Merge whalegate.ndb.mysql.com:/home/tomas/mysql-5.0

into  whalegate.ndb.mysql.com:/home/tomas/mysql-5.0-ndb

BUILD/Makefile.am

@@ -28,6 +28,7 @@ EXTRA_DIST =		FINISH.sh \
 			compile-alpha-debug \
 			compile-amd64-debug-max \
 			compile-amd64-max \
+			compile-amd64-max-sci \
 			compile-darwin-mwcc \
 			compile-dist \
 			compile-hpux11-parisc2-aCC \
@@ -53,6 +54,7 @@ EXTRA_DIST =		FINISH.sh \
 			compile-pentium-valgrind-max \
 			compile-pentium64-debug \
 			compile-pentium64-debug-max \
+			compile-pentium64-max-sci \
 			compile-pentium64-valgrind-max \
 			compile-ppc \
 			compile-ppc-debug \

BUILD/compile-amd64-max-sci

+#! /bin/sh
+
+path=`dirname $0`
+. "$path/SETUP.sh"
+extra_flags="$amd64_cflags $fast_cflags -g"
+extra_configs="$amd64_configs $max_configs --with-ndb-sci=/opt/DIS"
+
+. "$path/FINISH.sh"

BUILD/compile-pentium64-max-sci

+#! /bin/sh
+
+path=`dirname $0`
+. "$path/SETUP.sh"
+
+extra_flags="$pentium64_cflags $fast_cflags -g"
+extra_configs="$pentium_configs $max_configs --with-ndb-sci=/opt/DIS"
+
+. "$path/FINISH.sh"

config/ac-macros/ha_ndbcluster.m4

@@ -22,7 +22,7 @@ AC_DEFUN([MYSQL_CHECK_NDB_OPTIONS], [
       if test -f "$mysql_sci_dir/lib/libsisci.a" -a \ 
               -f "$mysql_sci_dir/include/sisci_api.h"; then
         NDB_SCI_INCLUDES="-I$mysql_sci_dir/include"
-        NDB_SCI_LIBS="-L$mysql_sci_dir/lib -lsisci"
+        NDB_SCI_LIBS="$mysql_sci_dir/lib/libsisci.a"
         AC_MSG_RESULT([-- including sci transporter])
         AC_DEFINE([NDB_SCI_TRANSPORTER], [1],
                   [Including Ndb Cluster DB sci transporter])

libmysqld/Makefile.am

@@ -81,7 +81,7 @@ INC_LIB=	$(top_builddir)/regex/libregex.a \
 		$(top_builddir)/mysys/libmysys.a \
 		$(top_builddir)/strings/libmystrings.a \
 		$(top_builddir)/dbug/libdbug.a \
-		$(top_builddir)/vio/libvio.a
+		$(top_builddir)/vio/libvio.a @NDB_SCI_LIBS@
 
 
 #

libmysqld/examples/Makefile.am

@@ -35,7 +35,8 @@ INCLUDES =	-I$(top_builddir)/include -I$(top_srcdir)/include -I$(srcdir) \
 		-I$(top_srcdir) -I$(top_srcdir)/client -I$(top_srcdir)/regex \
 		$(openssl_includes)
 LIBS =		@LIBS@ @WRAPLIBS@ @CLIENT_LIBS@ $(yassl_libs)
-LDADD =		@CLIENT_EXTRA_LDFLAGS@ ../libmysqld.a @innodb_system_libs@ @LIBDL@ $(CXXLDFLAGS)
+LDADD =		@CLIENT_EXTRA_LDFLAGS@ ../libmysqld.a @innodb_system_libs@ @LIBDL@ $(CXXLDFLAGS) \
+                @NDB_SCI_LIBS@
 
 mysqltest_embedded_LINK = $(CXXLINK)
 mysqltest_embedded_SOURCES =	mysqltest.c

mysql-test/include/have_log_bin.inc

+-- require r/have_log_bin.require
+disable_query_log;
+show variables like "log_bin";
+enable_query_log;

mysql-test/r/have_log_bin.require

+Variable_name	Value
+have_log_bin	ON

ndb/include/mgmapi/ndbd_exit_codes.h

@@ -79,6 +79,8 @@ typedef ndbd_exit_classification_enum ndbd_exit_classification;
 #define NDBD_EXIT_NO_MORE_UNDOLOG             2312 
 #define NDBD_EXIT_SR_UNDOLOG                  2313 
 #define NDBD_EXIT_SINGLE_USER_MODE            2314 
+#define NDBD_EXIT_NODE_DECLARED_DEAD          2315 
+#define NDBD_EXIT_SR_SCHEMAFILE               2316
 #define NDBD_EXIT_MEMALLOC                    2327
 #define NDBD_EXIT_BLOCK_JBUFCONGESTION        2334
 #define NDBD_EXIT_TIME_QUEUE_SHORT            2335

ndb/src/common/transporter/Packer.cpp

@@ -20,7 +20,12 @@
 #include <TransporterCallback.hpp>
 #include <RefConvert.hpp>
 
+#ifdef ERROR_INSERT
+Uint32 MAX_RECEIVED_SIGNALS = 1024;
+#else
 #define MAX_RECEIVED_SIGNALS 1024
+#endif
+
 Uint32
 TransporterRegistry::unpack(Uint32 * readPtr,
 			    Uint32 sizeOfData,

ndb/src/common/transporter/SCI_Transporter.hpp

@@ -54,12 +54,12 @@
  *  local segment, the SCI transporter connects to a segment created by another 
  *  transporter at a remote node, and the maps the remote segment into its  
  *  virtual address space. However, since NDB Cluster relies on redundancy 
- *  at the network level, by using dual SCI adapters communica 
- * 
+ *  at the network level, by using dual SCI adapters communication can be
+ *  maintained even if one of the adapter cards fails (or anything on the
+ *  network this adapter card exists in e.g. an SCI switch failure).
  * 
  */ 
 
-
 /**  
  * class SCITransporter 
  * @brief - main class for the SCI transporter. 
@@ -84,16 +84,6 @@ public:
   sci_error_t createSequence(Uint32 adapterid);      
    
    
-  /** 
-   * starts a sequence for error checking. 
-   * The actual checking that a sequence is correct is done implicitly 
-   * in SCIMemCpy (in doSend).  
-   * @param adapterid the adapter on which to start the sequence. 
-   * @return SCI_ERR_OK if ok, otherwize something else. 
-   */ 
-  sci_error_t startSequence(Uint32 adapterid);          
- 
- 
   /** Initiate Local Segment: create a memory segment, 
    * prepare a memory segment, map the local segment  
    * into  memory space and make segment available. 
@@ -159,7 +149,6 @@ private:
   bool m_mapped; 
   bool m_initLocal; 
   bool m_sciinit; 
-  Uint32 m_swapCounter; 
   Uint32 m_failCounter; 
   /** 
    * For statistics on transfered packets  
@@ -195,7 +184,6 @@ private:
    */ 
   Uint32 m_reportFreq; 
  
- 
   Uint32 m_adapters;   
   Uint32 m_numberOfRemoteNodes; 
  

ndb/src/common/transporter/TCP_Transporter.hpp

@@ -102,6 +102,10 @@ private:
   virtual void updateReceiveDataPtr(Uint32 bytesRead);
 
   virtual Uint32 get_free_buffer() const;
+
+  inline bool hasReceiveData () const {
+    return receiveBuffer.sizeOfData > 0;
+  }
 protected:
   /**
    * Setup client/server and perform connect/accept

ndb/src/common/transporter/TransporterRegistry.cpp

@@ -841,28 +841,13 @@ TransporterRegistry::poll_OSE(Uint32 timeOutMillis)
 Uint32 
 TransporterRegistry::poll_TCP(Uint32 timeOutMillis)
 {
+  bool hasdata = false;
   if (false && nTCPTransporters == 0)
   {
     tcpReadSelectReply = 0;
     return 0;
   }
   
-  struct timeval timeout;
-#ifdef NDB_OSE
-  // Return directly if there are no TCP transporters configured
-  
-  if(timeOutMillis <= 1){
-    timeout.tv_sec  = 0;
-    timeout.tv_usec = 1025;
-  } else {
-    timeout.tv_sec  = timeOutMillis / 1000;
-    timeout.tv_usec = (timeOutMillis % 1000) * 1000;
-  }
-#else  
-  timeout.tv_sec  = timeOutMillis / 1000;
-  timeout.tv_usec = (timeOutMillis % 1000) * 1000;
-#endif
-
   NDB_SOCKET_TYPE maxSocketValue = -1;
   
   // Needed for TCP/IP connections
@@ -885,7 +870,26 @@ TransporterRegistry::poll_TCP(Uint32 timeOutMillis)
       // Put the connected transporters in the socket read-set 
       FD_SET(socket, &tcpReadset);
     }
+    hasdata |= t->hasReceiveData();
+  }
+  
+  timeOutMillis = hasdata ? 0 : timeOutMillis;
+  
+  struct timeval timeout;
+#ifdef NDB_OSE
+  // Return directly if there are no TCP transporters configured
+  
+  if(timeOutMillis <= 1){
+    timeout.tv_sec  = 0;
+    timeout.tv_usec = 1025;
+  } else {
+    timeout.tv_sec  = timeOutMillis / 1000;
+    timeout.tv_usec = (timeOutMillis % 1000) * 1000;
   }
+#else  
+  timeout.tv_sec  = timeOutMillis / 1000;
+  timeout.tv_usec = (timeOutMillis % 1000) * 1000;
+#endif
 
   // The highest socket value plus one
   maxSocketValue++; 
@@ -901,7 +905,7 @@ TransporterRegistry::poll_TCP(Uint32 timeOutMillis)
   }
 #endif
   
-  return tcpReadSelectReply;
+  return tcpReadSelectReply || hasdata;
 }
 #endif
 
@@ -937,8 +941,6 @@ TransporterRegistry::performReceive()
 #endif
 
 #ifdef NDB_TCP_TRANSPORTER
-  if(tcpReadSelectReply > 0)
-  {
   for (int i=0; i<nTCPTransporters; i++) 
   {
     checkJobBuffer();
@@ -946,21 +948,23 @@ TransporterRegistry::performReceive()
     const NodeId nodeId = t->getRemoteNodeId();
     const NDB_SOCKET_TYPE socket    = t->getSocket();
     if(is_connected(nodeId)){
-	if(t->isConnected() && FD_ISSET(socket, &tcpReadset)) 
+      if(t->isConnected())
       {
-	  const int receiveSize = t->doReceive();
-	  if(receiveSize > 0)
+        if (FD_ISSET(socket, &tcpReadset))
+	{
+	  t->doReceive();
+        }
+
+        if (t->hasReceiveData())
         {
           Uint32 * ptr;
           Uint32 sz = t->getReceiveData(&ptr);
-	    transporter_recv_from(callbackObj, nodeId);
           Uint32 szUsed = unpack(ptr, sz, nodeId, ioStates[nodeId]);
           t->updateReceiveDataPtr(szUsed);
 	}
       } 
     }
   }
-  }
 #endif
   
 #ifdef NDB_SCI_TRANSPORTER

ndb/src/kernel/blocks/cmvmi/Cmvmi.cpp

@@ -132,6 +132,7 @@ Cmvmi::~Cmvmi()
 
 #ifdef ERROR_INSERT
 NodeBitmask c_error_9000_nodes_mask;
+extern Uint32 MAX_RECEIVED_SIGNALS;
 #endif
 
 void Cmvmi::execNDB_TAMPER(Signal* signal) 
@@ -161,6 +162,22 @@ void Cmvmi::execNDB_TAMPER(Signal* signal)
     kill(getpid(), SIGABRT);
   }
 #endif
+
+#ifdef ERROR_INSERT
+  if (signal->theData[0] == 9003)
+  {
+    if (MAX_RECEIVED_SIGNALS < 1024)
+    {
+      MAX_RECEIVED_SIGNALS = 1024;
+    }
+    else
+    {
+      MAX_RECEIVED_SIGNALS = 1 + (rand() % 128);
+    }
+    ndbout_c("MAX_RECEIVED_SIGNALS: %d", MAX_RECEIVED_SIGNALS);
+    CLEAR_ERROR_INSERT_VALUE;
+  }
+#endif
 }//execNDB_TAMPER()
 
 void Cmvmi::execSET_LOGLEVELORD(Signal* signal) 

ndb/src/kernel/blocks/dbdict/Dbdict.cpp

@@ -79,6 +79,9 @@
 #include <NdbSleep.h>
 #include <signaldata/ApiBroadcast.hpp>
 
+#include <EventLogger.hpp>
+extern EventLogger g_eventLogger;
+
 #define ZNOT_FOUND 626
 #define ZALREADYEXIST 630
 
@@ -1069,17 +1072,36 @@ void Dbdict::readSchemaConf(Signal* signal,
 
   for (Uint32 n = 0; n < xsf->noOfPages; n++) {
     SchemaFile * sf = &xsf->schemaPage[n];
-    bool ok =
-      memcmp(sf->Magic, NDB_SF_MAGIC, sizeof(sf->Magic)) == 0 &&
-      sf->FileSize != 0 &&
-      sf->FileSize % NDB_SF_PAGE_SIZE == 0 &&
-      sf->FileSize == sf0->FileSize &&
-      sf->PageNumber == n &&
-      computeChecksum((Uint32*)sf, NDB_SF_PAGE_SIZE_IN_WORDS) == 0;
-    ndbrequire(ok || !crashInd);
-    if (! ok) {
-      jam();
-      ndbrequire(fsPtr.p->fsState == FsConnectRecord::READ_SCHEMA1);
+    bool ok = false;
+    const char *reason;
+    if (memcmp(sf->Magic, NDB_SF_MAGIC, sizeof(sf->Magic)) != 0)
+    { jam(); reason = "magic code"; }
+    else if (sf->FileSize == 0)
+    { jam(); reason = "file size == 0"; }
+    else if (sf->FileSize % NDB_SF_PAGE_SIZE != 0)
+    { jam(); reason = "invalid size multiple"; }
+    else if (sf->FileSize != sf0->FileSize)
+    { jam(); reason = "invalid size"; }
+    else if (sf->PageNumber != n)
+    { jam(); reason = "invalid page number"; }
+    else if (computeChecksum((Uint32*)sf, NDB_SF_PAGE_SIZE_IN_WORDS) != 0)
+    { jam(); reason = "invalid checksum"; }
+    else
+      ok = true;
+
+    if (!ok)
+    {
+      char reason_msg[128];
+      snprintf(reason_msg, sizeof(reason_msg),
+               "schema file corrupt, page %u (%s, "
+               "sz=%u sz0=%u pn=%u)",
+               n, reason, sf->FileSize, sf0->FileSize, sf->PageNumber);
+      if (crashInd)
+        progError(__LINE__, NDBD_EXIT_SR_SCHEMAFILE, reason_msg);
+      ndbrequireErr(fsPtr.p->fsState == FsConnectRecord::READ_SCHEMA1,
+                    NDBD_EXIT_SR_SCHEMAFILE);
+      jam();
+      infoEvent("primary %s, trying backup", reason_msg);
       readSchemaRef(signal, fsPtr);
       return;
     }

ndb/src/kernel/blocks/qmgr/QmgrMain.cpp

@@ -2816,7 +2816,7 @@ void Qmgr::failReportLab(Signal* signal, Uint16 aFailedNode,
   if (failedNodePtr.i == getOwnNodeId()) {
     jam();
 
-    Uint32 code = 0;
+    Uint32 code = NDBD_EXIT_NODE_DECLARED_DEAD;
     const char * msg = 0;
     char extra[100];
     switch(aFailCause){

ndb/src/kernel/error/ndbd_exit_codes.c

@@ -57,12 +57,15 @@ static const ErrStruct errArray[] =
     "error(s) on other node(s)"},
    {NDBD_EXIT_PARTITIONED_SHUTDOWN, XAE, "Partitioned cluster detected. "
     "Please check if cluster is already running"},
+   {NDBD_EXIT_NODE_DECLARED_DEAD, XAE,
+    "Node declared dead. See error log for details"},
    {NDBD_EXIT_POINTER_NOTINRANGE, XIE, "Pointer too large"},
    {NDBD_EXIT_SR_OTHERNODEFAILED, XRE, "Another node failed during system "
     "restart, please investigate error(s) on other node(s)"},
    {NDBD_EXIT_NODE_NOT_DEAD, XRE, "Internal node state conflict, "
     "most probably resolved by restarting node again"},
    {NDBD_EXIT_SR_REDOLOG, XFI, "Error while reading the REDO log"},
+   {NDBD_EXIT_SR_SCHEMAFILE, XFI, "Error while reading the schema file"},
    /* Currently unused? */
    {2311, XIE, "Conflict when selecting restart type"},
    {NDBD_EXIT_NO_MORE_UNDOLOG, XCR,

ndb/test/ndbapi/testNdbApi.cpp

@@ -1306,6 +1306,36 @@ int runTestExecuteAsynch(NDBT_Context* ctx, NDBT_Step* step){
 
 template class Vector<NdbScanOperation*>;
 
+int 
+runBug28443(NDBT_Context* ctx, NDBT_Step* step)
+{
+  int result = NDBT_OK;
+  int records = ctx->getNumRecords();
+  
+  NdbRestarter restarter;
+
+  restarter.insertErrorInAllNodes(9003);
+
+  for (Uint32 i = 0; i<ctx->getNumLoops(); i++)
+  {
+    HugoTransactions hugoTrans(*ctx->getTab());
+    if (hugoTrans.loadTable(GETNDB(step), records, 2048) != 0)
+    {
+      result = NDBT_FAILED;
+      goto done;
+    }
+    if (runClearTable(ctx, step) != 0)
+    {
+      result = NDBT_FAILED;
+      goto done;
+    }
+  }
+  
+done:
+  restarter.insertErrorInAllNodes(9003);
+
+  return result;
+}
 
 NDBT_TESTSUITE(testNdbApi);
 TESTCASE("MaxNdb", 
@@ -1396,6 +1426,10 @@ TESTCASE("ExecuteAsynch",
 	 "Check that executeAsync() works (BUG#27495)\n"){ 
   INITIALIZER(runTestExecuteAsynch);
 }
+TESTCASE("Bug28443", 
+	 ""){ 
+  INITIALIZER(runBug28443);
+}
 NDBT_TESTSUITE_END(testNdbApi);
 
 int main(int argc, const char** argv){

ndb/test/run-test/daily-basic-tests.txt

@@ -629,6 +629,10 @@ max-time: 500
 cmd: testNdbApi
 args: -n ExecuteAsynch T1
 
+max-time: 1000
+cmd: testNdbApi
+args: -n BugBug28443
+
 #max-time: 500
 #cmd: testInterpreter
 #args: T1 

sql/ha_ndbcluster.cc

@@ -2309,16 +2309,24 @@ int ha_ndbcluster::write_row(byte *record)
   {
     // Table has hidden primary key
     Ndb *ndb= get_ndb();
-    int ret;
     Uint64 auto_value;
     uint retries= NDB_AUTO_INCREMENT_RETRIES;
-    do {
-      ret= ndb->getAutoIncrementValue((const NDBTAB *) m_table, auto_value, 1);
-    } while (ret == -1 && 
-             --retries &&
+    int retry_sleep= 30; /* 30 milliseconds, transaction */
+    for (;;)
+    {
+      if (ndb->getAutoIncrementValue((const NDBTAB *) m_table,
+                                     auto_value, 1) == -1)
+      {
+        if (--retries &&
             ndb->getNdbError().status == NdbError::TemporaryError);
-    if (ret == -1)
+        {
+          my_sleep(retry_sleep);
+          continue;
+        }
         ERR_RETURN(ndb->getNdbError());
+      }
+      break;
+    }
     if (set_hidden_key(op, table->s->fields, (const byte*)&auto_value))
       ERR_RETURN(op->getNdbError());
   } 
@@ -4855,23 +4863,28 @@ ulonglong ha_ndbcluster::get_auto_increment()
            m_rows_to_insert - m_rows_inserted :
            ((m_rows_to_insert > m_autoincrement_prefetch) ?
             m_rows_to_insert : m_autoincrement_prefetch));
-  int ret;
   uint retries= NDB_AUTO_INCREMENT_RETRIES;
-  do {
-    ret=
-      m_skip_auto_increment ? 
-      ndb->readAutoIncrementValue((const NDBTAB *) m_table, auto_value) :
-      ndb->getAutoIncrementValue((const NDBTAB *) m_table, auto_value, cache_size);
-  } while (ret == -1 && 
-           --retries &&
+  int retry_sleep= 30; /* 30 milliseconds, transaction */
+  for (;;)
+  {
+    if (m_skip_auto_increment &&
+        ndb->readAutoIncrementValue((const NDBTAB *) m_table, auto_value) ||
+        ndb->getAutoIncrementValue((const NDBTAB *) m_table,
+                                   auto_value, cache_size))
+    {
+      if (--retries &&
           ndb->getNdbError().status == NdbError::TemporaryError);
-  if (ret == -1)
       {
+        my_sleep(retry_sleep);
+        continue;
+      }
       const NdbError err= ndb->getNdbError();
       sql_print_error("Error %lu in ::get_auto_increment(): %s",
                       (ulong) err.code, err.message);
       DBUG_RETURN(~(ulonglong) 0);
     }
+    break;
+  }
   DBUG_RETURN((longlong)auto_value);
 }
 
@@ -5011,27 +5024,36 @@ int ha_ndbcluster::open(const char *name, int mode, uint test_if_locked)
   set_dbname(name);
   set_tabname(name);
   
-  if (check_ndb_connection()) {
-    free_share(m_share); m_share= 0;
-    DBUG_RETURN(HA_ERR_NO_CONNECTION);
+  if ((res= check_ndb_connection()) || 
+      (res= get_metadata(name)))
+  {
+    free_share(m_share);
+    m_share= 0;
+    DBUG_RETURN(res);
   }
-  
-  res= get_metadata(name);
-  if (!res)
+  while (1)
   {
     Ndb *ndb= get_ndb();
     if (ndb->setDatabaseName(m_dbname))
     {
-      ERR_RETURN(ndb->getNdbError());
+      res= ndb_to_mysql_error(&ndb->getNdbError());
+      break;
     }
     struct Ndb_statistics stat;
     res= ndb_get_table_statistics(NULL, false, ndb, m_tabname, &stat);
     records= stat.row_count;
     if(!res)
       res= info(HA_STATUS_CONST);
+    break;
   }
-
+  if (res)
+  {
+    free_share(m_share);
+    m_share= 0;
+    release_metadata();
     DBUG_RETURN(res);
+  }
+  DBUG_RETURN(0);
 }