime, DateTimeStrFormat("%H:%M:%S", nRequestTime/1000000), id);
// Make sure not to reuse time indexes to keep things in the same order
int64_t nNow = GetTimeMicros() - 1000000;
static int64_t nLastTime;
++nLastTime;
nNow = std::max(nNow, nLastTime);
nLastTime = nNow;
// Each retry is 2 minutes after the last
nRequestTime = std::max(nRequestTime + 2 * 60 * 1000000, nNow);
if (it != mapAlreadyAskedFor.end())
mapAlreadyAskedFor.update(it, nRequestTime);
else
mapAlreadyAskedFor.insert(std::make_pair(inv.hash, nRequestTime));
mapAskFor.insert(std::make_pair(nRequestTime, inv));
}
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
bool CConnman::NodeFullyConnected(const CNode* pnode)
{
return pnode && pnode->fSuccessfullyConnected && !pnode->fDisconnect;
}
std::vector<unsigned char> CNode::CalculateKeyedNetGroup(CAddress& address)
{
if(vchSecretKey.size() == 0) {
vchSecretKey.resize(32, 0);
GetRandBytes(vchSecretKey.data(), vchSecretKey.size());
}
std::vector<unsigned char> vchGroup;
CSHA256 hash;
std::vector<unsigned char> vch(32);
vchGroup = address.GetGroup();
hash.Write(begin_ptr(vchGroup), vchGroup.size());
hash.Write(begin_ptr(vchSecretKey), vchSecretKey.size());
hash.Finalize(begin_ptr(vch));
return vch;
}
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
CDataStream CConnman::BeginMessage(CNode* pnode, int nVersion, int flags, const std::string& sCommand)
{
return {SER_NETWORK, (nVersion ? nVersion : pnode->GetSendVersion()) | flags, CMessageHeader(Params().MessageStart(), sCommand.c_str(), 0) };
}
void CConnman::EndMessage(CDataStream& strm)
{
// Set the size
assert(strm.size () >= CMessageHeader::HEADER_SIZE);
unsigned int nSize = strm.size() - CMessageHeader::HEADER_SIZE;
WriteLE32((uint8_t*)&strm[CMessageHeader::MESSAGE_SIZE_OFFSET], nSize);
// Set the checksum
uint256 hash = Hash(strm.begin() + CMessageHeader::HEADER_SIZE, strm.end());
memcpy((char*)&strm[CMessageHeader::CHECKSUM_OFFSET], hash.begin(), CMessageHeader::CHECKSUM_SIZE);
}
void CConnman::PushMessage(CNode* pnode, CDataStream& strm, const std::string& sCommand)
{
if(strm.empty())
return;
unsigned int nSize = strm.size() - CMessageHeader::HEADER_SIZE;
LogPrint("net", "sending %s (%d bytes) peer=%d\n", SanitizeString(sCommand.c_str()), nSize, pnode->id);
size_t nBytesSent = 0;
{
LOCK(pnode->cs_vSend);
if(pnode->hSocket == INVALID_SOCKET) {
return;
}
bool optimisticSend(pnode->vSendMsg.empty());
pnode->vSendMsg.emplace_back(strm.begin(), strm.end());
//log total amount of bytes per command
pnode->mapSendBytesPerMsgCmd[sCommand] += strm.size();
pnode->nSendSize += strm.size();
Backport Bitcoin PR#9441: Net: Massive speedup. Net locks overhaul (#1586) * net: fix typo causing the wrong receive buffer size Surprisingly this hasn't been causing me any issues while testing, probably because it requires lots of large blocks to be flying around. Send/Recv corks need tests! * net: make vRecvMsg a list so that we can use splice() * net: make GetReceiveFloodSize public This will be needed so that the message processor can cork incoming messages * net: only disconnect if fDisconnect has been set These conditions are problematic to check without locking, and we shouldn't be relying on the refcount to disconnect. * net: wait until the node is destroyed to delete its recv buffer when vRecvMsg becomes a private buffer, it won't make sense to allow other threads to mess with it anymore. * net: set message deserialization version when it's actually time to deserialize We'll soon no longer have access to vRecvMsg, and this is more intuitive anyway. * net: handle message accounting in ReceiveMsgBytes This allows locking to be pushed down to only where it's needed Also reuse the current time rather than checking multiple times. * net: record bytes written before notifying the message processor * net: Add a simple function for waking the message handler This may be used publicly in the future * net: remove useless comments * net: remove redundant max sendbuffer size check This is left-over from before there was proper accounting. Hitting 2x the sendbuffer size should not be possible. * net: rework the way that the messagehandler sleeps In order to sleep accurately, the message handler needs to know if _any_ node has more processing that it should do before the entire thread sleeps. Rather than returning a value that represents whether ProcessMessages encountered a message that should trigger a disconnnect, interpret the return value as whether or not that node has more work to do. Also, use a global fProcessWake value that can be set by other threads, which takes precedence (for one cycle) over the messagehandler's decision. Note that the previous behavior was to only process one message per loop (except in the case of a bad checksum or invalid header). That was changed in PR #3180. The only change here in that regard is that the current node now falls to the back of the processing queue for the bad checksum/invalid header cases. * net: add a new message queue for the message processor This separates the storage of messages from the net and queued messages for processing, allowing the locks to be split. * net: add a flag to indicate when a node's process queue is full Messages are dumped very quickly from the socket handler to the processor, so it's the depth of the processing queue that's interesting. The socket handler checks the process queue's size during the brief message hand-off and pauses if necessary, and the processor possibly unpauses each time a message is popped off of its queue. * net: add a flag to indicate when a node's send buffer is full Similar to the recv flag, but this one indicates whether or not the net's send buffer is full. The socket handler checks the send queue when a new message is added and pauses if necessary, and possibly unpauses after each message is drained from its buffer. * net: remove cs_vRecvMsg vRecvMsg is now only touched by the socket handler thread. The accounting vars (nRecvBytes/nLastRecv/mapRecvBytesPerMsgCmd) are also only used by the socket handler thread, with the exception of queries from rpc/gui. These accesses are not threadsafe, but they never were. This needs to be addressed separately. Also, update comment describing data flow
2017-08-23 16:20:43 +02:00
if (pnode->nSendSize > nSendBufferMaxSize)
pnode->fPauseSend = true;
Backport Bitcoin PR#8708: net: have CConnman handle message sending (#1553) * serialization: teach serializers variadics Also add a variadic CDataStream ctor for ease-of-use. * connman is in charge of pushing messages The changes here are dense and subtle, but hopefully all is more explicit than before. - CConnman is now in charge of sending data rather than the nodes themselves. This is necessary because many decisions need to be made with all nodes in mind, and a model that requires the nodes calling up to their manager quickly turns to spaghetti. - The per-node-serializer (ssSend) has been replaced with a (quasi-)const send-version. Since the send version for serialization can only change once per connection, we now explicitly tag messages with INIT_PROTO_VERSION if they are sent before the handshake. With this done, there's no need to lock for access to nSendVersion. Also, a new stream is used for each message, so there's no need to lock during the serialization process. - This takes care of accounting for optimistic sends, so the nOptimisticBytesWritten hack can be removed. - -dropmessagestest and -fuzzmessagestest have not been preserved, as I suspect they haven't been used in years. * net: switch all callers to connman for pushing messages Drop all of the old stuff. * drop the optimistic write counter hack This is now handled properly in realtime. * net: remove now-unused ssSend and Fuzz * net: construct CNodeStates in place * net: handle version push in InitializeNode
2017-07-27 16:28:05 +02:00
// If write queue empty, attempt "optimistic write"
if (optimisticSend == true)
nBytesSent = SocketSendData(pnode);
}
if (nBytesSent)
RecordBytesSent(nBytesSent);
}
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
bool CConnman::ForNode(const CService& addr, std::function<bool(const CNode* pnode)> cond, std::function<bool(CNode* pnode)> func)
2015-06-19 15:27:37 +02:00
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* found = nullptr;
LOCK(cs_vNodes);
for (auto&& pnode : vNodes) {
if((CService)pnode->addr == addr) {
found = pnode;
break;
}
2015-06-19 15:27:37 +02:00
}
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
return found != nullptr && cond(found) && func(found);
2015-06-19 15:27:37 +02:00
}
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
bool CConnman::ForNode(NodeId id, std::function<bool(const CNode* pnode)> cond, std::function<bool(CNode* pnode)> func)
2015-06-19 15:27:37 +02:00
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
CNode* found = nullptr;
LOCK(cs_vNodes);
for (auto&& pnode : vNodes) {
if(pnode->id == id) {
found = pnode;
break;
}
2015-06-19 15:27:37 +02:00
}
Backport Bitcoin PR#9609: net: fix remaining net assertions (#1575) + Dashify * Dont deserialize nVersion into CNode, should fix #9212 * net: deserialize the entire version message locally This avoids having some vars set if the version negotiation fails. Also copy it all into CNode at the same site. nVersion and fSuccessfullyConnected are set last, as they are the gates for the other vars. Make them atomic for that reason. * net: don't run callbacks on nodes that haven't completed the version handshake Since ForEach* are can be used to send messages to all nodes, the caller may end up sending a message before the version handshake is complete. To limit this, filter out these nodes. While we're at it, may as well filter out disconnected nodes as well. Delete unused methods rather than updating them. * net: Disallow sending messages until the version handshake is complete This is a change in behavior, though it's much more sane now than before. * net: log an error rather than asserting if send version is misused Also cleaned up the comments and moved from the header to the .cpp so that logging headers aren't needed from net.h * Implement conditions for ForEachNode() and ForNode() methods of CConnman. A change making ForEachNode() and ForNode() methods ignore nodes that have not completed initial handshake have been backported from Bitcoin. Unfortunately, some Dash-specific code needs to iterate over all nodes. This change introduces additional condition argument to these methods. This argument is a functional object that should return true for nodes that should be taken into account, not ignored. Two functional objects are provided in CConnman namespace: * FullyConnectedOnly returns true for nodes that have handshake completed, * AllNodes returns true for all nodes. Overloads for ForEachNode() and ForNode() methods without condition argument are left for compatibility with non-Dash-specific code. They use FullyConnectedOnly functional object for condition. Signed-off-by: Oleg Girko <ol@infoserver.lv> * Iterate over all nodes in Dash-specific code using AllNodes condition. Use AllNodes functional object as newly introduced condition argument for ForEachNode() and ForNode() methods of CConnman to iterate over all nodes where needed in Dash-specific code. Signed-off-by: Oleg Girko <ol@infoserver.lv>
2017-08-17 20:37:22 +02:00
return found != nullptr && cond(found) && func(found);
2015-06-19 15:27:37 +02:00
}
int64_t PoissonNextSend(int64_t nNow, int average_interval_seconds) {
return nNow + (int64_t)(log1p(GetRand(1ULL << 48) * -0.0000000000000035527136788 /* -1/2^48 */) * average_interval_seconds * -1000000.0 + 0.5);
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
std::vector<CNode*> CConnman::CopyNodeVector()
{
std::vector<CNode*> vecNodesCopy;
LOCK(cs_vNodes);
for(size_t i = 0; i < vNodes.size(); ++i) {
CNode* pnode = vNodes[i];
pnode->AddRef();
vecNodesCopy.push_back(pnode);
}
return vecNodesCopy;
}
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
void CConnman::ReleaseNodeVector(const std::vector<CNode*>& vecNodes)
{
Backport Bitcoin PR#8085: p2p: Begin encapsulation (#1537) * net: move CBanDB and CAddrDB out of net.h/cpp This will eventually solve a circular dependency * net: Create CConnman to encapsulate p2p connections * net: Move socket binding into CConnman * net: move OpenNetworkConnection into CConnman * net: move ban and addrman functions into CConnman * net: Add oneshot functions to CConnman * net: move added node functions to CConnman * net: Add most functions needed for vNodes to CConnman * net: handle nodesignals in CConnman * net: Pass CConnection to wallet rather than using the global * net: Add rpc error for missing/disabled p2p functionality * net: Pass CConnman around as needed * gui: add NodeID to the peer table * net: create generic functor accessors and move vNodes to CConnman * net: move whitelist functions into CConnman * net: move nLastNodeId to CConnman * net: move nLocalHostNonce to CConnman This behavior seems to have been quite racy and broken. Move nLocalHostNonce into CNode, and check received nonces against all non-fully-connected nodes. If there's a match, assume we've connected to ourself. * net: move messageHandlerCondition to CConnman * net: move send/recv statistics to CConnman * net: move SendBufferSize/ReceiveFloodSize to CConnman * net: move nLocalServices/nRelevantServices to CConnman These are in-turn passed to CNode at connection time. This allows us to offer different services to different peers (or test the effects of doing so). * net: move semOutbound and semMasternodeOutbound to CConnman * net: SocketSendData returns written size * net: move max/max-outbound to CConnman * net: Pass best block known height into CConnman CConnman then passes the current best height into CNode at creation time. This way CConnman/CNode have no dependency on main for height, and the signals only move in one direction. This also helps to prevent identity leakage a tiny bit. Before this change, an attacker could theoretically make 2 connections on different interfaces. They would connect fully on one, and only establish the initial connection on the other. Once they receive a new block, they would relay it to your first connection, and immediately commence the version handshake on the second. Since the new block height is reflected immediately, they could attempt to learn whether the two connections were correlated. This is, of course, incredibly unlikely to work due to the small timings involved and receipt from other senders. But it doesn't hurt to lock-in nBestHeight at the time of connection, rather than letting the remote choose the time. * net: pass CClientUIInterface into CConnman * net: Drop StartNode/StopNode and use CConnman directly * net: Introduce CConnection::Options to avoid passing so many params * net: add nSendBufferMaxSize/nReceiveFloodSize to CConnection::Options * net: move vNodesDisconnected into CConnman * Made the ForEachNode* functions in src/net.cpp more pragmatic and self documenting * Convert ForEachNode* functions to take a templated function argument rather than a std::function to eliminate std::function overhead * net: move MAX_FEELER_CONNECTIONS into connman
2017-07-21 11:35:19 +02:00
LOCK(cs_vNodes);
for(size_t i = 0; i < vecNodes.size(); ++i) {
CNode* pnode = vecNodes[i];
pnode->Release();
}
}