Make sure we always have a node to do IBD from

This introduces the concept of the 'sync node', which is the one we
asked for missing blocks. In case the sync node goes away, a new one
will be selected.

For now, the heuristic is very simple, but it can easily be extended
later to add better policies.

3 files changed, 60 insertions, 19 deletions

diff --git a/src/main.cpp b/src/main.cpp
index 7f408c4d1b..06957ba526 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -3235,18 +3235,6 @@ bool static ProcessMessage(CNode* pfrom, string strCommand, CDataStream& vRecv)
             }
         }
 
-        // Ask the first connected node for block updates
-        static int nAskedForBlocks = 0;
-        if (!pfrom->fClient && !pfrom->fOneShot && !fImporting && !fReindex &&
-            (pfrom->nStartingHeight > (nBestHeight - 144)) &&
-            (pfrom->nVersion < NOBLKS_VERSION_START ||
-             pfrom->nVersion >= NOBLKS_VERSION_END) &&
-             (nAskedForBlocks < 1 || vNodes.size() <= 1))
-        {
-            nAskedForBlocks++;
-            pfrom->PushGetBlocks(pindexBest, uint256(0));
-        }
-
         // Relay alerts
         {
             LOCK(cs_mapAlerts);
@@ -3855,6 +3843,12 @@ bool SendMessages(CNode* pto, bool fSendTrickle)
                 pto->PushMessage("ping");
         }
 
+        // Start block sync
+        if (pto->fStartSync && !fImporting && !fReindex) {
+            pto->fStartSync = false;
+            pto->PushGetBlocks(pindexBest, uint256(0));
+        }
+
         // Resend wallet transactions that haven't gotten in a block yet
         // Except during reindex, importing and IBD, when old wallet
         // transactions become unconfirmed and spams other nodes.
diff --git a/src/net.cpp b/src/net.cpp
index 0defa96e40..4727f1e2e6 100644
--- a/src/net.cpp
+++ b/src/net.cpp
@@ -44,6 +44,7 @@ static map<CNetAddr, LocalServiceInfo> mapLocalHost;
 static bool vfReachable[NET_MAX] = {};
 static bool vfLimited[NET_MAX] = {};
 static CNode* pnodeLocalHost = NULL;
+static CNode* pnodeSync = NULL;
 uint64 nLocalHostNonce = 0;
 static std::vector<SOCKET> vhListenSocket;
 CAddrMan addrman;
@@ -521,12 +522,16 @@ void CNode::CloseSocketDisconnect()
         printf("disconnecting node %s\n", addrName.c_str());
         closesocket(hSocket);
         hSocket = INVALID_SOCKET;
-
-        // in case this fails, we'll empty the recv buffer when the CNode is deleted
-        TRY_LOCK(cs_vRecvMsg, lockRecv);
-        if (lockRecv)
-            vRecvMsg.clear();
     }
+
+    // in case this fails, we'll empty the recv buffer when the CNode is deleted
+    TRY_LOCK(cs_vRecvMsg, lockRecv);
+    if (lockRecv)
+        vRecvMsg.clear();
+
+    // if this was the sync node, we'll need a new one
+    if (this == pnodeSync)
+        pnodeSync = NULL;
 }
 
 void CNode::Cleanup()
@@ -1546,24 +1551,64 @@ bool OpenNetworkConnection(const CAddress& addrConnect, CSemaphoreGrant *grantOu
 }
 
 
+// for now, use a very simple selection metric: the node from which we received
+// most recently
+double static NodeSyncScore(const CNode *pnode) {
+    return -pnode->nLastRecv;
+}
 
+void static StartSync(const vector<CNode*> &vNodes) {
+    CNode *pnodeNewSync = NULL;
+    double dBestScore = 0;
 
+    // fImporting and fReindex are accessed out of cs_main here, but only
+    // as an optimization - they are checked again in SendMessages.
+    if (fImporting || fReindex)
+        return;
 
-
+    // Iterate over all nodes
+    BOOST_FOREACH(CNode* pnode, vNodes) {
+        // check preconditions for allowing a sync
+        if (!pnode->fClient && !pnode->fOneShot &&
+            !pnode->fDisconnect && pnode->fSuccessfullyConnected &&
+            (pnode->nStartingHeight > (nBestHeight - 144)) &&
+            (pnode->nVersion < NOBLKS_VERSION_START || pnode->nVersion >= NOBLKS_VERSION_END)) {
+            // if ok, compare node's score with the best so far
+            double dScore = NodeSyncScore(pnode);
+            if (pnodeNewSync == NULL || dScore > dBestScore) {
+                pnodeNewSync = pnode;
+                dBestScore = dScore;
+            }
+        }
+    }
+    // if a new sync candidate was found, start sync!
+    if (pnodeNewSync) {
+        pnodeNewSync->fStartSync = true;
+        pnodeSync = pnodeNewSync;
+    }
+}
 
 void ThreadMessageHandler()
 {
     SetThreadPriority(THREAD_PRIORITY_BELOW_NORMAL);
     while (true)
     {
+        bool fHaveSyncNode = false;
+
         vector<CNode*> vNodesCopy;
         {
             LOCK(cs_vNodes);
             vNodesCopy = vNodes;
-            BOOST_FOREACH(CNode* pnode, vNodesCopy)
+            BOOST_FOREACH(CNode* pnode, vNodesCopy) {
                 pnode->AddRef();
+                if (pnode == pnodeSync)
+                    fHaveSyncNode = true;
+            }
         }
 
+        if (!fHaveSyncNode)
+            StartSync(vNodesCopy);
+
         // Poll the connected nodes for messages
         CNode* pnodeTrickle = NULL;
         if (!vNodesCopy.empty())
diff --git a/src/net.h b/src/net.h
index 30b9ac8663..2a8b7f09ee 100644
--- a/src/net.h
+++ b/src/net.h
@@ -202,6 +202,7 @@ public:
     CBlockIndex* pindexLastGetBlocksBegin;
     uint256 hashLastGetBlocksEnd;
     int nStartingHeight;
+    bool fStartSync;
 
     // flood relay
     std::vector<CAddress> vAddrToSend;
@@ -242,6 +243,7 @@ public:
         pindexLastGetBlocksBegin = 0;
         hashLastGetBlocksEnd = 0;
         nStartingHeight = -1;
+        fStartSync = false;
         fGetAddr = false;
         nMisbehavior = 0;
         fRelayTxes = false;