bk commit into 5.1 tree (jonas:1.2043) BUG#22013

[jonas]

Below is the list of changes that have just been committed into a local
5.1 repository of jonas. When jonas does a push these changes will
be propagated to the main repository and, within 24 hours after the
push, to the public repository.
For information on how to access the public repository
see http://dev.mysql.com/doc/mysql/en/installing-source-tree.html

ChangeSet@stripped, 2006-09-05 14:00:12+02:00, jonas@stripped +4 -0
  ndb - bug#22013
    Fix bug in event handling wrt early node shutdown

  storage/ndb/src/mgmsrv/MgmtSrvr.cpp@stripped, 2006-09-05 14:00:10+02:00,
jonas@stripped +35 -13
    Fix bug in event handling wrt early node shutdown

  storage/ndb/src/ndbapi/ClusterMgr.cpp@stripped, 2006-09-05 14:00:10+02:00,
jonas@stripped +5 -4
    Fix reportNodeFailed if only connected wo/ having received any API_REGCONF

  storage/ndb/src/ndbapi/ClusterMgr.hpp@stripped, 2006-09-05 14:00:10+02:00,
jonas@stripped +2 -2
    Fix reportNodeFailed if only connected wo/ having received any API_REGCONF

  storage/ndb/src/ndbapi/SignalSender.cpp@stripped, 2006-09-05 14:00:10+02:00,
jonas@stripped +9 -0
    Fix memleak

# This is a BitKeeper patch.  What follows are the unified diffs for the
# set of deltas contained in the patch.  The rest of the patch, the part
# that BitKeeper cares about, is below these diffs.
# User:	jonas
# Host:	perch.ndb.mysql.com
# Root:	/home/jonas/src/mysql-5.1-wl2325-5.0

--- 1.94/storage/ndb/src/mgmsrv/MgmtSrvr.cpp	2006-09-05 14:00:15 +02:00
+++ 1.95/storage/ndb/src/mgmsrv/MgmtSrvr.cpp	2006-09-05 14:00:15 +02:00
@@ -139,8 +139,11 @@
         m_started_nodes.erase(0, false);
         m_started_nodes.unlock();
 
-        setEventReportingLevelImpl(node, req);
-
+        if (setEventReportingLevelImpl(node, req))
+	{
+	  ndbout_c("setEventReportingLevelImpl(%d): failed", node);
+	}
+	
         SetLogLevelOrd ord;
         ord = m_nodeLogLevel[node];
         setNodeLogLevelImpl(node, ord);
@@ -157,10 +160,16 @@
       m_log_level_requests.erase(0, false);
       m_log_level_requests.unlock();
 
-      if(req.blockRef == 0){
+      if(req.blockRef == 0)
+      {
 	req.blockRef = _ownReference;
-	setEventReportingLevelImpl(0, req);
-      } else {
+	if (setEventReportingLevelImpl(0, req))
+	{
+	  ndbout_c("setEventReportingLevelImpl: failed 2!");
+	}
+      } 
+      else 
+      {
         SetLogLevelOrd ord;
         ord = req;
 	setNodeLogLevelImpl(req.blockRef, ord);
@@ -1189,9 +1198,6 @@
   NodeId nodeId = 0;
   NDB_TICKS maxTime = NdbTick_CurrentMillisecond() + waitTime;
 
-  ndbout_c(" %d", nodes.get(1));
-  ndbout_c(" %d", nodes.get(2));
-
   while(getNextNodeId(&nodeId, NDB_MGM_NODE_TYPE_NDB)) {
     if (!nodes.get(nodeId))
       continue;
@@ -1391,6 +1397,11 @@
     }
   }
 
+  if (nodes.isclear())
+  {
+    return SEND_OR_RECEIVE_FAILED;
+  }
+  
   int error = 0;
   while (!nodes.isclear())
   {
@@ -1407,16 +1418,24 @@
       error = 1;
       break;
     }
+      // Since sending okToSend(true), 
+      // there is no guarantee that NF_COMPLETEREP will come
+      // i.e listen also to NODE_FAILREP
+    case GSN_NODE_FAILREP: {
+      const NodeFailRep * const rep =
+	CAST_CONSTPTR(NodeFailRep, signal->getDataPtr());
+      NdbNodeBitmask mask;
+      mask.assign(NdbNodeBitmask::Size, rep->theNodes);
+      nodes.bitANDC(mask);
+      break;
+    }
+      
     case GSN_NF_COMPLETEREP:{
       const NFCompleteRep * const rep =
 	CAST_CONSTPTR(NFCompleteRep, signal->getDataPtr());
       nodes.clear(rep->failedNodeId);
       break;
     }
-    case GSN_NODE_FAILREP:{
-      // ignore, NF_COMPLETEREP will arrive later
-      break;
-    }
     default:
       report_unknown_signal(signal);
       return SEND_OR_RECEIVE_FAILED;
@@ -1713,7 +1732,10 @@
 
   theData[1] = nodeId;
   if (alive) {
-    m_started_nodes.push_back(nodeId);
+    if (nodeTypes[nodeId] == NODE_TYPE_DB)
+    {
+      m_started_nodes.push_back(nodeId);
+    }
     rep->setEventType(NDB_LE_Connected);
   } else {
     rep->setEventType(NDB_LE_Connected);

--- 1.25/storage/ndb/src/ndbapi/ClusterMgr.cpp	2006-09-05 14:00:15 +02:00
+++ 1.26/storage/ndb/src/ndbapi/ClusterMgr.cpp	2006-09-05 14:00:15 +02:00
@@ -405,6 +405,7 @@
   theNode.m_info.m_version = 0;
   theNode.compatible = true;
   theNode.nfCompleteRep = true;
+  theNode.m_state.startLevel = NodeState::SL_NOTHING;
   
   theFacade.ReportNodeAlive(nodeId);
 }
@@ -416,14 +417,13 @@
 
   noOfConnectedNodes--;
   theNodes[nodeId].connected = false;
-
   theNodes[nodeId].m_state.m_connected_nodes.clear();
 
-  reportNodeFailed(nodeId);
+  reportNodeFailed(nodeId, true);
 }
 
 void
-ClusterMgr::reportNodeFailed(NodeId nodeId){
+ClusterMgr::reportNodeFailed(NodeId nodeId, bool disconnect){
 
   Node & theNode = theNodes[nodeId];
  
@@ -434,10 +434,11 @@
   {
     theFacade.doDisconnect(nodeId);
   }
+  
   const bool report = (theNode.m_state.startLevel != NodeState::SL_NOTHING);  
   theNode.m_state.startLevel = NodeState::SL_NOTHING;
   
-  if(report)
+  if(disconnect || report)
   {
     theFacade.ReportNodeDead(nodeId);
   }

--- 1.10/storage/ndb/src/ndbapi/ClusterMgr.hpp	2006-09-05 14:00:15 +02:00
+++ 1.11/storage/ndb/src/ndbapi/ClusterMgr.hpp	2006-09-05 14:00:15 +02:00
@@ -92,8 +92,8 @@
   NdbMutex*     clusterMgrThreadMutex;
   
   void showState(NodeId nodeId);
-  void reportNodeFailed(NodeId nodeId);
-
+  void reportNodeFailed(NodeId nodeId, bool disconnect = false);
+  
   /**
    * Signals received
    */

--- 1.7/storage/ndb/src/ndbapi/SignalSender.cpp	2006-09-05 14:00:15 +02:00
+++ 1.8/storage/ndb/src/ndbapi/SignalSender.cpp	2006-09-05 14:00:15 +02:00
@@ -20,6 +20,14 @@
 #include <signaldata/NFCompleteRep.hpp>
 #include <signaldata/NodeFailRep.hpp>
 
+static
+void
+require(bool x)
+{
+  if (!x)
+    abort();
+}
+
 SimpleSignal::SimpleSignal(bool dealloc){
   memset(this, 0, sizeof(* this));
   deallocSections = dealloc;
@@ -146,6 +154,7 @@
 {
   SimpleSignal * s = t.check(m_jobBuffer);
   if(s != 0){
+    m_usedBuffer.push_back(s);
     return s;
   }