BUG#20727343: Fix problems in UNDO log applier when changing log files

Author: Mikael Ronström

storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp

@@ -16632,6 +16632,12 @@ Dblqh::execFSWRITEREQ(Signal* signal)
  /**
  * This is currently run in other thread -> no jam
  * and no global variables
+ *
+ * This method is called from NDB file system while initialising a REDO log
+ * file, so we need to ensure that we don't touch any block variables other
+ * than to read stable variables. This is only called during initial
+ * restart. The pages are allocated by NDBFS from DataMemory, so these can
+ * be written to safely since they are owned by the file system thread.
  */
  Ptr<GlobalPage> page_ptr;
  FsReadWriteReq* req= (FsReadWriteReq*)signal->getDataPtr();

storage/ndb/src/kernel/blocks/dbtup/Dbtup.hpp

@@ -3320,7 +3320,7 @@ struct TupHeadInfo {
  Dbtup::Apply_undo f_undo;
  Uint32 c_proxy_undo_data[20 + MAX_TUPLE_SIZE_IN_WORDS];
 
- void disk_restart_undo_next(Signal*);
+ void disk_restart_undo_next(Signal*, Uint32 applied = 0);
  void disk_restart_undo_lcp(Uint32, Uint32, Uint32 flag, Uint32 lcpId);
  void disk_restart_undo_callback(Signal* signal, Uint32, Uint32);
  void disk_restart_undo_alloc(Apply_undo*);

storage/ndb/src/kernel/blocks/dbtup/DbtupCommit.cpp

@@ -1,5 +1,5 @@
 /*
- Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+ Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
 
  This program is free software; you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published by
@@ -22,10 +22,12 @@
 #include <ndb_limits.h>
 #include <pc.hpp>
 #include <signaldata/TupCommit.hpp>
+#include <EventLogger.hpp>
 #include "../dblqh/Dblqh.hpp"
 
 #define JAM_FILE_ID 416
 
+extern EventLogger *g_eventLogger;
 
 void Dbtup::execTUP_DEALLOCREQ(Signal* signal)
 {
@@ -633,6 +635,8 @@ int Dbtup::retrieve_log_page(Signal *signal,
  signal->theData[0] = 1;
  return res;
  case -1:
+ g_eventLogger->warning("Out of space in RG_DISK_OPERATIONS resource,"
+ " increase config parameter GlobalSharedMemory");
  ndbrequire("NOT YET IMPLEMENTED" == 0);
  break;
  default:

storage/ndb/src/kernel/blocks/dbtup/DbtupDiskAlloc.cpp

@@ -603,6 +603,9 @@ Dbtup::disk_page_prealloc(Signal* signal,
  case 0:
  break;
  case -1:
+ g_eventLogger->warning("Out of space in RG_DISK_OPERATIONS"
+ " resource, increase config parameter"
+ " GlobalSharedMemory");
  ndbrequire("NOT YET IMPLEMENTED" == 0);
  break;
  default:
@@ -1619,10 +1622,12 @@ Dbtup::disk_restart_undo(Signal* signal, Uint64 lsn,
 }
 
 void
-Dbtup::disk_restart_undo_next(Signal* signal)
+Dbtup::disk_restart_undo_next(Signal* signal, Uint32 applied)
 {
  signal->theData[0] = LgmanContinueB::EXECUTE_UNDO_RECORD;
- sendSignal(LGMAN_REF, GSN_CONTINUEB, signal, 1, JBB);
+ /* Flag indicating whether UNDO log was applied. */
+ signal->theData[1] = applied;
+ sendSignal(LGMAN_REF, GSN_CONTINUEB, signal, 2, JBB);
 }
 
 void
@@ -1773,6 +1778,7 @@ Dbtup::disk_restart_undo_callback(Signal* signal,
 // key.m_file_no = pagePtr.p->m_file_no;
 
  Uint64 lsn = 0;
+ Uint32 applied = 0;
  lsn += pagePtr.p->m_page_header.m_page_lsn_hi; lsn <<= 32;
  lsn += pagePtr.p->m_page_header.m_page_lsn_lo;
 
@@ -1788,6 +1794,7 @@ Dbtup::disk_restart_undo_callback(Signal* signal,
  }
 
  update = true;
+ applied = 1;
  if (DBG_UNDO)
  ndbout_c("applying %lld", undo->m_lsn);
  /**
@@ -1830,7 +1837,7 @@ Dbtup::disk_restart_undo_callback(Signal* signal,
  << " tab: " << tableId << endl;
  }
 
- disk_restart_undo_next(signal);
+ disk_restart_undo_next(signal, applied);
 }
 
 void
@@ -1882,8 +1889,17 @@ Dbtup::disk_restart_undo_free(Apply_undo* undo)
  {
  abort();
  } 
- 
- ndbrequire(idx == undo->m_key.m_page_idx);
+
+ if (idx != undo->m_key.m_page_idx)
+ {
+ Uint64 lsn = undo->m_lsn;
+ jam();
+ jamLine(lsn & 0xFFFF);
+ jamLine((lsn >> 16) & 0xFFFF);
+ jamLine((lsn >> 32) & 0xFFFF);
+ jamLine((lsn >> 48) & 0xFFFF);
+ ndbrequire(false);
+ }
  const Disk_undo::Free *free = (const Disk_undo::Free*)undo->m_ptr;
  const Uint32* src= free->m_data;
  memcpy(ptr, src, 4 * len);

storage/ndb/src/kernel/blocks/dbtup/DbtupMeta.cpp

@@ -1,5 +1,5 @@
 /*
- Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+ Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
 
  This program is free software; you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published by
@@ -488,6 +488,8 @@ void Dbtup::execTUP_ADD_ATTRREQ(Signal* signal)
 signal->theData[0] = 1;
 return;
  case -1:
+ g_eventLogger->warning("Out of space in RG_DISK_OPERATIONS resource,"
+ " increase config parameter GlobalSharedMemory");
 ndbrequire("NOT YET IMPLEMENTED" == 0);
 break;
  }
@@ -2041,6 +2043,8 @@ void Dbtup::releaseFragment(Signal* signal, Uint32 tableId,
  jam();
  return;
  case -1:
+ g_eventLogger->warning("Out of space in RG_DISK_OPERATIONS resource,"
+ " increase config parameter GlobalSharedMemory");
  warningEvent("Failed to get log buffer for drop table: %u",
  tabPtr.i);
  lgman.free_log_space(sz);
@@ -2194,6 +2198,8 @@ Dbtup::drop_fragment_free_extent(Signal *signal,
  jam();
  return;
 case -1:
+ g_eventLogger->warning("Out of space in RG_DISK_OPERATIONS resource,"
+ " increase config parameter GlobalSharedMemory");
  ndbrequire("NOT YET IMPLEMENTED" == 0);
  break;
 default:

storage/ndb/src/kernel/blocks/dbtup/DbtupProxy.cpp

@@ -1,4 +1,4 @@
-/* Copyright (c) 2008, 2013, Oracle and/or its affiliates. All rights reserved.
+/* Copyright (c) 2008, 2015, Oracle and/or its affiliates. All rights reserved.
 
  This program is free software; you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published by
@@ -402,7 +402,8 @@ DbtupProxy::disk_restart_undo_finish(Signal* signal)
  if (undo.m_actions & Proxy_undo::SendUndoNext) {
  jam();
  signal->theData[0] = LgmanContinueB::EXECUTE_UNDO_RECORD;
- sendSignal(LGMAN_REF, GSN_CONTINUEB, signal, 1, JBB);
+ signal->theData[1] = 0; /* Not applied flag */
+ sendSignal(LGMAN_REF, GSN_CONTINUEB, signal, 2, JBB);
  }
 
  if (undo.m_actions & Proxy_undo::NoExecute) {