diff options
Diffstat (limited to 'net.cpp')
-rw-r--r-- | net.cpp | 1100 |
1 files changed, 1100 insertions, 0 deletions
diff --git a/net.cpp b/net.cpp new file mode 100644 index 0000000000..b8d96006fc --- /dev/null +++ b/net.cpp @@ -0,0 +1,1100 @@ +// Copyright (c) 2009 Satoshi Nakamoto
+// Distributed under the MIT/X11 software license, see the accompanying
+// file license.txt or http://www.opensource.org/licenses/mit-license.php.
+
+#include "headers.h"
+#include <winsock2.h>
+
+void ThreadMessageHandler2(void* parg);
+void ThreadSocketHandler2(void* parg);
+void ThreadOpenConnections2(void* parg);
+
+
+
+
+
+
+//
+// Global state variables
+//
+bool fClient = false;
+uint64 nLocalServices = (fClient ? 0 : NODE_NETWORK);
+CAddress addrLocalHost(0, DEFAULT_PORT, nLocalServices);
+CNode nodeLocalHost(INVALID_SOCKET, CAddress("127.0.0.1", nLocalServices));
+CNode* pnodeLocalHost = &nodeLocalHost;
+bool fShutdown = false;
+array<bool, 10> vfThreadRunning;
+vector<CNode*> vNodes;
+CCriticalSection cs_vNodes;
+map<vector<unsigned char>, CAddress> mapAddresses;
+CCriticalSection cs_mapAddresses;
+map<CInv, CDataStream> mapRelay;
+deque<pair<int64, CInv> > vRelayExpiration;
+CCriticalSection cs_mapRelay;
+map<CInv, int64> mapAlreadyAskedFor;
+
+
+
+CAddress addrProxy;
+
+bool ConnectSocket(const CAddress& addrConnect, SOCKET& hSocketRet)
+{
+ hSocketRet = INVALID_SOCKET;
+
+ SOCKET hSocket = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+ if (hSocket == INVALID_SOCKET)
+ return false;
+
+ bool fRoutable = !(addrConnect.GetByte(3) == 10 || (addrConnect.GetByte(3) == 192 && addrConnect.GetByte(2) == 168));
+ bool fProxy = (addrProxy.ip && fRoutable);
+ struct sockaddr_in sockaddr = (fProxy ? addrProxy.GetSockAddr() : addrConnect.GetSockAddr());
+
+ if (connect(hSocket, (struct sockaddr*)&sockaddr, sizeof(sockaddr)) == SOCKET_ERROR)
+ {
+ closesocket(hSocket);
+ return false;
+ }
+
+ if (fProxy)
+ {
+ printf("Proxy connecting to %s\n", addrConnect.ToString().c_str());
+ char pszSocks4IP[] = "\4\1\0\0\0\0\0\0user";
+ memcpy(pszSocks4IP + 2, &addrConnect.port, 2);
+ memcpy(pszSocks4IP + 4, &addrConnect.ip, 4);
+ char* pszSocks4 = pszSocks4IP;
+ int nSize = sizeof(pszSocks4IP);
+
+ int ret = send(hSocket, pszSocks4, nSize, 0);
+ if (ret != nSize)
+ {
+ closesocket(hSocket);
+ return error("Error sending to proxy\n");
+ }
+ char pchRet[8];
+ if (recv(hSocket, pchRet, 8, 0) != 8)
+ {
+ closesocket(hSocket);
+ return error("Error reading proxy response\n");
+ }
+ if (pchRet[1] != 0x5a)
+ {
+ closesocket(hSocket);
+ return error("Proxy returned error %d\n", pchRet[1]);
+ }
+ printf("Proxy connection established %s\n", addrConnect.ToString().c_str());
+ }
+
+ hSocketRet = hSocket;
+ return true;
+}
+
+
+
+bool GetMyExternalIP2(const CAddress& addrConnect, const char* pszGet, const char* pszKeyword, unsigned int& ipRet)
+{
+ SOCKET hSocket;
+ if (!ConnectSocket(addrConnect, hSocket))
+ return error("GetMyExternalIP() : connection to %s failed\n", addrConnect.ToString().c_str());
+
+ send(hSocket, pszGet, strlen(pszGet), 0);
+
+ string strLine;
+ while (RecvLine(hSocket, strLine))
+ {
+ if (strLine.empty())
+ {
+ loop
+ {
+ if (!RecvLine(hSocket, strLine))
+ {
+ closesocket(hSocket);
+ return false;
+ }
+ if (strLine.find(pszKeyword) != -1)
+ {
+ strLine = strLine.substr(strLine.find(pszKeyword) + strlen(pszKeyword));
+ break;
+ }
+ }
+ closesocket(hSocket);
+ if (strLine.find("<"))
+ strLine = strLine.substr(0, strLine.find("<"));
+ strLine = strLine.substr(strspn(strLine.c_str(), " \t\n\r"));
+ strLine = wxString(strLine).Trim();
+ CAddress addr(strLine.c_str());
+ printf("GetMyExternalIP() received [%s] %s\n", strLine.c_str(), addr.ToString().c_str());
+ if (addr.ip == 0 || !addr.IsRoutable())
+ return false;
+ ipRet = addr.ip;
+ return true;
+ }
+ }
+ closesocket(hSocket);
+ return error("GetMyExternalIP() : connection closed\n");
+}
+
+
+bool GetMyExternalIP(unsigned int& ipRet)
+{
+ CAddress addrConnect;
+ char* pszGet;
+ char* pszKeyword;
+
+ for (int nLookup = 0; nLookup <= 1; nLookup++)
+ for (int nHost = 1; nHost <= 2; nHost++)
+ {
+ if (nHost == 1)
+ {
+ addrConnect = CAddress("70.86.96.218:80"); // www.ipaddressworld.com
+
+ if (nLookup == 1)
+ {
+ struct hostent* phostent = gethostbyname("www.ipaddressworld.com");
+ if (phostent && phostent->h_addr_list && phostent->h_addr_list[0])
+ addrConnect = CAddress(*(u_long*)phostent->h_addr_list[0], htons(80));
+ }
+
+ pszGet = "GET /ip.php HTTP/1.1\r\n"
+ "Host: www.ipaddressworld.com\r\n"
+ "User-Agent: Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 5.1)\r\n"
+ "Connection: close\r\n"
+ "\r\n";
+
+ pszKeyword = "IP:";
+ }
+ else if (nHost == 2)
+ {
+ addrConnect = CAddress("208.78.68.70:80"); // checkip.dyndns.org
+
+ if (nLookup == 1)
+ {
+ struct hostent* phostent = gethostbyname("checkip.dyndns.org");
+ if (phostent && phostent->h_addr_list && phostent->h_addr_list[0])
+ addrConnect = CAddress(*(u_long*)phostent->h_addr_list[0], htons(80));
+ }
+
+ pszGet = "GET / HTTP/1.1\r\n"
+ "Host: checkip.dyndns.org\r\n"
+ "User-Agent: Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 5.1)\r\n"
+ "Connection: close\r\n"
+ "\r\n";
+
+ pszKeyword = "Address:";
+ }
+
+ if (GetMyExternalIP2(addrConnect, pszGet, pszKeyword, ipRet))
+ return true;
+ }
+
+ return false;
+}
+
+
+
+
+
+bool AddAddress(CAddrDB& addrdb, const CAddress& addr)
+{
+ if (!addr.IsRoutable())
+ return false;
+ if (addr.ip == addrLocalHost.ip)
+ return false;
+ CRITICAL_BLOCK(cs_mapAddresses)
+ {
+ map<vector<unsigned char>, CAddress>::iterator it = mapAddresses.find(addr.GetKey());
+ if (it == mapAddresses.end())
+ {
+ // New address
+ mapAddresses.insert(make_pair(addr.GetKey(), addr));
+ addrdb.WriteAddress(addr);
+ return true;
+ }
+ else
+ {
+ CAddress& addrFound = (*it).second;
+ if ((addrFound.nServices | addr.nServices) != addrFound.nServices)
+ {
+ // Services have been added
+ addrFound.nServices |= addr.nServices;
+ addrdb.WriteAddress(addrFound);
+ return true;
+ }
+ }
+ }
+ return false;
+}
+
+
+
+
+
+void AbandonRequests(void (*fn)(void*, CDataStream&), void* param1)
+{
+ // If the dialog might get closed before the reply comes back,
+ // call this in the destructor so it doesn't get called after it's deleted.
+ CRITICAL_BLOCK(cs_vNodes)
+ {
+ foreach(CNode* pnode, vNodes)
+ {
+ CRITICAL_BLOCK(pnode->cs_mapRequests)
+ {
+ for (map<uint256, CRequestTracker>::iterator mi = pnode->mapRequests.begin(); mi != pnode->mapRequests.end();)
+ {
+ CRequestTracker& tracker = (*mi).second;
+ if (tracker.fn == fn && tracker.param1 == param1)
+ pnode->mapRequests.erase(mi++);
+ else
+ mi++;
+ }
+ }
+ }
+ }
+}
+
+
+
+
+
+
+
+//
+// Subscription methods for the broadcast and subscription system.
+// Channel numbers are message numbers, i.e. MSG_TABLE and MSG_PRODUCT.
+//
+// The subscription system uses a meet-in-the-middle strategy.
+// With 100,000 nodes, if senders broadcast to 1000 random nodes and receivers
+// subscribe to 1000 random nodes, 99.995% (1 - 0.99^1000) of messages will get through.
+//
+
+bool AnySubscribed(unsigned int nChannel)
+{
+ if (pnodeLocalHost->IsSubscribed(nChannel))
+ return true;
+ CRITICAL_BLOCK(cs_vNodes)
+ foreach(CNode* pnode, vNodes)
+ if (pnode->IsSubscribed(nChannel))
+ return true;
+ return false;
+}
+
+bool CNode::IsSubscribed(unsigned int nChannel)
+{
+ if (nChannel >= vfSubscribe.size())
+ return false;
+ return vfSubscribe[nChannel];
+}
+
+void CNode::Subscribe(unsigned int nChannel, unsigned int nHops)
+{
+ if (nChannel >= vfSubscribe.size())
+ return;
+
+ if (!AnySubscribed(nChannel))
+ {
+ // Relay subscribe
+ CRITICAL_BLOCK(cs_vNodes)
+ foreach(CNode* pnode, vNodes)
+ if (pnode != this)
+ pnode->PushMessage("subscribe", nChannel, nHops);
+ }
+
+ vfSubscribe[nChannel] = true;
+}
+
+void CNode::CancelSubscribe(unsigned int nChannel)
+{
+ if (nChannel >= vfSubscribe.size())
+ return;
+
+ // Prevent from relaying cancel if wasn't subscribed
+ if (!vfSubscribe[nChannel])
+ return;
+ vfSubscribe[nChannel] = false;
+
+ if (!AnySubscribed(nChannel))
+ {
+ // Relay subscription cancel
+ CRITICAL_BLOCK(cs_vNodes)
+ foreach(CNode* pnode, vNodes)
+ if (pnode != this)
+ pnode->PushMessage("sub-cancel", nChannel);
+
+ // Clear memory, no longer subscribed
+ if (nChannel == MSG_PRODUCT)
+ CRITICAL_BLOCK(cs_mapProducts)
+ mapProducts.clear();
+ }
+}
+
+
+
+
+
+
+
+
+
+CNode* FindNode(unsigned int ip)
+{
+ CRITICAL_BLOCK(cs_vNodes)
+ {
+ foreach(CNode* pnode, vNodes)
+ if (pnode->addr.ip == ip)
+ return (pnode);
+ }
+ return NULL;
+}
+
+CNode* FindNode(CAddress addr)
+{
+ CRITICAL_BLOCK(cs_vNodes)
+ {
+ foreach(CNode* pnode, vNodes)
+ if (pnode->addr == addr)
+ return (pnode);
+ }
+ return NULL;
+}
+
+CNode* ConnectNode(CAddress addrConnect, int64 nTimeout)
+{
+ if (addrConnect.ip == addrLocalHost.ip)
+ return NULL;
+
+ // Look for an existing connection
+ CNode* pnode = FindNode(addrConnect.ip);
+ if (pnode)
+ {
+ if (nTimeout != 0)
+ pnode->AddRef(nTimeout);
+ else
+ pnode->AddRef();
+ return pnode;
+ }
+
+ /// debug print
+ printf("trying %s\n", addrConnect.ToString().c_str());
+
+ // Connect
+ SOCKET hSocket;
+ if (ConnectSocket(addrConnect, hSocket))
+ {
+ /// debug print
+ printf("connected %s\n", addrConnect.ToString().c_str());
+
+ // Set to nonblocking
+ u_long nOne = 1;
+ if (ioctlsocket(hSocket, FIONBIO, &nOne) == SOCKET_ERROR)
+ printf("ConnectSocket() : ioctlsocket nonblocking setting failed, error %d\n", WSAGetLastError());
+
+ // Add node
+ CNode* pnode = new CNode(hSocket, addrConnect, false);
+ if (nTimeout != 0)
+ pnode->AddRef(nTimeout);
+ else
+ pnode->AddRef();
+ CRITICAL_BLOCK(cs_vNodes)
+ vNodes.push_back(pnode);
+
+ CRITICAL_BLOCK(cs_mapAddresses)
+ mapAddresses[addrConnect.GetKey()].nLastFailed = 0;
+ return pnode;
+ }
+ else
+ {
+ CRITICAL_BLOCK(cs_mapAddresses)
+ mapAddresses[addrConnect.GetKey()].nLastFailed = GetTime();
+ return NULL;
+ }
+}
+
+void CNode::Disconnect()
+{
+ printf("disconnecting node %s\n", addr.ToString().c_str());
+
+ closesocket(hSocket);
+
+ // If outbound and never got version message, mark address as failed
+ if (!fInbound && nVersion == 0)
+ CRITICAL_BLOCK(cs_mapAddresses)
+ mapAddresses[addr.GetKey()].nLastFailed = GetTime();
+
+ // All of a nodes broadcasts and subscriptions are automatically torn down
+ // when it goes down, so a node has to stay up to keep its broadcast going.
+
+ CRITICAL_BLOCK(cs_mapProducts)
+ for (map<uint256, CProduct>::iterator mi = mapProducts.begin(); mi != mapProducts.end();)
+ AdvertRemoveSource(this, MSG_PRODUCT, 0, (*(mi++)).second);
+
+ // Cancel subscriptions
+ for (unsigned int nChannel = 0; nChannel < vfSubscribe.size(); nChannel++)
+ if (vfSubscribe[nChannel])
+ CancelSubscribe(nChannel);
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+void ThreadSocketHandler(void* parg)
+{
+ IMPLEMENT_RANDOMIZE_STACK(ThreadSocketHandler(parg));
+
+ loop
+ {
+ vfThreadRunning[0] = true;
+ CheckForShutdown(0);
+ try
+ {
+ ThreadSocketHandler2(parg);
+ }
+ CATCH_PRINT_EXCEPTION("ThreadSocketHandler()")
+ vfThreadRunning[0] = false;
+ Sleep(5000);
+ }
+}
+
+void ThreadSocketHandler2(void* parg)
+{
+ printf("ThreadSocketHandler started\n");
+ SOCKET hListenSocket = *(SOCKET*)parg;
+ list<CNode*> vNodesDisconnected;
+ int nPrevNodeCount = 0;
+
+ loop
+ {
+ //
+ // Disconnect nodes
+ //
+ CRITICAL_BLOCK(cs_vNodes)
+ {
+ // Disconnect unused nodes
+ vector<CNode*> vNodesCopy = vNodes;
+ foreach(CNode* pnode, vNodesCopy)
+ {
+ if (pnode->ReadyToDisconnect() && pnode->vRecv.empty() && pnode->vSend.empty())
+ {
+ // remove from vNodes
+ vNodes.erase(remove(vNodes.begin(), vNodes.end(), pnode), vNodes.end());
+ pnode->Disconnect();
+
+ // hold in disconnected pool until all refs are released
+ pnode->nReleaseTime = max(pnode->nReleaseTime, GetTime() + 5 * 60);
+ if (pnode->fNetworkNode)
+ pnode->Release();
+ vNodesDisconnected.push_back(pnode);
+ }
+ }
+
+ // Delete disconnected nodes
+ list<CNode*> vNodesDisconnectedCopy = vNodesDisconnected;
+ foreach(CNode* pnode, vNodesDisconnectedCopy)
+ {
+ // wait until threads are done using it
+ if (pnode->GetRefCount() <= 0)
+ {
+ bool fDelete = false;
+ TRY_CRITICAL_BLOCK(pnode->cs_vSend)
+ TRY_CRITICAL_BLOCK(pnode->cs_vRecv)
+ TRY_CRITICAL_BLOCK(pnode->cs_mapRequests)
+ TRY_CRITICAL_BLOCK(pnode->cs_inventory)
+ fDelete = true;
+ if (fDelete)
+ {
+ vNodesDisconnected.remove(pnode);
+ delete pnode;
+ }
+ }
+ }
+ }
+ if (vNodes.size() != nPrevNodeCount)
+ {
+ nPrevNodeCount = vNodes.size();
+ MainFrameRepaint();
+ }
+
+
+ //
+ // Find which sockets have data to receive
+ //
+ struct timeval timeout;
+ timeout.tv_sec = 0;
+ timeout.tv_usec = 50000; // frequency to poll pnode->vSend
+
+ struct fd_set fdsetRecv;
+ struct fd_set fdsetSend;
+ FD_ZERO(&fdsetRecv);
+ FD_ZERO(&fdsetSend);
+ SOCKET hSocketMax = 0;
+ FD_SET(hListenSocket, &fdsetRecv);
+ hSocketMax = max(hSocketMax, hListenSocket);
+ CRITICAL_BLOCK(cs_vNodes)
+ {
+ foreach(CNode* pnode, vNodes)
+ {
+ FD_SET(pnode->hSocket, &fdsetRecv);
+ hSocketMax = max(hSocketMax, pnode->hSocket);
+ TRY_CRITICAL_BLOCK(pnode->cs_vSend)
+ if (!pnode->vSend.empty())
+ FD_SET(pnode->hSocket, &fdsetSend);
+ }
+ }
+
+ vfThreadRunning[0] = false;
+ int nSelect = select(hSocketMax + 1, &fdsetRecv, &fdsetSend, NULL, &timeout);
+ vfThreadRunning[0] = true;
+ CheckForShutdown(0);
+ if (nSelect == SOCKET_ERROR)
+ {
+ int nErr = WSAGetLastError();
+ printf("select failed: %d\n", nErr);
+ for (int i = 0; i <= hSocketMax; i++)
+ {
+ FD_SET(i, &fdsetRecv);
+ FD_SET(i, &fdsetSend);
+ }
+ Sleep(timeout.tv_usec/1000);
+ }
+ RandAddSeed();
+
+ //// debug print
+ //foreach(CNode* pnode, vNodes)
+ //{
+ // printf("vRecv = %-5d ", pnode->vRecv.size());
+ // printf("vSend = %-5d ", pnode->vSend.size());
+ //}
+ //printf("\n");
+
+
+ //
+ // Accept new connections
+ //
+ if (FD_ISSET(hListenSocket, &fdsetRecv))
+ {
+ struct sockaddr_in sockaddr;
+ int len = sizeof(sockaddr);
+ SOCKET hSocket = accept(hListenSocket, (struct sockaddr*)&sockaddr, &len);
+ CAddress addr(sockaddr);
+ if (hSocket == INVALID_SOCKET)
+ {
+ if (WSAGetLastError() != WSAEWOULDBLOCK)
+ printf("ERROR ThreadSocketHandler accept failed: %d\n", WSAGetLastError());
+ }
+ else
+ {
+ printf("accepted connection from %s\n", addr.ToString().c_str());
+ CNode* pnode = new CNode(hSocket, addr, true);
+ pnode->AddRef();
+ CRITICAL_BLOCK(cs_vNodes)
+ vNodes.push_back(pnode);
+ }
+ }
+
+
+ //
+ // Service each socket
+ //
+ vector<CNode*> vNodesCopy;
+ CRITICAL_BLOCK(cs_vNodes)
+ vNodesCopy = vNodes;
+ foreach(CNode* pnode, vNodesCopy)
+ {
+ CheckForShutdown(0);
+ SOCKET hSocket = pnode->hSocket;
+
+ //
+ // Receive
+ //
+ if (FD_ISSET(hSocket, &fdsetRecv))
+ {
+ TRY_CRITICAL_BLOCK(pnode->cs_vRecv)
+ {
+ CDataStream& vRecv = pnode->vRecv;
+ unsigned int nPos = vRecv.size();
+
+ // typical socket buffer is 8K-64K
+ const unsigned int nBufSize = 0x10000;
+ vRecv.resize(nPos + nBufSize);
+ int nBytes = recv(hSocket, &vRecv[nPos], nBufSize, 0);
+ vRecv.resize(nPos + max(nBytes, 0));
+ if (nBytes == 0)
+ {
+ // socket closed gracefully
+ if (!pnode->fDisconnect)
+ printf("recv: socket closed\n");
+ pnode->fDisconnect = true;
+ }
+ else if (nBytes < 0)
+ {
+ // socket error
+ int nErr = WSAGetLastError();
+ if (nErr != WSAEWOULDBLOCK && nErr != WSAEMSGSIZE && nErr != WSAEINTR && nErr != WSAEINPROGRESS)
+ {
+ if (!pnode->fDisconnect)
+ printf("recv failed: %d\n", nErr);
+ pnode->fDisconnect = true;
+ }
+ }
+ }
+ }
+
+ //
+ // Send
+ //
+ if (FD_ISSET(hSocket, &fdsetSend))
+ {
+ TRY_CRITICAL_BLOCK(pnode->cs_vSend)
+ {
+ CDataStream& vSend = pnode->vSend;
+ if (!vSend.empty())
+ {
+ int nBytes = send(hSocket, &vSend[0], vSend.size(), 0);
+ if (nBytes > 0)
+ {
+ vSend.erase(vSend.begin(), vSend.begin() + nBytes);
+ }
+ else if (nBytes == 0)
+ {
+ if (pnode->ReadyToDisconnect())
+ pnode->vSend.clear();
+ }
+ else
+ {
+ printf("send error %d\n", nBytes);
+ if (pnode->ReadyToDisconnect())
+ pnode->vSend.clear();
+ }
+ }
+ }
+ }
+ }
+
+
+ Sleep(10);
+ }
+}
+
+
+
+
+
+
+
+
+
+
+void ThreadOpenConnections(void* parg)
+{
+ IMPLEMENT_RANDOMIZE_STACK(ThreadOpenConnections(parg));
+
+ loop
+ {
+ vfThreadRunning[1] = true;
+ CheckForShutdown(1);
+ try
+ {
+ ThreadOpenConnections2(parg);
+ }
+ CATCH_PRINT_EXCEPTION("ThreadOpenConnections()")
+ vfThreadRunning[1] = false;
+ Sleep(5000);
+ }
+}
+
+void ThreadOpenConnections2(void* parg)
+{
+ printf("ThreadOpenConnections started\n");
+
+ // Initiate network connections
+ int nTry = 0;
+ bool fIRCOnly = false;
+ const int nMaxConnections = 15;
+ loop
+ {
+ // Wait
+ vfThreadRunning[1] = false;
+ Sleep(500);
+ while (vNodes.size() >= nMaxConnections || vNodes.size() >= mapAddresses.size())
+ {
+ CheckForShutdown(1);
+ Sleep(2000);
+ }
+ vfThreadRunning[1] = true;
+ CheckForShutdown(1);
+
+
+ //
+ // The IP selection process is designed to limit vulnerability to address flooding.
+ // Any class C (a.b.c.?) has an equal chance of being chosen, then an IP is
+ // chosen within the class C. An attacker may be able to allocate many IPs, but
+ // they would normally be concentrated in blocks of class C's. They can hog the
+ // attention within their class C, but not the whole IP address space overall.
+ // A lone node in a class C will get as much attention as someone holding all 255
+ // IPs in another class C.
+ //
+
+ // Every other try is with IRC addresses only
+ fIRCOnly = !fIRCOnly;
+ if (mapIRCAddresses.empty())
+ fIRCOnly = false;
+ else if (nTry++ < 30 && vNodes.size() < nMaxConnections/2)
+ fIRCOnly = true;
+
+ // Make a list of unique class C's
+ unsigned char pchIPCMask[4] = { 0xff, 0xff, 0xff, 0x00 };
+ unsigned int nIPCMask = *(unsigned int*)pchIPCMask;
+ vector<unsigned int> vIPC;
+ CRITICAL_BLOCK(cs_mapIRCAddresses)
+ CRITICAL_BLOCK(cs_mapAddresses)
+ {
+ vIPC.reserve(mapAddresses.size());
+ unsigned int nPrev = 0;
+ foreach(const PAIRTYPE(vector<unsigned char>, CAddress)& item, mapAddresses)
+ {
+ const CAddress& addr = item.second;
+ if (!addr.IsIPv4())
+ continue;
+ if (fIRCOnly && !mapIRCAddresses.count(item.first))
+ continue;
+
+ // Taking advantage of mapAddresses being in sorted order,
+ // with IPs of the same class C grouped together.
+ unsigned int ipC = addr.ip & nIPCMask;
+ if (ipC != nPrev)
+ vIPC.push_back(nPrev = ipC);
+ }
+ }
+ if (vIPC.empty())
+ continue;
+
+ // Choose a random class C
+ unsigned int ipC = vIPC[GetRand(vIPC.size())];
+
+ // Organize all addresses in the class C by IP
+ map<unsigned int, vector<CAddress> > mapIP;
+ CRITICAL_BLOCK(cs_mapIRCAddresses)
+ CRITICAL_BLOCK(cs_mapAddresses)
+ {
+ int64 nDelay = ((30 * 60) << vNodes.size());
+ if (!fIRCOnly)
+ {
+ nDelay *= 2;
+ if (vNodes.size() >= 3)
+ nDelay *= 4;
+ if (!mapIRCAddresses.empty())
+ nDelay *= 100;
+ }
+
+ for (map<vector<unsigned char>, CAddress>::iterator mi = mapAddresses.lower_bound(CAddress(ipC, 0).GetKey());
+ mi != mapAddresses.upper_bound(CAddress(ipC | ~nIPCMask, 0xffff).GetKey());
+ ++mi)
+ {
+ const CAddress& addr = (*mi).second;
+ if (fIRCOnly && !mapIRCAddresses.count((*mi).first))
+ continue;
+
+ int64 nRandomizer = (addr.nLastFailed * addr.ip * 7777U) % 20000;
+ if (GetTime() - addr.nLastFailed > nDelay * nRandomizer / 10000)
+ mapIP[addr.ip].push_back(addr);
+ }
+ }
+ if (mapIP.empty())
+ continue;
+
+ // Choose a random IP in the class C
+ map<unsigned int, vector<CAddress> >::iterator mi = mapIP.begin();
+ advance(mi, GetRand(mapIP.size()));
+
+ // Once we've chosen an IP, we'll try every given port before moving on
+ foreach(const CAddress& addrConnect, (*mi).second)
+ {
+ //
+ // Initiate outbound network connection
+ //
+ CheckForShutdown(1);
+ if (addrConnect.ip == addrLocalHost.ip || !addrConnect.IsIPv4() || FindNode(addrConnect.ip))
+ continue;
+
+ vfThreadRunning[1] = false;
+ CNode* pnode = ConnectNode(addrConnect);
+ vfThreadRunning[1] = true;
+ CheckForShutdown(1);
+ if (!pnode)
+ continue;
+ pnode->fNetworkNode = true;
+
+ if (addrLocalHost.IsRoutable())
+ {
+ // Advertise our address
+ vector<CAddress> vAddrToSend;
+ vAddrToSend.push_back(addrLocalHost);
+ pnode->PushMessage("addr", vAddrToSend);
+ }
+
+ // Get as many addresses as we can
+ pnode->PushMessage("getaddr");
+
+ ////// should the one on the receiving end do this too?
+ // Subscribe our local subscription list
+ const unsigned int nHops = 0;
+ for (unsigned int nChannel = 0; nChannel < pnodeLocalHost->vfSubscribe.size(); nChannel++)
+ if (pnodeLocalHost->vfSubscribe[nChannel])
+ pnode->PushMessage("subscribe", nChannel, nHops);
+
+ break;
+ }
+ }
+}
+
+
+
+
+
+
+
+
+void ThreadMessageHandler(void* parg)
+{
+ IMPLEMENT_RANDOMIZE_STACK(ThreadMessageHandler(parg));
+
+ loop
+ {
+ vfThreadRunning[2] = true;
+ CheckForShutdown(2);
+ try
+ {
+ ThreadMessageHandler2(parg);
+ }
+ CATCH_PRINT_EXCEPTION("ThreadMessageHandler()")
+ vfThreadRunning[2] = false;
+ Sleep(5000);
+ }
+}
+
+void ThreadMessageHandler2(void* parg)
+{
+ printf("ThreadMessageHandler started\n");
+ SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_BELOW_NORMAL);
+ loop
+ {
+ // Poll the connected nodes for messages
+ vector<CNode*> vNodesCopy;
+ CRITICAL_BLOCK(cs_vNodes)
+ vNodesCopy = vNodes;
+ foreach(CNode* pnode, vNodesCopy)
+ {
+ pnode->AddRef();
+
+ // Receive messages
+ TRY_CRITICAL_BLOCK(pnode->cs_vRecv)
+ ProcessMessages(pnode);
+
+ // Send messages
+ TRY_CRITICAL_BLOCK(pnode->cs_vSend)
+ SendMessages(pnode);
+
+ pnode->Release();
+ }
+
+ // Wait and allow messages to bunch up
+ vfThreadRunning[2] = false;
+ Sleep(100);
+ vfThreadRunning[2] = true;
+ CheckForShutdown(2);
+ }
+}
+
+
+
+
+
+
+
+
+
+//// todo: start one thread per processor, use getenv("NUMBER_OF_PROCESSORS")
+void ThreadBitcoinMiner(void* parg)
+{
+ vfThreadRunning[3] = true;
+ CheckForShutdown(3);
+ try
+ {
+ bool fRet = BitcoinMiner();
+ printf("BitcoinMiner returned %s\n\n\n", fRet ? "true" : "false");
+ }
+ CATCH_PRINT_EXCEPTION("BitcoinMiner()")
+ vfThreadRunning[3] = false;
+}
+
+
+
+
+
+
+
+
+
+
+
+bool StartNode(string& strError)
+{
+ strError = "";
+
+ // Sockets startup
+ WSADATA wsadata;
+ int ret = WSAStartup(MAKEWORD(2,2), &wsadata);
+ if (ret != NO_ERROR)
+ {
+ strError = strprintf("Error: TCP/IP socket library failed to start (WSAStartup returned error %d)", ret);
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ // Get local host ip
+ char pszHostName[255];
+ if (gethostname(pszHostName, 255) == SOCKET_ERROR)
+ {
+ strError = strprintf("Error: Unable to get IP address of this computer (gethostname returned error %d)", WSAGetLastError());
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+ struct hostent* phostent = gethostbyname(pszHostName);
+ if (!phostent)
+ {
+ strError = strprintf("Error: Unable to get IP address of this computer (gethostbyname returned error %d)", WSAGetLastError());
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+ addrLocalHost = CAddress(*(long*)(phostent->h_addr_list[0]),
+ DEFAULT_PORT,
+ nLocalServices);
+ printf("addrLocalHost = %s\n", addrLocalHost.ToString().c_str());
+
+ // Create socket for listening for incoming connections
+ SOCKET hListenSocket = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+ if (hListenSocket == INVALID_SOCKET)
+ {
+ strError = strprintf("Error: Couldn't open socket for incoming connections (socket returned error %d)", WSAGetLastError());
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ // Set to nonblocking, incoming connections will also inherit this
+ u_long nOne = 1;
+ if (ioctlsocket(hListenSocket, FIONBIO, &nOne) == SOCKET_ERROR)
+ {
+ strError = strprintf("Error: Couldn't set properties on socket for incoming connections (ioctlsocket returned error %d)", WSAGetLastError());
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ // The sockaddr_in structure specifies the address family,
+ // IP address, and port for the socket that is being bound
+ int nRetryLimit = 15;
+ struct sockaddr_in sockaddr = addrLocalHost.GetSockAddr();
+ if (bind(hListenSocket, (struct sockaddr*)&sockaddr, sizeof(sockaddr)) == SOCKET_ERROR)
+ {
+ int nErr = WSAGetLastError();
+ if (nErr == WSAEADDRINUSE)
+ strError = strprintf("Error: Unable to bind to port %s on this computer. The program is probably already running.", addrLocalHost.ToString().c_str());
+ else
+ strError = strprintf("Error: Unable to bind to port %s on this computer (bind returned error %d)", addrLocalHost.ToString().c_str(), nErr);
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+ printf("bound to addrLocalHost = %s\n\n", addrLocalHost.ToString().c_str());
+
+ // Listen for incoming connections
+ if (listen(hListenSocket, SOMAXCONN) == SOCKET_ERROR)
+ {
+ strError = strprintf("Error: Listening for incoming connections failed (listen returned error %d)", WSAGetLastError());
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ // Get our external IP address for incoming connections
+ if (addrIncoming.ip)
+ addrLocalHost.ip = addrIncoming.ip;
+
+ if (GetMyExternalIP(addrLocalHost.ip))
+ {
+ addrIncoming = addrLocalHost;
+ CWalletDB().WriteSetting("addrIncoming", addrIncoming);
+ }
+
+ // Get addresses from IRC and advertise ours
+ if (_beginthread(ThreadIRCSeed, 0, NULL) == -1)
+ printf("Error: _beginthread(ThreadIRCSeed) failed\n");
+
+ //
+ // Start threads
+ //
+ if (_beginthread(ThreadSocketHandler, 0, new SOCKET(hListenSocket)) == -1)
+ {
+ strError = "Error: _beginthread(ThreadSocketHandler) failed";
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ if (_beginthread(ThreadOpenConnections, 0, NULL) == -1)
+ {
+ strError = "Error: _beginthread(ThreadOpenConnections) failed";
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ if (_beginthread(ThreadMessageHandler, 0, NULL) == -1)
+ {
+ strError = "Error: _beginthread(ThreadMessageHandler) failed";
+ printf("%s\n", strError.c_str());
+ return false;
+ }
+
+ return true;
+}
+
+bool StopNode()
+{
+ printf("StopNode()\n");
+ fShutdown = true;
+ nTransactionsUpdated++;
+ int64 nStart = GetTime();
+ while (vfThreadRunning[0] || vfThreadRunning[2] || vfThreadRunning[3])
+ {
+ if (GetTime() - nStart > 15)
+ break;
+ Sleep(20);
+ }
+ if (vfThreadRunning[0]) printf("ThreadSocketHandler still running\n");
+ if (vfThreadRunning[1]) printf("ThreadOpenConnections still running\n");
+ if (vfThreadRunning[2]) printf("ThreadMessageHandler still running\n");
+ if (vfThreadRunning[3]) printf("ThreadBitcoinMiner still running\n");
+ while (vfThreadRunning[2])
+ Sleep(20);
+ Sleep(50);
+
+ // Sockets shutdown
+ WSACleanup();
+ return true;
+}
+
+void CheckForShutdown(int n)
+{
+ if (fShutdown)
+ {
+ if (n != -1)
+ vfThreadRunning[n] = false;
+ if (n == 0)
+ foreach(CNode* pnode, vNodes)
+ closesocket(pnode->hSocket);
+ _endthread();
+ }
+}
|